1. Introduction

Natural oxygenic photosynthesis is a singular process that has been fundamentally shaping the image of the biosphere on earth since the advent of the first light-powered life forms. Solar-to-chemical energy conversion by living organisms involves carbohydrate molecules as chemical energy carrier that capacitate them to fuel their own life processes as needed through cellular respiration (Dau et al. 2017). A substantial part of the photosynthesized organic compounds has been fossilized in millions of years that nowadays represents an essential energy source as fossil fuel and feedstock for the modern human society (Balzani et al. 2008). While in many respects, fossil fuels mean safety and comfort, on the other hand the extensive use of these non-renewable energy carriers presents the greatest threat humankind ever faced caused by itself (Pachauri et al. 2015). The extent of greenhouse gas emission, pollutants released into the environment, wasteful thinking about the available resources all root from a wrong perception that has prevailed for decades, if not for a century, for we have thought that nature could withstand the increasing load.

In part because of the extensive use of non-renewable energy carriers, the search for clean and renewable energy resources has become more urgent than ever. However, the availability of renewable energy resources is diffuse and intermittent that demands advanced storage technologies. The straightforward use in fuel-cell technology to generate electricity and the clean combustion properties make hydrogen gas a promising storage material and true alternative to fossil fuels. Moreover, the energy density of hydrogen in a compressed gaseous or liquid form is compatible with a broad range of applications. The global hydrogen production accounts for approximately 7.7 EJ/year, which may rise to 10 EJ/year by 2050, but only 4% of the hydrogen is produced via water electrolysis (Arregi et al. 2018; Hamza et al. 2020). Therefore, increasing the hydrogen production based on renewable technologies is a timely need.

With the above reasons in mind, we note that without feedback or simply, suitable coupling neither natural nor artificial functionalities will create a balanced renewable system, no matter how efficient or profitable individual, functional parts were put together. We believe that this is an important, if not the most important piece of knowledge that can be deciphered from natural photosynthesis for observers, beyond the obvious aim to create artificial systems based on renewable resources, which we can use on our own benefit (Nocera 2012; Faunce et al. 2013; Pantazis 2021; Osman et al. 2021). Or should we rather be the plummet that needs to be taken off the scales to rebalance? The importance of studying artificial photosynthesis lies in the systematic approach that starts with understanding the natural process.

Plants harvest solar energy by photosystem, produce O2 from water molecules, and as a coupled system, convert CO2 into carbohydrates in the course of the Calvin cycle (Krewald et al. 2015; Pantazis 2021). A Z-shaped energy diagram is generally used to illustrate the cascade of electron transfer steps. The energy to drive this reaction cascade is supplied by light absorption of the photosystem II primary electron donor P680, and the photosystem I primary electron donor P700 (Grimes et al. 2008). Indeed, photons are absorbed by the chlorophyll-containing antenna system in the thylakoid membrane as shown in Fig. 1, furnishing photosystem II with very high quantum efficiency. The oxidized primary electron donor of photosystem II gives electrons to an external quinone to form hydroquinone. Electron vacancies are filled by water oxidation producing O2 as byproduct, carried out by the oxygen-evolving complex, which contains an oxo–bridged tetramanganese calcium cluster Mn4O5Ca.

Fig. 1
figure 1

The light-dependent reactions of photosynthesis at the thylakoid membrane. The electron transfer chain is displayed in blue. The molecular components of the system are embedded in the thylakoid membrane. The water oxidation process at the oxygen-evolving complex is triggered by the light absorption of photosystem II primary electron donor. Quinone molecules transport the released electrons to the cytochrome complex and to plastocyanin that functions as electron donor to photosystem I. This electron transfer requires another photoexcitation process by the primary electron donor of photosystem I. The electrons directed to the lumen are used up in nicotinamide adenine dinucleotide phosphate reduction. Abbreviations: PS II is photosystem II, P680 is the photosystem II primary electron donor, OEC is the oxygen-evolving complex of photosystem II, Q is quinone, H2Q is hydroquinone, b6f is the cytochrome complex, Pc is the blue copper protein plastocyanin, PS I is photosystem I, P700 is the photosystem I primary electron donor, Fd is ferredoxin, FNR is the ferredoxin-NADPH reductase, NADP+ is nicotinamide adenine dinucleotide phosphate and NADPH is the reduced form of nicotinamide adenine dinucleotide phosphate (Pantazis 2021)

