By contrast with the use of biologic material such as hydrogenase or PS I, the synthetic approach benefits from the creativity of chemists that exploit the versatile and modular character of coordination chemistry to develop efficient molecular photosensitizers and catalysts. The last few years have seen tremendous achievements in the area of light-driven hydrogen production, with the emergence of two families of efficient catalysts, the diiron mimics of the active site of [FeFe] hydrogenases (46,47), and the bioinspired cobalt complexes (48–50). As a consequence, sustained efforts have also been invested to optimize the performances of the synthetic photocatalytic systems with the use of a wide range of molecular photosensitizers (PS) highlighted in the next section.
Molecular photosensitizers. For many years, most of the work in the area of visible-light driven hydrogen production has relied on the use of polypyridyl ruthenium light-harvesting units, such as the prototype complex [Ru(bpy)3]2+ (PS1, Fig. 6).
This family of photoactive complexes displays very unique redox and photochemical properties. Ruthenium polypyridine complexes indeed efficiently absorb visible light photons which promote one electron from a metal d orbital to the antibonding π* orbital of a diimine ligand. The singlet excited state (S = 0 spin configuration) formed immediately upon excitation rapidly converts into a triplet state (S = 1) through inter-system crossing (isc in Fig. 7), due to the strong spin-orbit coupling effect of ruthenium. This yields a long-lived emissive metal to ligand charge transfer excited state (3MLCT, written PS* in the following descriptions), that can be described as a RuIII center coordinated to a reduced diimine radical anion ligand. Because of the presence of an electron in the π* antibonding orbital of one ligand, this excited state is a quite powerful reducing agent (Table 1) with a standard potential of −0.86 V vs NHE in the case of PS1. At the same time, due to the presence of an unoccupied d orbital at the metal center, PS* can also act as a powerful oxidizing species (Fig. 7) with E°=0.84 V vs NHE in the case of PS1 (51). The nature and concentration of redox partners then essentially control the reactivity of PS* as either a reducing or oxidizing species. These properties make [Ru(bpy)3]2+ and its derivatives ideal candidates for applications in photocatalytic processes (52).
Figure 7. Schematic description of the various states of a photosensitizer relevant for electron transfer processes (PS* indicates the lowest triplet excited state).
Download figure to PowerPoint
Table 1. Excitation wavelengths (nm), molecular absorption coefficients (m−1 cm−1), excited states energies (eV) and redox potentials (V vs NHE) of representative molecular photosensitizers.
A similar description holds for the excited states of the other metal-diimine photosensitizers (Fig. 8) and also for the organic dyes (Fig. 9) presented below. The photophysical and redox properties of a representative example of each family are given in Table 1.
Figure 8. Structure of the iridium-, platinum- and rhenium-based photosensitizers employed in light-driven hydrogen evolution.
Download figure to PowerPoint
The successful application of noble-metal free catalysts in light-driven hydrogen evolution with ruthenium photosensitizers (53,54) has triggered the use of other photoactive complexes, based on iridium (PS4-5), platinum (PS6-7) and rhenium (PS8-9) metal centers (Fig. 8). The iridium photosensitizers are characterized by a higher tunability of their photophysical properties compared with the ruthenium ones (55) and they proved to be more efficient light-harvesting units for hydrogen photoproduction in various photocatalytic systems (55–57).
The platinum chromophores PS6 and PS7 have been extensively studied in combination with the cobaloxime catalysts by Eisenberg et al. (see below), thus allowing to establish some specific trends in the series, such as the lower activity of cyclometallated structures PS7 compared with PS6 (58–60). In a similar study, Castellano et al. have established the influence of phenylacetylide π-conjugation length in PS6e-f on the photocatalytic activity (61).
To date, the most efficient photosensitizers for light-driven hydrogen production are the tricarbonyl-rhenium diimine complexes PS8-9, combined with cobaloximes 3 (57) or 5 (62). Detailed photophysical studies by the group of Hamm et al. have allowed to establish some mechanistic issues related to the cobalt-based photocatalytic systems (62–64). Exchanging the bromide ligand by a thiocyanate (PS9b) in the coordination sphere of Re greatly increased the stability of these systems (62).
