Unraveling the mechanistic origins of efficiency and stereoselectivity in EDA-mediated photocatalysis Authors: Mengchen Fu, Yuanyuan Luo, Luyang Qin, Nicole P Qin, Jialong Jie, Hongmei Su Published: 2026 ABSTRACT AbstractVisible-light-driven electron-donor-acceptor (EDA) photocatalysis holds transformative potential for asymmetric synthesis, yet its efficiency and selectivity are dictated by excited-state dynamics that have, until now, eluded direct observation. Here, we utilize femtosecond and nanosecond transient absorption spectroscopy to investigate a prototypical EDA system: asymmetric α-alkylation of aldehydes. Our real-time kinetic mapping successfully resolves the singlet charge-transfer state (¹CT), triplet charge-transfer state (3CT), and donor-acceptor radical pairs (D•+ and A•). We uncover two concurrent radical-generation pathways: direct decay of ¹CT and a 3CT-mediated route, which maximize reaction efficiency. Rapid Br- departure from A•- furnishes A•, suppresses unproductive charge recombination with D•+, while solvent-cage confinement preserves radical orientation, promoting in-cage coupling with high enantioselectivity. These findings establish a general mechanistic blueprint where ground-state pre-organization and excited-state dynamics are synergistically harnessed, offering a distinct paradigm to overcome diffusion-controlled limits for the rational design of advanced photocatalytic systems.Subject terms: Photocatalysis, Synthetic chemistry methodology Introduction In recent years, light-driven catalytic asymmetric transformations have garnered significant attention as a powerful tool for the synthesis of chiral products1,2. Among these, dual-catalysis strategies have emerged as a promising approach for achieving enantioselective reactions by combining transition metal-based photoredox catalysts (typically polypyridyl-coordinated Ir and Ru complexes) with chiral organic catalysts, such as amine3,4. In such systems, light excitation of the photoredox catalysts triggers single-electron transfer (SET) with substrates, generating radicals that undergo selective transformations under the control of chiral organic catalysts, which govern the position and direction of radical attack5,6. However, the reliance on toxic metal complexes in photoredox catalysis limits their applicability, particularly in medicinal synthesis, raising the question of whether photocatalytic transformations with high enantioselectivity can be achieved under metal-free conditions. To address this challenge, simpler and more sustainable methodologies utilizing only chiral organic catalysts have been developed, offering reduced toxicity and operational simplicity7–10. Inspired by these advances, promising strategies for visible light-induced asymmetric transformations have been established, leveraging electron-donor-acceptor (EDA) complexes, a distinct class of noncovalent interactions (e.g., nπ*, ππ*)11–18. EDA complexes typically arise from the association of an electron donor (e.g., amines, enamines, or enolates) and an electron acceptor (e.g., electron-deficient substrates)19–21. While these substrates are intrinsically non-photosensitive, the complex formation results in visible light absorption, enabled by a unique electronic transition characteristic of EDA interactions19–21. Upon visible light irradiation of the EDA complex, electron transfer from the donor to the acceptor occurs, generating reactive open-shell intermediates. These intermediates are expected to facilitate stereoselective transformations under the influence of chiral catalysts. EDA-mediated transformations thus represent a promising frontier for asymmetric synthesis, offering a sustainable and operationally simple platform for the preparation of chiral compounds19–21. A seminal example of this approach was reported by Melchiorre and co-workers in 2013, showcasing the potential of EDA-mediated asymmetric catalytic α-alkylation of aldehydes with alkyl halides under visible light irradiation17. In this system, an aldehyde substrate, in the presence of a chiral amine catalyst, forms an enamine that acts as the electron donor (D). This enamine interacts with an electron-deficient alkyl halide, acting as the electron acceptor (A), through nπ* interactions, assembling into an EDA complex. Visible light excitation of this complex induces a charge transfer (CT) transition between D and A, facilitating carbon–carbon bond formation with high enantioselectivity. This strategy accommodates a broad range of bromide-containing acceptors and enamines derived from extended aldehydes, delivering enantioenriched products in high yields. This landmark study highlights the key role of EDA complexes as visible light absorbers in driving stereoselective transformations, offering a practical and efficient approach to addressing stereochemical challenges in asymmetric catalysis. Despite these growing applications, it is still lack of deeper insights into the principles underlying how EDA complexes enable visible light-driven catalytic asymmetric transformations with both high product yields and excellent stereoselectivity. Specifically, the excited-state dynamics of EDA complexes have yet to be directly observed, leaving key aspects of the reaction cascade unclear, particularly how charge transfer and subsequent bond-forming events are efficiently orchestrated. Gaining deeper insight into these processes is essential for advancing the rational design of EDA complexes in photocatalytic asymmetric transformations. Fundamentally, this addresses a central challenge in photocatalysis: how to rationally design molecular systems that convert photon energy into chemical bonds with minimal energetic waste and maximal stereochemical precision. Herein, we employ femtosecond (fs) and nanosecond (ns) time-resolved transient absorption spectroscopy to investigate a representative EDA-mediated reaction: the asymmetric α-alkylation of aldehydes. Distinct transient absorption profiles enable the real-time identification of key reactive species (1CT, D•⁺, and 3CT). Upon photoexcitation, the directly populated 1CT (1[D•⁺…A•-]) undergoes selective cleavage of the C–Br bond in A•-, leading to Br- departure and the formation of radicals D•⁺ and A•. Alternatively, 1CT can undergo intersystem crossing (ISC) to form 3CT, which converges to the same radical pair (D•⁺ and A•) via a parallel dissociative pathway. These radicals predominantly undergo in-cage coupling, ultimately yielding the final α-alkylation product. The dual-pathway radical generation mechanism is pivotal in maximizing efficiency, as the concentration of D•⁺ and A• directly governs the reaction