Chiral hafnium phosphate catalyzed asymmetric de Mayo reaction via visible-light-induced dearomatization

Abstract

Asymmetric photocatalysis has emerged as a powerful strategy for constructing chiral molecules under mild conditions, yet the enantioselective activation of aromatic enol systems via visible light remains a significant challenge. In particular, the photochemical dearomatization of salicyclaldehydes through excited-state intramolecular proton transfer (ESIPT) offers unique reactivity but has rarely been harnessed in catalytic asymmetric synthesis. Herein, we report an asymmetric de Mayo reaction catalyzed by a chiral hafnium phosphate complex, enabling visible-light-promoted dearomatization and exo-aromatic C–C bond insertion of salicylaldehyde derivatives with alkenes. By leveraging the unique photophysical properties of ESIPT systems, we demonstrate that coordination of 1-hydroxy-2-naphthaldehyde to a Hf(IV)–chiral phosphate catalyst induces a significant red-shift in absorption, activating the substrate under 480 nm irradiation. This strategy not only enhances photochemical reactivity but also provides excellent stereocontrol, yielding insertion products in up to 95% yield and 98% ee. Mechanistic studies, including transient spectroscopy and DFT calculations, reveal that the reaction proceeds via a triplet-state [2 + 2] cycloaddition pathway facilitated by hafnium-centered ligand-to-metal charge transfer (LMCT). This work establishes a distinctive catalytic strategy for enantioselective transformations of aromatic enol systems under visible light and expands the synthetic utility of ESIPT-driven photochemistry.

The authors report an asymmetric de Mayo reaction catalyzed by a chiral hafnium phosphate complex, enabling visible-light-promoted dearomatization and exoaromatic C–C bond insertion of salicylaldehyde derivatives with alkenes.

Introduction

Photochemistry of carbonyl compounds and their enol/enamine tautomers underpins a wide range of synthetically and mechanistically significant transformations in classical and modern photoatalysis^1–3. Phenol derivatives such as salicylaldehyde, commonly regarded as aromatic “enols” in the ground state, readily undergo photoinduced dearomatization to keto tautomers via excited-state intramolecular proton transfer (ESIPT) (Fig. 1A)^4,5. Although ESIPT has been extensively exploited in the design of fluorescent sensors, imaging agents, and luminescent materials^6,7, catalytic steering of phenol derivative-based excited-state enol–keto interconversion remains underdeveloped^8–13, when compared to the well-established photoreactions of conventional photoenolization systems^14,15. Those studies, mechanistically varied from the typical ESIPT process, are mainly focused on the capture of highly active photo-induced quinone methide intermediates, as illustrated in a number of notable synthetic applications developed by Yang¹⁴, Melchiorre¹⁵, Gao^16–19 and other groups^20–30.

Fig. 1. Reaction design.

A General form of excited-state intramolecular proton transfer. B Lewis acid catalysis on ESIPT-derived photoinduced. C Hafnium phosphate catalyzed enantioselective de Mayo-type reaction of salicylaldehyde derivatives.

In this context, metal-centered Lewis acids offer a distinctive and tunable strategy for redirecting photoactivation pathways^6,31–34. Through coordination to aromatic enol substrates, such metal complexes may induce pronounced red-shifting of absorption profiles, enabling visible-light excitation and unlocking novel excited-state reactivity beyond classical ESIPT scenarios. Among a range of Lewis acid screened, we successfully identified, hafnium (Hf) complexes—endowed with strong Lewis acidity and distinctive electronic properties—as promising candidates for advancing photoinduced dearomatization and asymmetric catalysis³⁵ (Fig. 1B). Herein, we report chiral hafnium phosphate catalyzed asymmetric de Mayo-type reaction of salicylaldehyde derivatives and alkenes (Fig. 1C). The classical de Mayo reaction, a photoactivated coupling between enol and alkenes, has long served as a powerful C–C shuffling strategy for constructing complex molecular scaffolds^36–39. The recent development of visible light version^40–46 as well asymmetric catalysis^47–49 add new arsenals to its synthetic applicability. Despite of these advances, to the best of our knowledge, an aromatic version with aromatic enol as in salicyladehyde skeletons has not been reported. On the other hand, previous efforts on ESIPT systems mainly led to [4 + 2]⁵⁰, [4 + 3]⁸ or [4 + 4]⁹ cycloadduct and in a few cases, asymmetric catalysis was achieved^{15,23,29,30,51}. In fact, de Mayo-type photoreactivity with ESIPT systems remains virtually unknown. Through combined experimental and theoretical studies, it was found that the reaction proceeded through [2 + 2] cycloaddition with Hf-complexed keto int1 (Fig. 1C), followed by a retro-aldol pathway through a cycloadduct int2 and then a ring opening int3 to give an exo-aromatic C-C bond insertion product with high chemo- and stereoselectivity (Fig. 1C).

