Yuhang Qia,
Yiqiang Heb,
Yuxin Liua,
Zhe Zhanga,
Chunguang Lia,
Fanchao Mengc,
Shiyu Wangc,
Xiaobo Chen
d,
Zhan Shi
*a and
Shouhua Feng
a
aState Key Laboratory of Inorganic Synthesis and Preparative Chemistry, College of Chemistry, Jilin University, Changchun 130012, China. E-mail: zshi@mail.jlu.edu.cn
bCollege of Chemistry and Chemical Engineering, Jiangxi Normal University, Nanchang 330022, China
cKey Laboratory of High Performance Plastics, Ministry of Education, College of Chemistry, Jilin University, Changchun 130012, China
dSchool of Engineering, RMIT University, Carlton, VIC 3053, Australia
First published on 28th May 2025
Whilst defect engineering is a sound approach to enhance CO2 photoreduction based on metal organic frameworks (MOFs), the underlying mechanisms were not well understood on an atomic scale. This study aims to elucidate the mechanisms at the atomic level, to provide vital insights to enable the design and development of selectively introduced ligand defects to maximize the CO2 photoreduction capability of a classical MOF UiO-66–NH2 without the need for co-catalysts, sacrificial agents and photosensitizers. Defect-containing UiO-66–NH2 (Zr/Ce0.25) demonstrates superior charge separation and CO2 photoreduction than both regular UiO-66–NH2 (Zr/Ce0.25) and UiO-66–NH2. The doped Ce is a key contributor to managing the coordination environment of the linkers, enabling the formation of selectively introduced ligand defects. The selective loss of ligands exposes pyramid-shaped activated clusters and facilitates spatial charge separation. As a result, electrons are transferred through Ce–O–Zr and Ce–Ovacancy–Zr pathways, effectively narrowing the band gap and suppressing photoinduced charge recombination. These findings are expected to provide alternative perspectives on selective defect engineering for the design and manufacture of high-performance MOF photo-catalysts for a variety of value-added engineering applications.
MOFs are promising candidates with excellent photocatalytic performance due to their high specific surface area available for absorption, their numerous active metal sites for activating intermediates, and their tunable ligands for photoabsorption and electron injection.6–9 However, MOFs show a strong dependence on co-catalysts, sacrificial agents or even photosensitizers, indicating that the full potential of MOFs for photocatalysis is yet to be discovered.10–12 As such, there is a pressing need to figure out feasible approaches to develop the desired intrinsic properties of MOFs, such as via defect engineering, to maximize the catalytic functions of MOFs.13–16
Defect engineering refers to the introduction of incongruous units into a perfect crystalline structure to disrupt its long-range order either in metals (or clusters) or ligands.17 Defect engineering improves the photocatalytic performance on many levels, such as by expanding the pore volume, tuning the band gap and creating active sites.18–21 However, there are few reports on accelerating the charge separation in MOFs via defect engineering.22–24
Given that classical defect engineering is a random loss of metal/clusters or linkers without selectivity, it is challenging to tailor the coordination environment within MOFs to build a desired electron transfer pathway. A series of mixed-ligand (terephthalic acid and amino terephthalic acid) MIL-125 MOFs were prepared to determine the mechanistic roles of various ligands in creating selective defects and imparting relevant photoactive performance.25 In comparison with terephthalic acid, amino terephthalic acid is thermally labile, which leads to uniform ligand deficiencies. Such a concept inspired us to investigate if the physical and chemical differences between various parts of a catalyst can be utilized to introduce selective linker defects.
Herein, we employ a thermally induced strategy to create MOFs with site-selective ligand defects. We selected Ce-doped UiO-66–NH2 as the matrix and leveraged the differences in coordination environments to introduce defects. When Ce is doped into Zr-oxo clusters, their coordination environments are changed and the linkers coordinated to Ce are located at thermally labile sites. Extended X-ray absorption fine structure (EXAFS) analysis confirms the existence of selective ligand deficiencies and unsymmetrical pyramid clusters. X-ray photoelectron spectroscopy (XPS) demonstrates that the defects open up an additional electron transfer pathway between the Ce and Zr to facilitate charge transfer. Density-functional theory (DFT) calculations further support the origin of the selectively introduced defects and the photocatalysis mechanisms. This work is expected to provide alternative guidance to optimize MOFs coordination environments through selective defect engineering strategies.
