Main

The synthesis of cyclopropanes1,2,3,4 via metal-catalysed carbene transfer from diazo compounds to olefins5,6,7,8 is among the best developed and most general cyclopropanations available. Enantioselective versions are documented, both with chiral metal complexes as catalysts and with engineered metalloenzymes. High effectiveness and stereocontrol have been accomplished with catalysts based on copper9, rhodium10 and ruthenium11. In particular, excellent levels of enantioselectivity have been achieved with copper(I) bisoxazoline complexes12 and with dinuclear rhodium(II) complexes, featuring chiral bidentate carboxylate, amidate or phosphate ligands13,14,15. A unique nickel carbene mediated cyclopropanation has also been reported16.

Biocatalytic, new-to-nature versions of this reaction were reported by Arnold et al. in 201317, using an engineered cytochrome P450 enzyme that catalyses highly diastereo- and enantioselective cyclopropanations of styrenes with diazoesters. The Hartwig group developed iridium-modified myoglobins that catalyse the cyclopropanation of unactivated olefins18, and the Keasling group19 described a variant in which both the diazoester and the styrene were produced intracellularly. While the biocatalytic and the transition metal catalysed approaches are conceptually different, they share the initial formation of electrophilic metal carbenoids, which then engage in the actual olefin cyclopropanation (Fig. 1a). Despite these advancements, regioselectivity for substrates with multiple olefins remains a particularly challenging problem20,21,22,23. We became intrigued by the idea of designing a complementary organocatalytic approach that would not necessitate the involvement of metal carbenoids and could solve the regioselectivity problem. Organocatalytic cyclopropanations of olefins activated by an electron-withdrawing group have previously been reported24,25,26 but those of electronically unbiased olefins are unknown. Since organocatalytic activation modes for diazo alkanes or olefins, other than protonation by Brønsted acids, which appeared not to be suitable for the desired process, are not available, such a design immediately led to the challenge of how to organocatalytically activate simple olefins and/or diazo esters. We hypothesized that the abstraction of a single electron—arguably the most general strategy for activating any molecule—could be suitable here. Accordingly, we envisioned a catalytic cycle in which the olefin is activated as a radical cation upon single electron transfer to a photoexcited cationic organocatalyst associated with an enantiopure counteranion. The resulting highly reactive open-shell radical cation forms an ion pair with the enantiopure counteranion and is expected to undergo enantioselective carbon–carbon bond formation with the diazoalkane, accompanied by the loss of N2 to furnish a cyclopropane radical cation27,28,29. A final single electron transfer from the catalyst radical to this intermediate should then regenerate the cationic photocatalyst and deliver the desired cyclopropane product (Fig. 1b). This design is supported by a recently established radical cation-based photoredox catalytic cyclopropanation27 and by our own work on utilizing enantiopure counteranions in photoredox catalytic cycloadditions via olefin-derived radical cations30. The envisioned challenges associated with our concept are controlling the facial selectivity of a highly reactive open-shell radical cation intermediate and preventing the direct activation of diazoalkane to a free carbene via photolysis or energy transfer from the excited state photocatalyst, which could lead to racemic products31,32,33.

Fig. 1: Asymmetric cyclopropanation between olefin and diazo compound.
figure 1

a, State-of-the-art approaches to cyclopropanation of olefin with diazo compounds using metal catalysts, enzyme catalysts and the catalytic cycle. b, Our design, featuring catalytic asymmetric insertion of a diazo compound to an in situ-generated radical cation intermediate. L, ligand; M, metal.

Full size image

Herein, we report an organocatalytic regio- and stereoselective cyclopropanation of olefins, facilitated by asymmetric counteranion-directed photoredox catalysis (ACPC).

