Using commercial dibenzothiophene S-oxide 2 and adapting straightforward conditions used widely for the synthesis of more common aryl sulfonium salts, we have developed a robust protocol for the preparation of alkoxy sulfonium salts from a range of alcohols (Table 1). The alkoxy sulfonium salts were readily isolated and characterized, and the process was amenable to larger scale; 3.15 g of salt 1a was prepared in 86% yield. Furthermore, 1a was prepared on a 1-kg scale in an industrial setting (see 'Large-scale, continuous 1,2-alkoxy-hydroxylation under photoflow conditions' section). A range of alkoxy sulfonium salts -- derived from primary and secondary alcohols -- were prepared in high yield, although only a few tertiary alcohols could be engaged (for example, the formation of 1ai), thus this is an area for future refinement of the approach. Crucially, products of alcohol oxidation were not observed. The process embraced both inexpensive, simple feedstock alcohols (1a-h) -- including a deuterated alcohol (1b) -- and complex alcohols derived from an amino acid (1s) and a sugar (1t), or alcohols from the chiral pool (1aa and 1ab). The mild reaction conditions allowed the presence of a number of important functional groups in the alcohol substrates to be tolerated, for example, halide (1d,e,i,j,o,y,z and ad), nitrile (1k), ester (1s,x,aa,ab), sulfone (1l), alkene (1m), alkyne (1n,o,u), tertiary alcohol (1q), acrylate (1r) and phthalimido (1p,s) motifs were compatible with alkoxy sulfonium salt synthesis, while a cyclic ketal delivered a sulfonium salt bearing the corresponding ketone functionality (1af). Crucially, enantiomerically pure alcohols, with the stereochemistry residing at (1aa,ab), or remote from the hydroxyl group (1s,t), gave alkoxy sulfonium salts with no loss of stereochemical integrity (Fig. 2).
With alkoxy sulfonium salts in hand, we optimized the photocatalytic selective construction of two C-O bonds across an alkene (Table 2). Using alkoxy sulfonium salt 1a and 4-chlorostyrene 3a, we arrived at the optimized conditions highlighted in entry 1; 4a was obtained in 83% yield when using the photocatalyst PC1 (0.1 mol%) in acetone, in the presence of water (2.0 equiv.) and NaHPO (50 mol%), under light irradiation at 456 nm. The formation of 2 -- probably by hydrolysis of 1a -- and 2' -- by direct homolysis of 1a -- were minimized by reducing the reaction temperature and using a photocatalyst. At higher reaction temperatures, in the presence of more alkoxy sulfonium salt, the formation of the aryl ketone product arising from oxidation of 4a was observed. We ascribe this to HAT from the benzylic position of 4a by the alkoxy radical. Under our optimized conditions, only traces of aryl ketone byproduct are observed (Supplementary Table S1). The reaction is only marginally sensitive to the reaction temperature (entry 2) and the presence of a base (entry 3) -- aspects that we later exploited in the development of a photoflow process. In contrast, the reaction shuts down in solvents other than acetone (entry 4). Using alternative alkoxy sulfonium salts, conveniently derived from other sulfoxides (entry 5), or photocatalysts (entry 7), gave lower yields of product. Addition of water was not essential -- probably owing to adventitious water in the acetone solvent -- and higher loadings of water did not improve the yield (entry 6). A low yield of 4a was obtained in the absence of a photocatalyst (entry 9), and no product was observed without light (entry 10).
