Visible light-driven photocatalytic hydrogen production from water is an attractive strategy towards solar energy conversion and storage. Various heterogenous-semiconductor [1-6] and homogenous-molecule assemblies [7, 8] have experienced intensive exploration, and great progress has been achieved [9, 10]. A hydrogen-evolving photochemical molecular device self-assembled Pd₆(RuL₃)₈(BF₄)₂₈ (L=2-(pyridin-3-yl)- 1H-imidazo [4, 5-f][1, 10]-phenanthroline) cage (MOC-16) [11], incorporating multiple photo- and catalytic metal centers, had been developed in our previous report [7]. Although the directional electron transfers through multiple but independent channels leaded to efficient hydrogen production, MOC-16 suffer from the typical drawbacks of homogeneous catalysis, primarily catalyst deactivation to colloidal Pd nanoparticles. To circumvent the vulnerability of molecular catalysts under photocatalytic conditions, a plethora of approaches have been pursued to immobilize them onto a suitable matrix as heterogenous systems. For example, anchoring molecular catalysts on surfaces with multiple binding motifs is a usual strategy, especially in electrode, albeit cleavage or deactivation of the catalyst is inevitable [12-14].

Graphitic carbon nitride (g-C₃N₄) is a fascinating conjugated polymer in solar fuel generation through water splitting and CO₂ photoreduction [15-21]. The special layer structure with π-conjugated electrons and abundant edge termination to form H bonds engender an ideal platform for C₃N₄ to construct hybrid systems with various components. So far, a plenty of heterostructured g-C₃N₄ with second even third components, including metal, semiconductor, graphilic carbon, sensitizer et al., had been constructed to remedy its imperfection of wide bandgap, severe charge recombination, and lack of surface-active sites [18, 22-25].

Inspired by preceding heterostructured g-C₃N₄ photocatalysts, immobilizing molecule-MOC-16 on C₃N₄ is a compelling and feasible approach to heterogenize MOC-16. The opened-up 2D flat morphology of C₃N₄ could provide a platform to isolate MOC-16, which is expected to promote the photocatalytic performance comparing with homogenous MOC-16. From perspective of g-C₃N₄, the high-efficient electron transfers and multiple photo- and catalytic metal centers within MOC-16 are also anticipated to advance the performance of g-C₃N₄, in terms of expanding light absorption, promoting charge separation, and accelerating surface kinetics. Hereinafter, hybrid nanocomposite of MOC-16/g-C₃N₄ was constructed, and the influences of MOC-16 content and the comparison systems, such as RuL₃/g-C₃N₄, Pd/RuL₃/g-C₃N₄ and mixture of MOC-16 and g-C₃N₄, on photocatalytic H₂ evolution (PHE) was evaluated and discussed. The optimized PHE rate of 10 wt% MOC-16/g-C₃N₄ was 2021 μmol g^–1 h^–1, which was higher than those with the comparison catalysts. Moreover, the catalysts 10 wt% MOC-16/g-C₃N₄ could be repeatedly utilized 3 times with slight loss in hydrogen production activity, giving an enhanced turnover number of 517 for 15 h compared with homogenous MOC-16 system (638 in 48 h).

MOC-16 was synthesized according to the reported procedure. g-C₃N₄ was prepared as following. 10 g urea was placed in a 50 mL crucible, calcinated in the muffle furnace with a heating rate of 10 ℃/min to reach the temperature of 550 ℃, then held at this temperature for 2 h. After grinded, the resulting powder underwent a heat treatment at the rate of 10 ℃/min to reach 500 ℃, holding for 2 h. The resulting g-C₃N₄ powder was then stirring overnight in H₂O and CH₃CN mixed solvent (v:v=1:1) dissolving a series calculated amount of MOC-16 (5.0/7.5/10.0/12.5/15.0 wt%), followed by withdrawing solvent with rotary evaporation. The product underwent washing and filtrating with H₂O several times, which stayed colorless during the washing procedure. The samples were denoted as X wt% MOC-16/g-C₃N₄.

