Abstract

Radical chain initiation strategies are fundamental to the synthesis of small molecule drugs and macromolecular materials. Modern methods for initiation through one-electron reduction are largely dominated by photo- and electrochemistry but the large-scale industrial application of these methods is often hampered by scalability challenges. Here we report a general, thermally driven and scalable method for the reductive initiation of radical chains that involves reacting an inexpensive azo initiator with a formate salt to form a carbon dioxide radical anion. Substoichiometric quantities of this initiator system were used to form C(sp2)–C(sp3), C(sp2)–S, C(sp2)–H, C(sp2)–B and C(sp2)–P bonds from complex (hetero)aryl halides, with high chemoselectivity and under transition-metal-free conditions. The developed initiator system was also used to probe the mechanism of other radical reactions.

Main

The controlled initiation of radical chains is a subject of fundamental importance to polymer science1, organic synthesis2,3, atmospheric chemistry4 and biochemistry5. In the context of synthetic organic chemistry, a substantial proportion of radical reactions are driven by chains6,7, but this aspect is often obscured—especially if the length of the chain is short. If the chain length is short, the reaction must be continuously reinitiated for its duration8. This is a major reason why photo- and electrochemistry have emerged as leading methods of initiation: they provide flexible and programmable frameworks to promote electron transfer and generate radicals at a controllable rate9,10,11,12,13. One drawback of photo-/electrochemical initiation strategies is their non-trivial scalability in a process development and manufacturing setting (beyond the milligram to gram scales used in earlier stages of medicinal chemistry)14,15. Indeed, both strategies require specialist reactor technologies, which may increase process development times and costs, limiting applications on the industrial scale when compared to standard manufacturing techniques (Fig. 1a). However, for chains driven by one-electron reduction (electron-catalysed and electron-transfer chain processes)16,17 photo- and electrochemical methods still far surpass the general utility of chemical/thermal initiation strategies using ground state electron donors18,19. Considering this current state-of-the-art, we sought to develop a general reductive initiation system with the following characteristics: (1) strongly reducing (({{E}_{1/2}^{\circ}} < -2,{\rm{V}}) versus saturated calomel electrode (SCE)); (2) usable in substoichiometric quantities; (3) thermally controlled; (4) does not form chain-terminating persistent radicals (a traceless one-electron reductant); (5) compatible with a broad range of substrates; (6) applicable to a wide range of reactions; and (7) inexpensive and scalable using standard manufacturing vessels.

Fig. 1: Background and mechanistic hypothesis.

a, One-electron reduction initiation strategies. b, Proposed thermal strategy using azo initiators to generate carbon dioxide radical anion. PC, photocatalyst; D, donor; ({E}_{1/2}^{\circ}), half-wave potential.

In this study, we describe the realization of this goal by heating inexpensive azo initiators commonly used in the polymer industry20 in the presence of formate salts to generate a strong one-electron reductant, carbon dioxide radical anion (CO2**·−, ({{E}_{1/2}^{\circ }}=-{2.22},{\rm{V}}) versus SCE)20,21. Inspired by advances in the photochemical generation of CO2•− (refs. 22,23,24,25,26,27,28,29,30), we hypothesized that electrophilic α-cyano alkyl radicals (derived from the thermal decomposition of an azo initiator) would readily abstract hydrogen atoms from formate salts to from CO2·**− in a polarity-matched process (Fig. 1b). The proposed combination of an azo initiator and formate salt would provide an operationally simple ‘dump and stir’ thermal method of initiation, which would satisfy the aforementioned criteria and be widely applicable across synthetic chemistry.