Results and discussion

Screening and optimization

The coupling between 1-hydroxy-2-naphthaldehyde 2a and styrene 3a was selected as the model reaction for screening and optimization (see SI Table 6–16 for details). Among a range of Lewis acid screened, HfCl₄ and ZrCl₄ were two most active Lewis acid and other metal Lewis acids such as Sc, Mg, Al, Fe, Ti showed rather low or virtually inactive for the reaction (Table 1, entries 1 and 2 vs 3-5). HfCl₄ stood out in terms of both activity and stereoselectivity, the reaction gave the de Mayo adduct 4a in 92% yield and 88% ee under the optimized conditions with HfCl₄/1a (Table 1, entry 1). Perfluoro-aryl side chain was found to be critical for stereocontrol, and other fluorinated side chains such as 1b and 1c, or non-fluorinated derivative gave inferior enantioselectivity (Table 1, entries 6-8). Notably, a quinone additive Q-1 could consistently improve the enantioselectivity, with only a marginal effect on activity (Table 1, entry 1 vs 9), suggesting a delicate tuning effect likely through metal coordination (vide infra for details). The reaction preferred a light wavelength of 480 nm, and the performance decreased with lower or longer wavelength (Table 1, entry 10). In addition, a mixed solvent involving 1:1 volume ratio of hexane and dichloromethane was identified as the optimal reaction media with a balance of reactivity and stereoselectivity (Table 1, entries 11 and 12). Further increasing the molar ratio of ligand and metal to 2:1 didn’t lead to any improvement on enantioselectivity, suggesting a likely optimal mono-ligand mode in the key stereo-determining step (Table 1, entry 13). In control experiments, the reaction did not proceed at all in the absence of HfCl₄, light irradiation or chiral phosphoric acid 1a, pinpointing the light-responsive nature and the critical role of Lewis acid catalysis (Table 1, entries 14, 15 and 17). The dramatic effect of chiral phosphoric acid on activity is mainly ascribed to a solubility issue, as the Lewis acid is only sparsely soluble in hexane/DCM solution and the switch to acetonitrile solvent restore the reactivity to some extent (Table 1, entry 16).

Table 1.

Reaction Screening

Entry	Variation from the standard condition	Yield/%	ee/%
1	None	92	88
2	ZrCl4	75	80
3	MgCl2	N.R.	–
4	FeCl3	N.R.	–
5	TiCl4	N.R.	–
6	1b	37	37
7	1c	83	65
8	1d	52	5
9	w/o Q-1	81	68
10	400 nm or 520 nm	87/37	80/76
11	CH2Cl2	83	76
12	hexane	< 5	–
13	1:2 (Hf/1a)	81	87
14	w/o HfCl4	N.R.	–
15	w/o 1a	N.R.	–
16	w/o CPA (MeCN)	50	N/A
17	w/o light	N.R.	–

^aConditions: The reaction was performed with 2 (0.1 mmol), 3 (0.2 mmol), HfCl₄ (10 mol%), CPA 1a (11 mol%), benzoquinone Q1 (20 mol%) and in DCM/hexane (0.5 mL/0.5 mL) under 480 nm irradiation at room temperature for 4 h after being freeze-pumped-thawed for three times.

^bIsolated yield.

^cEnantiomeric excess: determined by HPLC with a chiral column.

