Quasi-one-dimensional enaminone-linked covalent organic frameworks for efficient CO₂ photoreduction

Abstract

Quasi-one-dimensional covalent organic frameworks have emerged as promising platforms for photochemical energy conversion due to their low density of basal sites, abundant edge sites, and dual-chain-like structure. However, the development of quasi-one-dimensional covalent organic frameworks in photocatalysis is still highly hindered by their limited linkage chemistry. Herein, we report an enaminone-linked quasi-one-dimensional covalent organic framework. The polar enaminone bonds together with dual-chain-like structure endow enaminone-linked quasi-one-dimensional covalent organic framework with broad light adsorption and effective excitonic dissociation abilities. Significantly, a high CO yield of 3045 μmol g^-1 with approximately 100% selectivity was achieved in a 24 h reaction under gas-solid conditions. More interestingly, the hydrogen atom on nitrogen site in enaminone bond could assist in the activation of CO₂ molecule via hydrogen-bond interaction. This interaction leads to the strongest adsorption ability for CO₂ and the lowest energy barrier for the rate-determining step during CO₂ reduction over enaminone-linked quasi-one-dimensional covalent organic framework compared to those over mixture-linked quasi-one-dimensional covalent organic framework and imine-linked quasi-one-dimensional covalent organic framework counterparts. All of these factors directly contribute to the enhanced activity of enaminone-linked quasi-one-dimensional covalent organic framework in the photocatalytic CO₂ reduction to CO.

This study reports a new kind of En-Q1DCOF. The polar enaminone bonds and dual-chain-like structure endow En-Q1DCOF excellent photoelectric conversion performance. Furthermore, the N-site H atom in enaminone facilitates CO₂ activation via H-bonding.

Introduction

Achieving carbon peaking and carbon neutrality are regarded as the pivotal objectives for sustainable development¹. To realize this target, numerous strategies have been developed to address the issue of carbon dioxide (CO₂), especially to capture and convert CO₂ into high-value chemical feedstocks. In this regard, artificial photosynthesis, which is capable of mimicking natural leaves for the conversion of CO₂ with H₂O into carbohydrate and O₂, may provide a facile means to avoid the use of any sacrificial agents, precious metals cocatalysts and photosensitizers². As a class of porous materials, covalent organic frameworks (COFs) with the π-conjugation, especially two-dimensional (2D) or three-dimensional (3D) COFs^3,4 have exhibited immense potential in artificial photosynthesis thanks to their highly controllable skeleton, channels, local structures, advantageous photoelectric properties and structural stability during photochemical processes^5–10. However, their 2D or 3D orientation extended frameworks provide limited sites on the outside of pores (located at edges), which limit their photocatalytic activity¹¹.

As an alternative, Quasi-one-dimensional covalent organic frameworks (Q1DCOFs) are recently characterized by a low density of basal sites and a high proportion of edge sites on their frameworks, which significantly enhance the exposure of active sites and the photoelectrochemistry active surface areas¹¹. Besides, the distinct dual-chain architecture of Quasi-one-dimensional configurations could also promote the directional electron transfer, thereby minimizing energy dissipation associated with exciton dissociation^12–14. Thus, this structural attribute of Q1DCOFs endows their significant capability to enhance the directed transfer of electrons toward active sites, thereby improving photoelectric conversion efficiency and interfacial catalytic performance. Nevertheless, it is an irrefutable fact that the occurrence of regulation of linkers in Q1DCOFs is rare and that research on their photocatalytic applications is scarce. Currently, imine is the most extensively studied linkage in Q1DCOFs¹⁵. Unfortunately, the moderate stability of most imine-based COFs, particularly under prolonged light irradiation, significantly restricts the advancement of Q1DCOFs and impedes their application in artificial photosynthesis¹⁶. Therefore, exploiting connection modes is urgently important to develop 1D COFs toward structural and functional diversification, but remains a great challenge.

Enaminone, known as an intrinsic dipole, light-harvesting, and photo-switchable organic unit, has shown high potential in photocatalysis^17,18. Theoretical and experimental studies have shown that the integration of enaminone structural units into 2D COFs can greatly improve their photocatalytic performance by accelerating exciton dissociation and optical adsorption. For example, Jiang et al. discovered that enaminone-linked COFs possessed broader light absorption and easier exciton dissociation than imine-linked COFs and quinoline-linked COFs, directly contributing to the advantageous photocatalytic activity in hydrogen evolution reaction¹⁹. Besides, our group also demonstrated that enaminone-linked COFs were able to promote the directional transport of photogenerated electrons to reduction sites during CO₂ photoreduction, thereby generating a very high CO generation rate²⁰. Though these successful applications of enaminone linkage in 2D COFs have been achieved, no studies have been reported on investigation of artificial photosynthesis through integrating enaminone bonds into the Q1DCOF backbone to improve photocatalytic performance.

Herein, we present the design and synthesis of a Q1DCOF of En-Q1DCOF with the enaminone linkages. For comparative analysis, we also prepare Mix-Q1DCOF (with enaminone and imine linkages) and Im-Q1DCOF (solely with imine linkages) (Fig. 1a). The resulting Q1DCOFs exhibit high crystallinity and good physicochemical stability. Under metal-, photosensitizer-, or sacrificial reagent-free conditions, En-Q1DCOF achieves the highest CO yield of 3045 μmol g⁻¹ in 24 h, which is 7 and 12 times of Mix-Q1DCOF and Im-Q1DCOF, respectively (Fig. 1b). Further investigation indicates that En-Q1DCOF exhibits the highest number of active sites, advantageous visible light absorption and the fastest exciton dissociation. More interestingly, the hydrogen atom on nitrogen site in enaminone bond could assist in the activation of CO₂ molecule via hydrogen-bond interaction. This interaction leads to the strongest adsorption ability for CO₂ and the lowest energy barrier for the rate-determining step during CO₂ reduction over En-Q1DCOFs compared to those over Mix-Q1DCOF and Im-Q1DCOF.

