Metal-organic cages (MOCs) are usually assembled from metal nodes (ions or clusters) and polydentate organic linkers, which are also known as metal-organic polyhedra or metal-organic supercontainers [1–3]. Although the composition and assembling principle of MOCs is similar to that of metal-organic framework (MOFs) [4–6], MOCs exhibit different characteristics from MOFs in terms of structure, dimension, molecular weight, solubility, and stability, etc. [7–9]. Polyhedral MOCs can be categorized as several subclasses, including tetrahedron, octahedron, cube, dodecahedron, icosahedron, cuboctahedron, truncated regular polyhedral, and prisms [10–16]. The properties of MOCs are highly dependent on their compositional metals, organic ligands, and molecular structures. Thus, delicate control of these factors could intentionally develop MOCs materials with intriguing functions and applications in molecular devices [17], catalysis [18,19], host-guest chemistry [20,21], gas/solution separation and storage [22–25], and photo-/electro-catalysis reaction [26].

Polyoxometalates (POMs), a unique class of nanoscale metal oxygen–anion clusters, are formed through the condensation of early transition metalates in their high oxidation states (Mo, W, V, Nb, Ta, etc.) [27–31]. POMs are desirable inorganic building blocks owing to their rich redox chemistry, tunable electronic structure, and thermally and oxidatively stability [32,33]. The design and synthesis of POM-organic cages (POCs) combined the advantages of their constituent POMs and organic linkers to generate nanocages with unprecedented properties are highly desired. To date, most POCs were constructed by polyoxovanadates and polyoxotungstates as secondary building units (SBUs) [12,34–38]. The construction of POCs using transition-metal-substituted POMs as building nodes have been far more less explored. Some representative examples include: Yang and co-workers reported the first polyoxotungstates Ni₆-substituted {PW₉} cluster-based POC, which linked by rigid 1,3,5-benzenetricarboxylate ligand under hydrothermal conditions [11] (Scheme 1, Ni₆SiW₉ + L₇). Then, Fang and co-workers have constructed a family of discrete POCs with the general formula POM_2nL_3n using Ni₄SiW₉ as nodes and rigid dicarboxylate as linkers (Scheme 1, Ni₄SiW₉ + L₁/L₂/L₃/L₅/L₆) [39], in which the photocatalytic CO₂ reduction activities of these POCs have been investigated. More recently, Zhan group reported the self-assembly of anionic POC architecture {Ni₆L₃SiW₉}₄ comprising hexanuclear nickel-substituted POM clusters and p-aminobenzoic acid ligands (Scheme 1, Ni₆SiW₉ + L₄) [40].

In addition to the structural diversity of POMs, transition-metal-substituted POMs have been energetically investigated as catalysts for light-driven hydrogen evolution [41–51]. In this scenario, we expected that the POCs constructed by transition-metal-substituted POMs nodes and organic ligands might exhibit a synergistic cooperation between the constituent POMs and distinct cage structures for better catalytic performance. Moreover, their unique cage architecture may work as perfect nanoreactor with efficient accessibility to catalytic components.

In this context, we report herein the successful construction of a tetrahedral POC, K₃Na₁₇H₁₂[(C₄H₆O₆)₆[Ni₄(OH)₃(A-α-SiW₉O₃₄)]₄]·96H₂O (Ni₁₆L₆(SiW₉)₄), using Ni₄SiW₉ as the SBUs and L-(+)-tartaric acid (L) as linkers. The resulting POC tetrahedron has been firstly investigated as efficient catalyst for visible-light-driven hydrogen production when coupling with [Ir(coumarin)₂(dtbbpy)][PF₆] (PS) as the photosensitizer, TEOA as the sacrificial electron donor, and H₂O as the proton source.