Results and discussion

Initiator and model reaction development

To evaluate the feasibility of our mechanistic proposal, we monitored the reaction of an azo initiator and sodium formate by electron paramagnetic resonance (EPR) spectroscopy in the presence of the spin trap, 5,5-dimethyl-1-pyrroline-N-oxide (DMPO). 4,4-Azobis(4-cyanovaleric acid) (ACVA) was selected as the azo initiator, as unlike azobisisobutyronitrile (AIBN), ACVA is not classified as an explosive. When reacting ACVA, HCO2Na and DMPO, a species matching previously reported data for a DMPO–CO2**·− adduct was detected (Fig. 2a)31. To verify if this species was indeed a CO2·− adduct, these experiments were repeated using H13CO2Na and the expected additional 13C hyperfine coupling was clearly observed (aiso13C = 33 MHz). These results strongly suggest that CO2·**− is formed from the reaction of ACVA and sodium formate. The feasibility of polarity-matched hydrogen-atom transfer (HAT)32 between the ACVA-derived α-cyano alkyl radical I and formate II was assessed by density functional theory (DFT), which indicated that HAT is both kinetically and thermodynamically viable (ΔG‡ = 14.5 kcal mol−1, ΔG = −3.6 kcal mol−1; Fig. 2b). It is possible that the distonic nature of I and interactions between the α-cyano radical and carboxylate anion may play a role in the efficiency of this electron upconversion process (the conversion of a weak/mild reductant into a stronger reductant product)17, but any such interactions are probably weakened by the highly polar dimethylsulfoxide (DMSO) solvent environment33,34,35,36.

Fig. 2: Hypothesis validation and reaction development.

a, EPR spectroscopy to probe the formation of CO2**·**−. b, Computational study of HAT conducted at the M062X-D3(0) def2-TZVP SMD(DMSO) level of theory. c, Optimization of the proposed carbonyl α-arylation reaction. giso, isotropic g value; aiso, isotropic a value (the hyperfine coupling constant); Ar, aryl; ACHN, 1,1’-azobis(cyclohexane-1-carbonitrile); AAPH, 2,2’-azobis(2-amidinopropane) dihydrochloride; AIPH, 2,2′-azobis[2-(2-imidazolin-2-yl)propane] dihydrochloride; NA, not applicable.

Having validated our key mechanistic hypothesis, model synthetic applications were explored to determine the generality of this initiation strategy. We first targeted C(sp2)–C(sp3) bond-forming carbonyl α-arylation reactions due to the synthetic utility and prevalence of the α-aryl carbonyl motif in biologically active compounds37,38. Moreover, we posited that a general transition-metal-free radical strategy for carbonyl α-arylation would: (1) complement the scope of existing methods37,38 by providing improved compatibility with heteroaromatic systems, which are common poisons for metal catalysts; and (2) avoid issues related to the cost and toxicity of transition metals, as well as the geopolitical, ethical and environmental concerns regarding their supply chains39,40. Here, we hypothesized that CO2**·− could initiate the electron-catalysed unimolecular radical-nucleophilic substitution (SRN1)41,42 of aryl halides with enolate nucleophiles43,44. Indeed, CO2·− is known to readily reduce aryl halides to form aryl radicals22,23,45. Thus, aryl halide 1 was reacted with ethyl acetoacetate 2 (4.0 equiv.), Cs2CO3 (4.5 equiv.), ACVA (0.25 equiv.) and potassium formate (0.5 equiv.) in DMSO at 80 °C for 4 h (Fig. 2c). Pleasingly, following the addition of ammonium chloride (to promote complete deacetylation of the intermediate 1,3-dicarbonyl 3), this one-pot procedure produced deacetylated α-aryl ester 4 in 79% yield and hydrodehalogenated side product 5 in 10% yield (Fig. 2c, entry 1). Hydrodehalogenated product 5 is proposed to form from a slower competing HAT chain in which the aryl radical abstracts a hydrogen atom from the formate salt (reforming CO2·−). Other commercially available azo initiators were trialled, but none proved superior to ACVA (entries 2–5). No product formation was observed in the absence of ACVA and HCO2K (entry 6). However, some product formation was observed with HCO2K alone (entry 7); we attribute this background reactivity to inefficient spontaneous initiation events—in which the anionic nucleophile itself serves as a weak one-electron reductant45—being amplified by CO2·**− formation. Repeating the optimal reaction with ACVA and HCO2K in the presence of air reduced the yield of product 4 to 54% (entry 8). Moreover, all reactivity was completely supressed in the presence of 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) (entry 9). Finally, repeating the reaction in the absence of the nucleophile 2 afforded only the hydrodehalogenated product 5 in 70% yield (entry 10). These findings are all consistent with an SRN1 electron-transfer chain process. Other compatible, but less effective formate salts and reaction conditions are described in Supplementary Tables 1–3.

