Rational design of metal node-modified Ti-based MOFs for selective photoreduction of carbon dioxide to ethanol by computational screening

Abstract

In this work, density functional theory (DFT) calculations were conducted to investigate a series of metal node-modified Ti-MOF catalysts using transition metals (Mn, Fe, Co, Ni, Cu, Zn, Zr, Nb, Mo, Ru, Rh, Pd, Hf, Ta, W, Os, Ir, and Au) introduced into Ti-ATA (ATA = 2-aminoterephthalic acid) for the photocatalytic reduction of CO₂ to C₂ products. CO₂ can be sufficiently activated on Ti(M)-ATA but the adsorption configuration depends on the nature of M. Over Ti(Nb)-ATA, Ti(Ta)-ATA, Ti(Zr)-ATA and Ti(Hf)-ATA, the two *CHO species undergo C–C coupling to form *CHOCHO, an important C₂ intermediate. Ti(Nb)-ATA and Ti(Ta)-ATA tend to generate ethanol, while Ti(Zr)-ATA and Ti(Hf)-ATA are more selective to ethylene. Among the Ti(M)-ATA candidates studied, Ti(Nb)-ATA was identified as the most active catalyst for CO₂ reduction to ethanol due to its smallest limiting free energy change (1.12 eV), over which the *CH₂CH₂O reduction to *CH₂CH₂OH was found to be the rate-determining step. The correlation curve analysis illustrates that the reduction activity of Ti(M)-ATA catalysts is highly dependent on the binding strength of CO₂ and key reaction intermediates such as *OCHOH. The analysis of electronic and optical properties indicates that the altered energy band structure and charge transfer behavior around the bimetallic nodes of Ti(Nb)-ATA account for its excellent catalytic activity for CO₂ reduction to ethanol.

---

1 Introduction

The utilization of fossil fuels for the production of various industrial products emits large amounts of carbon dioxide (CO₂) into the atmosphere, leading to serious environmental problems.^1,2 Coping with these challenges, catalytic conversions of CO₂ into valuable chemicals via sustainable strategies offer promising approaches.^3,4 Photocatalytic reduction driven by renewable solar energy can convert CO₂ into liquid fuels and value-added chemicals such as ethanol, olefins, and carboxylic acids, which is widely recognized as a sustainable and green approach for CO₂ conversion.^5–7 Photocatalytic reduction of CO₂ to C₂ chemicals is economically more valuable and the process involves multi-electron transfer and multi-proton activation processes. However, the insufficient interactions of CO₂ molecules with the active sites of catalysts, the difficulty of CO₂ activation, and retarded C–C coupling kinetics lead to the limited formation of C₂ products with lower selectivity and yields.⁸ Therefore, the rational design of highly efficient and selective photocatalysts with multi-functional active sites to exert synergistic effects is urgently desired to tune the catalytic activity and selectivity to C₂ products.

Over the past few years, a new class of semiconductor-like crystalline porous materials known as metal–organic frameworks (MOFs) has attracted increasing attention in the fields of heterogeneous catalysis, gas storage, and separation.^9,10 MOFs are composed of inorganic metal ions and organic ligands, featuring well-defined structures, high porosity, larger surface area, component diversity, and tailorability.¹¹ Therefore, MOF-based materials possess rich physicochemical properties and unique structural advantages among porous catalytic materials. In particular, the large surface areas and specially designed active sites of MOFs could enrich CO₂ efficiently with superior CO₂ adsorption capacity and stabilize the reaction intermediates, making them promising platforms for the photocatalytic reduction of CO₂ to chemicals and fuels.^12,13

However, only a limited number of MOF-based photocatalytic systems have been developed for CO₂ reduction and their efficiency to C₂ products is still far from satisfactory. Enhancing the photocatalytic efficiency of MOFs for C₂ formation mainly depends on their ability to generate and maintain photogenerated electrons as well as their ability to activate CO₂ and accelerate C–C coupling.¹⁴ Over the past decade, researchers have designed strategies to improve the photocatalytic performance of MOFs, as summarized below:

(1) By constructing efficient active sites and using the chelating effect of ligands to immobilize atomically dispersed single/dual metals, the photocatalytic reaction can be promoted.^15–17 For example, our group successfully synthesized a novel Ti-based MOF photocatalyst denoted as Fe/Ti-BPDC (BPDC = 2,2′-bipyridine-5,5′-dicarboxylic acid) with atomically dispersed Fe sites, which exhibited high activity and selectivity for CO₂ reduction to HCOOH, with a yield of 703.9 μmol g⁻¹ h⁻¹ and a selectivity greater than 99.7%.¹⁸ Inspired by the experimental work, we proposed a new strategy for designing metal single-atom-modified and dual-atom-modified Ti-BPDC photocatalysts through structural and electronic modulation for CO₂ reduction to C₁ and C₂ products using density functional theory (DFT) calculations.^19,20 The computational results of structure–activity volcano curves show that metal single-atom modified Fe/Ti-BPDC (ΔG_L = 0.40 eV) and Pd/Ti-BPDC (ΔG_L = 1.17 eV) are optimal catalysts for the photoreduction of CO₂ to HCOOH and CH₃OH, respectively,¹⁹ due to their relatively small limiting free energy changes compared to those reported in the literature.²¹ Furthermore, metal dual-atom-modified Ti-BPDCs such as Cu–Sn/Ti-BPDC (ΔG_L = 0.20 eV) and Cu–Os/Ti-BPDC (ΔG_L = 0.70 eV) show enhanced activity toward the generation of HCOOH and CH₃OH products from CO₂ photoreduction compared to metal single-atom modified Fe/Ti-BPDC and Pd/Ti-BPDC. In particular, the dual-metal atom-modified Cu–In/Ti-BPDC (ΔG_L = 1.37 eV) exhibits improved activity and selectivity towards the generation of C₂ products (mainly C₂H₄).²⁰

(2) Light absorption efficiency can be improved by choosing the right ligands. For example, in order to improve the optical properties of MIL-125 (Ti) MOFs, researchers successfully shifted MIL-125 (Ti) to the visible range by introducing the 2-aminoterephthalic acid (ATA) ligand with a broad spectral response, enhancing the visible light absorption of the MOF materials.²² Sun et al. reported that MIL-125-NH₂(Ti) with {110}/{111}-heterojunction yields 10 and 18 times more CO and CH₄ products, respectively, from CO₂ photoreduction compared with the single {001} facet; DFT calculations identified energetically favorable pathways and rate-limiting steps for CO₂ reduction on different low-index surfaces of MIL-125-NH₂(Ti) in their studies.²³

