Metal-organic frameworks (MOFs) composed of metal-oxo clusters and polydentate organic linkers represent an intriguing class of porous crystalline material with many attractive properties such as porosity, large surface area, and structural and functional tunability [1]. Because of these unique advantages, MOFs have been widely investigated for gas storage and separation, sensors, drug delivery, and catalysis [2]. Some photo-responsive MOFs exhibiting semiconductor behavior have been developed as photocatalysts for promising solar energy conversion. In particular, solar-driven water splitting is of great significance when considering that hydrogen evolution from water provides an environmentally friendly and sustainable alternative to fossil fuels [3].

To date, some MOF-based photocatalysts have been explored for hydrogen production. These contain UiO-66 [4], NH₂-UiO-66 [4], NH₂-MIL-125(Ti) [5], and Al-PMOF [6]. Although major progress has been achieved, the selection of metals is often confined based on the robustness and conduction band (CB) position of the as-obtained MOF material [7, 8]. Consequently, new photo-responsive MOFs composed of other metals with good robustness and a suitable band structure are desired. Bismuth, as a nontoxic and earth-abundant element, is a promising candidate for the construction of stable MOFs, as it has a relatively high valence state and flexible coordination geometry [9, 10]. However, the synthesis of bismuth-based MOFs remains challenging because of their structural sensitivity to the reaction conditions such as solvents, temperature, and reaction time. The complicated structural behavior of Bi³⁺ ions during the solvothermal synthesis makes controlling the process of crystal growth difficult [11]. Consequently, reports on the synthesis of bismuth-based MOFs with photocatalytic activity are limited [12–14]. To the best of our knowledge, no bismuth-based MOFs have been synthesized for photocatalytic water reduction.

In this study, we report on a new stable bismuth-organic framework (Bi-TBAPy) with a 3D framework structure by employing 1, 3, 6, 8-tetrakis(p-benzoic acid)pyrene (H₄TBAPy) as the organic linker. The as-obtained Bi-TBAPy is demonstrated to possess good stability and suitable band edge positions for photocatalytic water reduction. Based on our optimization, Bi-TBAPy exhibits the optimal H₂ evolution rate of 140 μmol h^‒1 g^‒1 when triethanolamine (TEOA) is used as the sacrificial reagent, demonstrating its promising future in solar fuel production.

The structure of a Bi-TBAPy single crystal was investigated by single-crystal X-ray diffraction (XRD) analysis, and details of the crystal structure and refinement data are provided in Table S1. As given in Fig. S1, the Bi-TBAPy is crystallized in monoclinic space group C2/c to form a 3D rod-like structure. The typical structure is introduced in Fig. 1(a), where the Bi atom is coordinated with eight oxygen atoms from four linkers with a bond length of ~2.435 Å. A dimethylamine cation serving as the counter ion is also coordinated with the Bi atom to maintain the charge balance, and the bond length of Bi–N is ~2.923 Å. Each TBAPy^4‒ linker coordinated to four bismuth centers shapes the backbone of Bi-TBAPy into a 3D framework (Fig. 1(b) and (c)). Two types of 1D channels with different shapes exist along the a axis (Fig. 1(c)). One channel has a rhomboid shape and is occupied by dioxane; the other has a quadrilateral shape and is occupied by dimethylamine solvent molecules. The length of the two sides of these two windows are both 7.37 × 11.65 Ų, although their shapes are different. The total solvent-accessible volume is approximately 8.4% upon solvent removal, which is calculated with PLATON software [15]. The BET surface area of Bi-TBAPy is measured to be 594 m² g^‒1. The coordination effect between Bi ions and H₄TBAPy linkers can be further confirmed by the results of FT-IR measurement (Fig. S2), where the hydroxyl vibration peak of –COOH in the range of 3000–3400 cm^‒1 disappears during the shift of the carbonyl vibration peak in the range of 1620–1750 cm^‒1. The experimental XRD pattern (Fig. 1(d)) of Bi-TBAPy conforms well with the simulated pattern, which is generated based on single crystal structure analysis, thus demonstrating a good phase purity of the bulk material.

To examine the physical and/or chemical stability of Bi-TBAPy, thermogravimetric analysis (TGA) was first conducted, with results given in Fig. S3. The results reveal that the Bi-TBAPy exhibits good thermal stability with the framework collapsed at approximately 350 ℃. Subsequently, the Bi-TBAPy sample was also demonstrated to have good solvent stability when considering the fact that no obvious change of XRD patterns could be observed in the samples before and after impregnation in water, DMF, dioxane, acetonitrile, methylene dichloride, or isopropanol for 12 h at room temperature (Fig. S4). On the whole, the as-obtained Bi-TBAPy sample has an unexpected thermal and solvent stability, demonstrating its good potential as a photocatalyst.

Fig. 2(a) shows its typical UV/vis DRS, based on which the Bi-TBAPy has a visible light response with an absorption edge of ca. 465 nm. The band gap of Bi-TBAPy was then estimated to be ca. 2.67 eV from the Kubelka-Munk function. To evaluate further the band edge positions of Bi-TBAPy, a Mott-Schottky measurement was conducted. Fig. 2(b) clearly reveals the n-type semiconductor property of Bi-TBAPy, and the flat band potential was calculated to be ca. –0.15 V vs. RHE. As the bottom of the CB of an n-type semiconductor is generally more negative by ca. 0.1 eV than the flat band potential [16], the CB of Bi-TBAPy was estimated to be ca. –0.25 eV. Together with the bandgap value estimated from the UV/vis DRS (Fig. 2(a)), the valence band (VB) was thus determined to be located at 2.42 eV. The estimated band edge positions of Bi-TBAPy are shown in the insert of Fig. 2(b), in which the CB exhibits sufficient potential in the water reduction process.