Electrophile scope and scale-up

With optimized conditions in hand, the scope and application of this thermally initiated transformation were explored (Fig. 3a). First, the importance of the nucleofuge was examined with para-halobenzonitrile derivatives. Pleasingly, the iodo, bromo and chloro derivatives were all converted into α-aryl ester 4 in 54%, 67% and 53% yield, respectively. Conversely, only 7% of 4 was formed from the corresponding fluoride, which indicates that the contribution of a polar SNAr mechanism to these reactions is probably minimal. Moreover, efficient reactivity was observed regardless of the arene substitution pattern because the meta- and ortho-substituted products 6 and 7 were both formed in good yields. Other electron-deficient nitrobenzene and aryl sulfone derivatives were similarly compatible and converted into products 8–13 in 30–85% yield. Interestingly, the reaction of 4-iodonitrobenzene formed product 11 in 66% yield in the absence of ACVA and HCO2K. We attribute this reactivity to spontaneous initiation, which is commonly observed when reacting anionic nucleophiles with easily reduced substrates (strongly electron-deficient aryl iodides)46,47. Carbonyl and ester derivatives were also tolerated and converted into α-aryl esters 14–20 in good to excellent yields. Notably, ester 20 could be formed in 77% yield from a dihalogenated substrate, illustrating the selectivity of this method for heavier halogens. Reactivity was still observed in the absence of strong electron-withdrawing resonance effects: bistrifluoromethyl derivative 21 was prepared in 66% yield. However, less electron-deficient systems displayed lower levels of reactivity as demonstrated by the formation of 22 in 15% yield. Moreover, no α-aryl ester product was formed from electrophiles bearing strong electron-donating groups such as 4-iodoanisole, but hydrodehalogenated product 23 was formed in 42% yield. This important result indicates that while the ACVA–formate initiator system is capable of reducing challenging electron-rich systems, the model SRN1 electron-transfer chain was not viable (potentially due to inefficient radical-anion coupling). Attention then turned towards our original target of heteroaryl halides. Pleasingly, pyridine, pyrazine and pyrimidine azaarenes, including the drug etoricoxib, with varying substitution patterns were all tolerated, furnishing α-aryl esters 24–35 in 26–63% yield. Finally, a variety of fused bicyclic heteroaryl halides, including quinolines, isoquinolines, quinoxalines, quinazolines, benzothiazoles, pyrazolopyrimidines and imidazopyridazines, were also compatible and substituted to form products 36–43 in 13–88% yield.

Fig. 3: Electrophile scoping and scale-up studies.

a, Scope of aryl and heteroaryl halides. b, Scale-up at AstraZeneca. aSee Supplementary Information, ‘Experimental Procedures and Characterisation Data’, for variations of the standard reaction conditions. bYields determined by 1H NMR spectroscopy against an internal standard (1,3,5-trimethoxybenzene). cNo ACVA or formate added.

Next, in collaboration with the process chemistry group at AstraZeneca, we sought to determine if the developed method and initiator system were sufficiently scalable to allow widespread use in the pharmaceutical and related industries. Therefore, the safety of the model reaction system was examined, in particular the use of ACVA, DMSO and base at elevated temperature. Differential scanning calorimetry and fall-hammer testing of ACVA, and high-rate Carius tube testing of the reaction mixture, all confirmed that the reaction can be safely performed at 80 °C. The developed reaction was further optimized via high-throughput experimentation48 and time-course analysis, before being performed on a 50-g scale, which provided product 4 in 61% isolated yield and >95% purity following a simple precipitation and filtration process (Fig. 3b). This result was particularly pleasing considering the heterogeneous nature of the reaction mixture (common to many reactions using inorganic bases in organic solvents) because a lack of reaction homogeneity can complicate scale-up49. In addition, the clean reaction profile and lack of side products was notable as azo initiators are known to also react as electrophiles with strong nucleophiles50. Thus, the low-cost ACVA–formate initiator system appears highly amenable to large-scale applications (current prices: ACVA, £1.57 g−1; HCO2K, £0.11 g−1 (Sigma Aldrich)).

