Recent Developments in Hydrogen Evolving Molecular Cobalt(II)-Polypyridyl Catalysts.

The search for efficient noble metal-free hydrogen-evolving catalysts is the subject of intense research activity. A new family of molecular cobalt(II)-polypyridyl catalysts has recently emerged. These catalysts prove more robust under reductive conditions than other cobalt-based systems and display high activities under fully aqueous conditions. This review discusses the design, characterization, and evaluation of these catalysts for electrocatalytic and light-driven hydrogen production. Mechanistic considerations are addressed and structure-catalytic activity relationships identified in order to guide the future design of more efficient catalytic systems.

1. Introduction

The conversion of solar energy into chemical energy through light-driven water splitting to produce hydrogen (H2) may provide a new method of energy delivery based on the exclusive use of renewable resources, water and sunlight. Hydrogen is indeed a carbon free fuel, and possesses the highest energy output relative to mass [1]. Fuel cell technology allows this energy source to be converted into electricity on demand, with very high energy conversion efficiencies and without the production of greenhouse gases. In the context of water splitting, photosynthesis has been a great source of inspiration for the molecular chemistry community, and the design of molecular systems for hydrogen evolution, inspired by the Photosystem I-Hydrogenase couple found in some hydrogen producing photosynthetic micro-organisms [2-4], is at the heart of the field of artificial photosynthesis [5-10]. Of particular interest is the fact that the activities of the biological systems rely on first-row transition metals, leading to intense research activity in the field of bio-inspired noble metal-free hydrogen evolution catalysis, which may provide an alternative to the use of rare and expensive metals such as platinum in state of the art technological applications [11-15].

While having no biological relevance for water splitting, cobalt has emerged as a major player in the field in the last ten years [16, 17]. The potential use of cobalt-based catalysts in light-driven hydrogen evolution was identified some time ago [17], yet the cobaloxime catalysts (Scheme 1, ) had to wait until the mid-2000s to experience a renewal in interest, thanks to several detailed electrocatalytic studies [18-21]. Numerous efficient homogeneous photocatalytic systems, either multicomponent [22-30] or based on supramolecular photosensitizer-catalyst assemblies [31-39], have since been reported, displaying activity in fully organic or mixed organic-aqueous solutions. However, the cobaloxime catalysts suffer from a low stability, particularly in the reduced state, which limits their long-term efficiency. This drawback has been partly circumvented by substituting the two bidentate dimethylglyoxime ligands (Scheme 1) for the tetradentate diimine dioxime ligand, giving complexes that are more stable toward ligand exchange reactions [40-44].

Scheme 1

Parent cobaloximes and [Co(bpy)3]2+ catalysts.

The 2010s saw the start of a new chapter in the cobalt-based hydrogen production saga, with the first reports on the hydrogen evolution activity of cobalt(II)-polypyridyl complexes. These studies took inspiration from the work of Sutin and coworkers in the 1980s on the cobalt-trisbipyridine complex (Scheme 1, ) [45-50]. The polypyridyl catalysts display an increased stability under reductive conditions compared to both the parent [Co(bpy)3]2+ compound and the cobaloximes, thus allowing a more systematic evaluation of their activity in fully aqueous media. This has stimulated the design of new electro- and photocatalytic systems based on cobalt(II)-polypyridyl complexes, particularly in the last four years.

In this review article, the design, characterization, and evaluation of molecular cobalt(II)-polypyridyl catalysts for electrocatalytic and light-driven hydrogen production will be discussed. Structure-catalytic activity relationships will be specifically highlighted in order to define guidelines for the future design of more efficient catalytic systems. For the sake of clarity, classification of the polypyridyl ligands and their corresponding cobalt(II) complexes has been made on the basis of their structural features (and therefore does not follow the chronological order of publication). They have been divided into two families: the bipyridine-based structures (Schemes 2 and 3, ligands L3a to L9, catalysts 3a to 9) and the pyridine-based structures (Scheme 4, ligands L10a to L12b, catalysts 10a to 12b), each family comprising both tetradentate and pentadentate ligands.

Scheme 2

Tetradentate bipyridine-based ligands and their corresponding cobalt(II) complexes.

Scheme 3

Pentadentate bipyridine-based ligands and their corresponding cobalt(II) complexes.

Scheme 4

Pyridine-based ligands and their corresponding cobalt(II) complexes.

2. Electrochemical and electrocatalytic studies

The electrochemical characterization and assessment of the electrocatalytic performance of a catalyst are important steps in catalyst development, benchmarking [51] and understanding. These experiments provide important information about the electrochemical potential and rate at which catalysis occurs, as well as mechanistic details such as the nature of the catalytic process (cobalt-centered vs ligand-assisted). Due to their relatively high stability, information about electrocatalytic activity under fully aqueous conditions is almost always available. This can then be used to rationalize the photocatalytic hydrogen evolution activities measured under such conditions, provided both electro- and photocatalytic experiments were performed at the same pH. Additional kinetic and mechanistic information can be obtained by performing electrocatalytic studies in organic media, wherein proton delivery to the catalyst can be controlled by the addition of precisely measured amounts of a proton source with a given pKa.

In the following we will describe the different systems, starting with those for which an exclusively cobalt-centered catalytic mechanism is demonstrated. Catalysts exhibiting a ligand-assisted catalytic process will be described in a second part. The electrocatalytic performances of these catalysts are summarized in Table 1. It should be noted that to the best of our knowledge, no electrocatalytic study has been reported for the parent [Co(bpy)3]2+ complex 2.

Table 1

Electrocatalytic performances of hydrogen evolving molecular cobalt(II)-polypyridyl catalysts.

		Conditions			Cyclic Voltammetry			Bulk electrolysis			Ref
	Catalyst	Electrode	Protonsource	Medium	Co(II/I)	Electrocatalyticpotential	Over Potential(±50 mV)	AppliedPotential(overpotential)	TON(reactiontime)	Faradaicyield
Organic electrolytes	3b	Glassy carbon	CF3COOH	CH3CN(0.1 M nBu4NPF6)	−0.81 V vs SCE	−0.81 V vs SCE (mid-wave)	400 mV	−1 V vs SCE	n.i.	99%	[52]
	3b	Glassy carbon	CF3COOH	0.1 M KNO3 in H2O/CH3CN 1:1	−1 V vs SCE	−1 V vs SCE (mid-wave)	–	–	–	–	[52]
	7	Glassy carbon	–	DMF (0.1 M nBu4NPF6)	−0.53 V vs SCE	–	–	–	–	–	[53]
	8a	Glassy carbon	CH3COOH	CH3CN(0.1 M nBu4NPF6)	−1.20 V vs Fc+/Fc	−1.72 & −1.83V vs Fc+/Fc (mid-wave)	530 mV	n.i.	n.i.	90%	[54]
	8b	Glassy carbon	CH3COOH	CH3CN(0.1 M nBu4NPF6)	−1.14 V vs Fc+/Fc	−1.60 & −1.85V vs Fc+/Fc (mid-wave)	460 mV	n.i.	n.i.	90%	[54]
	11b	Glassy carbon	CH3COOH[a]	CH3CN(0.1 M nBu4NPF6)	−1.47 V vs Fc+/Fc[b]	−1.74 V vs Fc+/Fc (mid wave)	500 mV	n.i.	n.i.	100%	[55]
	12b	Glassy carbon	CF3COOH	CH3CN(0.1 M LiClO4)	−1.31 V vs SCE	−1.31V vs SCE (mid wave)	–	–	–	–	[56]
Fully aqueous electrolytes	4a	Glassy carbon	H2O	Ascorbate buffer 0.3 M, pH 4	−0.9 V vs SHE	−0.9 V vs SHE (onset)	–	–	–	–	[57]
	7	Glassy carbon	H2O	Phosphate buffer, pH 7	−0.4 V vs SHE	−0.9 V vs SHE (onset)	480 mV	–	–	–	[53]
	8a	Glassy carbon	H2O	Ascorbate buffer 0.3 M, pH 4	−0.85 V vs SHE	–	–	–	–	–	[54]
	10a	Glassy carbon	H2O	Ascorbate 0.3 M, pH 7	−1.0 V vs SHE	−1.0 V vs SHE (onset)		−1.20 V vs SHE	60 (3h)	100%	[57]
	10b	Glassy carbon	H2O	Ascorbate 0.3 M, pH 7	−1.0 V vs SHE	−1.0 V vs SHE (onset)		−1.20 V vs SHE	40 (3h)	100%	[57]
	10c	Glassy carbon	H2O	Ascorbate 0.3 M, pH 7	−0.8 v vs SHE	−0.8 V vs SHE (onset)		−1.20 V vs SHE	~0 (3h)	-	[57]
	11a	Hg pool	H2O	2 M Phosphate buffer, pH 7	−0.84 V vs SHE	−0.84 V vs SHE & −1.00 V vs SHE	420 mV	–	–	–	[58]
	11a	Glassy carbon (RDE)	H2O	0.1 M Phosphate buffer, pH 7	–	–	–	−0.963 V vs SHE (550 mV)	–	95%	[59]
	11b	Hg pool	H2O	2 M Phosphate buffer, pH 7	−1.00 V vs SHE	−1.00 V vs SHE & −1.20 V vs SHE	660 mV	−1.30 V vs SHE	55 000 (60 h)	100%	[58]
	11c	Hg pool	H2O	2 M Phosphate buffer, pH 7	−1.12 V vs SHE	−1.12 V vs SHE & −1.30 V vs SHE	–	–	–	–	[58]

Anation process.