Fig. 1. Design of Q1DCOFs for CO₂ photoreduction.

a Synthetic routes and (b) CO₂ photoreduction ability of the three Q1DCOFs, where Im-Q1DCOF, Mix-Q1DCOF, and En-Q1DCOF are linked by the imine, imine & enaminone, and enaminone bond, respectively (The dashed-line region indicates structural fragments of the three Q1DCOFs).

Results and Discussion

Synthesis and characterizations of COFs

Q1DCOFs of En-Q1DCOF, Mix-Q1DCOF, and Im-Q1DCOF were successfully prepared via Michael addition-elimination reaction or Schiff-base polymerization between Tetrakis(4-aminobiphenyl)ethylene (TAE) and 1,1′-(1,3-Phenylene)bis[3-(dimethylamino)−2-propen-1-one (PDP), 1-(3-Benzaldehyde)−3-(dimethylamino)−2-propen-1-one (BDP), and 1,3-Benzenedicarboxaldehyde (BD), respectively (Figure S1–S7). The crystal structures of these Q1DCOFs were theoretically and experimentally characterized by using Materials Studio simulation and powder X-ray diffraction (PXRD). As illustrated in Fig. 2a, b, En-Q1DCOF only presents an eclipsed stacking model (AA), corresponding to a typical 1D 4-c sql topology¹². The simulated pore diameter and interlayer spacing for En-Q1DCOF are 15.07 Å and 6.09 Å, respectively. The interlayer spacings for En-Q1DCOF are larger than normal COFs, which might be attributed to the combined effects of interlayer repulsion induced by the highly polar enaminone bond and the non-planar conformation of biphenyl in TAE. The PXRD pattern of En-Q1DCOF exhibits a main peak at around 4.8°, which is assigned to the (110) facet. Besides, the appearance of other diffraction peaks further indicates the high crystallinity of En-Q1DCOF. Pawley refinements of the PXRD patterns provide three unit-cell parameters of a = 29.4992 Å, b = 27.3954 Å, and c = 6.0908 Å, α = β = γ = 90°, with factors of R_p = 2.96% and R_wp = 3.38% for En-Q1DCOF (Table S1 and Supplementary Data 1). As for Mix-Q1DCOF and Im-Q1DCOF, the three unit cell is well fitted according to the Pawley refinements toward the experimental PXRD patterns, where the parameters for Mix-Q1DCOF and Im-Q1DCOF are a = 28.7365 Å, b = 24.7115 Å, c = 6.2289 Å, R_p = 2.72%, R_wp = 3.85% and a = 31.2157 Å, b = 21.5565 Å, c = 6.0374 Å, R_p = 6.32%, R_wp = 8.28%, respectively (Figures S8–S14, Tables S2, S3 and Supplementary Data 2, 3).

Fig. 2. Material structure characterization of Q1DCOFs.

a Experimental and Pawley refined PXRD of En-Q1DCOF crystal structure on the basis of the AA stacking models. b Top and side view of the AA stacking structure for En-Q1DCOF based on PXRD and modeling (Grey for C atoms, white for H atoms, red for O atoms and blue for N atoms). c FTIR spectra of En-Q1DCOF, PDP and TAE. d The ¹³C CP/MAS spectra of En-Q1DCOF. e High-resolution XPS spectra of N 1 s in Q1DCOFs. f N₂ adsorption-desorption isotherms of Q1DCOFs. g SEM image of En-Q1DCOF. h HRTEM image of En-Q1DCOF (FFT images are shown in the insets). i Elemental mapping images of En-Q1DCOF.

The chemical structures of the three Q1DCOFs were further characterized through Fourier transform infrared (FTIR), ¹³C cross-polarization/magic-angle spinning solid-state nuclear magnetic resonance (CP/MAS ssNMR) spectroscopy, and X-ray photoelectron spectroscopy (XPS). As shown in the FTIR spectroscopy, the disappearance of N-Me vibration peak at ~1430 cm⁻¹ and the generation of the N-C vibration peak at ~1175 cm⁻¹ signify the successful formation of the enaminone bond in En-Q1DCOF and Mix-Q1DCOF (Figs. 2c, S15 and S17)²¹. In the FTIR spectra of Im-Q1DCOF and Mix-Q1DCOF, the emergence of the peak at ~1625 cm⁻¹ attributing to the stretching vibration for -C = N- bond is accompanied by the disappearance of the -NH₂ vibration band at 3200–3500 cm⁻¹ for TAE monomers (Figures S15–S17)²². The successful synthesis of the different linkages and skeletons in these Q1DCOFs was further confirmed by the analysis of solid ¹³C CP/MAS spectra, where all carbon peaks in the benzene units of all the Q1DCOFs are well identified (Figs. 2d, S18 and S19). Furthermore, we also performed XPS measurement to verify the formation of Q1DCOFs (Figs. 2e and S20). Obviously, the N 1 s XPS peak in En-Q1DCOF shifted to a higher binding energy compared to that in Im-Q1DCOF, suggesting a stronger interaction between C and N atoms in enaminone bond than that in imine bond. As for Mix-Q1DCOF, the binding energies of the N 1 s XPS peaks are located between those of En-Q1DCOF and Im-Q1DCOF, which suggests the coexistence of enaminone and imine bonds in Mix-Q1DCOF.

The N₂ adsorption-desorption measurements at 77 K were performed to evaluate their porous structures. The corresponding Brunauer-Emmett-Teller (BET) surface areas of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF are 594, 204, and 487 m² g⁻¹, respectively (Fig. 2f). The much lower surface area of Mix-Q1DCOF might be related to its more complex connections and more disordered structures compared with those in En-Q1DCOF and Im-Q1DCOF. According to the pore size distribution curves, the corresponding micropore sizes for En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF are 1.55, 1.67, and 1.59 nm, respectively (Figure S21). The pore sizes obtained by N₂ adsorption isotherms are closely matched with the model-theoretical micropore size, further confirming the successful synthesis of the three Q1DCOFs.