(3) In addition, metal substitution/doping has been demonstrated to be an effective method to improve the performance of MOF-based photocatalysts. One of the effective strategies is to construct bimetallic assemblies by partially replacing a node metal in MOFs with another metal. The bimetallic assemblies can harvest visible light and the doped metal cations can act as electronic mediators to promote charge transfer, facilitating the photocatalytic processes.²⁴ In particular, partial substitution of metal cations in MOFs can lead to the formation of oxygen-bridged heterometallic structures within the framework, which could exhibit enhanced photocatalytic performance due to the structural flexibility and tunability of the designed MOFs.²⁵ For example, Ti-substituted NH₂-UiO-66 (Zr/Ti) was prepared by Li et al. by using the post-synthesis exchange (PSE) method which exhibits good catalytic performance for the photoreduction of CO₂ to formic acid under visible light, with a yield of 5.8 mmol mol⁻¹ for 10 h; DFT calculations and electron spin resonance (ESR) results indicate that the introduction of Ti substituents as a mediator promotes electron transfer, which improves the photocatalytic performance.²⁵ In addition, Truhlar et al. used UiO-66 as the basic skeleton and replaced partial Zr atoms in the nodes with metals such as Hf, Th, Ti, U, Ce, etc.; DFT calculations show that UiO-66(Ce) effectively promotes electron–hole separation due to the fact that Ce⁴⁺ in the material has a low-energy 4f vacancy orbital, which can receive photogenerated electrons, thus more effectively driving the photocatalytic reaction.²⁶

Based on the literature survey and prior studies in our group, the structures and properties of ligands in MOF materials have important impacts on the photocatalytic performance of CO₂ reduction. On the one hand, metal single-atom and dual-atom assemblies can be stably anchored with N-containing ligands within the MOF framework to improve the efficiency of CO₂ activation and conversion.^18–20,27 On the other hand, metal-modified nodes can effectively regulate light absorption and electron–hole separation efficiency, and provide special active sites to further tune the activity and selectivity of MOF-based catalysts for CO₂ photoreduction.^25,26,28 However, the directional control of CO₂ reduction products remains challenging, especially for the generation of C₂ products. Due to the complex mechanisms of adsorption/activation of CO₂ and cleavage/reconstruction of C=O double bonds over the active sites of catalysts, the principles governing the formation and transformation of key intermediates and electron–proton transfer in the reaction processes are still not clear, which make it difficult to regulate the conversion paths and target products.

The limited efficiency of MOF-based photocatalysts in the reduction of CO₂ to C₂ products (e.g., ethanol and ethylene) stems from several intrinsic challenges, among which insufficient activation of CO₂, slow kinetics of C–C coupling, and generation of key intermediates required for C–C coupling are the major bottlenecks.²⁹ On the one hand, many MOFs lack robust adsorption sites (e.g., metal active sites or Lewis base sites) to activate the linear structure of CO₂, which prevents the reduction of CO₂ to the critical *HCOO or *COOH intermediate. On the other hand, C₂ product generation requires multiple electrons, but MOFs usually lack efficient charge transport channels to deliver electrons quickly. In addition, the lack of bimetallic site nodes in some MOF materials leads to low stability of C–C coupling intermediates. To summarize, the synergistic optimization of CO₂ adsorption/activation, electron transfer, intermediate stability, and C–C coupling kinetics inside MOF materials is a crucial factor in overcoming the inefficiency of MOF-based photocatalysts for the production of C₂ products.

Inspired by our previous experimental studies on Ti-based MOFs and the potential application of 2-aminoterephthalic acid (ATA) ligands in MOFs for photocatalysis,^30,31 the present work systematically investigates the adsorption/activation and reduction of CO₂ to C₂ products (mainly ethanol and ethylene) on a series of metal node-modified Ti(M)-ATA MOFs (ATA = 2-aminoterephthalic acid, M = Mn, Fe, Co, Ni, Cu, Zn, Zr, Nb, Mo, Ru, Rh, Pd, Hf, Ta, W, Os, Ir, Au) by means of DFT calculations. By screening these heterometallic node structures via structure–activity correlation, Ti(Nb)-ATA is identified as the most active photocatalyst for CO₂ reduction to ethanol (C₂H₅OH), with a limiting free energy change of 1.12 eV associated with the *CH₂CH₂O → *CH₂CH₂OH step. Electronic property analysis further demonstrates that the computationally designed Ti(Nb)-ATA MOF catalyst via node modification is more favorable for the photocatalytic reduction of CO₂ to C₂ compared to unmodified Ti-ATA and other Ti-based MOFs such as Ti/BPDC and MIL-125-NH₂(Ti). In addition, we also find that node-modified metals from different groups of the periodic table lead to different C₂ products. For example, Ti(Ta)-ATA and Ti(Nb)-ATA tend to generate ethanol, while Ti(Zr)-ATA and Ti(Hf)-ATA are more selective to ethylene.

2 Computational models and methods

The present work builds on recent experimental studies by He et al. in our group, where characterization by X-ray diffraction (XRD) showed that the synthesized Ti-BPDC (BPDC = 2,2′-bipyridine-5,5′-dicarboxylic acid) has high crystallinity; the Ti and O atoms are periodically aligned to form Ti–O layers, and the Ti–O layers connected to the organic ligand are parallel to each other.^18,30 To construct the computational models in this work, we chose 2-aminoterephthalic acid (ATA) as the linker instead of BPDC and the aforementioned Ti-MOF as the basic framework. Within the constructed Ti-MOF catalyst, the Ti and O atoms in the nodes of the MOF exhibit periodically arranged Ti–O layers, with each Ti atom connected to six O atoms to form Ti–O clusters, in which two of the O atoms belong to the ATA linker and four of the O atoms are from the Ti–O layers, as shown in Fig. 1(a) and (c). A periodic cell of Ti-ATA-based MOFs with a lattice constant of 6.44 Å × 21.75 Å × 7.40 Å was built for modelling (Fig. 1(a)). Then, representative metals including Mn, Fe, Co, Ni, Cu, Zn, Zr, Nb, Mo, Ru, Rh, Pd, Hf, Ta, W, Os, Ir, and Au were selected as doping metals for node modification (denoted as Ti(M)-ATA) (Fig. 1(b)). According to the literature, bimetallic assemblies with Ti–O–M oxygen-bridged structures are formed in typical Ti-ATA MOF materials through metal substitution in the modification processes.²⁵ Here, upon introducing the second metal atoms into Ti-ATA, two possible configurations were considered, i.e. one is the formation of a Ti–Ti diagonal structure in one parallelogram at the node center which separates the second metal atoms into different parallelograms, and the other is the formation of an M–M diagonal structure in one parallelogram at the node center within the MOF, separating the Ti atoms into different parallelograms, as illustrated in Fig. 1(d).

Fig. 1 (a) The optimized structure of the Ti-ATA (ATA = 2-aminoterephthalic acid) framework. (b) Representative 3d (Mn, Fe, Co, Ni, Cu, and Zn), 4d (Zr, Nb, Mo, Ru, Rh, and Pd), and 5d (Hf, Ta, W, Os, Ir, and Au) metals selected as doping metals for node modification in Ti(M)-ATA catalysts. (c) Structure of the 2-aminoterephthalic acid (ATA) linker. (d) Two possible node configurations formed in Ti(M)-ATA catalysts.