Encouraged by the good stability and suitable CB position of Bi-TBAPy, we examined the performance of Bi-TBAPy for water reduction by using TEOA as the hole scavenger. The photocatalytic water reduction activity of the Bi-TBAPy sample free of a cocatalyst was very low (Fig. S5). As shown in Fig. 2(c), the loading of a suitable cocatalyst is desired to promote the H₂ evolution rate, and the platinum modified sample appeared to promote the H₂ evolution rate most efficiently. The difference in the promotion effect originating from the cocatalyst species may derive from their distinct ability at trapping electrons from the CB of Bi-TBAPy and/or activating protons for H₂ evolution [17]. A volcanic-type curve of H₂ evolution dependence on the Pt content was observed (Fig. S5), and the optimal H₂ evolution rate of 140 μmol h^‒1 g^‒1 appeared for the sample with 2.0 wt% Pt loaded. The XPS pattern of Bi-TBAPy after hydrogen evolution verified the loading of metallic Pt based on the binding energies of Pt 4f_7/2 and Pt 4f_5/2 peaks, as shown in Fig. S6 [18]. The loading of a suitable amount of platinum could trap the photogenerated electrons and accelerate the conversion of water molecules, whereas an excess amount probably leads to the shielded light absorption of MOFs, thus resulting in decreased photocatalytic H₂ evolution performance. For comparison, the water reduction performance of 2 wt% Pt/H₄TBAPy was also examined under the same conditions as the 2 wt% Pt/Bi-TBAPy. The negligible hydrogen evolution indicates that the formation of a framework by incorporating Bi ions is vital to the photocatalytic performance of Bi-TBAPy. The typical time course of the photocatalytic water reduction on the optimized sample is shown in Fig. 2(d), in which the linear increase of H₂ evolution as a function of the reaction time demonstrates the good photochemical stability of Bi-TBAPy. In addition, the XRD patterns of Bi-TBAPy before and after H₂ evolution were not obviously altered (Fig. S7), also indicating the good stability of Bi-TBAPy during the photocatalytic reaction. Furthermore, the transient photocurrent response on the Bi-TBAPy photoelectrode in a three-electrode setup was conducted by using a 0.5-M sodium sulfate solution as an electrolyte solution. An obvious photocathode current could be observed from the chopped photocurrent profile of Bi-TBAPy (Fig. S8), indicating the generation of electron-hole pairs in Bi-TBAPy during light irradiation and the capacity for photoelectrochemical water reduction by photoinduced electrons.

To obtain insights into the mechanism of the charge transfer during the photocatalytic process, Bi 4f XPS patterns of Bi-TBAPy before and after H₂ evolution were measured and compared. As shown in Fig. 3(a), in addition to the characteristic peaks of Bi³⁺ ions located at 164.2 and 158.9 eV, two new peaks centered at 162.2 and 156.9 eV were deconvoluted after 3 h of irradiation and were consistent with the characteristic peaks of the lower valence state of the Bi ion [19, 20]. It should be noted that although the lower valence state of Bi ions was generated, the framework of Bi-TBAPy was well maintained based on the XRD patterns shown in Fig. S7, indicating the robustness of Bi-TBAPy. Based on the XPS analysis, a ligand-to-metal charge transfer (LMCT) process was observed to occur in Bi-TBAPy, where the photogenerated electrons transferred from the excited TBAPy group to Bi³⁺, resulting in partial reduction of bismuth ions. The LMCT mechanism is always observed for MOFs as photocatalysts because of the intrinsic charge separation process. As a result, steady-state photoluminescence (PL) measurements were conducted to evaluate the influence of this charge transfer process. The results are shown in Fig. 3(b). H₄TBAPy exhibits the strongest steady-state PL intensity, revealing that a great carrier recombination process occurs in organic linkers, and the construction of Bi-TBAPy could suppress this process, thus resulting in decreased PL intensity of pristine Bi-TBAPy. Compared to the pristine sample, the considerably diminished steady-state PL intensity of Bi-TBAPy after irradiation confirmed the separation of photoinduced carriers as a result of the LMCT process, which in turn inhibited the process of charge recombination. The proposed mechanism for photocatalytic H₂ over Bi-TBAPy is illustrated in Fig. 4. First, the TBAPy groups in Bi-TBAPy absorb light to generate electrons and holes. Then, the excited electrons can be utilized through two pathways. In one pathway, electrons first transfer to the near Bi³⁺ ions by following the LMCT mechanism and then transfer to the cocatalysts. In the other pathway, the excited electrons in TBAPy groups transfer directly to Pt particles. Subsequently, the deposited Pt act as active sites to perform the catalytic reduction of water while the photogenerated holes are simultaneously consumed by the sacrificial reagent TEOA.

In summary, we synthesized a novel 3D bismuth-organic framework (Bi-TBAPy) with good stability and suitable band edge positions for promising photocatalytic H₂ evolution. The influence of cocatalysts and the amount of loading on the performance of water reduction were investigated and discussed, and the optimal H₂ evolution rate of 140 μmol h^‒1 g^‒1 was obtained. The LMCT mechanism that occurred in Bi-TBAPy during the photocatalytic process was confirmed by XPS spectra, and this intrinsic charge separation process was favorable to the photocatalytic reactions because the charge recombination was suppressed. Our study not only adds a new member to the family of MOF material, it also offers new insights into the design and manipulation of MOFs for applications in photocatalytic water splitting.

Acknowledgments

This work was financially supported by National Natural Science Foundation of China (21633009, 21522306, 21633010), jointed project between DICP and QIBEBT(UN201805), and Dalian Science Foundation for Distinguished Young Scholars (2017RJ02). F. Zhang thanks the priority support from the "Hundred Talents Program" of Chinese Academy of Sciences.