The morphologies of the as-synthesized Q1DCOFs were investigated by the scanning electron microscope (SEM) and transmission electron microscope (TEM) characterizations (Figures S22–S29). All Q1DCOFs exhibit nanosheet morphology (20 nm–3 μm). Crystalline nature of En-Q1DCOF is confirmed by the fast Fourier transform (FFT) images showing distinct diffraction spots (Fig. 2h) and the high-resolution transmission electron microscope (HRTEM) revealing 0.41 nm lattice fringes corresponding to (040) planes. The energy-dispersive X-ray (EDX) elemental mapping further demonstrates a homogeneous distribution of all elements throughout the frameworks (Figs. 2i, S26–S29).

To further investigate the band structures of the as-acquired Q1DCOFs, we recorded their ultraviolet-visible diffuse reflectance spectra first. As seen in Fig. 3a, En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF behave a strong absorption in the visible region, which is mainly caused by the increase in along the plane conjugation and system polarization intensity. As for En-Q1DCOF, its intrinsic UV/vis absorption edge is about 460 nm. Besides, an obvious red shift of absorption edge for En-Q1DCOFs signifies its smaller band gaps (E_g) and enhanced light adsorption compared with those of Mix-Q1DCOF and Im-Q1DCOFs. It is noting that the lack of strongly conjugated and highly symmetrical units (such as pyrene and porphyrin) enables the Q1DCOFs to exhibit the characteristic of interlayer π-π stacking and less conjugated bonding modes (imine linkages and enaminone linkages). This character induces the folding and displacement of their energy bands in k-space, thus transforming them into indirect bandgap semiconductors⁹. Accordingly, the E_g values of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF are determined to be 2.42, 2.45, and 2.50 eV according to the transformed Kubelka-Munk function, respectively (Fig. 3b). Combining with the Mott-Schottky analysis (Figure S30) and VB-XPS measurements (Figure S31), the conduction band (CB) and valence band (VB) positions of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF were estimated. Obviously, all the COFs exhibit sufficient driving force to trigger CO₂ reduction and water oxidation reactions (Fig. 3c). Furthermore, we also carried out the density functional theory (DFT) calculations by using the VASPKIT code to Q1DCOFs²³. The spatial separation of CBM-VBM is more pronounced in En-Q1DCOF, which reduces the possibility of electron-hole pair recombination (Fig. 3d–f). The calculated band gaps of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF are 2.08, 2.16, and 2.47 eV, respectively, which is consistent with those obtained from experiment. As a result, the introduction of enaminone bonds could not only modulate the band gap but also induce the local charge delocalization and distribution.

Fig. 3. Material photophysical characterization of Q1DCOFs.

a UV/vis spectra of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF (corresponding photographs of the materials are shown in the inset). b The optical band gaps of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF determined by Tauc plot. c Schematic energy band structures of En-Q1DCOF, Mix-Q1DCOF and Im-Q1DCOF. Kohn-Sham orbitals of CBM and VBM of (d) En-Q1DCOF, (e) Mix-Q1DCOF and (f) Im-Q1DCOF (Brown for C atoms, white for H atoms, red for O atoms, light blue for N atoms, blue for positive diffuse and yellow for negative diffuse).

Photocatalytic CO₂ reduction reaction performances of COFs

The catalytic activity of the as-acquired Q1DCOFs was evaluated in the photocatalytic CO₂ reduction reaction (CO₂RR) with gaseous H₂O under visible light irradiation in the absence of photosensitizer, sacrificial agents, and cocatalysts (Fig. 4a). As shown in Fig. 4b, the yields of CO are 426 and 263 μmol g⁻¹ in 24 h when the reaction is catalyzed by Mix-Q1DCOF and Im-Q1DCOF, respectively. Interestingly, the highest CO yield of 3045 μmol g⁻¹ is achieved in 24 h over En-Q1DCOF, which is 7 and 12 times of Mix-Q1DCOF and Im-Q1DCOF, respectively. This performance of En-Q1DCOF is also competitive among the reported metal-free COF photocatalysts and photocatalytic systems up to date (Table S4). Besides, we also simultaneously detected the evolution rates of O₂ during CO₂RR (Figure S32). The ratio of CO to O₂ is about 2:1, suggesting that CO is produced by coupling with the oxidation of H₂O. To further confirm the origin of O₂, we also implemented the reduction of CO₂ by utilizing H₂¹⁸O as the water source. After the reaction, we analyzed the gases by gas chromatography and mass spectrometry, where a peak of ¹⁸O₂ with m/z = 36 was detected, confirming that O₂ was originated from the water (Figure S33). CO₂RR is significantly impeded in the absence of catalyst, CO₂, H₂O, or light illumination, demonstrating their indispensability for photocatalytic CO₂ reduction to CO (Fig. 4c). Moreover, ¹³CO₂ isotopic labeling experiment shows a dominant peak of ¹³CO (m/z = 29) in the gas chromatography-mass spectrometry (GC-MS) plot, indicating that CO is generated from the photoreduction of CO₂ instead of the organic species (Fig. 4d).

Fig. 4. Photocatalytic performance.

a Schematic diagram of gas-solid conditions photocatalytic CO₂ reduction reactor. b Time course of photocatalytic CO evolution for various photocatalysts (2 mg catalyst, λ > 420 nm). c Catalytic activity of En-Q1DCOF under various reaction conditions. d Mass spectrum for photocatalytic reduction of ¹³CO₂ to ¹³CO driven by En-Q1DCOF. e Wavelength-dependent AQY of photocatalytic CO evolution for En-Q1DCOF. f Time courses of the photocatalytic activity and durability over Q1DCOFs. Error bars in (c, e) represent the standard deviation from three independent measurements.