Single crystal X-ray analyses reveal that Ni₁₆L₆(SiW₉)₄ crystallizes in the cubic space group I23 (Table S1 in Supporting information) and exhibits a POM-organic tetrahedral cage constructed by four Ni₄SiW₉ SBUs and six L linkers (Fig. 1a). The four Ni-substituted Ni₄SiW₉ SBUs are located at the vertex of the tetrahedron, while the L ligands are disposed on the six edges to form an interior cavity with a void space of about 102 ų as illustrated by a simulated yellow ball (Fig. 1a, see calculation method in Supporting information). The angles between nearby L ligands are measured as ~76° from each other, which is close to the value of the acetate-chelated derivative as reported by Mialane et al. [52]. From a different point of view, the polyoxoanion Ni₁₆L₆(SiW₉)₄ can be structurally regarded as a central {Ni₄(HO)₃}₄(C₄H₆O₆)₆} moiety capped by four lacunary A-α-{SiW₉O₃₄} ligands (Fig. 1b). In the C-symmetric Ni₄SiW₉ SBU, three crystallographically-equivalent Ni^Ⅱ ions (denoted as Ni2) in the pseudo cubane-type {Ni₄O₃} cluster are stabilized in the trivacant sites of A-α-{SiW₉O₃₄} ligand, while another one Ni^Ⅱ ion (denoted as Ni1) caps on three Ni2 atoms via three acetate bridges and three OH ligands with a C axis passing through Ni1 and the Si heteroatom (Fig. 1a). Each Ni₄SiW₉ SBU is coordinated by three organic ligands, all hydroxyl groups of L-(+)-tartaric acid ligands in the molecular structure point into the exterior cavity (Fig. 1a). The bond lengths of Ni–O bonds in Ni₄SiW₉ range from 2.002 to 2.065 Å. Bond valence sum (BVS) calculations (Table S2 in Supporting information) and charge balance requirements confirm that the +2 oxidation states of Ni1 and Ni2, and the valence of W and Si are +6 and +4, respectively. BVS values of the µ₃-O atoms (pink balls in Fig. S1 in Supporting information) are calculated as about 1.14, indicating the monoprotonation of these µ₃-O atoms (Table S2). It is worth mentioning that the synthesis of Ni₄SiW₉-based POC using the organic succinic acid ligand was unsuccessful, which could be attributed to the increased rigidity of the organic chain due to the presence of hydroxyl groups in L ligand. Recently, the structurally-similar tetrahedral POC has also been reported Fang and co-workers [39], Na₃₀Ni[{(A-α-SiW₉O₃₄)Ni₄(OH)₃}₄(OOC(C₄H₂S)COO)₆]·72H₂O, in which four Ni₄SiW₉ SBUs were also used as the nodes of POC, instead six more rigid dicarboxylate ligands (2,5-thiophenedicarboxylate) were used as linkers (Scheme 1, Ni₄SiW₉ + L₃).

The molecular structure of Ni₁₆L₆(SiW₉)₄ was characterized by FT-IR spectrum. The characteristic bands in the range of 1200–400 cm^–1 can be assigned to the Si–O, W–O, and W–O–W vibrational bands (Fig. S2 in Supporting information). The IR bands in the range of 1200–400 cm^–1 also show the characteristic vibrations of L ligands as expected. In addition, the FT-IR spectra before and after counter cations exchange from Na⁺/K⁺ to TBA⁺ showed all the characteristic bands of polyoxoanion Ni₁₆L₆(SiW₉)₄, confirming the retained molecular skeleton of Ni₁₆L₆(SiW₉)₄. Also, thermogravimetric analysis showed a 22% weight loss percentage, corresponding to a total of 96 crystallization water molecules per formula unit (Fig. S3 in Supporting information). The elemental composition of Ni₁₆L₆(SiW₉)₄ complex was also characterized by scanning electron microscopy and energy dispersive X-ray spectroscopy (SEM/EDS), revealing the microscopic morphology of the Ni₁₆L₆(SiW₉)₄ crystals and the presence of Ni, W, and Si elements (Fig. S4 in Supporting information). X-ray photoelectron spectroscopy (XPS) has also been utilized to assess the presence of Si, Ni, and W elements (Fig. 2a) as well as their corresponding chemical oxidation states (Figs. 2b-d). The binding energies at 855.8 and 873.7 eV were assigned to the signals of Ni 2p_3/2 and Ni 2p_1/2, respectively (corresponding satellite peaks at 861.9 and 879.9 eV), indicating the oxidation state of Ni is +2. While the W 4f XPS signal can be deconvoluted into W 4f_5/2 and 4f_7/2 at binding energies of 37.5 and 35.8 eV, respectively, indicating a +6 oxidation state. The Si 2p signal exhibits a binding energy of 101.8 eV, corresponding to a +4 oxidation state.