Substrate scopes

Under the optimized reaction conditions, we examined the scopes of substrates (Fig. 2). The reaction showed good compatibility with different styrenes. Styrenes bearing alkyl group (t-Bu, 5b), electron-donating alkyloxyl groups such as 5c, 5 d and 5q, halogens such as 5e and 5 f, and electron-withdrawing groups such as -CF₃ (5 g), -NO₂ (5 h), -CN (5i) and -CO₂Me (5j) all worked well to give the desired adducts in moderate to good yields and with 73%-98% ee (Fig. 2, entries 1–10 and 15). Other functional groups such as acetoxyl (5 l), hexafluoro-isopropyl 5 m, o-Boc (5n) and borate 5 u were also well tolerated, leading to alkene insertion adducts with good yields and 80%-84% ee (Fig. 2, entries 12–14 and 16).

Fig. 2. Substrate scope.

^a Conditions: The reaction was performed with 2 (0.1 mmol), 3 (0.2 mmol), HfCl₄ (10 mol%), CPA 1a (11 mol%), benzoquinone Q-1 (20 mol%) and in DCM/hexane (0.5 mL/0.5 mL) under 480 nm irradiation at − 40 °C for 4 h after being freeze-pumped-thawed for three times. ^b Isolated yield. ^c Enantiomeric excess: determined by HPLC with a chiral column. ^d For 48 h. ^e With 15 mol% CPA 1a and DCM/hexane (0.25 mL/ 0.75 mL). ^f The Crystallographic data generated in this study have been deposited in the Cambridge Crystallographic Data Center (CCDC) under accession code CCDC 2487538.

Furthermore, when it comes to 3,5-dimethyl-4-methoxystyrene (5o), piperylene (5p), and 3,4-dimethoxystyrene (5r), the reaction achieved excellent results with well-balanced yields and enantioselectivities (74–90% yield, 90–95% ee) (Fig. 2, entries 17–19). Meanwhile, styrene derivatives bearing sterically hindered substituents such as 3,5-di-tert-butyl (5 v) also exhibited good reaction performance, with 75% yield and 85% ee (Fig. 2, entry 22). Polyfluorinated styrenes (5 s, 5t and 5w) also enabled good yields and high enantioselectivities (Fig. 2, entries 20-21 and 23). In addition, internal alkene such as indene reacted smoothly to furnish the expected insertion adduct 5x with 76% yield and 77% ee as a single diastereoisomer (Fig. 2, entry 24). Moreover, substrates bearing a naphthyl ring (5 y), aromatic heterocycles such as benzofurane (5z) and thiophene (5aa) showed good compatibility in the reaction (Fig. 2, entries 25–27). Notably, a methylenecyclobutane derived alknene could be successfully incorporated in the reaction to afford a spiro-bicyclic adduct (5ab) with 79% yield and 78% ee (Fig. 2, entry 28). Unfortunately, the reactions did not work with other di-substituted alkenes or simple alkyl-substituted alkenes (such as 1-hexene and 1-octene) (Supplementary Table 5). We attribute this lack of reactivity to the electronic properties of the alkenes, where the conjugation of the aryl group is essential to facilitate the key bond-forming step. Nevertheless, the reaction worked well with alkenes bearing complexed functional skeletons derived from Celestolide (5ac), Ibuprofen (5ad) and Tyrosine (5ae) (Fig. 2, entries 29–31), demonstrating its synthetic applicability.

Subsequently, the scope of aromatic aldehydes was also investigated. For different alkyl substituents on 1-hydroxy-2-naphthaldehydes (6a-d), the reaction could afford the desired adducts with good yields and in high enantioselectivities (Fig. 2, entries 32–35). Among them, the 8-methyl-naphthaldehyde showed significantly lower conversion (6 d), which might be attributed to the steric hindrance. The reactions with 6-methoxy (6e), 6-halogen (6 f and 6 g) and 4-aryl (6 h) substituted naphthaldehydes proceeded well to give the expected products in good yields (63%-94%) and high enantioselectivities (83–90% ee) (Fig. 2, entries 36-39). Swap the aldehyde moiety to acetyl (6i) led to a smooth reaction with 30% yield and 42% ee (Fig. 2, entries 40), however, no reactivity was observed with a benzoyl derivative (Fig. 2, entry 41). 2-hydroxy-1-naphthaldehdye (6j) also reacted, albeit with a moderate 73% yield and 33% ee (Fig. 2, entries 42). Swap the naphthalene ring to phenanthrene (6k), the reaction can also proceed smoothly, with 20% yield and 30% ee (Fig. 2, Entry 43). Simple salicyladehydes were also attempted, with unfortunately no reactivity (SI, Table 5)

