While fuel and natural gas prices have skyrocketed worldwide and global warming has alarming impact on our society and the future of our planet, the transformation of greenhouse gases into high-value molecules has become an impelling priority1,2,3. As such, the valorisation of carbon dioxide (CO2) and its conversion into synthetically valuable C1 building blocks represents a challenge tackled with extensive efforts by the scientific community4,5,6,7,8,9. CO2 reduction to carbon monoxide (CO) is particularly appealing to support a circular and carbon-neutral economy. Among the established methodologies, electrochemical10,11 and photochemical (for representative reviews, see refs. 12,13,14) strategies for CO2 reduction may provide a future opportunity for sustainable industrial applications. Conceptually, harvesting sunlight to drive CO2 conversion, a process often referred to as artificial photosynthesis, is highly appealing to achieve low carbon footprint. Nonetheless, gaps remain between the photoreduction of CO2 into CO and its subsequent valorisation.

In general, carbonylation reactions require relatively high CO concentration, but the CO evolved from the photoreduction of CO2 usually does not meet such prerequisite. Furthermore, unreacted excess of CO2 might be parasitic to the subsequent CO functionalisation with various side-reactions, such as the formation of ammonium bicarbonates/carbonate by CO2−amine reaction in metal-catalysed amino-carbonylations15,16. In this context, achieving high CO2-to-CO conversion and high purity of CO (>90%) would save additional energy-intensive purification steps. Unfortunately, completeness of CO2-to-CO conversion is most often a neglected parameter17. Attention has been focused on the improvement of catalyst performances under high excess of CO2, rather than on complete exhaustive CO2 conversions. There are only limited examples in the literature where this factor has been taken into account18. He and co-workers have shown that a moderate CO2-to-CO conversion (<10%) was observed in presence of Re(bpy)(CO)3Cl photocatalyst19, while we failed to achieve >30% CO2 conversion on a scale as low as 0.5 mmol20. A similar outcome was encountered for the electrocatalytic reduction of CO2. In 2021, the groups of Cantat and Fontecave reported that CO2 electroreduction, coupled with propylene oxide carbonylating thermal catalysis, provided the corresponding β-butyrolactone in only 1.7% yield from CO221. From a societal perspective, such results rise a challenge on the future applicability of photochemical CO2 reduction.

Aiming to develop effective and stoichiometric functionalisation reactions using CO2, herein we report on a photocatalytic process that unlocks full and fast CO2-to-CO conversion (<10 min, at room temperature, on 0.4 mmol scale). Beside the fundamental interest of the process, the transformation has utmost implications in the field of carbon isotope radiolabeling, where [14C]CO and [11C]CO remain underexploited radioactive C1 building blocks (Fig. 1). In the process, we show the concrete application of the tandem CO2 reduction/carbonylation on a variety of substrates, including pharmaceutically relevant bioactive compounds. The method is suitable to all carbon isotopes (13C, 14C and 11C).

Fig. 1: Current state-of-the-art to access carbon isotopologues of carbon monoxide.
figure 1

a Three general strategies for formal CO2-to-CO reduction. b Our work and opportunities on the rapid and full CO2-to-CO photoreduction and direct valorisation of CO.

Results and discussion

Carbon monoxide in radiochemistry

Carbon monoxide is an excellent C1 building block in organic synthesis and carbonylation reactions are valuable tools to elaborate organic molecules22,23,24,25,26. While [12C]CO is cheap and available at an industrial scale, this is not the case for its carbon-labeled isotopologues. While [13C]CO is the primary 13C source, high isotopic purity is obtained by energy-intensive cryogenic distillation from natural CO (1.1% 13C abundance)27. For radioactive 14C (β- emitter, half-life 5730 years) and 11C (β+ emitter, half-life 20.4 min), access to radiolabeled CO is challenging. [14C]CO is unstable and undergoes radiolysis28,29, and it must be generated in or ex situ and immediately utilized. Three routes for preparing [14C]CO are described: (a) zinc-bed reduction of [14C]CO2 at 385 °C30, (b) dehydration of 14C-labeled formic acid with concentrated sulphuric acid at 70 °C31,32 and (c) ex situ decarbonylation of acid chloride using [14C]COGen (Fig. 1a)33. The thermal zinc reduction has the advantage of using directly [14C]CO2, (price: 1860 $/mmol) as the primary source of 14C, but it has inherent safety and practical limitations. On the other hand, use of formic acid and COgen are milder and easier to implement, but require secondary 14C-reagents that are synthetized from [14C]CO2. For short-lived 11C, access to [11C]CO is mainly granted by direct reduction of [11C]CO2 by zinc (400 °C) and molybdenum (850 °C) (for recent reviews on the topic, see refs. 34,35). Due to the 11C short half-life, the use of secondary formate derivatives is unsuitable, while the reduction of [11C]CO2 with silyl reagents was recently reported36,37,38. Additionally, in 2017 a proof-of-concept showing the use of electrochemical reduction of [11C]CO2 was reported by Gee and Long39. In light of the state-of-the-art, unlocking high conversion in the photocatalytic reduction of CO2 would have a significant impact on the field of CO2 valorisation and major applications in carbon isotope chemistry.