As controls, equivalent RuL₃ with 10 wt% MOC-16/g-C₃N₄ was immobilized on g-C₃N₄ with same procedure to get RuL₃/g-C₃N₄. Following that, equivalent Pd with 10 wt% MOC-16/g-C₃N₄ was co-immobilized on RuL₃/g-C₃N₄ via photo-deposition under Xenon lamp irradiating for 1 h with H₂PdCl₄ as Pd source. The resulting Pd-RuL₃/g-C₃N₄ powder was collected and washed with H₂O, and then dried under 40 ℃. In addition, equivalent MOC-16 and g-C₃N₄ with 10 wt% MOC-16/g-C₃N₄ were physical mixture to contrast with 10 wt% MOC-16/g-C₃N₄.

The photocatalytic H₂ evolution experiments were carried out in a closed gas circulation and evacuation system equipped online GC (Agilent 7820A) for H₂ detection. In detail, 10 mg photocatalyst was dispersed in 20 mL solvent containing 18 mL distilled water and 2 mL TEOA as sacrificial reagent. The reaction was carried out under visible light (λ > 420 nm) irradiation with 300 W Xenon lamp (PLS-SXE-300C, Beijing Perfectlight). The apparent quantum yields (AQYs) were tested by online GC (Agilent 7820A) and Zolix LED (λ > 420 nm, lighting area was 0.8 cm²), using 5 mg photocatalyst in 10 mL solvent containing 9 mL distilled water and 1 mL TEOA.

UV-vis absorption spectrum (UV-Vis) was measured by UV-3600 spectrometer. The fluorescence spectra were measured by Edinburgh FLS980 combined fluorescence spectrometer. Powder X-ray diffraction (PXRD) was recorded on a Rigaku Smart Lab diffractometer (Bragg-Brentano geometry, Cu-K_α1 radiation, λ = 1.54056 Å). The Fourier transform infrared (FT-IR) spectra were obtained by KBr tablet in Nicolet avatar 330 FT-IR spectrometer with the unit of cm^-1. Scanning electron microscopy (SEM) micrographs were measured by Hitachi ultra-high resolution FE-SEM SU8010 microscope. Transmission electron microscopy (TEM) images and EDX elemental mapping were obtained on FEI Tecnai G2F20S-TWIN (US) 200 KV transmission electron microscope. The specific surface area of 0.25 cm² was measured by the Cantal-automatic adsorption-iQ2-MP analyzer. X-ray photoelectron spectroscopy (XPS) measurements were carried out using an XR6 monochromatic Al K_α X-ray source (hν = 1486.6 eV) with a 900 mm spot size in an ultrahigh vacuum chamber (ESCALAB 250Xi). The pass energy was 20 eV. The amounts of MOC-16 loaded on samples were detected by inductively coupled plasma-atomic emission spectrometry (ICP-AES) (spectra range: 120-800 nm, holographic grating, 2924 line/mm, SPECTRO CIROS VISION). The photocurrent-time (i-t) curve was conducted on ChengHua CHI660E workstation with 0.1 M Na₂SO₄, FTO working electrode, glassy carbon electrode and Ag/AgCl as reference electrode.

Prior to photocatalytic test, the composition and structure of hybrid compounds were determined by ICP-AES, PXRD and FT-IR spectra. The percentage compositions of a serial of X wt% MOC-16/g-C₃N₄ samples were 3.9, 6.0, 8.4, 9.0, and 10.5, respectively, listed in Table 1, and all of them were slightly lower than original feeding. Fig. 1(a) gives XRD patterns of MOC-16/g-C₃N₄ composites with different MOC-16 content. All samples showed two distinct diffraction peaks at 27.7^o and 13.0^o related to (002) and (100) planes of g-C₃N₄ respectively, indicating g-C₃N₄ remained intact during hybrid process. Due to molecular feature of MOC-16, XRD is blind for MOC-16. Instead, FT-IR is sensitive to molecular MOC-16 and is used to confirm the presence of MOC-16 in hybrid composition. As shown in Fig. 1(b), pristine g-C₃N₄ exhibited characteristic IR bands of CN heterocycles at 1251, 1325, 1419, 1571, 1639 cm⁻¹, apart from broad peaks between 3000 and 3500 cm^-1 contributed to the N-H bond and breathing mode of triazine units at 810 cm⁻¹ [26]. The characteristic bands of MOC-16 were observed at 1000-1200 and 1400-1660 cm⁻¹ regions, accompanying with two broad bands at 3100-3500 cm⁻¹ region. For hybrid composite of 10 wt% MOC-16/g-C₃N₄, the IR features of g-C₃N₄ retained and overlapped with those of MOC-16. No obvious shifts were observed in composite of MOC-16/g-C₃N₄.