An all-electronic approach implemented in the Vienna Ab initio Simulation Package (VASP) program was used to perform spin-polarized DFT calculations.^32,33 The projector-augmented wave (PAW) pseudopotentials were used to describe the electron–ion interactions. The Perdew–Burke–Ernzerhof (PBE) functional in the generalized gradient approximation (GGA) was used to calculate the exchange–correlation energies of electrons.³⁴ The Coulomb and exchange interactions were corrected by setting the U_eff parameter (U_eff = Coulomb U − exchange J) for each transition metal.^35,36 The U_eff parameter for titanium atom was set to 3.0 eV, and those for remaining metals are given in Table S1 in the ESI.† To include van der Waals (vdW) interactions, the PBE + D3 method was used.^37,38 The valence electrons were described using a plane wave basis set with a cutoff energy of 450 eV. The convergence criteria for all calculated electronic energies and atomic forces were set to 10⁻⁴ eV and 0.03 eV Å⁻¹, respectively. Both Ti-ATA and Ti(M)-ATA structures were optimized using a 4 × 1 × 4 k-point mesh for sampling the Brillouin zone in the VASP calculations.

The adsorption energy (E_ads) of an adsorbate onto the catalyst was calculated using the equation E_ads = E_total – E_catalyst – E_adsorbate, where E_total, E_catalyst, and E_adsorbate represent the total energy of the system containing the adsorbate and catalyst, the energy of the bare catalyst, and the energy of the adsorbate in the gas phase, respectively. In order to evaluate the catalytic activity of CO₂ photoreduction over these Ti(M)-ATA catalysts, the free energy change (ΔG) was calculated for each elementary step involved in the CO₂ reduction process using the formula ΔG = ΔE + ΔE_ZEP – TΔS,^39,40 where ΔE is the total electronic energy change obtained through DFT calculations, ΔE_ZEP is the zero-point energy contribution, T is the temperature at 298. 15 K, and ΔS is the entropy change. The zero-point energy and entropy were obtained from vibrational frequency calculations using DFT. The zero-point energies of all adsorbed species are provided in Table S2.† In this work, we investigated the reaction pathways for the formation of different C₂ products (mainly C₂H₄ and C₂H₅OH) over Ti-ATA and Ti(M)-ATA catalysts, in which protons and photogenerated electrons are added to the reaction intermediates progressively in the presence of a catalyst, ultimately resulting in the formation of C₂ products.

3 Results and discussion

3.1 Structural stability of Ti(M)-ATA catalysts and the adsorption of CO₂

For clarity, we grouped the introduced second metals (M) into three categories based on their d electrons as 3d (Mn, Fe, Co, Ni, Cu, and Zn), 4d (Zr, Nb, Mo, Ru, Rh, and Pd), and 5d (Hf, Ta, W, Os, Ir, and Au). Before exploring the adsorption of CO₂ molecules on the designed Ti(M)-ATA catalysts, we first investigated the structural stabilities of the two types of metal-modified nodes of MOFs (Fig. 1(d)). The calculated DFT energies are presented in Fig. S1,† and the results show that Ti(M)-ATA candidates, except Ti(Zn)-ATA, exhibit a periodic enhancement of DFT energies from left to right according to the periodic table of elements. Notably, the absolute energy values of these two different node configurations are not very different. Therefore, we further calculated the substitution energies of bimetallic assemblies in Ti(M)-ATA as shown in Fig. S2,† and the specific values are given in Table S3.† The results show that considerable negative substitution energies are obtained for Ti(M)-ATA (M = Zr, Nb, Hf, Ta), which range from −0.93 eV to −3.70 eV. However, all other candidate catalysts have positive substitution energy values, indicating that Zr, Nb, Hf, and Ta atoms can bind strongly to the node Ti atom in the MOFs, thus ensuring their high structural stability.

It is worth noting that DFT energy and substitution energy alone are not sufficient to determine the optimal doping model for metal modification. Therefore, we further explored all possible adsorption configurations of CO₂ on the two different node models, considering four different initial adsorption configurations, including *COO, *OCO, O*CO, and O*OC (* indicates adsorption at the doped metal active site). With model structure 1, as shown in Fig. 1(d), after comprehensive structural screening and optimization, it was found that CO₂ is adequately activated on Ti(M)-ATA (M = Nb, Zr, Ta, Hf), with one O atom of CO₂ bonded to the doped metal site in the node and the C atom of CO₂ bonded with the N atom in the ATA ligand, forming a stable *O(M)–C(N)–O angular adsorption structure, as shown in Fig. 2(a). However, on the catalyst models of Ti(M)-ATA (M = Mn, Fe, Co, Ni, Cu, Zn, Mo, Ru, Rh, Pd, W, Os, Ir, Au), CO₂ shows a linear adsorption configuration and is not bonded to other atoms in the MOF catalysts, as illustrated in Fig. S3.† With model structure 2 as shown in Fig. 1(d), CO₂ exhibits the *O(Zr)–C(N)–O adsorption mode merely on Ti(Zr)-ATA whereas it shows a linear adsorption pattern on the rest of the candidate catalysts, as shown in Fig. S4.† The adsorption energies of CO₂ on pristine Ti-ATA and metal-modified Ti(M)-ATA (model structure 1 and model structure 2) were calculated, as shown in Fig. 2(b–d) and S5,† while the specific values of adsorption energies are provided in Table S4.† The calculation results show that the adsorption energy values of CO₂ on Ti(M)-ATA (M = Nb, Zr, Ta, Hf) are 0.15, 0.31, −0.10, and 0.26 eV, respectively, which are significantly lower than that obtained over the pristine Ti-ATA catalyst (1.45 eV) and those obtained over other metal-modified Ti(M)-ATA candidate catalysts. Therefore, the Ti(M)-ATA (M = Nb, Zr, Ta, Hf) candidates were chosen for subsequent exploration of the reaction pathways and reactivities for CO₂ photoreduction to C₂ products. It is worth noting that the adsorption energies of CO₂ on model structure 1 are generally lower than those on model structure 2, as shown in Fig. S5.† Based on the calculated CO₂ adsorption configurations and adsorption energies, the metal-modified model structure 1 exhibits good structural stability, sufficient interactions with CO₂, and better ability to activate CO₂; therefore it was chosen as the representative node-modification model for subsequent studies.

Fig. 2 (a) The optimized stable adsorption configurations of CO₂ molecules on metal-modified model structure 1 of Ti(M)-ATA (M = Nb, Zr, Ta, Hf); (b) the adsorption energies of CO₂ on pristine Ti-ATA and metal-modified model structure 1 of Ti(M)-ATA (M = Mn, Fe, Co, Ni, Cu, Zn); (c) the adsorption energies of CO₂ on pristine Ti-ATA and metal-modified model structure 1 of Ti(M)-ATA (M = Zr, Nb, Mo, Ru, Rh, Pd); (d) the adsorption energies of CO₂ on pristine Ti-ATA and metal-modified model structure 1of Ti(M)-ATA (M = Hf, Ta, W, Os, Ir, Au).

