Accessing Alkoxy Radicals via Frustrated Radical Pairs: Diverse Oxidative Functionalizations of Tertiary Alcohols

Abstract

Alkoxy radicals are versatile reactive intermediates in organic synthesis. Here, we leverage the principle of frustrated radical pair to provide convenient access to these highly reactive species directly from tertiary alcohols via oxoammonium-mediated oxidation of the corresponding alkoxides. This approach enabled various synthetically useful transformations including β-scission, radical cyclization, and remote C–H functionalization, giving rise to versatile alkoxyamines that can be further elaborated to various functionalities.

Alkoxy radicals are high-energy species because they are not stabilized by mesomeric effects or delocalization,¹ and their unique reactivities have been utilized for the synthesis of complex molecules.² Alkoxy radicals react with substrates via mainly one of three pathways: hydrogen-atom abstraction from C–H bonds, addition to unsaturated C–C π-bonds, or fragmentation via β-C–C bond scission; in each case, a carbon-centered radical is generated that can further participate in diverse transformations. Despite their synthetic potential, strategies for the convenient generation of alkoxy radicals are limited. Traditional methods rely on the preinstallation of oxygen-activating groups followed by thermolysis or UV photolysis.³ However, this approach often requires multiple steps, relatively harsh reaction conditions, and the activated precursors typically exhibit low stabilities.

In recent years, pioneering work has led to the development of alternative strategies that can afford access to alkoxy radicals directly from alcohols (Figure 1a). These reports focused on β-C–C bond cleavage of cycloalkanols via oxidation using a high-valent transition metal (Figure 1a, approach 1),⁴ activation using a hypervalent iodine agent (approach 2),⁵ light-induced ligand-to-metal charge transfer (approach 3),⁶ and proton-coupled electron transfer by means of photo-catalysis (approach 4) or electrochemistry (approach 5).⁷ These approaches have been successful in providing alkoxy radicals directly from free alcohol substrates, although approaches 1 and 2 predominantly require that the alcohol substrates contain strained ring systems and pendant electron-rich arene groups, respectively, which limits their scope. Approaches 3–5 are more general platforms for accessing alkoxyl chemistry from alcohols and have been elegantly applied in C–H functionalization or β-scission reactions. However, the downstream transformations of the incipient alkyl radical are currently limited to trapping with Michael acceptor alkenes (approaches 3 and 4), H-atom-donating thiols (approach 4), or halides and alkoxides (approach 5). In addition, challenges associated with the efficient delivery of photons at large scales may limit the scalability of the photochemical methods using traditional batch reactors.⁸ Thus, there remains a need for a generally applicable, operationally convenient, and scalable approach for accessing alkoxy radicals.

Figure 1.

(a) Established routes for accessing alkoxy radicals directly from alcohols (Approach 1: oxidation using a high-valent transition metal; Approach 2: activation using a hypervalent iodine agent; Approach 3: light-induced ligand-to-metal charge transfer; Approach 4: proton-coupled electron transfer). [Ru] = [Ru(bpy)₃](PF₆)₂; [Ir] = [Ir(dF(CF₃)ppy)₂(5,5′-d(CF₃)bpy)]PF₆; BI–OAc = 1-acetoxy-1,2-benziodoxol-3(1H)-one; DBAD = di-tert-butyl azodicarboxylate. (b) Prior work from our laboratory suggesting the prospect of using FRPs to access alkoxy radicals. (c) Synthetic approach developed in the present work.