We next assessed the scope of the photocatalysed alkoxy-hydroxylation of alkenes using methoxy sulfonium salt 1a and varying the alkene partner (Table 3). Styrene derivatives with electron-withdrawing and electron-donating groups at the para- (4a-m,z and 4aa-ac), meta- (4n-t) and ortho-positions (4u-y) were effective substrates. Products with polyfunctionalized aromatic rings were obtained in good to high yields regardless of the position and electronic properties of the substituents (4ad-ah). The benzocyclobutane motif (4ai) and both 1- and 2-substituted naphthyl-units were also tolerated (4aj and 4ak). Engaging heteroaryl-containing substrates in photocatalysis often proves challenging owing to the sensitivity of heteroaromatic motifs to light. Pleasingly, the use of heteroaryl-containing alkenes delivers the corresponding medicinally relevant products (4al-aq), albeit in lower yields. The protocol extends to more challenging polysubstituted alkenes (4ar-be), for example, compound 4ar was obtained in 83% yield despite the presence of weak allylic C-H bonds, which could be prone to competing HAT by the alkoxy radical intermediate. Other 1,2-disubstituted alkenes could be effectively engaged, including those bearing primary alkyl bromide (4as), ester (4at), ketone (4au), tertiary alkyl hydroxyl (4ay) and nitrile (4az) motifs. Cyclic 1,2-disubstituted alkenes afforded the corresponding products 4aw and 4ax in good yields. The process also proved effective for the 1,2-hydroxy-alkoxy functionalization of trisubstituted alkenes (4av). The formation of 4ay as a single diasteroisomer may be rationalized by invoking the interaction of the pendant hydroxyl group with one face of the intermediate benzylic carbocation (Fig. 3b), with subsequent nucleophilic attack by water on the opposite face. Interestingly, 4bc was obtained in a moderate yield with no sign of radical ring-opening of the cyclopropyl substituent. This suggests a rapid oxidation of the intermediate benzyl radical to the corresponding carbocation in the catalytic cycle. Finally, the benzazepine motif present in arginine vasopressine receptor antagonists can be conveniently decorated; 4be was prepared in good yield and its structure confirmed by X-ray crystallographic analysis. Overall, this method tolerates various substitution patterns and functional groups, highlighting its wide applicability. The process is, however, currently specific for styrene-type alkenes, and no product formation was observed when alkyl substituted alkenes were employed; this may be because of parasitic HAT from allylic C-H bonds outcompeting the addition of the alkoxy radical to the C=C bond.
We next explored the scope of the photocatalytic process with respect to the alkoxy sulfonium salt using alkene 3a (Fig. 2a). The reaction tolerated wide variation in the activated alcohol partner. Derivatives of simple alcohols, such as deuterated methanol (4bf) and ethanol (4bg), or more functionalized β-substituted derivatives (4bh-bx) were well-tolerated in the alkene difunctionalization. The reaction is not limited to primary alcohols; secondary (4by-cl) and tertiary alcohol derivatives (4cm) were also viable substrates. Importantly, products were obtained in good-to-high yield throughout, despite the known propensity of β-branched alkoxy radicals to undergo fragmentation. A range of functional groups in the alcohol-derived partner were tolerated: for example, primary halides (4bn,bo,bu,cc,cd), nitriles (4bm), sulfones (4bp), phthalimides (4bq and protected amino acid 4bx), hydroxyls (4br), ketones (4cj), α,β-unsaturated esters (4bs), alkenes (4bt), alkynes (4bu,bv,cb), esters (4bx,ca,ce), lactones (4cf) and ketals (in protected sugar 4bw).
By varying the nucleophile used to quench the carbocation intermediate, we expanded our approach to other 1,2-difunctionalization reactions. Pleasingly, minimal tuning of the reaction conditions delivered an alkene dialkoxylation procedure. The synthesis of unsymmetrical 1,2-diethers -- with different alkyl groups on the two oxygen atoms -- from diols, is non-trivial owing to the similar reactivity of the two hydroxy groups and the need for efficient and toxic alkylating agents (Fig. 2b, left). Using methoxy sulfonium salt 1a and alkene 3a, with a range of alcohol nucleophile partners, allowed selective access to several 1,2-diethers. Careful drying of the alcohol nucleophile ensured the formation of only traces of the 1,2-alkoxy-hydroxylation products. The alcohols used ranged from feedstock alcohols -- such as methanol (5a), benzyl alcohol (5e), cyclobutanol (5g), tert-butanol (5h) and 2-methoxyethanol (5d), through alcohols containing functionality -- such as bromoethanol (5b) and 2-(trimethylsilyl)ethanol (5c) -- to complex alcohols such as a protected serine (5f). Highlighting the generality of this platform, the use of inexpensive nitriles as co-solvents resulted in the 1,2-alkoxy-amidation of alkene 3a using alkoxy sulfonium salt 1a; amides 6a-h were isolated in moderate yields alongside traces of the 1,2-alkoxy-hydroxylation products. Finally, an analogue of the paclitaxel side chain 6i (Fig. 1a) was prepared from ethyl cinnamate when using benzonitrile as co-solvent (Fig. 2b, right).