In addition, we also investigated the effect of the amount of catalyst on the performance of photocatalytic CO₂RR (Figure S34). The slight decrease in the CO evolution rate when increasing the amount of catalyst might be caused by light not penetrating the thicker catalyst layer. To evaluate the light utilization efficiency, the apparent quantum yield (AQY) of En-Q1DCOF was measured under the irradiation of monochromatic light (Fig. 4e). En-Q1DCOF exhibited AQY values of 0.66% at 450 nm and 0.14% at 650 nm. Optimizing the photocatalyst loading to 20 mg enhanced its AQY to 1.94% at 450 nm (Figure S35). However, further mass increases failed to produce linear AQY improvements due to light shielding effects.

Recycling tests of the En-Q1DCOF catalyst showed no significant attenuation after three 24-hour photocatalytic cycles (Fig. 4f), confirming its high stability. The sample recovered after catalytic tests was named as En-Q1DCOF-irradiation. In addition, we also analyzed the FTIR spectra of En-Q1DCOF before and after photocatalytic reaction (Figure S36), where not any change in the intensity or wavenumber of the C = O group at 1632 cm⁻¹ was observed. This result further confirms that the product of CO does not originate from the C = O groups within En-Q1DCOF. Besides, we also collected the PXRD patterns of En-Q1DCOF before and after photocatalytic reaction. As shown in Figure S35, the XRD patterns of En-Q1DCOF before and after the reaction are almost identical, demonstrating that En-Q1DCOF did not undergo structural collapse. In other words, all the bonds in En-Q1DCOF including C = O bond are stable. All the above-mentioned results convincingly demonstrate that the CO is generated from the reduction of CO₂, not the degradation of the materials. In addition, the high chemical stability of En-Q1DCOF was also demonstrated by treating it with the solution of 1 M HCl, 1 M NaOH, and 1 M DMF, respectively, where the resulting COFs retained the crystallinity and skeletal structure of En-Q1DCOF (Figure S37).

Photophysical and photochemical properties study of COFs

To reveal the reason for the difference in the photoelectric performance of the enaminone and imine bonds in COFs, we performed DFT calculations. Figure 5a–d present the density of states (DOS) for Q1DCOFs. Prominent DOS peaks at the VB and CB edges reflect their molecular origins. Notably in En-Q1DCOF (Fig. 5b), the CB edge is dominated by the EN structure, while the VB edge originates primarily from TBE. Furthermore, EN in En-Q1DCOF contributes more significantly to the CB than IM does in Mix-Q1DCOF and Im-Q1DCOF (Fig. 5c, d), indicating the more electron-deficient nature of enaminone bonds compared to imine bonds, which favors band gap narrowing²⁴.

Fig. 5. DFT calculations study of photophysical properties.

a Calculated density of states for molecular fragments of QIDCOFs. Calculated density of states for (b) En-Q1DCOF, (c) Im-Q1DCOF and (d) Mix-Q1DCOF. En-COF and Im-COF in the ground state (e); blue and red represented electron accumulation and depletion, respectively. Localized orbital locator of (f) En-COF and (g) Im-COF (The dashed regions denote the N-site region of Im-COF and the O-site region of En-COF, respectively). Charge distribution and centroids diagram for (h) En-COF and (i) Im-COF in the excited state (hole is blue region and electron is green region).

Figure S38 shows the optimized COFs molecular fragment structures. Ground state electrostatic potential (ESP) analysis of COFs (Fig. 5e) shows that the enaminone bond has greater polarity compared to the imine bond. This facilitates the local polarization of the π-system, generating a strongly polarized electric field. Such a strongly polarized electric field promotes the delocalization of electrons within the system (dipole moments, En-COF:3.4769 Debye, Im-COF:2.2511 Debye). In addition, the maximum negative charge distribution of COF is located on the ketone group and imine group, indicating that both of these chemical bonds can effectively adsorb CO₂. Interestingly, the charge of O-site on the ketone group is more negative than that of the N-site on the imine group, which proves that the ketone group prefers to capture CO₂ (Figure S39). In addition, the localized orbital locator (LOL) indicates that En-COF has weaker localization in the O-site region compared to Im-COF in the N-site region, proving that En-COF is more conducive to the extraction and output of hot π electrons (Figs. 5f, g). The electron-hole distributions of the model COFs molecular fragment constructed by Multiwfn²⁵ and Visual Molecular Dynamics (VMD)²⁶ disclose the role of enaminone and imine bonds in the electronic structures. The calculated charge transfer distances are 2.161 Å and 3.173 Å for Im-COF and En-COF, respectively, indicating the remarkably improved electron-hole separation after enaminone polarization (Fig. 5h, i).

To further reveal the essence for the high CO₂RR performance of En-Q1DCOF, several photoelectrochemical measurements were conducted, including electrochemical impedance spectra (EIS), photocurrent profiles, and polarization curves. As expected, En-Q1DCOF demonstrates the highest photocurrent response and the smallest semicircle diameter in the EIS plots, followed by Mix-Q1DCOF and Im-Q1DCOF (Figs. 6a and S40). This result suggests the effective electrical conductivity at the interface and high photoreduction capacity of En-Q1DCOF, which is consistent with the result of CO₂RR. Notably, the polarization of the π-system could also decrease the overpotential of CO evolution in the Q1DCOF-based system, as evidenced by the polarization curves in Figure S41. Furthermore, electron paramagnetic resonance (EPR) analysis was conducted to evaluate the behavior of photoexcited electrons. As shown in Fig. 6b, En-Q1DCOF exhibits a stronger DMPO-•O^2- signal under visible light irradiation, indicating its greater photoexcited electron reduction ability compared with Mix-Q1DCOF and Im-Q1DCOF. No obvious DMPO-•O^2– signals are observed under dark conditions, further supporting the nature of the photoinduced activity for En-Q1DCOF (Figure S42).