The scale-up reaction was conducted on a 3 mmol scale, the target product 5r was finally obtained with a yield of 61% and an ee value of 96% (Fig. 2, entry 19), verifying its synthetic applicability. Then, we carried out derivatization studies with the initial insertion adduct 5r’ (Fig. 3). Specifically, hemiacetal 5r’could be readily transformed to dihydropyran 7a after dehydration (88%, 94% ee), to alcohol 7b (86%, 94% ee) or tetrahydropyran 7 d (79%, 96% ee) upon reduction or reduction-Mitsunobu sequence, respectively; and to amide 7c (87%, 96% ee) via oxidation and amidation. In the above transformations, the enantioselectivity of the products was well maintained.

Fig. 3. Derivatization of the product.

Synthetic transformations to diverse functionalized products via dehydration (7a), reduction (7b), oxidative amidation (7c) and reduction-Mitsunobu cascade (7 d).

Control experiments

Mechanistic studies were then conducted to disclose the origin of the current hafnium catalysis with respect to the role of metal, chiral phosphoric ligand and the quinone additive.

(1)The role of metal Lewis acid

Hafnium(IV) chloride (HfCl₄) and zirconium(IV) chloride (ZrCl₄) stood out as the two working metal Lewis acids, with the former exhibiting superior performance in terms of both activity and enantioselectivity (Fig. 4a). In sharp contrast, other tetravalent metal TiCl₄, trivalent metals such as AlCl₃, ScCl₃, CeCl₃ and InCl₃, and divalent metals such as ZnCl₂ and MgCl₂ showed rather poor or virtually inert for the reactions in most of the examined cases (Fig. 4a). This marked difference in catalytic efficiency underscores the decisive role of metal center selection in balancing reactivity and stereochemical control.

Fig. 4. Control experiments.

a Lewis acid effect in the model reaction. b UV-visible spectroscopy of 2a upon the addition of Lewis acid catalyst (All Tests were conducted in DCM at 3.3*10^-4M, with being freeze-pumped-thawed for three times). c quinone additive effect. d Non-linear effect of phosphoric ligand 1a and (e) triplet quenching experiments.

UV-vis absorption was conducted to probe the complexation behavior between the Lewis acid and substrate. 1-Hydroxy-2-naphthaldehyde (2a) in dichloromethane showed a characteristic π-π* absorption at ca. 360 nm (Fig. 4b). Adding chiral chiral phosphoric acid 1a caused no obvious shift. Notably, the addition of HfCl₄ together with 1a led to a significant decrease of 360 nm absorption and the appearance of a strong red-shift peak at 450 nm. Similar changes were also noted with ZrCl₄, but to a less extent. In contrast, no obvious red shift was observed with other Lewis acids. These observations add support to the observed visible light photoactivity with Hf and Zr Lewis acids, suggesting distinctive photophysical and photochemical properties associated with the Hf-2a complex.