Using transient absorption spectroscopy, we previously described the photo-induced electron transfer steps, from a ruthenium (II) trisbipyridine photosensitizer (Ru PS) to a rhenium (I) bipyridine triscarbonyl catalyst (Re cat)20. We have shown that higher efficiencies for the photocatalytic reduction of CO2-to-CO were achieved using 1,3-dimethyl-2-phenylbenzimidazoline (BIH), as two-electron and one-proton sacrificial donor and water as additional proton source in dimethylformamide (DMF). A first proof-of-concept for the consecutive utilisation of the photo-produced [13C]CO in an aminocarbonylation reaction could be obtained. However, we clearly highlighted the importance of the often-overlooked criterion of full CO2-to-CO conversion for this valorisation strategy, in contrast with the commonly used criterion of turnover number (TON = amount of CO/amount of catalyst). Indeed, we found that the lower CO2-to-amide yield was due to a low CO2-to-CO conversion (<30%) in the first reaction (i.e., the photocatalytic reduction of CO2 by the Re catalyst). NMR monitoring has revealed that the low yield was in part due to a concomitant formation of bicarbonate during the photocatalytic production of CO, as a result of CO2 acting also as an oxygen atom acceptor (2 CO2 + 2 e + H+ → CO + HCO3). To address this issue, guided by kinetic simulations (vide infra), we performed a systematic screening of different components of the catalytic system such as solvents, proton sources and additives that may play the role of oxygen atom acceptor.

Reaction optimisation

The optimisation of the photocatalytic CO2 reduction was performed using a two-chamber reactor system (Supplementary Table 1). The photoreduction took place in the first chamber (Ch.1), where a precise amount of stable [13C]CO2 was introduced with an RC-TRITEC carboxylation manifold, to guarantee a precise gas loading. The scale of labeled CO2 was decided to be 0.3 and 0.4 mmol. This decision was consciously made for the application to the radioisotopes 14C and 11C for the following reasons: (a) this scale is a good compromise in terms of costs of the 14C-radioactive material; (b) it provides suitable amounts of labeled product for in vivo applications; (c) it limits the possible generation of long-lasting radioactive waste after the completion of the carbonylation reactions; (d) it avoids high pressure in the reactor and provides a more suitable safety profile for the implementation to radioactive carbon.

After the reduction and gas diffusion, the [13C]CO generated was systematically quantified using an established palladium-catalysed aminocarbonylation reaction in the second chamber (Ch.2; yield based on [13C]CO2 as limiting reagent). Our initial reaction conditions utilizing Re cat (0.15 mol%), Ru PS (0.15 mol%) under blue light irradiation (Kessil A160WE Tuna Blue LED lamps, Supplementary Figs. 25 and 26), BIH and water as proton source allowed to obtain [13C]1 in a disappointing 35% yield (Fig. 2A, entry 1). Extensive optimisation pointed out the primary importance of the proton source in enhancing the efficiency of the transformation. The use of phenol revealed an improvement in comparison with the commonly used TEOA (triethanolamine) and water (Fig. 2A, entries 1–5). To our delight, by increasing the loading of Re cat and Ru PS from 0.15 mol% to 0.45 mol% (entry 8), the desired amide [13C]1 was obtained in 70% yield. In contrast, no beneficial effect was observed in presence of TEOA, under the same conditions (entry 4).

