Introduction

Unnatural α-tertiary amino acids (ATAAs) are frequently employed as biologically active molecules and building blocks in synthetic organic chemistry1,2,3. Over the past decade, various methodologies have been developed for synthesizing unnatural ATAAs, such as enzymatic4 or transition metal catalysis5,6, as well as radical methods7,8,9,10,11, among others. Among these strategies, the catalytic C(sp3)–H/C(sp3)–H cross-dehydrogenative coupling (CDC) reaction represents one of the most straightforward routes for accessing complex ATAAs with superior properties from easily accessible compounds12,13,14,15,16,17. Despite their versatility, these reactions primarily rely on metal catalysts (e.g., Pd, Ni, Cu and others), often assisted by a chemical oxidant, harsh conditions (temperatures up to 140 °C), and/or large amounts of solvent and C-H feedstock (Fig. 1a)18,19,20,21,22,23,24. These factors hinder the late-stage functionalization of bioactive pharmaceuticals and industrial application. Undoubtedly, the development of a mild and general strategy for accessing diverse ATAAs derivatives by CDC reaction with abundant hydrocarbon feedstocks as C-H donors is a highly challenging and underexplored area. We envision an alternative C(sp3)–H/C(sp3)–H CDC reaction employing a photoinduced hydrogen atom transfer (HAT) strategy, instead of a metal catalyst.

Fig. 1: Development of CDC reactions for the synthesis of unnatural ATAAs.
figure 1

a Transition-metal-catalyzed CDC reaction for the synthesis of ATAAs. b The direct use of alkoxyphtalimides as precursors of alkoxy radicals. c This work: Metal-free C(sp3)-H/C(sp3)-H CDC reaction for the synthesis of ATAAs. NPhth N-(acyloxy)phthalimide, PC photocatalyst, TM transition metal, SET single electron transfer, HAT hydrogen atom transfer.

Full size image

Photoinduced hydrogen atom transfer (HAT) has recently emerged as a powerful and selective strategy for C(sp3)–H functionalization that is practical, inexpensive, and environmentally friendly25,26,27,28,29,30,31,32,33,34. Among the various HAT intermediates35,36,37,38,39,40,41,42,43, alkoxy radicals are widely regarded as one of the most versatile44,45,46, with particular focus on intramolecular 1,5-HAT47,48,49,50 or 1,2-HAT51,52,53 reactions. The Zuo54,55 and Stahl56 groups have developed alkoxy radicals as hydrogen atom transfer (HAT) agents for C(sp3)–H functionalization. Alkoxyphthalimides57, which are convenient precursors of alkoxy radicals, have been less studied due to their higher reduction potentials compared to acyloxyphthalimides (alkoxy derivatives: E1/2red = −1.8 V, acyloxy: E1/2red = −1.6 V)58. Moreover, the direct use of N-alkoxyphtalimides as precursors of HAT agents for intermolecular HAT to generate alkyl radicals has not been extensively explored (Fig. 1b). Intramolecular 1,2-HAT or β-scission of alkoxy radicals usually occurs before their intermolecular HAT with substrates, presenting a common challenge51,52,53,59,60. There is only one recent example, from the Aggarwal group, that described a metal-free C(sp3)–H borylation of alkanes using a HAT strategy61. In their study, the alkoxy radical produced by N-(trifluoroethoxy)phthalimide played a critical role in the transformation, despite the identification of a radical ‘ate’ complex as the crucial HAT agent (Fig. 1b).