(2) The role of quinone additive

Control experiments showed that increasing the loading of Q1 to one equivalent could completely inhibit the reaction (Fig. 4c). This inhibition arises because quinone incorporation is limited to precisely one molecule per Hf center due to steric crowding; excess quinone leads to overcoordination, disrupting the delicate coordination sphere required for substrate activation and catalysis. A survey of different quinones clearly indicated a favorable steric effect on the enantioselectivity by increasing the size of the substituent from H (Q2), methyl (Q3), to tert-butyl group (Q1) (Fig. 4c, right). The UV-visible light absorption band becomes more intense and a little red-shifted in the presence of Q1, indicating an electronic contribution through quinone complexation (Fig. 4b). DFT calculation reveals a decrease of HOMO-LUMO energy gap from ~ 3.4 eV in Int1 to ~ 2.5 eV in Int1 + Q due to quinone’s involvement (Fig. 5a). Further analysis suggests quinone additive may contribute electronically by mixing its low-lying LUMO with the hafnium d orbital, decreasing the overall LUMO energy (Fig. 5a, right), in line with experimental observations (Table 1, entry 9 vs 1). Specifically, this orbital mixing in the quinone-involved complex (Int1 + Q) alters the nature of the electronic transition from a ligand-to-metal charge transfer (LMCT; from the π orbital of ligand 2a to the Hf center d orbital) in the quinone-free complex (Int1) to a ligand-to-ligand charge transfer (LLCT; from the π orbital of 2a to the π* orbital of quinone, Supplementary Fig. S10).

Fig. 5. Emission spectroscopic investigations.

a Comparison of HOMO–LUMO energies for the keto form of 2a (II), enol form of 2a (I), Hf- 2a complex w/o quinone (Int1), and Hf- 2a complex with Q1 (left). Isosurfaces of HOMO and LUMO orbitals for the Int1. b Steady-state emission (upon 370 nm excitation) spectroscopy studies. All Tests were conducted in DCM at 3.3*10^-4M, with being freeze-pumped-thawed for three times. c Transient spectroscopic studies (upon 450 nm excitation). (y-axis is logarithmic(up)) All Tests were conducted in DCM at 0.0125 M, with being freeze-pumped-thawed for three times. d Quenching experiments of 2a against 3a. (upon 450 nm excitation) The test was conducted in DCM (4.0 equiv styrene was added) at 0.0125 M, with being freeze-pumped-thawed for three times.

In addition, transient absorption spectroscopy (TAS) experiments demonstrate that the triplet (T1) state lifetime in the quinone-involved complex is extended from > 300 ns to > 400 ns (Supplementary Fig. S7, SI, right). This prolonged excited-state lifetime reduces non-radiative decay pathways and allows for more effective energy transfer or electron transfer processes in the reaction, contributing to the observed improvement in yield and selectivity. Furthermore, spin-orbit coupling (SOC) calculations reveal a larger SOC value in the quinone-involved complex for the S1 and T1 states (< T1|H_soc | S1 > ), leading to faster intersystem crossing (ISC) from S1 to T1 compared to the quinone-free complex. This enhanced ISC rate promotes more efficient population of the T1 state, complementing the extended lifetime and improving the photocatalytic performance. These findings underscore quinone’s dual role as both a steric modulator and an electron acceptor/stabilizer of the excited state.

(3) The complexation mode

Our previous work⁴⁷ as well as others⁵² have proved that chiral phosphoric acids weakly coordinate with metal Lewis acid in 1:1 or 2:1 ligand/ metal ratio. In this reaction, a 1:1 ratio was found to give the optimal enantioselectivity (Fig. 1, entry 13). A survey of the non-linear effect of chiral phosphoric acid 1a revealed a linear correlation with the observed enantioselectivity (Fig. 4d), excluding the possible involvement of high-order Hf-1a complex, and adding support to a 1:1 complex as the catalytic active species. On the basis of the known coordination behavior of ortho-hydroxynaphthaldehyde structural motif⁵³, we proposed a 1:1 bidentate chelation between Hf-1a and 2a as illustrated in Fig. 5a.

Efforts to crystallize the complex have been in vain. Nevertheless, we were able to optimize the Hf-1a-2a complex structure using density functional theory (DFT) and calculate the excitation energy via time-dependent density functional theory (TD-DFT) Fig. 5a. The calculated UV spectra based on the optimized structure showed good agreement with the experimental spectra, with a strong red shift absorption to 450 nm (Supplementary Fig. S4 in SI). HOMO and LUMO orbitals and TD-DFT calculations confirm its ligand-metal charge transfer (LMCT) character with the HOMO residing on the ligand’s enolate/phenyl π-system, while the LUMO localizes on hafnium d-orbitals (Fig. 5a). The LMCT process involves promotion of an electron from the ligand’s highest occupied molecular orbital (HOMO, π-orbitals) to the metal’s lowest unoccupied molecular orbital (LUMO, d-orbitals).