Converted from E(Fc+/Fc) = +0.64 V vs SHE.

2.1. Brief overview of hydrogen evolution mechanisms in cobalt-based catalysts

Hydrogen evolution catalyzed by molecular cobalt complexes generally implies the formation of a cobalt hydride species by protonation of a Co(I) intermediate [17]. The catalytic system may then evolve through two distinct mechanisms. In the heterolytic mechanism (Figure 1), the intermediate metal hydride, either Co(III)–H or Co(II)–H, decomposes by proton attack and evolves hydrogen via an intermediate dihydrogen metal σ-complex. During the catalytic cycle, two electrons are thus transferred to the complex, either consecutively (Figure 1, ECCE or EECC pathways, with E corresponding to electron transfer and C to protonation steps) or alternating with the two protonation steps (Figure 1, ECEC pathway). The distinction between these two mechanisms has so far proven very difficult to establish, except in a few cases where the second reduction step occurs at a more cathodic potential than the first [60, 61]. In the case of the well-studied cobaloximes and their diimine dioxime derivatives, DFT calculations favour the heterolytic mechanism involving a Co(II)–H as the intermediate species from which hydrogen is evolved [62-65]. In the alternative mechanism, two metal-hydride complexes evolve hydrogen through a reductive elimination reaction, involving homolysis of the Co–H bonds (homolytic pathways, inset Figure 1). Thus, in the course of the catalytic cycle, each metal center undergoes a single monoelectronic reduction process, either before protonation or once the Co(III)–H species is formed. It is important to note that both pathways can coexist [66], and that switching from a heterolytic to a homolytic pathway can simply be a matter of experimental conditions, such as the relative concentrations of protons and catalysts or acid strength, as recently demonstrated by Costentin and Savéant in the case of an iron-porphyrin catalyst [67]. Of note is the fact that strong acids are required for Co(III)–H species to evolve hydrogen in a heterolytic manner [68].

Figure 1

Heterolytic and homolytic (inset) mechanisms for hydrogen evolution catalyzed by a molecular cobalt(II) complex.

The exact nature of the hydrogen evolution mechanism has not yet been investigated with the cobalt(II)-polypyridyl catalysts. As for related amino-polypyridinyl cobalt systems [69], a heterolytic mechanism involving a Co(II)–H species is likely, following the conclusions gained from studies on cobaloximes and related compounds. In addition, the introduction of redox-active moieties such as bipyridine [70] in the polypyridyl coordination sphere of the cobalt center has two major consequences: (i) it stabilizes reduced states of the metal center and thus shifts the cobalt-centered reduction process to more positive potentials. As a consequence, the reduced Co(I) state is less basic and therefore less nucleophilic, and thus requires more acidic conditions for cobalt-centered catalysis to occur; (ii) at the same time, it introduces new ligand-centered reduction processes generating Co(I)(L·−) species, sometimes referred to as a formal Co(0) state. This process is observed at potentials less negative than that which would be required to reduce the cobalt center one step further, i.e. to the spectroscopic Co(0) state. Protonation of this Co(I)(L·−) species can then occur, likely generating a Co(II)–H species (Figure 1, EECC pathways) competent for hydrogen generation.

2.2. Cobalt(II)-polypyridyl catalysts displaying exclusively cobalt-centered catalytic processes

The first hydrogen evolving cobalt(II)-polypyridyl electrocatalyst, the octahedral cobalt(II) complex 3b, was reported in 2010 by Chang and coworkers [52]. These authors designed L3b bearing pyridine moieties to stabilize the metal center in its reduced states, and binding in a tetradentate manner, thus leaving two open coordination sites in cis position on which substrate binding and catalysis may occur. The cyclic voltammogram of 3b recorded in acetonitrile displays a reversible cathodic wave at −0.81 V vs SCE and a quasi-reversible oxidation at +0.87 V vs SCE, assigned to the Co(II/I) and the Co(III/II) couples, respectively. Addition of trifluoroacetic acid (TFA) triggers the appearance of a catalytic wave, near the Co(II/I) couple (the mid-wave of the electrocatalytic wave corresponds to an overpotential for hydrogen production of ~400 mV), attributed to the catalytic reduction of protons. Controlled potential electrolysis experiments at −1.0 V vs SCE confirm formation of hydrogen with 99% faradaic yield [52]. The catalytic wave plateaus at high acid concentrations, with catalytic current independent of the scan rate. This indicates that the system operates in the ‘pure kinetic’ catalytic regime with negligible consumption of the substrate (i.e. protons) [71]. As expected, the plateau current varies linearly with the concentration of the catalyst 3b. However, a similar linear variation is found with regards to TFA concentration, indicating that the catalytic rate is second order in acid. Although such a dependency has been observed relatively frequently for other catalysts, including cobalt-based ones [56, 72-75], it does not fit with any of the classical mechanisms for hydrogen evolution [76]. Nevertheless, a turnover frequency (TOF) of 40 mol H2/mol catalyst/hour at 60 mM TFA could be determined. The catalyst 3b is not soluble in water in the millimolar range, but the authors were able to demonstrate the compatibility of this polypyridyl platform with aqueous conditions by working in a 50:50 water acetonitrile mixture. Under these conditions, electrocatalytic proton reduction is also triggered by addition of TFA at a potential very close to the Co(II/I) reduction wave (−1.0 V vs SCE), with similar current densities to those observed in pure acetonitrile. Of note is a second, possibly catalytic process observed at a slightly more negative potential (−1.2 V vs SCE), but which has not been investigated in detail. We note, however, that this behavior, with consecutive catalytic processes, finds equivalence in the electrochemistry of compounds 10 and 11 discussed below. Also of note is the fact that L3b contains a bipyridine moiety with potential redox activity.

Water solubility was obtained through modification of the ligand [57]. Catalyst 4a is closely related to catalyst 3b, but with a methyl substituent instead of a methoxy group, and a triflate residual ligand on the cobalt center instead of an acetonitrile molecule. Electron-withdrawing groups have also been introduced e.g. in L4b, following the observation of the beneficial effect of such substituents on the series of catalyst 11a-c (see below) [58]. The cyclic voltammograms of complexes 4a and 4b (0.3 mM) were recorded in pH 4 ascorbic acid/ascorbate buffer, the same conditions employed for photocatalytic testing (see section 3.2). Complex 4a shows electrocatalytic current enhancement at approximately −0.90 V vs SHE with no discernible pre-catalytic features. Comparison with 3b indicates that the onset potential for the electrocatalytic wave corresponds to the reduction of the complex to the Co(I) state. In contrast, 4b displays a small pre-catalytic feature at −0.75 V vs SHE, which has been assigned to the Co(II/I) reduction and is likely non-catalytic. In addition, hydrogen evolution catalyzed by 4b occurs at potentials more negative than −0.90 V vs SHE (4a), which contrasts with the trend observed in the series 11a-c (see below). While no detailed mechanistic data are provided in the original paper, it is tempting to propose that the introduction of the electron-withdrawing group in L4b results in the inversion of the redox potentials of the Co(II/I) and the Co(III)–H/Co(II)–H couples of 4b (the latter triggering catalysis) as compared to the situation found for 4a. Also of note is the fact that the catalytic current obtained with 4b is significantly lower than that obtained with 4a. Indeed, the introduction of electron-withdrawing groups in cobalt polypyridyl complexes often leads to diminished activity of the corresponding Co(I) species [54]. Such a trend is confirmed in photocatalytic assays.