Regarding the form that Ce adopts in the clusters, previous studies reveal three available categories of cluster—(Zr6O4(OH)4, CeZr5O4(OH)4 and Ce6O4(OH)4)—and only two of these can coexist.26 Table S1† shows that the mole percentage of Ce in all the MOF samples is below 1/6. Under such circumstances, Zr6O4(OH)4 and CeZr5O4(OH)4 can coexist. Ce6O4(OH)4 only forms when Ce addition exceeds 1/6.26 As both Ce and Zr can both form M6O4(OH)4 clusters, the substitution of Zr with Ce minimizes the impact on the XRD patterns of the final products, which is consistent with the experimental results in Fig. 1a. After pyrolysis, D-UiO-66–NH2 (Zr/Ce0.25) retains one broad peak at 2θ values of 7.4° and 8.6°, corresponding to the main peaks of UiO-66–NH2, indicating that the framework and fcu-topology of UiO-66–NH2 were maintained despite ligand loss. Further research indicated that if the CeCl3 addition exceed 0.67, D-UiO-66–NH2 (Zr/Ce0.67) loses its framework structure (Fig. S4†).
Fig. 1b presents the Fourier-transform infrared (FT-IR) spectra of the as-prepared samples. Bands at 3460, 3343 and 1622 cm−1 are ascribed to the asymmetrical and symmetrical stretching vibrations of the free –NH2 groups on the linkers.27 Bands at 1571 and 1437 cm−1 correspond to the carboxylate and C–C bonds in the benzene ring skeleton.28 The stretching vibration of CO and the symmetrical vibration of the carboxylate are evident at 1656 and 1387 cm−1, respectively.29 Bands at 665 and 487 cm−1 are assigned to the M–Oμ3 and M–OHμ3 stretching while the M–(OC) asymmetric stretching is observed at 565 cm−1.30 These observations confirm the coexistence and successful coordination of the Zr6O4(OH)4 clusters and amino terephthalic acid in the samples. Furthermore, the FT-IR spectra of UiO-66–NH2 (Zr/Cex) and D-UiO-66–NH2 (Zr/Cex) exhibit similar bands (Fig. S5 and S6†), demonstrating their successful synthesis. Compared to UiO-66–NH2 (Zr/Cex), D-UiO-66–NH2 (Zr/Cex) shows broader and weaker bands corresponding to the –C
O, M–(OC) and M–OHμ3 groups, providing evidence of ligand deficiencies and dihydroxylation of the Zr-oxo clusters.31
The morphology of UiO-66–NH2 (Zr/Ce0.25) and D-UiO-66–NH2 (Zr/Ce0.25) was analyzed using scanning electron microscopy (SEM) and transmission electron microscopy (TEM) (Fig. 1c and S7†). After pyrolysis, no significant morphological changes were observed, and D-UiO-66–NH2 (Zr/Ce0.25) retained a similar surface to that of UiO-66–NH2 (Zr/Ce0.25). High-resolution TEM (HRTEM) did not reveal distinct lattice fringes, suggesting that D-UiO-66–NH2 (Zr/Ce0.25) does not undergo transformation into cerium oxide or zirconium oxide agglomerates. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy-dispersive spectroscopy (EDS) elemental mapping confirmed the uniform distribution of all elements (Fig. 1d–j).
