Hydrogen (H₂) is a promising fuel source for future energy use. Reduction of water into H₂ using visible light is one of the most promising scenarios for a sustainable energy supply. In this arena, molecular reaction systems can provide a high degree of tunability toward the optimization of this reductive half-reaction. An efficient homogeneous molecular system for photochemical reduction of protons to H₂ generally requires at least three basic components: a photosensitizer (PS), a proton-reduction catalyst (PRC), and a sacrificial electron donor. Among the PRCs, molecular species based on earth-abundant cobalt centers have received considerable attention [1-4]. It is noted that these cobalt-based PRCs often feature square planar or pyramidal structures with labile axial ligands (typically aqua or halide) [1-5], whereas cobalt complexes with octahedral geometries remain unexplored. Furthermore, in these reported homogeneous H₂-evolving systems, most cobalt PRCs usually operate in mixtures of organic solvents and water because of their poor solubility and aqueous activity [1-4]. Thus, the search for efficient cobalt-based molecular catalysts that can function in purely aqueous solution remains a great challenge [6-8].

In contrast, compounds based on thiosemicarbazide (NH₂-NH-C(=S)-NH₂, denoted as Htsc), which belong to the group of thiourea derivatives, can form stable complexes with many transition and main-group metal ions [9]. In the past, Htsc-containing complexes have attracted much attention because of their highly desirable biological and pharmaceutical properties, such as antioxidant, antibacterial, antiviral, or antimalarial activities [10-12]. However, the use of such thiosemicarbazide-based complexes as H₂-evolving molecular catalysts remains unexplored. Herein, we report a pair of octahedral geometric isomers with fac- and mer-geometry of tris(thiosemicarbazide) cobalt(Ⅲ), i.e., fac-[Co(Htsc)₃]Cl₃·3H₂O (C1) and mer-[Co(Htsc)₃]Cl₃·4H₂O (C2). They are confirmed to be active molecular PRCs for electrocatalytic and photocatalytic reduction of protons.

We chose the two cobalt complexes according to the following considerations: (1) There is no report of using octahedral geometric isomers for catalytic hydrogen production. Thus, the catalytically H₂-evolving activity of the two cobalt complexes with fac- and mer-configuration will be compared. (2) The two ionic complexes have sufficient solubility in water, which will facilitate the evaluation of their catalytic properties in H₂O without organic additives. (3) The amino nitrogens of Htsc can, in principle, serve as internal proton transfer sites, as previously hypothesized for a pendant nitrogen base in nickel bis(diphosphine) for protonation. (4) The results may represent a new paradigm for the construction of noble-metal-free photocatalytic systems using thiosemicarbazide-containing complexes for solar hydrogen generation.

All chemical reagents used in the syntheses were of analytical grade and used without further purification, unless otherwise indicated. Water was purified using a Milli-Q system. Triethylamine (TEA) and all solvents were dried by standard literature methods [13]. The elemental analyses (C, H, and N contents) were determined on a Vario EL Ⅲ analyzer. Co was analyzed on a PLASMA-SPEC ICP atomic emission spectrometer. UV/Vis absorption spectra were recorded on a Shimadzu UV-2600 spectrophotometer. Photoluminescence (PL) spectra were obtained using a Thermo Scientific Lumina fluorescence spectrophotometer. Fourier transform infrared spectra (FT-IR) were recorded on a Nicolet iS50 FT-IR spectrometer as KBr pellets in the frequency range of 4000-400 cm^-1. Mass spectra (MS) experiments were carried out on an Agilent Infinity Ⅱ QTOF G6545.

