Scalable and sustainable electrochemical allylic C-H oxidation.

New methods and strategies for the direct functionalization of C-H bonds are beginning to reshape the field of retrosynthetic analysis, affecting the synthesis of natural products, medicines and materials. The oxidation of allylic systems has played a prominent role in this context as possibly the most widely applied C-H functionalization, owing to the utility of enones and allylic alcohols as versatile intermediates, and their prevalence in natural and unnatural materials. Allylic oxidations have featured in hundreds of syntheses, including some natural product syntheses regarded as "classics". Despite many attempts to improve the efficiency and practicality of this transformation, the majority of conditions still use highly toxic reagents (based around toxic elements such as chromium or selenium) or expensive catalysts (such as palladium or rhodium). These requirements are problematic in industrial settings; currently, no scalable and sustainable solution to allylic oxidation exists. This oxidation strategy is therefore rarely used for large-scale synthetic applications, limiting the adoption of this retrosynthetic strategy by industrial scientists. Here we describe an electrochemical C-H oxidation strategy that exhibits broad substrate scope, operational simplicity and high chemoselectivity. It uses inexpensive and readily available materials, and represents a scalable allylic C-H oxidation (demonstrated on 100 grams), enabling the adoption of this C-H oxidation strategy in large-scale industrial settings without substantial environmental impact.

Electrochemical oxidation presents an attractive alternative to traditional chemical reagents for large-scale applications, in large part due to the generation of less toxic waste than that produced by current chemical processes[4,5]. In addition, electrochemical conditions are compatible with a wide range of functional groups[6-10], tend to have higher overall energy efficiency as compared to thermal processes, and, due to their limited use, offer new intellectual property space for small-molecule synthesis[4]. The first electrochemical allylic oxidation was reported by Shono and coworkers in 1968[11,12] (). Direct oxidation of α-pinene (1) led to the fragmentation of the cyclobutane ring with incorporation of methanol or acetic acid (depending on the solvent) to give products 2 in 22–24% yield. A major advance in this field came in 1985, when Masui and coworkers reported the indirect oxidation of 1 using N-hydroxyphthalimide (NHPI) as an electrochemical mediator[13-15] (). Unfortunately, verbenone (3) was isolated in only 13–23% yield. Although these reactions are not useful in a preparative sense, they were a proof of concept that served as a foundation for our work.

In our own laboratory, systematic and extensive experimentation led to the identification of three modifications of the original Masui precedent which transformed this process into a synthetically useful electrochemical allylic C–H oxidation (). As described below, these modifications include the addition of a simple co-oxidant, the identification of a new electrochemical mediator, and the design of a reliable and inexpensive setup.

From the outset of this work, we avoided the use of expensive electrodes such as precious metals (e.g. Pt or Au), focusing our efforts exclusively on carbon. Initial optimization was undertaken using graphite rods, but despite clean conversion of starting material to desired product, mass recovery was typically low. We considered that this may in part be due to absorption of the substrate onto the graphite. Switching to reticulated vitreous carbon (RVC, 100 ppi available from K.R. Reynolds Co. for ca. $3/electrode) electrodes proved to be far more productive.

In our hands, the original Masui conditions[14] applied to valencene (4) led to only 6% isolated yield of nootkatone (5), the principal fragrance component of grapefruit aroma (). Our hypothesis was that air was the oxygen-atom source in this transformation, which was qualitatively confirmed by bubbling O2 gas in the reaction – resulting in an improved isolated yield of 18%. However, NHPI/O2 systems[16-18] have been explicitly avoided by the pharmaceutical industry, along with other oxygen-mediated reactions, due to the obvious challenges with flammability and other issues arising from reliably and safety in performing oxygen-mediated reactions on large scale[19,20]; as such, applications of aerobic oxidations in the pharmaceutical and fine chemical industries remain sparse[21-23]. Indeed, within Bristol-Myers Squibb such processes are explicitly avoided. Thus, a number of co-oxidants were evaluated, using NHPI as a mediator, and tert-butyl hydroperoxide (BuOOH) led to substantial increases in reaction conversion and reproducibility, delivering 5 in 51% yield. Using BuOOH without the NHPI mediator led to only 18% isolated yield under otherwise identical conditions.