In recent work, our group demonstrated that frustrated radical pairs (FRPs)—in this case, generated through a single-electron transfer event between a disilazide donor and an N-oxoammonium acceptor—can be used to achieve aliphatic C–H functionalization in a regioselective manner.⁹ In the course of this work, we found that potassium tert-butoxide (KO^tBu) can be oxidized by TEMPO⁺BF₄⁻ to form a ^tBuO^•/TEMPO^• frustrated radical pair (Figure 1b). Density functional theory calculations carried out on an ^tBuO–TEMPO adduct predicted that it would feature an exceedingly weak N–O bond (BDE: 11.6 kcal/mol) and spontaneously dissociate in a homolytic fashion to form the radical pair. Significant steric repulsion between the two components (closest H···H contacts are ∼2.1 Å) is also anticipated to contribute to the low stability of the adduct and its spontaneous dissociation (ΔG_dis = −7.5 kcal/mol). In the absence of an alkane substrate, a small amount of Me–TEMPO product was detected via β-scission of ^tBuO^• (Figure 1b).¹⁰ This finding suggested to us the possibility of employing such FRP chemistry to access alkoxyl radicals and the transformations that they enable. We envisioned that alkoxy radicals could be conveniently generated from free tertiary alcohols through a deprotonation–single-electron transfer sequence (Figure 1c). Upon formation of the alkoxy radical and subsequent β-scission, the incipient carbon-centered radical would be captured by simultaneously generated TEMPO^•, installing a versatile aminoxyl group that could be readily converted to various functionalities, including carbonyl and hydroxyl motifs using well-established methods.^9,11 Of note, although TEMPO⁺ and related oxoammonium ions have been extensively studied in the oxidation of primary and secondary alcohols,¹² the reactivity of oxoammonium ions with tertiary alcohols has not been established.

We set out to explore this proposed reaction strategy in the context of radical-induced C–C bond cleavage using 1-phenylcyclohexanol as a model substrate (1a; Table 1), both as a proof-of-principle and to facilitate a comparison with the established routes described above (Figure 1a). We initially evaluated conditions similar to those we previously reported for C–H functionalization using radical pairs, namely potassium hexamethyldisilazide (KHMDS) as the base, TEMPO⁺PF₆⁻ as the electron-acceptor species, in benzotri-fluoride (PhCF₃) solvent. In this system, KHMDS deprotonates the alcohol, and single-electron transfer from the alkoxide intermediate to the oxoammonium generates the desired alkoxy radical–TEMPO^• FRP (Figure 1c). This type of O-centered radical α to a saturated carbocycle is known to undergo facile ring-opening via β-scission,¹³ and the resulting carbon-centered radical can be trapped by TEMPO^• to furnish a new C–O bond.

Table 1.

Conditions Explored for Reaction Optimization^a

Entry	Additive	Yield of 1b	Yield of 1c
1	none	14%	18%
2	THF	13%	12%
3	18-crown-6	3%	2%
4	Ph3N	9%	13%
5	pyridine	7%	9%
6	1,10-phen	9%	1%
7	HMPA	9%	8%
8	benzophenone	32%	ND
9	benzophenoneb	57%	2%
10 c	benzophenone	72%	2%

^aOptimization was conducted on a 0.1 mmol scale. Yield was determined by ¹H NMR using mesitylene internal standard.

^b0.25 equiv of benzophenone.

^c2.5 equiv of TEMPO⁺PF₆⁻, 2.25 equiv of KHMDS, 0.45 equiv of benzophenone.

Under these initial conditions, 14% of the desired β-scission product 1b was formed, featuring an appended aminoxyl group (Table 1, Entry 1). However, we also isolated side product 1c in 18% yield, featuring additional functionalization of the α-position of the carbonyl group, presumably through a polar pathway involving α-deprotonation and nucleophilic addition to TEMPO⁺. We hypothesized that the single-electron oxidation of the alkoxide could be facilitated—and side product formation reduced—with the addition of a Lewis base that would coordinate to the potassium counterion. However, addition of tetrahydrofuran (THF), triphenylamine (Ph₃N), pyridine, and hexamethylphosphoramide (HMPA) did not improve the reaction outcome (Entries 2, 4, 5, 7), while readily oxidized 18-crown-6 was detrimental (Entry 3). Addition of 1,10-phenanthroline was effective in suppressing side product formation, although the overall yield of 1b was low in this case (9%; Entry 6). Interestingly, the use of benzophenone as an additive successfully suppressed side product formation and afforded the desired product in 32% yield (Entry 8). Preliminary NMR data suggest that benzophenone may also help attenuate the basicity of KHMDS and thus suppress the formation of 1c (see the SI for details). Further optimization of the reaction by changing the KHMDS/benzophenone ratio from 1:1 to 5:1 improved conversion to 57% (Entry 9), and finally, increasing the equivalence of TEMPO⁺PF₆⁻ and KHMDS further improved yield to 72% (Entry 10).¹⁴

