Main

The importance of organoboron compounds has triggered intensive research into methods for introducing boron into organic molecules, the most direct of which is C–H borylation (Fig. 1a)2,3. This approach benefits from introducing a synthetically valuable boron moiety into inherently unreactive feedstock chemicals by substitution of ubiquitous C–H bonds. The C(sp2)–H borylation of aromatic compounds is well established (for example, using iridium catalysts)6,7, but C(sp3)–H borylation of alkanes is less developed4. The majority of reported C(sp3)–H borylations use activated substrates8,9,10,11,12,13 or those possessing heteroatom directing groups14,15,16,17,18, whereas borylations of non-activated C(sp3)–H bonds require the use of precious-metal catalysts and harsh reaction conditions19,20,21,22,23. Furthermore, the higher reactivity of aromatic C–H bonds prohibits selective borylation of non-activated C(sp3)–H bonds in the presence of sterically accessible C(sp2)–H bonds23.

Fig. 1: Catalytic C–H borylation reactions.
figure 1

a, Transition-metal-catalysed C–H borylations. These reactions proceed via cleavage of the C–H bond by the metal catalyst to form alkyl metal intermediates. b, Proposed radical-mediated C(sp3)–H borylation using HAT catalysis. c, Photoinduced decarboxylative borylation of N-hydroxyphthalimide esters 1 (ref. 27) and proposed C(sp3)–H borylation using N-alkoxyphthalimides 7. d, Reaction development. Yields are based on molar equivalents of 2. LED, light-emitting diode; cat, catecholato; pin, pinacolato; GC, gas chromatography.

Full size image

We envisioned an alternative C(sp3)–H borylation using a hydrogen atom transfer (HAT) strategy5. Here, the C–H bond is cleaved by intermolecular reaction with a heteroatom-centred radical (X•), rather than with a metal catalyst, and subsequent homolytic substitution of a diboron reagent by the resulting alkyl radical intermediate forms the C–B bond (Fig. 1b). An important feature of this proposed mechanism is that regioselectivity should be determined by the bond dissociation energy (BDE) of the C–H bond, and thus there will be high selectivity for borylation of alkyl groups (BDE = 98 kcal mol−1 for cyclohexyl–H)24 over aromatic rings (BDE = 113 kcal mol−1 for phenyl–H)25.

Herein we report a photoinduced borylation of non-activated C(sp3)–H bonds using chloride as a HAT catalyst. The use of a radical-mediated strategy enables a wide range of functionalized alkanes to be borylated under mild conditions (ambient temperature, violet-light irradiation) and circumvents the high C(sp2)–H selectivity inherent in transition-metal-catalysed borylations. Furthermore, direct photoexcitation of an N-alkoxyphthalimide oxidant allows these reactions to occur in the absence of a metal catalyst26, therefore providing room-temperature and metal-free C(sp3)–H bond borylations.

Our initial design plan for C–H borylation was inspired by our decarboxylative borylation of N-hydroxyphthalimide esters 1 with bis(catecholato)diboron (2, B2(cat)2; Fig. 1c)27. In that process, photoinduced electron transfer (PET) under blue-light irradiation provided phthalimide radical anion intermediates 3. In the case of aliphatic esters (R1 = alkyl), decarboxylative fragmentation furnished phthalimide anion 4 and an alkyl radical, 5, which was borylated by reaction with B2(cat)2 to form boronic ester 6. We reasoned that if N-alkoxyphthalimides 7 were used in place of esters 1, fragmentation of the radical anion intermediate 8 by N–O bond cleavage would produce a highly reactive oxygen-centred radical (9) that is capable of cleaving strong C(sp3)–H bonds via HAT28,29,30. Intermolecular HAT between 9 and an alkane 10 (R2 = alkyl) would generate alcohol 11 and alkyl radical 12, which would be rapidly borylated by reaction with B2(cat)2 to give C–H borylation product 1331.

We postulated that the nature of the oxygen-centred radical 9 would be crucial for successfully implementing our reaction design. For productive C–H borylation, 9 must (1) be stable to β-fragmentation to alkyl radicals, (2) be electrophilic, therefore favouring reaction with hydridic aliphatic hydrogen atoms over direct reaction at the electrophilic boron atoms of 2, and (3) form strong O–H bonds (BDE > 100 kcal mol−1) upon HAT, thus allowing borylation of strong C(sp3)–H bonds. Considering these requirements, we selected N-(2,2,2-trifluoroethoxy)phthalimide (14) (Fig. 1d) as this provides the electrophilic 2,2,2-trifluoroethoxy radical, which has previously been reported to cleave strong C(sp3)–H bonds32.