yield. Remarkably, the irreversible departure of Br-, minimizes charge recombination between D•⁺ and A•-, while preserving the original donor and acceptor orientation for the correspondingly generated radicals. This precise spatial organization facilitates the formation of the stereogenic center in the α-alkylation product with high stereoselectivity. By leveraging these unique structural and mechanistic features, EDA complexes achieve significant enhancements in both photocatalytic yield and stereoselectivity. In contrast to traditional photocatalytic systems that rely on radical escape and diffusion, the EDA strategy provides a highly efficient and well-defined approach to photocatalytic reactions. More broadly, these findings establish a powerful blueprint for photocatalyst design, where the synergy between ground-state pre-organization and tailored excited-state dynamics provides a controllable, metal-free strategy to overcome the classical diffusion-limited paradigm and achieve exceptional stereochemical precision. Results Formation and characterization of the EDA complex In this study, we first synthesized reactive nucleophilic enamine species, which are well-established to form via efficient condensation of amines with aldehydes in a 1:1 ratio under dehydration conditions, releasing a molecule of water (Fig. 1a)11,13,17,18,22. The UV-vis absorbance data collected for the chiral secondary amine (catalyst), aldehyde (trans-2-Pentenal), and their mixture show the appearance of a new absorption spectral feature within the range of 275 ~ 380 nm, clearly indicating the formation of enamine (Supplementary Fig. 1)11,13,17,18,22. Formation and steady-state characterization of the EDA complexes. a Schematic illustration of EDA complex formation and the subsequent α-alkylation process under visible light irradiation. b UV-vis absorption spectra of electron acceptors (bromides A1, A2, A3), electron donors (enamines D1, D2), and their corresponding EDA complexes (A1/D1, A1/D2, A2/D1, A3/D2). The concentrations of electron acceptors are ~ 5 mM, and electron donors are ~ 1 mM, with methyl tert-butyl ether (MTBE) as the solvent. Following enamine formation, the synthesis of an electron-donor-acceptor (EDA) complex was carried out, as previously reported17. The pyrrolidine ring within the enamine structure possesses a lone pair of electrons (n orbital), enabling it to act as a nucleophile and electron donor17,23,24. In contrast, certain alkyl halides exhibit high electron affinity (EA) due to the presence of π* orbitals and electron-withdrawing groups (EWGs), enabling their function as electrophiles and electron acceptors17,25. These complementary electronic characteristics facilitate interactions between the two species through nπ* interactions17,26,27, resulting in the formation of the EDA complex (Fig. 1a). As depicted in Fig. 1b, the phenacyl bromide (A1) displays absorption below 400 nm, with a characteristic band centered around 330 nm. Upon addition of enamine to a solution of A1, a distinct absorption spectrum emerges, extending up to 650 nm, with two prominent bands observed at 490 nm and 535 nm. Literature attributes the absorption above 450 nm to the charge transfer (CT, n→π* transition) band17. The appearance of this CT band serves as further evidence for the formation of the EDA complex11–18,21,28–30. Notably, the absorption profiles of these EDA complexes are significantly modulated by the substituents’ electronic effect (Fig. 1b). Since the CT band originates from an electronic transition between the donor’s HOMO and the acceptor’s LUMO, the band position is dictated by the HOMO-LUMO gap. Our DFT calculations (Supplementary Fig. 2) reveal that the introduction of a strong electron-withdrawing nitro group in acceptor A2 significantly stabilizes its LUMO (− 0.13251 Hartree) compared to the phenacyl (A1, − 0.09180 Hartree) and naphthyl (A3, − 0.09422 Hartree) analogs. Conversely, the extended conjugation in the donor D1 raises its HOMO energy (− 0.18429 Hartree) relative to the donor D2 (− 0.19639 Hartree). Consequently, the A2/D1 complex possesses the narrowest HOMO-LUMO gap, resulting in the most pronounced bathochromic shift of the CT band that possesses the largest overlap with the visible light spectrum, as shown in Fig. 1b. To further quantify the strength of the donor-acceptor interaction, we performed spectrophotometric titration experiments. By monitoring the increase in CT band intensity with increasing acceptor concentration, we determined the association constants (KEDA) for the EDA complexes to be in the range of 0.9 – 9.8 M−1 using the Benesi-Hildebrand method (Supplementary Fig. 3). Given this weak interaction, an excess of bromide electron acceptor was used (donor/acceptor = 1:5) to drives the equilibrium toward complex formation. To delve into the underlying photochemical mechanism that effectively drives stereoselective catalytic α-alkylation of aldehydes, we performed systematic femtosecond and nanosecond transient absorption (fs-TA and ns-TA) spectroscopy to monitor the excited-state decay dynamics of the EDA complex. The pump wavelength of fs-520 nm or ns-532 nm is chosen to selectively excite the CT band of the EDA complex. Ultrafast photodynamics of EDA complex Monitoring excited-state dynamics via femtosecond transient absorption spectroscopy Figures 2a, b display the fs-TA spectra of the EDA complex (A1/D1) upon 520 nm photoexcitation. Immediately after the pump pulse, an excited-state absorption (ESA) emerges, spanning from 375 to 480 nm, with one maximum around 430 nm, alongside a negative signal extending from 475 up to 780 nm with a distinct negative band at 545 nm. Both features are developed within the instrumental response time (∼ 200 fs). As the photoexcitation of the CT band directly leads to the formation of the 1CT state, these transient signals observed in the initial time 0.3 ps should correspond to the ESA of the 1CT state (positive 430 nm band) and the ground-state bleach (GSB, negative 545 nm band) of EDA. Afterward, the 1CT features begin to decay, almost back to baseline within 7 ns. Concomitantly, the GSB band at 545 nm gradually recovers, nearly returning to the baseline on the same timescale. This evolution suggests the decay of 1CT likely involves charge recombination to the ground state, which actually is a well-known decay pathway for the 1CT state31. Femtosecond transient absorption spectroscopy of the EDA complex A1/D1. a Transient absorption 2D contour map. b Absorption spectra at typical delay times. c Time traces at typical wavelengths (430 nm and 545 nm, circles) and fits (lines) from single-wavelength