Fig. 6. Charge separation dynamics in Q1DCOFs.

a Photocurrent profiles of Q1DCOFs. b Electron paramagnetic resonance spectra of DMPO-•O^2– of the Q1DCOFs under light. c PL spectra of En-Q1DCOF, Mix-Q1DCOF, and Im-Q1DCOF. Temperature-dependent PL spectra and corresponding fitting curves (inset) of (d) En-Q1DCOF, (e) Mix-Q1DCOF and (f) Im-Q1DCOF. Transient absorption spectra (inset and excitation at 400 nm) and the decay kinetics of (g) En-Q1DCOF, (h) Mix-Q1DCOF and (i) Im-Q1DCOF.

Figure 6c shows the photoluminescence (PL) spectra of these Q1DCOF. Obviously, En-Q1DCOF exhibits a lower photoluminescence (PL) peak intensity than Mix-Q1DCOF and Im-Q1DCOF, indicating that enaminone bonds effectively contribute to the separation and transfer of photogenerated carriers. Furthermore, temperature-dependent PL measurements were also conducted to investigate the exciton dissociation kinetics, which is one of the essential factors affecting the photocatalytic activity of COFs photocatalysts (Fig. 6d–f). The integrated photoluminescence intensities increase gradually with decreasing temperature from 300 K to 85 K. Besides, the corresponding exciton binding energies (E_b) can be calculated based on the Arrhenius equation²⁷. Obviously, the E_b values decrease rapidly with the increase of enaminone bonds, suggesting that the introduction of enaminone bonds in Q1DCOFs could effectively suppress the exciton effect.

Femtosecond time-resolved transient absorption spectra (fs-TAS) probed excited-state kinetics in Q1DCOF photocatalysts. Spectra recorded at 400 nm excitation (Figs. 6g–i) revealed negative signals (480–650 nm) across all time delays. The negative signal can be assigned to the ground state bleaching (GSB) and stimulated emission (SE) according to UV-vis and steady state PL spectra. Notably, En-Q1DCOF exhibited accelerated SE decay relative to Im-Q1DCOF and Mix-Q1DCOF, aligning with the observed exciton effects²⁸. The kinetics fitting results show biexponential decay processes, namely the short lifetime (τ₁) and long lifetime (τ₂), corresponding to electron trapping and electron transfer kinetics, respectively. As shown in Fig. 6g, i, the decay time constants for Im-Q1DCOF are τ₁ = 0.31 ps and τ₂ = 25.4 ps. As for En-Q1DCOF, the decay time constants of τ₁ and τ₂ decrease to 0.05 and 6.8 ps, respectively. This decrease in τ₁ and τ₂ indicates different interfacial charge transfer efficiency in Q1DCOF^28,29. A similar trend was observed in Mix-Q1DCOF (Fig. 6h). These observations from fs-TAS provide compelling evidence that enaminone bonds can effectively regulate the efficiency of interfacial charge transfer.

As the adsorption and activation of CO₂ on the catalyst surface is a prerequisite for its further reduction, we conducted in situ attenuated total reflection Fourier transform infrared (ATR FTIR) to investigate the adsorption of CO₂ by exposing En-Q1DCOF to dry CO₂ under dark. Interestingly, a new broad band centered at 3630 cm^-1 attributing to the H-O bending of N-H-O = C is observed (Figure S43)^30–32. In addition, the hydrogen atoms on the nitrogen atoms in the enaminone unit have higher positive charges, indicating their stronger electrostatic attraction than the hydrogen atoms on other carbon atoms, which is more conducive to forming hydrogen bonds with CO₂ (Fig. 7a). This formation of the hydrogen bond could weaken the O = C bond in CO₂ molecule^33,34. Accordingly, the hydrogen atom on nitrogen atom of the enaminone unit could assist in the adsorption of CO₂ molecule. To further directly visualize the process of CO₂ photoreduction and water photooxidation over En-Q1DCOF, we further measured the in situ ATR FTIR spectroscopy under the simulated reaction conditions, where En-Q1DCOF was exposed to saturated H₂O vapor of CO₂ under light irradiation. As shown in Fig. 7b, several peaks gradually appeared between 1100 and 2000 cm⁻¹ with the irradiation time from 0 to 60 min. A new broad band centered at 1634 cm⁻¹ is attributed to the H-O-H bending of water molecules³⁵. Due to the presence of water molecules in the reaction process, the adsorbed CO₂ molecules can be converted into various intermediate substances, such as bidentate carbonate (b-CO₃²⁻) at 1596 and 1208 cm⁻¹, and monodentate carbonate (m-CO₃²⁻) at 1482 cm^−1 36. The infrared peaks at 1288 cm⁻¹ is attributed to the symmetrical bending of the intermediate *COOH³⁷. Additionally, the *COOH species, a crucial intermediate in reducing CO₂ to CO, gradually increased during the photocatalytic process. Besides, we also characterized the intermediates in the H₂O oxidation reaction by using in situ ATR FTIR. As depicted in Fig. 7c, the broad band centered at 950 cm⁻¹ is attributed to *OH, a crucial intermediate in the photooxidation of water to O₂. The peaks of *OOH intermediates at 1051 cm⁻¹, are also gradually enhanced during the photocatalytic process. Thus, a possible photocatalytic water oxidation process for En-Q1DCOF is proposed as follows: H₂O → *OH→ *OOH → *O₂ → O₂³⁸.