Our initial investigations focused on the borylation of norbornane (15), which was irradiated in the presence of B2(cat)2 and alkoxyphthalimide 14 in acetonitrile (Fig. 1d), but unfortunately no C–H borylation product was observed (Fig. 1d, entry 1). On the basis of a recent report on alkoxy-radical-mediated C(sp3)–H functionalization using cerium catalysis32, we investigated the effect of cerium salts on the borylation reaction (Supplementary Table 1)33. Pleasingly, in the presence of cerium(iv) chloride tetraethylammonium chloride complex (CeCl6(NEt4)2), irradiation with violet light gave boronic ester 17 in 36% yield (Fig. 1d, entry 2). Unexpectedly, a control experiment with just tetraethylammonium chloride (NEt4Cl) as catalyst resulted in the formation of 17 in a similar yield, suggesting that chloride is catalysing the reaction (Fig. 1d, entry 3). Further improvements were made through analysis of a range of other chloride sources (Supplementary Table 2), with 20 mol% B-chlorocatecholborane (ClB(cat)) proving optimal, enabling formation of 17 in 61% yield (Fig. 1d, entry 4). Owing to the instability of the initially formed catechol boronic ester 16 to hydrolysis, in all cases in situ transesterification to the stable pinacol boronic ester 17 was performed before isolation.

With these metal-free C(sp3)–H borylation conditions, we proceeded to explore the scope of the reaction (Fig. 2). Unfunctionalized cyclic alkanes, which are commonly used as solvents in metal-catalysed C–H borylations owing to their low reactivity23, were successfully borylated to provide alkylboronic esters 17–22. In addition, borylation of cycloalkanes functionalized with halides or nitrile groups was also possible (23–25). High regio- and diastereoselectivities were obtained in the reactions of norbornane (17), trans-decalin (23) and exo-2-chloronorbornane (24), with no borylation of tertiary C–H bonds observed. Cyclopentane carbonitrile (26) reacted selectively at the methylene distal to the electron-withdrawing nitrile group.

Fig. 2: Photoinduced C–H borylations of alkanes.
figure 2

Reactions were performed with 0.3 mmol of B2(cat)2. Yellow and purple spheres represent protons or alkyl groups. Yields are of isolated products. Regioisomeric ratios (r.r.) and diastereomeric ratios (d.r.) were determined by GC analysis. Numbers in parentheses show yields obtained using different alkane stoichiometry: a20 equiv., b5 equiv., c3 equiv. dr.r. and d.r. were determined by 1H NMR analysis after oxidation to the corresponding alcohol. eThe yield was determined by GC analysis. fr.r. and d.r. were determined by 1H NMR analysis. gUsing 1.0 equiv. alkane and 1.2 equiv. B2(cat)2. N(Phth), N-phthalimide; tBu, tert-butyl.

Full size image

We next investigated the effectiveness of our protocol for the borylation of acyclic alkanes. Although various unfunctionalized substrates were successfully borylated (26–31), we were surprised to observe selective borylation of methyl groups over methylene groups. This selectivity is the opposite of that expected in homolytic cleavage of C–H bonds via HAT, where selectivity is typically determined by the strength of the C–H bond, thus favouring reaction at methylene (BDE = 99 kcal mol−1 for iPr–H) over methyl groups (BDE = 101 kcal mol−1 for Et–H)25. Pentane was borylated with a methyl/methylene selectivity of approximately 1:1 (26), whereas for substrates containing sterically hindered methylene groups, complete methyl selectivity was observed (28–31). Notably, 2,2,4,4-tetramethylpentane (31) was borylated efficiently; to our knowledge this is the first borylation of a tert-butyl group in the absence of directing groups11,22,26. Selective reaction at sterically hindered methyl groups was also observed for various functionalized acyclic alkanes, including those containing halides (32–36), protected amines (37, 38) and nitriles (39). The regioselectivity appeared to be sensitive to electronic effects, with no borylation observed at sterically unhindered methylene groups proximal to electron-withdrawing groups (33, 36–39). For substrates possessing even weaker methine C–H bonds, in no cases were tertiary boronic ester products observed.