The syntheses of fac-[Co(Htsc)₃]Cl₃·3H₂O (C1) and mer-[Co(Htsc)₃]Cl₃·4H₂O (C2) were achieved using literature methods [14] and the products were characterized by electronic absorption spectroscopy, elemental analysis, FT-IR spectroscopy, and single-crystal X-ray diffraction (SCXRD) to determine their actual molecular structures.

fac-[Co(Htsc)₃]Cl₃·3H₂O (C1). MS (m/z): 332.3, corresponding to the cationic unit. UV/Vis, λ_max/nm (ε/L mol^-1·cm^-1), (H₂O): 257 nm (6261), 414 nm (97), 547 nm (93). IR (KBr): v(cm^-1) = 3387 (s), 3144 (s), 1625 (s), 1401 (s), 1328 (m), 1219 (s), 1154 (m), 1011 (w), 754 (w), 703 (s), 613 (s), 575 (s), 520 (m). Anal. Calcd. (found) for CoC₃H₂₁O₃N₉S₃Cl₃: C, 7.31 (7.43); H, 4.30 (4.36); N, 25.58 (25.66); Co, 11.96 (12.28).

mer-[Co(Htsc)₃]Cl₃·4H₂O (C2). MS (m/z): 332.3, corresponding to the cationic unit. UV/Vis, λ_max/nm (ε/L mol^-1·cm^-1), (H₂O): 248 nm (4179), 420 nm (309), 575 nm (79). IR (KBr): v(cm^-1) = 3272 (s), 3100 (s), 1634 (s), 1434 (m), 1393 (m), 1253 (m), 1152 (w), 1012 (w), 760 (w), 700 (m), 589 (w), 439 (w). Anal. Calcd. (found) for CoC₃H₂₃O₄N₉S₃Cl₃: C, 7.05 (7.16); H, 4.54 (4.61); N, 24.68 (24.76); Co, 11.54 (11.82).

Two other cobalt complexes, [Co(dmgH)₂pyCl] (C3, where dmgH = dimethylglyoximate monoanion and py = pyridine) and [Co(bpy)₃Cl₂] (C4, where bpy = 2, 2′-bipyridine), were prepared according to the literature procedures [15, 16].

Electrochemical experiments were performed using a CHI 650E electrochemical analyzer. Cyclic voltammograms (CVs) were recorded in an electrolytic cell with a 3 mm glassy-carbon working electrode, saturated calomel electrode (SCE) as reference electrode, and a platinum-wire counter electrode as auxiliary electrode. The glassy-carbon working electrode was polished with alumina (1.0 μm, 0.3 μm, and 0.05 μm) on a polishing cloth before each measurement. The solution was purged with Ar gas and dried by standard literature methods prior to measurements. The measurements were performed in methanol with 0.1 mol/L tetrabutylammonium hexafluorophosphate (n-Bu₄NPF₆) as a supporting electrolyte. Controlled potential coulometry was done in an electrolytic cell with a glassy-carbon working electrode, SCE reference electrode, and a platinum-wire counter electrode. After electrolysis, a sample of the headspace was injected into a gas chromatograph (GC).

The photocatalytic water splitting experiments were performed according to the previous work [17, 18]. In a typical experiment, 10 mL of a 5% TEA aqueous solution containing fluorescein (FL) (1 × 10^-3 M) and either C1 or C2 (1 × 10^-5 mol/L) was added to a Schlenk bottle. The solution was then freeze-pump-thaw degassed three times and irradiated using a xenon light source (300 W, Beijing Perfectlight Co., Ltd.), with an optical filter employed to cut off light with wavelengths below 420 nm, under constant stirring. The pH of the solution was adjusted to the desired value using hydrochloric acid or sodium hydroxide as required and determined by an OHAUS ST5000 pH meter. During irradiation, the gas phase of the reaction system was analyzed by a GC (Shimadzu GC-2014C) with a thermal conductivity detector, a 5 Å molecular sieve column, and with N₂ as carrying gas. The amounts of hydrogen were quantified by the external standard method.