Fig. 7. Study of artificial photosynthesis mechanism in Q1DCOFs.

a En-COF hydrogen atom charge distribution. In situ ATR FTIR spectra of (b) CO₂ and (c) H₂O interactions with En-Q1DCOF under subsequent light irradiation. d Schematic illustration of the mechanism of En-Q1DCOF photoinduced electron transfer (PET) for CO₂RR coupled with H₂O oxidation (Blue and red circles indicate the regions of holes and electrons, respectively). e Gibbs free energy at different CO₂ adsorption sites of Q1DCOFs. f DFT-calculated Gibbs free energy profiles for CO₂ photoreduction over En-Q1DCOF and Im-Q1DCOF. g DFT-calculated Gibbs free energy profiles for H₂O oxidation reaction on En-Q1DCOF.

On the basis of the above experiments and analysis, a reasonable mechanism was proposed for CO₂ reduction coupled with H₂O oxidation process over En-Q1DCOF (Fig. 7d), which was further supported by DFT calculations. As shown in Fig. 7e, the corresponding CO₂ adsorption energy for sites 1–3 was calculated. The results show that the enaminone bond is more favorable for CO₂ adsorption than the imine bond. Interestingly, the adsorption energy could be further reduced when hydrogen atom in the enaminone unit is involved in the adsorption of CO₂, suggesting that the hydrogen atom on nitrogen site could assist in the activation of CO₂ molecule. This result is consistent with that of in situ ATR FTIR. Considering the reaction was conducted under the atmosphere of water vapor, the H₂O-mediated solvation effect was weak and transient, thus exerting minimal influence on CO₂ adsorption/activation. Furthermore, the experimental conditions for CO₂ photoreduction were identical for both enaminone COFs and imine COFs, while the enaminone COFs demonstrated advantageous performance. This result could further exclude the H₂O-mediated solvation effect on CO₂ reduction. Besides, we also calculated the corresponding O₂ adsorption energy for 1,1,2,2-tetra(biphenyl-4-yl)ethene sites 4-6. As shown in Figure S44, the vinyl is more favorable for O₂ adsorption than the benzene motif. Combining with in situ ATR FTIR analysis, we finally calculated the Gibbs free energy for the reaction processes of CO₂ reduction and water oxidation over En-Q1DCOF and Im-Q1DCOF (Fig. 7f, g). Accordingly, the rate-determining step for the reduction of CO₂ to CO is determined to be the transformation from *CO₂ to *COOH. The energy barrier for the generation of *COOH on En-Q1DCOF is 2.33 eV, which is lower than that of Im-Q1DCOF of 2.54 eV. The H₂O oxidation process on the 1,1,2,2-tetra(biphenyl-4-yl)ethene fragment can be divided into four steps: H₂O → *OH→ *OOH → *O₂ → O₂³⁸.

Discussion

In summary, a Q1DCOF of En-Q1DCOF featuring a characteristic 4-c sql topology by adopting a connection mode of enaminone is designed and synthesized for CO₂ photoreduction under visible-light irradiation. The En-Q1DCOF exhibit high crystallinity and effective physic-chemical stability. During photocatalytic CO₂ reduction, En-Q1DCOF achieves the highest CO yield of 3045 μmol g⁻¹ in 24 h, which is much higher than those analogous COFs with partial imine linkages (Mix-Q1DCOF, 426 μmol g⁻¹) and full imine linkages (Im-Q1DCOF, 263 μmol g⁻¹). Further investigation indicates that the enaminone bond is more polar compared to the imine bond, favoring π-system polarization and promoting exciton dissociation. Thus, En-Q1DCOF displays advantageous visible light absorption and lower exciton binding energy capability. More interestingly, the hydrogen atom on nitrogen site in enaminone bond could assist in the activation of CO₂ molecule via hydrogen-bond interaction. Such interaction leads to the strongest adsorption ability for CO₂ and the lowest energy barrier for the rate-determining step during CO₂ reduction over En-Q1DCOF compared to those over Mix-Q1DCOF and Im-Q1DCOF. This work represents the initial instance of Q1DCOFs via a connection mode of enaminone for artificial photosynthesis. We believe our research not only enriches the data base of Q1DCOFs but also serves as a valuable reference for the design of Q1DCOFs for various applications.

Methods

Materials

All chemicals and reagents were of analytical grade and used as received without further purification. The following materials were sourced as indicated: BD (1,3-Benzenedicarboxaldehyde, 98%), TAE (Tetrakis(4-aminobiphenyl)ethylene, 98%), 3-acetylbenzaldehyde (98%), and 1,3-diacetylbenzene (99%) from Shanghai Haohong Biomedical Technology Co.; O-DCB (1,2-Dichlorobenzene, 99%), anhydrous n-But (n-Butanol, 99.4%), acetic acid (>99.0%), and N,N’-dimethylformamide dimethyl acetal (>99.0%) from Macklin Chemicals; tetrahydrofuran (THF, 99.5%) and acetone (ACE) from Tianjin Komeo Chemical Reagent Co.

Synthesis of 1,1′-(1,3-phenylene)bis[3-(dimethylamino)−2-propen-1-one (PDP)

In a typical synthesis, 1.0 g of 1,3-diacetylbenzene was dissolved into 20 mL of N,N’-dimethylformamide dimethyl acetal. The reaction mixture was heated to 90 °C and stirred under nitrogen for 12 h. After evaporating the solvent and the subsequent addition of Et₂O, the crude product was obtained by filtration. The crystals of PDP were acquired after carefully washing with Et₂O and pentane as well as the drying under vacuum. ¹H NMR (400 MHz, CDCl₃): 8.40 (s, H), 8.00 (q, 2H), 7.83 (d, 2H), 7.46 (t, H), 5.78 (d, 2H), 3.04 (d, 12H) (Figure S1). ¹³C NMR (100 MHz, CDCl₃): δ 188.26, 154.49, 140.39, 129.98, 128.03, 126.50, 92.32, 45.09, 37.41 (Figure S2).