3.2 CO₂ reduction mechanisms and catalytic performance of Ti-ATA and Ti(M)-ATA (M = Zr, Nb, Hf, Ta)

Since CO₂ can be sufficiently activated on Ti(M)-ATA (M = Zr, Nb, Hf, Ta), we proceeded to investigate the reaction properties of CO₂ reduction over these candidate catalysts. The reduction of CO₂ to C₂ products can be divided into hydrogenation processes and C–C coupling reaction. In particular, the formation of stable C–C coupling intermediates on the catalyst is a crucial step in the reduction of CO₂ to C₂ products; thus we first examined all possible hydrogenation pathways for CO₂ reduction before C–C coupling on pristine Ti-ATA and metal-modified Ti(M)-ATA (M = Zr, Nb, Hf, Ta) candidates, and mapped out the Gibbs free energy diagrams as shown in Fig. 3. The first proton may attack the O or C atom of *OCO, forming the *OCOH or *OCHO intermediate, respectively. For the IVB metals in the periodic table, including single metal Ti, bi-metal Ti(Zr) and Ti(Hf), the *OCOH intermediate cannot be stabilized even after structural optimization and spontaneously transforms into the *OCHO intermediate. Moreover, it is found that the first hydrogenation step of *OCO → *OCHO is thermodynamically quite unfavorable on Ti (Zr) and Ti (Hf), with ΔG values of 0.91 and 1.18 eV, respectively, as shown in Fig. 3(b) and (c). Although the hydrogenation of *OCO to the *OCHO intermediate is slightly uphill (ΔG of 0.17 eV) over pristine Ti-ATA, the CO₂ adsorption/activation process is largely endothermic (ΔG of 1.45 eV) on this catalyst, as illustrated in Fig. 3(a). In the next scenario, the continued hydrogenation of *OCHO to form the *OCHOH species is significantly exothermic, releasing energies of 2.92, 2.77, and 2.81 eV, respectively on Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA. Then, the formed *OCHOH species needs to overcome large uphill energy barriers (ΔG = 1.48, 2.04, and 2.10 eV on Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA) to dissociate into the key *CHO intermediate species, releasing a H₂O molecule simultaneously. Overall, the formation of the key *CHO intermediate on Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA is quite energy consuming and thus energetically unfavorable. By contrast, for the VB metals including Ti(Nb) and Ti(Ta), both intermediates can be formed from the first hydrogenation step of *OCO. Clearly, the *OCHO intermediate is much more stable than the *OCOH intermediate, as illustrated in Fig. 3(d) and (e); thus the pathway for further conversion of the *OCOH intermediate is not considered here. Subsequently, the hydrogenation of the *OCHO intermediate on Ti(Nb)-ATA and Ti(Ta)-ATA requires overcoming moderate uphill energies of 0.68 and 0.72 eV, respectively, leading to the formation of the *OCHOH intermediate. In the next step, the continued reduction of *OCHOH on Ti(Nb)-ATA to form the key intermediate *CHO is thermodynamically favorable, releasing an energy of 0.71 eV, whereas the formation of *CHO species on Ti(Ta)-ATA still requires overcoming an energy barrier of 0.35 eV. The above calculation results reveal that the formation of the key *CHO intermediate is thermodynamically more preferred on the Ti(Nb)-ATA catalyst, which plays an important role in subsequent C–C coupling and further reduction to generate C₂ products.

Fig. 3 The calculated Gibbs free energy diagrams for CO₂ reduction steps before C–C coupling on (a) Ti-ATA, (b) Ti(Zr)-ATA, (c) Ti(Hf)-ATA, (d) Ti(Nb)-ATA, and (e) Ti(Ta)-ATA catalysts.

Starting with the formed *CHO species, all possible pathways and intermediates for the reduction of CO₂ to C₂H₄ or C₂H₅OH products on pristine Ti-ATA and metal-modified Ti(M)-ATA (M = Zr, Hf, Nb, Ta) were considered, as illustrated in Fig. 4. By comparing the reaction Gibbs free energies, the optimal pathway for CO₂ reduction to C₂H₄ or C₂H₅OH over each of the examined catalyst was identified, as shown in Fig. 5(a) and (b), and the optimized structures of relevant intermediates are given in Fig. 5(c) and (d). Other pathways considered and investigated are provided in Fig. S6.† In this scenario, we focus on discussing the reaction mechanisms starting from the C–C coupling step, as detailed below:

Fig. 4 Possible reaction pathways considered for the reduction of CO₂ to C₂ (C₂H₅OH and C₂H₄) products on (a) Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA and (b) Ti(Nb)-ATA and Ti(Ta)-ATA.

Fig. 5 The calculated Gibbs free energy diagrams of the optimal paths for CO₂ reduction to the C₂H₄ product on (a) Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA and to the C₂H₅OH product on (c) Ti(Nb)-ATA and Ti(Ta)-ATA. Structural illustrations of all stable intermediates associated with the optimal pathways on (b) Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA and (d) Ti(Nb)-ATA and Ti(Ta)-ATA. The MOF framework structure is omitted for clarity.