We also tested substrates possessing both C(sp3)–H bonds and sterically unhindered aromatic C(sp2)–H bonds, which give high selectively for aromatic C–H borylation with iridium catalysis23. Isopropylbenzene (40) was borylated with complete selectivity for the methyl groups over both the aromatic and the much weaker benzylic C–H bonds. Similar selectivity was observed for several other isopropylbenzene derivatives (41–42); notably 4-isopropyltoluene (41), which possesses an unhindered benzylic methyl group8. The reaction was extended to sterically hindered methyl groups of tert-butylbenzenes (43–44), which are unreactive under metal-catalysed conditions. Borylation of C(sp3)–H bonds distal to aromatic rings was also possible, including neopentyl (45) and cyclohexyl benzene (46), and various heteroaromatic carboxylate esters (47, 48). In addition, terminal alkenes were tolerated, with 4,4-dimethyl-1-pentene reacting at the tert-butyl group to provide boronic ester 49. More complex substrates were also successfully borylated, including ibuprofen methyl ester (50), a galactose derivative (51), and phthalimide-protected derivatives of the amino acids leucine (52) and tert-leucine (53). Although reactions were routinely run using 10 equivalents of alkane, the excess alkane could be recovered in >80% yield, and in several cases we showed that 5 or 3 equivalents could also be used without a substantial reduction in yield (17, 29, 30, 32, 36 and 53). Alternatively, improved yields could be obtained with 20 equivalents of alkane (17).

Usually, C(sp3)–H borylations require substrates to be used in excess, and it was only very recently this limitation was overcome using a modified iridium/phenanthroline ligand system23. This is especially valuable for complex substrates. Therefore, we tested our methodology using the alkane as the limiting reagent and found that, under slightly modified conditions, the complex amino acid-derived product 53 could be isolated in comparable yield with 85% recovery of unreacted alkane (for further alkane stoichiometry studies, see Supplementary Information, section 2.7).

Organosilanes are an important compound class that are used extensively in organic synthesis and materials science. It has been previously demonstrated that iridium-catalysed C(sp3)–H borylation of methyl silanes provides a convenient and direct method to access synthetically useful (borylmethyl)silanes11. However, that method uses 5–10 mol% iridium, and methylsilanes cannot be selectively borylated in the presence of sterically unhindered aromatic C–H bonds. By contrast, our metal-free, HAT-mediated borylation provided complete selectivity for reaction of α-silyl C(sp3)–H bonds (54–56, 62–65; Fig. 3). In addition, good functional-group tolerance was demonstrated by the successful borylation of substrates containing halides (56–58, 61), carboxylate esters (59–66), enoates (64), ketones (65), and pinacol boronic esters (67). Permethyloligosilanes were also borylated in good yield, including hexamethyldisilane (68) and bis(trimethylsilyl)methane (69). Interestingly, tetraethylsilane (70) reacted selectively at the methylene (α) position, again contrasting with the iridium-catalysed selectivity, which gives only the product of methyl (β) borylation11. As before, we explored the effect of stoichiometry of organosilane and found that although the yields declined with reduced loading (54, 65 and 68), a 35% yield of boronic ester 54 could still be achieved using the organosilane as the limiting reagent.

Fig. 3: Photoinduced C–H borylations of silanes.
figure 3

Reactions were performed with 0.3 mmol of B2(cat)2. Yields are of isolated products. r.r. was determined by 1H NMR analysis. Numbers in parentheses show yields obtained using different silane stoichiometry: a3 equiv., b5 equiv., c20 equiv. dUsing 1.0 equiv. silane and 1.2 equiv. B2(cat)2. er.r. was determined by GC analysis.

Full size image

To gain insight into the mechanism of this metal-free C–H borylation, we performed a series of experiments to confirm the identity of the HAT species and determine how it is generated. Testing other N-alkoxyphthalimides showed that replacing the trifluoroethoxy group of 14 with a phenoxy group caused only a small reduction in reaction efficiency (Supplementary Table 4). This indicated that alkoxy radicals are unlikely to be responsible for HAT because the relatively weak O–H bond of phenol (BDE = 90 kcal mol−1), compared to non-activated C(sp3)–H bonds (BDE = 97–101 kcal mol–1) makes HAT thermodynamically disfavoured25. Instead, the vital role of chloride led us to suspect the involvement of chlorine radicals. This was confirmed upon subjecting 1,6-heptadiene (71) to modified reaction conditions using 1 equivalent of ClB(cat), which gave chlorinated (cyclopentylmethyl)boronic ester 72 derived from a chlorine radical addition–5-exo-trig cyclisation–borylation sequence (Fig. 4a). Through fluorescence quenching experiments (Supplementary Fig. 9), we determined that chlorine radicals could be formed via PET between 14 and chloride anions, where photoexcitation of 14 produces a strongly oxidising excited state (Ep/2(14*/14•−) = 1.46 V versus saturated calomel electrode (SCE) in MeCN, where Ep/2 is the half-peak potential) capable of undergoing exergonic single-electron transfer (SET) with chloride anions (Ep/2(Cl•/Cl−) = 1.00 V versus SCE in MeCN for NEt4Cl). The formation of chloride anions from ClB(cat) is supported by NMR experiments showing hydrolysis of ClB(cat) by trace water in acetonitrile (Supplementary Figs. 18–20). Interestingly, fluorescence quenching experiments showed that the excited state of 14 was quenched with similar efficiency by ClB(cat), NEt4Cl and B2(cat)2 (Supplementary Figs. 9–11). The initial concentration B2(cat)2 is high with respect to ClB(cat) or chloride anions, and so the predominant PET process is probably between 14 and B2(cat)2.