As a major breakthrough, the first homogeneous systems exclusively based on earth-abundant elements were reported in 2009 (65). The noble metal-based photosensitizers have been replaced by low cost organic dyes (Fig. 9), such as Eosin Y (PS10) or Rose Bengal (PS11). Other xanthene derivatives such as fluoresceine, lacking bromide or iodide substituents, did not prove active under photocatalytic conditions. Actually, the presence of heavy atoms facilitates inter-system crossing and thus production of a long-lived triplet state, required for an efficient electron transfer to the acceptor. The excited state of these xanthene dyes displays redox properties similar to the noble-metal based PS (Table 1). Photophysical studies have established the feasibility of both reductive quenching in the presence of amine electron donors and oxidative quenching process with a cobaloxime acceptor (66).
The successful implementation of inexpensive organic dyes in light-driven hydrogen production now opens new perspectives for the design and synthesis of novel metal-free photosensitizers with tunable photophysical properties.
Diiron-based catalysts. Following the first syntheses of biomimetic models of the [FeFe] hydrogenase active site covalently linked to a ruthenium photosensitizer by Åkermark, Sun et al. in 2003 (67,68), various diiron dithiolate [2Fe2S] complexes have been combined to a photosensitizing unit (69–75). However, these early generations were shown to be unsuccessful for light-driven hydrogen evolution. The first catalytically active homogeneous system was reported in 2008 by Sun et al. (76). This three-component system, based on the ruthenium photosensitizers PS2, the diiron dithiolate complexes 1 (Fig. 10) and ascorbic acid as the sacrificial electron donor, generated up to 86 turnover number (TON) based on PS (only 4 TON based on the diiron catalyst) in a CH3CN/H2O mixture, within 3 h under visible-light irradiation. The P(Pyr)3-monosubstituted complex 1b (Fig. 10) proved to be the most efficient catalyst of the series, with an improved photostability. Photostability is a critical issue as many of these organometallic complexes suffer from decomposition by light-induced CO release (77). Recently, replacement of PS2 by the cyclometallated iridium photosensitizer PS4 and of the ascorbic acid by TEA conferred a longer lifetime and a higher activity to the previous photocatalytic system, with 466 TONFe measured over 8 h of irradiation (56). To the best of our knowledge, this result represents the highest TON ever reported with a diiron-based catalyst system.
In a different approach, Ott et al. have demonstrated that a subtle modification of the catalyst structure could improve its photocatalytic activity. Indeed, under visible-light irradiation, the H2-evolving catalyst 2 (Fig. 10) achieves ca 200 TON of hydrogen in a 2.5 h experiment, in the presence of [Ru(bpy)3]2+ (PS1) and ascorbic acid (sacrificial electron donor) in a 1:1 DMF/H2O solution at pH = 5.5 (78). Replacing the azadithiolate bridge by a 3,6-dichlorobenzene-1,2-dithiolate moiety is proposed to confer a higher stability to the reduced form of the diiron complex, as evidenced by the reversibility of the first reduction featured in its cyclic voltammogram (79).
Cobaloxime and cobalt diimine-dioxime catalysts. Another class of first-row transition metal complexes that have been successfully employed as H2-evolving catalysts are the cobaloxime series. These cheap nonbiomimetic pseudomacrocyclic bis(dialkylglyoximato)cobalt complexes have been recognized as powerful electrocatalysts for H2 production, notably in terms of working potential and catalytic rate (80–84). The first study using a cobaloxime catalyst in combination with PS1 as a photosensitizer was reported in 1983 by Lehn et al. (85). This family of catalysts for light-driven hydrogen production was revisited by us in 2008, with a series of supramolecular photocatalysts based on a cobaloxime center (54,57). Following this, various photosensitizers have been combined with cobaloxime complexes, either in a supramolecular mode or in multicomponent systems. Nowadays, these cobalt-based entities are among the most employed catalysts for the construction of light-driven systems for H2 evolution.
The structures of various cobaloxime complexes 3–6 evaluated in three-component homogeneous systems are presented in Fig. 11. The substitution pattern has a great influence on their photocatalytic activity. Although displaying a more positive reduction potential, complex 4, bearing four phenyl substituants, is 13 times less active than 3 (66). Comparison of catalysts 5 and 3, in the presence of PS1 and TEA, has established the superiority of the BF2-bridged structures (20 TON in a 1 h experiment) on their H-bridged analog (2 TON in a 4 h experiment, in the presence of an excess of dmgH2 ligand) (54). This has been attributed to the higher stability toward hydrolysis of the BF2-bridged cobaloximes.