Single crystals of C1 and C2 with appropriate dimensions were chosen and mounted on a glass fiber for data collection. Data were collected on an Oxford Diffraction Gemini A with a CCD area detector, and the CrysAlisPro and CrysAlis RED software packages were used for data collection and data integration. The collected data were corrected for absorbance by an analytical numeric absorption correction using a multifaceted crystal model based on expressions derived from the Laue symmetry using equivalent reflections [19]. Structural solution and full-matrix least-squares refinement based on F² were performed with the SHELXS-97 [20] and SHELXL-97 [21] program packages, respectively. All the non-hydrogen atoms were refined anisotropically. Hydrogen atoms of organic ligands were generated geometrically and refined with isotropic temperature factors. The hydrogen atoms attached to nitrogen and oxygen for C1 and C2 were located in the difference Fourier maps and refined with all other atoms fixed. Selected crystallographic data for C1 and C2 are summarized in Table S1. Selected bond lengths and angles for C1 and C2 are listed in Table S2. CCDC-1576996 (C1) and 1577003 (C2) contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

All the theoretical calculations were performed with the Gaussian 09 package [22]. For the purpose of accordance with the experimental results, the geometric parameters taken from our SCXRD analysis were used as the starting points for the geometry optimizations. Ground-state geometries were fully optimized in the framework of density functional theory (DFT) using the hybrid functional B3LYP, which combines Becke's 3-parameter exchange functional [23] and Lee, Yang, and Parr's correlation functional [24]. The split-valence and triple-zeta basis set TZVP [25] and the polarizable continuum model (PCM) [26] of water solvent were employed in all calculations. All geometries were deemed to be energy minima, as no negative frequencies were found. Time-dependent DFT (TD-DFT) [27] was applied to compute the vertical transition energies.

The complexes of cobalt thiosemicarbazide (C1and C2) were synthesized as reported in the literature [14]. Their structures were characterized by elemental analyses, MS, IR spectroscopy, and UV/Vis spectroscopy in the solution and solid state, as well as X-ray crystallography in the solid state. Single crystals of C1 and C2 suitable for X-ray diffraction were grown through the diffusion of concentrated hydrochloric acid into aqueous solutions of C1 and C2. Single-crystal X-ray diffraction (SCXRD) analysis revealed that C1 is composed of a [Co(Htsc)₃]³⁺ cation, three Cl^- anions, and three lattice water molecules, while C1 consists of a [Co(Htsc)₃]³⁺ cation, three Cl^- anions, and four lattice water molecules (see Fig. S1 in the Supporting Information (SI)). In C1 and C2, the cation [Co(Htsc)₃]³⁺ is a six-coordinate Co^Ⅲ complex with a distorted octahedral geometry. The cobalt center is ligated by three bidentate Htsc ligands through the sulfur and hydrazinic nitrogen atoms. The average Co-N distance of 1.988 Å in C1 is close to the 1.980 Å observed in C2, whereas the average Co-S distance of 2.220 Å in C1 is slightly less than the corresponding distance of 2.251 Å observed for C2. In the case of complexes of general formula [Co(N, S)₃], there are two possible geometric isomers: facial (fac)- and meridional (mer)-isomers. Symmetry considerations show that in the fac-isomer the three ligands are equivalent, which is not the case in the mer-isomer. As can be seen from Fig. 1, the [Co(Htsc)]³⁺ unit in C1 adopts a fac-arrangement, which is distinct from the mer-geometry of C2 (Fig. 1).

The UV/Vis absorption spectra of C1 and C2 measured in water are shown in Fig. 2. The electronic absorption spectra of the Co^Ⅲ complexes contain two visible bands at 545 and 425 nm for C1 and at 585 and 410 nm for C2. These bands are characteristic of low-spin Co^Ⅲ complexes in which the metal is in an octahedral environment [28]. The first of these bands is assigned to the transition ¹A_1g →¹T_1g, and the second to the ¹A_1g →¹T_2g transition. In addition, the third, strong, high-energy absorption band at ~255 nm is attributed to the ligand-to-metal charge transfer. The diffuse reflectance spectra of C1 and C2 bear very similar features to the solution spectra of the ionic complexes in water (Fig. S2), indicating that the same species with fac- and mer-configuration exist in the solid state and in solution. The frontier molecular orbitals of C1 and C2 were also analyzed by DFT. The optimized geometric parameters for the lowest-energy conformer are comparable to those determined crystallographically for C1 and C2 (Fig. S3). The main contribution to the highest occupied molecular orbital (HOMO) calculated for the optimized complex C1 or C2 is from the thiosemicarbazide ligands (Fig. 3), while the lowest unoccupied molecular orbital (LUMO) is an antibonding combination of a Co-centered orbital (of d symmetry) with the orbitals of the thiosemicarbazide ligands. A time-dependent DFT study was also performed to calculate the electronic transition in water. The calculated spectra of C1 and C2 are similar to the experimental spectra (Fig. S4).

