Photolysis of CO₂ Carbamate for Hydrocarboxylation Reactions

Abstract

The conversion of carbon dioxide into value-added products has emerged as an alternative method to achieve net-zero emissions. While technologies that transform CO₂ into fuels and chemical feedstocks have made great strides, the direct use of CO₂ as a C1 synthon for the formation of new carbon–carbon bonds remains a critical challenge. Herein, we present a new catalytic CO₂ activation mode for hydrocarboxylation reactions. Key to this methodology is the formation of a CO₂ carbamate with a phenothiazine catalyst, which sets the required trigonal geometry for the release of CO₂ ^•– via photolysis upon absorption of visible light. The polarity-reversed CO₂ ^•– is employed in the hydrocarboxylation reactions of alkenes and heterocycles. This protocol is distinguished by its mild reaction conditions, wide substrate scope and broad applicability, even in the context of pharmaceutical cores. Our chemistry can also be utilized for the synthesis of carbon-13 labeled spirolactones using ¹³CO₂. Mechanistic experiments support the photolysis of the CO₂ carbamate as the main productive pathway under our optimized reaction conditions.

Introduction

The direct application of CO₂ as a C1 synthon for creating new carbon–carbon bonds continues to pose a significant challenge in synthetic chemistry. − Although CO₂ is considered an ideal feedstock because of its wide availability, the relative stability of this linear small molecule has stimulated chemists to design various activation strategies to exploit its synthetic potential. Traditionally, these methods required the use of strong nucleophiles such as organometallic reagents or transition-metal catalysts (Figure A). − Lately, the single electron reduction of carbon dioxide to generate the CO₂ radical anion (CO₂ ^•–) has attracted considerable attention as an alternative approach to access new reactivity. However, this is a thermodynamically and kinetically demanding process, due to the extremely low reduction potential of CO₂ (E _1/2 = −2.2 V vs SCE) and the high reorganization energy required to accommodate the new trigonal geometry of CO₂ ^•– during the electron transfer. Given the high energetic requirement, researchers first utilized an electrochemical setup to achieve the single electron reduction of carbon dioxide. While various works reported the successful activation of CO₂, high negative overpotentials and currents were crucial to overcome the slow kinetics and efficiently generate the CO₂ ^•–. − Recently, several methods have emerged for the photochemical single-electron reduction of the CO₂. High-energy UV light, , stoichiometric photoreductants, iridium or heterogeneous photocatalysts, − often in combination with stoichiometric bases and additives, were essential to enable the formation of CO₂ ^•–, arguably due to the high kinetic barrier caused by the required change in geometry. In a significant effort, Maiti and Audisio recently reported an elegant procedure to obtain CO₂ ^•– from CO₂ via in situ generation of formate, merging photoredox catalysis with stoichiometric hydride and hydrogen atom transfer (Figure A).

1.

CO₂ as a feedstock in carboxylation reactions. (A) CO₂ activation modes in organic synthesis. (B) This work: New CO₂ activation mode for hydrocarboxylation reactions.

Herein, we report a new CO₂ activation mode for hydrocarboxylation reactions (Figure B). Exploiting the classical polar reactivity of carbon dioxide with nucleophiles, we designed a catalytic system in which a photoactive carbamate is formed in situ between CO₂ and a phenothiazine catalyst. Upon irradiation with visible light, this transient carbamate is photolyzed to release CO₂ ^•–, which is then employed in hydrocarboxylation reactions. A variety of alkenes and heterocycles react smoothly with CO₂ to afford the corresponding carboxylic acids. Feedstock and complex cyclic ketones can be readily converted with a two-step process into valuable spirocyclic structures in good yields, even in the context of pharmaceutical cores. Finally, this new method was utilized to synthesize ¹³C-labeled lactones using ¹³CO₂, showcasing the potential for isotopic labeling applications.

Design plan

From the outset of our investigation, we recognized that the formation of a carbamate intermediate through the covalent interaction between CO₂ and a N-centered nucleophile would already set the trigonal geometry for the potential generation of CO₂ ^•–, avoiding the high kinetic barrier due to the geometry transition in the case of direct electron transfer reduction.

