Radical Chain Reduction via Carbon Dioxide Radical Anion (CO₂^•–)

Abstract

We have developed an effective method for reductive radical formation that utilizes the radical anion of carbon dioxide (CO₂^•–) as a powerful single electron reductant. Through a polarity matched hydrogen atom transfer (HAT) between an electrophilic radical and a formate salt, CO₂^•– formation occurs as a key element in a new radical chain reaction. Here, radical chain initiation can be performed through photochemical or thermal means, and we illustrate the ability of this approach to accomplish reductive activation of a range of substrate classes. Specifically, we have employed this strategy in the intermolecular hydroarylation of unactivated alkenes with (hetero)aryl chlorides/bromides, radical deamination of arylammonium salts, aliphatic ketyl radical formation, and sulfonamide cleavage. We show that the reactivity of CO₂^•– with electron-poor olefins results in either single electron reduction or alkene hydrocarboxylation, where substrate reduction potentials can be utilized to predict reaction outcome.

Graphical Abstract

Single electron reduction is a key mechanistic step in the formation of radical species from many different electron-poor substrate classes. In this context, dissolving metal conditions^1,2 and electrochemical techniques^3,4 are outstanding approaches for accomplishing challenging reductions (those of substrates with very negative reduction potentials). However, these strategies are limited in their ability to promote free radical reactivity. Because sequential two-electron reduction occurs rapidly at metal surfaces, overreduction of transient radicals to the corresponding anions occurs. This principle is particularly acute in reductive activation of organic halides,⁵ as depicted in Figure 1. In contrast, single electron reduction using photoredox catalysis^6,7 allows for precise tuning of kinetic factors (e.g. reduction potential, effective reductant concentration), which enables the interception of transient radicals. Although the use of visible light as an energy source is highly appealing, reducing power is inherently limited by the energy contained in a blue photon (up to 2.8 eV).⁸ To overcome this, strategies that involve photoexcitation of radical anions have been recently developed to yield superreductants that readily perform a range of challenging reductions.^9–11 We were interested in developing a new radical chain approach to single electron reduction. We understood that the value of this technology would largely depend on its ability to engage unreactive substrates and grant reliable access to radical intermediates in a simple manner.

Figure 1.

Novel approach to SET through reactivity of CO₂^•–

The carbon dioxide radical anion (CO₂^•–) is a highly reactive intermediate that is routinely accessed through electrochemical reduction of CO₂.^12,13 Given the combination of its very negative reduction potential (E_1/2° = –2.2 V vs SCE)¹⁴ and reactive radical character, CO₂^•– is a potentially valuable synthetic intermediate. However, it has not yet been generally deployed in this context. Early studies by Kubiak¹⁵ and Barba¹⁶ demonstrated that single electron reduction of CO₂ can be utilized in carboxylation processes under photochemical or electrochemical conditions, respectively. More recently, the Jamison group developed a continuous flow-based approach that allows for photochemical production and interception of CO₂^•–.¹⁷ We have developed a novel system that utilizes the reducing power of CO₂^•– to accomplish difficult reduction events that is centered around the ability of thiol catalysts to shuttle hydrogen atoms. This radical chain mechanism allows for direct production of CO₂^•– through hydrogen atom transfer (HAT) from formate. Under these conditions, CO₂^•– acts as a potent reductant for single electron activation of many different substrate classes that are typically redox inert (shown in Figure 1). This method utilizes inexpensive inputs (no metal required), and accomplishes a range of valuable radical processes.

We envisioned the mechanistic scenario that is outlined in Figure 2A, where the radical chain is comprised of hydrogen shuttling from a thiol catalyst. Reaction of thiyl radical with formate affords the key reductant (CO₂^•–) that is able to accomplish general substrate reduction. For example, single electron transfer to an aryl chloride would give rise to the corresponding arene radical anion, along with carbon dioxide (providing a thermodynamic driving force). This HAT event is essentially thermoneutral (formate C–H BDE: 88 kcal/mol, thiol S–H BDE: 88–90 kcal/mol, see SI). Mesolytic C–Cl fragmentation would generate the corresponding aryl radical, which would undergo reduction through H-atom abstraction from the catalytic thiol, propagating the radical chain. Importantly, as indicated in Figure 2B, the principal feature of this approach is thiyl formation in the presence of formate. This can be conveniently accomplished from a thiol using the combination of a photoredox catalyst and visible light. In addition, chain initiation can be easily conducted by other means. For example, direct photolysis of disulfides or thermal decomposition of persulfate (in the presence of catalytic thiol) are equally effective.