(1) According to the free energy diagram shown in Fig. 5(a), over pristine Ti-ATA, metal-modified Ti(Zr)-ATA and Ti(Hf)-ATA, the two *CHO species undergo C–C coupling to form the important C₂ intermediate *CHOCHO (Fig. 5(b)), releasing the energies of 0.98, 1.91, and 1.99 eV, respectively on the three catalysts. The subsequent conversion of *CHOCHO species leads to the formation of four possible intermediates, namely *CHOCHOH, *CHOCH₂O, *CHOHCHO, or *CH₂OCHO. The calculated Gibbs free energies in Fig. S6(a–c)† show that the formation of these four intermediates is thermodynamically unfavorable due to the uphill free energy changes, with the *CHOCHOH species formation involving a relatively smaller energy increase compared to the other three species (ΔG = 0.31 (Ti-ATA), 1.17 (Ti(Zr-ATA)), and 1.20 eV (Ti(Hf)-ATA)). The formed *CHOCHOH intermediate is then reduced by proton-coupled electron transfer to form *CHOHCHOH, *CHOCH + H₂O, *CHOCH₂OH, or *CH₂OCHOH species. Notably, the ΔG values for the formation of *CHOCH + *H₂O on Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA are −0.78, −1.00, and −1.07 eV, respectively, which are lower than those for the formation of the other three intermediates (Fig. S6(a–c)†), suggesting that *CHOCH + *H₂O is the preferred intermediate from *CHOCHOH reduction. In the next step, further reduction of *CHOCH leads to the formation of *CH₂OCH, *CHOCH₂, or *CHOHCH. As demonstrated in Fig. 5(a) and S6(a–c),† the formation of *CH₂OCH, *CHOCH₂, and *CHOHCH species is thermodynamically unfavorable due to uphill free energy changes, but *CHOCH₂ is relatively more stable than the other two species, with ΔG values of 0.67 (Ti-ATA), 0.53 (Ti(Zr)-ATA), and 0.63 eV (Ti(Hf)-ATA), respectively. Subsequently, the *CHOCH₂ intermediate undergoes further reduction to either *CH₂OCH₂ or *CHOHCH₂ species, with the formation of *CHOHCH₂ species being thermodynamically more favorable, as shown in Fig. S6(a–c).† In contrast, the reduction of *CHOCH₂ to *CH₂OCH₂ species is thermodynamically unfavorable, with ΔG values of −0.26 (Ti-ATA), −0.47 (Ti(Zr)-ATA), and −0.48 eV (Ti(Hf)-ATA), as shown in Fig. 5(a). For the generated *CHOHCH₂ intermediate, two possible hydrogenation pathways lead to different C₂ intermediates: either the formation of *CHCH₂ species by releasing a H₂O molecule (ΔG = 0.21 (Ti-ATA), 0.37 (Ti(Zr)-ATA), and 0.38 eV (Ti(Hf)-ATA)), which is then reduced to *CH₂CH₂ (ΔG = −0.44 (Ti-ATA), −0.54 (Ti(Zr)-ATA), and −0.61 eV (Ti(Hf)-ATA)), or the generated *CHOHCH₂ intermediate continues to be hydrogenated to form *CH₂OHCH₂ species which is further reduced to *CH₂OHCH₃. Finally, the formed *CH₂OHCH₃ (*C₂H₅OH) or *CH₂CH₂ (*C₂H₄) desorbs from the catalysts. Apparently, the reduction of CO₂ to *C₂H₄ over Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA catalysts is more selective (energetically favorable) than to *C₂H₅OH according to the calculation results given in Fig. S6(a–c).† However, the desorption of formed *CH₂CH₂ from the catalyst proceeds slowly due to the substantial free energy changes on the three catalysts examined, as shown in Fig. 5(a). Based on the above results, the optimal pathway for the photocatalytic reduction of CO₂ to the preferred C₂H₄ product on pristine Ti-ATA, metal-modified Ti(Zr)-ATA and Ti(Hf)-ATA catalysts can be summarized as *CO₂ → *OCHO → *OCHOH → *CHO → *CHOCHO → *CHOCHOH → *CHOCH + H₂O → *CHOCH₂ → *CHOHCH₂ → *CHCH₂ → *CH₂CH₂ → *+C₂H₄. According to the free energy diagram in Fig. 5(a), the reduction step *OCHOH → *CHO (ΔG = 1.48 (Ti-ATA), 2.04 (Ti(Zr)-ATA), and 2.10 eV (Ti(Hf)-ATA)) has the largest free energy change and is considered to be the rate-determining step for the overall reaction. On these three catalysts, C–C coupling is not difficult to achieve with a substantially downhill free energy change; however the hydrogenation reactions before C–C coupling proceed slowly, especially the generation of the key *CHO intermediate.

(2) In contrast, the mechanisms of CO₂ reduction to C₂ products on Ti(Nb)-ATA and Ti(Ta)-ATA catalysts are different. According to the free energy diagrams given in Fig. 5(c), for Ti(Nb)-ATA and Ti(Ta)-ATA, the two *CHO species undergo C–C coupling to form *CHOCHO (Fig. 5(d)), an important C₂ intermediate, releasing the energies of 1.02 and 1.45 eV, respectively on the two catalysts. Further reduction of *CHOCHO species proceeds via different pathways. As illustrated in Fig. S6(d) and (e),† among the four possible intermediates produced from *CHOCHO reduction, the formation of *CHOCH₂O is thermodynamically the most favorable, with ΔG values of 0.45 and 0.46 eV, on Ti(Nb)-ATA and Ti(Ta)-ATA, respectively. The formed *CHOCH₂O species can be further reduced to *CH₂OCH₂O, *CHOHCH₂O or *CHOCH₂OH. From Fig. S6(d) and (e),† it can be seen that *CHOHCH₂O formation has a lower Gibbs free energy change compared to *CH₂OCH₂O and *CHOCH₂OH. In the next step, the formed *CHOHCH₂O species can be further reduced to three possible intermediates including *CH₂OHCH₂O, *CHCH₂O + H₂O or *CHOHCH₂OH. Based on the Gibbs free energy diagrams given in Fig. 5(c) and S6(d), (e),† it can be observed that the formation of the *CH₂OHCH₂O intermediate is thermodynamically much more favorable (ΔG = −0.60 (Ti(Nb)-ATA) and −0.42 eV (Ti(Ta)-ATA)), whereas the formation of *CHCH₂O + H₂O and *CHOHCH₂OH intermediates is associated with large free energy barriers. Subsequently, the resulting *CH₂OHCH₂O species can either be reduced to *CH₂OHCH₂OH species which is thermodynamically unfavorable, or to *CH₂CH₂O and a H₂O molecule which is thermodynamically favorable. The ΔG values for *CH₂OHCH₂O reduction to *CH₂CH₂O + *H₂O on Ti(Nb)-ATA and Ti(Ta)-ATA are calculated to be 0.40 and 0.74 eV, respectively. After the release of H₂O molecule, the remaining *CH₂CH₂O species can be further reduced by proton coupled electron transfer to produce the *CH₂CH₂OH species. As shown in Fig. 5(b), remarkably, the Gibbs free energy change for this hydrogenation process increases by 1.12 and 1.62 eV for Ti(Nb)-ATA and Ti(Ta)-ATA, respectively. In the last step, further reduction of *CH₂CH₂OH generates either *CH₂CH₂ + H₂O or *CH₃CH₂OH. Unlike Ti-ATA, Ti(Zr)-ATA, and Ti(Hf)-ATA, it is clear that the generation of *CH₃CH₂OH (*C₂H₅OH) on Ti(Nb)-ATA and Ti(Ta)-ATA is energetically more favorable than the formation of *CH₂CH₂ (*C₂H₄), as shown in Fig. S6(d) and (e),† and the corresponding ΔG values for *C₂H₅OH formation on these two catalysts are −0.78 and −0.54 eV, respectively. It is worth noting that there is a small uphill energy change involved in the desorption of *CH₃CH₂OH species on Ti(Nb)-ATA; however, the *CH₃CH₂OH desorption is not a highly energy-consuming step compared to the preceding step of *CH₂CH₂O reduction to *CH₂CH₂OH with a large positive ΔG value, as illustrated in Fig. 5(c). To sum up, for the reduction of CO₂ to the preferred C₂H₅OH product on Ti(Nb)-ATA and Ti(Ta)-ATA catalysts, the optimal pathway is identified as *CO₂ → *OCHO → *OCHOH → *CHO → *CHOCHO → *CHOCH₂O → *CHOHCH₂O → *CH₂OHCH₂O → *CH₂CH₂O + H₂O → *CH₂CH₂OH → *CH₃CH₂OH → *+C₂H₅OH, in which the rate-determining step is found to be the reduction of *CH₂CH₂O → *CH₂CH₂OH as it involves the highest free energy change (ΔG = 1.12 (Ti(Nb)-ATA) and 1.62 eV (Ti(Ta)-ATA)) among all of the elementary steps.

Moreover, we also calculated the kinetic barrier of the rate-determining step in the optimal energy pathway for CO₂ reduction to ethanol on Ti(Nb)-ATA and Ti(Zr)-ATA, as shown in Fig. S7.† In addition to considering the energy barrier of the rate-determining step, the corresponding reaction rate constant was also calculated and is provided in Table S5.† The results show that the CO₂ reduction has an obvious kinetic advantage on Ti(Nb)-ATA compared to Ti(Zr)-ATA based on both the Gibbs free energy barrier and reaction rate constant calculations, further consolidate the above screening results.