Fig. 2: Optimisation of the CO2-to-CO reduction coupled with aminocarbonylation.
figure 2

A Optimisation of the transformation. See Supplementary Tables 1 and 2, for full experimental details. [a] Ch.1: [13C]CO2 (0.365 mmol), BIH (0.78 mmol); DMF, room temperature; Ch.2: 4‐iodoanisole (0.724 mmol), n‐hexylamine (1.45 mmol), DABCO (1.45 mmol), Pd(dba)2 (2 mol%), P(tBu)3HBF4 (4 mol%), THF (0.24 M), 25 °C, 48 h. [b] 1H-NMR yields calculated using 1,3,5-trimethoxybenzene as internal standard. [c] Yield of isolated product. DMF dimethylformamide, THF tetrahydrofuran, BIH 1,3-dimethyl-2-phenylbenzimidazoline, TEOA triethanolamine, PPh3 triphenylphosphine, DABCO 1,4-diazabicyclo[2.2.2]octane, Pd(dba)2 Palladium(0) bis(dibenzylideneacetone), P(tBu)3HBF4 Tri-tert-butylphosphine tetrafluoroborate. B Screening of phosphine additives: conditions reported in Entry 12 (A) were used (0.036 mmol). C Screening of phenol proton donors: phosphine P2 was used (0.036 mmol). D Pressure measurements relating to the rate of CO formation. Conditions: [13C]CO2 (0.40 mmol), BIH (0.87 mmol); proton source (2.22 mmol); Ru PS (0.45 mol%); Re cat (0.45 mol%) in DMF. The reaction mixture was stirred at room temperature under blue light irradiation (low intensity) for 1 h. TEOA triethanolamine, TFE trifluoroethanol.

To prevent the collateral formation of HCO3, triphenylphosphine (PPh3, 1 equiv.) was used as a potential oxygen atom acceptor40,41,42,43 and we were pleased to see an improvement in the yield of ca. 5–10% (entries 2–3, 6–7). Interestingly, PPh3 was effective even when used in catalytic amount (10 mol%), thus disproving its role as a stoichiometric oxygen atom acceptor (entry 12). Spectro-electrochemical investigations showed no significant effect of the PPh3 additive in the reduction potentials of the Re cat, nor in the electrocatalytic reduction activity of the catalyst (Supplementary Figure 19). These results suggested that PPh3 does not play a role in the catalytic redox cycle of Re cat. At present, the origin of such beneficial effect remains unclear. Additionally, we observed that the dual Ru PS - Re cat sensitised catalysis offered a tangible advantage in terms of reaction time, allowing the full conversion of [13C]CO2 into [13C]CO within only 1 h (entry 10). In absence of Ru PS the photoreduction was much slower (after 1 h only 40% yield of [13C]1 is observed). Finally, by adjusting the amount of phenol in solution (5.5 equiv.) and using a catalytic amount of PPh3 (10 mol%) an overall 83% yield of [13C]1 from [13C]CO2 was achieved (entry 12). The high isotopic purity of the compound [13C]1 (97.3%) highlighted that only a negligible isotopic dilution took place in the process4. Control experiments performed in the absence of BIH, Re Cat or light irradiation highlighted no product formations (entries 29, 30, 34, Supplementary Table S1). In the absence of PPh3, [13C]1 was observed in 45% yield under identical catalyst loading, while in 55% yield when the Ru PS–Re cat charge was increased to 0.75 mol% (entries 32, 33, Supplementary Table S1). Experiments performed by replacing the Re Cat by organic photocatalyst 4CzIPN were unsuccessful (Supplementary Table 1, entries 35–37) (for a recent review on organophotoredox catalysis, see ref. 44).

Next, we screened a series of phosphines (Fig. 2B). Electron-rich phosphines (P4-P5) resulted to inferior outcomes while dimeric ones (P6-P7) even dramatically decreased the yield. In contrast, electron-poor phosphines (P2-P3) behaved analogously to PPh3. P2 was selected for further experiments, as a good compromise between yield and cost. We then examined the impact of substituents on the proton donor (Fig. 2C): p-methoxyphenol gave a lower yield (56%), p-nitrophenol led to no conversion (possibly be explained by interference in light absorption due to the intense yellow-colored solution); while 2,4-difluorophenol was comparable with unsubstituted phenol. A control experiment using sodium phenoxide did not show any conversion and confirmed the crucial role of phenol as a proton donor.