The same group has also synthesized a pentadentate pyridine-based “Py5Me2” ligand, L11, and first demonstrated the impressively high efficiency and robustness of the oxo- and disulfide-molybdenum complexes of 11b for hydrogen generation from neutral water [77, 78]. A series of cobalt complexes 11 was subsequently prepared and evaluated in pH 7 phosphate buffer solutions. Two electrocatalytic processes are observed at mercury drop or mercury pool electrodes. The first is peak-shaped and develops at a potential close to the Co(II/I) couple (−1.0 V vs SHE for complex 11b) as independently measured in acetonitrile solutions [7]. The second process occurs at a potential 200 mV more cathodic and seems to be limited only by mass transport of protons to the electrode surface. Controlled potential electrolysis coupled to GC measurements, carried out at a potential below that of the second process (−1.30 V vs SHE), established the stability of 11b during hydrogen evolution, with a 100% faradaic yield and 55000 TONs achieved during a 60 hour experiment. Substitution on the central pyridine ring by either electron-withdrawing (L11a) or electron-donating (L11c) groups allows for tuning of the potential of both hydrogen evolution processes, highlighting the role of molecular design in catalyst optimization [58]. While no mechanistic information is available for this second process, which possibly involves a Co(0) state, rotating disk electrode (RDE) voltammetry allowed further insights into the first catalytic process in the case of complex 11a [59]. From the voltammograms obtained at different rotation rates in 0.1 M pH 7 phosphate buffer, a Levich plot was constructed from the current density at −0.9 V vs SHE, and its linearity indicates catalysis operating under diffusion control. Despite the fact that no bulk electrolysis experiment has been performed at the exact potential of this first wave, the catalytic nature of the process was demonstrated by comparing the current density measured at −0.9 V vs SHE to that of the Co(III/II) process observed at +0.4 V vs SHE. This ratio (almost 16) was much higher than the values of 1 to 8 reported for a series of cobalt complexes at a similar overpotential of 500 mV in pH 2.2 aqueous buffer [79]. The cobalt-centered mechanism of proton reduction by catalysts 11 was also investigated in acetonitrile in the presence of acetic acid [55]. Compound 11b displayed two well-defined reversible redox events at +0.815 V and −0.830 V vs SHE, assigned to the Co(III/II) and Co(II/I) events, respectively. Surprisingly, the addition of acetic acid caused a decrease in the current at the Co(II/I) couple and the appearance of a catalytic wave at a slightly more negative potential (−1.10 V vs SHE). This behavior could be explained through anation of the complex by the conjugate base produced during catalysis, the acetate anion. Indeed, addition of tetrabutylammonium acetate (n-Bu4N(OAc)) to a solution of 11b in acetonitrile caused the appearance of a new couple at −1.18 V vs SHE, with complete disappearance of the original wave after the addition of one equivalent. Furthermore, titration of a 1:1 mixture of n-Bu4N(OAc):AcOH into a solution of 11b showed the development of a catalytic peak at the potential of the Co(II/I) couple of the acetate-bound species, the peak current vs acid concentration demonstrating first order kinetics at all concentrations, while addition of excess n-Bu4N(OAc) to a solution of the acetate complex with acetic acid inhibits catalysis. Addition of acids of various strength to 11b, and also to the CF3 (11a) and NMe2-substituted (11c) compounds showed that in all cases, current enhancement at the Co(II/I) couple, as well as evidence for anation by the conjugate base, were present in the cyclic voltammograms.

Interestingly, no anation was observed by Scandola and coworkers for the cobalt complex with the “Py5” ligand L12b, with a structure very similar to L11b, varying only in the substitution on the methylene linkers [56]. They investigated the electrocatalytic activity of 12b in acetonitrile in the presence of increasing amounts of TFA, the conjugate of which is expected to be a weaker ligand than acetate. A catalytic wave developed at potential slightly more positive than that of the Co(II/I) redox couple (−1.31 V vs SCE) and has been assigned to hydrogen production. It should be noted, however, that no bulk electrolysis experiment coupled to GC measurements is reported in order to confirm the formation of hydrogen. The authors tentatively proposed a mechanism based on the formation of a Co(III)–H intermediate followed by reduction to Co(II)–H. At high acid concentration, however, the observation of catalytic current from very positive potentials (onset at −0.50 V vs SCE) suggests detachment and protonation of one of the pyridine groups on the pentadentate ligand, since reduction of the more positively charged protonated complex is expected to occur at a less cathodic potential. Such a mechanism has already been suggested by Wang and coworkers for a related complex [80].

Ligands L10a-c are structurally related to ligands L11a-c, from which a single pyridine moiety has been removed, and thus allowed for direct comparison of the activity of pentadentate “Py5” versus tetradentate “Py4” ligand complexes [57]. Cyclic voltammograms recorded using a glassy carbon electrode at pH 7 show a plateau-shaped catalytic process with onset potentials ranging from −0.8 V (10c) to −1.0 V (10a-b) vs SHE, preceding direct proton reduction at the glassy carbon electrode. Controlled potential electrolysis (100% faradaic efficiency) carried out at −1.2 V vs SHE confirmed the catalytic nature of these pre-waves with moderate TONCo of 60 and 44 achieved after 3 hours for 10a and 10b, respectively. Surprisingly, 10c proved almost entirely inactive for hydrogen evolution during electrolysis, to account for which the authors have suggested competing pathways in its reductive chemistry.

2.3. Cobalt(II)-polypyridyl catalysts displaying ligand-assisted catalytic processes

Ligand-centered redox processes have been characterized in two series of complexes, 7 and 8a-b, which bear redox-active moieties in the ligand backbone.

Ligand L7 accommodates Co(II) in a square planar environment in 7, with two axial chloride ligands completing the coordination sphere [53]. The cyclic voltammogram of 7 in DMF reveals a reversible oxidation process (Co(III/II), +0.43 V vs SHE) and three reversible reductive systems at −0.53, −0.78 and −1.13 V vs SHE. The first reduction process is assigned to the Co(II/I) couple. The second and third waves, occurring at potentials close to those observed in a voltammogram of the free ligand L7, are likely to be ligand-centered. The electrochemistry of 7 is significantly altered in aqueous electrolytes. Aquation of the two axial chloride ligands readily occurs, and the Co(III/II) system is coupled to deprotonation of water ligands in a proton coupled electron transfer (PCET) process. The first cobalt-centered reduction is still observed as a quasi-reversible process occurring at −0.4 V vs SHE at all pH values. A pH-dependent catalytic wave assigned to hydrogen evolution is observed at more cathodic potentials, thus it is likely triggered by a ligand-centered reduction process. These results show that ligand-assisted catalytic processes can occur at quite positive potentials in aqueous conditions, and are thus particularly relevant under light-driven conditions.

A similar situation is found with the complexes (8a-b) of two pentadentate ligands, L8a-b, based on two bipyridine and one pyridine unit [54]. Cyclic voltammetry of the complex 8a, recorded in acetonitrile, displays three cathodic processes at −1.20 V, −1.79 and −1.94 V vs Fc+/Fc, assigned to the Co(II/I) event and the successive reduction of the two bipyridine units in L8a. The Co(III/II) couple appears as a broad event centered at +0.235 V vs Fc+/Fc. The CF3-substituted complex 8b shows the expected positive shift of the metal-centered Co(III/II) and Co(II/I) redox couples, by 75 and 61 mV, respectively. More unexpectedly, the ligand based reductions also moved to more positive potentials, by around 80 mV. In both cases, addition of acetic acid triggers the appearance of electrocatalytic waves corresponding to the reduction of protons into hydrogen, with faradaic yields greater than 90%, as confirmed by bulk electrolysis experiments. Two distinct systems, at potentials matching the redox events assigned to ligand-centered reductions, are observed. In the case of 8a, these peaks are concentration dependent up to 90 mM acetic acid and plateau afterwards, with the second peak giving even higher current densities. For 8b, the first event shows relatively low current enhancement and levels off quite quickly with the acid concentration. The second event, however, shows much greater current enhancement, although not as high as in the case of 8a. The authors also recorded cyclic voltammograms in aqueous electrolyte (0.3 M ascorbate buffer, pH 4) on glassy carbon electrodes. Under these conditions, only the Co(II/I) system is observed, the ligand-centered redox processes occurring at potentials where direct proton reduction is readily achieved at the glassy carbon electrode.