Several in-depth analyses have been published discussing the structure and operating mechanism of photosystem II and that of the oxygen-evolving complex (McEvoy and Brudvig 2006; Umena et al. 2011; Kärkäs et al. 2014; Shen 2015; Zhang and Sun 2019; Zabret et al. 2021). Herein, we would like to emphasize two details that are worth considering. First, an oxygen-evolving complex has a certain average lifetime of circa 30 min under full sunlight, because light quanta induce degradation, and the molecular integrity eventually breaks down (Inoue et al. 2011). Thus, the oxygen-evolving complex has to be protected but also continuously rebuilt by repair mechanisms strongly coupled to energy conversion. Note that the Mn4O5Ca cluster itself binds to the protein and assembles spontaneously from free Mn2+ and Ca2+ in solution under visible light, a process called photoactivation (Dasgupta et al. 2008).

From the perspective of the many artificial catalytic systems, where the longest possible lifespan is desired, a similar feedback mechanism at the expense of robustness and maximal energy efficiency might seem rather contra-productive. Nevertheless, the question rises, why this natural system has subsisted a very long evolutionary process, but still remains the best practice in natural energy conversion. Perhaps this feature of the natural oxygen-evolving complex inspired the idea of a Co-based self-healing oxygen-evolving catalyst (Costentin and Nocera 2017) that has to do historically with the century-old attempt of determining the potential for the CoIII/II couple (Coehn and Gläser 1902). In any case, the electrochemical behavior of cobalt has certainly not changed in a hundred years, which may give cause for optimism in terms of the flexibility of the human approach.

The second detail, which is the starting point of our focused review, is that photosystem II contains redox-active organic cofactors close to the inorganic cluster. As integral part of the oxygen-evolving complex a redox-active tyrosine-Z residue coupled to a hydrogen-bonded histidine-Z unit is found near the Mn4O5Ca cluster, shuttling proton-coupled electron transfer and synchronizing the light-induced Kok cycle with the distal transport of the charge carriers between the oxidized primary electron donor of photosystem II and the cluster (Kok et al. 1970). Another tyrosine–histidine pair, tyrosine-D and histidine-D is in redox contact with both tyrosine-Z and Mn4O5Ca. Both redox cofactors are present in every known oxygen-evolving organism; therefore, no photosynthesis as we witness today would be possible without these small molecular fragments.

Our review focuses on the catalysis of the water oxidation reaction that lies at the heart of artificial photosynthesis. Irrespective of the design of the whole system, that is photocatalytic, photoelectrocatalytic or photovoltaic plus electrocatalytic, the function of the unit that facilitates the water oxidation reaction remains the same, which is identical with the function of the oxygen-evolving complex in photosystem II. Since the tyrosine residues serve as a natural archetype and inspiration for the utilization of redox cooperating ligands in artificial systems, in the next section we will discuss the role of the tyrosine–histidine pairs in photosystem II (Styring et al. 2012). Then, artificial molecular catalysts for the oxygen evolution reaction will be discussed.

Molecular catalysts may be regarded as workpieces that can be rationally tuned to achieve ultimate efficiency at an atomic level. However, molecular entities are often intrinsically prone to degradation due to the oxidative reactivity of the organic ancillary ligands. At the same time, the oxidative activation of such ligands can contribute to the water oxidation reactivity of a metal–ligand complex. Therefore, we focus on what can be gained of such a cooperation and bring examples on some known scenarios of redox state changes in ancillary ligands. Analogies to the redox-mediators found in photosystem II will also get attention.

2. Effect of tyrosine on the mechanism of the natural oxygen evolution reaction

In nature metal cofactors and perhaps to a lesser extent, yet, of comparable effect redox-active organic moieties are key components to enzymatic catalysis to facilitate biochemical redox reactions. In this context, the evolutionary role of the phenolic sidechain of tyrosine cannot be over-emphasized. The oxygen-evolving complex strongly relies on two tyrosine residues found in its protein backbone, tyrosine-Z and tyrosine-D (Siegbahn and Blomberg 2004; McEvoy and Brudvig 2006; Styring et al. 2012; Narzi et al. 2014; Nakamura et al. 2020). The tyrosine residues are involved in the regulation of the sequential transfer of electrons and protons between the oxidized primary electron donor of photosystem II and the Mn4O5Ca.

The oxidation of H2O molecules may take place in four different ways producing HO·, H2O2, O2·−, or O2, depending on the number of electrons removed. The 4e oxidation pathway yielding triplet O2 as product is the lowest in energy although, still energetically uphill and the most complex reaction of all at the molecular level, since four O–H bonds from two water molecules have to be broken and an O–O bond must be formed. This 4e oxidation occurs at the oxygen-evolving complex of photosystem II.