Since the adsorption and activation of CO₂ is an extremely important initial step in CO₂ reduction, as demonstrated in the previous studies,^19,20,41 here we correlated the adsorption energy of CO₂ with the limiting free energy change for the generation of C₂ products (C₂H₄ and C₂H₅OH) on the four metal-modified Ti(M)-ATA (M = Zr, Hf, Nb, Ta) catalysts, and plotted the corresponding volcano type curve as displayed in Fig. 6(a). The calculation results show that the Ti(Nb)-ATA catalyst has the lowest limiting free energy change and a moderate CO₂ adsorption intensity, positioning it at the top of the volcano-type curve. In addition, we note that the stability of the *OCHOH intermediate affects the next step of dehydration to form the key C–C coupling intermediate of *CHO, and thus we plotted the volcano type curve of G_*OCHOH and the limiting free energy change on the four metal-modified Ti(M)-ATA (M = Zr, Hf, Nb, Ta) catalysts, as displayed in Fig. 6(b). The results show that the Ti(Nb)-ATA catalyst remains at the top of the volcano curve, further indicating that the moderate stability of the *OCHOH intermediate over Ti(Nb)-ATA is quite favorable for the subsequent generation of the *CHO intermediate. Therefore, the Ti(Nb)-ATA catalyst would be the best candidate for C₂ production among the four candidates studied. These results further clarify that the adsorption strength of the CO₂ reactant and key reaction intermediates (e.g.*OCHOH) directly affects the CO₂ reduction activity, and the relatively mild bonding of *CO₂ and *OCHOH over Ti(Nb)-ATA is responsible for the highest activity of Ti(Nb)-ATA among the screened candidate catalysts. This finding is consistent with our previous studies on CO₂ photoreduction over other Ti-MOF-based catalytic materials.^19,20

Fig. 6 Volcano-type curves between (a) CO₂ adsorption energy, (b) G_*OCHOH and limiting free energy change for C₂ products generation on Ti(M)-ATA (M = Zr, Hf, Nb, Ta) catalysts.

In addition, we note that CO₂ molecules can be adequately activated when the doped nodal metals belong to the IVB and VB groups and could provide favorable active sites for CO₂ adsorption and activation. For IVB group metal-modified Ti-ATA catalysts, such as Ti(Zr)-ATA and Ti(Hf)-ATA, CO₂ photocatalytic reduction tends to generate the C₂H₄ product and follows similar reaction pathways. In contrast, for VB group metal-modified Ti-ATA catalysts, such as Ti(Nb)-ATA and Ti(Ta)-ATA, CO₂ photoreduction to the C₂H₅OH product is energetically more preferred. The calculated limiting free energy change (ΔG_L = 1.12 eV) over Ti(Nb)-ATA for C₂H₅OH generation is even smaller than that (ΔG_L = 1.37 eV) over the dual metal-atom-modified Cu–In/Ti-BPDC catalyst which was identified to be a good catalyst for C₂ (C₂H₄) generation in our previous DFT work.²⁰ Therefore, Ti(Nb)-ATA can be proposed as a promising candidate catalyst for the photocatalytic reduction of CO₂ to C₂ products, especially C₂H₅OH. Overall, the calculation results indicate that the ATA ligand with a Ti–O–Nb dual-metal-node structure as well as with a broad absorption spectrum effectively improves the photocatalytic efficiency for CO₂ reduction to C₂ products, which are also essential for regulating the catalytic activity and product selectivity of Ti-MOF-based photocatalysts.

The CO₂ photoreduction was typically carried out in aqueous solution. Under the same conditions, *H species could be adsorbed on the active site and reduced by the proton–electron pair (H⁺ + e⁻), which would affect the photocatalytic efficiency of CO₂ reduction. Here, we calculated the Gibbs free energy diagram for the hydrogen evolution reaction (HER) on the Ti(Nb)-ATA catalyst, as shown in Fig. S8(a),† which shows that the *H species is preferentially adsorbed on the Nb atom in the node of Ti(Nb)-ATA. It is worth noting that the limiting free energy change for the HER on Ti(Nb)-ATA is calculated to be 1.29 eV, which is higher than that (1.12 eV) for CO₂ reduction; thus the competitive effect of HER is minor. In addition, we also calculated the Gibbs free energy diagrams for CO and HCOOH production from CO₂ reduction on Ti(Nb)-ATA, as shown in Fig. S8(b),† and the results show that the stability of the *CHO intermediate is higher than that of CO and HCOOH products; thus the ethanol formation pathway through the CHO* intermediate is mainly considered in this work.

3.3 Electronic structure and optical property analysis of Ti-ATA and Ti(M)-ATA (M = Zr, Nb, Hf, Ta)

In order to gain fundamental insights into the optical properties of the node-modified Ti-ATA catalysts, we further analyzed the electronic structures of pristine Ti-ATA and the best active catalyst identified, Ti(Nb)-ATA, by calculating the total and projected density of states (TDOS and PDOS) and band structures, as shown in Fig. 7(a) and (b). The results for other catalysts (Ti(M)-ATA (M = Zr, Hf, Ta)) are provided in Fig. S9(a) and (b).† The indirect band gap of unmodified Ti-ATA is 1.54 eV. Compared to the unmodified Ti-ATA catalyst, the band gap of metal-modified Ti(Nb)-ATA (1.25 eV) and Ti(Ta)-ATA (1.31 eV) (VB metals in the periodic table) becomes significantly narrower and several new energy bands appear near the Fermi energy level. By contrast, the band gaps of Ti(Zr)-ATA (1.77 eV) and Ti(Hf)-ATA (1.79 eV) (IVB metals in the periodic table) are wider and there are no new energy bands near the Fermi energy level. Importantly, a new energy band consisting of transition metal d orbitals emerges near the Fermi energy level of Ti(Nb)-ATA. This energy band has an important contribution near the Fermi energy level and can be used as a doping energy level to promote the separation of photoelectrons and holes by trapping electrons, thus increasing the stability of reaction intermediates and lowering the barriers in the reduction of CO₂ to C₂H₅OH via photocatalysis. Charge separation is a key step in photocatalysis to convert light energy into chemical energy, and it directly impacts the performance of the catalyst. In order to further improve the activity and selectivity of CO₂ photoreduction to ethanol, we constructed bimetallic active sites by introducing a second metal component into the metal node of Ti-ATA, effectively regulated the energy band structure of the MOF material, and optimized the separation efficiency of photogenerated electrons. To further analyze the electron transfer behavior during the photocatalytic process, we further calculated the valence band maxima (VBM) and conduction band minima (CBM) of Ti-ATA and Ti(M)-ATA (M = Zr, Hf, Nb, Ta), as shown in Fig. 7(c) and S9(c).† For the unmodified Ti-ATA catalyst, the VBM is mainly located on the π-bond orbitals of the ligand, whereas the CBM is mainly located on the Ti-oxo cluster. For Ti(Nb)-ATA, electrons are transferred from the excited ATA ligand to the Nb centers, and subsequently, the Nb ions transfer electrons to Ti, forming photocatalytically active Nb ions. The calculation results suggest that the modified Ti nodes may act as electronic mediators to facilitate the transfer of electrons from the ATA ligand to the Nb site. The emergence of new energy levels in Ti(Nb)-ATA effectively inhibits the rapid electron–hole complexation and provides a more stable platform for photogenerated electrons with superior photocatalytic performance for CO₂ reduction to C₂ products. In addition, according to Bader charge analysis (Fig. S10†), the doped metal atoms transfer electrons to bridging oxygen atoms between the Ti atom and doped metal atoms, which leads to the accumulation of negative charges around the doped metal center. This result indicates that there is a significant charge transfer in the Ti–O–M (M = Zr, Hf, Nb, Ta) structure, with the Nb dopant exhibiting the greatest charge transfer, as evidenced in Fig. S10.† Thus, doping another metal into the Ti-MOF can result in a significant charge redistribution at the nodes of MOF materials. Overall, the electronic structure characteristics of pristine Ti-ATA and metal-modified Ti(M)-ATA (M = Zr, Hf, Nb, Ta) are generally consistent with the predicted activity of these candidate catalysts.