No cobalt-centered catalytic process is observed for the complexes 8a-b, which contrasts with the related pentadentate catalysts 11a-b, which lack any redox-active ligand. This is likely due to the fact that the introduction of bipyridine moieties in the ligand stabilizes the cobalt center in the reduced state through increased π-back-bonding. This both shifts the Co(II/I) redox couple to more positive potentials and reduces the nucleophilicity of the Co(I) state. Hence a stronger acid is required for catalysis to be observed. It is likely that a catalytic process could be observed with complexes 8a-b if TFA was used as a proton source, i.e. under conditions similar to those used for assaying the catalytic activity of 3b, the ligand of which also contains a bipyridine unit, although only binding in a tetradentate manner. Comparing the activity of 3b and 8a under similar conditions would also be very interesting in order to verify if the presence of two open or labile coordination sites in cis position is beneficial for catalysis.

Despite the large number of electrochemical and electrocatalytic studies available for cobalt(II)-polypyridine complexes, it appears difficult to derive structure-function relationships for this class of compounds, principally because of the large range of experimental conditions used. In particular, data recorded in aqueous solutions are sometimes not exploitable, because the solutions are not properly buffered or because direct hydrogen evolution at the surface of the electrode readily competes with the catalytic process [54].

3. Photocatalytic studies

The light-driven activities of several cobalt(II)-polypyridyl catalysts and of the parent [Co(bpy)3]2+ (2) have been evaluated under homogeneous conditions, using a photosensitizer able to harvest light and initiate photoinduced electron transfer, together with a sacrificial electron donor able to regenerate the photosensitizer to its initial state, thus rendering the system catalytic. The structures of the main photosensitizers employed in combination with catalysts 2-12 are presented in Scheme 5, and the photocatalytic performances of these systems are reported in Table 2. Note that unless specifically stated, the formation of cobalt colloids or nanoparticles has been excluded on the basis of the mercury poisoning test or dynamic light scattering measurements in the photocatalytic studies described below.

Scheme 5

Main photosensitizers employed in combination with [Co(bpy)3]2+ (2) and cobalt(II)-polypyridyl catalysts.

Table 2

Photocatalytic performances of homogeneous systems based on hydrogen evolving [Co(bpy)3]2+ and cobalt(II)–polypyridyl catalysts.

	Cat	[Cat]	PS	Ratio PS : Cat	Conditions	pH	Light	Irr. Time	Amount of H2	TON (H2/Co)	TOF (h−1)	Φ	Ref.
Organic and mixed organic/aqueous medium	2	5 mM	PS2	1 : 20	CH3CN/H2O, 1:1TEOA (0.5 M)LiCl (0.25 M)	8.0	≥ 400 nm	–	–	–	–	29%	[48]
	2	2.5 mM	PS1	1 : 50	CH3CN/H2O, 1:1TEOA (0.57 M)LiCl (0.27 M)	– [a]	LED465 nm	o/n	50 μmol	2	–	–	[81]
	2	2.5 mM	PS2	1 : 50	CH3CN/H2O, 1:1TEOA (0.57 M)LiCl (0.27 M)	– [a]	LED465 nm	o/n	290 μmol	12	–	–	[81]
	2	2.5 mM	PS3	1 : 50	CH3CN/H2O, 1:1TEOA (0.57 M)LiCl (0.27 M)	– [a]	LED465 nm	o/n	400 μmol	16	–	–	[81]
	2	5 mM	PS2	1 : 33	CH3CN/H2O, 1:1TEOA (0.57 M)LiCl (0.27 M)	– [b]	LED465 nm	40 h	145 μmol	2.9[c]	–	–	[82]
	2	5 mM	PS4	1 : 33	CH3CN/H2O, 1:1TEOA (0.57 M)LiCl (0.27 M)	– [b]	LED465 nm	40 h	265 μmol	5.3[c]	–	–	[82]
	2	0.4 mM	PS6	1 : 1	CH3CN/H2O, 1:1TEA (10% v/v)	10	> 450 nm	2 h	382 μmol	191	–	–	[83]
	2	40 μM	PS6	10 : 1	CH3CN/H2O, 1:1TEA (10% v/v)	10	>450 nm	2 h	126 μmol	631	–	–	[83]
	2	0.6 mM	PS3	1 : 20	CH3CN, TEOA (0.2 M)p-cyanoanilinium (0.025 M)	–	>390 nm	20 h	72 μmol	12[c]	–	–	[84]
	[Co(bpy)2(OH2)2]2+	0.6 mM	PS3	1 : 20	CH3CN, TEOA (0.2 M)p-cyanoanilinium (0.025 M)	–	> 390 nm	20 h	66 μmol	11[c]	–	–	[84]
	5a	0.6 mM	PS3	1 : 20	CH3CN, TEOA (0.2 M)p-cyanoanilinium (0.025 M)	–	> 390 nm	20 h	162 μmol	27[c]	–	–	[84]
	5a	0.6 mM	PS4	1 : 20	CH3CN, TEOA (0.2 M)p-cyanoanilinium (0.025 M)	–	> 420 nm	20 h	306 μmol	51[c]	–	–	[84]
	5a	0.6 mM	PS4	1 : 20	CH3CN/H2O, 95:5 TEOA (0.2 M)p-cyanoanilinium (0.025 M)	–	> 420 nm	20 h	348 μmol	58[c]	–	–	[84]
	5a	0.6 mM	PS4	1 : 20	CH3CN/H2O, 95:5 TEOA (0.2 M)	–	> 420 nm	20 h	270 μmol	45[c]	–	–	[84]
Fully aqueous medium	L2a + Co(II)	2 mM L2a + 10 mM Co(II)	PS1	min. 1 : 4	H2A/HA− (0.7 M)	5.0	Visible light	1-3 h	–	–	–	2%	[45]
	L2b + Co(II)	2 mM L2b + 10 mM Co(II)	PS1	min. 1 : 4	H2A/HA− (0.7 M)	5.0	Visible light	1-3 h	–	–	–	13%	[45]
	3a	0.1 μM	PS5	5000 : 1	H2A/HA− (1 M)	4.1	LED385 nm	20 h	30 μmol	9000	–	–	[85]
	3a	0.5 mM	PS5	1 : 16.7	H2A/HA− (1 M)	4.1	LED385 nm	120 h	520 μmol	104[c]	–	–	[85]
	3a’	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.0	LED452 nm	14 h	190 μmol	950[c]	–	–	[57]
	3b	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.5	LED452 nm	14 h	205 μmol	1025[c]	–	–	[57]
	4a	40 μM	PS1	8.25 : 1	H2A/HA− (0.5 M)	4.0	LED452 nm	–	–	–	–	7.5 ± 0.8%	[57]
	4a	1 μM	PS1	330 : 1	H2A/HA− (0.3 M)	4.0	Simulated Sunlight (100 mW.cm−2)	2.5 h	–	~4200	~3160	–	[57]
	4a	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.0	LED452 nm	14 h	370 μmol	1850[c]	–	–	[57]
	4b	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.5	LED452 nm	14 h	80 μmol	400[c]	–	–	[57]
	5a	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	5.5	LED452 nm	14 h	40 μmoL	200[c]	–	–	[57]
	6	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	5.0	LED452 nm	14 h	30 μmoL	150[c]	–	–	[57]
	7	3 μM	PS1	133 : 1	H2A/HA− (0.3 M)	4.0	LED469 nm	3 h	10 μmoL	333	586	–	[53]
	8a	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.0	LED452 nm	13 h	326 μmol[d]	1630	660	3.6%	[54]
	8b	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	4.5	LED452 nm	13 h	278 μmol[d]	1390	500	2.7%	[54]
	9	0.1 μM	PS5	5000 : 1	H2A/HA− (1 M)	4.1	LED385 nm	–	–	10800	–	–	[86]
	9	5 μM	PS5	100 : 1	H2A/HA− (1 M)	4.1	LED385 nm	15 h	69 μmol	1380	920	–	[86]
	9	1 μM	PS1	500 : 1	H2O, TCEP[e] (0.1 M)NaHA (100 mM)	5.0	LED385 nm	11 h	290 μmol	33300	5900	–	[87]
	9	100 μM	PS1	5 : 1	H2O, TCEP[e] (0.1 M)NaHA (100 mM)	5.0	LED385 nm	35 h	970 μmol	1080	70	–	[87]
	10a	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	5.0	LED452 nm	–	310 μmol[d]	1550	–	–	[57]
	10a	1.25 μM	PS1	264 : 1	H2A/HA− (0.3 M)	7.0	LED452 nm	–	–	2400	–	–	[57]
	10b	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	5.5	LED452 nm	–	50 μmol[d]	250	–	–	[57]
	10c	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	5.0	LED452 nm	–	45 μmol[d]	225	–	–	[57]
	11a	250 μM	PS1	1 : 1.25	Phosphate buffer (1 M)H2A/HA− (0.1 M)	7.0	≥ 455 nm	2 h	–	–	–	0.6%	[59]
	11a	50 μM	PS1	4 : 1	Phosphate buffer (1 M)H2A/HA− (0.1 M)	7.0	≥ 455 nm	8 h	0.50 mL	–	–	–	[59]
	11a	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	6.0	LED452 nm	13 h	~60 μmol	~300	–	–	[54]
	11b	50 μM	PS1	4 : 1	Phosphate buffer (1 M)H2A/HA− (0.1 M)	7.0	≥455 nm	8 h	0.42 mL	–	–	–	[59]
	11b	20 μM	PS1	16.5 : 1	H2A/HA− (0.3 M)	6.0	LED452 nm	13 h	~60 μmol	~300	–	–	[54]
	11c	50 μM	PS1	4 : 1	Phosphate buffer (1 M)H2A/HA− (0.1 M)	7.0	≥455 nm	8 h	0.25 mL	–	–	–	[59]
	12a	5 μM	PS5	100 : 1	H2A/HA− (1 M)	4.1	LED385 nm	40 h	59 μmol	1180	–	–	[86]
	12b	50 μM	PS1	10 : 1	Acetate buffer (1 M)H2A/HA− (0.1 M)	4.0	>400 nm	1 h	45 μmol	187	486	–	[56]