With these optimized conditions, we explored the scope of the FRP-mediated functionalization of tertiary alcohols via β-scission with a diverse array of cycloalkanol substrates. On a synthetically relevant scale (0.2 mmol), the reaction proceeded smoothly with 1-phenylcyclohexanol (1a in 81% yield), 1-phenylcyclopentanol (2a in 74% yield), and 1-phenyl-cyclobutanol (3a in 72% yield) (Figure 2). Other electronically differentiated arenes α to the hydroxyl group, such as 4-Cl-C₆H₄ (4a), 4-F-C₆H₄ (5a), 4-t-Bu-C₆H₄ (6a) were well-tolerated (66–77% yield), although the product yield from a substrate containing electron-rich 4-OMe-C₆H₄ (7a) was slightly lower (52%). A substrate featuring pyridine as a substituent (8a) gave the product in a relatively low yield (33%), presumably due to limited solubility of the substrate in PhCF₃. Notably, good reactivity was also found with 1-methyl substituted cycloalkanols (9a and 10a), indicating that 1-aryl substitution leading to conjugated ketone product is not a prerequisite for this reaction, as is the case for some earlier methods.^7a,c,15 Functionalities such as difluoro (11a), acetal (12a), trifluoromethyl (13a), and tetrahydropyran (14a) were also compatible. Derivatives of adamantanone (15a), camphor (16a), and norcamphor (17a) containing bridged tri- and bicyclic structures all furnished the desired β-scission product, with substrate 16a showing a preference for the generation of a secondary over primary carbon-radical (regioisomeric ratio, r.r., of 2.5:1).

Figure 2.

Scope of β-scission reactions explored in this work. When a diastereomeric ratio (d.r.) is given, a * symbol indicates the carbon at which the diastereoisomerism is concerned. ^aElimination side-product observed (see the SI). ^bYield was determined by ¹H NMR. For substrates 7a and 18a, the alkoxide was preformed by mixing the alcohol with KHMDS and benzophenone followed by addition of TEMPO⁺PF₆⁻ (see SI); these substrates were found to undergo side reactions with TEMPO⁺ directly, and pre-mixing with the base to form the alkoxide reduces the influence of this process.

Natural terpene compounds sabinene hydrate (18a) and globuol (19a), which contain fused ring systems, were also viable substrates; these yielded products in which one of the two possible β-C–C bonds was preferentially cleaved, presumably the one which leads to the more stable carbon-centered radical. A similar result was found when estrone derivative 21a was used (i.e., β-C–C bond cleavage preferentially generated a tertiary versus a primary carbon-centered radical), while for cholesterol derivative 20a, there was no distinction between the two β-C–C bonds, which would both form primary radicals. Finally, our method can also accommodate acyl radical generation, as evidenced from the success of reactions carried out using the essential oil 2-hydroxy-3-pinanone (22a) and an isatin derivative 23a, which yielded acyl aminoxyl derivatives 22b and 23b, respectively. Finally, acyclic tertiary alcohols 24a and 25a also underwent β-scission to yield the desired alkoxyamine products.

We next went on to explore the utility of our approach for carrying out other transformations known for alkoxy radicals. For example, we investigated several tertiary alcohol substrates featuring pendent alkenes that could undergo 3-exo-trig and 5-exo-trig cyclizations via homolytic alkoxyl addition to the double bond (Figure 3a). For substrates with 1,1-dimethyl substitutions (26a and 27a), three-membered epoxide and five-membered tetrahydrofuran rings were formed in 81% and 85% yield, respectively, without further optimization of the conditions developed for the β-scission reaction. Interestingly, cycloalkanols 28a and 29a exclusively underwent radical cyclization without any sign of β-scission, furnishing spirocyclic products. When protected sclareol 30a was used as a substrate, the base-sensitive acetate group was tolerated under the reaction conditions. When the acyclic monoterpene linalool 32a was employed, 5-exo-trig radical cyclization was favored over the 3-exo-trig pathway (products formed in an 8.0:1 ratio). Notably, a sesquiterpene alcohol α-bisabolol 31a exclusively reacted at the prenyl alkene and did not yield any cyclization product with the endocyclic alkene, presumably due to unfavorable geometric constraints.