We hypothesized that if a proper photoactive molecule is employed as the nucleophile, then the carbamate could undergo photolysis to liberate the CO₂ ^•–, along with the radical species of the nucleophile. However, given the propensity of CO₂ ^•– to form undesired products such as oxalate and multicarboxylated species, the concentration of this transient carbamate must be kept low for its application in synthetic reactions. Therefore, we proposed the use of a photoactive nucleophilic catalyst, which would comprise the following features: 1) easily obtainable in the reaction media from a stable precursor; 2) exhibits absorption in the visible region; 3) able to perform catalytic redox cycles and 4) deliver a persistent radical after the photolysis process, providing a driving force for the homolytic fragmentation. We identified benzophenothiazine (BPTZ) anion 1 as a suitable photoactive nucleophile for our novel activation mode toward the generation of CO₂ ^•– (Figure ). Previous work showed the ability of phenothiazine derivatives to absorb visible light and participate in redox catalytic cycles. − We anticipated that in a saturated atmosphere of carbon dioxide, a carbamate intermediate is initially obtained between BPTZ anion 1 and CO₂. It is reasonable to assume that the absorption properties of carbamate 2 would be similar to those of neutral BPTZ. Therefore, upon absorption of visible light, excited carbamate 3 undergoes photolysis to generate CO₂ ^•– along with open-shell species 4, which is known to be a persistent radical. , The BPTZ anion 1 is then restored via a single electron transfer by an appropriate reductant.

2.

Design plan.

To demonstrate this conceptually distinct activation mode as a generic platform for installation of CO₂ into organic molecules, we have initially selected to perform a carboxylation reaction of hydroxyalkenes 6 for the direct synthesis of γ- and δ-spirolactones. Spirolactones, especially the γ- and δ- analogues, have attracted considerable interest in scientific research due to their diverse biological activities and potential therapeutic applications. − Several synthetic methods are available for the construction of the γ-spirolactones: (i) esterification, (ii) Baeyer−Villiger oxidation, (iii) radical cyclization of oxalates, , (iv) alkenes reaction with carboxylic acid derivatives − or CO₂ ^•–/carbonyl compounds, (v) alcohols/ketones addition-cyclization with α,β-unsaturated carbonyl compounds. , On the contrary, only a few protocols allow the spirolactonization in the case of the δ-spirocyclic structures: (i) esterification, (ii) Baeyer–Villiger oxidation, and (iii) organocatalytic cyclization of isatins with in situ generated conjugated enolates. Despite the considerable number of synthetic reactions, the lactone carbonyl group is always preinstalled in the starting materials, normally by multistep synthesis and oxidation, except for limited examples of metal-catalyzed carbonylation using high pressure of toxic carbon monoxide. , Therefore, the development of a general unified strategy for the synthesis of γ- and δ-spirolactones using simple starting materials such as cyclic ketones and CO₂ is highly desirable yet elusive.

We sought to address this synthetic challenge with a novel catalytic activation strategy. Initially, cyclic ketone 5 is converted to the corresponding hydroxyalkene 6 via a high-yielding Grignard or Barbier reaction, depending on whether the final molecule is a γ- and δ-spirolactone (Figure ). The addition of highly reactive CO₂ ^•– to alkene 6 in an anti-Markovnikov fashion furnishes carbon-centered radical intermediate 7. The hydroxycarboxylate 8 is then obtained via hydrogen atom transfer (HAT) with γ-terpinene, followed by an intramolecular esterification to provide the desired spirolactone 9. Simultaneously, the persistent radical 4 (E _1/2 = 0.18 V vs SCE, please see Supporting Information (SI) for details) is reduced by the γ-terpinene radical (E _1/2 = −0.1 V vs SCE), restoring the benzophenothiazine anion 1.