pH adjusted by addition of 0.4 mL of 12 M HCl.

pH adjusted by addition of 0.125 mL of 37% HCl.

TON(H2/Co) not given in the reference paper but calculated from the reported TON(PS) or from the reported amount of produced hydrogen.

Amount of hydrogen calculated from the reported TON value, the catalyst concentration and the solution volume.

Tris(2-carboxylethyl)phosphine (TCEP).

3.1. Photocatalytic systems based on the parent [Co(bpy)3]2+ complex

Until the late 1970s, most of the work in the area of photocatalytic hydrogen evolving systems relied on the use of [Ru(bpy)3]2+ (PS1, Scheme 5) as the photosensitizer in combination with platinum colloids as the heterogeneous hydrogen evolving catalyst [88, 89]. In these systems, cobalt complexes were initially employed as alternative electron relays to methyl-viologen [90-92]. However, blank experiments revealed that hydrogen could also be evolved in the absence of any Pt catalyst [92, 93]. Importantly, these early reports established that in the case of homogeneous cobalt-based hydrogen photoproduction, no electron mediator is required, in contrast to systems based on platinum colloids. In this context, cobalt-bipyridine systems have been extensively studied by Sutin and coworkers [45-50]. Visible light-induced hydrogen evolution was demonstrated as early as 1981, when Co(II) ions and bipyridine were mixed in water at the optimum pH of 5, together with PS1 and ascorbate in large excess as the sacrificial electron donor [45]. Formation of dihydrobipyridine as a consequence of the reduction of bipyridine was noticed as a side reaction. When bipyridine (L2a) was replaced by 4,4′-dimethylbipyridine (L2b) in the catalytic system, a higher quantum yield for hydrogen production was obtained (13% with L2b vs 2% with L2a) [45]. This work has been further improved by using ruthenium tris-dimethylphenanthroline [Ru(dmphen)3]2+(PS2, Scheme 5) as the photosensitizer, triethanolamine (TEOA) as the sacrificial electron donor and [Co(bpy)3]2+(2) as the catalyst precursor in a 1:1 mixture of acetonitrile and water at pH 8 [48]; under these conditions, a maximum quantum yield of 29% was obtained.

More recently, the chloride salt of 2 was investigated by the group of Bernhard in combination with substituted heteroleptic cyclometallated iridium complexes of the [Ir(ppy)2(bpy)]+ family (PS3, Scheme 5), prepared by a combinatorial chemistry approach [81]. In water acetonitrile mixtures, and in the presence of LiCl and TEOA as the sacrificial electron donor, 16 turnovers based on 2 (TONCo) were achieved by the photocatalytic system, with no significant dependence on the choice of iridium-based photosensitizer (±2 TONs), while the Ru-based PS1 only allowed for 2 turnovers. Use of PS2 instead of PS1 improved the efficiency by a factor of 6, as already observed by Sutin and co-workers [48]. Following this study, 2 has been employed as a reference catalytic system for the evaluation of novel iridium-based photosensitizer structures [82, 94-96], such as charge-neutral amidinate [95] or tricyclometallated derivatives [96]. The reader should note that in all the aforementioned studies, 2 is employed in relatively high excess compared to the iridium-based PS (PS:Cat ratio of 1:20 to 1:50; see table 2), thus accounting for the low TONCo reported in Table 2.

In 2011, Wang, Sun and coworkers reported fully noble metal-free hydrogen evolving photocatalytic systems [83]. Commercially available organic dyes from the xanthene family (PS6-8) have been successfully combined with 2 and triethylamine (TEA) as the sacrificial electron donor in an acetonitrile water mixture at pH 10. Under these conditions, the hydrogen evolution efficiency proved to be considerably higher (up to 631 TONCo) than that of similar systems based either on PS1 or on neutral cobaloximes (1b,c) as catalysts. Electrostatic attraction between the negatively charged organic dyes and the cationic catalyst has been indicated by UV-vis spectroscopy and proposed as an explanation for the superior performance of these photocatalytic systems.

Finally, a supramolecular strategy, developed by Sakai and coworkers exploited the spontaneous self-assembly of bipyridine-appended cyclometallated iridium photosensitizers in the presence of Co(II) ions, thus generating a [Co(bpy)3]2+ center in situ, bearing one, two or three appended photosensitizing units [97]. In the presence of TEOA, these assemblies mediated light-driven hydrogen production in acetonitrile water mixtures with up to 20 TONCo. Multi-component systems tested under the same experimental conditions proved to be half as efficient, demonstrating the importance of the supramolecular design.

3.2. Photocatalytic systems based on cobalt(II)-polypyridyl catalysts

Since the first report in 2010 by Chang and coworkers on the cobalt(II)-polypyridyl electrocatalyst 3b [52], a great deal of effort has been devoted to the design of original polypyridyl ligands and the evaluation of their cobalt complexes in light-driven hydrogen evolution. In contrast with the cobaloximes, no supramolecular or covalent photosensitizer-catalyst assembly has to date been reported for cobalt(II)-polypyridyl catalysts, to the best of our knowledge.

In 2012, Lau and coworkers reported the hydrogen evolution activity of the bis-aqua catalyst 5a in combination with either PS3 or the fluorinated derivative PS4, in presence of TEOA as sacrificial electron donor and p-cyanoanilinium tetrafluoroborate as proton source [84]. Under optimized conditions (PS4:Cat ratio of 1:20), 58 TONCo have been measured in a 95:5 acetonitrile water mixture. After 20 hours, addition of extra photosensitizer could reinitiate catalysis. In contrast, catalyst 2 and the related [Co(bpy)2(OH2)2]2+ display much lower TONs, and catalysis could not be reinitiated by addition of PS. It should be noted that this study is the only one of the polypyridyl series reported under organic conditions and producing a clear comparison with the parent catalyst 2. In addition, catalyst 5a was recently reevaluated by Long, Chang, Castellano and coworkers under fully aqueous conditions (PS1, ascorbic acid as sacrificial electron donor at pH 5.5) and performed up to 200 TONCo which ranks it among the less active polypyridyl catalysts of the tested series (from 150 to 1850 TONCo for catalysts 3a-6 and 10a-11c tested under identical conditions; see section 3.3 below for more detailed comparison) [57].