Given the potential impact of developing a process that can be applied on industrial scale, the reaction was adapted from a batch process to an easily scaled photoflow process -- a flow chemistry setting under light irradiation. We selected the reaction between methoxy sulfonium salt 1a and styrene 3a as a model system. Firstly, the synthesis of methoxy sulfonium salt 1a was optimized for a 1-kg scale (Supplementary Information). Secondly, a rapid optimization campaign was performed taking into consideration key aspects of the method to make it compatible with the process chemistry flow reactor used and fulfilling requirements for its use in the pharmaceutical industry (Fig. 3a). Switching from an inorganic to an organic base removed any solubility concerns and ensured reaction homogeneity, while a 5-min residence time was found sufficient to achieve a good conversion and an excellent yield. The reaction temperature was raised from -30 °C to 0 °C, thus avoiding cryogenic conditions and unnecessary light diffraction from frost build-up on the reactor surface. As a stability study on the starting mixture indicated a slow decay of 1a in the reaction mixture over 24 h, a two-feed system was established that consisted of two equal volume solutions of 1a in acetone and styrene 3a with photocatalyst, base and water, respectively. Both streams were mixed at the same flow rate and cooled to 0 °C, allowing for fresh mixing and pre-cooling of the active reagents before irradiation. Process safety experiments (differential scanning calorimetry, and accelerating rate calorimetry) for starting mixtures did not reveal safety concerns (Supplementary Information). With these minor modifications to the standard conditions, we achieved good yields of 4a in the gram-scale photoflow reactor. The reaction was then scaled up to the kilogram-scale photoflow reactor, starting with 1.0 kg of 1a, and obtaining 4a in a 64% yield over 3.5 h.
Building on previous studies on the photochemistry of sulfonium salts by our team and by others, we propose the catalytic cycle set out in Fig. 3b. Photoexcitation of PC1 ( of -0.81 eV versus saturated calomel electrode (SCE)) under blue light triggers single electron transfer (SET) reduction of the alkoxy sulfonium salts 1 with the generation of alkoxy radicals I. The highly reactive alkoxy radicals add to the alkenes, generating benzylic radicals II, which are oxidized to the corresponding carbocations III (Eof +0.74 eV versus SCE) by PC1 ( +1.29 eV versus SCE), closing the photocatalytic cycle (path a). Interception of the carbocation intermediates III by a nucleophile -- water, alcohol or nitrile -- furnishes the products of alkene 1,2-difunctionalization 4-6. Mechanistic studies were conducted to probe the proposed catalytic cycle (Fig. 3c,d). Stern-Volmer fluorescence quenching studies revealed that the alkoxy sulfonium salt (for example, 1a) is the only quencher of the excited state of PC1, supporting the SET reduction of the salt in the first step of the reaction. Cyclic voltammetry studies on 1a found a more positive reduction potential (Eof -0.14 eV versus Ag/AgCl) than the photoexcited PC1, further supporting the proposed SET process. Addition of alkoxy radical I to the alkene was supported by a radical trapping experiment conducted under the standard reaction conditions in the presence of the radical scavenger 2,2,6,6-tetramethylpiperidine-1-oxyl. The 2,2,6,6-tetramethylpiperidine-1-oxyl-adduct 7, resulting from interception of radical intermediates II, was detected by high-resolution mass spectrometry (HR-MS). The involvement of alkoxy radicals is also supported by the results obtained using sulfonium salt 1aj, bearing a β-branched alkoxy motif; the introduction of 4-methyl-quinoline 8, under otherwise standard reaction conditions, gave Minisci-type coupling product 9 and aldehydes 10a,b arising from radicals formed by β-scission of the alkoxy radical species. The intermediacy of benzylic carbocations III is also corroborated by the detection by HR-MS of a sulfonium salt corresponding to IV formed by reaction of III with dibenzothiophene 2' liberated in the early stages of the catalytic cycle. A quantum yield (ø) of 1.02 was measured for the coupling of 1a and 3a to give 4a, thus suggesting the involvement -- at least partially -- of a radical-chain mechanism, in which II reduces the starting alkoxy sulfonium salt 1a (Fig. 3b, path b).