To further disclose distribution of MOC-16 on g-C₃N₄, HRTEM and EDX elemental mapping were recorded and images were shown in Fig. 2. In Fig. 2(a) and (b), laminar morphology with bending edges was observed, and no features related to MOC-16 were observed due to its molecular nature. In elemental mapping in Fig. 2(c)-(h), all elements were uniformly distributed in the measured domain, including noble metals of Pd and Ru, which evidenced that MOC-16 was homogenously dispersed on g-C₃N₄.

Photocatalytic H₂ evolution in MOC-16/g-C₃N₄ with different MOC-16 contents were shown in Fig. 3(a), and the corresponding evolution rates were listed in Table 1. It can be seen that H₂ evolution was significantly enhanced in hybrid catalysts comparing with pristine g-C₃N₄, and the improvement was strongly correlated with MOC-16 content. When MOC-16 content increased from 5 to 10 wt%, the H₂ evolution rate increased from 1105 μmol g^–1 h^–1 to the maximum of 2021 μmol g^–1 h^–1, further increasing MOC-16 content, the rate declined, instead further improvement. That is, there existed an optimum MOC-16 content of 10 wt% for H₂ evolution. As controls, pristine g-C₃N₄, RuL₃/g-C₃N₄, Pd/RuL₃/g-C₃N₄ and physical mixture of MOC-16 and g-C₃N₄ were tested for H₂ generation and compared with the optimized sample in Fig. 3(b). The negligible H₂ evolution was observed in pristine g-C₃N₄, and the hybrid of RuL₃ with g-C₃N₄ did not lead to activity increase as expected, although optical absorption was also extended to longer wavelength with RuL₃. Further loading Pd metal on the surface of RuL₃/g-C₃N₄, the activity was significantly improved several magnitudes, but lower than the optimized 10 wt% MOC-16/g-C₃N₄. It indicated that metal Pd as an active site was important for activity promoting, and the linkage of RuL₃ and Pd in MOC-16 assembly with high-efficient electron transfer was favorable for H₂ evolution. The activity of physical mixture of MOC-16 and g-C₃N₄ was inferior to that of 10 wt% MOC-16/g-C₃N₄, exemplified the interaction (vide infra) between MOC-16 and g-C₃N₄ was inevitable. The stability of 10 wt% MOC-16/g-C₃N₄ was also test in three cycles as shown in Fig. 3(c). In the following two cycles, the rate slightly declined but still sustained 1731μmol g^–1 h^–1. The accumulated turn over number (TON) and turn over frequency (TOF) based on Pd was calculated and shown in Fig. 3(d). The TON achieved 517 within 15 h, significantly improved comparing with homogenous MOC-16 system of 638 (48 h) [7]. Its TOF value of ~35 h^–1 was comparable with that of ~30 h^–1 for homogenous MOC-16. The apparent quantum yield of 10 wt% MOC-16/g-C₃N₄ was 0.29%, while that of homogenous MOC-16 was 0.01% under the same condition, in agreement with the results in photocatalytic hydrogen production experiments.

In order to unveil the reason for activity improvement in hybrid compounds, optical absorption of a serial of MOC-16/g-C₃N₄ samples were measured, and relevant results were shown in Fig. 4(a). The pristine g-C₃N₄ exhibited a typical feature of semiconductor absorption with onset band edge at about 460 nm. For solid-state MOC-16, a broad absorption band at 300-400 nm UV region and a visible absorption peak centered at 460 nm were observed along with a tail till to 700 nm, partly inconsistent in UV region with that dissolved in solvent [7] due to different condensed states. For hybrid compounds of MOC-16/g-C₃N₄, absorption bands showed the overlapped features of MOC-16 and g-C₃N₄, the sharp band edge absorption of g-C₃N₄ plus the visible absorption band of MOC-16. In other words, the optical absorption of g-C₃N₄ was extremely extended from 460 to 700 nm after combining with MOC-16, allowing a more efficient utilization of solar light to create photogenerated electrons and holes. Moreover, the loading amount of MOC-16 increased in the MOC-16/g-C₃N₄ hybrid, the absorbance intensities of the hybrids was enhanced. The optimum loading content was always reported in a plenty of composites related to g-C₃N₄ due to shielding effect existence [22]. In our case, it is reasonable that lower content MOC-16 could be isolated on the surface of g-C₃N₄, corresponding to smaller amounts of active sites, on the contrary, excessive MOC-16 could engender aggregation on 2D surface and the following decline of activity.