CVs of C1 recorded in a solution of n-Bu₄NPF₆ in CH₃OH (0.1 mol/L) show the coupled Co^Ⅲ/Co^Ⅱ and Co^Ⅱ/Co^Ⅰ reduction process at ~-0.68 and ~-0.096 V (vs SCE) [29, 30], respectively (Fig. 3a). The Co^Ⅱ/Co^Ⅰ potential falls within the range of that of proton reduction, indicating that the reduced state of C1 may be capable of directly reducing protons. To investigate whether C1 is a potential electrocatalyst for proton reduction, acetic acid (AcOH, pK_a^MeOH = 9.7) [31] was employed as a very inexpensive proton source. As displayed in Fig. 4(a), upon successive addition of AcOH to CH₃OH solutions of C1, the cathodic current at -1.28 V increases, indicative of electrocatalytic reduction. At the concentration of 40 equivalents of AcOH, the observed catalytic current has a linear correlation with the added AcOH concentration at a scan rate of 200 mV/s (Fig. 4(b)). Beyond this acid concentration, the current is acid-independent, indicating that the PRC C1 is acid-saturated. Control experiments using AcOH or Co(NO₃)₂ in AcOH in the absence of C1 showed no hydrogen evolution activity within the potential window. The catalytic wave grew out of the Co^Ⅱ/Co^Ⅰ redox couple at a more cathodic potential in these control experiments, implying that the first step for H₂ evolution involves sequential electrochemical reductions to Co(Ⅰ) species, which is followed by protonation events. Electrodeposits as potential heterogeneous catalysts are excluded on the basis of negative standard rinse experiments (Fig. S5). Controlled potential electrolysis was performed at -1.28 V (vs SCE) to confirm the identity of the gaseous product as hydrogen. C1 consumed more charge than the blank during the bulk-electrolysis experiment, which supports that the complex is responsible for catalysis (Fig. S6).

To evaluate the electrocatalytic property of C1, the turnover frequency (TOF) and the overpotential (η) of the reaction were determined. Using data in this acid-independent regime, the rate constant for hydrogen evolution (k) can be calculated from the expression:

in which v is the scan rate in V/s, n is the number of electrons transferred (two for hydrogen evolution), R is the gas constant, T is the temperature in K, and F is the Faraday constant. The i_cat/i_p value of 23 in the acid-independent region (for a scan rate of 200 mV/s) indicates that the TOF, which is also the observed rate constant, is 210 s^-1. Using the method of Appel and Helm [32], η is determined to be 640 mV (-1.28 V vs SCE). Under identical experimental conditions, hydrogen evolution using the other PRC, C2, yielded an equivalent TOF of 208 s^-1 and η of 636 mV (Fig. S7), demonstrating that the two geometric isomers of C1 and C2 have the same electrochemical behavior.

As shown in Fig. 5, electrochemical hydrogen evolution reaction catalyzed by C1 was further evaluated in a phosphate buffer solution (pH = 7.0; 0.1 mol/L), which also showing distinct flow of cathodic current. The overpotential η(E_cat/2) in the aqueous solution is defined by the difference between the standard electrode potential for electrochemical hydrogen evolution reaction defined by the equation 1 (equation 1: E_1/2(2H⁺/H₂) = -0.241 -0.059 pH/V vs. SCE) and the potential (E_cat/2) at which the catalytic current becomes exactly the half of the maximum catalytic current. From E_cat/2 value determined for the catalytic current, the η(E_cat/2) value for electrochemical hydrogen evolution reaction catalyzed by C1 at pH = 7.0, is measured to be 560 mV.

