δ-C–H Mono- and Di-Halogenation of Alcohols

Abstract

Alkoxy radicals have long been known to enable remote C-H functionalization via 1,5-hydrogen atom abstraction. However, methods for their generation traditionally relied upon highly oxidizing metals, ultraviolet radiation or preformed peroxide intermediates, which has prevented the development of many desirable transformations. Herein, we report a new, bench-stable precursor that decomposes to free alkoxy radicals via a previously unreported single electron oxidation pathway. This new precursor enables the fluorination and chlorination of remote C-H bonds under exceptionally mild conditions and in exceedingly high mono-selectivity. Iterative use of this precursor enables the introduction of a second halogen atom, granting access to remote dihalide motifs including CF₂ and CFCl.

Graphical Abstract

The development of new chemical reactions to functionalize inert C-H bonds holds great promise in organic chemistry. Directed metallation has been extensively exploited to functionalize sp³ C-H bonds in the past decade^1–3. However, this process remains largely limited to C-H bonds that are three bonds away from the directing atom due to the preferential formation of five-membered metallocycles. While the development of ligands and directing groups to favor more distal C-H palladation has been recently demonstrated⁴, heteroatom-centered free radicals provide a potentially powerful alternative means to access these more remote C-H bonds via the process of 1,5-hydrogen atom transfer (HAT)⁵, the Hofmann-Freytag-Löffler reaction and Barton nitrite ester reactions being early examples of such reactions.

In light of our lab’s previous efforts in remote radical C-H halogenation reactions via nitrogen-centered radicals^6,7, we became interested in using alkoxy radicals to accomplish the remote C-H halogenation of alcohols. Hydrogen atom transfer reactions of alkoxy radicals to produce remotely functionalized alcohols are particularly desirable due to the prevalence of alcohols in natural products and drug targets. However, existing methods to accomplish the remote halogenation of alcohols pose practical limitations (Scheme 1A)^8–12, since alkoxy radicals are typically generated using stoichiometric organotin¹³ or lead¹⁴ reagents, hypervalent iodine¹⁵ or unstable precursors^8–12, which are often not compatible with a variety of desirable transformations.

Scheme 1.

In this context, photoredox chemistry has furnished new means to generate heteroatom-centered radical intermediates from a range of convenient, synthetic precursors^16–18. These precursors undergo either a one-electron oxidation or reduction in the presence of a photoredox catalyst, and subsequently decompose to yield the heteroatom-centered radical. In our efforts to develop alcohol-directed δ-C-H halogenation, several examples of visible light-mediated alkoxy radical generation caught our attention^18–26. However, since common electrophilic halogenation reagents such as Selectfluor and N-chlorosuccinimide (NCS) are quenched by single electron reduction following halogen atom transfer, many of the previously reported alkoxy radical precursors, which decompose by single electron reduction (e.g. N-alkoxyphthalimides^19–21, N-alkoxypyridinium salts^22,25,26), are unsuitable for a redox-neutral cycle. Thus, for a redox-neutral photoredox cycle to be feasible with the desired halogenation reagents, the alkoxy radical must be generated via a single electron oxidation of the precursor. In certain cases, oxidative pathways have been used to directly generate alkoxy radicals from free alcohols using either cerium catalysis²³ or proton-coupled electron transfer (PCET)^27,28; however, the narrow redox window of these cerium catalysts has limited their application to new transformations while 1,5-HAT reactions of PCET-generated alkoxy radicals remain unreported. In order to overcome these limitations, we embarked on the development of a new alkoxy radical precursor with appropriate redox properties to enable remote alcohol C-H halogenation.

Therefore, drawing inspiration from previously reported decarboxylating precursors designed for nitrogen-centered radical HAT^29,30, we questioned whether the alcohol-derived oxyimino-acid directing group 1 recently reported by our lab⁴ for palladium-catalyzed C-H activation might serve as the required alkoxy radical precursor. One-electron oxidation of the deprotonated directing group could initiate a decarboxylative sequence to yield carbon dioxide, acetonitrile and an alkoxy radical competent for hydrogen atom abstraction (Scheme 1B³¹). Moreover, this directing group is easily prepared in a one-step condensation from the corresponding hydroxylamine using pyruvic acid, an inexpensive natural product.