We recently developed a metal-free method for the synthesis of valuable unnatural α-amino acid derivatives using aliphatic carboxylic acids as alkyl radical precursors62. In that process, the redox-active esters could be reduced by the excited state of oxazolones (E1/2red = −2.12 V vs Ag/Ag+ in DMF), followed by decarboxylative fragmentation to generated alkyl radicals. We hypothesized that a highly reactive alkoxy radicals would be produced by N–O bond cleavage, that is capable of cleaving strong C(sp3)–H bonds via HAT, when the N-alkoxyphthalimides were used instead of the redox-active esters. Followed intermolecular HAT between an alkoxy radical and an alkane (R = alkyl) would generate an alkyl radical, which would be rapidly couple with the persistent oxazolone radical to give a CDC product. However, several challenges must be addressed in order to realize this reaction. Firstly, previous studies have shown that azaallyl radical tends to dimerize under oxidizing conditions, which can lead to competitive side reactions63,64,65. Secondly, the regioselectivity of the reaction (C4/C2) must be carefully controlled. Thirdly, it remains unclear whether the alkoxy radicals will undergo the desired intermolecular HAT with an alkane prior to the β-scission process and intramolecular 1,2-HAT51,52,53,59,60. Finally, a suitable solvent is required to facilitate the enolization of the oxazolone, without participating in the reaction itself. Herein, we report a photoinduced metal-free C(sp3)–H/C(sp3)–H cross dehydrogenative coupling (CDC) reaction using a hydrogen atom transfer (HAT) strategy. Over 85 unnatural ATAAs with vastly diversified functional groups were synthesized with excellent regioselectivity (C4/C2 > 20/1) from abundant hydrocarbon feedstocks under mild conditions (Fig. 1c; metal-, PC-, and additive-free, ambient temperature). This metal-free strategy provided excellent functional group tolerance and Late-stage applicability. Mechanistic studies revealed the reactive trifluoro ethoxy radical act as a HAT agent in the reaction, abstracting a hydrogen atom from hydrocarbon to generate an alkyl radical.

Results and discussion

Initial optimization studies

We initiated an investigation into the metal-free CDC reaction using oxazolone 1 and commercially available ethylbenzene 2 (4.4$ /100 mL) as model substrates (Table 1). After systematically screening various reaction parameters, we achieved 70% yield of the desired product 3 with 18:1 regioselectivity upon irradiation with purple light-emitting diode (LED) in t-BuCN for 12 hours, in the presence of alkoxyphthalimide H1 (Table 1, entry 1). Other electrophilic oxygen-centered radical precursors, such as H2, H3, and H4, showed lower reactivity and selectivity (Table 1, entries 2-4). The reaction exhibited lower efficiency when more electron-rich H5 and H6 were used (entry 5). Using other solvents instead of t-BuCN resulted in decreased yield and regioselectivity, and no product 3 was observed in DMF and PhCF3, respectively (entries 6-9). Control experiments confirmed that light irradiation and H1 were essential for the reaction (entries 10 and 11). No CDC product was detected in the absence of visible light, even upon heating at 120  °C (Table 1, entry 12). Furthermore, traditional oxidants or radical initiators, including (NH4)2S2O8, 2,2’-azobis(2-methylpropionitrile) (AIBN), t-butyl hydrogen peroxide (TBHP), benzoyl peroxide (BPO), and di-tert-butyl peroxide (DTBP), could not facilitate this transformation (Table 1, entry 13; see Supporting materials for details).

Table 1 Optimization of the reaction conditionsaView full size image
Full size table

Reaction scope

With the optimized conditions established, we investigated the substrate generality of the present metal-free CDC reaction (all examples with 5.0–10.0 equiv of C–H donor substrates). We started by investigating oxazolone derivatives with different substituted groups (Fig. 2). Various phenyl glycine derivatives with electron-withdrawing and donating substituents (F, Cl, Br, Me, OMe, Ph), as well as heterocycle and condensed ring, were all compatible with the reaction and gave the corresponding CDC products in moderate to high yields (41–81%, 4–11). Next, we explored the scope of the C(sp3)–H substrates. Benzylic and heterobenzylic C–H bonds are ubiquitous in bioactive molecules, and site-selective functionalization of such positions could have a broad impact66. As shown in Fig. 2, various primary C-H benzyl substrates bearing an electron-withdrawing or an electron-donating group at the para or meta position were identified as suitable primary C(sp3)–H donors for this methodology (12–22). The list of suitable groups included OMe (13), F (15), Cl (16), Br (18), and CF3 (19). This method also tolerated various easily hydrolyzed functional groups, such as OCF3 (20), SCN (21), and Bpin (22).