Microscale parallel screening

Inspired by these promising findings, we sought to test the limits of the developed method and initiator system by exploring their compatibility with complex (hetero)aryl drug-like intermediates through microscale parallel screening at AstraZeneca (Fig. 4) (see Supplementary Information for more details). We had confidence in the ultrahigh-performance liquid chromatography (uHPLC)/mass spectrometry (MS)-based analysis because four substrates previously tested showed close agreement between isolated yields and uHPLC/MS area/area%. Pleasingly, 13 of the remaining 20 complex substrates, dense with Lewis basic-nitrogen functionality, afforded the desired products in synthetically useful yields under unoptimized conditions. To further validate these results, two of these ‘hits’ (49 and 59) were performed on a preparative scale and isolated in comparable yield. These results demonstrate the broad synthetic utility of the developed initiator system and α-arylation methodology, which both appear ready for immediate application in industrial drug discovery programmes.

Fig. 4: Microscale parallel screening of complex aryl halides.

The deacetylation step with ammonium chloride was not performed to facilitate a workflow without additional solid handling. Yields shown are combined uHPLC/MS area/area% of acetylated and deacetylated products (if applicable) at 220 nm (identified by MS analysis). Screening yields are visually summarized in a heatmap-style graphic (higher yields are brighter in colour). Isolated yields at 0.3-mmol scale for validation are reported in parentheses. aAccurate reporting of the reaction outcome was hampered by poor resolution of the reaction components across multiple analytical methods. Ts, p-toluenesulfonyl; Bn, benzyl; PMB, p-methoxybenzyl; Boc, t-butyloxycarbonyl.

Nucleophile scope

Having sufficiently explored the scope with respect to the electrophile, the nucleophile scope was investigated using 1 as a model substrate (Fig. 5a). First, closely related methyl and tert-butyl acetoacetates 67 and 68 were reacted to afford the corresponding α-arylated esters 69 and 70 in 77% and 42% yield, respectively. The ester functionality was not essential for reactivity because acetylacetone 71 was also converted into 72 (following a modified deacetylation protocol). Amide 73 was also tolerated and selectively converted into dicarbonyl 74 in 50% yield. More sterically congested nucleophiles such as 75 could be used to form α-methyl ester 76 in 55% yield. This result was particularly pleasing considering the prevalence of the α-methyl carboxylic acid moiety in non-steroidal anti-inflammatory drugs, as exemplified by the conversion of aryl bromide 77 into suprofen ethyl ester 78 in 42% yield. Cyclic ketone 79 was also successfully arylated to form 80 with a quaternary carbon centre in 27% yield. The ketone functionality was not essential for reactivity because ethyl cyanoacetate 81 was arylated to form 82 in 66% yield. In a similar fashion, malonates 83–85 were converted into 86–88 in 24–68% yield. Pleasingly, benzophenone glycine derivative 89 could also be used to form the corresponding N-protected unnatural α-aryl amino acid 90 in 43% yield. Finally, we examined non-anionic nucleophiles such as enamines using modified conditions based on the work of Gianetti and co-workers who arylated enamines under photoredox-catalysed conditions51. Here, the pyrrolidine-derived enamine of cyclohexanone 91 was formed in situ and arylated to afford 92 in 36% yield. This result supports the mechanistic hypothesis of Gianetti and co-workers who proposed that a chain mechanism was potentially operative alongside a photoredox catalytic cycle. In addition to applications in methodology development, the ACVA–formate initiator system may therefore be used as a general mechanistic tool to probe cases in which a radical chain mechanism is suspected.