analysis. d Normalized absorption spectra at 0.3 ps and 7000 ps. e Difference spectra obtained by subtracting the spectrum at the initial time 0.3 ps from those at different time delays. f Time traces at typical wavelengths (circles) and fits (lines) from global analysis. These data were obtained for the EDA complex A1/D1 in MTBE upon fs-520 nm excitation. The concentration of electron acceptor A1 is ~ 50 mM, and that of electron donor D1 is ~ 10 mM. Representative kinetics traces at 430 nm and 545 nm corresponding to the decay of 1CT in Fig. 2c were analyzed. Interestingly, these traces could not be fitted with a single-exponential decay model. Instead, the most satisfactory fitting results can be obtained by a three-exponential function (y = A1e-x/t1 + A2e-x/t2 + A3e-x/t3), indicating the presence of multiple lifetimes for the 1CT. The single-wavelength fitting parameters are summarized in Table 1. In principle, when only one conformation is present, regardless of the number of 1CT relaxation pathways, the decay should follow a single-exponential process. However, EDA complexes are assembled via weak non-covalent interactions, which inherently lack rigid structural constraints and possess significant conformational flexibility. This structural flexibility may generate a heterogeneous ensemble of ground-state conformations characterized by varying donor-acceptor distances, relative orientations, and electronic coupling strengths32–34. Consequently, upon photoexcitation, this structural heterogeneity translates into distinct ¹CT populations with varying relaxation dynamics, directly giving rise to the observed multi-exponential decay kinetics. Such behavior has been documented in similar non-covalent EDA systems, where conformational distributions dictate complex decay profiles32–34. Therefore, the observed multiexponential decay process in the fs-TA measurements is expected to be a natural consequence of the inherent conformational flexibility of the EDA complex. In particular, the shortest time constant (τ1 ~ 4 ps) may correspond to the most compact conformation, which is expected to be the most favorable for the charge-recombination decay of 1CT. In contrast, the longest time constant (τ3 ~ 1.5 ns) may represent a much looser conformation, while the intermediate constant (τ2 ~ 100 ps) may correspond to a conformation that lies between these two extremes. Table 1Lifetimes of 1CT obtained from single-wavelength and global analysisτ1 (ps)A1τ2 (ps)A2τ3 (ps)A3λ = 430 nm3.812.9%10037.8%147749.3%λ = 545 nm3.614.3%9531.9%151453.8%Global Fitting3.5–100–1573– Lifetimes of 1CT obtained from single-wavelength and global analysis Spectral evidence for long-lived species generation Strikingly, the gradual recovery of the GSB band at 545 nm involves not only a decrease in intensity but also progressive bandwidth narrowing. Moreover, the spectral features at 7 ns, though weak, are distinct and markedly different from those observed at 0.3 ps, indicating the formation of a new long-lived species (Fig. 2d). Specifically, while the negative GSB band around 545 nm persists, the positive maximum at 430 nm shifts to 450 nm. Most notably, a switch from negative signals to positive signals is observed in the spectral region spanning 570–780 nm. These positive signals exhibit two maxima: a broad band centered around 730 nm and a relatively weaker one near 580 nm. To better visualize the temporal evolution, difference spectra are obtained by subtracting the spectrum at the initial time 0.3 ps from those at different time delays (Fig. 2e). The difference spectra clearly reveal that the decay of 1CT spectral features corresponds to the buildup of a new spectral shape with two maxima around 580 nm and 730 nm. Although this evolution is not pronounced during the fastest process (τ1 ~ 4 ps), it becomes increasingly evident during the two later processes, particularly during the slowest process (τ3 ~ 1.5 ns) (Supplementary Fig. 4). These spectral evolutions suggest that the 1CT with the shortest lifetime (τ1 ~ 4 ps) primarily decays to the ground state via charge recombination. In contrast, for the 1CT with the middle (τ2 ~ 100 ps) and longest (τ3 ~ 1.5 ns) lifetimes, especially the longest-lived state, an additional decay pathway exits, in which the 1CT state transforms into the new long-lived species. Crucially, the amplitudes for the three single-wavelength fitted time components are also show in Table 1. Compared to τ1, the longer time components of τ2 and τ3 exhibit the dominant contribution to the decay dynamics of 1CT. This amplitude distribution implies that the majority EDA complex adopts the looser conformation, which are less prone to rapid charge recombination. Consequently, a sufficient fraction of the 1CT state survives to drive the subsequent generation of the long-lived species. According to the above spectral analysis, a target model involving three time constants is used for global analysis of the fs-TA spectra (Fig. 2f). The fitted data show satisfactory agreement with the experimental data and are consistent with the results of the above single-wavelength multiexponential fitting (Table 1), thereby validating the reliability of the global fitting model. Characterization of long-lived species via nanosecond transient absorption spectroscopy To further investigate the newly observed long-lived species in the fs-TA measurements, nanosecond transient absorption (ns-TA) spectroscopy was performed. Figure 3a presents the ns-TA spectra of the EDA complex in a deaerated solution excited at 532 nm. A negative absorption band at ~ 540 nm is initially observed, corresponding to the ground-state bleaching of the EDA complex. Simultaneously, the spectrum displays positive absorption features with a peak at 580 nm and a broad band centered around 730 nm. These spectral characteristics are consistent with the 7 ns fs-TA results (Fig. 2d) and are assigned to the long-lived species. Nanosecond transient absorption spectroscopy of the EDA complex A1/D1. a Transient absorption spectra under N2 -saturated conditions. b Normalized decay kinetics traces for the transient absorptions at 730 nm under air- (red) and N2-saturated (black) conditions, with fitted curves represented by red and black solid lines, respectively. c Kinetics traces at longer timescales for the transient absorption signals at 580 nm under N2-saturated conditions, with the solid line showing the fit. d The SEC spectrum of the oxidized donor D1 (D•+, red) and the reduced acceptor A1 (A•, blue), measured in 0.1 M TBAPF6 acetonitrile solution using a Pt mesh working electrode. e Normalized kinetics traces for the transient absorption