The TBA salt of polyoxoanion Ni₁₆L₆(SiW₉)₄ (TBA-Ni₁₆L₆(SiW₉)₄) was subsequently employed as the molecular catalyst for visible-light-driven hydrogen evolution in a well-established three-component catalytic system using [Ir(coumarin)₂(dtbbpy)]⁺ [53] as the photosensitizer, TEOA as the sacrificial electron donor, and H₂O as the proton source. The photocatalysis was performed in a CH₃CN/DMF (v/v = 1/3) solution under a Xe-lamp light source equipped with a 400 nm cut off filter at 20 ℃. As shown in Figs. 3a-c, varying the concentration of each essential component can also greatly adjust the efficiency of H₂ production. For instance, increasing the concentration of TBA-Ni₁₆L₆(SiW₉)₄ from 0 µmol/L to 25 µmol/L leads to the enhancement of H₂ yield from ~1.98 µmol to ~321.47 µmol in 5 h (Fig. 3a), corresponding to the TONs of 4619 (5 µmol/L catalyst), 3618 (10 µmol/L catalyst), 3403 (15 µmol/L catalyst), 2677 (20 µmol/L catalyst), and 2143 (25 µmol/L catalyst), respectively (Fig. S5a in Supporting information). The optimized TON (~4619) is calculated at a TBA-Ni₁₆L₆(SiW₉)₄ concentration of 5 µmol/L, because a given catalyst at relatively low concentration may possibly work as the limiting parameter for reaching a higher TON [54]. Specifically, during photocatalytic reaction, the photosensitizer is excited to generate photogenerated electrons and holes, which are harvested by TEOA and the catalyst, respectively. When the catalyst concentration is very low, the effective numbers of electrons received by per catalyst become the rate-limiting factor for catalysis, thereby leading to a higher TON value. In other control experiment, turning the concentration of [Ir(coumarin)₂(dtbbpy)]⁺ photosensitizer (PS) from 0 to 0.3 mmol/L increased the H₂ yield 64.13 µmol to 162.96 µmol (Fig. 3b). It is also noted that the increment of TEOA concentration from 0.35 mol/L to 0.50 mol/L does not greatly enhance the H₂ production (Fig. 3c), indicating that the concentration of TEOA higher than 0.35 mol/L is no longer the rete-limiting factor for photocatalysis. After 5-h photocatalysis, a TON of as high as 6834 has been achieved, which represents one of the highest values in terms of known Ni-substituted POM-catalyzed H₂ production systems (Table S3 in Supporting information). Therefore, an optimal TEOA concentration of 0.35 mol/L is used for further study. To better understand the catalytic system, several control experiments were conducted to assess the importance of each component for photocatalytic hydrogen generation. As shown in Fig. 3d, the absence of any components (catalyst, sacrificial reagent, or photosensitizer) resulted in negligible hydrogen production, and the optimal photocatalytic system exhibited the highest TON value among various control experiments (Fig. S5b in Supporting information). The catalytic system only produced very little amount of H₂ gas while replacing the TBA-Ni₁₆L₆(SiW₉)₄ catalyst with TBA-SiW₉, revealing the significance of the Ni active sites. In addition, the replacement of TBA-Ni₁₆L₆(SiW₉)₄ catalyst with molar equivalents of Ni(NO₃)₂·4H₂O, TBA-SiW₉, and L ligands also led to inefficient H₂ production. These results confirm that the catalyst, photosensitizer, and sacrificial reagent are all indispensable factors, and the special molecular skeleton of TBA-Ni₁₆L₆(SiW₉)₄ is extremely crucial for efficient photocatalysis.