To accomplish this oxidation, we tested a range of photocatalysts possessing different excited-state redox potentials using irradiation with 467 nm light. Using Selectfluor as a radical trapping reagent^30,32,33, the imino acid 1i underwent decarboxylation and intramolecular HAT to provide a remotely-generated carbon radical which was trapped to forge a carbon-fluorine bond. The choice of base was crucial to the success of this reaction, with a combination of cesium fluoride and cesium carbonate providing the product 2i in 67% yield (see SI for details) - we attribute the effectiveness of this base combination to a fine-tuning of the pH of the reaction in the mixed acetonitrile-water solvent system. Notably, the organic photosensitizer (4s,6s)-2,4,5,6-tetra(9Hcarbazol-9-yl)isophthalonitrile (4CzIPN)³⁴ could supplant the expensive iridium complexes typically reported for decarboxylation reactions³³. Control experiments conducted in the absence of photocatalyst and in the absence of light concluded that both light and photocatalyst were crucial to the reaction’s success.

Building on the success of this fluorination reaction, we questioned whether remote C-H chlorination might also be achieved. Testing NCS as a chlorine source and cesium carbonate as the base in acetonitrile as solvent, the desired remotely chlorinated product 3i could be obtained in 22% yield. By switching to ethyl trichloroacetate (ETCA) as a chlorine atom source, as recently disclosed by Reisman³⁵, the yield was improved to 63%.

All reactions were carried out on 0.2 mmol scale. All yields reported are isolated yields. PG = CONH(p-NO₂)C₆H₄. Certain products were isolated in the presence of another regioisomer: 2b (95:5), 2c (88:12), 2d (95:5), 2e (97:3), 2f (91:9), 2g (95:5), 2h (95:5), 2i (97:3), 2o (96:4), 2p (95:5), 2q (99:1), 2s (97:3). Trace difluorinated product was observed in nearly all cases (see SI for further details).

We then proceeded to test the scope of these halogenation reactions with respect to both fluorination (Scheme 2) and chlorination (Scheme 3). In order to reduce the volatility and ease the purification of the products, many of the resulting alcohol products were derivatized to carbamates using 4-nitrophenylisocyanate. Methylene and methine C-H bonds could be halogenated in moderate to good yields among a range of substrates bearing heterocycles (2h, 2i, 2o, 2q, 3h, 3i, 3o, 3q), azides (2g, 3g), phenolic ethers (2p, 3p) and nearby benzylic C-H bonds (2c, 2d, 3c, 3d). The natural product derivatives from tetrahydrogeraniol (2f, 3f), norbornane (2k, 3k) and 2-methylvaleric acid (2n, 3n) were all amenable to this halogenation procedure. Due to the propensity of Selectfluor to react with unsaturated systems, olefin- (3u) and alkyne-containing (3v) substrates were successful only for chlorination. In certain substrates, minor amounts of alternate regioisomers were observed - control experiments (see SI) indicate that these regioisomers arise mainly from 1,6-HAT as opposed to non-directed C-H abstraction processes.

Scheme 2.

Scope of the δ-fluorination reaction.

Scheme 3.

Scope of the δ-chlorination reaction

All reactions were carried out on 0.2 mmol scale. All yields reported are isolated yields. ETCA = ethyl trichloroacetate. PG = CONH(p-NO₂)C₆H₄. 3d was isolated in the presence of another regioisomer (94:6). 3q was isolated in the presence of product with eliminated chlorine (14 mol%).

To test the scalability of our protocol, we attempted the remote fluorination reaction using one gram of azide-containing starting material 1g. This reaction proceeded smoothly, providing the fluorinated azido-alcohol in 61% isolated yield in only 15 minutes (see SI for procedure). This process is particularly attractive since the oxyimino acid precursor is bench-stable for months, thermally stable up to 100 °C and requires no transition metal catalysts for radical generation. Moreover, this protocol uses Selectfluor, one of the cheapest and most easily handled electrophilic fluorine sources³⁶.