signals at 730 nm (green), 580 nm (pink), and 530 nm (blue) under N2-saturated conditions, with fitted curves shown as solid lines, respectively. f Normalized kinetics traces for the transient absorptions at 580 nm under air- (red) and N2-saturated (black) conditions. These transient data were obtained for the EDA complex A1/D1 in MTBE upon ns-532 nm excitation. The concentration of electron acceptor A1 is ~ 5 mM, and that of electron donor D1 is ~ 1 mM. Characterization of 3CT as one of long-lived species The ESA absorption band at 730 nm exhibits single-exponential decay kinetics with a lifetime of 7.8 μs under N2-saturated conditions (Fig. 3b). However, this lifetime is significantly shortened to approximately 1.3 μs under air-saturated conditions. This observation demonstrates that the long-lived species associated with the 730 nm band is oxygen-sensitive, supporting its assignment as a triplet state (Fig. 3b and Supplementary Fig. 5). To investigate the possible involvement of triplet-state species, subsequent control experiments were conducted. Upon selective 355 nm photoexcitation of donor D1 alone, no transient absorption signal was observed, possibly due to the lack of chromophore in the enamine-type D1 species. In contrast, 355 nm excitation of the acceptor A1 alone produced a characteristic triplet absorption band around 460 nm (Supplementary Fig. 6a). However, the EDA complex’s 730 nm band should not be directly assigned to this localized triplet state (3A*). Compelling evidence against a localized triplet assignment is provided by comparing complexes with varying donor/acceptor structures, such as A1/D1, A1/D2, and A3/D2 (Fig. 3a and Supplementary Figs. 7d, 9d). Despite the naphthalene-derivative A3 possessing distinct intrinsic triplet features located at ~ 425 nm (Supplementary Fig. 6c), compared to the phenyl-based A1 at ~ 460 nm, the resulting EDA complex retained an identical 730 nm absorption. Such an independence of the spectral feature on the variation of the acceptor core strongly suggests that the 730 nm absorption is dominated by charge-transfer character rather than the acceptor’s local excitation. Alternatively, the 730 nm feature is most likely assigned to the triplet charge-transfer state (3CT, 3[D•+…A•-]). In such EDA systems, ISC to 3CT is a thermodynamically favorable process. Moreover, the large separation between donor and acceptor in conventional EDA results in weak electron spin–spin interaction and a small electron exchange integral, which will allow radical pair ISC (i.e., 3CT) while rendering the formation of the localized excited triplet state (3LE) inefficient35. In this specific system, the presence of the heavy bromine atom in the acceptor moiety likely further facilitates this spin-forbidden transition through enhanced spin-orbit coupling, thereby promoting the population of the 3CT state. Lastly, the assignment of this 730 nm feature to 3CT well explains the spectral similarity across different acceptor complexes, as the absorption is dominated by the charge-transfer character rather than the localized acceptor orbitals. Characterization of radicals as coexisting long-lived species Interestingly, the decay of the 3CT state, represented by the 730 nm band, is not accompanied by a simultaneous decay of the 580 nm band. Under air-saturated conditions, the 730 nm band is rapidly quenched by oxygen with a lifetime of 1.3 μs, whereas the 580 nm band persists on the millisecond timescale (Fig. 3b, c and Supplementary Fig. 5). The different spectral evolution of the 730 nm and 580 nm strongly suggests that they belong to two different transient species. To assign the observed transient species responsible for the 580 nm band, we analyze the possible reaction pathways following the decay of the 1CT state. The 1CT state is essentially a chiral contact radical ion pair within a solvent cage (1[D•+…A•-]), consisting of the donor enamine radical cation (D•+) and the acceptor phenacyl bromide radial anion (A•-) (Fig. 4). The presence of bromine in A•-, a suitable leaving group, can trigger cleavage of the C-Br bond, leading to the rapid formation of the phenacyl radical (A•)36. Subsequently, A• and D•+ may couple to generate a radical addition intermediate. This intermediate can react with a molecule of water to produce the stable target product, while the chiral secondary amine catalyst is regenerated to complete the catalytic cycle17. Based on these pathways, the long-lived species responsible for the 580 nm band in the ns-TA spectra could correspond to D•+, A•-, A•, or the radical addition intermediate. Reaction mechanism for EDA-mediated asymmetric catalytic photochemical processes. The schematic illustrates the complete elementary reaction pathway, from the absorption of visible light by the assembled EDA complex to the final formation of the asymmetric α-alkylation product. Square brackets denote that the electron acceptor and donor are confined within a solvent cage. Abbreviations: CT charge transfer; CR charge recombination; ISC intersystem crossing. We first excluded A•- and the radical addition intermediate based on stability and kinetic considerations. Our DFT optimization indicates that A•- is inherently unstable and spontaneously eliminates Br- to form A• (Supplementary Fig. 10). Furthermore, the radical addition intermediate is unlikely to be the species observed on this timescale; its subsequent hydrolysis is a thermal process typically occurring on the order of seconds to minutes37,38, orders of magnitude slower than the millisecond decay dynamics observed here for the 580 nm band. To definitively identify the remaining candidates (D•+ and A•), we conducted spectroelectrochemistry (SEC) measurements (Fig. 3d and Supplementary Fig. 11). As shown in Fig. 3d, the SEC spectrum of the oxidized donor (D•+) exhibits a strong absorption peak at ~ 350 nm, accompanied by a shoulder band extending from 500 to 700 nm. In contrast, the spectrum of the reduced acceptor (assigned to A• following rapid debromination of A•-) shows an absorption band at ~ 305 nm, with quite weak absorption in the visible region. These SEC results support the assignment of the 580 nm transient absorption primarily to the D•+, while A• contributes very little. It is noticeable that the primary radical absorption features identified by SEC are largely obscured in the TA spectra due to significant spectral superposition with GSB. Specifically, the radical absorption at wavelengths shorter than 580 nm are masked by the intense GSB signal (Figs. 2d and 3a). The EDA complex possesses a high molar extinction coefficient (10000–20000 M−1 cm−1)39–42 for the CT