Since the primary motivation for H₂-producing photocatalysis is to reduce protons from water, organic additives and their waste by-products should be minimized or completely eliminated. The ionic complexes C1 and C2 have sufficient solubility (> 40 mg/mL at 20 ℃) in water for their catalytic light-assisted proton reduction activity to be suitably evaluated in fully aqueous solution. Toxic solvents like acetonitrile or methanol commonly used in other homogeneous H₂-evolving systems are not required in our system [1-4]. The visible light-driven catalytic activity of C1 and C2 for H₂ production was investigated using a three-component artificial photosynthesis system: fluorescein (FL) as PS, TEA as sacrificial electron donor, and C1 (or C2) as PRC. Quantitative determination of the generated H₂ was conducted by GC analysis, as described in the Experimental Section. We chose FL as the PS because its excited state and reduced state provide a high driving force (¹FL^* ~-1.7 V and FL^·- ~-1.3 V vs SCE [33]) for successive reduction of the catalyst C1 or C2. When employing TEA as the sacrificial reductant, the net reaction being driven photochemically can be expressed by the equation: NEt₃ + H₂O → HNEt₂ + H₂ + CH₃CHO. Thermochemical data indicate that the overall reaction is thermodynamically unfavorable and must be driven by light energy. During initial observations using the PRC C1, the pH of the reaction solution had a marked effect on hydrogen production from an aqueous solution of TEA (Fig. 6). The amount of H₂ generated in 15 h photolysis maximized at pH = 11.5 and quickly decreased below pH = 9 and above pH = 13. For the photolysis, the component concentrations were 1 × 10^-3 mol/L FL, 1 × 10^-5 mol/L C1, and 0.36 mol/L TEA in water. The pH value of each run was adjusted by addition of HCl or NaOH. Below pH 9 and above pH 13, the amounts of H₂ produced were moderate (less than 15% of that obtained at pH = 11.5 during the same period of irradiation). The great decrease in production of hydrogen at lower pH values was probably because of the protonation of TEA, which resulted in poor electron-donating ability. At higher pH values, the decreased concentrations of protons impair the efficiency of photogeneration of hydrogen.

The photocatalytic efficiencies of C1 and C2 were compared under the optimized pH condition. As shown in Fig. 7, the systems composed of either catalyst C1 or C2 produced copious amounts of H₂ with a turnover number (TON) of 900 for C1 and 890 for C2, corresponding to an estimated initial TOF of 128 h^-1 with respect to catalyst C1 and 125 h^-1 for C2. The very similar values of both TON and TOF indicate that the fac-C1 and mer-C2 have approximately the same photocatalytic activity in our system. Control experiments revealed that the absence of light activation or deletion of any single molecular constituent resulted in negligible H₂ production under the same conditions. In all of the hydrogen production experiments that did yield H₂, induction periods of about 0.5-1 h were observed before substantial amounts of hydrogen evolved. These induction phases may be ascribed to the formation of Co^Ⅰ species or cobalt nanoparticles from the photodecomposition of cobalt complexes. A similar result was observed by Eisenberg and co-workers for cobaloxime-based molecular catalysts [34]. The homogeneity of the photocatalytic system utilizing C1 or C2 was ascertained through the Hg poisoning test. When a large excess of Hg was added to the system, the catalytic activity was unaffected and remained stable throughout the photolysis process (Fig. 8 and Fig. S8). This finding suggests a molecular catalyst for the photocatalytic hydrogen production in each of the C1- and C2-based systems.