To prove the important molecular cage structure of TBA-Ni₁₆L₆(SiW₉)₄ catalyst, we have compared its catalytic performance with that of our previously-reported Ni₁₆-containg POM catalyst [42], TBA-Ni₁₆(PO₄)₄(PW₉)₄, which is also characterized by FT-IR spectra (Fig. S6 in Supporting information). Under otherwise identical conditions, the hydrogen production of TBA-Ni₁₆L₆(SiW₉)₄ is much higher than that of TBA-Ni₁₆(PO₄)₄(PW₉)₄ (Fig. S7 in Supporting information). Such distinct catalytic performance could be attributed to the unique cage structure of the Ni₁₆L₆(SiW₉)₄ POCs, which could not only provide good transportation of catalytic components via a possible substrate channeling effect, but also offer more accessible catalytic active sites for H₂ production.

Then, the stability of catalyst and robustness of the catalytic system have been further investigated by various experiments. Firstly, a Hg-poisoning test by adding 20 mg Hg to the photocatalytic system does not significantly decrease the hydrogen evolution, eliminating the formation of potential metal nanoparticles from the decomposition of TBA-Ni₁₆L₆(SiW₉)₄ catalyst (Fig. S8 in Supporting information). Secondly, the long-term catalytic reaction up to 96 h produced ~450 µmol of H₂, corresponding to a TON of ~15,000, indicating the long-term robustness of the TBA-Ni₁₆L₆(SiW₉)₄ catalyst (Fig. S9 in Supporting information). The declining hydrogen production rate at prolonged reaction time could be attributed to (1) the consumption of TEOA that may change the environment of the photocatalytic reaction system, (2) the slight decomposition of [Ir(coumarin)₂(dtbbpy)]⁺ photosensitizer that leads to the decreased utilization efficiency of incident photons. Thirdly, to better assess the stability of the catalyst, DLS measurement was also utilized to examine the homogeneity of the Ni₁₆L₆(SiW₉)₄-catalyzed reaction. The DLS signal centered at 1.2 nm was observed after 5-h photocatalysis (Fig. S10 in Supporting information), which is consistent with the hydrodynamic diameter of TBA-Ni₁₆L₆(SiW₉)₄ (5 µmol/L) before catalysis. Moreover, the structural integrity of the TBA-Ni₁₆L₆(SiW₉)₄ catalyst has also been characterized by their FT-IR spectra before and after photocatalysis. Basically, an excessive amount of [Ru(bpy)₃]²⁺ was added to the reaction solution to separate the catalyst after photocatalysis, forming a {[Ru(bpy)₃]-Ni₁₆L₆(SiW₉)₄} precipitate. The FT-IR spectra of isolated {[Ru(bpy)₃]-Ni₁₆L₆(SiW₉)₄} precipitate before and after photocatalysis retained all characteristic bands of TBA-Ni₁₆L₆(SiW₉)₄ catalyst, further confirming the structural stability of the TBA-Ni₁₆L₆(SiW₉)₄ catalyst (Fig. S11 in Supporting information).

