Phosphoryl Radicals from Trivalent Iminyl Phosphines: A Photocatalytic Approach to N‐Phosphoryl Azetidines

ABSTRACT

We report that iminyl phosphines, a class of bench‐stable and readily accessible trivalent P‐centered radical precursors, enable the synthesis of N‐phosphoryl azetidines via formation of a key phosphoryl radical under visible‐light single‐electron transfer (SET) catalysis. Mechanistic investigations reveal that SET oxidation of the iminyl phosphine engages a previously unexplored SET‐triggered polar pathway that channels the reaction away from classical phosphine oxidation toward the controlled generation of a P‐centered radical. A combination of experimental evidence, including electron paramagnetic resonance (EPR) spectroscopy and optical studies, supported by density functional theory (DFT) calculations, were used to shed light on this unconventional reaction manifold. This cooperative radical–polar mechanism facilitates the radical ring‐opening of azabicyclo[1.1.0]butanes (ABBs), granting access to previously inaccessible P–N bonds.

Photochemically generated phosphoryl radicals from iminyl phosphines enable strain‐release addition to azabicyclobutanes, providing access to elusive N‐phosphoryl azetidines. Combined experimental, computational, and EPR studies reveal an unconventional water‐assisted photoredox pathway.

1. Introduction

Organophosphorus compounds are integral to a wide range of applications, from pharmaceuticals [1] and agrochemicals [2] to functional materials [3] as well as ligands in transition‐metal catalysis [4]. As such, the development of efficient and selective methods for P‐heteroatom bond formation, particularly under mild and functional‐group‐tolerant conditions, remains a longstanding objective in synthetic chemistry [5]. In recent years, photoredox catalysis has gained prominence as a powerful platform for generating phosphorus‐centered radicals, especially phosphoryl (R₂P(═O)·) [6] and phosphinyl (R₂P·) [7] species, useful in radical‐mediated phosphorylation reactions. However, despite substantial progress, these methods are still constrained by several limitations: (i) a narrow pool of structurally diverse P(III)‐radical precursors (1‐4) [6, 7], (ii) the dependence on air‐ and moisture‐sensitive (1‐ 4), pyrophoric or toxic reagents (1), which often necessitate the use of a glove‐box (Figure 1A) [8, 9, 10, 11], and (iii) the need for post‐reaction treatment with crystalline sulfur [8, 9, 10, 11, 12, 13, 14, 15], or oxidizing reagents (m‐CPBA, H₂O₂) [9, 10, 11, 16, 17] when employing trivalent phosphorus precursors. Furthermore, these reagents are often used with transition‐metal salts and organometallic reagents, raising concerns about cost, and environmental sustainability of these processes [6, 7, 16, 17]. Recent advances in energy‐transfer (EnT) photochemistry have shown that oxime carbonates and esters 5 [18, 19], as well as sulfonamides 6 [20, 21, 22, 23], can be used to the generation of N, O, and S‐centered radicals (Figure 1B, left). The utility of these radical precursors arises from their relatively low bond dissociation energies (BDEs) and unique electronic properties, such as electron‐deficient nitrogen centers and the lone‐pair repulsion between nitrogen and adjacent heteroatoms (O, S) [24]. In stark contrast, the development of analogous bifunctional reagents 7 featuring nitrogen–phosphorus (N–P) bonds has proven considerably more challenging (Figure 1B, right) [25]. The primary obstacle lies in the remarkably high bond strength of the N–P linkage, which resists homolytic cleavage. Even when phosphorus‐ centered radicals are generated successfully through direct photoexcitation or EnT, their reactivity remains hindered by fast radical recombination [26]. For this reason, potential of N–P bond for radical transformation still remains as an underexplored area in photocatalysis unlike other N‐heteroatom bonds (O, S).

FIGURE 1.

Development and application of stable trivalent phosphorus‐centered radical precursors. (A) Conventional limits of trivalent P‐radical precursors. (B) Imine based radical precursor and EnT limits for iminyl phosphines 7. (C) Background on SET oxidation of P(III) compounds and possible reaction outcomes. (D) Challenge associated with oxidative quenching cycle for 7 (top). Our Strategy: Merged SET/polar process to access P‐radical 24 (bottom). (E) Current art for strain‐release of azabicyclo[1.1.0]butanes (left). This work: SET‐enabled N‐phosphoryl azetidines synthesis via P‐N bond formation with 17 (right).