2.1 Mechanism of the oxygen-producing Kok cycle

The catalytic center has been conceived as progressing in a cycle of five oxidation states, S0 to S4, since Kok has proposed that a trapping center or an associated catalyst successively accumulates oxidizing equivalents to link the 1e progression in photoexcitation to the 4e¯ and 4H+ abstraction from two H2O molecules to yield O2 (Kok et al. 1970) (Fig. 2). In the S1 state the system is stable in the dark, while O2 is produced only after reaching the S4 state, during a single photon-induced S4 → S0 transition. The e to H+ stoichiometry over the four steps corresponds to 1:0:1:2 (Förster and Junge 1985). The photo-induced charge separation in photosystem II generates the strongest oxidant known in biology, the oxidized form of the primary electron donor of photosystem II, for which the estimated redox potential is circa +1.2 V versus the normal hydrogen electrode (Rappaport and Diner 2008). This potential is enough to drive each of the transitions in the Kok cycle and ultimately extract electrons from water molecules. At the same time, interfacial tyrosine regulators are required to adjust the sequential charge accumulation in the Mn4O5Ca cluster.

Fig. 2
figure 2

The Kok cycle illustrates the increasingly oxidized S0-S4 states for the Mn4O5Ca cluster in the oxygen-evolving complex. The system utilizes photon energy (hv) to carry out each 1e oxidation step of the cluster. The primary electron donor of photosystem II undergoes oxidation by excitation at 680 nm. The oxidized form is a sufficiently strong oxidizer to drive the cycle up to the S4 state. Oxygen evolution is triggered by the absorption of the fourth photon until which point the Mn4O5Ca cluster accumulates the electron vacancies. Each step is mediated by the tyrosine-Z/tyrosine-Z·+ couple (Kok et al. 1970; Nocera 2012). Abbreviations: Tyr-Z is tyrosine-Z, Tyr-Z·+ is the oxidized form of tyrosine-Z, P680 is the photosystem II primary electron donor, and P680·+ is the oxidized primary electron donor of photosystem II

Outstanding achievements have led to the presently known detailed models for O2 formation from water by the oxygen-evolving complex, including valence and spin state analysis of the Mn ions in various S states (Cox et al. 2013; Krewald et al. 2016; Pantazis 2018). In conjunction with spectroscopic and kinetic investigations, structural analysis using femtosecond X-ray pulses, with improved spatial and temporal resolution, selective characterization of the different states resulted in the native structure of photosystem II (Suga et al. 2015) and the structural analysis of the intermediates in the Kok cycle (Kern et al. 2018).

A thorough overview on the evolution of structural and kinetic models of photosystem II and oxygen-evolving complex has been provided recently (Junge 2019). We rely on the current works cited therein, including the stepwise S state structural analysis (Krewald et al. 2016; Kern et al. 2018), and further considerations made on tyrosine-Z (Nakamura et al. 2020), in order to explain the role of the tyrosine–histidine pairs at the different stages of the Kok cycle.

2.1.1. Structure of the Mn4O5Ca cluster and the initial S states

The structure of the Mn4O5Ca cluster is often depicted as a distorted chair since the first atomic resolution crystal structure of photosystem II has been reported (Umena et al. 2011) (Fig. 3a). This metal cluster along with the surrounding protein scaffold is very stable, and the expected structural changes during the cycle are subtle (Zhang et al. 2017a); thus, the positions of the primary electron donor of photosystem II, tyrosine-Z, tyrosine-D and the Mn4O5Ca cluster vary very little. In states S0 and S1 the cluster adopts low spin configurations with 1/2 and 0 total spin, respectively. In S0 the O5 bridge (Fig. 3a) is protonated, one MnIII and the Ca2+ ion both have aqua ligands.

