Photocatalytic Deoxygenation of Sulfoxides Using Visible Light: Mechanistic Investigations and Synthetic Applications

Abstract

The photocatalytic deoxygenation of sulfoxides to generate sulfides facilitated by either Ir[(dF(CF₃)ppy)₂(dtbbpy)]PF₆ or fac-Ir(ppy)₃ is reported. Mechanistic studies indicate that a radical chain mechanism operates, which proceeds via a phosphoranyl radical generated from a radical/polar crossover process. Initiation of the radical chain was found to proceed via two opposing photocatalytic quenching mechanisms, offering complementary reactivity. The mild nature of the radical deoxygenation process enables the reduction of a wide range of functionalized sulfoxides, including those containing acid-sensitive groups, in typically high isolated yields.

The deoxygenation of sulfoxides to generate sulfides is a fundamental transformation in organic synthesis¹ and biochemistry.² Established methods to convert sulfoxides into sulfides³ involve the use of low-valent metallic species,⁴ metal hydride reagents,⁵ halide ions,⁶ and phosphorus compounds.⁷ However, these reaction systems can suffer from potential disadvantages, including the use of expensive and/or toxic reagents, difficult workup procedures, and use of harsh reaction conditions, which often limit their functional group tolerance. Consequently, this is an area of continued research, and new, efficient procedures for the reduction of sulfoxides into their corresponding sulfides are desirable.

Over the past decade, photoredox catalysis has evolved into a vitally important method able to address long-standing challenges in synthetic chemistry,⁸ in large part due to the mild conditions by which reactive radicals can be generated. However, photocatalytic methods for deoxygenation of sulfoxides have rarely been explored.⁹ The cleavage of C–O bonds via β-scission of phosphoranyl radicals was initially recognized in the early 1970s by Bentrude,¹⁰ and since then, the groups of Zhu and Xie,¹¹Doyle,¹² and Rovis¹² have extended the synthetic application of this strategy to incorporate photoredox catalysis (Figure 1A), establishing valuable methods for deoxygenation of alcohols and carboxylic acids. More recently, this work was extended to include cleavage of N–O bonds by Yang et al.¹³ and also by Lardy and Schmidt,¹⁴ who employed more traditional radical initiation methods.

Figure 1.

Photocatalytic deoxygenation methods.

Inspired by these works, we speculated that direct cleavage of S=O bonds could be accomplished via a polar/radical crossover process between phosphine radical cations, generated from a photocatalyst (PC) initiator, and sulfoxides, resulting in mild deoxygenation of sulfoxides (Figure 1B). Based on existing phosphoranyl radical studies (Figure 1A) and the reported oxidation potentials of sulfides (e.g., diphenyl sulfide {E_1/2 = +1.43 V versus saturated calomel electrode [SCE]})¹⁵ relative to those of phosphines (e.g., PPh₃ {E_1/2 = +0.98 V versus SCE}),¹² a radical chain mechanism was proposed¹⁶ (Figure 1C, see later for a description).

We postulated that initiation of a radical chain deoxygenation process could be promoted by single-electron oxidation of PPh₃1 using a suitable oxidizing photocatalyst (initiator)¹⁷ to afford a catalytic amount of phosphine radical cation 2. Polar nucleophilic addition of sulfoxide 3 to radical cation 2 would generate phosphoranyl radical 4, which upon β-scission, would afford sulfide radical cation 6 and triphenylphosphine oxide 5. Finally, reduction of the sulfide radical cation 6 by PPh₃1 would afford the desired sulfide 7, as well as propagating the radical chain via regeneration of phosphine radical cation 2. Herein, we describe the realization of this radical chain process for the high-yielding deoxygenation of sulfoxides under mild, visible light-driven reaction conditons.¹⁸

Studies began by surveying the ability of a series of photocatalyst initiators (PC1–PC8, Scheme 1) to promote the reduction of 4-bromophenyl methyl sulfoxide 8a into sulfide 8b using PPh₃ as the terminal reductant and CH₂Cl₂ as the solvent, irradiating with a 60 W blue LED light¹⁹ under an argon atmosphere. In line with related literature,^11,12 both PC1 and PC2, which have excited-state oxidation potentials (M*/M^–) greater than that of PPh₃ (E_1/2 = +0.98 V versus SCE), afforded sulfide 8b in excellent yields.²⁰ Moreover, PC3 and PC6, which have excited-state oxidation potentials lower than PPh₃, resulted in much lower yields of sulfide 8b (11 and 5%, respectively) as expected. However, PCs possessing a far lower oxidation potential than PPh₃ (e.g., PC7 and PC8; the PCs that we originally considered to be the least likely to promote effective deoxygenation of 8a) unexpectedly promoted the formation of sulfide 8b in high yields. We noticed that all four PCs (PC4, PC5, PC7, and PC8) are able to initiate the reaction effectively despite their low excited-state oxidation potentials and relatively high excited-state reduction potentials (M⁺/M*). In contrast, the PCs with relatively low excited-state oxidation and reduction potentials (i.e., PC3 and PC6, for which both potentials are within the white area in Scheme 1) did not perform well in the reaction. These observations suggested that two mechanistic pathways may be viable, based on either a reductive or oxidative photocatalyst quenching cycle, with the route taken dependent on the redox potentials of the PC initiator used.