Figure 2.

A) Proposed mechanistic scenario for radical chain SET activation of aryl chloride substrates. B) Methods for radical chain initiation. C) Experimental data in support of this radical chain proposal.

To interrogate this proposal, we considered the reductive dechlorination of methyl-2-chloro benzoate (Figure 2C). Subjecting this aryl chloride to sodium formate (5 equiv) in the presence of different combinations of photoredox catalysts and HAT catalysts resulted in radical hydrodechlorination. With mesna (an inexpensive and odorless alkyl thiol), a range of photoredox catalysts accomplished this transformation in nearly quantitative yield (as determined by ¹H NMR). These data illustrate the value of this radical chain approach; while SET from CO₂^•– to substrate would be expected to occur with ease (based on reduction potential), SET from these photocatalysts would be highly unfavored. Indicated in Figure 2C are the most strongly reducing states (ground state radical anions, given by reductive quenching), which are at least ca. 600 mV more positive than that of the chloride (–2.1 V vs SCE).¹⁸A range of thiols were found to perform this reaction as was DABCO (albeit to a lesser extent), and no conversion was observed in the absence of HAT catalyst. While light and formate were critical to substrate conversion under these conditions (see the SI for details), quantitative reduction was observed when thiyl generation was performed via photolysis of dimethyl disulfide (no photoredox catalyst) or with thiol and persulfate (20 mol%) at 100 °C. However, because trace amounts of thiol impurities that are present in DMSO^19–21 are capable of promoting this reaction (indeed, these experiments were conducted with freshly purified DMSO (indeed, solvent from particularly old bottles was capable of promoting this reactivity without added thiol co-catalyst). We found the quantum yield (Φ) of the reaction under our standard conditions (P1, mesna) to be 2.63, which strongly indicates radical chain character. Using transient absorption spectroscopy (see SI for details), we determined the rate of thiyl formation under the standard (photoredox initiation) conditions (4.0 × 10³ M^–1•s^–1), as well as the quantum yield of the triplet excited state (Φ= 0.0072). These rates indicate the average radical chain length to be greater than or equal to 365, under standard conditions.

With the understanding that CO₂^•– can be accessed under mild conditions, we sought to utilize this radical chain reduction in other valuable synthetic processes. First, we considered the intermolecular coupling of aryl radical species with unactivated olefins (radical hydroarylation). As shown in Table 1, reaction of 4-chlorobenzonitrile (E_p/2 = –2.1 V vs SCE)¹⁸ with 1-octene afforded alkylation product 1 in 66% yield with photoredox initiation. Though slightly less efficient for intermolecular radical coupling, alternative initiation systems also promoted the desired reactivity (disulfide: 50% yield, persulfate: 42% yield). We have reported a number of radical hydroarylation systems;^22–24 however, these conditions are unique in that they allow for coupling of benzene-derived radicals with simple olefins. Monosubstituted alkenes bearing alcohol and alkyl chloride functional groups reacted to give alkylated products 2 and 3 in 46% and 85% yield, respectively, where the remaining material was direct arene reduction. The more electron-rich olefins 2-methyl-2-butene and isopropenyl acetate also underwent smooth hydroarylation to yield 4 (66%) and 5 (81%). Evaluation of the aryl radical scope with 1-octene revealed that a number of electron-poor aryl and heteroaryl chlorides can be engaged under this protocol (6–14, 32–80% yield) with the primary limitation being the diminished yield with ortho-chloromethylbenzoate (10, 32% yield). In addition, aryl radical coupling reactions employing bromobenzene, 4-chlorobenzonitrile, and 2-bromothiazole all smoothly engaged tert-butylvinyl carbamate (3 equiv) to afford the corresponding phenethylamines (15–17, 74–98% yield).