Thermogravimetric analysis (TGA) under an air atmosphere and 1H nuclear magnetic resonance (NMR) spectroscopy were employed to qualitatively and quantitatively assess the defect levels. As shown in Fig. 2a, the number of remaining ligands in D-UiO-66–NH2 (Zr/Cex) is significantly lower than in UiO-66–NH2, indicating the successful introduction of ligand defects through pyrolysis. This finding is further supported by the quantitative 1H NMR spectra. Generally, hydrofluoric acid is applied to digest the as-prepared samples while preserving the integrity of ligands. All digested samples show the same peaks (δ = 7.88, 7.43, 7.18) as a commercially available NH2–H2BDC sample (Fig. S8†). Notably, additional peaks appeared in the spectra of D-UiO-66–NH2 (Zr/Cex) when compared to UiO-66–NH2 (Zr/Ce0.25) and pristine UiO-66–NH2, which can be attributed to derivatives produced via the decarboxylation of NH2–H2BDC during pyrolysis (Fig. S9†).
By combining the 1H NMR data with the inductively coupled plasma optical emission spectroscopy (ICP-OES) results, the ligand-to-cluster (L/C) ratios of the samples were determined (Table S2†). In ideal UiO-66, there are 12 dicarboxylate ligands coordinated to each Zr6O4(OH)4 cluster with two clusters sharing one ligand, resulting in an L/C ratio of 6. For UiO-66–NH2, the L/C ratio is measured to be 5.82, indicating the presence of minor ligand defects. Upon CeIII doping, one of the 12 ligands (BDC–NH22−) is expected to dissociate and compensate for the charge imbalance caused by Ce substitution. Consequently, the L/C ratio of UiO-66–NH2 (Zr/Ce0.25) decreases slightly to 5.71. After pyrolysis, the defect level of D-UiO-66–NH2 (Zr/Ce0.25) increased significantly, reducing the L/C ratio to 4.47, which is mainly attributed to the thermal treatment.
In contrast, UiO-66–NH2 and UiO-66–NH2 (Zr/Ce0.167), exhibit minimal cluster distortion and do not show significant pyrolysis below 300 °C. These findings further support the inference that the defect levels vary systematically with the Ce dosage. Quantitative 1H NMR and ICP-OES data confirm that the L/C ratio of D-UiO-66–NH2 (Zr/Cex) gradually decreases as the Ce dosage increases, consistent with the earlier assumption (Table S2†). In addition, ultraviolet-visible diffuse reflectance spectroscopy (UV-vis DRS) analysis of UiO-66–NH2 (Zr/Cex) and D-UiO-66–NH2 (Zr/Cex) reveals a similar trend (Fig. S10 and S11†). For UiO-66–NH2 (Zr/Cex), the visible absorption expands with increasing cerium content, which can be ascribed to the level of Ce doping. The selectively introduced defects further enhance the impact of cerium on visible light absorption. The band gap shown in the Tauc plots also confirms this regular variation (Fig. S12 and S13†), indicating that the defect level in these materials is systematically modulated by the Ce dosage. Unlike the defect-containing MOFs induced by mixed ligands mentioned above, the ligands in UiO-66–NH2 (Zr/Cex) are selectively removed not due to the intrinsic properties of the ligands, but due to their coordination environment.25
To further investigate the local coordination environment around Ce, EXAFS analysis was performed. Both wavelet transformation EXAFS (WT-EXAFS, Fig. S14–S16†) and Fourier-transform EXAFS (FT-EXAFS, Fig. 2d–f and S17–S19†) were applied. The peak at R = 2.55 Å is assigned to the Ce–O bonds in the ZrCe-oxo clusters and the peak at R = 3.60 Å is ascribed to the Ce–Zr distance (Fig. 2d and Table S3†). For the Zr K-edge of D-UiO-66–NH2 (Zr/Ce0.25) and UiO-66–NH2, two paths Zr–OCOOH and Zr–Oμ3 are applied to fit the first shell, and the second shell is attributed to the Zr–Zr path (Fig. 2e, f, Tables S4 and S5†). The corresponding models are provided in Fig. S20–S22.†