Fig. 3
figure 3

a Structural changes of the Mn4O5Ca cluster during the S0-S4 states. The most reduced state is S0 with a single MnIV and three MnIII ions. One MnIII in S0 is oxidized by one electron to provide S1 with a proton dissociating from the O5 oxygen atom. The next oxidation generates another MnIV ion to give a low and a higher spin S2 structural isomer. Only S2B can progress to S3 that is facilitated by tyrosine-Z·+. In this step a water molecule coordinates to the cluster that loses one proton. In the S3 state two oxygen atoms get close to each other and participate in O–O bond formation upon reaching subsequent S4 state (tyrosine-Z·+)MnIV4 upon the final photoexcitation. The cycle is closed by the evolution of O2 and the coordination of H2O. b The redox transition of tyrosine-Z/tyrosine-Z·+ and the proximal histidine-Z residue that acts as neighboring base. Abbreviations: ST is the total number of unpaired electrons that defines the spin configuration, hv represents the incident photons, P680 is the primary electron donor of photosystem II, P680·+ is the oxidized primary electron donor of photosystem II, Tyr-Z and Tyr-D are the respective tyrosine residues, His-Z and His-D are the respective histidine residues

All remaining ligands in the first coordination sphere are carboxylate groups of glutamic acid, aspartic acid and alanine derived from structural polypeptides except for a nitrogen-donor histidine ligand, which coordinates to a Mn ion. The peptide ligands were omitted from Fig. 3a for clarity. The changes taking place in the coordination mode of O5 are of particular importance because spectroscopic results implicated that O5 originates from one of the two reacting substrate water molecules.

The photogenerated oxidized primary electron donor cation is coupled to the Mn4O5Ca cluster via the intermediary redox-active tyrosine-Z, which works as a 1e rectifier (Fig. 2). Initial electron transfer between the oxidized primary electron donor cation and tyrosine-Z occurs on a nanosecond time scale, whereas the subsequent electron transfer between tyrosine-Z·+ and Mn4O5Ca occurs on a micro- to millisecond time scale, depending on the actual S-state. Fast reversible redox transition of tyrosine-Z/tyrosine-Z·+ is an intrinsic property originating from the proximal histidine-Z residue that acts as base, stabilizing the deprotonated tyrosine-Z (Fig. 3b), while structured water molecules bound to the Ca2+ ion also tune the redox potential of this molecular moiety. Tyrosine-Z·+ is a strong oxidizer with a redox potential greater than +0.9–1.0 V versus the normal hydrogen electrode (Vass and Styring 1991). Due to the proximity of the Mn4O5Ca cluster tyrosine-Z·+ is reduced very fast (Babcock et al. 1976), by means of an S-state-dependent kinetics; therefore, a direct detection is complicated.

2.1.2 The S2 state

The most characterized state of the catalytic cycle is undoubtedly S2, in which the cluster is oxidized by 1e compared to S1. The manganese oxidation states in S2 are MnIIIMnIV3, leading to a 1/2 spin configuration that is detectable by electron paramagnetic resonance spectroscopy. Importantly, S2 can adopt a low and a higher spin form exhibiting 1/2 total spin for the open cube-like form S2A and 5/2 total spin for the closed cube form S2B (Fig. 3a). At this point in the cycle a spin state interconversion takes place on a microsecond timescale under physiological conditions with a low energy barrier for the O5 translocation, thus enabling the conversion of S2A to S2B that in turn can progress to S3 (Styring et al. 2012; Pantazis et al. 2012; Bovi et al. 2013). The low spin configuration up to the S2A state of the oxygen-evolving complex has been suggested to prevent short circuiting of the 4e oxidation at a 2e level of peroxide formation that may help to increase the lifetime of the complex (Rutherford 1989).

In order to proceed further from the low spin S2 tyrosine-Z·+ state Ca2+ is also indispensable, facilitating the S2A to S2B interconversion and the stabilization of the S3 state. Deprotonation of the coordinating water molecule is coupled to the MnIII → MnIV oxidation in S2B which occurs after formation of tyrosine-Z·+ (Retegan et al. 2016). The reorientation of the electric dipole of the Mn4O5Ca cofactor supports the coupled charge transport.

2.1.3. The S3 and S4 states

In the S3 state, all Mn ions adopt oxidation state IV. Note that the structure of S3 in Fig. 3 may not represent the final state. This is because the S2 to S3 transition is a multistep process including manganese oxidation, substrate deprotonation, internal rearrangement and water binding. The resulting S3 has been described as (tyrosine-Z)MnIV4 or (tyrosine-Z·+)MnIV3MnIII (Retegan et al. 2014; Cox et al. 2014). In S3 two oxygen sites are close to each other and may couple together in the subsequent S4 state (tyrosine-Z·+)MnIV4 upon the final photoexcitation.