Results and Discussion

Initially, we performed a series of experiments to verify the feasibility of our new activation mode. We conducted UV–vis absorption and emission studies to characterize and compare the spectroscopic properties of benzophenothiazine (BPTZ) 10, the corresponding anion 1 and CO₂ carbamate 2 (Figure A and B). When an excess of potassium tert-butoxide was added to a solution of benzophenothiazine 10, we observed a change in color from pale green to wine red, indicating the formation of the benzophenothiazine anion 1. As expected, when CO₂ was bubbled through solution containing anion 1, the wine-red color changed to bright green, showing a relatively similar yet different absorption spectra to that of the neutral benzophenothiazine 10, revealing the formation of carbamate 2. Once certain of the ability of carbamate 2 to absorb visible light, we exposed presynthesized carbamate 11 to 390 nm LED light in the presence of γ-terpinene and DMF under argon atmosphere (Figure C). After 16 h, we were pleased to observe the formation of formate in 30% yield, supporting the generation of CO₂ ^•–. Formate is the result of only one of the possible CO₂ ^•– termination pathways and therefore does not represent the complete mass balance of CO₂ ^•– formation. Importantly, when the same experiment was performed without light, no formate was detected. Next, we sought to gain additional evidence for the formation of CO₂ ^•– without preforming the carbamate, but simply using a solution of benzophenothiazine 10, potassium tert-butoxide, γ-terpinene and CO₂ (Figure D). The reaction was monitored over time with an FT-IR spectrometer. After 10 s, the IR spectra clearly showed the characteristics bands of carbamate 2 (1661 and 1644 cm^–1, see SI for details), confirming its formation in solution. Over the course of the reaction, carbamate 2 and CO₂ (2360 cm^–1) IR bands gradually disappeared in favor of a new band at 1610 cm^–1, assigned to the CO₂ ^•– or formate. − The presence of formate was also confirmed by NMR studies at the end of the reaction (see SI for details). When the same experiment was repeated in the dark, the only detected species was carbamate 2. Moreover, we were able to detect oxalate using FT-IR in the absence of γ-terpinene. Taken together, these preliminary experiments strongly suggest the formation of CO₂ ^•– via photolysis of carbamate 2 under visible light irradiation.

3.

Preliminary experiments. (A) UV–vis studies of benzophenothiazine 10, corresponding anion 1 and CO₂ carbamate 2. (B) Emission studies of benzophenothiazine 10, corresponding anion 1 and CO₂ carbamate 2. (C) Detection of formate upon irradiation of presynthesized CO₂ carbamate 11. (D) FT-IR monitoring of the reaction. BPTZ = benzophenothiazine.

Having confirmed the generation of CO₂ ^•– using benzophenothiazine 10 and carbon dioxide, we proceeded to study the catalytic anti-Markovnikov carboxylation of hydroxyalkene 12 (Table ). To our delight, when a DMF solution of 12, benzophenothiazine 10 (10 mol %), t-BuOK (20 mol %) and γ-terpinene was backfilled with 1 atm of CO₂ and exposed to 390 nm LED light, the desired γ-spirolactone 13 was obtained in 75% isolated yield (entry 1). Catalyst, solvent, base, and HAT source were key parameters for the optimization studies. Variation on the phenothiazine core revealed the superiority of conjugated BPTZ 10 compared to analogues 14–16 in terms of yield (entries 2–4). Polar solvents such as DMSO and NMP showed the formation of the product in comparable yields with DMF, while in acetonitrile the γ-spirolactone was only detected in 37% yield (entries 5,6). t-BuOK was found to be the best base in terms of reaction outcome (entries 7, 8), while full conversion of the starting hydroxylalkene was achieved at 0.2 M concentration (please see SI for details). As expected, γ-terpinene and 1,4-cyclohexadiene were the only competent HAT sources under the reaction conditions, arguably because of their matched potential to reduce persistent radical 4 (entries 9, 10). Additional control reactions demonstrated that purple light and catalyst are fundamental for the observed reactivity (entries 11 and 12), consistent with the mechanistic blueprint outlined in Figure . The reaction without base provided the product in 38% yield (entry 13), probably due to the inefficient formation of the carbamate. Other well-established CO₂ carboxylation methods have failed to furnish product 13 in yields higher than 30%, demonstrating the superiority of our protocol for the synthesis of spirolactones (please see SI for details).

1. Optimization and Control Reactions .

^aReaction conditions: 12 (0.2 mmol), photocatalyst (10 mol %), base (20 mol %), HAT donor (0.6 mmol), in solvent (0.2 M) at rt for 21 h, 390 nm LED.

^bNMR yields using methyl 3,5-dinitrobenzoate or 1,1,2,2-tetrachloroethane as internal standard. BPTZ = benzophenothiazine.