FT-EXAFS fitting reveals a reduction in the coordination number (CN) of the Ce–OCOO bonds, as indicated by the low intensity of the corresponding peaks, which means all linkers around the Ce atoms are gone. Meanwhile, the CN of Zr–OCOOH for D-UiO-66–NH2 (Zr/Ce0.25) and UiO-66–NH2 are 3.36 and 4, respectively (Tables S3–S5†). The CN of the Ce–O bonds is 2.63, which is significantly lower than that of Zr–Oμ3 in both D-UiO-66–NH2 (Zr/Ce0.25) and UiO-66–NH2. In a typical Zr6O4(OH)4 cluster, the Zr atoms are coordinated with 2Oμ3, 2Oμ3-OH and 4 OCOO. Upon pyrolysis, Zr6O4(OH)4 will lose two H2O (derived from two Oμ3-OH and four Hμ3-OH).30 If it is assumed that CeZr5O4(OH)4 loses either two Oμ3-OH or loses nothing during pyrolysis, 68.5% of CeZr5O4(OH)4 will give up two O atoms coordinated to Ce. Thus, the CN of Ce–O could decrease to 2.63 and the Ce-doped clusters turn into D (defect-containing)-ZrCe-oxo clusters. In conclusion, the FT-EXAFS analysis confirms that the Ce atoms lose all surrounding ligands and some Oμ3 during pyrolysis. The ligand defects are site-selectively introduced into D-UiO-66–NH2 (Zr/Ce0.25). Thus, by combining the EXAFS fitting and ligand quantification (Table S2†), we present a model in Fig. 3.
![]() | ||
Fig. 3 COHP of (a) the Zr-oxo clusters and (b) the ZrCe-oxo clusters; models of (c) the Zr-oxo clusters and (d) the ZrCe-oxo clusters. |
The Ce–O and Ce–Zr distances are longer than those of Zr–O and Zr–Zr due to the larger atomic radius of Ce. Such differences in atomic size lead to distortions in the clusters. Coupled with the weaker electrostatic interactions resulting from the lower valency of CeIII, the linkers around Ce are prone to dissociate, thus explaining the origin of the selectively introduced defects.
To elucidate the origin of selectively introduced defects from a theoretical perspective, crystal orbital Hamilton population (COHP) calculations were conducted to evaluate the strength of the M–OC bonds between the linkers and metals (Fig. 3). Generally, a negative integrated COHP value indicates the formation of a chemical bond and a higher absolute value corresponds to stronger bonds. Obviously, the Ce1–O13 bond in the ZrCe-oxo clusters is weaker than the Zr–O bond in the Zr-oxo clusters. Though the Zr1–O9 bond in ZrCe-oxo is stronger, the total integrated COHP values of Ce1–O13 and Zr1–O9 in the CeZr-oxo clusters still indicate a weaker interaction compared with the two Zr–O bonds in the Zr-oxo clusters. As a result, there is a lower energy barrier for the ligands coordinated with Ce to dissociate. As a consequence, the selective defect engineering creates D-ZrCe-oxo clusters with an unsymmetrical pyramid. The top of the pyramid is a Ce atom and the base consists of four Zr atoms.
To verify the photocatalytic stability of D-UiO-66–NH2 (Zr/Ce0.25), we conducted 12 h continuous photocatalytic experiments. D-UiO-66–NH2 (Zr/Ce0.25) maintains a steady CO2 photoreduction rate as shown in Fig. 4b, demonstrating high stability under irradiation for a long time period. Furthermore, TEM and XRD analysis of D-UiO-66–NH2 (Zr/Ce0.25) after 12 h of irradiation shows negligible changes, indicating good stability (Fig. S29 and S30†). 1H NMR quantitative defect level characterization also confirms that the L/C ratio shows a negligible decrease (Table S2†). No substantial CO was detected in control experiments without illumination and catalysts. To confirm the carbon source of the generated CO, an isotope labelling experiment was conducted by replacing CO2 with 13CO2 during the photocatalysis. As shown in the inset of Fig. 4b, it is clear that 13CO (m/z = 29) is the dominant product, indicating that the CO mainly originated from CO2 photoreduction instead of decomposition of the ligands or other sources. The high stability of the photocatalyst can be assigned to strong Zr–O bonds and high coordination number of the Zr6O4(OH)4 clusters.32
We systematically compared our results with recently reported photocatalytic performances of defect-containing MOFs (Table S6†). The comparison reveals that most reported MOF-based systems rely on unsustainable photosensitizers and organic sacrificial agents. In contrast, this work focuses on enhancing the intrinsic photocatalytic activity of UiO-66–NH2 without requiring additional photosensitizers or sacrificial agents.