The protonation states of the water-derived ligands have not been established yet for each intermediate and vivid ongoing research aims at resolving the deprotonation sequences (Yang et al. 2021). According to computational studies on the O–O bond formation, an oxyl radical is likely to be formed in the S4 state, most probably a terminal oxyl, which subsequently couples with an oxo bridge to form the O–O bond (Siegbahn 2009). Upon O2 formation the release of the first proton takes place in the time range of 100 μs followed by a slower deprotonation step at 1 ms (Förster et al. 1981). The latter timescale also limits a set of inseparable events from the deprotonation, namely the 1e transfer to tyrosine-Z·+...H+-histidine, the reduction of three manganese ions, the release of O2 and the uptake of an H2O molecule. The exact order of the events is subject to ongoing research.

2.2 Role of tyrosine residues

All photosynthetic events could take place without tyrosine-D, yet this structural component is found in every natural photosynthesizing species. Tyrosine-D has a redox potential of circa +0.7–0.8 V versus the normal hydrogen electrode (Vass and Styring 1991). This residue is very stable as a radical located farther from the cluster than tyrosine-Z and remaining in the oxidized state from the first sunlight into the dark night hours.

In the D2 protein of photosystem II, the oxidized tyrosine-D·+...H+-histidine-D can elevate the potential energy of the oxidized primary electron donor of photosystem II via columbic interaction (Rutherford et al. 2004). Thus, the oxidation of tyrosine-Z is accelerated, and the highly oxidizing charges will be located only in the D1 protein side of photosystem II, suppressing side events. Further, tyrosine-D·+ can oxidize the S0 state to S1 in the dark but can also deactivate the S2 and S3 states by reduction, preventing other unwanted reductants from interfering with the S state cycle (Feyziyev et al. 2003). Finally, tyrosine-D also plays an important role in the repair cycle of photosystem II, since in the absence of electrons from the Mn4O5Ca cluster tyrosine-D can serve as electron–hole carrier in equilibrium with the oxidized primary electron donor of photosystem II and tyrosine-Z·+ (Magnuson et al. 1999). As conclusion, nature provides two examples at a time of redox-active organic functions that can be paralleled with those existing in artificial systems to be discussed herein.

3. Redox-active ligands in artificial water oxidation catalysts

3.1. General considerations on the mechanism of the catalytic oxygen evolution reaction

Synthetic catalysts for the oxygen evolution reaction must deal with the removal of four electrons and four protons from two molecules of water according to Eq. (1).

$$2{\text{H}}_{2} {\text{O}} \to {\text{O}}_{2} + \, 4{\text{H}}^{ + } + \, 4e^{ - }$$

Transition metal complexes have rich redox chemistry due to a variable d electron configuration making the utilization in catalysts for the oxygen evolution reaction appealing. Upon reversible 1e oxidation of a metal-aqua complex of general formula Mn+–OH2 the aqua ligand becomes more acidic and may lose a proton; thus, multiple-site electron–proton transfer can generate the corresponding oxidized form, M(n+1)+–OH (Weinberg et al. 2012). In a next oxidation M(n+2)+=O can be obtained in a similar fashion (Fig. 4). For example, the aqua-Ru-polypyridyl complexes can reach high oxidation states within a narrow potential range in this way, due to the σ- and π-donation character of the M=O group (Sala et al. 2014; Matheu et al. 2019). The possibility of multiple-site electron–proton transfer prevents the buildup of positive charges that would otherwise strongly destabilize the oxidized forms and separate the consecutive oxidation steps by high differences in potentials. Instead, the coupling between electron and proton transfer steps allows for a redox leveling that favors rapid kinetics for the whole oxygen evolution process. An excellent example is the histidine base effect in the tyrosine–histidine pairs of the natural oxygen-evolving complex.

Fig. 4
figure 4

Electron and proton transfer processes during the oxygen evolution reaction catalyzed by transition metal complexes and leading to the potentially reactive species. In the vertical direction, protic equilibria are shown that affect the coordinated aquo, hydroxide and oxo ligands. In the vertical equilibria, the oxidation states do not change. In the horizontal direction, only the oxidation state changes. Diagonal equilibria are multiple-site electron–proton transfer reactions, in which the proton transfer is strongly coupled to the transfer of the electron. This process is often favored energetically. Water oxidation occurs if the M(n+2)+=O moiety is thermodynamically competent to undergo water nucleophilic attack, or reaction with another M(n+2)+=O unit. MS-EPT stands for multiple-site electron–proton transfer

Without histidine acting as acceptor site for the proton, electron transfer is unfavorable and slowed down by a factor of 200 (Diner et al. 1991) and the oxidation of tyrosine cannot outcompete the reverse electron transfer from the quinone radical anion to the oxidized primary donor of photosystem II (Hoganson et al. 1995) (see the electron transfer chain in Fig. 1). Another example taken from photosystem II is the proposed multiple-site electron–proton transfer reaction in the S0 to S1 transition of the Kok cycle (Fig. 3a). The S0 to S1 step involves 1e oxidation of the Mn4O5Ca cluster with electron transfer to tyrosine-Z•+ in concert with two proton transfer events: one taking place from a Mn–OH2 moiety to a neighboring aspartate base and another at the electron acceptor site tyrosine-Z·...+H–histidine-Z providing tyrosine-Z-H...histidine-Z.