To evaluate in real-time the effect of reaction conditions (proton sources, solvents and light intensities) on the [13C]CO production, we decided to undertake a series of pressure studies. With carbon monoxide being much less soluble than CO2 in organic solvents, the increase in pressure measured in the headspace of the manifold directly correlates to the CO2-to-CO conversion (Supplementary Fig. S10). As shown in Fig. 2D, the presence of phenol allowed reaching a plateau within only 40 min in DMF. This result shows that phenol is significantly more effective compared to other proton sources (TEOA, TFE or water, Supplementary Fig. S14). Interestingly, acetonitrile provided even better performances compared to DMF and DMSO in terms of reaction rate, as the plateau was reached within 27 min (Supplementary Fig. S15). The blue light intensity of commercially available Kessil LED lamps has been evaluated and the use of higher intensity allowed the full CO2-to-CO conversion in only 7 min (Supplementary Fig. S16). Gas chromatography monitoring of the reaction headspace further confirmed these results (Supplementary Figs. S21 and 22). Interruption of the light irradiation by a series of dark/light cycles, showed that the catalyst is still active and the reduction proceeds after short pauses (Supplementary Fig. S11). Unfortunately, attempts to recycle the photocatalytic set-up by recharging the reactor with additional [13C]CO2 after the first run were unsuccessful.

These results revealed that a significant enhancement of the CO2-to-CO conversion (up to 83% of [13C]1, over two steps) was achieved when phenol was used as a proton source instead of water (Fig. 2D). A prior electrocatalytic evaluation of the catalytic system has also shown that for a CO2-purged DMF solution of the Re catalyst, higher catalytic rates were observed when PhOH was used instead of water (Supplementary Figs. S17 and S18). Similar observations have been reported for the electrocatalytic activity in acetonitrile solutions45. However, the effect of phenol has rarely been documented to improve the activity and CO2 conversion efficiency of photocatalytic systems involving the Re bipyridine catalysts, where most of the studies use triethanolamine, triethylamine, or water as additives in DMF or ACN46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66. From this optimisation, we found that the most important parameters to influence the CO2 to CO conversion are the use of phenol as proton donor, the stoichiometry of phenol and the increase of the catalyst loading from 0.15 mol% to 0.45 mol%.

Mechanistic studies

To build a more comprehensive picture and to understand the origin of the beneficial effect of phenol, we combined thermodynamic data (redox potentials) obtained from electrochemical experiments and DFT calculations reported in the literature, along with known pKa values, kinetic rate constants, and solubility constants of the gases (CO2 and CO), to implement a reaction-model-based kinetic simulation (Supplementary pages 3337). The relevance of these simulation results was experimentally validated by monitoring the pressure changes in the headspace during the photocatalytic reduction of CO2, which gives a direct access to a time-resolved CO production profile (Fig. 3a).

Fig. 3: Mechanistic investigation.
figure 3

a Experimental pressure reading (circles) during the photocatalytic reduction of CO2 to CO from Ch.1 containing 0.57 mM Ru PS, 0.57 mM Re cat, 279 mM BIH, 12.7 mM sodium phosphinidynetris(benzene sulfonate) and 712 mM phenol (or H2O) in dimethylformamide, irradiated using a Blue LED lamp (irradiance of 117 W m−2). Fitted simulated pressure data (solid line) is shown based on a reaction-model-based kinetic simulation. Concentration profiles of CO, CO2, and HCO3 are shown based on the results of the kinetics simulation (see Supplementary pages 2838) distinguishing the effect of adding b phenol or c water to the photocatalytic solution. d Proposed photocatalytic cycle for the Ru-Re bimolecular system on which the kinetic simulation is based upon. Effect of water as proton source is indicated in thick black arrows accumulating bicarbonate while the changes when phenol is used is indicated in green arrows.