Furthermore, density of state (DOS) calculations were performed to elucidate the mechanism responsible for the band variations (Fig. S34†). The band gap of the Zr-oxo-clusters is calculated as 0.126 eV while the D-ZrCe-oxo clusters do not have a band gap. Obviously, the projected DOS of Ce contribute greatly to the total DOS (TDOS) and crosses the Fermi level, and thus the bandgap is significantly narrowed. Moreover, Ce doping and the selectively introduced defects change the coordination environment of Zr. Zr contributes to the TDOS and expands the lowest unoccupied crystal orbits. Although the bandgap energy is underestimated by the restricted calculation method, the trend remains in the results obtained from the Tauc plots.
During carrier separation and migration, recombination is inevitable and significantly reduces photocatalytic performance. To investigate the charge separation and migration efficiency, photoelectrochemical experiments were conducted. Electrochemical impedance spectroscopy (EIS) Nyquist plots (Fig. 5a) reveal that D-UiO-66–NH2 (Zr/Ce0.25) exhibits the smallest semicircle diameter, indicating minimal resistance to carrier migration. The photocurrent responses (Fig. 5b) demonstrate that D-UiO-66–NH2 (Zr/Ce0.25) has the highest photocurrent, suggesting that it has the most efficient charge separation and migration of all the samples.
Photoluminescence (PL) emission spectroscopy and time-resolved photoluminescence (TRPL) spectroscopy were performed to evaluate the charge recombination. In Fig. 5c, UiO-66–NH2 exhibits the highest photoluminescence intensity at 460 nm, while D-UiO-66–NH2 (Zr/Ce0.25) shows negligible PL emission near 480 nm. A lower PL intensity indicates reduced carrier recombination in D-UiO-66–NH2 (Zr/Ce0.25). This finding is further supported by the TRPL results (Fig. 5d). The average fluorescence lifetime (τ) is calculated using the following equation:
τ = (B1τ12 + B2τ22 + B3τ32)/(B1τ1 + B2τ2 + B3τ3) |
To further explore the charge transfer, femtosecond transient absorption spectroscopy (fs-TAS) was conducted. As is demonstrated in Fig. 5e–g, UiO-66–NH2 and UiO-66–NH2 (Zr/Ce0.25) show broad positive features at 640–700 nm. These positive features can be ascribed to excited-state absorption (ESA) induced by charge transfer and charge separation.34 LCCT (ligand to cluster charge transfer) might be the main cause of the signal at 640–700 nm.35 However, D-UiO-66–NH2 (Zr/Ce0.25) only shows a slightly positive signal at 640–700 nm while it demonstrates a stronger positive signal at 590 nm. Obviously, the ESA signal of D-UiO-66–NH2 (Zr/Ce0.25) undergoes a faster dynamics process within 20 ps while it takes approximate 100 ps for the ESA signals of UiO-66–NH2 and UiO-66–NH2 (Zr/Ce0.25) to vanish. Consistent with the TRPL, the TAS kinetics also demonstrate that D-UiO-66–NH2 (Zr/Ce0.25) has the shortest lifetime, further confirming the faster decay of the ESA signal of D-UiO-66–NH2 (Zr/Ce0.25) (Fig. 5h). These observations suggest that D-UiO-66–NH2 (Zr/Ce0.25) displays weak LCCT because of the loss of ligands. Meanwhile, another stronger electron transfer pathway is opened up and induces faster charge separation and transfer.