With optimal conditions in hand, we first examined the generality of our new activation mode in terms of radical carboxylation for the synthesis of γ- and δ-spirolactones. As is evident from the results compiled in Figure , our mild carboxylation via CO₂ ^•– could be conducted on a wide variety of hydroxyalkenes derived from feedstock cyclic ketones. Starting ketones of different sizes (13, 17–20) and heterocyclic analogues (23–26) were readily converted to the corresponding γ-spirolactones in moderate to good yields. Substituents on the starting cyclic ketone (21) and sterically hindered structures such as adamantanone (22) were also tolerated. Viable motifs in this transformation include N-Boc protected azetidinone (23), thiethanone (24) and tetrahydro­(thio)­pyranone (25, 26) derivatives. Notably, the use of molecules already characterized by the presence of a spirocenter allows the synthesis of geometrically intricated structures bearing multiple spirocenters (27, 28), thus elevating molecular three-dimensionality and complexity, key attributes known to enhance the potential of new structures for pharmaceutical applications. Both electron-rich and electron-poor substituents on the phenyl moiety posed no problems (30–34), including a heteroaromatic ring such as thiophene (35), providing the corresponding spirocyclic product in moderate yields. Interestingly, we also observed that the vinylsulfonylbenzene motif can function as a masked C2-synthon for the synthesis of an unsubstituted spirolactone. When this feature was installed on 3,3-diphenylcyclobutanone, the corresponding hydroxyalkene (36) smoothly underwent radical carboxylation and reductive desulfonylation, resulting in the formation of the unsubstituted γ-spirolactone (37) in 42% yield after only 3 h.

4.

Synthesis of γ-spirolactones. Reaction conditions as in Table , entry 1, 0.2 mmol scale. Isolated yield unless otherwise noted. ^‡5 mmol scale. ^†2 lamps, 60 °C, 48 h. ^§NMR yield using 1,1,2,2-tetrachloroethane as an internal standard. ^#0.05 M, 3 h. *Isolated as mixture of diastereoisomers. BPTZ = benzophenothiazine.

Importantly, we were pleased to find that this method is also amenable to substrates containing pharmaceutical cores such as ibuprofen (38), gemfibrozil (39) and ciprofibrate (40). Our reaction could also be applied to the steroid derivative estrone (41), leading to the synthesis of a new potential candidate in the class of the antimineralocorticoid 17α-spirolactosteroids. , The excellent diastereoselectivity observed using estrone is arguably under kinetic control due to the irreversibility of the diastereodetermining HAT process.

Remarkably, our protocol was found to be applicable to the more challenging synthesis of δ-spirolactones (Figure ). To prevent polymerization and degradation of the starting material, a combination of higher dilution and catalytic amount of tetrabutylammonium bromide was necessary. Pleasingly, we were able to derivatize most of the cyclic ketones used in Figure and convert them to the desired δ-spirolactones. Ring sizes up to 12 members (42–47), molecules bearing multiple spirocenters (49) and different heterocyclic structures (53–55) were obtained in moderate to good yields. The methodology showed good compatibility with a variety of moieties such as geminal difluoro substituents (50), ketone (51) and unactivated alkene (52). Our method could be successfully employed for the synthesis of fused lactone 58 and the derivatization of linear acyclic ketones (59–61), showing progressive improvements in yields with an increase in the alkyl chain length.

5.

Synthesis of δ-spirolactones and carbon-13 labeling. Reaction conditions: hydroxyalkene (0.2 mmol), BPTZ 10 (10 mol %), t-BuOK (20 mol %), TBAB (30 mol %) γ-terpinene (0.6 mmol), in DMF (0.05 M) at rt for 21 h, 390 nm LED. Isolated yield unless otherwise noted. ^†Reaction conditions are the same as Table , entry 1. *Isolated as mixture of diasteroisomers. BPTZ = benzophenothiazine. TBAB = tetrabutylammonium bromide.