A comparison of the HOMO-LUMO energy levels for the Hf-coordinated complexes (Int 1 and Int1 + Q) and the ESIPT process (keto form II) is also presented in Fig. 5a. While ESIPT decreases the LUMO level by only 0.2 eV resulting in a modest reduction of the HOMO-LUMO gap, 2a and CPA coordination to Hf incorporates a significant portion of the Hf d-orbitals into the π* orbital of 2a (Fig. 5a), leading to substantial stabilization of the LUMO and an overall narrowing of the HOMO-LUMO gap by approximately 0.4 eV compared to the enol form of 2a. The additional coordination of quinone further enhances this effect by lowering the LUMO by an additional ~ 2.0 eV, accounting for the significant red-shift observed in the UV-vis spectrum.

Steady emission and transient absorption spectroscopic studies

Quenching experiments were carried out and the reactions were largely inhibited in the presence of typical triplet quenchers such as 2,5-dimethyl-2,4-hexadiene, pyrene, perylene or oxygen (Fig. 5e), suggesting that the reaction proceeded through a triplet excited state. To gain deeper insights into the excited-state, we first investigated the steady-state emission spectra of 2a upon 370 nm excitation (Fig. 5b). The dual fluorescence emission of compound 2a in DCM, observed at ~ 465 nm (enol form) and ~ 489 nm (keto tautomer), serves as a signature of the ESIPT mechanism⁵⁴. Adding HfCl₄-1a complex diminished the overall emission, particularly the enol band, suggesting coordination alters the ESIPT pathway.

To elucidate the photophysical origin of the observed photo-reactivity, we investigated the femtosecond to nanosecond transient absorption spectroscopy (TAS) of Hf-1a-2a complex upon excitation at 450 nm (Fig. 5c). The time-resolved spectra showed that an initial negative signal at ~ 520 nm (III), which decayed within about 100 ps, was replaced by a positive absorption band lasting into the nanosecond timescale (II), accompanied by a weaker feature at ~ 600 nm that decayed rapidly in less than 1 ns (I). In addition, a distinct band at ~ 800 nm (IV) rose from the picosecond timescale and persisted up to the experimental limit of 8 ns, with an extrapolated lifetime exceeding 300 ns (Supplementary Table 20, SI).

The strong band near 500 nm can be assigned to the S₁ states of keto (II)-enol (I) equilibration of 2a. This band is characteristic of ESIPT emission³⁶ and was also detected in the emission spectrum under 450 nm excitation (Supplementary Fig. S3, SI). This observation suggests the free 2a coexists in solution due to the dynamic feature of hafnium coordination, and it mainly participates in an emission pathway. The band at ca. 600 nm, likely originated from T1 state of 2a (III), which should be negligible for photoreactivity due to its rather short lifetime (< 1 ns). The near-IR band at ~ 800 nm, with its long lifetime, could be ascribed to the triplet state of the Hf-complex system (IV). This prolonged lifetime may result from the heavy atom effect of hafnium, which enhances spin-orbit coupling, thereby facilitating intersystem crossing to a triplet state that inherently exhibits slower decay rates compared to singlet states.

To further validate the above deduction, we also measured the transient absorption spectra in the presence of styrene (Fig. 5d). Notably, the lifetime of this long-lived ~ 800 nm band shortened dramatically from > 300 ns to about 1 ns (Fig. 5d). This quenching effect implies that the long-lived excited state is directly involved in the reaction with the alkene. This observation further supports that the reaction proceeds via the triplet Hf-complex (Int 1), where the metal coordination enables efficient ISC and triplet-mediated cycloaddition under visible light.