Literature evidence demonstrates that strategies based on SET oxidation of phosphines 8 are inherently challenging (Figure 1C), because the resulting phosphoranyl radical intermediates 11 preferentially undergo α‐scission‐1, α‐scission‐2, or more favorable β‐scission pathways, leading predominantly to neutral phosphine products 12, 8, and 15, respectively [27, 28]. Taking this into consideration, we sought to develop an alternative strategy to activate P‐heteroatom bonds in iminyl phosphines 17 via single electron transfer (SET) approach (Figure 1D). We initially evaluated an oxidative quenching manifold. However, there are still quite a few roadblocks and challenges to successfully access P‐centered radicals from P(III) precursors. Prior studies suggest that the phosphinyl radical 20 has very low reactivity toward alkenes and strained cyclic compounds and is characterized by high activation energy barriers (ΔG^‡) compared to the phosphoryl radical 24 [9, 10, 11]. This reduced reactivity may limit the effectiveness of these scaffolds when employed via an oxidative quenching (Figure 1D, top). Further, even if such a reaction would proceed, a subsequent oxidation step is needed to render the products stable and readily isolable [8, 9, 10, 11, 12, 13, 14, 15]. Building on recent studies that demonstrate the water activation of phosphine radical cations 9 and their utility as tunable intermediates for hydrogenation of alkenes [29] and to synthesize P‐chiral tertiary phosphine oxides (Figure 1C; R = H, β‐scission pathway) [30], we envisioned to develop an alternative activation strategy based on an opposite reductive quenching pathway. Here, iminyl phosphine 17 is oxidized via SET and followed by a polar activation mode, where the imine is acting as a leaving group thereby enabling a new route for the generation of reactive phosphoryl radical 24 (Figure 1D, bottom). However, several intrinsic constraints complicate the realization of this strategy. Foremost, the reduction potential of the iminyl phosphine 17 must fall within a regime that enables efficient photoinduced oxidation by the organophotocatalyst, while avoiding the oxidation of the substrate(s). In parallel, hydrolysis must proceed slower than the initial SET event to prevent premature degradation of the starting iminyl phosphine 17. Finally, the reaction manifold must suppress the thermodynamically favorable phosphine oxide 7 formation, a dominant pathway that typically outcompetes productive radical generation [28, 29, 30].

To test this hypothesis, we selected azabicyclo[1.1.0]butanes (ABBs) 26 [31, 32] as model substrates. Although significant progress has been achieved in the radical ring‐opening of bicyclo[1.1.0]butanes and related [n.1.1]propellanes (n = 1, 3, 4) families [33], analogous transformations of ABBs have remained largely unexplored. Recently, our group reported the radical ring‐opening of ABBs via an energy transfer (EnT) strategy employing sulfonyl imines 6 as bifunctional radical precursors (Figure 1E, left) [23]. Subsequent studies by Tlili, Qin, Pitts, Tantillo as well as Hong and co‐workers have marked further advances in ABB activation, while demonstrating the high interest and potential of this chemistry [34, 35, 36, 37]. Despite these advances, existing ABB radical ring openings are only confined to S‐centered radicals (25 and 27) [23, 34, 35, 36, 37], whereas the addition of P‐centered radicals to ABBs via an unconventional P‐N bond forming radical process will grant access to N‐phosphoryl azetidines 28 (Figure 1E, right). Given the scarcity of these motifs and the lack of general synthetic methods for their preparation [38, 39], we envision that the iminyl phosphines 17 could serve as robust, bench‐stable, phosphorus‐centered radical precursors useful for a variety of chemical transformations.

2. Results and Discussion

2.1. Reaction Development

We commenced our investigation by synthesizing iminyl phosphines 29, a bench stable and easily accessible source of P‐centered radicals [40]. Subsequent oxidation of 29 with elemental sulfur [8, 9, 10, 11, 12, 13, 14, 15] yielded the corresponding phosphine sulfide 30. Further oxidation of 30 with meta‐chloroperbenzoic acid (m‐CPBA) [9, 10, 11] afforded the phosphine oxide 31 as a yellow solid. (See Supporting Information, section 2.0). Preliminary studies involved determining the triplet energy (T₁) and spin distribution for compounds 29, 30, and 31 utilizing the freely accessible EnT decker machine learning‐based (ML) platform [41]. Bond dissociation energies (BDEs) were calculated through density functional theory (DFT) at the UωB97X‐D/Def2‐SVP level of theory in DCM, employing the SMD solvation model [10, 11].