Scheme 1. Photocatalyst Initiator Screening.

To probe this possibility further, comparative control reactions were conducted, with Ir[(dF(CF₃)ppy)₂(dtbbpy)]PF₆ (PC2) chosen as a representative oxidizing photocatalyst and fac-Ir(ppy)₃ (PC8) as a representative reducing photocatalyst. In the absence of PC and light (entries 2 and 3, Table 1), no reaction occurred in either system. In the absence of PPh₃, no reaction occurred when employing the oxidizing PC2, although contrastingly, a small amount of conversion into sulfide 8b was observed when employing the reducing PC8. TEMPO drastically suppressed the efficiency of both reaction systems, supporting a free-radical reaction pathway (entry 5); triphenylphosphine oxide was the major product formed in these TEMPO reactions, presumably via the pathway described by Bentrude.¹⁰ Both reactions could be performed in other solvents or under an atmosphere of air but a reduction in yield was generally observed (entries 6–8). Other readily available phosphines, phosphites, and phosphinites were also able to promote sulfoxide reduction, albeit in reduced yields compared with PPh₃ (entries 9–11).

Table 1. General Reaction Condition Optimization^a.

entry	deviation from standard conditions	yield (PC2) (%)	yield (PC8) (%)
1	none	99	99
2	no PC	0	0
3	no light	0	0
4	no PPh3	0	9
5	3 equiv of TEMPO	2	3
6	under air	69	46
7	THF	85	69
8	toluene	44	99
9	PCy3	48	33
10	PPh2OEt	68	5
11	P(OPh)3	28	8

^aReaction conditions: 8a (0.20 mmol), PC2 or PC8 (1 mol %), PPh₃ (0.24 mmol) in CH₂Cl₂ (1.0 mL) at RT, 24 h. ¹H NMR yields reported based on a trimethoxybenzene internal standard.

Mechanistically, the single-electron oxidation of PPh₃ by the most oxidizing catalysts (e.g., PC1 and PC2, Scheme 1) is an established concept.^11,12 Consequently, initiation of the radical chain cycle when using such oxidizing PCs is proposed to occur via reductive quenching of the excited-state PC to generate the key phosphorus radical cation 2 required to initiate the proposed radical chain mechanism (Scheme 2A). To support this, Stern–Volmer quenching studies were conducted, confirming that the emission of the excited-state PC2 is quenched by PPh₃ (Scheme 2B). Furthermore, when DDQ, an organic oxidant, was used in sub-stoichiometric amounts (≥5 mol %) in place of the PC, sulfide 8b was produced in excellent yields (Scheme 2C), further supporting the notion that the generation of phosphorus radical cation 2 promotes an efficient radical chain process, as depicted in Figure 1C.

Scheme 2. Initiation via Phosphine Oxidation (PC2).

Stern–Volmer quenching study of Ir[(dF(CF₃)ppy)₂(dtbbpy)]PF₆ with PPh₃ in degassed CH₂Cl₂.

¹H NMR yields reported based on a trimethoxybenzene internal standard.

In contrast, PCs possessing lower oxidation potentials (M*/M^–) such as fac-Ir(ppy)₃ (PC8) should not be able to oxidize PPh₃, which is supported by the absence of emission quenching of the excited state of PC8 by PPh₃ (see the Supporting Information). It has also been documented that sulfoxides such as DMSO are unable to quench the emission of PC8.²¹ Nonetheless, contrary to these observations, which suggest that no reaction should occur, it was found that DMSO can be reduced to DMS in high yields (99%) when reacted with PC8 and PPh₃ under our standard reaction conditions (for conditions, see Table 1). We initially postulated that this may be a result of energy transfer from the excited state of PC8 to the sulfoxide,²² thus forming an excited-state sulfoxide species able to undergo deoxygenation. The low-yielding deoxygenation of 8a in the absence of PPh₃ (see earlier control reactions, entry 4, Table 1) offers some support for this hypothesis, and the direct deoxygenation of sulfoxides under UV irradiation has also been reported.²³ However, we could find no evidence of emission quenching of the excited state of PC8 by either sulfoxides 8a or 18a in Stern–Volmer quenching studies (in line with literature precedent).²¹ Furthermore, if energy transfer is involved, it is not clear why the redox properties of the various PCs would have such a pronounced influence on the observed reactivity. Based on these observations, it was considered more likely that initiation is mediated by a redox process. It is also clear from the synthetic results that phosphine plays a key role in the reaction. Thus, an alternative mechanism was postulated, in which PPh₃ and the sulfoxide interact to form an adduct that can initiate the radical chain mechanism via an initial oxidative quench of the PC (Scheme 3A).