Given the simplicity of our newly developed method, we set out to explore the possible utilization of ¹³CO₂ for the synthesis of carbon-13 labeled spirolactones. In particular, the direct generation of labeled reactive species such as ¹³CO₂ ^•– represents a powerful tool toward direct and versatile site-specific incorporation of ¹³CO₂ into organic backbones, that only recently has shown its first applications with the synthesis of labeled carboxylic acids. , − The isolation of ¹³C-labeled organic compounds is indeed crucial for a precise tracking of the molecular transformations of a target molecule in fields such as fundamental biology, metabolomics, and hyperpolarized magnetic resonance imaging. − Given the practical flexibility of the reaction setup, we anticipated that our new radical carboxylation protocol could be used for the streamline synthesis of ¹³C enriched spirolactones. Replacing the atmosphere of CO₂ with ¹³CO₂, we successfully provided the ¹³C-labeled products via direct generation of ¹³CO₂ ^•–, including γ- and δ-spirolactones (62, 63), fused δ-lactone (64) and spirolactones derived of pharmaceutical cores (65, 66).

Finally, we were delighted to see that our protocol is not limited to hydroalkenes, demonstrating the broad applicability of our CO₂ activation mode (Figure ). More conventional alkenes such as styrenes (67–71), acrylates (74–76) and acrylamides (77) effectively undergo hydrocarboxylation. Carboxylic acid and the fluoxetine core were also well tolerated, furnishing desired products 72 and 73 in synthetically useful yields.

6.

Substrate scope for the radical hydrocarboxylation of more conventional alkenes and heterocycles. Reaction conditions: hydroxyalkene (0.2 mmol), BPTZ 10 (10 mol %), t-BuOK (20 mol %), γ-terpinene (0.6 mmol), in DMF (0.2 M) at rt for 21 h, 390 nm LED. Isolated yield unless otherwise noted. BPTZ = benzophenothiazine.

Unfortunately, only traces of the product were observed when unactivated alkenes were used, likely due to the combination of slow and reversible addition of the CO₂ ^•– to unactivated olefins and the polarity mismatch of the HAT process. However, electron-rich and electron-poor heterocycles could be readily converted to the corresponding semiunsaturated carboxylic acids (78–80) in good yields.

Mechanistic Studies

While the preliminary studies described in Figure strongly suggest that the formation of CO₂ ^•– via photolysis of carbamate 2 is feasible, we conducted additional experiments to confirm that this process is indeed the main pathway for the generation of CO₂ ^•– and the hydrocarboxylated product under our optimized reaction conditions.

Formate was detected both in the absence and in the presence of hydroalkene 12, with TON of 5 and 35 respectively, supporting the catalytic generation of CO₂ ^•– (Figure A, entries 1 and 2). Control experiments in the dark without CO₂ or γ-terpinene did not show any formate. These results confirm that (i) formate is generated only as a consequence of the CO₂ ^•– and (ii) γ-terpinene is not a hydride donor under our reaction conditions, excluding any possible formation of formate directly from CO₂. To rule out the possibility of formate as productive intermediate in our hydrocarboxylation strategy, we performed a control experiment where 2 equivalents of sodium formate were used instead of CO₂ (Figure B). As expected, we were unable to detect product 13. The formation of product 82 in a radical clock experiment using the cyclopropyl styrene substrate 81 confirmed the radical nature of our method (Figure C).

7.

Mechanistic studies. (A) Additional experiments supporting the catalytic generation of CO₂ ^•–. (B) Control experiment using sodium formate instead of CO₂. (C) Radical clock experiment. (D) Control experiments using a BPTZ unable to generate the carbamate. (E) Estimated contribution of different reaction mechanisms. BPTZ = benzophenothiazine. TON = turnover number.

Once it was established that CO₂ ^•– is catalytically generated under our optimized reaction conditions, we conducted a series of control experiments to confirm that carbamate photolysis is the main pathway for the formation of the hydrocarboxylated products (Figure D). When N–H benzophenothiazine 10 was replaced with N-phenyl benzophenothiazine 83, which is unable to generate the key photoactive carbamate 2, we detected γ-spirolactone 13 in a low 33% yield. The same experiment was repeated using benzothiophene 84 as the starting material, where we could only detect traces of the corresponding hydrocarboxylated product 79. While the second experiment clearly supports the crucial role of the carbamate and the proposed photolysis, the first experiment indicates that another minor mechanism is contributing to the formation of product 13. Formate was not observed in either reaction, suggesting that the alternative mechanism for the formation of product 13 is not the direct reduction of CO₂, but the reduction of styrene 12. Notably, although styrene substrates having a lower reduction potential than CO₂ (−2.58 V vs −2.2 V vs SCE respectively), the photocatalyst is only able to reduce 12, confirming the prohibitive kinetic requirement to directly reduce CO₂. When the optimized reaction was repeated under argon, we observed the formation of product 85 in only a 24% yield (Figure D).