This approach facilitates the assessment of dissociation probabilities under specific energy transfer conditions, particularly when the substrate's BDE is less than or equal to T₁ values [42]. According to the model predictions, compounds 29, 30, and 31 possess triplet energies of 45.0, 43.7, and 60.3 kcal·mol⁻¹, respectively (Figure 2A), which are within the accessible range of commonly used photosensitizers such as thioxanthone (TXO; T ₁ = 65.5 kcal·mol⁻¹) [42]. However, the calculated BDEs for these compounds are 52.3, 58.1, and 67.0 kcal·mol⁻¹, respectively, exceeding their predicted T ₁ values. Our experimental focus was directed towards validating these results for the radical ring opening of azabicyclo[1.1.0]butanes (ABBs). Specifically, iminyl phosphine 29 (1 eq.) was subjected to irradiation with a 400 nm LED light source in the presence of TXO (5 mol%) as photocatalyst, along with the model azabicyclobutane 32 (1.0 eq.). At the end of the reaction, the mixture was treated with 1.2 equivalents of sulfur under an argon atmosphere to oxidize the phosphine‐to‐phosphine sulfides, thereby facilitating the purification of products (Figure 2A). The formation of N‐thiophosphoryl azetidine 33 or the difunctionalized products were not observed. Instead, compound 34 was obtained in 15% yield, The presence of other side products such as the iminyl phosphine oxide 31 and hydrolysis products such as benzophenone 35, diphenyl phosphine oxide 36, diphenylphosphine amide 37 and diphenylphosphinic acid 38 [43] were observed by UPC² analysis. No dimers of the corresponding iminyl radical (Ph₂═N·) [23] or phosphinyl radical (Ph₂P·) [10] were detected in the reaction media, suggesting that those species do not form during the reaction. We then carried out the reaction of 32 with 30 and 31 under similar reaction conditions. However, both 30, having similar T₁ and BDEs as 29, and 31 remain unreacted. These observations along with the mismatch between the BDEs and T₁ values suggest the operation of a SET manifold over a classical EnT catalysis. This prompted us to conduct further screening of a library of PCs, additives and optimization of other reaction parameters such as solvent, concentration while avoiding sulfur treatment at the end of the reaction (See Tables S3–S6). Pleasingly, using diphenylphosphinic acid 38 (20 mol%), with 4DPAIPN (5 mol%) in 1,2‐dichloroethane (DCE, 0.05 M) under blue LED irradiation (456 nm) afforded N‐phosphoryl azetidine 34 with 80% ¹H NMR yield and a 75% isolated yield (Figure 2B, entry 1). Omitting additive 38 resulted in a reduced yield of 49% (entry 2). Substituting 38 with acetic acid (entry 3) or other acids did not improve the reaction outcome (See Table S4). Replacing 38 with addition of external water (H₂O) at 1.0 equivalent maintained the yield at approximately 50% (entry 4). Furthermore, addition of 20 mol% diphenylphosphine oxide 36 instead of 38 led to a significant decrease in yield to 32% (entry 5), likely due to its known propensity for single‐electron transfer (SET) oxidation, which can compete with the desired photochemical pathway for product formation [6]. Increasing the amount of additive 38 from 20 mol% to 1.0 equivalent promoted excessive hydrolysis of the starting material 29 (entry 6). Reaction performed with highly oxidizing Acr⁺‐MesBF₄ ⁻ instead of 4DPAIPN still provided the N‐phosphoryl azetidine 34 albeit with lower yield of 45% (entry 7), while supporting the activity of a reductive quenching process. Lastly, no reaction was observed in the absence of the PC, light or water (entry 8), indicating the key presence of all these three ingredients. It is worth noting that with this approach, it is possible to directly access N‐phosphoryl azetidine 34 avoiding two conventional approaches to synthesize phosphine oxide from trivalent phosphine reagent [8, 9, 10, 11, 16, 17] such as (i) initial oxidation of phosphine‐to‐phosphine sulfides (33) by treatment with sulfur (ii) and subsequent oxidative exchange of the S atom to O atom via treatment with oxidizing agents like m‐CPBA or H₂O₂.

