Aliphatic C–H Functionalization Using Pyridine N-Oxides as H-Atom Abstraction Agents

Abstract

The alkylation and heteroarylation of unactivated tertiary, secondary, and primary C(sp³)–H bonds was achieved by employing an acridinium photoredox catalyst along with readily available pyridine Noxides as hydrogen atom transfer (HAT) precursors under visible light. Oxygen-centered radicals, generated by single-electron oxidation of the Noxides, are the proposed key intermediates whose reactivity can be easily modified by structural adjustments. A broad range of aliphatic C–H substrates with electron-donating or -withdrawing groups as well as various olefinic radical acceptors and heteroarenes were well tolerated.

Graphical Abstract

Visible-light-mediated photoredox catalysis has emerged as a powerful and reliable strategy for the functionalization of aliphatic C–H bonds in the past two decades. Typical approaches employ HAT catalysts that, upon single electron transfer (SET) with an excited photoredox catalyst, generate alkyl radicals by C(sp³)–H abstraction.¹ The majority of previous C–H functionalization studies focused on substrates containing C–H bonds alpha to heteroatoms² or at aldehyde,³ benzylic,⁴ and allylic⁵ positions with common bond dissociation energies (BDE) of <95 kcal/mol.⁶ Although these reactions benefit from high selectivities toward the weakest hydridic bonds, the HAT catalysts employed often lack reactivity for substrates with C–H BDEs ≥ 95 kcal/mol including unactivated tertiary, secondary, and, in particular, primary C–H bonds.⁷

Aside from halogen-⁸ and nitrogen-centered radicals,⁹ oxygen-centered radicals are most widely known to abstract strong unactivated C(sp³)–H bonds because of their high electrophilicity.¹⁰ Typical precursors of these reactive species comprise, among others,¹¹ oxo compounds such as aromatic ketones¹² and inorganic derivates (e.g., tetrabutyl ammonium decatungstate/TBADT),¹³ alcohols,¹⁴ benzoates,¹⁵ and phosphates,¹⁶ with the latter ones being employed by our group as well (Figure 1a).¹⁷

Figure 1.

Oxygen-centered radicals in C–H functionalization reactions.

Our continuous interest in developing photoredox catalyzed C–H functionalization reactions that are operationally simple, sustainable, highly efficient and selective, and allow structural fine-tuning of the catalysts led to the discovery of pyridine N-oxides as organic precursors for oxygen-centered radicals that allow HAT processes (Figure 1b).¹⁸ Although these compounds had previously been utilized as substrates in a limited number of visible light-mediated transformations,¹⁹ their catalytic application under photoredox catalysis remained unprecedented to the best of our knowledge. Wu and coworkers first described the generation of pyridine N-oxy radicals by single electron oxidation with an acridinium photoredox catalyst.^19c Based on this report and a study by the Zhang group who demonstrated the oxidation of water with pyridine N-oxides as HAT catalysts using electro-chemitry,²⁰ we hypothesized that they could also abstract H atoms from strong aliphatic C–H bonds using photoredox catalytic conditions. Moreover, pyridine N-oxides are commercially available and inexpensive or can be obtained in one step by simple oxidation of the corresponding pyridines with mCPBA or H₂O₂,²¹ resulting in a broad library of readily accessible and fine-tunable organic HAT catalysts.

Based on this vision, we found that cyclohexane as C–H substrate reacted with benzylidene malononitrile in the presence of 5 mol % of the photoredox catalyst Mes-AcrBF₄ and 20 mol % pyridine N-oxide A under irradiation with blue light (456 nm Kessil), providing product 1 in high NMR yield (Table 1, entry 1). Following this promising result, several other N-oxide derivatives were synthesized or purchased and subjected to the same reaction conditions. In general, electronrich pyridine N-oxides with E_ox < +1.8 V showed incomplete conversion of the radical acceptors (entries 2–4), while more electron-deficient ones with E_ox > +1.8 V readily provided product 1 in high NMR yields (entries 5–7), except for HAT H which exceeded the redox window of photoexcited Mes-AcrBF₄ (E*_red = +2.08 V) (entry 8). These results were also supported by DFT calculations²² providing lower BDEs of the O–H bond for the protonated HAT catalysts B-D (93–96 kcal/mol), while A and E-H showed significantly higher BDEs (99–109 kcal/mol) to efficiently abstract an H atom from cyclohexane with BDE(C–H, calc.) = 95 kcal/mol.