Fig. 2: The reaction scope.
figure 2

a Reaction conditions: oxazolone (0.1 mmol), (0.2 mmol), and alkanes (0.5 mmol) in t-BuCN (0.1 M), irradiation with a 10 W purple LED (395 nm) under N2 at room temperature for 12 h. Isolated yield. b dr ≈ 1: 1. c Alkanes details are provided in the supplementary materials.

Full size image

Various representative secondary C-H benzyl substrates were compatible with the reaction, delivering the corresponding products in good yields with excellent regioselectivity (23–28). The reaction conditions demonstrated excellent tolerance towards longer alkyl chains, including primary alkyl halide substituents (29-31). Tetralin (35) and indan (32–34) derivatives, substructures commonly present in many drugs such as indinavir, sertraline, and nepicastat67, were also well-tolerated, affording the target products in moderate to high yields (51-80%). The C(sp3)–H substrate bearing a naphthalene ring structure gave the desired product in 75% yield (36). Diarylmethane derivatives (37–40), pervasive as key motifs in a wide range of pharmaceutical agents68, delivered the CDC products in good yields (70–75%) The corresponding sterically hindered ATTAs derivatives (41–42) were obtained in moderate yield when tertiary C-H benzyl substrates were used as C(sp3)–H donors, which would be difficult to construct using traditional methods.

Functional groups such as SCF3, F, Cl, CN, and SiR3 not only influence the hydrogen atom transfer (HAT) process of the reaction but also readily participate in subsequent reactions, leading to the complexity of the reaction69,70,71,72. Consequently, substrates containing these functional groups on the reaction center carbon (the carbon of C-H bond) are less likely to undergo further multi-functional group transformation through C-H activation. Pleasingly, the substrates with these functional groups were also well tolerated in this method due to its mild reaction onditions, yielding the multifunctional products in reasonable yields (43–49). The CDC reaction also proceeded effectively in the late-stage functionalization of a number of complex molecules, including the retinoic acid receptor agonist derivatives (50–51) and dapagliflozi derivatives (52–53). In this redox-neutral system (Fig. 3), not only olefin (54), ether (55), and readily oxidized thioethers (56–57) could be employed as alkylation agents, but also aliphatic C(sp3)-H bonds with higher bond energy (≈ 100 kcal/mol)73 were demonstrated to be compatible with this methodology (58-68). Moreover, this CDC system exhibits excellent compatibility with simple silanes, displaying exclusive regioselectivity at the α-silyl position and leading to the formation of synthetically valuable products (69–75). However, compared to traditional C(sp3)–H substrates, it remains a great challenge to convert substrates with alkynyl (76–79) and tertiary ether (82) into corresponding allenic and sterically hindered amino acids due to the multiple reaction sites offered by C-H substrates74,75. Nevertheless, it is delightful to observe that our system can still efficiently convert such substrates into the corresponding CDC products in moderate yields (49-63%). This CDC reaction also had a good regioselectivity, HAT preferentially occurring at the more electron-rich and less steric benzyl sites (such as 24, 34, 38, 40, and 50). Furthermore, the CDC product 12 could be conveniently transformed into the corresponding free amino acid 83 in high yield with a one-step procedure (Fig. 3a). Treating 12 with pyrrolidine, the ring-opening amide product 84 was obtained in 71% yield (Fig. 3b). An asymmetric CDC strategy to construct chiral amino acid compounds was also investigated. When we added chiral phosphoric acids (CPA) to the standard conditions, the CDC product 85 was afforded in 40% yield with a promising enantiomeric ratio of 68.5:31.5 (see Supporting materials for details; Figure 6). To our delight, the target CDC products (86-88) were obtained in moderate yields with excellent stereoselectivities (up to 97:3 er) using chiral phosphoric acid L-1 as the catalyst under adjusted reaction conditions (Fig. 3c). This one-pot strategy involved a multi-step synthesis, utilizing 2,6-dichloro-4-nitropyridine as the hydrogen atom transfer (HAT) reagent (see Supporting materials for more details; Figure 8).