With a suitable co-oxidant selected, attention turned to the optimization of base, solvent, and electrolyte. A variety of organic and inorganic bases were evaluated, with pyridine proving to be ideal. The use of acetone as solvent led to a slightly increased yield (56% 5 isolated) and was chosen as a general solvent for this reaction due to its ability to solubilize a wide range of organic substrates. Acetonitrile, dichloromethane, pyridine, or mixtures of these four solvents could also be used. The electrolyte LiBF4 could be used in place of LiClO4 with little decrease in yield, but tetraalkylammonium salts were not competent electrolytes.

Although most of the other mediators studied were inferior to the original NHPI, we reasoned that the addition of electron-withdrawing groups to the phthalamide scaffold would improve the reactivity of the catalyst[24]. Thus, tetrachloro-N-hydroxyphthalimide (Cl4NHPI) was chosen due to its ease of preparation from tetrachlorophthalic anhydride, an industrial non-toxic flame retardant (ca. $30/kg from VWR). The expectation of increased reactivity was supported by cyclic voltammetry data. In the case of NHPI, a reversible redox couple is observed at 0.78 V vs. Ag/AgCl in the presence of excess pyridine, whereas Cl4NHPI shows a redox couple at 0.87 V vs. Ag/AgCl under identical conditions. This slightly increased oxidation potential is consistent with the generation of a higher-energy and more reactive phthalimido-N-oxyl radical, and to our delight, the use of Cl4NHPI as a mediator led to a cleaner reaction profile and an isolated yield of 77%.

The final optimized conditions for oxidation of 4 to 5 is as follows: 20 mol% Cl4NHPI, pyridine (2.0 equiv), BuOOH (1.5 equiv), and LiClO4 as the supporting electrolyte (0.1 M) in acetone (6 mL/mmol substrate) under constant current conditions in an undivided cell. Notably, no precautions to exclude oxygen or water were undertaken, technical-grade solvents and reagents were employed, and a simple setup of two reticulated vitrous carbon (RVC) electrodes separated by a glass slide was used (see Supporting Information for a photographic guide of the experimental setup).

Our initial explorations into the tolerance focused on several cycloalkene-derived substrates of interest to Bristol-Myers Squibb process chemistry as part of ongoing drug discovery efforts (). Thus, tert-butyl cyclohexenone 6 could be prepared as a single regioisomer in 52% yield, while phenyl cyclohexenone 7 was prepared in 55% yield. Pleasingly, cyclopentenone 8 could be prepared from TMS-protected cyclopentenol in 52% yield. Unprotected tertiary alcohols 9–12 could be prepared from the corresponding aryl-substituted cyclohexenols in reasonable yields. Most notably, the Lewis basic pyridine-containing enone 12 was prepared in 60% yield. No alcohol elimination and aromatization was observed, and no allylic transposition occurred; this complimentary reactivity to most chromium-based oxidants and the use of unprotected alcohols is particularly notable. Furthermore, the unsubstituted cyclohexenone products 13 and 14 were prepared in 56% and 51% yield, comparing favorably to our previously reported CrV-mediated oxidation for the synthesis of 13. Substitution at the alpha position of the enone was tolerated as well, with 15 being isolated in 58% yield. In addition to the cyclic substrates described above, several acyclic alkenes of various chain lengths were successfully oxidized under the reaction conditions. Enones 16 and 17 were prepared in 46% and 51% yields, respectively. Interestingly, propargylic alcohol 18 was formed in 52% yield from its corresponding alkyne.