Table 1.

Optimization Reactions with Cyclohexane^a

entry	conditions	conversionb [%]	yield 1b [%]
1	HAT A	>99	89
2	HAT B	31	18
3	HAT C	12	9
4	HAT D	64	40
5	HAT E	>99	95
6	HAT F	>99	84
7	HAT G	>99	97 (92)c
8	HAT H	21	16
9	HAT G (10 mol %)	>99	91
10	HAT G (5 mol %)	>99	91
11	HAT G, DCMd	>99	83
12	HAT G, PhCFsd	66	48
13	HAT G, no light	0	0
14	no HAT	5	traces
15	HAT G, no Mes-AcrBF4	0	0

^aScale: 0.100 mmol of benzylidene malononitrile.

^bDetermined by ¹H NMR of the crude mixture with HMDSO as internal standard.

^cAverage isolated yield in parentheses (n = 2).

^dInstead of MeCN as solvent.

4-Acetylpyridine N-oxide G proved to be the most efficient catalyst generating 1 in 92% isolated yield (Table 1, entry 7). The catalytic activity of G was also confirmed by reducing its loading to 10 and 5 mol %, respectively, without loss of reactivity, albeit with slightly lower NMR yields (entries 9–10). Additionally, the choice of solvent is crucial in order to achieve high yields (entries 11–12), and control experiments revealed the necessity of light, the HAT catalyst, and the photocatalyst to observe reactivity (entries 13–15).

Various radical acceptors were investigated next (Chart 1a). Except for 1 and 2, which were synthesized in high yields, less electrophilic radical acceptors such as phenyl vinyl sulfone (product 3) provided poor results under the optimized conditions (Method A). It is noteworthy that previously reported C–H alkylation reactions using HAT and acridinium photoredox catalysts were generally limited to a very narrow radical acceptor scope, with mainly benzylidene malononitrile and other highly electron-deficient olefins being the only acceptors applicable.^8d,f,9a After extensive additional optimization (see Table S3 in the Supporting Information (SI)), 50 mol % pyridine N-oxide A in a mixture of MeCN/HFIP (7:3) afforded sulfone 3 with an excellent isolated yield of 85% (91% NMR yield). Although similar results were obtained with only 20 and 10 mol % of A, we settled on 50 mol % due to partial deoxygenation/decomposition of the HAT catalysts during the course of the reactions. This observation including NMR experiments has been discussed in more detail in the SI.

Chart 1.

Substrate Scope of the C–H Alkylation^a

^a Scales: 0.100 or 0.200 mmol of radical acceptor. Average isolated yields are reported (n = 2). Regio- and diastereoselectivities were determined by ¹H NMR of the crude mixtures. Only diastereomeric ratios other than d.r. ~1:1 are shown. ^bYields determined by ¹H NMR with HMDSO as internal standard. ^cWas isolated in its cyclized form 12′ (not shown, indicated by dashed arrow, see SI). ^dIn CH₂Cl₂ instead of MeCN/HFIP (7:3), isolated as the methyl ester upon treatment with trimethylsilyldiazomethane (10 equiv). ^ePartly contains cyclized product 26′ as a single diastereoisomer (not shown, indicated by dashed arrow, see SI). ^f42 h. ^g66 h.

The newly established reaction conditions (Method B) also provided better results for other olefins. Thus, simple vinyl ketones afforded products 4 and 5 in good to moderate yields. Moreover, less reactive 4-vinylpyridine was presumably activated by protonation with HFIP to deliver 6 in 37% yield. Other electron-poor olefins with substituents in the α- or β-position were well tolerated and gave rise to their corresponding products 7–9. However, the additional steric hindrance of two methyl groups at the reactive β-position led to a diminished yield of dinitrile 10.