Fig. 3: The reaction scope and derivatization reaction.
figure 3

a Reaction conditions: oxazolone (0.1 mmol), H1 (0.2 mmol), and alkanes (0.5 mmol) in t-BuCN (0.1 M), irradiation with a 10 W purple LED (395 nm) under N2 at room temperature for 12 h. Isolated yield. b dr ≈1: 1. c Alkanes or Silanes (1.0 mmol). d Regioisomer ratio of C4/C2, 56 (3:1), 58 (3:1), 59 (2.5:1), 60 (5.7:1). e Operational details are provided in the supplementary materials.

Full size image

Mechanistic studies

A series of control experiments were conducted to gain insight into the nature of this new light-driven, metal-free CDC reaction. As shown in Fig. 4a, the cross-coupling product 3 was not observed in the presence of radical scavengers, such as (2,2,6,6-tetramethylpiperidin-1-yl)oxyl (TEMPO) and BHT, and the radical capture product 89 was detected by HRMS. The formation of 3 was completely inhibited when the reaction was performed under air, and the oxidation products acetophenone and 1-phenylethan-1-ol were detected by GC-MS (more details see Supporting materials; Figures 10 and 11). Moreover, the dimerization product of the oxazolone was obtained in 48% NMR yield without adding phenethane 2 under standard conditions, indicating that the oxazolone radicals were formed in the reaction (more details see Supporting materials; Figures 12 and 13). These results indicated that the benzyl radical and the oxazolone radical were formed in the transformation, and the reaction underwent a radical pathway. Furthermore, the UV/vis spectrum indicated that the absorption onset of the mixture of oxazolone 1 with H1 did not significantly differ from that of oxazolone 1 alone, thus excluding the possibility of an EDA complex formation (see Supporting materials for details; Figures 18–22). Additional evidence was obtained from in-situ 1H-NMR spectroscopy and 19F-NMR spectroscopy of the mixture of oxazolone 1 with H1, which showed no interaction between H1 and oxazolone 1 (see Supporting materials for details; Figures 15 and 17). Stern-Volmer experiments demonstrated the efficient quenching of the luminescence emission of oxazolone by H1. In contrast, no quenching was observed with 2, suggesting the absence of an interaction between H1 and the excited state of oxazolone during the transformation (see Supporting materials for details; Figure 23).

Fig. 4: Control experiments and proposed mechanism.
figure 4

a Radical inhibition experiment. b In-situ 19F-NMR spectroscopy. c Trifluoroethoxy radical trapping experiment. d Cyclic voltammetry experiment. e Proving photosensitive intermediates. f Intermolecular competition experiment. g KIE experiment. h The reaction quantum yield. i Proposed mechanism. TEMPO, 2,2,6,6-tetramethylpiperidin-1-oxyl; BHT, butylated Hydroxytoluene; KIE kinetic isotope effect; n.d. not determined.

Full size image

Previous reports have demonstrated that alkoxy radicals can act as hydrogen atom transfer (HAT) reagents, abstracting a hydrogen atom from alkane substrates to generate alkyl radicals and corresponding alcohols45,54. Upon visible light irradiation of the mixture of 1, H1 and 2 under standard conditions, trifluoroethanol was detected by in-situ 19F-NMR spectroscopy (Fig. 4b). In addition, when PhMe-d8 was used as a substrate and MeCN-d3 as a solvent, a new fluorine signal was observed by in-situ 19F-NMR spectroscopy, indicating that the trifluoroethoxy radical abstracted a hydrogen atom from the C-H substrate (see Supporting materials for details; Figure 16). Furthermore, under standard conditions, a three-component product 91 was obtained in 65% yield when 90 was used instead of phenethane 2, which was formed via an oxygen radical addition pathway (Fig. 4c). These results demonstrate that a trifluoroethyl oxygen radical is produced in this metal-free CDC reaction and can act as a HAT agent, abstracting a hydrogen atom from phenethane 2 to generate a benzyl radical.