Hydrogen evolving photocatalytic activity of cobalt(II)-polypyridyl catalysts in fully aqueous medium was first reported in early 2013 with catalysts 11a-c [59] and 3a [85]. Long, Chang and coworkers have demonstrated that the family of complexes 11a-c, previously described as highly efficient electrocatalysts in neutral water [58], were also active under visible light-driven conditions in the presence of PS1 and ascorbate in aqueous phosphate buffer at pH 7. Catalyst 11a exhibited the highest performance in the series, in good agreement with its low overpotential for electrocatalytic hydrogen evolution [58]. The authors established that photosensitizer decomposition was the primary reason for the cease of activity after 8 hours of photolysis. Hydrogen evolution was also demonstrated in combination with a semiconductor nanowire photosensitizer, which is, however, out of the scope of this review [59]. At the same time, Hamm, Alberto and coworkers have prepared and tested the dibromo complex 3a in combination with the rhenium-based photosensitizer PS5 in ascorbate buffer at pH 4.1 [85]. Up to 9000 TONCo were obtained, however at very low concentration of catalyst (0.1 μM) and extremely high PS:Cat ratio (5000:1). The long-term stability of the system has been studied using a higher concentration of catalyst (0.5 mM) producing 0.52 mmol hydrogen (104 TONCo) over 120 hours. At the end of the photocatalysis experiments, LC-MS analysis showed the presence of both catalyst and PS5 (≥ 90%), demonstrating their stability. Two new pentadentate ligands and their cobalt catalysts have also been reported by the same authors [86]. The first, L9, displays a “Bpy2Py” coordination sphere whereas the second ligand, L12a, relies on a “Py5” scaffold closely related to L11b [58, 59]. From a series of first-row transition metal complexes (Mn(II), Fe(II), Co(II), Ni(II), Cu(II), Zn(II)) evaluated as water reduction catalysts, the cobalt complexes alone displayed some hydrogen evolution activity. TONCo as high as 1380 for 9 and 1180 for 11a were recorded, using 100 equivalents of PS5 in aqueous solution of ascorbate (pH 4.1). Loss of catalytic activity under these conditions was assigned to competitive back electron transfer from the reduced photosensitizer PS− to dehydroascorbic acid (DHA), formed by oxidation of ascorbic acid [85, 86]. Such self-inhibition by the oxidized form of the sacrificial electron donor, accumulated during the course of the catalysis, has been also identified in other studies as a limitation of the ascorbate-based systems [56, 57]. This drawback has been recently elegantly circumvented by using the reversible ascorbate system as an electron relay between the photosensitizer (PS1 or PS5) and a tertiary phosphine-based irreversible electron donor (tris(2-carboxyethyl)phosphine, TCEP) [87]. Higher hydrogen evolution rates and total amounts of hydrogen could thus be obtained. This work from Alberto and coworkers opens new perspectives for the optimization of light-driven hydrogen evolution under fully aqueous medium.

Two other examples of “Bpy2Py”-type ligands, L8a-b, were later reported by Chang, Long, Castellano and coworkers [54]. In good agreement with the electrocatalytic studies, catalyst 8b, bearing electron-withdrawing CF3 groups, displayed a lower light-driven activity than 8a (1630 TONCo with Φ 3.6% versus 1390 TONCo with Φ 2.7%, respectively). The authors demonstrated that under the reported experimental conditions, photocatalytic activity was solely limited by photosensitizer stability.

In 2014, Thummel and coworkers tested complex 7, a hydrogen evolving electrocatalyst active under neutral aqueous conditions, in combination with the classical PS1-ascorbate system, and up to 333 TONCo were measured during the course of a three hour experiment under LED irradiation [53]. Meanwhile, Scandola and coworkers have evaluated catalyst 12b in presence of PS1 in aqueous ascorbate buffered solution (pH 4) [56]. Under visible light irradiation, 187 TONCo were achieved in one hour. The authors have shown that TONCo values were insensitive to catalyst concentration, in contrast with what is usually observed for the other reported polypyridyl cobalt-based systems.

Finally, Long, Chang, Castellano and coworkers have undertaken the benchmarking of 10 distinct catalysts (3-6, 10-11) under identical conditions (2×10−5 M catalyst, 3.3×10−4 M PS1 in 0.3 M aqueous ascorbate buffer; irradiation at 452 nm) in order to provide comprehensive findings and guidelines for the design of optimized systems [57]. A systematic adjustment of the pH has been made for each individual catalyst in order to take into account the variations in relative basicity of the Co(I) intermediate likely to be protonated during the cobalt-based catalytic cycle (see section 2.1 for mechanistic details). In this study, new ligands (L4a,b and L10a-c) were synthesized to more accurately evaluate parameters such as electronic effects of substituents (L4 series) and the tetradentate versus pentadentate nature of the ligands (L10 series). The results from this study are highlighted in the next section.

3.3. Structure-photocatalytic activity relationships

Specific catalyst structure-photocatalytic activity relationships arise from the reported studies on cobalt(II)-polypyridyl photocatalytic systems. Although a straightforward comparison of the activities reported in Table 2 cannot be made for all the systems due to differences in the experimental conditions employed, a strong influence of the ligand structure and coordination sphere of the cobalt center is observed. A series of structural relationships identified in the cobalt-polypyridyl catalyst series are listed below and their influence on the photocatalytic activity tentatively assigned; as more than one relationship often exists between the catalysts considered, the discussion necessarily overlaps at some points.

Tris-bipyridine vs polypyridyl

Going from the simple bipyridine ligand L2a to polypyridine structures with higher denticities is proposed to confer higher stability to the cobalt complex in its reduced states by virtue of the chelate effect, and as a consequence improve the efficiency of the catalytic system. The parent complex 2 is known to be active upon the loss of one bipyridine ligand [49], whereas penta- and tetradentate ligands allow one and two coordination sites, respectively, to remain vacant on the cobalt center for catalysis to occur via formation of a cobalt hydride species (see section 2.1). There is only one example in the literature where 2 and a cobalt(II)-polypyridyl catalyst are compared under similar conditions: Lau and coworkers have shown that the tetradentate quaterpyridine catalyst 5a displays much higher TONs than 2 and the related [Co(bpy)2(OH2)2]2+; moreover, whereas catalysis could be reinitiated by addition of photosensitizer in case of 5a, full decomposition of the catalyst was observed with 2 and [Co(bpy)2(OH2)2]2+ [84]. This result can be correlated with the formation of dihydrobipyridine by reduction of the bipyridine ligand observed in the early works from Sutin and coworkers [45]. It should be noted, however, that the relative arrangement of the two aqua ligands differs between these complexes, trans in 5a versus presumably cis in [Co(bpy)2(OH2)2]2+, which could also influence the reactivity (see below). In contrast, the stability of catalyst 3a has been assessed by LC-MS analysis of the photocatalytic system, and it was established that 96.5% of 3a is still present after 120 hours of irradiation [85]. Similarly, HPLC control experiments established that 90% of 4a remained intact at the end of the photocatalytic test (~2 hours) [57]. Thus, no reduction of the bipyridine moieties occurred on the tetradentate bpy-based ligands L3a and L4a.

Bipyridine-based vs pyridine-based ligands

Two main consequences stem from the presence of a bipyridine moiety in the ligand structure as compared with single pyridine moieties: (i) more sterically constrained geometries around the cobalt center are obtained and (ii) ligand-centered redox processes can occur (see section 2.3). The former effect has been put forward by Hamm, Alberto and coworkers to explain the higher activity of 9 compared to 12a [86]. Complex 9 indeed displays a much more strongly distorted octahedral structure compared to that of 12a, and was shown to catalyze hydrogen evolution at faster rate. However, these two series of complexes also differ due to more subtle effects linked to the orientation of the pyridyl rings of the ligand within the complexes: while π-π overlap between the pyridyl rings in the 11-12 series is possible, in the series 8-9, the positioning of the four rings in the equatorial plane prevents it. The photocatalytic activity of pentadentate catalysts 8a-b, bearing a redox-active bipyridine, clearly outperformed catalysts 11, relying on a “Py5” structure [54]. The superiority of bipyridine-based structures is, however, less marked in the tetradentate series, where “BpyPy2” ligands L3a-L4b have been compared to “Py4” ligands L10a-c under similar conditions [57]. Catalyst 4a displayed the highest activity with 1850 TONs achieved in 14 hours, however it is closely followed by 10a (1550 TONs). Electronic and steric effects of the substituents also come into play when considering the other complexes of the two series, 3a’ (950 TONs), 3b (1025 TONs), 4b (400 TONs), 10b (250 TONs) and 10c (225 TONs). The contribution of ligand-centered redox processes to hydrogen evolution activity is thus difficult to properly demonstrate under light-driven conditions.

