Intermolecular 2+2 imine-olefin photocycloadditions enabled by Cu(I)-alkene MLCT

2 + 2 Photocycloadditions are idealized, convergent construction approaches of 4-membered heterocyclic rings, including azetidines. However, methods of direct excitation are limited by the unfavorable photophysical properties of imines and electronically unbiased alkenes. Here, we report copper-catalyzed photocycloadditions of non-conjugated imines and alkenes to produce a variety of substituted azetidines. Design principles allow this base metal-catalyzed method to achieve 2 + 2 imine-olefin photocycloaddition via selective alkene activation through a coordination-MLCT pathway supported by combined experimental and computational mechanistic studies.

Transition metal compounds capable of photoexcitation via metal to ligand charge transfer (MLCT) to achieve intermolecular electron or energy transfer have gained significant recognition for their use in chemical synthesis [19][20][21][22] . While the majority of these reports use photocatalysts that access the needed excited states in the absence of any substrate, the formation of the active photocatalytic species in situ through selective substrate engagement is a comparatively underdeveloped approach. Seminal contributions from Salomon, Kochi, and co-workers identified that cyclobutane formation from alkene photodimerization was significantly improved by the addition of Cu(I)-trifluoromethanesulfonate (CuOTf) due to the in situ formation of a Cu(I)-olefin compound serving as the active chromophore 23 . Meggers and co-workers reported photochemical cyclobutane synthesis by in situ formation of the active chromophore as a chiral at metal, O-bound Rh-enone compound 24 . We recently reported a Cu-catalyzed 2 + 2 carbonylolefin photocycloaddition (COPC) approach for oxetane formation by inverting conventional Paternò-Büchi reactivity and achieved alkene activation via Cu(I)-coordination followed by MLCT 25 .
In this work we show that azetidines may be accessed by substrate coordination and selective activation via MLCT with the alkene π-component in a 2 + 2 imine-olefin photocycloaddition (Fig. 1e).