XPS was applied to analyze the surface states of UiO-66–NH2 (Zr/Ce0.25) and D-UiO-66–NH2 (Zr/Ce0.25). XPS confirms the presence of C, N, O, Zr, and Ce in both samples (Fig. S35†). In Fig. S36 and Table S9,† the Zr 3d spectrum of D-UiO-66–NH2 (Zr/Ce0.25) revealed that the Zr 3d5/2 and 3d3/2 peaks appear at 182.47 and 184.88 eV, corresponding to the ZrIV oxidation state.36 Meanwhile, the Ce 3d spectrum (Fig. S37 and Table S8†) of D-UiO-66–NH2 (Zr/Ce0.25) displays two peaks at 886.74 and 882.46 eV for Ce 3d5/2 and two peaks at 905.41 and 901.10 eV for Ce 3d3/2, confirming the presence of CeIII.37 Interestingly, compared with UiO-66–NH2 (Zr/Ce0.25), the Zr 3d peaks of D-UiO-66–NH2 (Zr/Ce0.25) are shifted to a lower binding energy (BE) while the Ce 3d peaks are shifted to a higher BE. These findings indicate that Zr is partially reduced, while Ce is partially oxidized, with a similar shift in magnitude of approximately 0.6 eV for both elements, thus suggesting that an electron transfer path exists between Ce and Zr.38 In photocatalysis, Zr acts as a reduction site, while the Ce serves as an oxidation site. The Ce L3-edge XANES spectra of UiO-66–NH2 (Zr/Ce0.25) and D-UiO-66–NH2 (Zr/Ce0.25) reveal their bulk oxidation states (Fig. S38†). Both materials exhibit Ce(III) signatures, while UiO-66–NH2 (Zr/Ce0.25) additionally shows a characteristic Ce(IV) peak at 5738 eV, indicating mixed-valence states.39 We propose a Ce(III)-terminated surface with bulk Ce(III)/Ce(IV) mixed valence in UiO-66–NH2 (Zr/Ce0.25). This phenomenon has been documented previously,40 where formic acid addition during synthesis suppresses Ce(III) oxidation to Ce(IV).40 On the one hand, monocarboxylic acid can introduce linker defects, whilst on the other hand, the surface of UiO-66–NH2 is usually capped by monocarboxylic acid or water, which can also be seen as a linker defect state. As a result, we attribute the Ce(IV) → Ce(III) transformation (Fig. S39†) during pyrolysis to linker deficiency-induced stabilization of the Ce(III) state. Ce(IV) reduction creates a Zr–Ce charge imbalance, facilitating Zr–O–Ce electron transfer pathways.
To verify this electron transfer mechanism, we performed ex situ XPS measurements under dark and illuminated conditions (Fig. 6a, b and Table S10†). High-resolution Zr 3d and Ce 3d spectra were acquired for D-UiO-66–NH2 (Zr/Ce0.25) under both conditions. The peaks of Zr 3d shift to lower BE, while the peaks of Ce 3d shift to higher BE compared with the results obtained in darkness. To further understand the electron transfer pathway, the O 1s spectrum was obtained and is displayed in Fig. 6c and Table S11.† The ZrCe-oxo clusters and Zr-oxo-clusters both have three typical oxygen species: carboxyl oxygen, μ3-O and μ3-OH. Therefore, the spectrum of UiO-66–NH2 (Zr/Ce0.25) was deconvoluted into three peaks. However, a new peak at 533.87 eV, attributed to uncoordinated carboxyl oxygen, appears in the spectrum of D-UiO-66–NH2 (Zr/Ce0.25).38 Obviously, the proportion of μ3-O increases while that of carboxyl oxygen reduces in the change from UiO-66–NH2 (Zr/Ce0.25) to D-UiO-66–NH2 (Zr/Ce0.25), indeed, massive linker defects result in this case.
However, because water is adsorbed in the form of –OH, the proportion of M-μ3-OH only decreases from 23.4% to 12.2%. In contrast, the BE of μ3-O shows the reverse tendency, indicating an increase in the electron density in D-UiO-66–NH2 (Zr/Ce0.25). Interestingly, the BE shift of μ3-O between UiO-66–NH2 (Zr/Ce0.25) and D-UiO-66–NH2 (Zr/Ce0.25) is merely −0.21 eV, which might be due to the effect of the electron transfer pathway.