Results

Reaction development

To initiate our investigation, we studied the cyclopropanation reaction of trans-anethole (1a) and diazoester 2a in a (1:1 v/v) solvent mixture of dichloromethane (CH2Cl2) and pentane at −100 °C under irradiation by 6 W of green light. We began by exchanging the counteranion of the synthesized thioxanthylium triflate34 organophotoredox catalyst with different confined imidodiphosphorimidate (IDPi)35 anions (Fig. 2). Encouraged by our previous studies on IDPi-catalyzed transformations, we initially selected from our assortment of privileged IDPi anions. An IDPi catalyst possessing a para-tert-butyl group as 3,3′ substituent and a trifyl core (IDPi-A) delivered the corresponding cyclopropanated product with moderate yield, good enantioselectivity (80:20 er) and poor diastereoselectivity. An analogous meta-biphenyl IDPi (IDPi-B) gave improved yield (90%) and enantioselectivity (90:10 er). Using catalyst 4C where the 3,3′ substituents of the IDPi catalyst feature a combination of meta- and para-substitution, led to a further enhancement of the enantioselectivity (95:5 er) with a dr of roughly 3:1. Having identified an optimal catalyst towards enantioselectivity, we were keen to further improve the diastereoselectivity. In this regard, meta-,para-substituted catalysts were extensively evaluated (Supplementary Table 1), and among them, spirocyclic fluorenyl substitution on the 3,3′-positions of the BINOL was found to be suitable towards this aim (4D–F). To our delight, the modified spirocyclopentyl fluorenyl substitution on the 3,3′-positions of the BINOL (IDPi-E) delivered the product not only with excellent yield and enantioselectivity but also with excellent diastereoselectivity of >20:1. Finally, fine-tuning of the inner core sulfonamide from trifluoromethyl (IDPi-E) to n-perfluoroethyl (IDPi-F) enhanced the enantiomeric ratio from 97:3 to 99:1, maintaining the excellent yield and diastereoselectivity. Remarkably, even at −100 °C, catalyst 4F rapidly furnished product 3a in a short reaction time. Lowering the catalyst loading from 2.5 to 0.5 mol% still gave full conversion with high yield even though a slight reduction of enantioselectivity (96:4 er) was observed.

Fig. 2: Reaction development.
figure 2

Optimization of the catalytic reaction for the evaluation of IDPi anions was carried out, and the yield was determined by 1H NMR using an internal standard.

Full size image

Substrate scope

Next, a broad range of olefins with different double bond ___location and substituents at different ring positions were evaluated (Fig. 3) in the reaction with diazoester derivatives, and the cyclopropane products were obtained with good to excellent yields and enantioselectivities. Moreover, most of these products were obtained with excellent diastereoselectivities. 3-Ethyl anethole was found to be a suitable reaction partner and furnished product 3b with excellent enantioselectivity. Interestingly, 3,3-dimethyl anethole also delivered the desired product (3c) with excellent enantioselectivity, albeit with low diastereoselectivity. An olefin bearing an tert-butyldimethylsilyloxy (OTBS) substituent was also tolerated, providing product 3d with an er of 97:3 in 86% yield. Similarly, olefins featuring electron-withdrawing substituents and a heterocyclic ring furnished products 3e and 3f in excellent enantioselectivities. It is worth mentioning that other electron-rich heterocyclic olefins successfully engaged in our cyclopropanation and furnished products 3g and 3h with good to excellent level of enantioselectivity. Next, we focused on using electron neutral styrene derivatives as potential olefins for cyclopropanation. As their redox potential (>1.8 V) is higher than that of the used thioxanthylium photocatalyst (1.76 V), a photocatalyst featuring a higher redox potential was required. Gratifyingly, acridinium photocatalyst 5 (2.04 V), under red light irradiation, delivered products 3i and 3j with excellent enantioselectivity. Internal olefins gave products 3k and 3l in good yields and excellent enantioselectivity. We explored different diazoesters and found that all of the different diazoesters gave the desired products (3m and 3n) with an excellent level of enantioselectivity. Additionally, a disubstituted diazoester was found to be a suitable partner for our cyclopropanation reaction and yielded the desired product 3o in good enantioselectivity and moderate diastereoselectivity. We also investigated a diazoalkanes featuring an N-H amide-, aromatic or aliphatic keto functionality, and all furnished the desired products (3p–r) with trans-anethole in moderate to good level of enantioselectivity. Finally, even trimethylsilyl-diazomethane reacted to product 3s with promising results.

Having successfully developed an organocatalytic cyclopropanation of olefins, we wondered whether our confined catalyst design may enable selectivity also with substrates bearing multiple olefins. For example, 2,5-dimethylhexa-2,4-diene (1t) is a symmetrical diene that has previously been used in the synthesis of chrysanthemic acid, a natural product also found in a variety of insecticides. In this case, the chemoselectivity issue towards obtaining the corresponding monocyclopropane has previously only been overcome by using a several fold excess of the diene23. Remarkably, using our organocatalyst system, chrysanthemic acid derivative 3t could be obtained selectively in good yield and good enantioselectivity for both diastereomers without requiring an excess of diene. Subsequently, a simple hydrolysis of ester 3t gave chrysanthemic acid (3u), preserving the enantioselectivity (Fig. 3b). We further explored the reactions of unsymmetrical conjugated dienes and trienes, delivering the desired products in good yield and enantioselectivities.