Results
Reaction optimization. We irradiated a 1.1:1 mixture of N-butyl-2-methylpropan-1-imine (1) and norbornene with catalytic hydrotris(pyrazolyl)borate copper(I) (TpCu) using a 100 W Hg lamp for 24 h resulting in formation of the corresponding azetidine 1A in 55% yield with exclusive exo selectivity ( Table 1, entry 1). Removing either TpCu or light resulted in no formation of 1A (entries 2 and 3). Using a threefold excess of 1, produced 1A in 79% yield (entry 4). Application of a 300 nm cut-on filter resulted in a significant decrease in the yield of 1A to 39% even after a prolonged 36 h reaction time suggesting that product forming irradiation occurred at wavelengths shorter than 300 nm (entry 5). Using CuCl or CuOTf in place of TpCu did not result in any detectable quantities of 1A (entries 6 and 7), underscoring the importance of the coordination environment of the Cu(I) ion source.
IOPC was particularly notable in the presence of a γ-lactam (11), showcasing the chemoselectivity of olefin activation even in the presence of other potentially competitive chromophores. Imines derived from ketones also reacted efficiently to produce the corresponding azetidines even when the imine also contained Lewisbasic 3°amine (13), acetal (14), or ether (15) functionalities.
Biologically active azetidines such as penaresidin A, nicotianamine, medicanine, and mugineic acid bear N-alkyl or no substitution, but access to 2°azetidines can be challenging using existing cyclization methods that require N-substituents that are stubborn or not amenable to cleavage. This Cu-catalyzed 2 + 2 IOPC is uniquely capable of engaging N-alkyl imines directly as  well as N-substituents that are easily removed. N-Allyl azetidine 13 underwent deallylation to reveal the corresponding 2°a zetidine in 82% yield using 2 mol % Pd(PPh 3 ) 4 and 2-mercaptobenzoic acid. Alternatively, O-methyl oxime-derived azetidine 19 underwent facile reductive N-O bond cleavage using activated Zn dust in aqueous acetic acid to provide the corresponding unsubstituted azetidine in 93% yield. Diversification of the olefin component allowed norbornene derivatives containing acetyl, vinyl, cyano, and tert-butyltrimethyl silyl ether groups to be successfully coupled with N-allyl cyclopentanimine (Fig. 2, 20-24). The identity of azetidine 23 was unambiguously confirmed through determination of the solid-state structure of its hydroiodic salt using X-ray diffraction (23-HI salt).
While maleic anhydride is known to undergo selective cyclobutane forming 2 + 2 photocycloaddition with ethylene using acetophenone as a photosensitizer 28 , direct excitation of maleic anhydride with N-allyl cyclopentanimine resulted in the formation of azetidine 25 in 51% yield along with a complex mixture of unidentified byproducts. Subjecting the same reactant pair to irradiation in the presence of 20 mol % TpCu resulted in the exclusive formation of 25 in 98% yield. Similarly, other imines were converted to the corresponding maleic anhydride derived azetidines (26 and 27) and use of TpCu uniformly increased both reaction efficiency and selectivity compared to uncatalyzed reactivity. N-Methyl maleimide was also successfully converted to the corresponding azetidine 28 with N-allyl cyclopentanimine in 49% yield.
Mechanism investigations. We initially hypothesized that 2 + 2 IOPC could occur through activation of either the imine or alkene π-bond. Through a combination of 1 H NMR and electronic absorption spectroscopies we observed the requirement of alkene coordination to TpCu for successful azetidine formation. Three olefin adducts of TpCu were successfully prepared and isolated through equimolar mixing experiments (Fig. 3a). TpCu-NB was isolated as a white crystalline solid and the solid-state structure determined by single crystal X-ray diffraction (Fig. 3b). The coordination geometry is best described as pseudo-tetrahedral about the Cu-ion with the carbon-carbon bond distance of the coordinated olefin (1.387 Å) in the range of other Cu(I)-norbornene compounds (1.36-1.39 Å) indicative of a predominating σ-bonding, rather than π-back-bonding, metal-olefin coordination interaction [29][30][31][32] . Examination of the Cu-N(pyrazolyl) bond distances show that while two were nearly equivalent (Cu-N(2) = 1.999(2), Cu-N(4) = 2.010(2)), the third was significantly longer (Cu-N(6) = 2.274(2)), breaking the C 3 symmetry axis of the hydrotris(pyrazolyl)borate chelate.
The solid-state structure of TpCu-MA shows an analogous coordination geometry to TpCu-NB and while the carbon-carbon bond length of the coordinated maleic anhydride (1.390 Å) is elongated compared to unligated maleic anhydride (1.3032 Å) 33 , it is significantly shortened compared to the only other structurally characterized Cu(I)-maleic anhydride compound supported by a strongly donating iminophosphanamide chelate (1.49 Å) 34 . This suggests that despite the electronic disparities between maleic anhydride and norbornene, the bonding in TpCu-MA is analogous to that observed in TpCu-NB. While ligand exchange was observed by 1 H NMR spectroscopy from TpCu-NB in the presence of the imines used in our study, isolation of a TpCu-imine coordination compound was unsuccessful likely due to the observed rapid, reversible ligand exchange. However, TpCu-MA did not undergo ligand exchange when subjected to imines of this study. We observed that irradiating equimolar amounts of TpCu-MA, N-allyl cyclopentanimine, and norbornene for 24 h resulted in the exclusive formation of azetidine 25 in 95% yield. This suggests that olefin coordination is necessary for 2 + 2 IOPC to proceed and that the coordinated alkene is selectively incorporated into the cycloadduct produced. We noted throughout our investigations that when additional strained, cyclic alkenes were assayed, azetidine formation was not observed when TpCu-alkene coordination was not detected by 1 H NMR.
The transmission spectrum of the borosilicate glass reaction vessels used in this study showed that light below 280 nm is effectively filtered from reaching the reaction mixture (Fig. 3c, gray area) and the electronic absorption spectra of ethereal solutions of most individual 2 + 2 IOPC reaction components showed no distinct absorption features 250-400 nm (Fig. 3c). While neither TpCu (green line), 1 (dark blue line), nor norbornene (not pictured) showed significant absorptions above 250 nm, maleic anhydride did (light blue line), explaining the small amount of non-selective background photocycloaddition observed with this alkene. Conversely, a stoichiometric solution mixture of TpCu and maleic anhydride (pink line) resulted in an electronic absorption spectrum with a distinctive broad feature with a maximum at 263 nm, which is similar to the spectrum observed of TpCu-NB which shows a maximum at 272 nm 25 , and is nearly identical to the spectrum of a stoichiometric mixture of TpCu, maleic anhydride, and imine 1 (orange line). In combination with the diminished azetidine formation when a 300 nm cuton filter was applied (Table 1, entry 6), these data support azetidine formation occurring via excitation between 280-300 nm.
A calculated electronic absorption spectrum of TpCu-NB best approximated the experimentally observed spectrum using timedependent density functional theory (DFT) calculations of at the M06 level of theory with the Tamm-Dancoff approximation. This enabled the construction of qualitative molecular orbital diagrams for TpCu-NB at the observed absorption feature 250-300 nm linked to azetidine formation (see the Supporting Information for additional details). The highest occupied molecular orbitals (HOMO-1) and the lowest unoccupied molecular orbital (LUMO) of TpCu-NB are primarily TpCu-based and olefin(π*)-based, respectively (Fig. 4). This is supported by previous DFT calculations which concluded an MLCT of idealized cationic Cu(I)-ethylene compounds was responsible for cyclobutane forming 2 + 2 photocycloadditions 35 and recent optical and X-ray transient absorption spectroscopic insights of Cu(I)mediated photodimerization of norbornene 36 . The selective imine C = N capture of this excited state affords the cycloadduct azetidine. Chemoselectivity of azetidine formation over alkene dimerization pathways is attributed to the rapid rates of C-centered radical addition to C = N double bonds 37 and preference over alkene addition 26 .
Analogous to our observations in 2 + 2 COPC, luminescence data collected from excitation between 280-300 nm did not indicate linear relationships between TpCu and the concentration of added norbornene or imine 1. However, a quenching relationship was observed with maleic anhydride, indicating its potential as a sensitizing agent in this process. While photosensitization cannot be ruled out for all substrates, the impact of olefin coordination ability in azetidine formation and chemoselectivity suggests that a shared sensitization process in all examples is unlikely to be operative. Stereochemical probes often used to infer 2 + 2 photocycloaddition spin states are not applicable in this 2 + 2 IOPC system as identical stereoisomers would be produced irrespective of the proceeding spin state due to the geometric constraints of the cyclic alkenes used.