Due to their prevalence in the drug discovery, flavor, and fragrance industries, subsequent efforts focused on a variety of representative terpene classes. The oxidation products of monoterpenes are amongst the most widely used and correspondingly valuable substances not only as fragrances and flavors, but also as building blocks in synthesis. As such, we evaluated the efficiency with which the electrochemical allylic oxidation could be applied to these substances. Verbenone (3), prepared in 13–23% yield in Masui's original report, was prepared in 67% yield. Both isomers of the food additive theaspirane could be oxidized to the natural products cis- and trans-theaspirone 19 and 20 in 63% and 49% yields. Carvone-derived enone 21 was prepared in 47% yield, while carvone (22) was prepared in 42% yield. Myrtenol acetate and nopol acetate were converted to the corresponding oxo-myrtenol and oxo-nopol compounds 23 and 24 in 64% and 43% yield. Furthermore, aza-nopol analogues 25 and 26 were prepared in 42% and 53% yield, further highlighting the tolerance for nitrogen-containing functionalities under these reaction conditions. Notably, in all cases, isolated yields are comparable to those in the literature using other methods, and low product yields were obtained in all cases in our hands under Ishii's NHPI/O2 conditions after prolonged heating[16-18]. To put these results in context, an extensive comparative survey of literature conditions and yields is included in the Supporting Information.

Sesquiterpenoid and diterpenoid natural products, many of which are components of essential oils, have provided inspiration for countless novel strategies and methods in synthesis primarily due to their complex and dense structures as well as their promising biological activities[25]. Allylic oxidation of valencene gave nootkatone (5) in 77% yield (vide supra). Sclareolide-derived terpene 27 was synthesized in 75% yield using the electrochemical method. Eudesmane natural products carissone (28, as its TMS ether) and cyperone (29) were prepared in 54% and 51% yields, respectively. The related aubergenone skeleton was successfully oxidized to give enone 30 in 67% yield. Isolongifolenone (31) was prepared from the feedstock chemical isolongifolene in 91% yield. Natural products in the guaiane family were oxidized to afford the natural products pancherione acetate (32) and rotundone (33) in 41% and 44% yields, respectively, and the complex [3.2.1]-bicyclic system in cedrene was oxidized to give cedren-10-one (34) in 53% yield. Under electrochemical conditions, abietic acid derivative 35 could be prepared in 66% isolated yield along with 12% recovered starting material. Interestingly, while the starting material contained the Δ[7,8] bond, isomerization to the Δ[8,9] olefin occurred, and no Δ[7,8]-enone was observed. This anomalous result may be due to hydrogen-atom abstraction at the C-9 position followed by by trapping of the allylic radical at the C-7 position.

The oxidation of steroid and triterpene substrates has been shown to improve properties such as solubility and pharmacokinetics, and the development of tools to modify their “oxidation barcodes” are of immediate importance[26]. Electrochemical oxidation of acetate-protected dehydroepiandrosterone (DHEA) gave enone product 36 in 81% yield. Notably, unprotected DHEA could also be oxidized to give enone 37 in 72% yield. Diosgenin acetate underwent smooth oxidation to give 38 in 74% yield. Encouraged by the tolerance for free alcohols and sensitive acetals, glycosylated derivatives of DHEA were evaluated. The tetrabenzoyl protected glycoside 39 was prepared in 60% yield, and, remarkably, the unprotected glycoside 40 could be prepared in 38% yield, demonstrating chemoselectivity that would be out of reach using classical oxidants such as chromium. Free hydroxyl-containing methyl oleanolate derivative 41 was prepared in 43% isolated yield, while oxidation of the ketone-containing substrate gave the desired enone 42 in 46%. Methyl glycyrrhetinate (43), an important starting material for a variety of oxidized medicinally relevant triterpenes, was also prepared in 41% yield.