band, thus generating a massive negative signal at ca. 530 nm that overwhelms the weaker radical absorptions. Therefore, only a small positive feature of the radical is exposed around 580 nm. For the same reason, in the 375– 460 nm region where SEC indicates the strong radical absorption (particularly for D•+), only a weak positive signal of radicals is exposed in the TA spectra. Thus, the observed TA spectrum represents a composite of radical absorption and ground-state bleaching. Nevertheless, the 580 nm band can serve as the critical diagnostic feature for the long-lived radicals, as it lies in the spectral window least obscured by the dominant GSB signals. Mechanistic Insights and implications of dual radical-generation pathways After identifying both 3CT and the radicals in the transient absorption spectra, we carefully examined the temporal evolution of the spectra. Prominently, under N2-saturated conditions, the decay of the 730 nm band corresponds to an increase in the signal intensity of the 580 nm band. Kinetic analysis (Fig. 3e and Supplementary Table S1) reveals that the decay time constant of the 730 nm band (τ ~ 7.8 μs) matches the growth time constant of the 580 nm band (τ ~ 7.6 μs). Furthermore, under air conditions, while the decay timescale of the 730 nm band (τ ~ 1.3 μs) still matches the growth timescale of the 580 nm band (τ ~ 1.5 μs), the magnitude of the increase in the 580 nm band is significantly weaker compared to N2-saturated conditions (Fig. 3f). These results indicate that the decay of 3CT is accompanied by the gradual formation of the radicals (D•+ and A•), implying an additional pathway for the radicals (D•+ and A•) generation from 3CT beyond the direct 1CT pathway. For the 3CT-mediated pathways of radical generation (D•+ and A•), the process is outlined as follows: the decay of 1CT generates 3CT via ISC. Unlike the singlet state, this 3CT species is spin-forbidden to undergo charge recombination, relaxing to the ground state, rendering it a long-lived reservoir of the radical ion pair. Within this state, the unstable A•- could eliminate Br- to produce A•. Notably, we observed that the conversion of 3CT into D•+ and A• occurs relatively slowly, on the microsecond timescale, possibly due to the higher thermodynamic stability of the 3CT state and the ensuing weaker driving force for the Br- elimination to generate radicals. Nevertheless, the fs-TA and ns-TA measurements clearly demonstrate the existence of two distinct radical generation pathways in this EDA system: one occurring directly via the decay of the 1CT state and the other mediated by the 3CT state (Fig. 5). Reaction scheme for the dual radical-generation pathways. The key to maximizing the efficiency of this system lies in the dual radical generation pathways involving both the 1CT and 3CT states. Square brackets denote that the electron acceptor and donor are confined within a solvent cage. Abbreviations: 1CT, singlet charge-transfer state; 3CT, triplet charge-transfer state; ISC, intersystem crossing. Most interestingly, similar fs-TA and ns-TA data are obtained for three additional representative types of EDA complexes (A2/D1, A3/D2, and A1/D2) containing bromides A1, A2, or A3 with enamines (Supplementary Figs. 7–9), generated via the condensation of the aminocatalyst with either saturated (butyraldehyde) or unsaturated (trans-2-Pentenal) aldehydes (Fig. 1b). These findings suggest that both the direct radical formation pathway from the 1CT state and the 3CT-mediated pathway are broadly applicable across these systems, thereby providing mechanistic evidence for the generality of the dual-pathway process in the EDA-based stereoselective photocatalytic a-alkylation of aldehydes. Expanding on these findings, the additional triplet-state-mediated pathway identified here might not be limited to this specific EDA-based photocatalytic stereoselective system. Instead, the photoinduced 1CT states that generate 3CT appear to be a common feature across a wide range of such systems, highlighting the broad applicability of these mechanistic insights35,43–49. The reactive radicals (D•+ and A•), generated via the dual-pathway process, subsequently participate in a coupling reaction, forging a C–C bond in the radical addition intermediate. This intermediate undergoes hydrolysis to regenerate the chiral secondary amine catalyst and afford the final α-alkylation product. Notably, our ns-TA results clearly demonstrate that molecular oxygen acts as a quencher of the 3CT state, significantly suppressing radical formation through this pathway (Fig. 3b, f and Supplementary Fig. 5). In contrast, under deaerated conditions, both 1CT and 3CT pathways are fully utilized, resulting in higher radical concentrations. These results are consistent with the reported decrease in reactivity under aerobic conditions. For example, for the EDA complex A2/D2, a 98% yield of the desired product was obtained after 6 h of irradiation under deaerated conditions, whereas only a 78% yield was achieved after 40 h of irradiation under air17. These findings underscore the importance of carefully controlling reaction conditions to optimize EDA-based photocatalytic reactions. Specifically, maintaining an oxygen-free environment is critical to maximize efficiency by enabling full utilization of both 1CT and 3CT pathways for radical generation. Besides, the intrinsic photophysical properties of the EDA complex, modulated by substituent modification, also play a role in determining photocatalytic efficiency. As illustrated in Fig. 1b, the A2/D1 complex, which features the narrowest HOMO–LUMO gap, exhibits the most pronounced bathochromic shift of the CT band, resulting in the largest overlap with the visible light spectrum. This enhanced spectral overlap is expected to improve visible-light photon harvesting and thus increase the catalytic reactivity. As evidenced by the synthetic experiments17, the system involving the A2-based EDA superior absorber achieved almost unity product yields, whereas only 70% conversion was obtained for the A1-based EDA system under identical irradiation time (9 h). These results underscore that tuning substituents to maximize visible light absorption is another important factor for optimizing photocatalytic efficiency. Estimation of selective Br- departure timescale of 1CT via fs-TA measurements Nucleophilic substitution reactions mediated by electron transfer represent a transformative strategy that overcomes the inherent limitations of traditional polar reactions, offering a powerful approach for inert bond activation and selective functional group transformations50–53. The design of leaving