Scheme 3. Initiation via Adduct Reduction (PC8).

Stern–Volmer quenching study of fac-Ir(ppy)₃ with 1:1 solution of sulfoxide 8a:PPh₃ in degassed CH₂Cl₂.

This alternative mechanism would proceed via an electron transfer from the excited-state PC to a sulfoxide–PPh₃ adduct, thus accessing a ground-state Ir⁴⁺ complex (M⁺/M, Scheme 1). This Ir⁴⁺ species (which is considerably more oxidizing than the corresponding *Ir³⁺ state) could afford the key phosphine radical cation 2 via phosphine oxidation (1 → 2), thus enabling the earlier proposed radical chain reaction to proceed. Following sulfoxide attack (2 → 4) and β-scission (4 → 6), the resultant sulfide radical cation 6 could then undergo reduction in a number of ways: (1) reaction with PPh₃, thus regenerating phosphine radical cation 2 and propagating the radical chain (depicted in Scheme 3A); (2) reaction with the reduced sulfoxide–PPh₃ adduct; and (3) reaction with the excited-state PC to form the oxidizing ground-state Ir⁴⁺ complex, which would then go on to propagate initiation via phosphine oxidation (1 → 2) (2 and 3 are not depicted in Scheme 3A). We first sought to identify the formation of the proposed sulfoxide–PPh₃ adduct spectroscopically, but regrettably, no evidence for phosphine–sulfoxide interaction was evident using ¹H/³¹P NMR or UV–vis spectroscopy (see the Supporting Information). Stern–Volmer quenching studies also revealed that a 1:1 mixture of PPh₃ and sulfoxide 8a did not quench the emission of excited-state fac-Ir(ppy)₃ (PC8) even at concentrations far greater than that found in the reaction (Scheme 3B). We therefore turned to cyclic voltammetry to see if we could observe a reduction potential consistent with oxidative quenching of the excited state of PC8. More encouragingly, a unique reduction process was observed (with an onset potential of approximately −0.8 V vs Ag/AgCl) when both the sulfoxide and PPh₃ were present in solution, which was absent when either of these reagents was omitted (Scheme 3C). This electrochemical data certainly suggests that the redox chemistry of the sulfoxide and PPh₃ is affected by the presence/absence of the other. At present, these findings still leave some questions unanswered (most pertinently, what the structure of the hypothetical sulfoxide–PPh₃ adduct could be), but the synthetic and mechanistic results do support the notion that an alternative mechanism for deoxygenation operates when a PC with a sufficiently reductive potential is used.

Next, attention was turned to probing the synthetic utility of the deoxygenation. A preliminary substrate screen was conducted from which the relatively oxidizing photocatalyst PC2 was identified as the most broadly effective PC (see the Supporting Information) and was taken into further substrate scoping studies (Scheme 4). Diaryl sulfoxides 9a–11a were all well tolerated; notably, sulfoxide 11a, incorporating an acid-sensitive Boc group, was converted into its sulfide 11b in 99% yield. Various sulfoxides bearing a single functionalized aryl group also worked well, including halogenated systems (e.g., 8b and 15b). Sulfoxide 14a, which contains an acid-labile silyl ether, was also an excellent substrate for this transformation, providing the corresponding sulfide 14b in 90% yield.

Scheme 4. Substrate Scope of Sulfoxide-to-Sulfide Reduction.

Reaction conditions: sulfoxide (0.30 mmol), Ir[(dF(CF₃)ppy)₂(dtbpy)]PF₆ (1 mol %), PPh₃ (0.36 mmol) in CH₂Cl₂ (1.5 mL) at RT, 24 h.

¹H NMR yields reported based on a trimethoxybenzene internal standard; isolated yields of products after column chromatography are reported in parentheses.

1 mol % Ir[(dF(CF₃)ppy)₂(d(CF₃)bpy)]PF₆ (PC1) and 4 day reaction time employed.