FIGURE 2.

Iminyl phosphines as P‐ centered radical precursors. (A) Preliminary studies on possible P‐radical precursors. (B) Optimized reaction condition. ^[a]Reaction performed at 0.1 mmol scale, 29 (4.0 eq.), 32 (1.0 eq.), under 456 nm LED irradiation, see Supporting Information, section 3.2. ^[b]1H NMR yield using 1,3,5‐trimethoxybenzene as internal standard. ^[c]Isolated yield.

2.2. Mechanistic Investigations

To elucidate the reaction mechanism and the roles of all the participating species, we conducted a series of mechanistic studies (Figure 3 and Supporting Information, section 4.0). We assessed the feasibility of the initial SET step between the excited 4DPAIPN* and the reaction partners such as iminyl phosphine 29 and azabicyclo[1.1.0]butane 32, via Stern–Volmer quenching studies. Here, 29 is the sole quencher of the PC, as 32 showed no signs of quenching 4DPAIPN in DCE (Figure 3A). Consequently, changing the solvent to ethyl acetate and toluene resulted in different quenching constants and can therefore serve as evidence for an underlying SET mechanism over EnT process [42]. Although 4DPAIPN is known for its strong reducing capabilities, its exited state oxidation potential 4DPAIPN* (E _ox = +1.1 V, Figure 4, electrochemical data) [44] as well as the formation of N‐phosphoryl azetidine 34 using photocatalysts such as highly oxidizing Acr⁺‐MesBF₄ ⁻ further support that the reaction proceeds through the pathway involving SET oxidation of 29. Recorded UV/visible absorption spectra of 29, 30, and 31 and their mixtures indicates that the only absorbing species is 29 in the range above 427 nm (see Supporting Information, section 4.3). However, direct excitation studies without the PC showed no signs of product (see Supporting Information, section 4.4), in agreement with the activity of a SET manifold versus EnT. Proved the feasibility of the first oxidative step, we next investigated the presence of the key radical 24. A trapping experiment with TEMPO was conducted which completely suppressed the formation of N‐phosphoryl azetidine 34 (see Supporting Information, section 4.5), although, no TEMPO‐trapping adduct were detected by UPC² analysis. To gain further mechanistic insight, we conducted electron paramagnetic resonance (EPR) studies utilizing the spin‐trapping agent α‐phenyl‐tert‐butylnitrone (PBN). For this, a solution of spin trap (PBN, N‐tert‐Butyl‐α‐phenylnitrone) was incubated with the reaction mixture (the photocatalyst 4DPAIPN, 29, 32, and 38). The solution was thoroughly degassed and analyzed before and after 20 h of irradiation using blue LED light (λ = 456 nm). The solution showed a modest presence of trapped radicals immediately after preparation, the amount of which increased more than ten times with irradiation. The EPR spectrum obtained after irradiation (Figure 3B—shown with the global simulation in red) shows the presence of two radicals. The dominant species (75%) can be attributed to the trapping of the phosphoryl radical 24 based on the simulation (Figure 3B– green simulation) performed with EasySpin: [45] the hyperfine couplings and g _iso (a _14N = 14.5 G; a _1H = 3.6 G; a _31P = 16.5 G; g _iso = 2.0059) are comparable to what is reported in the literature (a _14N = 14.0–14.5 G; a _1H = 2.9‐4.3 G; a _31P = 14.9‐19.0 G) [46, 47, 48, 49]. The phosphoryl radical 24 is trapped also when either 32 or 38 are not present in solution (see Supporting Information, section 4.6). However, in the second case it is much less abundant, suggesting a key kinetic role of 38 in favoring the generation of the phosphoryl radical 24. Following confirmation of the formation of the