Systems based on cobaloxime 3 as the H2-evolving catalyst were further improved in combination with other metal-based photosensitizers such as the cyclometallated iridium diimine PS5 (Fig. 17), or the tricarbonyl-rhenium diimine PS8 (Fig. 8), resulting in H2 production with 165 (15 h) and 273 (15 h) TON respectively (57). The organic dye PS11 has also been successfully employed with up to 327 TON H2 in a 5 h experiment (CH3CN/H2O, 1:2; 10% TEA) (66).
Cobaloximes 6 bearing an axial pyridine ligand have been studied in combination with the platinum-based photosensitizers (PS6-7) (58,59,61). Under the best experimental conditions established by Eisenberg et al. (CH3CN/H2O, 24:1; TEOA 0.27 m; pH = 8.5), 120 TONCo (2150 turnovers with respect to the Pt-sensitizer) are produced during a 10 h experiment. Replacing the Pt-based photosensitizers by commercially available Eosin Y (PS10) resulted in noble-metal free systems able to achieve up to 180 TONCo (CH3CN/H2O, 1:1; 5% TEOA; pH = 7) (65). The stability of photocatalytic system has been further improved by replacing the fluorescein dyes by synthetic rhodamine derivatives containing S or Se atoms in place of O in the xanthene ring (PS12a-c) (86). The performances of a commercially available organic dye, the acriflavine PS13, have also been recently established in combination with catalyst 6a, resulting in H2 production with 66 TONCo (2 h of irradiation; 110 TON with respect to PS13) in the presence of TEOA (2.5%) in a 3:1 mixture of DMF and H2O (87).
Clearly, these studies have demonstrated the potential of cobalt-based catalysts in light-driven hydrogen production. However, the stability of cobaloxime complexes still needs to be improved for the development of long-term stable systems. We specifically addressed this issue with a second generation of catalysts, the cobalt diimine-dioxime [Co(DO)(DOH)pnBr2] complexes bearing a tetradentate ligand instead of two bidentate dioxime ligands and which proved to display excellent electrocatalytic properties for hydrogen evolution and greater stability under acidic conditions as compared with the cobaloximes (88). Recently, complexes 7-8 (Fig. 12) were combined with the cyclometalated iridium photosensitizer PS4 in a 1:1 ratio, to light-drive hydrogen production in a CH3CN/H2O mixture (10% TEA, pH = 10) (89). Catalyst 8 was shown to be the most efficient one with 307 TONCo achieved in a 4 h experiment. However, it has been established that the performance of the photocatalytic system was still limited by the catalyst decomposition. To improve its stability, a fine tuning of the coordination sphere of the cobalt center was successfully undertaken. Addition of two equivalents of triphenylphosphine allowed to reach up to ca 700 TONCo in a 10 h experiment, thanks to the formation of spectroscopically identified Co(I)(DO)(DOHpn)(PPh3) (10) catalytically active species (89).
Finally, Alberto et al. have recently re-examined the use of cobaloxime 5 in combination with tricarbonyl-rhenium diimine sensitizers (62,63). In the presence of TEOA as the sacrificial electron donor and acetic acid as the proton source in DMF, 75 TON were achieved after 9 h irradiation when PS9a is employed (63). Exchanging the bromide ligand by a thiocyanate one in PS9b conferred a higher stability to the photocatalytic system, thus able to achieve up to 353 TON in a 120 h experiment (62). Under optimized experimental conditions, turnovers up to 1850 per cobalt ion (ca 300 per dimethylglyoxime ligand (dmgH2) added to the solution to restore the integrity of the catalyst) could be achieved, which represents the highest efficiency reported so far for a cobaloxime-based photocatalytic system. Alternatively, tricarbonyl-rhenium diimine complexes have also been successfully used as light-harvesting units in combination with diimine-dioxime cobalt catalysts, resulting in H2 production with up to 550 TONRe (17 TONCo) for complex 9 (64).
Mechanistic issues. The general mechanisms for light-driven H2-evolution catalyzed by synthetic systems are illustrated in Fig. 13. The process is, in any case, initiated by the absorption of a photon by the photosensitizer, yielding the PS* excited state. From the latter, a first photoinduced electron transfer may take place by an oxidative quenching process involving the catalyst as the electron acceptor and generating the active catalytic reduced species. A second electron transfer process then occurs between the sacrificial electron donor and the resulting oxidized photosensitizer PS+, thus regenerating the PS. An alternative mechanism implies first a reductive quenching of the PS* by the sacrificial electron donor to yield the reduced PS−, which subsequently reduces the catalyst to its active state. The following steps, that are metal-centered mechanisms for H2 evolution catalyzed either by a diiron center (47) or by a cobalt catalyst (48,50) have been recently reviewed and are beyond the scope of this article.