Given the power of this methodology for site-selective monohalogenation, we questioned whether re-installation of the oxyimino acid directing group might enable the remote C-H halogenation reaction to be performed a second time to form valuable dihalogenated methylene groups. Difluoromethylene groups, known for their isosteric and isopolar relationship to oxygen, have played an important role in devising more potent protease inhibitors³⁷ and nucleoside analogs³⁸. Difluoromethylene groups are typically formed by reacting deoxyfluorination reagents such as (diethylamino)sulfur trifluoride (DAST) with pre-installed ketones. Given the hazards associated with aminosulfuranes and the reaction requirements for cryogenic temperatures and long reaction times, other methods to incorporate difluoromethylene groups are highly sought after^39–41. Using our fluorination methodology, it was possible to install difluoromethylene groups at remote positions in moderate yields (Scheme 4).

Scheme 4.

Scope of the dihalogenation reactions

This dihalogenation strategy could also be extended to generate chlorofluoromethylene groups, albeit in lower yields (Scheme 4). Aliphatic chlorofluoromethylene groups are typically encountered in refrigerants and fire retardants. While methods to introduce the chlorofluoromethylene group into complex molecules as part of a cyclopropane ring are well established through additions of chlorofluorocarbene to olefins⁴², acyclic chlorofluoromethylene groups are typically synthesized as mixtures of regioisomers using elemental halogens⁴³, making their selective introduction to complex molecules extremely difficult. Our strategy is, to our knowledge, the first method for the regioselective introduction of the chlorofluoromethylene motif to unfunctionalized aliphatic chains.

All reactions were carried out on at least a 0.1 mmol scale. ETCA = ethyl trichloroacetate. PG = CONH(p-NO₂)C₆H₄. All substrates were isolated as inseparable mixtures with their monofluorinated analogs (X=H). Certain products were isolated as regioisomers: 5c (96:4), 5f (98:2), 6c (99:1). Yields of the stated product were estimated using either ¹H or ¹⁹F NMR of the isolated mixture. Reaction conditions: X=F: 1 mol% 4CzIPN, CsF (2.20 eq.), Cs₂CO₃ (0.5 eq.), Selectfluor (2.0 eq.), MeCN:H₂O (0.05 M, 4:1) 467 nm hv, 10 min, 40 °C, argon atmosphere; (ArNCO, catalyst, r.t., 12 – 18 h). X=Cl: 3 mol% 4CzIPN, Cs₂CO₃ (1.1 eq.), ETCA (2.0 eq.), MeCN (0.05 M), 467 nm hv, 18 h, 40 °C, argon atmosphere; (ArNCO, catalyst, r.t., 12 – 18 h).

The stark differences in the times required for the fluorination and chlorination reactions led us investigate the reaction mechanism. Suspecting that a radical chain might be in operation in the fluorination reaction⁴⁴, the quantum yields of the two reactions were measured (Figure 5A) as Φ = 0.070 for the C-H fluorination and Φ = 0.016 for the C-H chlorination. Although these results are inconclusive evidence for a radical chain in either reaction, they do show that despite a lower photocatalyst loading, the remote C-H fluorination reaction is photochemically a more efficient process than the C-H chlorination.

We also compared the rate of quenching of the carbon-centered radical from 1,5-HAT through a radical clock experiment⁴⁵ (Scheme 5B). Using a neophyl-type substrate as the clock⁴⁶, we found that the rate of quenching for the carbon-centered radical was over 54 times faster in the fluorination reaction than in the chlorination reaction. Taken together with the quantum yield measurements, these mechanistic results may help justify the large discrepancy in the times required for these reactions.

Scheme 5.

Mechanistic studies

In summary, we have developed a new precursor to access alkoxy radicals via photoredox catalysis. Using these alkoxy radicals, remote carbon radicals generated by hydrogen atom transfer may be quenched with fluorinating and chlorinating reagents to forge new carbon halogen bonds. By reinstallation of this directing group, valuable dihalomethylene groups may be generated at remote positions in the molecule. We expect that this new radical precursor will enable a diverse range of new transformations to be developed, both by expansion of the remote C-H transformations reported herein as well as by taking advantage of the other reaction modes of alkoxy radicals, including beta-scission and addition across unsaturated bonds.