To reveal the photocatalytic mechanism, the emission quenching of [Ir(coumarin)₂(dtbbpy)]⁺ by either TEOA or TBA-Ni₁₆L₆(SiW₉)₄ was monitored in an inert atmosphere by using steady-state and time-resolved luminescence spectroscopy. The addition of either 5 µmol/L TBA-Ni₁₆L₆(SiW₉)₄ catalyst or 0.35 mol/L TEOA electron donor can cause the emission quenching of [Ir(coumarin)₂(dtbbpy)]^+* (Fig. 4a). The single-exponential decay kinetics of excited-state [Ir(coumarin)₂(dtbbpy)]^+* luminescence in the presence of TEOA and TBA-Ni₁₆L₆(SiW₉)₄ exhibited a decreased lifetimes from 1131.6 ns to 887.8 ns and 927.8 ns (Fig. 4b), respectively. These results proved the presence of both oxidative and reductive quenching processes during photocatalysis. The emission intensity of [Ir(coumarin)₂(dtbbpy)]^+* was gradually quenched with the addition of TBA-Ni₁₆L₆(SiW₉)₄ (from 0 to 60 µmol/L) or TEOA (from 0 to 0.35 mol/L) (Figs. 4c and d). The linear Stern-Volmer plots give rise to an oxidative quenching rate constant (k_oq) by TBA-Ni₁₆L₆(SiW₉)₄ of 4.1 × 10⁹ L mol^–1 s^–1 (Fig. S12 in Supporting information) and a reductive quenching rate constant (k_rq) by TEOA of 2.2 × 10⁶ L mol^–1 s^–1 (Fig. S13 in Supporting information). Even though the koq value is three times orders of magnitude higher than the k_rq value due to the strong electrostatic interaction between the cationic [Ir(coumarin)₂(dtbbpy)]⁺ and the anionic Ni₁₆L₆(SiW₉)₄ catalyst, the reduction quenching pathway is still regarded as a dominant one due to the much higher concentration of TEOA (0.35 mol/L) than that of TBA-Ni₁₆L₆(SiW₉)₄ (5 µmol/L), corresponding to a calculated quenching rate of 7.8 × 10⁵ s^–1 and 2.1 × 10⁴ s^–1, respectively.

In summary, an unprecedented tetrahedral POC architecture K₃Na₁₇H₁₂[(C₄H₆O₆)₆[Ni₄(OH)₃(A-α-SiW₉O₃₄)]₄]·96H₂O has been successfully synthesized using flexible L ligand and Ni₄SiW₉ POM SBUs. Such Ni₁₆L₆(SiW₉)₄ tetrahedron was constructed by four Ni₄SiW₉ SBUs and six L linkers to form a cage structure with a void space of about 102 ų. Under minimally optimized conditions, the TBA-Ni₁₆L₆(SiW₉)₄-catalyzed hydrogen production system produces a TON of 6834 after 5-h reaction, and the TON value steadily increases to an extraordinary level of approximately 15,500 while prolonging the reaction time to 96-h. Spectroscopic analyses confirmed that the photocatalytic mechanism includes both the reductive and oxidative quenching pathways. In addition, a series of experimental evidence (e.g., Hg-poisoning test, DLS measurements and FT-IR spectra of the isolated catalyst before and after photocatalysis) confirmed the structural stability of the TBA-Ni₁₆L₆(SiW₉)₄ catalyst during the photocatalytic process. Due to the unique cage structure of the Ni₁₆L₆(SiW₉)₄ POCs, more efficient catalytic component accessibility would be expected in the case of TBA-Ni₁₆L₆(SiW₉)₄-catalyzed reaction system compared to that of the cage-free TBA-Ni₁₆(PO₄)₄(PW₉)₄-catalyzed system. This work not only enriches the family of POM-based organic cages, but also extends new catalytic applications of POCs in solar-driven hydrogen production.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was financially supported by the National Natural Science Foundation of China (Nos. 21871025 and 21831001), the Recruitment Program of Global Experts (Young Talents) and BIT Excellent Young Scholars Research Fund. The instrumental support from the Analysis and Testing Center of Beijing Institute of Technology is also highly appreciated.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2023.108736.