Based on the above experimental investigations, we validated the key reaction pathway, particularly the photochemical steps (Int1 to Int2) through computational studies (Supplementary Fig. S9a). Density functional theory (DFT) calculations were conducted at the (U)PBE0/6-311 + G(2 d,p), SDD // (U)PBE0/6-31 G(d), SDD level of theory, incorporating SMD solvation in dichloromethane (DCM). The reaction initiates with the coordination of substrate 2a to the Hf(IV) center in the presence of chiral phosphoric acid (CPA), resulting in the formation of complex Int1 (Fig. 1). Irradiation with a 480 nm LED promotes multiple singlet excited states of Int1, leading to efficient intersystem crossing (ISC) facilitated by the heavy-atom effect of Hf and subsequent relaxation to the first triplet state (³Int1). The resulting triplet manifold drives a “stepwise-pseudo” [2 + 2] cycloaddition with alkene 3a (Supplementary Fig. S9, SI), beginning with C–C bond formation to yield biradical ³pre-Int2. Subsequent ring closure occurs via a minimum energy crossing point (MECP) to generate singlet intermediate ¹Int2. This step may also govern the enantioselectivity, as illustrated by the proposed schematic representation of the key transition state (Supplementary Fig. S9b, SI). The observed enantioselectivity could arise from enhanced π–π interactions and steric effect from the aromatic rings of 2a, 3a and the perfluorinated arene of phosphoric ligand 1a.

Conclusion

In summary, this study demonstrates an enantioselective de Mayo reaction enabled by a chiral hafnium phosphate catalyst under visible light irradiation. By exploiting the unique photophysical properties of ESIPT-capable salicylaldehyde derivatives, we have developed a mild and efficient method for exo-aromatic C–C bond insertion with broad substrate scope and excellent stereocontrol. Mechanistic investigations reveal that the reaction proceeds through a visible-light-activated triplet manifold via a hafnium-stabilized ligand-to-metal charge transfer (LMCT) process, which not only facilitates dearomatization but also ensures high enantioselectivity. This work bridges the gap between classical ESIPT photochemistry and modern asymmetric catalysis, offering a distinctive strategy for the stereocontrolled functionalization of aromatic enol systems. We anticipate that this hafnium-based photocatalytic platform will inspire further developments in excited-state asymmetric synthesis and expand the toolbox for constructing complex chiral architectures under sustainable conditions.

Methods

Unless otherwise stated, all reactions were carried out on a 0.1 mmol scale. In a glovebox, 1-hydroxy-2-naphthaldehyde 2a (0.1 mmol, 1.0 equiv.), CPA 1a (11 mol%), quinone additive Q-1 (20 mol%), and HfCl₄ (10 mol%) were added to an oven-dried 10 mL or 25 mL Schlenk tube. The reaction tube was then removed from the glovebox, and styrene 3a (0.2 mmol, 2.0 equiv.), anhydrous dichloromethane (0.5 mL), and anhydrous n-hexane (0.5 mL) were added under an argon atmosphere. The reaction mixture was freeze-pump-thawed for three times. The Schlenk tube was placed in a cooling bath pre-cooled to − 40 °C, and after the temperature stabilized, the reaction system was irradiated with 480 nm light for 4 hours. Upon completion of irradiation, the mixture was warmed to room temperature, and Dess-Martin reagent (0.1 mmol, 1.0 equiv.) was added, followed by stirring for 6 h. After the reaction was complete, the crude mixture was directly purified by flash column chromatography to afford the target product 5a.

Author contributions

X.M. and S.L. conceived the projects and designed the experiments. C.Y. and Y.W. conducted synthetic and mechanistic experiments with the help of C.H., X.L., and J.Y. Y.Z. conducted DFT calculations. S.L., X.M., and C.Y. wrote the manuscript with contributions from all authors.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available from the corresponding author upon request. Supplementary Information is available and includes general information, substrate and reagent synthesis, optimization details, general experimental procedures, and compound characterization, determination of the absolute configuration, mechanistic studies, HPLC, NMR spectra. Crystallographic data generated during this study have been deposited in the Cambridge Crystallographic Data Center (CCDC) under accession number CCDC: 2487538 (5 f). These data can be obtained free of charge from the CCDC at http://www.ccdc.cam.ac.uk/data_request/cif. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72827-5.