To further confirm the enhanced photoresponse, photocurrent-time (i-t) curves of 10 wt% MOC-16/g-C₃N₄ and MOC-16 electrode have been measured under illumination with several on/off switches (Fig. 4(b)). The MOC-16/g-C₃N₄ electrode showed higher photocurrent than MOC-16, hinting that effective interfacial electron transfer between MOC-16 and g-C₃N₄ in the hybrid.

Although H₂ evolution was significant improved as expected, the activity still declined in sequence reaction cycles (Fig. 3(c)). XPS was carried out to disclose the deactivated reason and results were shown in Fig. 4(c). The double peaks of Pd²⁺ species corresponding to 3d_5/2 and 3d_3/2 are observed in both MOC-16 and 10 wt% MOC-16/g-C₃N₄ with comparable binding energy at 338.0 and 343.1 eV. For the recycled sample after three cycles, the preceding peaks moved to 337.3 and 342.6 eV, accompanying with the appearance of zero valent Pd with binding energy at 335.0 and 340.0 eV. It is indicated that MOC-16 immobilized on the surface of g-C₃N₄ went through dissociation of Pd-N bonds during photocatalytic process. The photogenerated electron continuously flowed to Pd active sites for H₂ generation, and the sluggish dynamic of consumption of electrons engendered part reduction of Pd²⁺ and then breakage of Pd-N.

As mentioned in preceding text, the interaction between MOC-16 and g-C₃N₄ was essential to the highest activity observed in 10 wt% MOC-16/g-C₃N₄, apart from the optical absorption extending. In previous reports [21], negative polarity surface on 2D g-C₃N₄ was evidenced by Z-potential measurement, on the contrary, MOC-16 with Pd₆(RuL₃)₈(BF₄)₂₈ composition possessed +28 charges in solvent [11]. It is supposed that electrostatic attraction were the main driving force to immobilize MOC-16 on g-C₃N₄. The location of MOC-16 on the surface defects of g-C₃N₄ via electrostatic and π−π stacking interaction and on the edge of g-C₃N₄ through hydrogen bonding were favorable. Thus, an interface between MOC-16 molecule and g-C₃N₄ semiconductor was constructed. A possible reaction mechanism based on band edge alignment of MOC-16 and g-C₃N₄ with respect to standard water redox potential was proposed in Fig. 5. The HOMO and LUMO energy levels of MOC-16 were respectively determined from the oxidation and reduction peaks of cyclic voltammogram, 1.55 and –0.95 V vs. NHE [7]. The energy band position of pristine g-C₃N₄ was slightly different in previous reported references [22, 27]. The conductive band (CB) changed from –1.09 to –1.3 V, concomitantly, valence band (VB) altered from 1.53 to 1.4. Anyway, the photogenerated electron transfer from g-C₃N₄ to MOC-16 was thermodynamically feasible. The destination of photoelectrons were active sites of Pd to be consumed for H₂ evolution, meantime, photo holes were consumed by sacrificial reagent of TEOA.

In summary, MOC-16 was immobilized on g-C₃N₄ with open-up 2D flat morphology to fulfill the heterogenization of molecular catalyst and the high-efficient electron transfer in MOC-16 with photosensitized RuL₃ centers to catalytic Pd centers was inherited in hybrid MOC-16/g-C₃N₄. A significantly-promoted H₂ evolution with TON (517) was obtained comparing with pristine MOC-16 and g-C₃N₄. In addition to the support function, g-C₃N₄ also possessed thermodynamic electron donating ability to facilitate activity enhancement. This work provides a promising strategy to promote H₂ evolution via constructing molecule and semiconductor interface.

Acknowledgments

We gratefully acknowledge the financial support from the National Natural Science Foundation of China (21875293, 21821003, 21890380, 21720102007, 21572280), the Natural Science Foundation of Guangdong Province (2016A030313268), the STP Project of Guangzhou (201804010386, 201707010114), the Fundamental Research Funds for the Central Universities (17lgzd18, 17lgzd01) and the Research Fund Program of Key Laboratory of Fuel Cell Technology of Guangdong Province.