1 mol % fac-Ir(ppy)₃ (PC8).

48 h reaction time employed.

Yields of the corresponding sulfide observed by ¹H NMR spectroscopy based on a trimethoxybenzene internal standard are presented.

Importantly, the freedom to vary the PC (and in particular, to vary its redox properties) allows deoxygenation to be performed on a wide range of substrates. For example, when using the most oxidizing photocatalyst Ir[(dF(CF₃)ppy)₂(d(CF₃)bpy)]PF₆ (PC1), we were pleased to discover that sulfoxide 13a, which contains an unprotected alcohol, afforded sulfide 13b in 78% yield, which was a significant improvement upon the yield using PC2.²⁴ Sulfoxide 17a also reacted poorly with PC2 under optimized conditions (34% conversion), with this attributed to competing oxidation of the benzothiazole moiety in this substrate. To address this, we tested the deoxygenation of sulfoxide 17a using the less oxidizing fac-Ir(ppy)₃ photocatalyst PC8, and gratifyingly, the corresponding sulfide 17b was isolated in near-quantitative yield, further demonstrating the value of having complementary synthetic protocols based on both oxidizing and reducing catalysts (see the Supporting Information for more comparisons between the reactivities of PC2 and PC8).

Sulfoxide reduction was also performed on a wide range of dialkyl sulfoxides with varying alkyl chain lengths; all reactions progressed cleanly to furnish the desired linear (18b, 20b, and 21b) and cyclic (19b) sulfide products in excellent yields. Acetal protecting groups are also well tolerated by this procedure, with sulfide 24b generated in 90% yield. Complete reduction of sulfoxide 25a derived from N-Boc-protected methionine was also achieved, furnishing the corresponding sulfide 25b in 99% isolated yield. A list of low-yielding or unreactive substrates is presented at the bottom of Scheme 4 (26a–32a). We believe that the low reactivities of these substrates can generally be attributed to poor solubility of the sulfoxide starting material in CH₂Cl₂ or low nucleophilicity of the sulfoxide/sulfone starting material. Interestingly, aryl carboxylic acid-containing sulfoxide 29a undergoes deoxygenation when using PC2 (38% yield), but incomplete conversion of the sulfoxide starting material is observed alongside the formation of a side product.²⁵ When performing the deoxygenation reaction using PC8, the corresponding sulfide is formed cleanly in 67% yield, with the remaining mass balance composed of unreacted sulfoxide. In all the above scoping studies, the only byproduct formed is triphenylphosphine oxide, and no discernible side products were isolated except where explicitly stated.

Finally, to further demonstrate the functional group tolerance and utility of the procedure, the deoxygenation of a sulfoxide-containing agrochemical (33a) and drug molecule (34a) was investigated (Scheme 5). In these examples, low to moderate yields of sulfide products were observed when using PC2 under a range of conditions. However, upon switching to PC8 (with a greater reduction potential), far superior reactivity was observed. Thus, agrochemical agent fensulfothion 33a, which contains a phosphorothioate moiety, was cleanly reduced to its corresponding sulfide 33b in 96% isolated yield. Furthermore, sulmazole 34a, a cardiotonic drug containing an imidazopyridine ring, was converted into sulfide 34b in 62% yield under the same conditions; this was a more challenging substrate due to its limited solubility in a range of solvents.

Scheme 5. Biologically Active Sulfoxide Reduction^,

Reaction conditions: sulfoxide (0.30 mmol), fac-Ir(ppy)₃ (1 mol %), PPh₃ (0.36 mmol) in CH₂Cl₂ (1.5 mL) at RT.

¹H NMR yields reported based on a trimethoxybenzene internal standard and isolated yields of products after column chromatography are shown in parentheses.

48 h reaction time.

24 h reaction time.

Scalability can be a concern in photoredox-catalyzed processes, with an increased photon flux needed for large-scale photochemical reactions. Advances in flow chemistry technology have come a long way in addressing this problem; however, larger-scale photochemical reactions are still typically less straightforward to achieve experimentally compared to the scale-up of thermal reactions.²⁶ To demonstrate that this phosphine radical cation strategy can easily be adopted by researchers who do not have access to the necessary equipment to perform photochemical flow reactions, deoxygenation of sulfoxide 8a was performed on a 1 g scale, using DDQ as the radical chain initiator, affording sulfide 8b in 96% yield, with the rest of the mass balance consisting of unreacted sulfoxide 8a.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.0c00690.

• Experimental procedures and characterization data for all new compounds (PDF)

Author Contributions

^‡ A.K.C. and J.A.R.-A. contributed equally to this work.

The authors declare no competing financial interest.