Fig. 4: Mechanistic studies.
figure 4

a, Evidence for the formation of chlorine radicals. b, Proposed mechanism. c, Effect of B2(cat)2 and borate 78 concentration on regioselectivity in the borylation of pentane. d, Trapping of tertiary and hindered secondary alkyl radicals.

Full size image

On the basis of these observations, we propose the mechanism outlined in Fig. 4b. After photoexcitation of 14 to 73, reductive quenching by B2(cat)2 and subsequent β-scission of radical anion 74 gives the trifluoroethoxy radical 75. Oxygen-centred radicals are known to react with catechol boronic esters to form radical ‘ate’ complexes, where the unpaired electron is delocalized onto the catecholate ligand34. Therefore, we propose that chlorine radicals are formed through the reaction of 75 with ClB(cat) via radical ‘ate’ complex 76. HAT from 15 (BDE = 99 kcal mol−1)35 to either a chlorine radical or complex 76 generates hydrochloric acid (BDE = 103 kcal mol−1)25, alkyl radical 77 and trifluoroethyl borate 78. Borylation of 77 with B2(cat)2 proceeds via radical complex 79 (as previously described)31, with cleavage of the B–B bond facilitated by reaction with chloride. As Lewis base adducts of boryl radicals have been reported to be strong single-electron reductants36, regeneration of 76 can occur through SET between chloride-stabilized boryl radical 80 and 14 (Supplementary Figs. 32–34).

Given that regioselectivities in HAT processes from simple alkanes to chlorine radicals follow the reactivity trend of methine > methylene > methyl37, we proceeded to investigate the origin of the unexpected regioselectivity in our borylation reaction. Under our standard conditions, pentane was borylated in a 54:46 primary:secondary ratio. However, the regioselectivity was found to be dependent on the stoichiometry of B2(cat)2, with higher equivalents leading to increased secondary selectivity (Fig. 4c). Conversely, adding trifluoroethyl borate 78 to the reaction resulted in an increase in primary selectivity. These reactions provide evidence that the regioselectivity is determined during both the HAT and C–B bond-forming steps. Higher primary selectivity with lower B2(cat)2 concentration is indicative of competing reaction pathways for secondary alkyl radical intermediates, whereby a slower rate of borylation for these more hindered radicals results in deleterious pathways (for example, single-electron oxidation). Higher primary selectivity with higher borate 78 concentration provides indirect evidence for the formation of radical ‘ate’ complex 76, which undergoes primary-selective HAT reactions. To probe the high methyl selectivity for substrates possessing both methylene and methine groups, we performed the borylation of 2,5-dimethylhexane (82) in the presence of an alkene radical trap (N-methyl-N-phenyl-methacrylamide (83); Fig. 4d). This provided boronic ester 29 in 21% yield with >97:3 methyl (α) selectivity, as well as 3% of oxindole 84, which was formed with high selectivity for methine (β) C–H functionalization. This confirms the non-productive formation of sterically hindered secondary and tertiary alkyl radicals. However, the low yield of 84 and the lack of other tertiary radical-derived side products suggest that the regioselectivity is largely determined during the HAT step. We propose that this is a result of HAT occurring directly to complex 76, rather than a ‘free’ chlorine radical, where complexation of the chlorine radical with borate 78 makes a more sterically demanding HAT species that is able to selectively functionalize less hindered but stronger C–H bonds38, thus providing unusually high primary C–H selectivity.

We have introduced an approach to borylations of non-activated C(sp3)–H bonds for the synthesis of alkylboronic esters from simple alkanes. Using a photoinduced HAT strategy, a broad range of alkanes were borylated under mild conditions and with regioselectivities distinct from those of established metal-catalysed protocols, namely high selectivity for C(sp3)–H over C(sp2)–H bonds. The reaction also provides a rare example of a radical-mediated C–H functionalization proceeding with high selectivity for substitution of methyl C–H bonds over weaker secondary, tertiary and benzylic C–H bonds. Although further developments are needed to improve the modest yields and high substrate loading, this radical-mediated C–H borylation enables the transformation of feedstock chemicals into valuable organoboron products with selectivities that complement those of existing methods.