In order to fully explore the synthetic potential of the oxygen-centered radicals generated from pyridine N-oxides, we subjected several electronically and sterically different C–H substrates to our reaction conditions with benzylidene malononitrile as the radical acceptor. C–H substrates with α-heteroatoms including ethers, alcohols, and amides (11–13), as well as aldehydes (14), unsubstituted cyclic (15–16), and acyclic alkanes (17) readily reacted using Method A in typically excellent yields (Chart 1b,c). As expected, tertiary C–H bonds were favored over secondary ones (17), but only with moderate regioselectivity.

Applying the reaction conditions of Method A to substrates containing short alkyl chains and electron-withdrawing groups such as esters, however, resulted in lower product yields (Chart 1d). Unexpectedly, Method B also provided better results in these cases as shown by comparison of both Methods for benzoate 19 and methyl ester 23.

1-Chloro-3-methylbutane selectively afforded two detectable regioisomers of product 18, with the tertiary position (C3) being favored over the primary ones (C4). In contrast to compound 17, the adjacent secondary position (C2) was not functionalized due to its close proximity to the electron-withdrawing chloride.

Compounds 19 and 20 were synthesized from n-butyl benzoate and n-butyl acetate, respectively, with all four possible regioisomers and with the C3-alkylated product being the major one followed by functionalization at C1. Notably, previous H atom abstraction reactions of O-alkyl esters predominantly occurred at the most remote and least deactivated methylene site, while stronger C–H bonds (e.g., the CH₃-group) often remained untouched.²³ Thus, our results imply that pyridine N-oxy radicals can be classified as rather strong HAT species. Moreover, a difference in regioselectivity was obtained when ester 19 was synthesized by Method A or B, with the latter one providing higher selectivities toward secondary over primary C–H bonds than under the conditions of Method A. The same trend was also observed for product 23.

Reducing the alkyl chain from n-butyl to n-propyl and ethyl acetate led to slightly lower yields of the corresponding esters 21 and 22 along with a change in regioselectivity favoring the C1–H bonds. Methyl valerate afforded 23 as three regioisomers in excellent overall yield using Method B and with good C4-selectivity. However, valeronitrile, carrying a more electron-withdrawing cyano instead of an ester group, reacted much slower to give rise to 24 in moderate yield. In this case, only two regioisomers (C4+C5) were detected and isolated since no functionalization occurred at the electronically less activated C3-position. Notably, other strong HAT species like TBADT provided a mixture of all three regioisomers (C3, C4 and C5) for valeronitrile, resulting in overall lower site-selectivity than under our conditions.^13c

Aliphatic carboxylic acids are typically unsuitable substrates for photoredox catalyzed C–H functionalization reactions with basic HAT catalysts due to their propensity toward decarboxylation upon deprotonation and SET.²⁴ Since pyridine N-oxides are slightly basic (pK_a(HAT A in MeCN) = 10),²⁵ mixtures of C–H alkylated and decarboxylated products were generally obtained using both Methods. However, the decarboxylation pathway was completely prevented by switching the solvent to CH₂Cl₂ for the reaction with n-butanoic acid. Hence, product 25 was isolated in high yield and with a regioselectivity of C3:C4 = 1.5:1. The comparatively higher amount of primary C4–H functionalization in this case implies that the solvent also plays a crucial role and MeCN or MeCN/HFIP (7:3) overall enables better regioselectivities.

Cyclohexanone furnished 26 in excellent yield and with high C3-selectivity.²⁶ Other groups have reported a similar trend regarding the regioselectivity of related transformations with this C–H substrate.^9a

The efficiency of pyridine N-oxides as strong HAT catalysts was eventually evaluated by subjecting two substrates with only primary C–H bonds to the reaction conditions of Method B for prolonged reaction times (Chart 1e). Both tert-butyl benzene and pivalic acid provided 27 and 28, respectively, in acceptable yields. The latter one was isolated without any decarboxylated byproduct, again highlighting the high chemoselectivity of this reaction in CH₂Cl₂.