Our proposed mechanism for the photocatalytic CO2-to-CO reduction is depicted in Fig. 3d. Using transient absorption spectroscopy, we have previously shown that the catalytically active species (Re) is generated by two successive electron transfers20. The first one comes from the reduced Ru photosensitizer (the formal RuI), to form the singly-reduced Re (Re) catalyst (Reaction 3 in Fig. 3d), and the second one from the highly reducing BI radical (Reaction 4) formed during the photo-induced first steps (Reactions B and C)67. The next steps (Reactions 5–11) are mainly based on reported DFT calculations68,69. In Reaction 5, the catalytically active species Re is nucleophilic enough to react with CO2 and form a ReCO2 intermediate which undergoes a first protonation (Reaction 6) to give ReCO2H followed by an electron transfer step (Reaction 7) to yield ReCO2H. These last two steps are considered to be fast and precede a ‘dehydroxylation’ reaction, which is the rate-limiting step of the catalytic system. A proton can ‘dehydroxylate’ the ReCO2H intermediate to form water (Reaction 8), or another CO2 substrate can ‘dehydroxylate’ this intermediate to form bicarbonate (Reaction 9). We believe that the latter is the critical step for the CO2-to-CO conversion efficiency. Indeed, in the presence of water, the carbonic acid (H2CO3) produced by the equilibrium between water and CO2 (Reaction 10) can be a source of proton (consequently producing bicarbonate as shown in Reaction E), and together with Reaction 9 can generate another equivalent of HCO3. This explains the low CO2-to-CO conversion observed initially in presence of water because, virtually, 50% of CO2 is converted to HCO3 (Fig. 3c). By using phenol as a proton source in the catalytic system, the HCO3 can be recycled back to H2CO3 (Reaction E) then to CO2 (Reaction 10). It is worth mentioning that both the simulation and the pressure monitoring experiment show two distinct kinetic processes in presence of phenol, i.e., the first 20 min where the CO2 originally introduced in the reaction is consumed and the second phase where the CO2 recycled from bicarbonate is consumed (Fig. 3b). Phenol also plays a direct role as a proton donor in Reactions 8 and D, as further shown in Supplementary Fig. S24. In the absence of phenol, the concentration of the HCO3 formed in the first phase remains unchanged (Fig. 3c). In the last step (Reaction 11), one more electron transfer (possibly coming from Reaction C) is needed to release the CO and regenerate the active species Re.

Substrate scope

Next, we attempted to apply this technology for the direct preparation of high-value 13C-isotopically labeled compounds exploiting the effective ex situ generation of [13C]CO (Fig. 4). Aminocarbonylation gives access to a straightforward synthesis of amides by using a sub-stoichiometric amount of carbon monoxide (Fig. 4A). Combinations of (hetero)aryl iodides and alkyl amines allowed the preparation of amides containing relevant motifs, such as morpholine, piperidine and more sterically hindered adamantyl in good yields from [13C]CO2 (compounds [13C]1–6). When higher stoichiometry of [13C]CO2 was used (0.6 mmol), the isolated yield of amide [13C]6 increased from 53% to 65%. A visible-light-enabled aminocarbonylation of alkyl iodide was attempted, as well. This non-optimised result allowed observing the labeled amide [13C]3b in 18% yield in the reaction crude. Using alcohol as a nucleophilic partner, a series of 13C-labeled esters was synthesized. In the two-chamber system, the coupling reaction catalyzed by Pd(dba)2 and CataCXium-A gave lower yields when aryl bromides were used (compounds [13C]7-9, 27–35%)70. On the other hand, with (hetero)aryl iodides improved yields were observed and compounds [13C]10–13 could be isolated in 64 to 75% yields.

Fig. 4: Investigation of the scope of the photocatalytic CO2-to-CO conversion.
figure 4

Green colored circles and numbers denote the positions of the carbon atoms labeled and the percent incorporation of the carbon isotope. See Supplementary pages 4045. Conditions in Ch1, for A, B, C, E, F: [13C]CO2 (0.36 mmol), BIH (0.87 mmol, 2.4 equiv.); phenol (2.0 mmol, 5.5 equiv.); Ru PS (0.45 mol%, 1.62 µmol); Re(CO)3(bpy)Br Re cat (0.45 mol%, 1.62 µmol), phosphine P2 (0.036 mmol, 0.1 equiv.) in ACN. Conditions in Ch1, for D: [13C]CO2 (0.44 mmol), BIH (0.95 mmol, 2.12 equiv.); phenol (2.44 mmol, 5.5 equiv.); Ru PS (0.45 mol%, 1.95 µmol); Re(CO)3(bpy)Br Re cat (0.45 mol%, 1.95 µmol), phosphine P2 (0.044 mmol, 0.1 equiv.) in ACN. aConditions in Ch1: [13C]CO2 (0.60 mmol), BIH (1.56 mmol, 2.60 equiv.); phenol (4.0 mmol, 6.6 equiv.); Ru PS (0.54 mol%, 3.3 µmol); Re(CO)3(bpy)Br Re cat (0.53 mol%, 3.2 µmol), phosphine P2 (0.072 mmol, 0.12 equiv.) in ACN. bSee Supplementary page 94.