Figure 13. Schematic representation of the two plausible photosensitizer-based mechanisms ([A] Reductive quenching process; [B] Oxidative quenching process).
Download figure to PowerPoint
By contrast with multicomponent systems based on platinum colloids or hydrogenase, early studies have revealed that an electron relay between the PS and the molecular synthetic catalyst was not needed to establish an efficient electron transfer (90,91). This is true for all the diiron and cobalt photocatalytic systems described before.
The establishment of one mechanism vs the other will essentially depend on the redox properties of different couples (PS+/PS*, PS*/PS−, Cat/Cat−, D•+/D), and their relative concentrations. It is well-established that a diiron catalytic center cannot be directly reduced by an excited photosensitizer such as *[Ru(bpy)3]2+ or *[Ir(ppy)2(bpy)]+, because its reduction potential is more negative than the oxidation potential of excited PS*, making the direct electron transfer thermodynamically unfavorable (76). Thus, light-driven hydrogen production with diiron catalysts does occur via pathway (A) (Fig. 13). This reductive quenching mechanism also applies when PS1 and cobaloxime 5 are combined for light-driven hydrogen production (85). However, the mechanistic analysis is more complex for the other cobalt-based photocatalytic systems as both pathways (A) and (B) are in many cases thermodynamically feasible (65,66,89). Morevover, when CoIII catalysts (6–9) are employed, two successive electron transfer steps have to be taken into account. A reductive quenching mechanism has been nevertheless evidenced for the hydrogen-evolving systems comprising catalyst 5 and the rhenium photosensitizers PS9 (62) or for the combination of catalyst 6a with the organic dye PS12c (86). By contrast with the photosensitizers PS6, their cyclometallated platinum analogs PS7 cannot be reductively quenched by TEOA and consequently hydrogen production occurs via the oxidative pathway (B) (59). In the other cases, a reductive quenching cannot be excluded, in particular because of the large excess of sacrificial electron donor employed in the photocatalytic experiments.
The nature of sacrificial electron donor also needs to be considered from a mechanistic point of view. Indeed, in most of the previously described photocatalytic systems, an amine (triethanolamine, TEOA, or triethylamine, TEA) is employed. The radical cation species generated after a first electron transfer to the PS is a strong reducing agent, thus able to furnish a second equivalent of electrons to the system, yielding an aldehyde and a secondary amine by-product (Scheme 2) (62,90,92). For instance, a yield of 70% for this second electron transfer has been determined from kinetic studies on a system comprising PS9b and TEOA (62). This dark process should be systematically taken into account in the determination of quantum yield for the photoproduction of H2.
Figure Scheme 2.. Decomposition pathways for the sacrificial electron donors TEA (R = H) and TEOA (R = OH) generating one equivalent of electrons (dark process).
Download figure to PowerPoint
It is clear from the photocatalytic studies described above that optimization of the experimental conditions for a given system is not a trivial process. Indeed, its efficiency strongly relies on a broad range of parameters that cannot be independently optimized. The acidity of the medium is perhaps the most important one to consider. The pH is strongly correlated to the choice of the sacrificial electron donor. Generally, with TEA or TEOA, no H2 production was observed at pH below 6–8 or above 12–13. Indeed, at low pH values, protonation of the amine reduces the amount of redox-active basic form acting as sacrificial electron donor for the regeneration of photosensitizer, whereas at high pH protonation of the catalyst is less favorable. Pronounced solvent effects are also observed for all the photocatalytic systems described in the literature. DMF, CH3CN or acetone generally appear as solvents of choice, in combination with water (up to 66%). Even though such solvent mixtures provide a first step toward using water as a solvent in applications for water splitting, only a limited number of homogeneous photocatalytic systems display activity in pure water (53,64,66). Finally, deactivation of the system is often observed after a few hours of activity, but generally not understood at the molecular level. Bleaching of the dye is the major cause of inactivation in the case of systems based on organic dyes. Metal-polyimine photosensitizers, such as PS1, are also known to photodecompose but rhenium-based photosensitizers are significantly more stable. More generally, it appears crucial to determine whether the deactivation process is an intrinsic feature of the photocatalytic system or is specifically due to the use of a sacrificial electron donor generating potentially damaging radical species. The fact that a mixture of dyes PS12b-c and TEOA rapidly bleaches in the absence of catalyst supports the latter hypothesis.