key phosphoryl radical intermediate 24 in the reaction medium, we sought to establish that this species originates specifically from iminyl phosphine 29 upon SET oxidation to 29˙⁺. To exclude alternative sources, a series of control experiments was performed in which iminyl phosphine 29 was replaced with phosphine‐derived side products isolated from the reaction media (36–38). Among these, 36 is known to undergo SET oxidation under photoredox conditions to generate the corresponding phosphoryl radical 24 [6]. Accordingly, we evaluated whether 36 could serve as a viable precursor of 24˙ using 4DPAIPN or thioxanthone (TXO). However, no formation of N‐phosphoryl azetidine 34 was observed (Figure 3C). Similarly, replacement of iminyl phosphine 29 with either 37 or 38 failed to afford product 34 (see Supporting Information, section 4.7). To discern whether radical 20 or 24 is responsible for the strain release of compound 32, sulfur was subsequently added to the reaction after irradiation under argon atmosphere, which was then stirred for 1 h at ambient temperature. If the phosphinyl radical 20 is responsible for the ABB opening, the formation of compound 33 via sulfur oxidation should be observed. The absence of 33 along with generation of 5% of 30 (formed from the oxidation of unreacted 29) suggests that the phosphoryl radical 24 is the only species responsible for strain release of 32 (Figure 3D). This finding is further corroborated by density functional theory (DFT) calculations performed at the UωB97X‐D/Def2‐SVP level of theory in dichloromethane (SMD solvation model) [10, 11]. The addition of a phosphinyl radical (Ph₂P·) 20 to azabicyclo[1.1.0]butane 32 exhibits a relatively high activation barrier (ΔE ^‡ = 14.1 kcal·mol⁻¹; ΔG ^‡ = 13.6 kcal·mol⁻¹). In contrast, the phosphoryl radical 24 displays markedly enhanced reactivity toward azabicyclobutane 32 (Figure 3E), with a substantially lower activation barrier (ΔE ^‡ = 8.9 kcal·mol⁻¹; ΔG ^‡ = 10.4 kcal·mol⁻¹) and a pronouncedly exergonic reaction profile (ΔE = –22.0 kcal·mol⁻¹; ΔG = –21.8 kcal·mol⁻¹ vs. ΔE = –43.2 kcal·mol⁻¹; ΔG = –42.1 kcal·mol⁻¹). To shed light on the second part of the catalytic manifold, we performed a reaction using 1.0 eq. of D₂O, that resulted in 40% deuterium incorporation at C3 of the azetidine core. This finding suggests that upon radical opening of the ABB the transiently generated tertiary radical is reduced into the corresponding carbanion that reacts with H₂O, as a proton source (Figure 3F). No reaction was observed without water (see Figure 2b, Supporting Information section 4.7.5). With the key steps of the catalytic cycle established, we finally turned to the transformation of the phosphine radical cation 29˙⁺ into the phosphoryl radical 24. The superior performance of diphenylphosphinic acid 38 in enhancing the reaction yield, relative to other acid additives, led us to consider its potential nucleophilic attack on 29˙⁺ , followed by a β‐scission pathway to generate the radical 24 (Figure 3G, also see Supporting Information section 4.11). This process would be analogous to acyl radical generation from carboxylate anions and phosphine radical cations [50]. To probe this possibility, the standard iminyl phosphine 29 was replaced with the phosphite derivative 42. The machine learning model predicts a triplet energy of 48.9 kcal·mol⁻¹ for 42 with a DFT‐calculated bond dissociation energy (BDE) of 57.6 kcal·mol⁻¹, consistent with the trend observed for 29. Under these conditions, the N‐phosphoryl azetidine 34 was not detected in the reaction mixture; instead, azetidine 45 was obtained in 50% ¹H NMR yield.