Subsequent efforts were aimed at verifying whether this photochemical process was initiated by the direct photoexcitation of either oxazolone 1 or alkoxyphthalimide H1. Our recent findings demonstrate that the in situ-formed oxazolone’s excited state can function as a strong reductant upon direct photoexcitation by visible light. Moreover, the cyclic voltammetry studies revealed that the C(sp3)–H of alkanes (>1.5 V in t-BuCN) cannot be oxidized by the excited states of H1 (Eox = +1.46 V vs SCE)61. The redox potential of H1’s ground state was around −1.24 V vs SCE61, which could be easily reduced by the excited state of oxazolone 1 (E1/2red = −2.12 V vs. Ag/Ag+)62 to produce a trifluoroethoxy radical by N–O bond cleavage of H1. TMS-phenethane 92 was used as a substrate, and no benzyl radical-adding product 12 was detected, excluding the aryl radical cation mechanism in this process (Fig. 4d)76. Aggarwal’s work showed that H1’s absorption was around 400 nm and H1 could not be excited at around 460 nm in MeCN61. Moderate yields of product 3 were obtained under irradiation at 440 nm and 460 nm (the solution was yellow), respectively, in our control experiments. When toluene 93 was used as a C-H donor and solvent under standard conditions (the solution was colorless), only 10% of the target product 94 was obtained due to oxazolone 1 was difficult to enolize in a non-polar solvent and the reaction cannot been initiated (Fig. 4e). These observations suggest that the photochemical process was most likely initiated by excited state 1* rather than excited state H1*. However, it can not be completely ruled out that direct photoinitiation of H1 may also have played a role (see Supporting materials for details; Figure 31).

Recently, Houk and co-workers demonstrated that the CF3CH2O• radical exhibits superior efficiency in cleaving strong C(sp3)–H bonds via HAT compared to other alkoxy radicals (such as tBuO• and CH3O•)77. Furthermore, the trend of HAT capability for CF3CH2O• with different C-H bonds of alkanes was PhCH2Me > PhCH(Me)2 > PhMe (Fig. 4f). The results obtained in our system are consistent with those observed in Houk’s studies, further supporting the idea that the CF3CH2O• radical plays a critical role in the reaction as a HAT agent (more details see Supporting materials; Figure 32). The KIE study results indicate that the C–H cleavage step is the rate-determining step of this CDC reaction (kH/kD = 2.4, Fig. 4g). Moreover, the quantum yield of the coupling reaction between 1 and 2 was found to be 0.15, suggesting that the reaction probably undergoes a radical cross-coupling pathway (Fig. 4h). However, the radical-chain process cannot be rigorously excluded in this transformation.

Based on a series of control experiments and previous reports54,61,62,78,79, we propose the mechanism outlined in Fig. 4i. Upon visible light irradiation, the in situ-generated oxazolone enolate II is directly excited, resulting in the formation of excited state II*. An intermolecular single electron transfer (SET) between alkoxyphthalimide H1 generates trifluoroethoxy radical A and the persistent oxazolone radical C. Trifluoroethyloxygen radical A acts as a hydrogen atom transfer (HAT) reagent, abstracting a hydrogen atom from alkane B to generate alkyl radical D. The alkyl radical D subsequently couples with oxazolone radical C to generate CDC product E.

In summary, this manuscript presents a facile strategy for synthesizing valuable α-tertiary amino acid derivatives through CDC reactions with abundant hydrocarbon feedstocks. This method enables rapid construction of sterically hindered α,β-tetrasubstituted α-amino acids under mild conditions, including metal-, PC-, and redox agent-free reaction conditions at room temperature. This metal-free strategy exhibits excellent functional group tolerance and broad substrate scope. Moreover, it can employ simple alkanes to provide highly C(sp3)-enriched products. Overall, this simple radical approach offers a less expensive and less toxic alternative to classical methods for synthesizing α-tertiary amino acids.

Methods

General procedure for synthesis of α‑tertiary amino acid derivatives via CDC reaction

An oven-dried 10-mL Schlenk tube equipped with a stirrer was charged with the oxazolones (0.1 mmol, 1.0 equiv.) and the phthalimide H1 (0.2 mmol, 2.0 equiv). Then, 1.0 mL t-BuCN (0.1 M) was added followed by the alkane (0.5 mmol, 5 equiv) in glove box. The tube was sealed with a screw cap and took out from glove box. The reaction mixture was inserted into the PhotoSyn 3.0 reactor and irradiated using a 10 W LED lamp (395 nm) for 12 h. The reaction mixture was concentrated in vacuo and purified by flash column chromatography (petroleum ether/EA = 20/1).