Despite the major focus of this study toward electron-deficient olefins as radical acceptors, we were able to further extend the substrate scope by employing heteroarenes instead which required the use of a terminal oxidant (Chart 2). Based on reports by Molander and co-workers,²⁷ a quick optimization study proved to be fruitful (see Tables S4 and S5 in the SI). In the presence of K₂S₂O₈ as the terminal oxidant and trifluoroacetic acid (TFA) to activate the heteroarenes by protonation, isoquinoline was alkylated with cyclohexane, toluene, and N-methylacetamide to generate 29–31 in good yields. Benzaldehyde provided alcohol 32 in the absence of K₂S₂O₈ according to a spin-center shift mechanism.²⁸ Additionally, several heteroarenes including quinoline, pyridines, and benzothiazole afforded products 33–36 in acceptable to high yields. Control experiments (see Table S5 in the SI) revealed that sulfate radical anions (SO₄^•–) generated from K₂S₂O₈ upon homolysis under visible light²⁹ or thermal conditions³⁰ do not act as the HAT species in this approach, at least for cyclohexane as the C–H substrate. However, other possible pathways to generate pyridine N-oxy radicals either by oxidation with SO₄^•–,31 or by formation of an electron donor–acceptor (EDA) complex between pyridine N-oxide and the protonated heteroarene³² might take place as a background reaction but furnished significantly diminished yields without the photocatalyst under our optimized conditions.

Chart 2.

Substrate Scope of the C–H Heteroarylation^a

^a Scale: 0.200 mmol of heteroarene. Average isolated yields are reported (n = 2). ^bTFA (1.5 equiv), 18 h. ^cWithout K₂S₂O₈. Benzaldehyde used as C–H substrate.

In accordance to previous reports^8f,17 and kinetic studies (see SI), the proposed mechanism of the C–H alkylation reaction is displayed in Scheme 1.

Scheme 1.

Proposed Mechanism of the C–H Alkylation

After generating highly oxidizing Mes-Acr⁺* by photoexcitation, pyridine N-oxide undergoes SET to become an N-oxy radical. This electrophilic species (BDE = 99 kcal/mol) can abstract a hydrogen atom from a C–H substrate to afford an alkyl radical, which then reacts with an electron-deficient olefin. In the presence of slightly acidic HFIP, less reactive radical acceptors were preactivated by hydrogen-bonding of the protic cosolvent. The resulting electrophilic radical alpha to the EWG was reduced by the acridine radical Mes-Acr followed by protonation from the N-hydroxy pyridinium to deliver the C–H alkylated product along with the photoredox catalyst Mes-Acr⁺ and pyridine N-oxide, closing both the photo- and HAT catalytic cycle. These final steps were presumably also enhanced by hydrogen-bonding of HFIP to the EWG which increases the reduction potential of the radical alpha to EWG to oxidize Mes-Acr back to its ground state. A proposed mechanism for the C–H heteroarylation is provided in the SI.

In summary, we have accomplished a highly efficient protocol for C–H alkylation reactions using a synergistic combination of an acridinium photoredox catalyst and readily available pyridine N-oxides as HAT precursors. This purely organic approach allows the abstraction of tertiary, secondary, and even strong primary C–H bonds in the presence of electron-donating and -withdrawing moieties. A broad range of functional groups were tolerated, including aliphatic carboxylic acids that are otherwise prone to readily undergo decarboxylation, and the substrate scope regarding the olefinic radical acceptors was significantly extended compared with previously reported methods. Additionally, the same catalytic system was also applied to Minisci-type reactions demonstrating the high versatility of the established chemistry. The straightforward structural modification of pyridine N-oxides allows the fine-tuning of their electronic and steric properties, including a broad accessible range of BDEs (ca. 90–110 kcal/mol) and, thus, provides the opportunity for highly regioselective transformations which is currently ongoing in our laboratory.