N-Directed fluorination of unactivated Csp3–H bonds†

Site-selective fluorination of aliphatic C–H bonds remains synthetically challenging. While directed C–H fluorination represents the most promising approach, the limited work conducted to date has enabled just a few functional groups as the arbiters of direction. Leveraging insights gained from both computations and experimentation, we enabled the use of the ubiquitous amine functional group as a handle for the directed C–H fluorination of Csp3–H bonds. By converting primary amines to adamantoyl-based fluoroamides, site-selective C–H fluorination proceeds under the influence of a simple iron catalyst in 20 minutes. Computational studies revealed a unique reaction coordinate for the catalytic process and offer an explanation for the high site selectivity.

Due to the pervasiveness of uorine atoms in industrially relevant small molecules, all practicing organic chemists appreciate the importance of this element. As a result of its unusual size and electronegativity, uorine imparts unique physicochemical properties to pendant organic molecules. 1 For example, the strong C-F bond can prevent biological oxidation pathways, thereby thwarting rapid clearance and potentially improving pharmacokinetics of molecules. 2 Moreover, the installation of uorine or triuoromethyl groups, with their strong inductive effects, 2 can have a profound effect on the pK a of nearby hydrogen atoms. 3 These attributes, among others, have solidi-ed the importance of uorinated molecules in the medicinal, 1-4 material, 5 and agrochemical 6 industries. Yet, the same unique properties that make uorine atoms attractive chemical modiers also make their installation difficult. Consequently, new methods for site-selective uorine incorporation remain highly desirable. 7 Methods to construct Csp 2 -F bonds traditionally make use of the Balz-Schiemann uorodediazonization 8 and halogen exchange ("Halex" process). 9 Advances in transition metalmediated uorination have broadened access to Csp 2 -Fcontaining molecules, 10 but methods to access aliphatic uorides remain limited. Conventional methods to make Csp 3 -F bonds-such as nucleophilic displacement of alkyl halides 11 and deoxyuorination 12 -can have limited functional group compatibility and unwanted side reactions. A more efficient route to form aliphatic C-F bonds would target the direct uorination of Csp 3 -H bonds (Scheme 1). 13 Recent efforts with palladium catalysis employ conventional C-H-metallation strategies to target Csp 3 -H bonds for uorination. 14 Alternatively, radical H-atom abstraction can remove the transition metal from the C-H-cleavage step, thereby offering a promising approach for Csp 3 -H-bond functionalization. 15 With undirected C-H uorination, 16 however, selectivity remains a challenge in molecules without strengthdifferentiated Csp 3 -H bonds. 17 To overcome this, our group pioneered the directed uorination of benzylic Csp 3 -H bonds through an iron-catalyzed process that involves 1,5 hydrogenatom transfer (HAT) to cleave the desired Csp 3 -H bond. 18 Since this work, other groups have demonstrated directed Csp 3 -H uorination based on radical propagation that proceeds through an interrupted Hofmann-Löffler-Freytag (HLF) 19 reaction (Scheme 1a). These examples employ various radical precursors such as enones, 20 ketones, 21 hydroperoxides, 22 and carboxamides 23 to direct uorination to specic Csp 3 -H bonds. Since amines are ubiquitous in natural products and drugs, we sought to use amines as the building block of our directing group to achieve uorination of unactivated Csp 3 -H bonds (Scheme 1b). By using amines as the starting point, one could use the approach in straightforward synthetic planning for the late-stage functionalization of remote C-H bonds.
In the design phase of the project, we needed to devise a synthetically tractable N-F system that would enable 1,5-HAT and allow for uorine transfer (Scheme 1b). To begin, we decided to examine common amine activating groups that would support 1,5-HAT while avoiding undesired radical reactions. The chosen activating group would provide the ideal steric and electronic properties to enable both N-F synthesis and N-F scission for 1,5-HAT. We rst examined common acyl groups (e.g., acetyl-, benzoyl, and tosyl-based amides), but these proved unsatisfactory. For example, uoroamide synthesis was either not achieved or low yielding, and the desired uorine transfer proceeded with signicant side reactions or returned starting material. We then turned our attention to more sterically hindered amides-which allow for higher yielding uoroamide synthesis. For uorine transfer, we hypothesized that the increased steric bulk could slow intermolecular H-atom transfer, thereby leading more efficient intramolecular 1,5-HAT. To that end, we were delighted that pivaloyl-based uoroamide 1a proceeded in 64% yield to form product 2a (Scheme 2a). Interestingly, 7% of 1a underwent uorination at the tertbutyl group of the pivaloyl-presumably through a 1,4-HAT reaction (2aa, Scheme 2a). 24 The problem is further exacerbated when the pivaloyl group is homologated by one methyleneproviding only 7% yield of desired 2b with 32% of the uorination taking place on the iso-pentyl group (2bb, Scheme 2a). In an attempt to "tie back" the pivaloyl group and prevent the undesired uorination, we employed a cyclopropylmethylbased uoroamide but observed no improvement.