After approximately 8 h of irradiation, the rate of hydrogen evolution began to noticeably decrease, indicating that at least one of the system components had been consumed. The addition of fresh FL or TEA produced a negligible restoration of activity. However, upon addition of C1, the initial activity was nearly fully restored, indicating that the system was mainly limited by the decomposition of the PRC (Fig. 9). Two other known cobalt complexes, cobaloxime C3 and the tris(2, 2′-bipyridine)cobalt(Ⅱ) complex C4 (Scheme 1), were reported to be efficient photocatalysts both in basic solutions of organic solvents and in water [35, 36]. We also studied the catalytic performance of those cobalt complexes for photoinduced hydrogen generation under our conditions. However, under the same conditions specified in Fig. 6, C3 and C4 gave TONs of 151 and 20, respectively, much lower than C1, after 15 h irradiation. In summary, C1 and C2 displayed efficient photocatalytic hydrogen evolution activity in pure water solution (Fig. 10).

In the three-component photodriven catalytic systems, the PS excited state can function as either an oxidant or a reductant, and can be quenched by an electron donor or an acceptor [37]. To decide between these two possibilities in our system, the luminescence of the excited FL^*, in deaerated H₂O at pH = 11.5, was monitored as a function of both TEA and, separately, C1 concentration (Fig. 11 and 12). Stern-Volmer analyses yielded the quenching constants for the system. The rate constants for reductive quenching of FL^* by TEA and oxidative quenching by C1 are 1.1 × 10⁷ and 1.2 × 10¹⁰ mol L^-1 s^-1, respectively. These results show that the excited state of FL^* can be both oxidatively quenched by the catalyst and reductively quenched by TEA, and the corresponding quenching processes are diffusion-controlled. Although the rate constant for oxidative quenching is about 10³ times larger than that of reductive quenching, the reductive quenching process is still dominant given the much higher concentration of TEA (0.36 mol/L) relative to C1 (0.01 mmol/L). It should be noted that C2 also shows similar quenching constants on the basis of the Stern-Volmer plots (Fig. S9).

Based on the preceding electrochemical and photochemical observations, a possible, although still ambiguous, catalytic mechanism can be postulated (Scheme 2). After irradiation by visible light, FL is promoted to its excited state FL^*, and then quenched by TEA to generate its anionic form FL^·-. The mechanism for the reduction of protons to H₂ by C1 or C2 involves two-electron reduction from FL^·- or FL^*, which is followed by two protonation events, most likely protonation at an amino nitrogen atom of the thiosemicarbazide ligand, accompanied by dechelation. Protonation of a pendant nitrogen base would be consistent with the proposed mechanism for the hydrogen-evolving catalysts nickel bis(diphosphine) [38] and nickel tri(pyridine-2-thiolate) [33]. After the above steps, H₂ is released from the system through the heterocoupling between Co-H^- and N-H⁺, though a homolytic pathway cannot be ruled out. Further studies including theoretical calculations are needed to describe the mechanism in detail.

In summary, two geometric isomers, fac- and mer-tris(thiosemicarbazide) cobalt(Ⅲ) (C1 and C2), have been synthesized and characterized by X-ray single-crystal diffraction. The two cobalt complexes were tested for activity in electrocatalytic H₂ production by employing acetic acid as a very inexpensive proton source in CH₃OH. Electrocatalytic experiments in CH₃OH showed that both C1 and C2 are active electrocatalysts having the same activity, providing a TOF of 210 s^-1 with a moderate overpotential of 0.64 V when using glassy carbon as electrode material. C1 also exhibits electrocatalytic activity for hydrogen evolution reaction in aqueous media (pH = 7.0) with a moderate overpotential (560 mV). The photocatalytic water reduction properties of C1 and C2 were further evaluated in purely aqueous solution using FL as the PS with TEA as the sacrificial reductant. Efficient H₂ evolution (up to 900 turnovers vs catalyst) was achieved under visible-light irradiation. C1 and C2 showed similar electrocatalytic and photocatalytic H₂ evolution activities, indicating that fac/mer isomerism does not affect the H₂ evolution activity of this complex.

Acknowledgments

Yang Zhao thanks Subsidized Project for Cultivating Postgraduates' Innovative Ability in Scientific Research of Huaqiao University. The Instrumental Analysis Centre of Huaqiao University for analytical characterization is also acknowledged.