FIGURE 3.

Mechanistic investigations.

FIGURE 4.

Proposed reaction mechanism. Density function theory (DFT) calculations performed at UωB97X‐D/Def2‐SVP level of theory including SMD solvation model. Distribution of the unpaired electron in Int‐1 and Int‐3 is shown in dark blue.

2.3. Computational Studies

To support our experimental findings while shedding light on the mechanism and the nature of key catalytic intermediates, DFT calculations were performed at the UωB97X‐D/Def2‐SVP level of theory [51, 52] in implicit dichloromethane (SMD solvation model) [53] using Gaussian16 Rev C.02 [54]. On occasion, we ran preliminary calculations using ORCA (v. 5.0.4) [55] (see Supporting Information section 5.0) at the UωB97X‐D3/Def2‐SVP level of theory, in the gas phase [52, 56]. Consistent with the experimental observations (Figure 3G), the pathway involving nucleophilic attack of 38 on 29˙⁺ , followed by direct β‐scission of the resulting iminophosphoranyl intermediate to generate radical 24, is confirmed not to be accessible at room temperature, since the predicted free energy barrier (∆G ^‡) is over 22.5 kcal·mol⁻¹ (see Supporting Information section 5.0 and calculations online) [57, 58, 59]. In our search for alternative reliable transition states involving the key reactants H₂O and/or 38, we reasoned that a pre‐association complex between 29˙⁺ and 38 should still be the mechanistically relevant starting point for subsequent investigation (Figure 4). Formation of this complex (29˙⁺–38) is slightly exergonic, with ∆G estimated at –2.6 kcal·mol⁻¹. This complexation leads to the formation of iminophosphoranyl intermediate (Int‐1) via proton transfer from 38 through TS1. This step proceeds with a calculated free‐energy barrier of ΔG ^‡ = +8.8 kcal·mol^{− 1} and overall endergonicity of ΔG = +8.7 kcal·mol^{− 1}. This alternative mechanistic route proceeding from Int‐1 bears a lower overall energetic cost compared to the β‐scission pathway (∆∆G ^‡ = at least –2.7 kcal·mol⁻¹). Notably, Int‐1 features the unpaired electron mainly confined to the P atom, pointing directly away from the P–N bond and with marginal delocalization to the adjacent O and C atoms. This electronic arrangement enables heterolytic cleavage of the P–N bond, resulting in the formation of 19 and the phosphoranyl radical cation (Int‐2) with a ∆G^‡ of +11.3 kcal·mol⁻¹ and again an overall ∆G of +5.5 kcal·mol⁻¹ (+14.0 kcal·mol⁻¹ with respect to the initial 29˙⁺–38 complex). Accessing the transition state TS2 on the way from Int‐1 to the Int‐2–19 complex only requires a modest elongation of the P–N bond (from 1.78 to 1.96 Å). Although the detailed mechanism for nucleophilic attack on Int‐2 was not fully explored, Merz–Kollman–Singh charge fitting [60] on the optimized uncomplexed Int‐2 (Figure 4) indicates that the non‐radical phosphorus center carries a higher positive charge (+0.65e) than the radical phosphorus (+0.42e), identifying it as the more favorable site for nucleophilic attack by H₂O. Consistently, free‐energy comparisons between the optimized Int‐2–H₂O complex and the resulting 24–38 complex show that the process is strongly exergonic (ΔG = –16.1 kcal·mol^{− 1}). This charge distribution further supports the fact that phosphoryl radical 24 originates from the iminyl phosphine 29 frameworks rather than from 38 (Refer to Figure 3G). The resulting phosphoryl radical 24 then undergo radical strain release with azabicyclo[1.1.0]butane 32 via TS3 to generate a tertiary‐alkyl radical intermediate Int‐3. This process is characterized by a ∆G ^‡ of +10.4 kcal·mol⁻¹ and a very high exergonic reaction profile (∆G = –42.1 kcal·mol⁻¹). The resulting intermediate Int‐3 then undergoes a facile single‐electron transfer (SET) reduction by the PC radical anion to generate the anionic intermediate Int‐4, which subsequently yield the final N‐phosphoryl azetidine 34.

2.4. Generality of the Method

We next assessed the generality of the investigated P–N bond forming reaction between 17 and ABBs 26 (Figure 5). ABBs featuring aryl substituents with EDGs were well tolerated, delivering azetidines 46–49 in yields of up to 85%. In contrast, para‐EWGs (CF₃, Cl, and F) afforded the corresponding products 50–52 in slightly diminished yields, up to 67%. Replacement of the phenyl group in ABB 32 with a naphthyl moiety, as well as a range of other heteroaromatic substituents including pyridyl (ortho and meta), furan, thiophene, and triazole proceeded smoothly, furnishing N‐phosphoryl azetidines 53–58 in yields ranging from 52% to 70%. The method was next applied to more complex, biologically relevant scaffolds. Remarkably, azabicyclo[1.1.0]butanes derived from a 1,3‐naphthoxazine derivative, as well as from pharmaceutically relevant compounds such as febuxostat and tafamidis were efficiently converted into the respective azetidines 59–61 in yields of up to 72%. Introduction of trifluoromethyl groups at the 3,5‐positions of the phosphine aryl ring led to a drastic reduction in yield, affording azetidine 62 in only 11%. In contrast, (bis(3,5‐dimethyl) on the phosphine 63 are tolerated in the reaction, albeit with moderate efficiency (52% yield).