Figure 3.

(a) Scope of radical cyclization. (b) Scope of remote C–H functionalization. When a diastereomeric ratio (d.r.) is given, a * symbol indicates the carbon at which the diastereoisomerism is concerned. ^aStereochemistry undetermined due to the small amount of product obtained (for 32c) or the flexible nature of 7-membered rings (for 37a). ^bBis-aminoxylated side-product observed (see SI).

Another transformation of interest is intramolecular H-atom transfer (HAT), wherein a radical center is effectively transposed from oxygen to a carbon atom. Alkoxy radicals are known to undergo highly regioselective δ-C–H bond abstraction (1,5-HAT) via a six-membered ring transition state.¹⁶ We explored the utility of our approach for achieving such reactivity, starting with simple, linear substrates containing activated benzylic (34a) and allylic C–H bonds (33a) (Figure 3b). In both cases, δ-functionalized products were obtained albeit in moderate yields, which we attribute to the fact that the alcohol functionalities in the substrate and product are not effectively distinguished in the linear scaffold. In further scope study, we found that the use of rigidified substrates (35a–38a) resulted in improved yields (52–61%) even though the reactive C–H bonds exhibit higher bond dissociation energies. The nuclear magnetic resonance spectra of these products revealed deshielding of the O–H proton presumably through hydrogen bonding with the adjacent alkoxyamine group, which thus protected it from competing with the substrate’s hydroxy group. Other strained substrates containing benzylic (39a and 40a) and allylic δ-C–H bonds (41a) were also tolerated. Notably, starting from sesquiterpene cedrol (42a), 1,5-HAT was found to take place at a predisposed methyl group to furnish δ C–H aminoxylation product, while a side-reaction also generated a ring-opening product through β-scission. Finally, we note that primary and secondary alcohols are currently not suitable substrates for this reaction strategy due to the competing alcohol oxidation reactivity of TEMPO⁺.¹²

To demonstrate the synthetic utility of our FRP-mediated approach for accessing alkoxy radicals, we performed a gram-scale reaction with 2-hydroxy-3-pinanone (22a) (Figure 4a). Notably, the product was obtained in good yield (72%) without any optimization. Furthermore, this and other alkoxyamine products could be transformed to diverse useful functionalities under simple oxidizing or reducing conditions (Figure 4b). Treatment of the O-acyl TEMPO product 22b, which features a remote acetyl group, with diisobutylaluminum hydride (DIBAL-H)¹⁷ resulted in the selective formation of the corresponding hydroxy aldehyde without further reduction to a diol. The C–H aminoxylation product generated from cedrol (42b) was readily converted to natural product cedranediol (42d)¹⁸ via zinc mediated reductive N–O bond cleavage.¹⁹ This facile synthetic route stands in contrast to a previously reported synthesis of this cedranoid sesquiterpene compound, which required 10 steps and yielded a racemic mixture.²⁰ We also demonstrated the oxidation of alkoxyamine products to carbonyl compounds using 3-chloroperxybenzoic acid (mCPBA) as an oxidant. When sclareol derivative 30b was used as a substrate, the epoxide ring was preserved and aldehyde 30d was obtained in good yield. Under the same conditions, the major ring-opening product from (−)-globulol (19b) was efficiently converted into 1,10-dioxotayloriane 19d,²¹ affording another concise late-stage approach to natural product synthesis from abundant precursors.¹⁷

Figure 4.

(a) Gram-scale reaction. (b) Product derivatization reactions.

In summary, we have described a convenient and scalable method for accessing alkoxy radical chemistry from free alcohols. The simple deprotonation–single-electron transfer sequence can be used to furnish three different alkoxy radical-mediated transformations (β-scission, radical cyclization, and 1,5-HAT) using a broad scope of feedstock and complex alcohol substrates. The aminoxyl functionality of the products can be further converted to C–H, C–OH, and C=O groups. This approach thus provides convenient access to rare terpenes from abundant natural products.