## Introduction

Alkyl radical (R·), one of the most fundamental intermediates in organic synthesis, constitutes important approaches toward rapid molecular construction. Its generation from functionalised precursors such as aliphatic carboxylic acids, boronic reagents, halides, and others have been well-established and broadly applied as efficient paradigms in routine synthesis1,2,3,4,5,6,7,8,9,10,11,12. Of equal importance, direct C(sp3) radical generation from the readily available non-functionalised alkanes represents a more straightforward and sustainable method13. However, due to the high bond dissociation energies (BDEs)14,15, current strategies involving the homolysis of strong C(sp3)–H bonds mostly rely on the hydrogen atom transfer (HAT) with electrophilic heteroatom radicals, for instance, bromo-16,17,18, nitrogen-15,19, or oxygen-centred radicals19,20,21,22 (Fig. 1a). In this context, robust redox catalysts or strongly oxidising reagents are required, whereas precise control over the site selectivity among ubiquitous C(sp3)–H bonds in a molecule with broad substrate scope imposes grand challenges.

Chlorine radical (Cl·) is an efficient HAT agent that could cleave various C(sp3)–H bonds. Using chloride (Cl) in Cl· generation brings several benefits in organic syntheses because it is innocuous and abundant in diverse salt forms. Nevertheless, the unfavourable chloride-to-chlorine oxidation ($${{{E}}}^{^\circ }$$ = 1.36 V vs NHE)23 and untamed reactivity of Cl· compared with other halide analogues17,18,24,25,26 make chlorine radical-promoted alkylation rarely explored. In this endeavour, few strategies have been disclosed for the efficient usage of Cl·, including (a) the direct single-electron transfer (SET) from Cl to photocatalyst under photothermal conditions27,28,29,30; (b) the ligand-to-metal charge transfer (LMCT), which has been employed for the coupling of alkanes and organohalides by metallophotoredox catalysis31,32,33,34,35,36; (c) the photolysis of in situ generated Cl2 via electrooxidation of HCl37; (d) the bimolecular homolytic substitution (SH2) between chloroborate and an oxy radical for the alkane borylation38. These pioneering examples demonstrate the potential of Cl to realise the HAT process via Cl· intermediate (Fig. 1b).

The Minisci alkylation involves the coupling between heteroarenes and nucleophilic alkyl radicals. In views of the diversity of heteroarenes and their countless applications in material science, agrochemicals, and the pharmaceutical industry, the Minisci alkylation plays a pivotal role in synthetic chemistry. Since the key of the Minisci alkylation is the alkyl radical generation, it prompts us to consider the possibility of merging the chlorine radical-mediated alkyl radical generation manifold with the Minisci alkylation, which could allow the heteroarene diversification simply through the coupling with widely available C(sp3)-H feedstocks39. Furthermore, an H2-evolving Minisci alkylation is more desirable for its highest step- and atom-economy40,41. Owing to the strong aliphatic C–H bonds and the net oxidative nature, excessive alkane (normally as solvent) and oxidant loadings, high temperature, or precious catalysts are not uncommon. Therefore, a catalytic cross-dehydrogenative Minisci alkylation without stoichiometric chemical oxidants is long-sought-after39. Our group has a long-term interest in the arene functionalisation, which was facilitated by the excited aromatics under catalyst-free conditions42,43,44,45,46,47. Recently, we documented a simple and clean Minisci alkylation reaction via formal dehydrative coupling of heteroarenes with alcohols. Taking advantage of the superior redox properties of the excited heteroarenes, metallophotoredox catalysts could be strategically avoided48.

Based on these literature precedents49,50, we envisioned that heteroarene itself could be an intrinsic photosensitiser, allowing efficient Cl oxidation in the Minisci reaction. In this work, a dehydrogenative Minisci alkylation using catalytic Cl under photochemical conditions is presented (Fig. 1c). Owing to the strong hydrogen atom affinity of Cl·, a wide range of inactivated and activated C(sp3)–H bonds could be arylated with good functional group tolerance and substrate diversity, and notably, with the strategic introduction of the cobaloxime catalyst, we formulate a chemical oxidant-free heteroarene alkylation protocol by releasing H2.

## Results

### Reaction design and optimisation

The proposed dehydrogenative alkylation pathway is depicted in Fig. 1d. It was hypothesised that the excited heteroarenes I could oxidise the Cl under irradiation to generate the Cl· for the aliphatic C–H abstraction from alkane 2. The addition of the so-formed alkyl radical to another equivalent of heteroarene 1-H+ followed by hydrogen atom removal would give the desired alkylated heteroarene 3-H+. Meanwhile, to efficiently quench the generated radical intermediate II, a readily accessible cobaloxime catalyst [Co(dmgH)2(py)]Cl was introduced to the system51,52,53,54,55 not only to prevent the over-reduction of intermediate II but also to serve as a terminal oxidant for the rearomatisation of alkylated intermediate III though H2 evolution.