Discussion
In conclusion, we report herein an intermolecular 2 + 2 photocycloaddition of aliphatic imines and alkenes to form azetidines. The use of hydrotris(pyrazolyl)borate copper(I) as a pre-catalyst allows for coordination of the olefin component and selective absorption of light 280-300 nm. A combination of solution 1 H NMR and electronic absorption spectroscopies along with DFT calculations from solid-state structural data supports a Cu(d) to olefin(π*) MLCT pathway to azetidine formation. This work highlights a catalyst controlled 2 + 2 imine-olefin photocycloaddition strategy that selectively activates π-components at red-shifted wavelengths compared to uncoordinated alkenes to access novel azetidine containing compounds.

Methods
Representative procedure. In a nitrogen filled glovebox a borosilicate culture tube was charged with imine (1.1 equiv), TpCu (0.2 equiv), alkene (1.0 equiv), and diethyl ether (0.15 M), capped then sealed with electrical tape and irradiated with a UVP Blak-Ray B-100A UV lamp in a fume hood. After the indicated time, the reaction mixture was quenched by exposure to air for 15 min, then concentrated under reduced pressure. The residue was taken up in methanol and purified by passage through a portion of basic alumina before solvent removal under reduced pressure to afford pure azetidine compounds.

Data availability
Experimental data as well as 1 H and 13 C NMR spectra for all new compounds prepared in the course of these studies are provided in the Supplementary Information Fig. 4 Mechanistic proposal of the title reaction. The proposed 2 + 2 IOPC mechanism commences with olefin substrate coordination to TpCu. Ligand exchange experiments indicate that imine substrates undergo rapid and reversible coordination to TpCu. The active TpCu-olefin chromophore then undergoes a MLCT from computationally visualized TpCu-NB HOMO-1 which is primarily Cu-based to TpCu-NB LUMO which is olefin(π*)-based. Capture by the imine C = N double bond leads to azetidine formation and release of TpCu.