Since product 85 is generated via direct reduction of substrate 12, we can estimate that the maximum contribution of this pathway to the formation of spirolactone 13 is 32%, confirming that the photolysis of carbamate 2 is the major productive mechanism under our optimized reaction conditions (Figure E). Finally, a quantum yield of 0.07 was obtained using the ferrioxalate chemical actinometer, consistent with the proposed mechanism (see SI for details).

To further evaluate the kinetic feasibility of the proposed activation mode, we performed DFT calculations on the photoinduced formation of 4 and CO₂ ^•–K⁺ from potassium carbamate 2, involving excitation from the ground state (S₀) to the singlet excited state (S₁), intersystem crossing to the triplet state (T₁) and subsequent evolution along the T₁ surface (Figure , please see SI for details). Formation of potassium carbamate 2 from the corresponding BPTZ anion 1 and CO₂ was found to be almost barrierless and exergonic at the employed level of theory, affording four conformers (2 _A‑D ) in fast equilibrium, depending on the position of the potassium cation. This energetic profile indicates that, under a CO₂ atmosphere and steady-state photocatalytic conditions, the most stable conformer, carbamate 2 _A , is the dominant ground-state resting species and therefore the main photoactive intermediate, while the free BPTZ anion 1 is likely present only transiently as it is rapidly intercepted by CO₂. Upon light irradiation, carbamate 2 _A undergoes vertical excitation to the lowest singlet excited state (S₁) which after vibrational relaxation yields the corresponding S₁ minimum.

8.

Computed free energy profile for the formation of potassium carbamate 2 in the ground state (S₀, violet), photoinduced excitation to the singlet excited state (S₁, pink), intersystem crossing to the triplet excited state (T₁, red), and subsequent evolution to form 4 and CO₂ ^•–K⁺. Computational studies were performed at wB97X-D/6-31G­(SMD = DMF)// wB97X-D/6-31G theory level. All energies are DG_sol reported in kcal mol^–1 relative to those of 2 _A (S₀). Energy values marked with an asterisk correspond to electronic energies (DE_sol) reported in kcal mol^–1 relative to 2 _A (S₀). BPTZ = benzophenothiazine. hν = photon. MECP = minimum energy crossing point. ISC = intersystem crossing.

A minimum energy crossing point (MECP) between the S₁ and T₁ potential energy surface was located 9.8 kcal mol^–1 above the 2(S₁) minimum, demonstrating the feasibility of the S₁–T₁ intersystem crossing. Optimization of the potassium carbamate from the MECP geometry on the triplet potential energy surface afforded species 2(T₁), from which two nearly barrierless pathways were identified for homolytic cleavage of the C–N bond leading to 4(S₀) and CO₂ ^•–K⁺ (ΔG^‡ = 2.9 and 1.2 kcal mol^–1, respectively; ΔG⁰ = −37.0 kcal mol^–1). These computational results, together with the experimental findings presented in Figure and , support a mechanism in which the photolysis of carbamate 2 constitutes the main pathway for the hydrocarboxylation of unsaturated systems.

Conclusions

In summary, we have developed a new catalytic CO₂ activation mode for hydrocarboxylation reactions. The formation of photoactive CO₂ carbamate, which set the required trigonal geometry, allows the generation of CO₂ ^•– under mild reaction conditions. The protocol exhibits a wide functional group tolerance and broad substrate scope, even in the context of biologically active molecules. Furthermore, this technology was successfully applied to the carbon-13 isotope labeling of γ- and δ-spirolactones. The general principle of photolytic nucleophilic activation of electrophiles to generate radicals is expected to pave the way for the development of new synthetic reactions. Exploration of new avenues is currently underway in our laboratories.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c21208.

• General information; experimental procedures; characterization data for all new compounds; mechanistic experiments; DFT calculations; and relevant coordinates of optimized geometries. (PDF)

The authors declare no competing financial interest.