Fig. 7 The calculated (a) total and projected density of states (TDOS and PDOS), with the Fermi energy level set to zero. (b) Energy band structures of pristine Ti-ATA and metal-modified Ti(Nb)-ATA. (c) The calculated valence-band maximum (VBM) (left side) and conduction-band minimum (CBM) (right side) of pristine Ti-ATA and metal-modified Ti(Nb)-ATA.

Finally, we calculated the optical absorption spectrum of Ti-ATA and Ti(Nb)-ATA, as shown in Fig. 8(a) and (b). The optical absorption of Ti-ATA and Ti(Nb)-ATA is mainly located in the ultraviolet region. However, the light absorption activity of the modified nodes was slightly improved. Due to the narrow band gap, the response of Ti(Nb)-ATA in visible and infrared bands is slightly stronger than that of Ti-ATA. The absorption spectra of other Ti(M)-ATA catalysts (M = Zr, Hf, Ta) were also calculated, as shown in Fig. S11.† In addition, the work function (Φ) is an important parameter to measure the electron gain/loss ability of materials, which is defined as the minimum energy required for electrons to escape from the interior of materials to the surface.^42,43 In DFT calculations, the work function is obtained from the electrostatic potential distribution, which is calculated using the equation Φ = E_V – E_f, where E_v and E_f are the potentials of the vacuum energy level and the Fermi energy level, respectively. As shown in Fig. 8(c) and (d), the work function of Ti-ATA is calculated to be 4.23 eV (Fig. 8(c)). After the introduction of Nb atoms into the Ti node, the work function is reduced to 2.91 eV (Fig. 8(d)). The work function of other Ti(M)-ATA catalysts is also lower than that of pristine Ti-ATA (Fig. S12†). The values of the work function, vacuum energy level and Fermi energy level for Ti-ATA and Ti(M)-ATA are provided in Table S6.† Further analysis reveals that the decrease in work function is caused by an up shift in the Fermi level. As mentioned above, the electrons from the Nb atom transfer to the O atom at the node, and the Fermi level of Ti(M)-ATA is enhanced due to Nb–O bonding. The reduced work function is conducive to the transfer of electrons from Ti(M)-ATA to the surface-adsorbed species in the photocatalytic reaction, thereby improving photocatalytic activity.

Fig. 8 The optical absorption spectra for (a) Ti-ATA and (b) Ti(Nb)-ATA. The electrostatic potential profiles of (c) Ti-ATA and (d) Ti(Nb)-ATA.

4 Conclusions

In summary, by means of DFT calculations, we systematically investigated the structure and stability of a series of metal node-modified Ti(M)-ATA (Mn, Fe, Co, Ni, Cu, Zn, Zr, Nb, Mo, Ru, Rh, Pd, Hf, Ta, W, Os, Ir, Au) MOF catalysts and their applications in the photocatalytic reduction of CO₂ to C₂ chemicals. Our results show that Ti(M)-ATA (M = Zr, Hf, Nb, Ta) can sufficiently activate CO₂, and the energetically optimal pathway for the reduction of CO₂ to C₂ products has been determined by reaction route screening. Over Ti(Nb)-ATA, Ti(Ta)-ATA, Ti(Zr)-ATA and Ti(Hf)-ATA, the two *CHO species undergo C–C coupling to form *CHOCHO, an important C₂ intermediate. Ti(Nb)-ATA and Ti(Ta)-ATA tend to generate ethanol, while Ti(Zr)-ATA and Ti(Hf)-ATA are more selective to ethylene.

Remarkably, Ti(Nb)-ATA is the best photocatalyst for C₂H₅OH production among all the candidates studied, and the corresponding optimal reduction pathway is identified as: *CO₂ → *OCHO → *OCHOH → *CHO → *CHOCHO → *CHOCH₂O → *CHOHCH₂O → *CH₂OHCH₂O → *CH₂CH₂O + H₂O → *CH₂CH₂OH → *CH₃CH₂OH → *+C₂H₅OH, in which the hydrogenation of *CH₂CH₂O species is the rate-determining step for the overall reaction with a limiting free energy change of 1.12 eV. By analyzing the correlation, it is found that the catalytic activity of these metal node-modified catalysts is highly dependent on their binding strength to the CO₂ reactant and key reaction intermediates (e.g. *OCHOH).

In addition, the analysis of electronic and optical properties indicates that the altered energy band structure and charge transfer characteristics at the bimetallic node of Ti(Nb)-ATA are responsible for its superior catalytic activity towards CO₂ reduction to C₂H₅OH compared to pristine Ti-ATA and other metal-modified Ti(M)-ATA candidates. The facile substitution or doping of metals in Ti-MOFs and their structural flexibility and diversity enable us to design a variety of Ti-based MOF photocatalysts with desirable properties. Our findings will stimulate further in-depth experimental studies of Ti-based MOF materials and open up new avenues for developing Ti-MOF-based catalysts for CO₂ photoreduction, especially for the synthesis of C₂ chemicals.

Data availability

The data supporting this article have been included as part of the ESI.†

Author contributions

Shuang Wang: conceptualization, methodology, data acquirement, writing – original draft preparation. Xiaowa Nie: conceptualization, supervision, methodology, data analysis, writing – review & editing. Chunshan Song: investigation, writing – review & editing. Xinwen Guo: supervision, resources, project administration, writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 22278056), the Fundamental Research Funds for the Central Universities (No. DUT22LAB602), and the Liaoning Revitalization Talents Program (No. XLYC3008032). We also acknowledge the Dalian University of Technology Supercomputing Center, Tianjin Supercomputing Center, and Beijing Paratera Co., Ltd. Supercomputing Center for providing computing resources for this work.