Both Ce and Zr exist in UiO-66–NH2 (Zr/Ce0.25) and D-UiO-66–NH2 (Zr/Ce0.25). However, the electron transfer pathway is closed up before the site-selective defect engineering is conducted. To further investigate the pathway, electron paramagnetic resonance (EPR) spectroscopy was conducted. As depicted in Fig. 6d, a sharp and symmetrical peak at g = 2.003 is ascribed to the unpaired electrons trapped by the O vacancies while a peak at g = 2.021 indicates the existence of Zr3+ species.41–43 Furthermore, Ce L3-edge EXAFS fitting of D-UiO-66–NH2 (Zr/Ce0.25) suggests a significantly lower coordination number for the Ce–O path compared to the Zr-oxo clusters. This finding indicates that the μ3-oxygen atoms near Ce in the ZrCe-oxo clusters are partially lost during pyrolysis, leading to the formation of numerous oxygen vacancies. This could explain why D-UiO-66–NH2 (Zr/Ce0.25) exhibits the most intense EPR signal. Based on these observations, we can further infer that the trapped electrons do not localize evenly between Ce and Zr due to their different valencies. Instead, the trapped electrons prefer the Zr atoms. Upon illumination, the excited electrons transfer from Zr to the adsorbed CO2, and then the trapped electrons will transfer to Zr, promoting charge separation. To identify the critical intermediates during the photoreduction, in situ diffuse reflection infrared Fourier-transform spectroscopy (DRIFTS) was conducted. As shown in Fig. 6e, after aerating CO2 into the system, the spectrum has obvious CO2 asymmetric stretching peaks (2340 cm−1, Fig. S39†), confirming CO2 adsorption.15 There exist few peaks corresponding to intermediates until illumination occurs. The peaks at 1664 and 1483 cm−1 are correlated to bicarbonate (HCO3−).44 The peak corresponding to bidentate carbonate (b-CO32-) is found at 1593 cm−1.45 Most importantly, the peaks at 1550 and 1200 cm−1 provide strong evidence for the presence of critical intermediate *COOH.46 Based on this spectrum, we propose the following mechanism for the photocatalytic conversion of CO2 to CO:
CO2(g) → *CO2 |
H2O → H+ + OH− |
*CO2 + e− + H+ → *COOH |
*COOH + e− + H+ → *CO + H2O |
*CO → CO(g) |
Notably, after selective defect engineering, the Zr 4d-orbitals redistribute and the d-band center of Zr 4d drops from 1.3 to 0.16 eV, indicating that the latter has better catalytic performance (Fig. 6g and h).47,48 Charge density difference and Bader charge analysis also give similar results whereby the Zr in the Zr-oxo-clusters only transfers 0.004 and 0.131 electrons to *CO2 and *COOH, respectively. Meanwhile, the Zr in the D-ZrCe-oxo clusters transfers 1.304 and 0.768 electrons to *CO2 and *COOH, respectively, indicating higher activity for the Zr in the D-ZrCe-oxo-clusters (Fig. S41 and S42†).
Therefore, we propose the following mechanism: according to a “LCCT” mechanism, upon irradiation, electrons are injected from the amino terephthalic acid moieties to the clusters.35 Simultaneously, electrons in the metal-oxo-clusters can be excited. However, the Zr-oxo-cluster with 12 linkers is highly symmetrical, and therefore, the injected electrons are likely to recombine with the holes left in the clusters. Conversely, linkers with selectively introduced defects give the D-ZrCe-oxo-clusters an unsymmetrical pyramidal structure. The Ce atoms located at the top of pyramid cannot receive electrons from the linkers but act as the accumulation sites of the holes. The excited electrons can transfer through the Ce–O–Zr and Ce–Ovacancy–Zr pathways to Zr, realizing spatial charge separation. Finally, the adsorbed CO2 and H2O is reduced and oxidized to finish the photoreaction.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5sc01194a |
This journal is © The Royal Society of Chemistry 2025 |