Table 1:

Radical Reactivity via CO₂^•– Radical Chain Conditions: Coupling of Radicals with Olefins and Hydrogen Atom

^aReaction conditions: substrate (1 equiv), olefin (5 equiv), P1 (1 mol%), mesna (20 mol%), sodium formate (5 equiv), MeCN/DMSO (1:1, v/v), blue light, 16 h.

^bYield determined by ¹H NMR with internal standard.

^cReaction conducted with aryl bromide.

^dReaction conducted at 100 °C.

^eDMSO used as solvent.

^f20% H₂O/DMSO (v/v) used as solvent.

To further illustrate the value of this radical chain, we applied it to the reduction of other challenging substrates. For example, aryl trimethylammonium salts have been demonstrated as electrophilic coupling partners²⁵ that also undergo deamination under Birch conditions. Under standard conditions, we found that electron-poor arylammonium salts undergo clean deamination (18–20, 70–94% yield). At present, this process is limited to substrates that contain electron-withdrawing groups in the meta-position. This limitation stems from competitive nucleophilic demethylation processes and challenges associated with trialkylammonium salt formation. Nicewicz recently demonstrated a consecutive photoinduced electron transfer process of acridinium salts that effectively cleaves arylsulfonamide N–S bonds.⁹ We found that CO₂^•– is a competent reductant in achieving the same, where radical chain deprotection gave 21–23 in 52–84% yield. This system also accomplishes defluoroalkylation of trifluoromethyl aromatics, a strategy that our group has developed,^26,27 to give difluoroalkyl products 24–26 (30–59% yield). Finally, we found that these conditions effectively achieve reduction of aliphatic aldehydes, where ketyl formation is presumably followed by protonation and HAT to deliver the corresponding alcohol products (27–29, 70–87% yield).

Photochemical reduction of electron-deficient olefins has been established as a powerful method for directly accessing radical character at the β-position of Michael acceptors.^28,29 We posited that our radical chain method would reductively engage a broad range of olefin substrates, through SET (Figure 3). Indeed, we observed efficient transformation of a pyridyl cinnamate ester (E_p/2 = –1.6 V vs SCE) to the hydrogenated product 34 (95% yield). To understand this reductive process, we prepared a β-cyclopropyl enone substrate with a reduction potential within range of CO₂^•–. Reaction of this system under standard conditions resulted in effective ring opening, giving rise to 33 (79% yield), in strong support of the proposed SET event. With the analogous ester, a substrate outside the reducing power of the CO₂^•–, we instead observed very clean formation of Giese type addition product 32 (66% yield), where cyclopropyl ring opening was not observed. These data indicate that olefin reduction potential and reaction outcome are correlated, where more electron-poor olefins (E_1/2° ≥ – 2.1 V vs SCE) undergo SET and less electron-poor olefins (E_1/2° ≤ – 2.1 V vs SCE) undergo radical hydrocarboxylation. In line with this assertion, CO₂ incorporation was observed with both an α,β-unsaturated nitrile and amide to afford the subsequent hyrdocarboxylated products (31 to 30, 67–69% yield) with complete regiocontrol. While electrophilic carboxylation reactions are commonplace (e.g. nucleophile trapping with CO₂), this approach leverages nucleophilic behavior of CO₂^•– in the construction of 1,4-dicarbonyls, motifs that are valuable in the synthesis of hetoercycles,^30–32 and bioactive small molecules.³³

Figure 3.

Reaction of carbon dioxide radical anion with electron-deficient alkenes: Divergent reactivity based on reduction potential

In summary, we have developed a novel radical chain mechanism in which CO₂^•– is easily produced from formate through thiol-mediated HAT. This highly-reducing intermediate readily activates a broad array of challenging substrates, where single electron reduction grants access to radical reactivity. This radical chain can be initiated using either photochemical or thermal pathways, where SET reactivity generically occurs with substrates with reduction potentials ≥ – 2.1 V vs SCE. In the presence of Michal acceptors that possess more negative reduction potentials, we observed clean hydrocarboxylation, demonstrating the nucleophilic character of CO₂^•–.