groups, particularly halogen-based moieties, remains a cornerstone in the development of such nucleophilic substitution processes36,50–53. For substrates bearing halogen leaving groups, the reaction typically involves electron capture by the halide acceptor to form a negatively charged radical anion, while C–halogen bond cleavage proceeds rapidly as a thermodynamically driven process36,50–53. Despite the importance of this process, the precise kinetics of halogen departure remain poorly understood, with rare studies reporting the departure rates of halogen leaving groups. Here, using fs-TA measurements, we estimated the timescale of Br- departure from the 1CT state. As observed in the fs-TA spectral evolution (Fig. 2e), the fastest decay (τ1 ~ 4 ps) of the tight conformation 1CT state manifests merely as a decrease in the intensity of ESA and the recovery of GSB without noticeable new species formation, indicating a dominant charge recombination (CR) to the ground state. This ultrafast CR likely arises from the most compact conformation, where the donor-acceptor proximity renders CR highly favorable, thereby effectively outcompeting debromination, making its contribution negligible during the τ1 phase. Conversely, in looser conformations with the middle (τ2 ∼ 100 ps) and especially the slowest (τ3 ∼ 1.5 ns) decay, the CR rate is slowed down, allowing debromination to effectively compete. This kinetic competition is well supported by the fs-TA difference spectra, where these slower decay phases (τ2 and τ3) are accompanied by the clear formation of the radicals (D•+/A•). Since the formation of D•+/A• radicals is inherently coupled with the departure of Br- from A•-, the departure of Br- can thus be estimated from τ2 and τ3, occurring on a sub-nanosecond to nanosecond timescale. Furthermore, the fs-TA difference spectra reveal that the formation of the radicals (D•+/A•) is concurrent with the emergence of 3CT. This simultaneous formation suggests that ISC and debromination act as parallel, competitive pathways originating from the same 1CT precursor (specifically, the loose conformation 1CT characterized by τ2 and τ3). Consequently, similar to the Br- departure process, the 1CT → 3CT ISC is estimated to occur alongside the τ2 and τ3 decay phases, spanning a timescale of sub-nanoseconds to nanoseconds. This estimated ISC timescale is in agreement with the 1CT → 3CT ISC timescales reported for other organic donor-acceptor complexes54,55. Although there are rare literature reports on the timescale of this debromination process, the rapid dissociative cleavage of Br- is further supported by our DFT calculations. As shown in Supplementary Fig. 10, when optimizing the structure of A•-, the resulting geometry reveals that Br-, carrying a negative charge, has already dissociated from the species. Note that while the DFT calculation predicts spontaneous dissociation of Br- from the isolated free radical A•-, the fs-experiments measure the Br- departure timescale from a contact radical ion pair of D•+…A•- (i.e., 1CT). The electrostatic attraction between D•+ and A•- in the EDA complex may slightly stabilize the C–Br bond, introducing a kinetic barrier that is absent in the free radical anion A•-. Therefore, the actual timescale of Br- departure is observed to be hundreds to thousands of picoseconds. Overall, these results indicate that A•- is inherently unstable and releases Br- rapidly. Critically, without this rapid departure of the leaving group (Br-), the primary pathway for photoinduced 1CT of EDA complexes would be dominated by unproductive charge recombination to the ground state. This inefficiency would preclude the swift generation of key radical species, thereby undermining the efficiency and stereoselectivity of product formation. In-cage radical coupling as the dominant pathway A variety of evidence supports the occurrence of radical coupling prior to the diffusive separation of the radical pair within the solvent cage. Paolo Melchiorre and co-workers conducted a series of control experiments to rule out the possibility of out-of-cage radical chain propagation as the operative pathway17. First, intermittent irradiation experiments showed that the reaction ceased completely in the absence of light and resumed upon re-illumination, confirming the indispensable role of light in driving the reaction. Second, a radical clock-containing aldehyde substrate revealed that the enamine radical cation directly participates in C–C bond formation, a process fundamentally incompatible with a radical-chain mechanism. Third, a FeBr2-mediated simulation conducted in the dark yielded radical byproducts and a low yield (26%), in stark contrast to the photo-driven reaction, which delivered a nearly unity product yield (98%) with high selectivity. These findings effectively exclude the viability of a sustainable radical-chain mechanism under the reaction conditions. Most compellingly, the proposed in-cage radical coupling mechanism is supported by the analysis of the product distribution17. If a significant fraction of D•+ and A• escaped the solvent cage, characteristic by-products of random diffusive encounters (e.g., homocoupling dimers or hydrogen abstraction derivatives) would be expected. However, the reaction yields the cross-coupling product with almost unity yields, whereas no homocoupling or H-abstraction by-products were detected. This clean product profile strongly corroborates that in-cage coupling is the dominant reaction pathway. Moreover, the bulky substitution group on D•+ likely impedes the cage escape, consistent with the favorable in-cage radical cross-coupling product yield. It is worth noting that the formation of the EDA complex is primarily driven by the nπ* interaction between D and A. However, after photoinduced charge transfer and subsequent Br- departure, the nπ* interaction is disrupted, creating the potential for D•+ and A• to escape the solvent cage. Nevertheless, the covalent C–C bonding of D•+ and A• (radical coupling) should be much more competitive than the cage escape, such that cross-coupling products with almost unity yields were observed17. This implies that although the cage-escaped radical signals are also observed in the ns-TA spectra sustaining millisecond timescale, they correspond to only a minor fraction. EDA Strategy as a blueprint for co-optimizing efficiency and stereoselectivity Traditional photocatalytic systems often struggle to simultaneously achieve high reaction efficiency and excellent stereoselectivity. Efficiency is limited by the competition between productive pathways (e.g., cage escape) and rapid charge recombination, while stereoselectivity