Pentadentate vs tetradentate ligands

During the course of their benchmarking of cobalt(II) catalysts, Long, Chang, Castellano and coworkers have designed three new tetradentate ligands L10a-c, based on the parent pentadentate structure L11b, but with one fewer pyridine moiety [57]. These two series of ligands thus allow for an accurate comparison of the influence of the ligand denticity on the hydrogen evolution activity of the corresponding cobalt catalysts. It was established that tetradentate ligands led to significantly more active catalysts (1550 TONCo for 10a versus 300 TONCo for 11b under the same experimental conditions), but this effect cannot be generalized since more sterically hindered catalysts 10b and 10c display activities 6 to 8 times lower than 10a, reflecting the trend observed in the course of electrochemical studies [57]. The structures of catalysts 3a [85] and 9 [86] also only differ by the denticity of their ligands, tetradentate for L3a versus pentadentate for L9. Under similar photocatalytic conditions (5 μM in catalyst, 100:1 PS5:Cat ratio, ascorbate buffer pH 4.1), 3a proved to be slightly superior, with ~2000 TONs achieved vs 1380 TONs for 9. From this analysis, it is not possible to demonstrate a clear advantage of tetradentate ligands over pentadentate ligands. It should also be mentioned that a closely related pentadentate system [80] undergoes decoordination of one pyridine arm in the course of catalysis (described in more detail later in the discussion), thus leading to a pseudo-tetradentate system.

Cis vs trans open coordination sites

Catalysts bearing tetradentate bipyridine-based ligands (3-7 in Scheme 2) can be distinguished by the cis/trans stereochemistry of the two labile positions on the cobalt complex. The cis stereochemistry in complexes 3a-4b is controlled by ligands bearing one bipyridine moiety plus two appended pyridines, whereas trans complexes 5a-7 rely on two diimine moieties (bipyridine or phenanthroline) in their coordination sphere. In order to establish the influence of the cis/trans geometry on the light-driven hydrogen evolution activity, the groups of Castellano, Long and Chang have benchmarked catalysts 3a-6 (initially described either in their groups or in other research groups) under identical experimental conditions. This allowed them to establish that catalysts with cis open coordination sites significantly outperform those with trans sites (up to 1850 TONCo for 4a versus 200 and 150 TONCo for 5a and 6, respectively) [57].

Nature of the labile monodentate ligand(s) on the cobalt center

Subtle variations of the monodentate ligand(s) that complete the coordination sphere of the catalyst are observed. They can be monoanions, such as Cl−, Br−, OTf−, or solvent molecules (CH3CN; H2O), thus leading to either neutral, monocationic or dicationic cobalt catalysts. The nature of these ligands is generally dictated by the choice of the cobalt salts used for the metallation reaction and also by some solubility issues in water. Although ligand L3a has, for instance, been employed by two independent groups, with different labile ligands on the cobalt center, the dicationic hexacoordinated bis-acetonitrile complex 3a’ [57] and the monocationic pentacoordinated bromo complex 3a [85], a direct comparison of their relative efficiencies is precluded as very different conditions for their evaluation have been used (PS1 for 3a’ versus PS5 for 3a, in different PS:Cat ratios). Chloride substitution by water has been electrochemically observed for 7 under aqueous conditions. Anation by the acetate anion occurred during the course of the electrocatalytic study on catalysts 11a-b, shifting the redox potential of the Co(II/I) couple toward more cathodic potentials. If such a monodentate ligand remains coordinated in the Co(I) state, it may also influence the protonation steps. These labile positions are thus highly sensitive to their environment.

During the course of photocatalysis, formation of a cobalt hydride intermediate (see section 2.1) is expected to occur with displacement of one monodentate ligand; at the end of the hydrogen evolution catalytic cycle, the cobalt center will be recovered either in a solvated form or coordinated by a new monodentate ligand, depending on the anion present in the medium. The situation could be a little bit trickier for catalysts based on pentadentate ligands. It has been established on a closely related pentadentate aminopyridinic system that formation of Co(III)–H occurs via decoordination of one pyridine arm of the ligand, rather than by removal of the chloride monodentate ligand from the cobalt center [80]. Indeed, four catalysts differing only by the nature of the monodentate ligand (Cl−, NO3−, OTf−, H2O) displayed very different activities, the chloride derivative being the most active and the nitrate five times less active. Furthermore, activity was enhanced in 0.3 M NaCl solution, and decoordination of one pyridine arm was established by NMR under acidic conditions. This highly counter-intuitive result emphasizes the necessity to correctly establish the cobalt-center hydrogen evolution mechanism in order to optimize the photocatalytic activity of such systems. The influence of the so-called labile positions on the cobalt center deserves to be more systematically studied in the future.

Ligand substituent electronic effects

The first pentadentate family of ligands L11a-c studied by Chang and coworkers consisted of three structures differing only in the central pyridine para-substituent: electron-withdrawing -CF3 group in L11a, -H in L11b and electron-donating -NMe2 group in L11c. In excellent agreement with the electrocatalytic studies, catalyst 11a proved to be the most efficient of the series, producing twice the amount of hydrogen of 11c after 8 hours of visible light irradiation (LED 452 nm) [59]. On the basis of these results, the same authors designed two new sets of ligands 4a,b [57] and 8a,b [54], 4b and 8b bearing CF3 substituents on the pyridine moieties. However, the latter displayed lower activities than the parent H-substituted 3a and 8a under both electro- and photocatalytic conditions [54, 57]. Of note is the fact that these two series of complexes possess redox-active bipyridine moieties, in contrast to 11a-c, with a ligand-assisted catalytic process being clearly established for 8a-b (see section 2.3). Substitution of CF3 on the bipyridine rather than on the pyridine moiety would therefore be interesting to investigate, in order to electronically modulate the ligand-assisted process.

Another type of substituent effect is addressed in the 3a’,b-4a series of catalysts, the ligand structures of which vary only in the substitution on the central quaternary carbon (Me, OH or OMe). The recent comparative study by Long, Chang, Castellano and coworkers established that the presence of the methyl substituent confers a greatly enhanced catalytic activity in comparison with the OH or OMe derivatives (1850 TONCo for 4a vs 950 TONCo for 3a’ and 1025 TONCo for 3b) [57]. However, such a variation in activity is difficult to rationalize on the basis of simple electronic effects. Modification of the substitution of a bridging carbon is also encountered in the series 10a-c. The increase in steric hindrance, observed in going from 10a (bridging CH2) to 10c (bridging C(Me)2), drastically influences the light-driven activity of these catalysts, with 10a being one of the more efficient in the field while 10b and 10c display low activities [57]. This effect is also observed in the electrocatalytic studies.

3.4. Mechanistic considerations

One important development with the cobalt(II)-polypyridyl catalysts has been the ability to perform photocatalytic tests under fully aqueous conditions, thanks to the increased stability of the cobalt coordination sphere towards both hydrolysis and reductive conditions. This has allowed the commonly used amine sacrificial electron donors (TEA, TEOA), which require working at basic pH, to be avoided. The ascorbic acid/ascorbate couple, active at the optimal pH of 4-5, can be employed instead, allowing conditions for the photocatalytic tests to better match those favoring hydrogen evolution. It should be noted, however, that the pH of the medium is only precisely defined in fully aqueous conditions; strong uncertainty exists for both the pH value and pKa of the sacrificial electron donor when pure organic or mixed organic/aqueous solutions are employed.

The photosensitizer-based mechanisms are presented in Figure 2. In contrast with electrocatalysis performed under the control of a potentiostat, the reductive power of the medium cannot be adjusted on demand under light-driven conditions, but is fixed by the redox levels of the photosensitizer. In addition, the concentration of reducing equivalents is controlled by sequential monoelectronic processes induced by photon absorption and thus limited by the photon flux. The reductive species for the Co(II) center can be either PS* via an oxidative quenching pathway (Figure 2, right) or PS− via a reductive quenching mechanism (Figure 2, left). Mechanistic studies are therefore mostly focused on the elucidation of the photosensitizer-based processes, through the use of time-resolved spectroscopic measurements that allow observation of signatures associated with the intermediate forms of the photosensitizer.