It is noteworthy that an unsymmetrical conjugated diene delivered product 3v with remarkable regioselectivity (>20:1 rr) and diastereoselectivity (>20:1 dr). A diene substrate bearing an additional pendant olefin functionality (1w) was also found to be suitable for our cyclopropanation. Even the reaction with conjugated and unsymmetrical triene 1x, posing a formidable regioselectivity challenge, selectively delivered product 3x (Fig. 3c). On the other hand, we also recognized the limitations of our reaction (Supplementary Fig. 2). Olefins bearing electron-withdrawing groups yielded the desired product in low amounts but with high enantioselectivity. Unfortunately, substrates containing amino-substituted aryl groups, indole-derived olefins, enol ethers and enamines remained unreactive under our reaction conditions. Additional diazo compounds were tested, furnishing the desired product in moderate to good yields with satisfactory enantioselectivity. However, we observed that acceptor–acceptor diazoalkanes did not undergo the transformation. Additionally, enantioenriched product 3a was derivatized to carboxylic acid product 6 without any loss of selectivities. Furthermore, a sequential reduction of the ethyl ester of 3a to the corresponding alcohol, followed by Steglich esterification gave ferrocene ester 7, the absolute configuration of which could be determined by single crystal X-ray diffraction.

Fig. 3: Substrate scope of the cyclopropanation methodology and derivatizations.
figure 3

a, Alkene and diazoalkane variations. b, Synthesis of chrysanthemic acid. c, Scope of the dienes and trienes. d, Derivatizations dr was determined by 1HNMR. aUsing catalyst 4E. bUsing catalyst 4C. cUsing 5 equiv. 2a for 18 h. dYield determined by 1H NMR. eUsing catalyst 5F under 6 W of red light. fUsing catalyst 4B. gUsing catalyst 4D. hUsing catalyst 5C under 6 W of blue light. TBS, tert-butyldimethylsilyl; LAH, lithium aluminium hydride; EDCI, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimid; DMAP, 4-dimethylaminopyridine.

Full size image

Mechanistic investigations

During our reaction development, lowering the catalyst loading from 2.5 to 0.5 mol% still gave full conversion with a high yield. However, a slight reduction of enantioselectivity was observed (Fig. 4a). Furthermore, the amount of diazoester was evaluated. Surprisingly, increasing the amount of the diazoester also led to a reduction of enantioselectivity (Fig. 4b). These two results hinted at a non-enantioselective, photochemical background pathway31,32,33. Since such a background reactivity may require a different wavelength, we were intrigued to explore light of different wavelengths. Indeed, a red shift was associated with a notable increase of enantioselectivity (Fig. 4c). To gain further insight into this phenomenon, we irradiated our model substrate 1a and diazoester 2a with blue or yellow light separately. Remarkably, irradiation with yellow light led to higher enantioselectivity, from 80% ee with blue light to 97% ee with yellow light. Similarly, when olefin substrates 1k and 1l were excited with yellow instead of blue light under similar conditions, an increase in enantioselectivity was observed. For example, for product 3l, the enantiomeric excess increased from 41.8% with blue light to 95.4% with yellow light (Fig. 4d).

Fig. 4: Reaction development for wavelength of light.
figure 4

a, Reaction development under lower catalyst loading conditions. b, Reaction development on the basis of the diazo compound 2a. c, Reaction development under different light sources. d, Reactions with different olefins under blue and yellow light.

Full size image

To shed light on this unusual observation, we conducted absorption spectroscopy of diazoester and photocatalyst, where diazoester 2a has a maximum wavelenth of absorption (λmax) at 397 nm (Fig. 5a) and photocatalyst 4F has a λmax at 472 nm (Fig. 5b). Since the absorption band of photocatalyst 4F is very broad, the opportunity arose to excite the photocatalyst at different wavelengths and study the corresponding emission spectra. Photocatalyst 4F was excited at 472 nm (λmax) and also at a lower wavelength of 350 nm. Surprisingly, in addition to the usual emission spectra maximum at 627 nm, another maximum emission band was detected at 395 nm upon excitation at 350 nm (Fig. 5c). Interestingly the absorption band of the diazoester significantly superimposes with the emission band of the photocatalyst at lower wavelength (Fig. 5d), which led us to the hypothesis that it is this lower-wavelength excitation mode of the diazoester that is involved in the photochemical background pathway.