As aldehydes are versatile functional groups in organic synthesis and found even as active pharmaceutical ingredients71, we explored whether this technology could be used to access them through a reductive carbonylation. By applying a reported protocol in combination with our photoreduction, it was possible to obtain benzaldehydes 14–19 (Fig. 4C)72. Notwithstanding the volatility issue encountered for some products, we were able to prepare a set of aldehydes using [13C]CO as limiting reagent (compounds [13C]14–16). For compounds [13C]17–19, excess of [13C]CO was utilized in order to obtain satisfactory yields. Carbonylative Suzuki–Miyaura coupling was also explored for preparing nonsymmetrical benzophenones (Fig. 4D)73. Compounds [13C]20–25 were isolated in 34 to 91% yields applying slightly modified conditions with respect to the literature. In particular, the presence of PPh3 favored the carbonylative process, thus reducing classical non-carbonylative Suzuki–Miyaura coupling. Once more, with higher stoichiometry of [13C]CO2 (0.6 mmol), the isolated yield of [13C]21 increased from 34% to 88%. Interestingly, the presence of aryl chlorides was tolerated under different coupling conditions (see products [13C]15 and [13C]24). When we attempted replacing aryl iodides with the corresponding aryl chlorides, no desired product was observed (see SI for details).

The emergence of Carbon Isotope Exchange (CIE)74,75 provided a paradigm change in the preparation of carbon-labeled molecules76,77,78,79,80,81. The work by Gauthier and co-workers is particularly attractive in this area82. Utilizing acyl chlorides and palladium catalysis under [13C]CO atmosphere, the corresponding carboxylic acids were obtained without the need for time-consuming syntheses of precursors. By adapting our photocatalytic CO2-to-CO conversion, we could directly access isotopically enriched carboxylic acids (Fig. 4E). Benzoic acids [13C]26–29 were isolated in moderate to good yields, in 19 to 50% isotopic enrichments (i.e., 13C/(13C+12C) ratios). Phenyl acetic acid [13C]30 was also obtained in a similar degree of efficiency and finally, terephthalic acid [13C]31 was isolated in 43% single 13C-labeled and 12% double 13C-labeled form.

Further, we wished to implement a carbonylative Sonogashira coupling protocol (Fig. 4F). This key transformation generates synthetically useful ynones, starting from aryl halides and alkynes, whose reactivity is suitable for the construction of biologically active heterocycles. In presence of a catalytic amount of PdCl2 (5 mol%) and Xantphos (5 mol%), compounds [13C]32–34 were isolated in 32 to 66% yield. Ynone [13C]34 was further converted into the benzofuran derivative [13C]35, by means of an efficient one-pot procedure starting from o-iodophenol, and the pyrazole [13C]36, which enabled the insertion of the carbon tag in the heterocycle core.

Having established the versatility of the photocatalytic CO2-to-CO reduction over multiple carbonylative transformations (Fig. 4), we looked into the application to the labeling of biologically active molecules. isoFCA-4 [13C]38 (Fig. 5A), an antimitotic agent developed by the group of Alami83, was prepared in two steps with an overall 37% yield, using a carbonylative Suzuki coupling and subsequent Wittig olefination. The Rofecoxib analogue [13C]39 was synthesized by an intramolecular alkoxycarbonylation in 71% yield, while the intermolecular coupling reaction allowed to isolate Butoxycaine [13C]40 in 96% yield and Moclobemide [13C]41 in 61% yield. Finally, we carried out a CIE on a Probenecid [13C]42 starting from the corresponding unlabeled acid with a good 36% isotopic enrichment observed.

Fig. 5: Application of the photocatalytic CO2-to-CO conversion to 13C, 14C and 11C carbonylation.
figure 5

A Labeling of pharmaceutically relevant scaffolds. B Carbon-14 radiolabeling. C Carbon-11 radiolabeling. See Supplementary pages 86109, for detailed conditions. Blue-colored circles and numbers denote the positions of the 14C atoms labeled and the percent incorporation of the isotope. Orange-colored circles denote the positions of the 11C atoms labeled.