Figure 2

Photosensitizer-based processes involved in light-driven hydrogen evolution catalyzed by cobalt(II)-polypyridyl complexes.

When [Co(bpy)3]2+ (2) is used as the catalyst in a photochemical hydrogen evolving system, it can be reduced to [Co(I)(bpy)3]+ either by PS* (PS1* or PS2*) via the oxidative quenching process or by PS−, generated from the reductive quenching of PS* by D [50]. The work of Sutin and coworkers established that the reductive quenching process takes place when ascorbate is used as sacrificial electron donor [45, 46, 50], whereas the oxidative quenching mechanism is at work with TEOA [48, 50]. Generation of cobalt(I) complexes from 2 under irradiation is evidenced by the blue color of the solution, and dissociation of bipyridine ligands from [Co(I)(bpy)3]+ has been observed under catalytically relevant conditions [49]. In fact, the hydride complex [Co(bpy)2(OH2)H]2+, independently prepared by pulse radiolysis of aqueous CoSO4-bipyridine mixtures in the presence of radical scavengers [47], is likely the key catalytic intermediate during hydrogen photogeneration from aqueous solutions [48, 50]. It should also be noted that the major cobalt(II) complexes formed in solution are the mono and bis-bipyridine [Co(bpy)n]2+ (n = 1 and 2) species if the catalytic mixture is prepared by mixing a Co(II) salt with the diimine ligand (L2a or L2b; Co(II):L = 5:1, see Table 2) [45].

Reductive quenching of PS1* by ascorbate is also recognized as the predominant process in most of the photocatalytic systems based on polypyridyl catalysts [57]. PS1* is, for instance, not quenched by catalyst 8a at the concentrations used in the photocatalytic experiments, whereas a reductive quenching constant of 2.6×107 M−1·s−1 has been measured with ascorbate at pH 4 [54]. Oxidative quenching of PS1* by catalyst 4a has been calculated to be thermodynamically disfavored from the potentials of the two redox couples [57]. In contrast, both oxidative and reductive processes are allowed for the photocatalytic system based on 12b, PS1 and ascorbate; bimolecular rate constants of 3.1×108 M−1·s−1 and 1.0×107 M−1·s−1 for 12b and ascorbate, respectively, have been obtained from steady-state quenching experiments [56]. However, despite a difference of one order of magnitude between these two constants, the reductive quenching process is dominant under the employed photocatalytic conditions (0.1 M in ascorbate versus 10-100 μM in 12b). Formation of reduced [Ru(bpy)3]+ has moreover been confirmed by laser flash photolysis experiments, and subsequent electron transfer to 12b occurred with a bimolecular rate constant estimated to be 5.7×109 M−1·s−1 [56]. Such high values, close to the diffusion limit, have also been measured for closely related systems based on aminopyridyl ligands [98, 99]. Fast electron transfer from PS− to the catalyst is required to optimize the photocatalytic performances of a system by limiting decomposition of the reduced photosensitizer PS−.

Nevertheless, these photophysical studies generally only address the first electron transfer process occurring in the catalytic cycle, that is the initial reduction of a Co(II) complex to a Co(I) state (Figure 1). A second photoinduced electron transfer is required to produce hydrogen, either to reduce a Co(III)–H species (ECEC pathway, Figure 1), for a second reduction of the Co(I) intermediate (EECC pathways, Figure 1) or, alternatively, to regenerate the initial Co(II) state of the catalyst (ECCE pathway, Figure 1). Very little is known about the cobalt-centered mechanism taking place under light-driven conditions. The hydrogen evolution rate is, in many examples [59, 85, 100], reported to be first-order with catalyst concentration for experiments performed at relatively low catalytic concentrations, which reflects the fact that the rate determining step is the electron transfer from the photosensitizer to the catalyst, the hydrogen evolution rate being generally proportional to the photon flux. At the same time, such a dependency likely excludes homolytic mechanisms that are strongly disfavored at low catalyst concentration. Such conditions, however, are not relevant from a practical point of view since very low amounts of hydrogen are produced (see Table 2), and substantial production of hydrogen is generally obtained at higher catalyst concentrations (see Table 2 and, for instance, reference [85]). The Co(III)–H based ECCE pathway (Figure 1) requires strong acids to operate and is thus very unlikely to take place under the reported photocatalytic conditions; pathways involving a Co(II)–H species are therefore favored. Thermodynamics indicates that the reductive strength of [Ru(bpy)3]+(−1.26 V vs NHE) is high enough to further reduce the Co(I) state of catalysts such as 8a,b [54], thus allowing the ligand-assisted EECC pathway, characterized in the electrocatalytic studies, to also operate under photocatalytic conditions. Of note are the studies from Long, Chang and coworkers where both electrocatalytic and photocatalytic experiments were undertaken in pH 7 aqueous phosphate buffer for catalysts 11a-c [58, 59] and in pH 4 ascorbate buffer for 4a,b. In both cases, the redox potential of [Ru(bpy)3]+ is more cathodic than the onset potential of the electrocatalytic wave which demonstrates the ability of PS1 to drive hydrogen production under visible light irradiation. In general however, comparison of the electrocatalytic and photocatalytic studies is precluded as they are performed under too dissimilar experimental conditions. It has also been recently highlighted that electrocatalytic and photocatalytic processes may proceed through distinct mechanisms, especially if the former involves two successive electron transfers [101]. In a very recent study published during the course of preparing this review, Hamm and coworkers addressed this mechanistic issue by performing spectroscopic studies combined with in-situ time-resolved hydrogen detection [102]; a three-component system based on PS5, 3a and TEOA was shown to evolve hydrogen via the heterolytic ECEC pathway (Figure 1), an identical mechanism to that previously established for cobaloximes.

4. Conclusion and outlook

The electro- and photocatalytic systems based on hydrogen evolving molecular cobalt(II)–polypyridyl catalysts are distinguished from the previous cobaloxime and [Co(bpy)3]2+ based catalysts by the robustness of the polypyridyl coordination sphere toward either reduction, hydrolysis or ligand exchange under reductive conditions. Thanks to this property, light-driven hydrogen evolution has been successfully achieved under fully aqueous conditions using ascorbate as the sacrificial electron donor. Obviously, working under aqueous acidic conditions is beneficial for the cobalt-centered mechanisms and represents an important development in the field. It should nevertheless be emphasized that combined experimental-theoretical mechanistic studies are still needed to fully understand the catalytic mechanism, with a special focus on the beneficial role played by redox-active centers often present in the ligand structure, such as the bipyridine moiety.

Thanks to the wide variety of polypyridyl ligands evaluated during the last four years, structure-activity relationships were tentatively established in order to shed light on some specific trends for the future design of better performing catalytic systems. First, the relatively high sensitivity of the light-driven hydrogen evolving activity to very subtle variations on the ligand backbone should be noted from these studies. Although catalysts based on tetradentate ligands were reported to be superior to pentadentate ones, no obvious advantage could be established here, as substitution on the ligand backbone also strongly influences the activity, without clear rationale. Only the presence of two cis open coordination sites in the tetradentate series can be recognized as beneficial for the catalytic activity. We nonetheless would like to draw the reader’s attention to the fact that a straightforward comparison of the reported numerical values such as TONs can be misleading in comparing different catalysts. TON values are indeed highly dependent on the photocatalytic experimental conditions that, in most cases, drastically vary from one study to another. High TONCo can be reached at low catalyst concentration but to the detriment of the long term stability and of the quantity of hydrogen produced. As a consequence, quantum yield together with long-term stability should be the two parameters to more systematically consider in the future.

These studies clearly open new perspectives in the field of light-driven hydrogen evolution by molecular cobalt catalysts. Up to now, the prototype ruthenium trisbipyridine was the main photosensitizer used in aqueous media, but new light-harvesting units can now be considered to improve both the activity and the sustainability of these photocatalytic systems. In addition, the phosphine (sacrificial electron donor) ascorbate (electron relay) combination developed by Alberto and coworkers [87] should be systematically evaluated to overcome the limitations imposed by the dehydroascorbate deactivation pathway.