Fig. 5: Spectroscopic analysis for mechanistic insights.
figure 5

a, The ultraviolet–visible (UV–vis) absorption spectra of 2a. b, The UV–vis absorption spectra of 4F. c, The emission spectra of 4F. d, Overlay of the UV–vis spectrum of 2a with the emission spectra of 4F. e, A Stern–Volmer plot at excitation of 472 nm. f, A Stern–Volmer plot at excitation of 350 nm. Q, quencher’s concentration.

Full size image

Steady-state fluorescence quenching experiments with the individual reactants were conducted to obtain Stern–Volmer plots and to identify the reagent that interacts with the excited state photocatalyst. Photocatalyst 4F was first excited at 470 nm, and quenching experiments were performed separately with different amount of anethole 1a and diazoester 2a. The corresponding Stern–Volmer plots (Fig. 5e) clearly indicate that anethole 1a is responsible for the fluorescence quenching, whereas diazoester 2a has virtually no effect at this concentration. A separate set of steady-state florescence quenching experiments was conducted where the photocatalyst 4F was excited at 350 nm. In contrast, the Stern–Volmer plot (Fig. 5f) shows significant fluorescence quenching with diazoester 2a. Further standard radical probe and radical clock experiments were conducted to rule out radical process and support the radical cation mechanism, which is in line with the mechanism proposed in literature (Supplementary Fig. 3). To gain further information regarding the stereo-specificity of our reaction, we conducted control experiment with Z-anethole (Supplementary Fig. 4). Interestingly, under our optimized conditions, all four possible diastereoisomers were obtained in good yields. Under the optimized condition, control experiments were performed where interconversion of anethole was not observed (Supplementary Fig. 4). These experiments further strengthen the stepwise mechanism, which is also in line with the literature. Additionally, the redox potentials of our reactants, catalyst and product were measured (Supplementary Fig. 5).

Conclusions

Key to the success of our organocatalytic cyclopropanation of olefins with diazoalkanes has been the use of ACPC, exploiting the positive charge of the olefin radical cation intermediate towards ion pairing with an enantiopure counteranion. Our methodology is useful for a broad range of substrates and complements previously reported metal and metalloenzyme catalyzed cyclopropanations. It also addresses the unsolved problem of regioselectivity in the reaction of substrates with different olefins. The observed wavelength-dependent enantioselectivity could serve as a valuable mechanistic probe for photoredox catalysis, potentially aiding in distinguishing between alternative reaction pathways. Observing enantioselectivity in ACPC at one wavelength confirms the involvement of radical cations in the enantiodifferentiating step, whereas its absence at this or another wavelength is consistent with alternative mechanistic pathways, such as energy transfer that proceed via neutral intermediates. Our results underscore the potential of ACPC for controlling the selectivity of radical cations in alkene manipulations, complementing previous reports.

Methods

General procedure for anion exchange

The achiral anion of the photocatalysts was exchanged with the desired enantiopure anion using the salt metathesis technique. In a gas chromatography vial, photocatalyst (25–30 mg, 1.0 equiv.) and a corresponding chiral Brønsted acid catalyst (1.05 equiv.) were measured. Then, dichloromethane (300 μl) was added, followed by saturated NaHCO3 solution (1.1 equiv., 400 μl). The reaction mixture was stirred vigorously for 30 min. Then, it was diluted with 1 ml of water and 1 ml of dichloromethane. The organic layer was separated, and the aqueous layer was extracted with dichloromethane (5 ml x3). The organic layer was collected and dried with Na2SO4, and the solvent was removed under reduced pressure. The obtained (>95% yield) solid colourful photocatalysts were used without further purification.

General procedure for the [2+1] cycloaddition reaction

A gas chromatography glass vial was charged with a magnetic stirring bar and photocatalyst (0.025 equiv., 2.5 mol%). The photocatalyst was dissolved in a 1:1 (v/v) solvent mixture of CH2Cl2 and pentane (0.1 M with respect to 1a). At room temperature, anethole 1a (1 equiv.) was added followed by diazoester 2a (2 equiv.) in the dark. Then, the glass vial was transferred to a −100 °C cryostat and kept at this low temperature for 30 min and then irradiated with 6 W of green light. After the desired time, the light was switched off, and after 10 min, the reaction was then quenched with saturated K2CO3 in ethanol and allowed to warm up to room temperature. The crude reaction mixture was analysed by proton nuclear magnetic resonance (1H NMR) using an internal standard and purified by SiO2 flash column chromatography using 5–10% acetone in pentane as eluent.