Radiolabeling with carbon-14 and carbon-11

To assess the importance of the photocatalytic CO2-to-CO conversion in the field of radiochemistry, we next explored its application to 14C carbonylation reactions. While this step (i.e., switching from stable labeled 13C to radioactive beta emitter 14C) might look trivial, the inherent electron emission of the isotope might trigger fast covalent bond cleavage (i.e., radiolysis). The fact that [14C]CO undergoes radiolysis is an emblematic example28,29. Furthermore, differences in reactivity between stable 12C- and 14C-labeled compounds has been reported. In 1986, Parker made the hypothesis that, “if a chemical reaction can proceed by two or more pathways, where one involves the participation of free radicals, the latter might be favored, when a high molar activity radiolabeled species is involved”84. Consequently, the application of this photocatalytic procedure provided a high degree of uncertainty. When the standard conditions were applied to the model amide, the desired amide [14C]1 could be isolated in 46% radiochemical yield (RCY) and high molar activity (Am) of 2.1 GBq mmol−1. This straightforward generation of [14C]CO was further showcased by the preparation of Butoxycaine [14C]40 and the isoFCA-4 precursor [14C]37, both of them isolated in high Am. Finally, CIE on Probenecid allowed to isolate the labeled drug [14C]42 in 73% yield and 37% IE (Am = 0.61 GBq  mmol−1).

At last, a proof-of-concept on 11C-photocarbonylation was realized on a Synthra module, to demonstrate the feasibility of the automation of the reaction and, therefore, the applicability to the synthesis of radiotracers for PET imaging studies. Cyclotron-produced [11C]CO2 was photochemically converted into [11C]CO within only 5 min under mild conditions compared to the standard methods reported in the literature34, and with a conversion of 30% (n = 4) comparable to the recent method described by Bongarzone et al.36,38,39 using the disilane strategy. The separation of [11C]CO and unreacted [11C]CO2 was easily realized by means of an Ascarite® column. Subsequent palladium-catalyzed aminocarbonylation afforded the desired compound [11C]1 in 83% conversion from [11C]CO and 25% from [11C]CO2 (n = 3) within only 2 min. Despite the presence of “cold” [12C]CO in the reaction mixture coming from the Re catalyst, [11C]1 was obtained in good molar activity of 30 GBq/µmol, as a consequence of the high-yielding process. These results demonstrate that the scope of this photoreduction–carbonylation process can be expanded to the preparation of 11C PET tracers for in vivo imaging.

In conclusion, we have reported an effective phototocatalytic approach to enable the full CO2-to-CO reduction within minutes and the direct use of the produced CO in different types of carbonylation reactions. The versatility of this reaction manifold has also shown its potential in the easier and straightforward preparation of radiotracers, which is nowadays essential in the field of human health, such as diagnosis and drug and agrochemical developments. The optimisation of this transformation has been rationalized using reaction-model-based kinetic simulations implementing photophysical and electrochemical data. The overall process has allowed the labeling of a structurally diverse library of derivatives including, amides, ester, ketones, aldehydes and carboxylic acids in one single step from CO2. This technology opens up new opportunities for the direct access to 11C- and 14C-labeled pharmaceuticals from their primary isotopic sources [11C]CO2 and [14C]CO2.


General procedure for catalytic for the [13C]CO2 photoreduction

In a two-chamber reaction, a suspension of Ru(bpy)3Cl2.6H2O stock solution (1.20 mL, 1.34 mM), Re(CO)3(bpy)Br stock solution (0.81 mL, 1.98 mM), ACN (0.8 mL), BIH (175 mg, 0.78 mmol), phosphine P2 (20.5 mg, 0.036 mmol), phenol (188 mg, 2.0 mmol) were transferred into Chamber 1 with a Pasteur pipette. The chambers were sealed with a screwcap fitted with a Teflon®. The adaptor was then connected to the RC Tritec® system. The solution in Chamber 1 was frozen with a liquid nitrogen bath and the chambers were degassed with vacuum pump connected with RC Tritec manifold for 10 min The stopcock was closed between the two chambers. [13C]CO2 (365 μmol) was then loaded into Chamber 1 using the RC Tritec® system and the stopcock was closed between Chamber 1 and the adaptor. The loaded Two-Chamber Glassware was then disconnected from the RC Tritec® system and the suspension was warmed to room temperature. Chamber 1 was placed ca. 2 cm away from a 40 W A160WE Tuna Blue Kessil® LED lamp and photo‐irradiated with the lower light intensity for 1 h. The [13C]CO produced is then used in the carbonylation reaction in Chamber 2.