is challenged by the stochastic nature of diffusion-controlled encounters between free radicals. Our findings reveal that the EDA strategy provides a mechanistic paradigm that circumvents these limitations, and this approach is not merely an alternative, but a superior design blueprint built upon three synergistic mechanistic features we have identified. Maximized photon utilization via a dual-pathway mechanism A primary bottleneck in photocatalysis is the loss of absorbed photon energy through unproductive decay channels. Our direct spectroscopic evidence for a dual radical-generation pathway: one proceeding directly from the ¹CT state and another mediated by a 3CT state, represents a fundamental mechanism for improving photon utilization efficiency. The triplet pathway acts as a salvage route, capturing population that would otherwise be lost from the singlet state and converting it into productive radicals. This mechanism we have elucidated, harnessing parallel, complementary excited-state pathways, represents a generalizable strategy for improving the photon-to-product conversion efficiency in any photocatalytic system where intersystem crossing is accessible, ensuring that a larger fraction of absorbed photons contributes to the desired chemical transformation. Kinetic gating via irreversible bond cleavage The most common failure point in photoinduced electron transfer is rapid back electron transfer (charge recombination), which returns the system to its ground state. The EDA system design brilliantly solves this problem through a kinetic gating mechanism. In the directly populated 1CT state, the subsequent departure of Br- occurs on a rapid timescale (hundreds to thousands of picoseconds), effectively outcompeting the allowed back electron transfer. Crucially, this efficiency extends to the 3CT manifold: although Br- elimination is slower within this state, the competing charge recombination to the ground state is spin-forbidden and thus significantly suppressed. Consequently, this irreversible chemical step of C-Br bond cleavage acts as a kinetic trap, fundamentally altering the potential energy surface and making the reverse physical process (charge recombination of the initial D•+…A•- pair) non-competitive. This kinetic gating mechanism we have characterized, coupling electron transfer to an irreversible bond cleavage to lock in a productive intermediate, constitutes a general design element for preventing energy-wasting recombination and can be applied to a wide range of transformations involving suitable leaving groups. Deterministic stereocontrol via in-cage pre-organization Achieving high stereoselectivity in radical reactions is notoriously difficult due to the transient and highly reactive nature of the intermediates. The EDA strategy transforms this challenge into an opportunity by translating ground-state pre-organization directly into a deterministic stereochemical outcome. The donor and acceptor are not random partners in solution; they are assembled into a well-defined EDA complex before photoexcitation. The rapid, in-cage generation of the radical pair (D•+ and A•) following Br- expulsion means they are born in close proximity and with a specific relative orientation inherited from the ground-state complex. This preserved geometry, as supported by our DFT calculations (Supplementary Fig. 12), funnels the reaction towards rapid in-cage coupling, precluding the need for diffusional reorientation and minimizing the formation of racemic or achiral byproducts from escaped radicals. This deterministic stereocontrol mechanism we have discovered, using non-covalent interactions to pre-organize reactants for stereochemically defined, in-cage reactions, offers a superior paradigm to traditional approaches that rely on probabilistic, diffusion-controlled encounters. Overall, by integrating these three mechanistic insights, the EDA strategy provides a holistic and highly controllable approach to photocatalysis. It establishes a clear blueprint for how to design molecular systems where photon utilization, product yield and stereoselectivity are not independent challenges to be addressed separately, but are synergistically and rationally co-optimized from the ground up. By employing fs-TA and ns-TA spectroscopy, we have elucidated the excited-state dynamics and reactive intermediates (1CT, 3CT, D•+) involved in the visible light-driven enantioselective α-alkylation of aldehydes, establishing a comprehensive mechanism for understanding this emerging photocatalytic approach. Upon visible light excitation, the EDA complex directly populates the 1CT state, namely 1[D•+…A•-], which undergoes either charge recombination or selective dissociative cleavage of the C–Br bond of A•- to A•. Concurrently, ISC of the 1CT state to the triplet state 3CT provides an additional pathway for the generation of reactive radical pairs (D•+ and A•). Importantly, the timescale of selective departure Br- is tightly coupled with the decay of the CT state. This irreversible Br- departure effectively prevents D•+ and A•- from undergoing the unproductive charge recombination, ensuring that the radicals (D•+ and A•) are efficiently generated, thereby facilitating the subsequent radical coupling, leading to the formation of the α-alkylation product with high stereoselectivity. Furthermore, theoretical calculations confirm that the spatial alignment of the initial EDA complex is inherently favorable for direct coupling, providing a rational explanation for the observed stereoselectivity. These results highlight the unique structural and mechanistic features of EDA complexes, including a dual-pathway for reactive radical generation, selective Br- departure, and the specific spatial orientation of the donor and acceptor, which collectively enhance photocatalytic efficiency and selectivity. These attributes confer a distinct advantage over traditional photocatalytic systems, which typically rely on radical escape and diffusion but are often hindered by charge recombination. Ultimately, this work shifts the paradigm from optimizing diffusion-based radical encounters to deterministically programming reaction outcomes through the rational design of ground-state molecular complexes. It provides a powerful and generalizable blueprint for the future of photocatalysis, where the precise control of excited-state dynamics can lead to optimized reaction pathways, ensuring high synthetic yield and stereoselectivity. Methods Materials (2 R)-2-[Bis[3,5-bis(trifluoromethyl)phenyl]-[(trimethylsilyl)oxy]methyl]pyrrolidine (chiral secondary amine catalyst, 97%, Apinno), butyraldehyde, trans-2-Pentenal (97%, Acros), phenacyl bromide (A1, 98%, Innochem), 2,4-dinitrobenzyl bromide (A2, 95%, Macklin) and 2-bromo-1-(naphthalene-2-yl) ethenone (A3, 98%, Innochem) were used as purchased. Methyl tert-butyl ether (MTBE, 99 + %, Extra Dry, Acros) was used as a solvent. Steady-state UV absorption spectra The steady-state UV absorption spectra were measured using a double-beam spectrometer (model U-3010, Hitachi). Each measurement was collected from 700 to 275 nm at a scan rate of 600 nm/min. The spectrum of a blank sample containing MTBE solvent was used as the background, which was subtracted from the averaged spectra. All measurements were performed in 1 cm quartz cuvettes at ambient temperature. Femtosecond transient absorption spectra The femtosecond transient absorption spectra were measured using a femtosecond transient absorption spectrometer (Harpia-TA, Light Conversion). The fundamental pulses (800 nm, 40 fs, 1 kHz) were generated with a Ti:Sapphire laser system (Coherent Astrella) and split by a beam splitter to generate the pump and probe beams. A small fraction of the fundamental pulses was sent to a computer-controlled optical delay line and then focused onto a CaF2 plate to generate the white light continuum probe beam. The rest of the fundamental pulses were sent to the optical parametric amplifier (TOPAS-C, Coherent Inc.) to generate the 520 nm pump beam, which was chopped at 500 Hz. To avoid anisotropic signals, the polarizations of pump and probe beams were set to the magic angle (54.7°) in all measurements. The focused signal and pump beams were overlapped into a fused silica cuvette with a 1 mm beam path length. The frequent movement of the electrically driven sample holder ensures that the sample in the probed volume was continuously renewed to avoid re-excitation of the excited volume by successive laser pulses, similar to the function of a magnetic stirrer. Moreover, considering the relatively low laser frequency of 1 kHz operated in our experiments, there is a negligible thermal effect, especially for solutions. The OD at 520 nm of the sample solution was maintained at ∼ 0.4. All experiments were carried out at room temperature. The differential absorbance ΔA (t, λ) obtained from the femtosecond transient absorption spectra as a function of wavelength and time delay was analyzed using the Glotaran program56. Time-resolved laser flash photolysis Nanosecond transient absorption spectra were measured using a flash photolysis setup, an Edinburgh LP980 spectrometer (Edinburgh Instruments Ltd.) combined with a Nd:YAG laser (Spectra-Physics Lab 170, Newport Corp.). Each measurement was performed in a 1 cm path length quartz cuvette at room temperature. The sample was excited by a 532 nm laser pulse (1 Hz, 10 mJ/pulse/cm2, fwhm ≈ 7 ns). The analyzing light came from a 150 W pulsed xenon lamp. A monochromator equipped with a photomultiplier to collect the spectral range from 300 to 800 nm was used to analyze the transient absorption spectra. The signals from the photomultiplier were displayed and recorded as a function of time on a 100 MHz (1.25 Gs/s sampling rate) oscilloscope (Tektronix, TDS 3012 C), and the data were transferred to a personal computer. The data were analyzed using the online software of the LP980 spectrophotometer. The fitting quality was judged by weighted residuals and the reduced χ2 value. Spectroelectrochemical measurements Spectroelectrochemical (SEC) measurements of D1 and A1 were performed in an acetonitrile (MeCN) solution containing 0.1 M tetrabutylammonium hexafluorophosphate (TBAPF6) as the supporting electrolyte. The experiments utilized a standard three-electrode configuration equipped with a Pt mesh working electrode, a Pt wire counter electrode, and an Ag/AgCl reference electrode. Cyclic voltammetry (CV) was initially conducted at a scan rate of 10 mV s−1 to determine the redox potentials by scanning A1 from 0 to – 1.8 V (reduction) and D1 from 0 to 1.9 V (oxidation). Subsequently, the donor enamine radical cation (D•+) was generated by applying a constant oxidative potential of + 1.0 V vs. Ag/AgCl, slightly beyond its single-electron oxidation potential. Conversely, the acceptor phenacyl radical (A•) was generated by holding the potential at a reductive value of − 1.0 V vs. Ag/AgCl. Theoretical calculations Geometry optimizations for all species were performed using Density Functional Theory (DFT) at the B3LYP/6–311 + + G(d,p) level of theory, incorporating Grimme’s D3 empirical dispersion correction. While dispersion effects were found to be negligible for isolated species lacking significant non-covalent interactions, this level of theory was applied uniformly to ensure methodological consistency with the electron donor-acceptor (EDA) complexes, where such interactions are critical. Solvation effects were included using the Polarizable Continuum Model (PCM), based on the solvent parameter of Diethyl ether as a proxy of methyl tert-butyl ether (MTBE) due to their structural similarities and comparable dielectric constants. All the calculations were carried out using the Gaussian 16 program package (Gaussian 16, Revision A. 03). Gauss View molecular modeling software was used to plot the molecular orbitals and draw the molecular structure. Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data Author contributions H.S. conceived the research. H.S. and J.J. supervised the project. M.F., Y.L., and J.J. performed the experiments and implemented the computation. M.F., Y.L., L.Q., N.Q., J.J., and H.S. discussed the results and analyzed the data. M.F., N.Q., J.J., and H.S. wrote the manuscript. 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. Funding This work was supported by the National Natural Science Foundation of China (No. 21933005 to H.S., 22473014 to J.J., and 21727803 to H. S. ), and the National Key R&D Program of China (No. 2022YFA1505400 to J.J.). Data availability The data generated in this study are provided in the Supplementary Information/Source Data files. Data supporting the findings of this manuscript are also available from the corresponding author upon request. Source data are provided in this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Jialong Jie, Email: jialong@bnu.edu.cn. Hongmei Su, Email: hongmei@bnu.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74293-5. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data Data Availability Statement The data generated in this study are provided in the Supplementary Information/Source Data files. Data supporting the findings of this manuscript are also available from the corresponding author upon request. Source data are provided in this paper.