References

1. X. Nie, W. Li, X. Jiang, X. Guo and C. Song, Adv. Catal., 2019, 65, 121–233.
2. X. Chang, T. Wang, P. Yang, G. Zhang and J. Gong, Adv. Mater., 2019, 31, 1804710.
3. W. Gao, S. Liang, R. Wang, Q. Jiang, Y. Zhang, Q. Zheng, B. Xie, C. Ying Toe, X. Zhu, J. Wang, L. Huang, Y. Gao, Z. Wang, C. Jo, Q. Wang, L. Wang, Y. Liu, B. Louis, J. Scott, A. Roger, R. Amal, H. Heh and S. Park, Chem. Soc. Rev., 2020, 49, 8584–8686.
4. N. Liu, Y. Zhao, S. Zhou and J. Zhao, J. Mater. Chem. A, 2020, 8, 5688–5698.
5. H. Shi, H. Wang, Y. Zhou, J. Li, P. Zhai, X. Li, G. Gurzadyan, J. Hou, H. Yang and X. Guo, Angew. Chem., Int. Ed., 2022, e202208904.
6. H. Liu, B. Sun, Z. Li, D. Xiao, Z. Wang, Y. Liu, Z. Zheng, P. Wang, Y. Dai, H. Cheng and B. Huang, Angew. Chem., Int. Ed., 2024, 63, e202410596.
7. Y. Zhao, N. Liu, S. Zhou and J. Zhao, J. Mater. Chem. A, 2019, 7, 16294–16303.
8. H. Zhou, Y. Hua, Y. Zhang, W. Jiang and J. Di, Sep. Purif. Technol., 2024, 16, 1383–5866.
9. S. Ma, W. Han, W. Han, F. Dong and Z. Tang, J. Mater. Chem. A, 2023, 11, 3315–3363.
10. J. Zhang, W. Yang, M. Zhang, H. Wang, R. Si, D. Zhong and T. Lu, Nano Energy, 2021, 80, 105542.
11. L. Wang, S. You, Y. Gong, J. Gu, J. Zhang, G. Shan, B. Zhu, W. Yang, C. Sun, X. Wang and Z. Su, Green Chem., 2024, 26, 9415.
12. M. Zhang, D. Zhang, X. Jing, B. Xu and C. Duan, Angew. Chem., Int. Ed., 2024, 63, e202402755.
13. R. Li, W. Zhang and K. Zhou, Adv. Mater., 2018, 30, 1705512.
14. Y. Zhao, Z. Shao, Y. Cui, K. Geng, X. Meng, J. Wu and H. Hou, Small, 2023, 19, 2300398.
15. X. Shi, L. Cao, M. Chen and Y. Huang, Chin. Chem. Lett., 2022, 33, 5023–5029.
16. H. Yang, G. Liu, L. Zheng, M. Zhang, Z. Guan, T. Liu and J. Yang, Appl. Catal. B Environ., 2024, 359, 124491.
17. B. Qiao, A. Wang, X. Yang, L. F. Allard, Z. Jiang, Y. Cui, J. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641.
18. X. He, X. Gao, X. Chen, S. Hu, F. Tan, Y. Xiong, R. Long, M. Liu, E. Tse, F. Wei, H. Yang, J. Hou, C. Song and X. Guo, Appl. Catal. B Environ., 2023, 327, 122418.
19. S. Wang, X. Nie, J. Lin, F. Ding, C. Song and X. Guo, ChemSusChem, 2024, e202301619.
20. S. Wang, X. Nie, C. Song and X. Guo, CCS Chem., 2024, 6, 3018–3029.
21. F. Xie, C. Bie, J. Sun, Z. Zhang and B. Zhu, J. Mater. Sci. Technol., 2024, 170, 87–94.
22. J. Olowoyo, U. Saini, M. Kumar, H. Valdes, H. Singh, M. Omorogie, J. Babalola, A. Vorontsov, U. Kumar and P. Smirniotis, J. CO2 Util., 2020, 42, 101300.
23. X. Cheng, Y. Gu, X. Zhang, X. Dao, S. Wang, J. Ma, J. Zhao and W. Sun, Appl. Catal. B Environ., 2021, 298, 120524.
24. I. Tsuji, H. Kato, H. Kobayashi and A. Kudo, J. Am. Chem. Soc., 2004, 126, 13406–13413.
25. D. Sun, W. Liu, M. Qiu, Y. Zhang and Z. Li, Chem. Commun., 2015, 51, 2056–2059.
26. X. Wu, L. Gagliardi and D. Truhlar, J. Am. Chem. Soc., 2018, 140, 7904–7912.
27. M. Ming, Y. Wang, W. Tao, W. Shi, D. Zhong and T. Lu, Green Chem., 2023, 25, 6207.
28. K. Mori, J. Matsuo, Y. Kondo, H. Hata and H. Yamashita, ACS Appl. Energy Mater., 2021, 4, 11634–11642.
29. D. Lee, S. Lee, I. Choi and M. Kim, Smart Mol., 2024, 2, e20240002.
30. X. He, Y. Ding, Z. Huang, M. Liu, M. Chi, Z. Wu, C. Segre, C. Song, X. Wang and X. Guo, Angew. Chem., Int. Ed., 2023, e202217439.
31. X. Cheng, X. Dao, S. Wang, J. Zhao and W. Sun, ACS Catal., 2021, 11, 650–658.
32. G. Kresse and J. Furthmüller, Phys. Rev. B:Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
33. G. Kresse and J. Furthmüller, Comp. Mater. Sci., 1996, 6, 15–50.
34. S. Grimme, J. Comput. Chem., 2006, 27, 1787–1799.
35. M. Yu, S. Yang, C. Wu and N. Marom, npj. Comput. Mater., 2020, 6, 180.
36. L. Gong, D. Zhang, C. Lin, Y. Zhu, Y. Shen, J. Zhang, X. Han, L. Zhang and Z. Xia, Adv. Energy Mater., 2019, 9, 1902625.
37. S. Li, Z. Lu, Y. Zhang, D. Ma and Z. Yang, Phys. Chem. Chem. Phys., 2017, 19, 9007–9015.
38. G. Jajko, P. Gryta, P. Kozyra, M. Szufla, D. Matoga, D. Majda and W. Makowski, J. Phys. Chem. C, 2022, 126, 9185–9194.
39. S. Wang, Y. Liu and J. Zhao, Phys. Chem. Chem. Phys., 2020, 22, 17639–17645.
40. Y. Sun, S. Wang, J. Jia, Y. Liu, Q. Cai and J. Zhao, J. Mater. Chem. A, 2022, 10, 14460.
41. X. Gao, Z. Wang, Q. Huang, M. Jiang, S. Askari, N. Dewangan and S. Kawi, Catal. Today, 2022, 402, 88–103.
42. Y. Huang, K. Dai, J. Zhang and G. Dawson, Chin. J. Catal., 2022, 43, 2539–2547.
43. B. Zhu, H. Tan, J. Fan, B. Cheng, J. Yu and W. Ho, J. Mater., 2021, 7, 988–997.

---

Footnotes

† Electronic supplementary information (ESI) available: Fig. S1–S8 and Tables S1–S4. See DOI: https://doi.org/10.1039/d5ta03415a

‡ These authors contributed equally to this work.

---