Synthesis of 1-(3-benzaldehyde)−3-(dimethylamino)−2-propen-1-one (BDP)

In a typical synthesis, 1.0 g of 3-acetylbenzaldehyde was dissolved into 20 mL of N,N’-dimethylformamide dimethyl acetal. The reaction mixture was heated to 90 °C and stirred under nitrogen for 12 h. Filtration removed insoluble impurities precipitated by the reaction. After evaporating the solvent and the subsequent addition of Et₂O, the crude product was obtained by cooling filtration at −78 °C. The oily liquid of BDP was acquired after carefully washing with Et₂O and pentane, and then removing the low-boiling-point solvents under vacuum conditions. ¹H NMR (400 MHz, CDCl₃): 10.09 (s, H), 8.38 (s, H), 8.20 (d, H), 7.97 (d, H), 7.87 (d, H), 5.76 (d, H), 3.19 (s, 3H), 2.97 (s, 3H). (Figure S3). ¹³C NMR (100 MHz, CDCl₃): δ 192.23, 154.86, 141.33, 136.25, 133.45, 131.67, 129.01, 128.77, 121.80, 91.60, 66.56, 45.24. (Figure S4).

Synthesis of En-Q1DCOF

A typical synthesis involved charging a Pyrex tube with PDP (19.8 mg, 0.07 mmol), TAE (24.5 mg, 0.035 mmol), 1,2-dichlorobenzene (0.5 mL), and 1-butanol (0.5 mL). After 5 min sonication, 0.1 mL of 6 M aqueous acetic acid was added under shaking, followed by another 5 min sonication. Subsequently, the mixture undergoes flash freezing with liquid nitrogen and is then sealed under vacuum. The reaction mixture was heated to 120 °C and held undisturbed for 72 h. The resulting precipitate was collected by filtration, washed with THF and ACE, and dried at 120 °C to yield the product as a yellow powder.

Synthesis of Mix-Q1DCOF

A typical synthesis involved charging a Pyrex tube with BDP (15.2 mg, 0.07 mmol), TAE (24.5 mg, 0.035 mmol), 1,2-dichlorobenzene (0.5 mL), and 1-butanol (0.5 mL). After 5 min sonication, 0.1 mL of 6 M aqueous acetic acid was added under shaking, followed by another 5 min sonication. Subsequently, the mixture undergoes flash freezing with liquid nitrogen and is then sealed under vacuum. The reaction mixture was heated to 120 °C and held undisturbed for 72 h. The resulting precipitate was collected by filtration, washed with THF and ACE, and dried at 120 °C to yield the product as a yellow powder.

Synthesis of Im-Q1DCOF

A typical synthesis involved charging a Pyrex tube with BD (9.4 mg, 0.07 mmol), TAE (24.5 mg, 0.035 mmol), 1,2-dichlorobenzene (0.5 mL), and 1-butanol (0.5 mL). After 5 min sonication, 0.1 mL of 6 M aqueous acetic acid was added under shaking, followed by another 5 min sonication. Subsequently, the mixture undergoes flash freezing with liquid nitrogen and is then sealed under vacuum. The reaction mixture was heated to 120 °C and held undisturbed for 72 h. The resulting precipitate was collected by filtration, washed with THF and ACE, and dried at 120 °C to yield the product as a yellow powder.

Characterizations

Powder X-ray diffraction (PXRD) patterns were acquired on a X’pert Powder diffractometer in reflection geometry using Cu Kα radiation (λ = 1.5 Å). Data collection spanned 2θ = 1-40° and 6°/min. Solid-state ¹³C NMR spectra, acquired via cross polarization magic-angle spinning (CP/MAS), were obtained on a JEOL-NM-ECZL G 600 MHz spectrometer. FT-IR spectra were acquired with a Thermo Fisher Nicolet Avatar 6700 spectrometer. UVDRS spectra were acquired by U-3010UV-V1S in the range of 200-1200 nm. The Brunauer-Emmett-Teller (BET) surface area was measured on an instrument of Micromertics ASAP 2020 M. The tube voltage and current used were 40 kV and 40 mA, respectively. XPS measurements were conducted on K-alpha (Binding energies were calibrated by the peak at 284.8 eV for C). TEM and the corresponding HRTEM were conducted on JEM2100 with an acceleration voltage of 200 kV. SEM images were obtained by using Hitachi SU8200 with an acceleration voltage of 10 kV. ESR measurements were conducted at 293 K using a Bruker EMX CW micro spectrometer operating in the X-band (microwave frequency ≈ 9.87 GHz). The instrument was equipped with an ER 4119HS-WI high-sensitivity optical resonator featuring a front-side grid. Samples were illuminated by a 300 W Xe lamp fitted with a 420 nm cut-off filter (LOT Oriel). All measurements used identical parameters: microwave power 6.74 mW, receiver gain 2 × 10⁴, modulation frequency 100 kHz, modulation amplitude 3 G, and sweep time 45 s. g-values were calculated from the resonance field (B₀) and frequency (ν) using the resonance condition hν = gβB₀. Calibration employed DPPH (2, 2-diphenyl-1-picrylhydrazyl) as a standard (g = 2.0036 ± 0.00004). The working electrode was fabricated by dispersing 5 mg of photocatalyst powder into 4 mL of ethanol containing 20 μL of Nafion, followed by 1 h of sonication. Subsequently, 0.5 mL of the resulting suspension was drop-cast onto a 1.9 cm × 3.3 cm ITO glass substrate, with an additional drop applied after the ethanol from the first application evaporated. Finally, the coated substrate was calcined at 150 °C for 1 h in a tube furnace under N₂ atmosphere, yielding the working electrode. Photoelectrochemical measurements were performed using a CHI660E electrochemical workstation configured with a standard three-electrode system: a catalyst-coated ITO working electrode, a platinum plate counter electrode, and an Ag/AgCl reference electrode. Illumination was provided by a 300 W Xe arc lamp equipped with a cut-off filter (λ > 420 nm), and 0.1 M Na₂SO₄(aq) served as the electrolyte. The Mott-Schottky plots were harvested at the frequencies of 1000, 1500 and 2000 Hz. Linear sweep voltammetry with a scan rate of 5 mV/s was carried out in a 0.1 M Na₂SO₄ aqueous solution. The ATR infrared spectrometric measurements were conducted by a Bruker INVENIO S spectrometer (Germany) with a KBr detector in the system. In a typical procedure, the COF membrane was placed in the chamber before sealing and then purging with N₂ gas for 30 min. Typical signals of several intermediates were captured after the introduction of a flowing dry CO₂ (g) or CO₂-H₂O (g) mixture under dark for 30 min and then visible light irradiation for 0–60 min. During the experiment of fs-TAS, transient absorption spectroscopy was recorded at the excitation wavelength of 400 nm. Before the measurement, 2 mg of the as-prepared COF was ultrasonically dispersed into 4 mL of glycol or H₂O to acquire the dispersion of COF with a concentration of 0.5 mg/mL. Transient absorption spectroscopy was performed using a Helios pump-probe system (Ultrafast Systems LLC) with an amplified femtosecond laser (Coherent; 35 fs, 1 kHz, 800 nm). Probe pulses (430-603 nm) were generated via focusing 10 μJ of the 800 nm output into a 1 mm CaF₂ crystal. Pump pulses (365 nm) originated from a TOPAS-800-fs optical parametric amplifier. The test container was in a 2 mm optical path colorimetric dish. The test was performed for at least three times. The result was the automatic average value of the equipment. Steady-state PL and temperature-resolved PL spectra were acquired by using Edinburgh Instruments, FLS980 spectrometer. The ¹³CO gas produced from ¹³CO₂ isotope experiments was examined by a gas chromatograph-mass spectrometer (GC-MS, Agilent 7890B-5977B).