In nearly all substrates evaluated, isolated yields compare favorably to literature precedent using traditional reagent-based oxidants. In some cases, such as the known conversion of 4 to 5 using 15 equivalents of CrO3•pyridine in 80% yield, not only is the isolated yield comparable, but our conditions obviate the need for the excessive use of toxic reagents and minimize the use of solvent and aqueous media for extraction/isolation.

In order to demonstrate the feasibility of adopting this technology in a process setting, the described conditions were applied to substrates of relevance to Asymchem Laboratories. These reactions were conducted using inexpensive graphite plate electrodes in a beaker open to air (, inset photograph), and LiBF4 was used as the supporting electrolyte. Using this modest setup, 22 was prepared on 27-gram scale (198 mmol) from limonene in 44% yield, and 3 was prepared in 55% yield on 27-gram scale. Further scale-up to 100 grams (734 mmol) gave 46% yield of 3. Notably, carrying out the identical transformation with traditional chromium reagents (e.g. toxic chromium hexacarbonyl) would require at least 81 grams of chromium reagent followed by extensive efforts to remove Cr-based contaminants. Sterol 37 and its acetate 36 were produced in 48% yield (100 g, 347 mmol) and 62% yield (100g, 303 mmol), respectively. Highlights of this successful external field test include operational simplicity, safe procedure, simple workup, and ease of product isolation.

To verify the improved environmental footprint of the electrochemical allylic oxidation, we compared the Process Greenness Scores (PGS) for the electrochemical preparation of 36 to two known literature methods[20,27] (). The PGS is a method often utilized by industrial companies, such as Bristol-Myers Squibb, to evaluate the potential environmental impact of chemical manufacturing processes. The scoring parameters are closely aligned with the 12 principles of green chemistry[28], amongst which limiting waste generation and maximizing process efficiency are two main metrics of environmentally friendly processes. A greener reaction exhibits a higher PGS. Oxidation reactions, especially aliphatic C-H oxidations, are generally associated with lower than average PGS due to typically low process yields and the common use of toxic metal mediators in stoichiometric quantities. Unsurprisingly, the CrO3-mediated oxidation of deoxy-36 scored lowest in terms of PGS (32.1%). The RuCl3-catalyzed oxidation had an improved, albeit still modest, PGS of 37.1%. We were pleased to find that the electrochemical allylic oxidation showed a dramatically improved PGS of 55.8% (an improvement of >50%). This difference is significant and shows the step-change in applicability of this new technology. For comparison, the PGS for a typical amide bond formation (EDC, HOBt) ranges from 55–70%, while the PGS of the widely employed Pd-catalyzed cross-coupling of aryl halides with boronic acids falls between 45–60%. As a further testament to its robustness, the electrochemical oxidation was carried out using a 6-volt lantern battery[29,30] (), with valencene being converted to nootkatone in 37% yield with 15% recovered starting material.

Although this reaction is useful for the oxidation of numerous natural and unnatural carbon skeletons, it is not without its limitations. For example, while cyclic substrates are all reactive, not all acyclic alkenes give very high conversion to enone products, nor do electron-deficient alkenes. In some cases, allylic alcohol products were isolated alongside enone products, though with prolonged reaction times these products are converted to the desired enones. Yields for some substrate classes were modest (e.g. 32 and 33), in part due to incomplete conversion, substrate decomposition, or adsorption to the electrode surface. In nearly all cases, however, isolated yields are comparable to alternative procedures present in the literature.

Mechanistic aspects of the initiation step of this new transformation may have parallels to other NHPI-catalyzed oxidations[16-18] (). Thus, deprotonation of Cl4NHPI by pyridine, followed by anodic oxidation, leads to the tetra-chlorophthalimido N-oxyl radical species. Olefinic substrate 44 would then undergo hydrogen atom abstraction, regenerating Cl4NHPI as well as the relatively stable allylic radical species 45. Reaction with electrochemically-generated BuOO• would then give allylic peroxide 46, which, upon elimination of BuOH, affords enone 47 (see Supporting Information for more details).