Fig. 5: Nucleophile and radical trap scoping studies.

a, Scope of carbon-based nucleophiles and synthesis of suprofen ethyl ester. b, Scope of other nucleophiles and radical traps. aYields determined by 1H NMR spectroscopy against an internal standard (1,3,5-trimethoxybenzene). EWG, electron-withdrawing group; Ac, acetyl; tBu, tert-butyl; Tol, p-tolyl; Pin, pinacol.

To conclude our studies, we chose to demonstrate the utility of the developed ACVA–formate initiator system beyond C(sp2)–C(sp3) bond formation. First, simply by exchanging the nucleophile for a thiol, 4-iodobiphenyl 93 was converted into diaryl thioether 94 in 62% yield (Fig. 5b). This reaction presumably proceeds via an SRN1 mechanism as thiolates are known to be excellent SRN1 nucleophiles42,52. Next, by omitting any nucleophile and increasing the equivalents of ACVA and HCO2K, aryl bromide 95 was converted into hydrodehalogenated product 96 in 61% yield. The addition of bis(pinacolato)diboron (B2Pin2) enabled aryl boronic ester 97 to be formed in 51% yield. In a similar fashion, phosphonate ester 98 was formed in 50% yield when using triethyl phosphite as a radical trap. These results illustrate that the developed initiator system can be broadly applied for reaction development.

Conclusion

We have developed a general, scalable and inexpensive method to initiate radical chains through one-electron reduction by reacting an azo initiator with a formate salt. The power of this initiation approach was demonstrated through the development of a general carbonyl α-(hetero)arylation protocol, which enabled challenging and valuable structural motifs to be formed under remarkably simple reaction conditions. The scalability of this initiation approach and carbonyl α-(hetero)arylation protocol was demonstrated on a 50-g scale in an industrial setting. Moreover, compatibility with a range of complex substrates was illustrated through microscale parallel screening. Finally, in addition to C(sp2)–C(sp3) bond formation, the wider synthetic potential of the developed initiation system was illustrated through the development of C(sp2)–S, C(sp2)–H, C(sp2)–B and C(sp2)–P coupling reactions. Considering the multidisciplinary importance of radical chain chemistry, we anticipate that this inexpensive initiation strategy will find widespread application in both industry and academia. We also speculate that alongside large-scale manufacturing opportunities, the disclosed initiator system may serve as a valuable mechanistic tool to probe the chain character of other radical reactions.

Methods

General procedure for carbonyl α-arylation

An 8-ml screw-cap vial was charged with ACVA (25.6 mg, 75 μmol, 0.25 equiv.), HCO2K (12.6 mg, 150 μmol, 0.50 equiv.), Cs2CO3 (440 mg, 1.35 mmol, 4.5 equiv.), and if solid, the aryl halide coupling partner (1.0 equiv.). To the solids was sequentially added a magnetic stirrer bar, anhydrous DMSO (1.5 ml), ethyl acetoacetate (156 mg, 1.20 mmol, 4.0 equiv.), and if liquid, the aryl halide coupling partner through the screw-cap septa. The reaction mixture was sparged with N2 for 15 min before being sealed with parafilm. The reaction mixture was then stirred at 80 °C in a metal heating block for 4 h. To promote complete deacetylation of intermediate 1,3-dicarbonyls, NH4Cl (80.3 mg, 1.50 mmol, 5.0 equiv.) was added. The reaction was then stirred at 50 °C in a metal heating block for 1 h. The reaction mixture was cooled to room temperature before being diluted with CH2Cl2 (10 ml) and water (10 ml). The organic phase was collected, and the aqueous phase was extracted with CH2Cl2 (3 × 10 ml). The organics were combined, washed with brine (25 ml), dried (MgSO4) and concentrated under reduced pressure. The crude product was then purified by column chromatography to afford the α-arylated product.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary Information files.

References

Theriot, J. C. et al. Organocatalyzed atom transfer radical polymerization driven by visible light. Science 352, 1082–1086 (2016).

Article CAS PubMed Google Scholar 1.

Fazekas, T. J. et al. Diversification of aliphatic C–H bonds in small molecules and polyolefins through radical chain transfer. Science 375, 545–550 (2022).

Article CAS PubMed PubMed Central Google Scholar 1.