Photocatalytic CO₂ reduction method and parameters

Photocatalytic CO₂ reduction was conducted in a 250 mL reactor under visible light (>420 nm, 100 mW cm⁻²) at 35 °C with stirring. Catalyst films were prepared by ultrasonically dispersing 2 mg catalyst in 5 mL H₂O (15 min), immobilizing on a 40 mm glass fiber membrane, and vacuum-activating (120 °C, 12 h). For reactions, catalyst films and 3 mL H₂O were loaded into a custom quartz vessel. A 300 W Xe lamp (PLS-SXE300C, Perfectlight) with 420 nm cutoff filter provided illumination. Pre-treatment included reactor evacuation, triple CO₂ purging, and 15 min CO₂ bubbling. Gaseous products were periodically sampled (1 mL) and analyzed by GC (FID: CO/CH₄; TCD: O₂/H₂). The apparent quantum yield (AQY) was measured under the same condition (λ = 420, 450, 500, 550, 600 and 650 nm, irradiation area: 0.785 cm², irradiation time: 1 h), and the following equation was used to calculate AQY:

$$AQYs\left(\%\right)=\frac{2M{N}_{A}hc}{SPt\lambda }\times 100\%$$ 1

Where M is the amount of CO produced (mol), N_A is the Avogadro constant (6.022 × 10²³/mol), ℎ is the Planck constant (6.626 × 10⁻³⁴ J·s), c is the speed of light (3 × 10⁸ m/s), S is the irradiation area of the incident light (m²), P is the intensity of the incident light (W/m²), t is the photoreaction time (h), and λ is the wavelength of the monochromatic light (nm). The computational details for AQY are summarized in Figure S45.

Calculations details and structure simulations

Molecular geometry optimizations and electronic structure characterizations were performed at the B3LYP/6-31 G level using Gaussian 09. TD-DFT calculations (TD-B3LYP/6-31 G, scrf) provided excited state energies and oscillator strengths for transient species, incorporating an implicit solvent model and dispersion correction. Multiwfn (version 3.8) was employed to analyze electron-hole distributions during excitation, electrostatic surface potentials (ESP), and localized orbital locators (LOL)²⁵. Multiwfn and Visual Molecular Dynamics (VMD) generated the electron-hole distributions (Fig. 5h, i), electrostatic surface potential (ESP, Fig. 5e), and localized orbital locator (LOL, Fig. 5f, g)²⁶. Structural modeling of COFs utilized the Materials Visualizer module in Materials Studio. First-principles calculations employed the Vienna ab-initio simulation package (VASP)³⁹. The Projector Augmented Wave (PAW) method described the electron-ion interactions⁴⁰. Exchange-correlation effects were treated with the Perdew-Burke-Ernzerhof (PBE) functional under the Generalized Gradient Approximation (GGA), including DFT-D3 correction for van der Waals interactions. Brillouin zone integration used A1 × 1 × 1 Gamma-centered k-point mesh⁴¹. Relaxation convergence criteria comprised: electronic self-consistent cycle energy tolerance of 10⁻⁶ eV, maximum atomic force below 0.02 eV/Å, and a plane wave cutoff energy of 520 eV. Post-processing of CBM-VBM and DOS data utilized VASPKIT²³. While DFT simulations provide valuable insights into catalytic mechanisms, all conclusions should be cautiously interpreted due to approximations in modeling real catalytic environments.

Author contributions

J.-X.B., T.-T.H., and L.-Y.W. conceived and designed the research. J.-X.B. performed the synthesis and conducted the experiments, J.-X.B, G.-J.L and Y.-X.H performed the characterizations. J.-X.B. conducted the theoretical calculations. F.-Y.S and Y.L supervised the project and discussed the experiments. J.-X.B., T.-T.H., and L.-Y.W. participated in drafting the paper and gave approval to the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Yuan-Biao Huang and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The source data generated in this study are provided in the Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69361-9.