In our initial evaluation, 2-phenylquinoline (1a) and cyclohexane (2a) were chosen as model substrates with trifluoroacetic acid (TFA) and a catalytic amount of Bu4NCl and cobaloxime [Co(dmgH)2(py)]Cl in CHCl3. Delightfully, the desired coupling product 3 could be obtained in good conversion and product yield. After considerable efforts, the optimal reaction conditions yielded 80% of 3 when 20 mol% Bu4NCl, 5 mol% [Co(dmgH)2(py)]Cl, and 3 equiv. of TFA were used in CHCl3 under photo-irradiation for 20 h (Table 1, entry 1). During the optimisations, three key reaction components including Bu4NCl, [Co(dmgH)2(py)]Cl, and CHCl348 were identified, all of which could potentially serve as the Cl· sources. The reaction could proceed smoothly in the presence of any of these chlorides; otherwise, no reaction occurred (entries 2-8). Importantly, the cobaloxime catalyst could improve the reaction productivity by minimising the over-reduction of quinoline to tetrahydroquinoline or other off-target decomposition reactivities. A significant solvent effect was observed in this transformation. In the chlorinated solvent CHCl3, heteroarene 1a was alkylated efficiently without side reaction detected; however, when other solvents were used, poor selectivity and side product formations were often observed (e.g. in entry 4, 10% tetrahydroquinoline was obtained). Additionally, we found that the reaction could be partially suppressed by oxygen (entry 9), and no reaction occurred without light (entry 10).

### Substrate scope

After obtaining the optimal reaction conditions, we approached the substrate scope to different C(sp3)–H species using 2-phenylquinoline (1a) as the coupling partner (Fig. 2). Simple cyclic alkanes containing five to twelve carbons afforded the corresponding alkylated heteroarenes 3 to 7 in moderate to good yields, and so did bridged alkanes norbornane and adamantane (8 and 9). Carbonyl compound cyclopentanone was functionalised at the β-position and afforded the desired product 11. Benzylic C–H abstractions of methylbenzene derivatives also provided benzylated heteroarenes (1114)56. It was not surprising that ethylated quinoline 15 was produced as the major product with diethyl ether (Et2O) as the alkane source48; and the similar C–O cleavage was observed with 1,2-dimethoxyethane (DME) to give the deethoxylated products 16; gratifyingly, the non-cleaved ethereal compounds 17 and 18 could be obtained by decreasing the acid amount to 1.2 to 2 equiv37. However, the C–O cleavage was still inevitable with tetrahydrofuran (THF) and methanol (1922), possibly due to the relatively high ring strain and oxidation potential of the alcoholic oxygen atom. Reactions with tetrahydropyran (THP) and its analogues proceeded smoothly and gave good yields of the products 2326. α-C(sp3)–H functionalisation of amine derivatives, for example, amides, sulfonamide and phosphoramide, were all successful (2733). Interestingly, the HAT of the N-methyl group of N,N-dimethylformamide (DMF) is more favourable than the formyl one and the alkylated heteroarene 29 was obtained as the major product. Few alkane substrates, for instance, cyclododecane, adamantane and cyclopentane were inert at ambient temperature, and a slight temperature increase to 55–60 °C was helpful for their transformations (7, 9, and 10). The unique regioselectivity of Cl· could be displayed from some substrates. For example, the HAT on adamantane is slightly favourable on the tertiary C–H bond in comparison to the secondary one (9), yet 2-methyltetrahydrofuran (MeTHF) only afforded the secondary carbon-functionalised product (20); C4 functionalisation of the N-methylpyrrolidinone (NMP) primarily occurred at the secondary C4 position rather the primary C5 one (31). Other than alkyl substrates, the amidation of heteroarene is also viable with formamide, which furnished the amido product 34. Noticeably, the potential application of this reaction was briefly demonstrated by a challenging gram-scale reaction of heteroarene 1a with 3 equiv. of unactivated alkane 3e, and delightfully the desired product 7 could be obtained in 54% isolated yield.

Further substrate and functional group tolerance were examined by coupling various heterocycles 1 with cyclohexane (2a). Satisfyingly, a broad variety of functional groups such as halides, cyano, acetyl, ester, amino, nitro, and sulfonamide are compatible with our method, as shown in Fig. 3. Quinoline, pyridine, and isoquinoline moieties afforded the corresponding alkylated products with up to 75% yield (3556); other heterocycles like pyrimidine (57 and 58), pyrazine (59), quinoxalinone (60), and purine (6163) were also proven to be feasible (substrates that are incompatible or with low reactivity in the reaction are listed in Supplementary Information Fig. 3). Complex heterocycles bearing different aliphatic C(sp3)–H bonds were examined, including alanine and menthol-derived pyridines (64 and 65), and nicotine (66) were successfully turned into the desired products with moderate yields. Alkylation of hydrocinchonine, a common chiral ligand candidate in synthetic chemistry, gave an appreciable yield of the product (67) with its hydroxy group preserved. Fasudil, a potent Rho-kinase inhibitor and vasodilator, also provided the desired product (68) after protecting its amino group.