FIGURE 5.

Generality of the developed method. Reactions were performed at 0.1 mmol scale. See Supporting Information, section 3.3. Isolated yields are given in brackets. ^[b]Reaction time 48 h. ^[c]Without acid additive. ^[*^]Iminyl phosphine generated in situ. CSD, Cambridge Structural Database (CSD 2527691, CSD 2527692, CSD 2527693).

We next explored the feasibility of replacing the diaryl phosphoryl radical with a dialkoxy phosphoryl radical (Figure 6). DFT calculations predicted a comparable activation barrier for radical addition to model ABB 32 towards 64 relative to 24 (See Figure 6 pink box, top right). Pleasingly, reaction of 42 with 32 and an o‐pyridyl‐substituted ABB afforded azetidines 45 and 65 in 50% yield. The versatility of the diethoxy phosphoryl radical 64 was further demonstrated through late‐stage functionalization of a febuxostat‐derived ABB, delivering azetidine 66 in 58% yield. Notably, we successfully introduced a cyclic phosphite moiety into the azetidine core, resulting in the naphthoxazine‐derived azetidine 67 with a yield of 44%. It is worth noting that, compounds 62, 63, and 67 were synthesized through a three‐component reaction via in‐situ generation of the corresponding iminyl phosphines (See Supporting Information section 3.4). Next, we evaluated the electronic effect on the aromatic phenyl rings on imine for the photocatalytic process (Figure 6, imine electronic effects). The p‐OMe‐substituted iminyl phosphine 68 afforded the N‐phosphoryl azetidine 34 in 63% yield, whereas the p‐Cl‐substituted iminyl phosphine 69 provided a lower yield of 38%. We finally performed some synthetic manipulations that enabled further selective modification of the carbonyl and phosphine functionalities (Figure 6, downstream transformations). Azetidine 34 was reduced to the corresponding secondary alcohol 70 using NaBH₄ in 76% yield. In addition, compounds 34 and 53 underwent selective reduction with air‐stable 1,3‐diphenyl‐disiloxane (DPDS) generating azetidine 71. Subsequent oxidation of the P(III) azetidine afforded the N‐thiophosphoryl azetidines 72 and 73 in up to 81% yield.

FIGURE 6.

Utility of the developed method and product manipulation. Reactions were performed at 0.1 mmol scale (See Supporting Information, section 3.3 and 3.6). Isolated yields are given in brackets. ^[*^]Iminyl phosphine generated in situ.

3. Conclusion

In summary, we have shown that iminyl phosphines can be efficiently engaged under photocatalytic conditions to access rare P–N bonds within azetidine scaffolds. The reaction proceeds through an unconventional mechanistic pathway in which initial single‐electron transfer (SET) oxidation of the iminyl phosphine triggers a cascade of events, assisted by hydrogen‐bond donors, that culminates in the generation of a high‐energy phosphoryl radical. This key intermediate was directly detected by EPR spectroscopy and identified as the species responsible for azabicyclo[1.1.0]butane (ABB) ring opening. Subsequent P–N bond formation, reduction of the resulting tertiary radical by the PC radical anion, and protonation deliver the N‐phosphoryl azetidine products. A series of control experiments and DFT calculations supported the proposed mechanism. The potential and generality of the method was demonstrated across a broad range of structurally diverse and biologically relevant substrates, providing N‐phosphoryl azetidines in high yields (up to 85%). Moreover, these products can be readily converted into N‐thiophosphoryl azetidines, enabling downstream transformation. Overall, this work establishes iminyl phosphines as a new class of bench stable trivalent P‐centered radical precursors and offers a mechanistic platform for the photochemical installation of P‐atom into diverse molecular frameworks.

Author Contributions

Chandu G. Krishnan conceived the project and devised the experiments with Luca Dell'Amico. Chandu G. Krishnan, and Gabriel Cormier carried out the reactions and isolated and characterized the products. Luca Dell'Amico, Chandu G. Krishnan, and Gabriel Cormier rationalized the experimental results. Marco Bortolus performed the spectroscopic investigations using EPR. Stefano A. Serapian performed the DFT studies. Luca Dell'Amico and Chandu G. Krishnan wrote the manuscript with contributions from all the authors. Luca Dell'Amico directed the work.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: anie72854‐sup‐0001‐SuppMat.pdf.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.