At this point, 1a proved most promising for efficient uorine transfer, as well as being the most synthetically accessible uoroamide. The increased steric hindrance minimizes N-sulfonylation during uorination with NFSI, a problem that plagued the synthesis of our previously targeted uoroamides. 18 Therefore, to further investigate how to improve uorine transfer from 1a, we decided to model H-abstraction computationally.
We hypothesized that the uorinated side product 2aa was formed aer 1,4-HAT. Since 1,4-HAT is rare, 24 we employed DFT (see ESI † for details) to calculate the 5-membered and 6-memebered transition-states for 1,4-and 1,5-HAT, respectively. Surprisingly, we found that the barrier for 1,4 C-H abstraction in 1a was 18.7 kcal mol À1 , which was only 2.6 kcal mol À1 higher in energy than the barrier calculated for 1,5 C-H abstraction in the same system (Scheme 2b). This suggested that both processes were competing at room temperature. We attributed the comparable barriers to the exibility of the tert-butyl group, which undergoes vibrational scissoring to accommodate the C-H abstraction. The transition state distortion is modest and allows the molecule to maintain bond angles close to the ideal 109.5 (Scheme 2b). Based on this insight, we sought to limit the scissoring of the tert-butyl group and prevent the 1,4-HAT that leads to the undesired side product. Aer investigating several possible candidates, the underutilized adamantoyl group appeared promising. To evaluate the rigidity of adamantane, we calculated the barriers for 1,4-and 1,5-HAT for the adamantoylcapped octylamine 1c (Scheme 2c). As expected, the barriers for 1,4-and 1,5-HAT differed signicantly-with 1,4 C-H abstraction proceeding with a barrier of 25.1 kcal mol À1 and the 1,5-HAT barely changed at 16.4 kcal mol À1 -an 8.7 kcal mol À1 difference. Consequently, we synthesized 1c and subjected it to the reaction conditions. Excitingly, the adamantoyl-capped system produced desired product 2c in 75% yield with no uorination of the adamantyl group (Scheme 2d). Using the newly devised adamantoyl-based uoroamides, the reaction conditions were optimized. While a range of metal salts, ligands, and radical initiators were evaluated, Fe(OTf) 2 proved unique in catalyzing uorine transfer with uoroamides. 18 Catalyst loading of 10 mol% allowed convenient setup and minor deviations above or below this loading had little effect on yield (see ESI †). Increasing the temperature to 40 C produced a slight increase in yield (entry 2, Table 1). Likewise, raising the temperature to 80 C resulted in full conversion of the starting material in 20 minutes with 81% yield of the desired product (entry 3, Table 1). It should be noted that uorine transfer occurs efficiently at a variety of temperatures with adjustments in reaction time (see ESI †). Increasing the reaction concentration or changing the solvent resulted in decreased yield (entries 4 and 5, Table 1). Furthermore, the absence of Fe(OTf) 2 leads to no reaction and quantitative recovery of starting material, attesting to the stability of uoroamides and the effectiveness of Fe(OTf) 2 (entry 6, Table 1).