Constantin, T. et al. Halogen-atom and group transfer reactivity enabled by hydrogen tunneling. Science 377, 1323–1328 (2022).

Article CAS PubMed Google Scholar 1.

Johansson, K. O., Head-Gordon, M. P., Schrader, P. E., Wilson, K. R. & Michelsen, H. A. Resonance-stabilized hydrocarbon-radical chain reactions may explain soot inception and growth. Science 361, 997–1000 (2018).

Article CAS PubMed Google Scholar 1.

Freitas, F. P. et al. 7-Dehydrocholesterol is an endogenous suppressor of ferroptosis. Nature 626, 401–410 (2024).

Article CAS PubMed Google Scholar 1.

Alabugin, I. V., Eckhardt, P., Christopher, K. M. & Opatz, T. The photoredox paradox: electron and hole upconversion as the hidden secrets of photoredox catalysis. J. Am. Chem. Soc. 146, 27233–27254 (2024).

Article CAS PubMed Google Scholar 1.

Motherwell, W. B. & Crich, D. Free Radical Chain Reactions in Organic Synthesis (Elsevier, 1992). 1.

Studer, A. & Curran, D. P. Catalysis of radical reactions: a radical chemistry perspective. Angew. Chem. Int. Ed. 55, 58–102 (2016).

Article CAS Google Scholar 1.

Shaw, M. H., Twilton, J. & MacMillan, D. W. C. Photoredox catalysis in organic chemistry. J. Org. Chem. 81, 6898–6926 (2016).

Article CAS PubMed PubMed Central Google Scholar 1.

Yan, M., Kawamata, Y. & Baran, P. S. Synthetic organic electrochemical methods since 2000: on the verge of a renaissance. Chem. Rev. 117, 13230–13319 (2017).

Article CAS PubMed PubMed Central Google Scholar 1.

Magenau, A. J. D., Strandwitz, N. C., Gennaro, A. & Matyjaszewski, K. Electrochemically mediated atom transfer radical polymerization. Science 332, 81–84 (2011).

Article CAS PubMed Google Scholar 1.

Fu, M.-C., Shang, R., Zhao, B., Wang, B. & Fu, Y. Photocatalytic decarboxylative alkylations mediated by triphenylphosphine and sodium iodide. Science 363, 1429–1434 (2019).

Article CAS PubMed Google Scholar 1.

Constantin, T. et al. Aminoalkyl radicals as halogen-atom transfer agents for activation of alkyl and aryl halides. Science 367, 1021–1026 (2020).

Article CAS PubMed Google Scholar 1.

Lovato, K., Fier, P. S. & Maloney, K. M. The application of modern reactions in large-scale synthesis. Nat. Rev. Chem. 5, 546–563 (2021).

Article CAS PubMed Google Scholar 1.

Petrović, N., Malviya, B. K., Kappe, C. O. & Cantillo, D. Scaling-up electroorganic synthesis using a spinning electrode electrochemical reactor in batch and flow mode. Org. Process Res. Dev. 27, 2072–2081 (2023).

Article Google Scholar 1.

Studer, A. & Curran, D. P. The electron is a catalyst. Nat. Chem. 6, 765–773 (2014).

Article CAS PubMed Google Scholar 1.

Syroeshkin, M. A. et al. Upconversion of reductants. Angew. Chem. Int. Ed. 58, 5532–5550 (2019).

Article CAS Google Scholar 1.

Zhou, S. et al. Identifying the roles of amino acids, alcohols and 1,2-diamines as mediators in coupling of haloarenes to arenes. J. Am. Chem. Soc. 136, 17818–17826 (2014).

Article CAS PubMed Google Scholar 1.

Rohrbach, S., Shah, R. S., Tuttle, T. & Murphy, J. A. Neutral organic super electron donors made catalytic. Angew. Chem. Int. Ed. 58, 11454–11458 (2019).

Article CAS Google Scholar 1.

Odian, G. Principles of Polymerization (Wiley, 2004). 1.

Koppenol, W. H. & Rush, J. D. Reduction pot