With the optimized conditions established, we evaluated the substrate scope of the reaction ( Table 2). The reaction proved quite general for the uorination of primary and secondary Csp 3 -H bonds (2c-l, Table 2), while tertiary Csp 3 -H abstraction led to greater side reactions and lower yields (2m). While all reactions resulted in complete consumption of the uoroamide, only a singly uorinated product is produced with the parent amide being the major side product (see ESI †). The reaction proved selective for d-uorination even in the presence of tertiary Csp 3 -H bonds (e.g., 2h, 2j, and 2k), thereby demonstrating selectivity counter to C-H-bond strength. Interestingly, transannular uorine transfer occurs with complete regioselectivity to produce 2l as the sole product. Additionally, benzylic C-H bonds can be uorinated under these conditions (2n). The reaction also exhibits good functional group compatibility, allowing access to a variety of uorinated motifs. In particular, the reaction proceeds in the presence of either free or protected alcohols (2o and 2p). Moreover, esters and halides are both tolerated to give uorinated products 2q and 2r in good yield. Notably, the reaction provides access to uorohydrin 2s-highlighting the unique ability of this methodology to access both uorohydrins and g-uoroalcohols such as 2o. In addition to these examples, terminal alkene 1t works quite well giving 2t in 67% yield. Furthermore, alkene functionalizations of 2t would provide access to a diverse range of uorinated motifs. To target diuoromethylene units with this methodology, uoroamide 1u was prepared and subjected to the reaction conditions. Pleasingly, 2u was observed in 20% yield.
While exploring the substrate scope, we were surprised to discover that the uoroamide N-F bond is unusually stable to a variety of common reactions. For example, uoroamide 1o was carried through an Appel reaction, PCC oxidation, and Wittig reaction with minimal loss of the uoroamide. With such robustness, it becomes obvious that uoroamides could act as secondary amide protecting group-being installed and carried through a multi-step synthesis until uorine transfer is desired. Moreover, the greater rigidity of adamantoyl-based amides relative to pivalamides offers greater stability to acid and base hydrolysis-another feature of this system. Fortunately, the amide can be cleaved using conditions reported by Charette et al. with no evidence of elimination or loss of the alkyl uoride (see ESI †). 25 To evaluate the differences between C-H bonds, we calculated the hypothesized minima and maxima en route to C-F bond formation for primary, secondary, and tertiary substrates (Fig. 1). To begin, we dened the start of the pathway with the  uoroamides as octahedral, high-spin Fe(OTf) 2 -DME complex (I). 18 Ligand dissociation results in the loss of DME to form II which is 7.2 kcal mol À1 higher in energy relative to I. This ligand loss opens a coordination site that allows Fe to enter the catalytic cycle via F-abstraction from the uoroamides. This proceeds with a barrier (II-TS) of $25 kcal mol À1 for all systems to form the corresponding N-based radical (III). This new Nbased radical is generally about À10 kcal mol À1 from the starting materials. The 1,5-HAT proceeds through a sixmembered transition state (III-TS) with 16.4, 12.6, and 9.7 kcal mol À1 barriers for primary, secondary, and tertiary substrates, respectively. This abstraction forms the corresponding C-based radicals (IV) that were À15.0, À19.9 and À22.4 kcal mol À1 relative to the starting materials for primary, secondary, and tertiary substrates, respectively. A barrierless transition allows for the abstraction of uorine from Fe(III)-uoride to simultaneously furnish the products (V) and regenerate catalyst II. Interestingly, this transition seems to proceed with an intermolecular electron-transfer from the alkyl radicals to the Fe(III) center. The overall process is highly exergonic at À53.7, À58.6, and À61.9 kcal mol À1 for primary, secondary, and tertiary substrates, respectively. We attribute the low yields for the tertiary example to rapid oxidation of the carbon radical, likely by Fe(III), that forms a tertiary carbocation and leads to unwanted side reactions. The turnover-limiting step is the N-F abstraction by Fe (II-TS). An alternative pathway, related to the classic HLF reaction, 19a,b would involve radical chain propagation. Although unlikely, we also evaluated this pathway computationally ( Fig. 1). Consistent with our previous report, 18 this process proceeds with an unfavorably high barrier of 30.0, 28.1, and 26.8 kcal mol À1 for primary, secondary, and tertiary substrates, respectively. Hence, this process cannot compete with the barrierless delivery of uorine from the Fe(III) uoride species.
In conclusion, we leveraged critical computational insights to enable the use of simple amines as a building block for the directed uorination of C-H bonds. The reaction targets unactivated Csp 3 -H bonds site selectively regardless of bond strength. The reaction proceeds under mild iron catalysis that allows broad functional-group compatibility and provides access to unique uorinated motifs. Moreover, we identied uoroamides as surprisingly stable functional groups with likely implications for biology and materials. Mechanistic evaluation of uorine transfer with DFT provided a detailed reaction coordinate that explains the observed reactivity. The overall reaction and mechanistic insights should provide chemists a more predictable approach to site-selective uorination of C-H bonds.

Conflicts of interest
The authors declare no conicts of interest.