In addition to the removal of electrons and protons, the oxidation of water to O2 also requires the formation of an O–O single bond that is generally a purely chemical step in nature and takes place after the second oxidation step of the catalyst. Molecular catalysts for the oxygen evolution reaction follow two main types of mechanism, depending on whether an external H2O molecule participates in the formation of the O–O bond, or two M–O units react. The former mechanism corresponds to water nucleophilic attack abbreviated as WNA, while the interaction of two M–O units is called I2M, as shown in Fig. 5. The chemical O–O bond formation step that creates a metal-peroxide intermediate is generally regarded as the bottleneck of the overall process, whereas the follow-up reaction steps proceed rapidly toward the closing of the catalytic cycle.

Fig. 5
figure 5

The two main types of mechanisms of the O–O bond formation step of the oxygen evolution reaction. The water nucleophilic attack (WNA) involves an external water molecule that reacts with the electrophilic oxo ligand. The interaction of two M–O units (I2M) occurs in cases, when the M(n+2)+=O unit has oxyl radical character and less electrophilic (Sala et al. 2014). The assistance of a base is typical in the case of the water nucleophilic attack mechanism with late transition metals. The interaction of two M–O units is favored by high catalyst concentration, bridging ancillary ligands and relatively electron populated metal d orbitals

A recent review covered the role of redox-active and redox non-innocent ligands in water splitting with a categorization based on the way of activation, that is electrochemical, photochemical or chemical, collecting examples of catalysts for both the hydrogen evolution reaction and the oxygen evolution reaction (Singh and Indra 2020). The same review paid attention to structural mimics and briefly discussed the possible effects of redox-active ligands on a catalytic process in general. Another work categorized molecular transition metal catalysts as mono- and binuclear complexes, containing redox non-innocent ligands, and emphasized their role in supplying electrons or binding protons at different stages of the corresponding mechanism, such as the water nucleophilic attack or the interaction of two M–O units, leading to a more efficient reaction (Sutradhar et al. 2021). Similarly to the previous reviews, we follow the definition of a ligand being non-innocent, if the oxidation states of the central atom cannot be defined unequivocally (Jørgensen 1966). According to this definition, intramolecular electron transfer may occur between the ligand and the metal center in complexes containing non-innocent ligands, allowing for different valence tautomeric forms (Tezgerevska et al. 2014). However, redox-active ligands are those, in which well-defined changes in the redox state also alter the reactivity of the complex in a favorable direction.

In the case of the oxygen evolution reaction, participation of redox-active ancillary ligands in one or more of the oxidation steps is likely, if this change in the electronic structure of the complex makes the concerned steps energetically favorable. Obviously, those ligands, which can accelerate O–O bond forming or reduce the potential barrier in this manner, are expected to pay most attention to. At the same time, catalytic efficiency is occasionally disfavored by intervening ligands. Attention should be paid to unexpected effects, as these may occur in new systems, and account for the specific behavior. Such redox activity also contributes to a more complete picture.

Herein we focus on catalysts for the oxygen evolution reaction, seeking analogies if apply, with the cooperation between the oxygen-evolving complex and the tyrosine–histidine pairs found in photosystem II with the notion that several redox-active ligands have been explored in hydrogen evolution reaction catalysis (Singh and Indra 2020). We bring representative examples on the possible outcome of the redox state change in ancillary ligands during catalysis, keeping in mind that there may be crosstalk between molecular catalysts and oxides, the other main group of synthetic catalysts for the oxygen evolution reaction. The following ligand types will be discussed, irrespective of the metal content, the nuclearity of the complex, or the way of activation:

  1. (i)

    Those accumulating electron vacancy to assist catalysis;

  2. (ii)

    Those possessing protic sites to facilitate proton-coupled electron transfer;

  3. (iii)

    Those participating in the O–O bond formation step and get recovered;

  4. (iv)

    Those undergoing redox transformation, thus making the complex an efficient precursor of the true catalyst.

3.2 Electron vacancy on redox-active ligands to assist catalysis

3.2.1. Robust ruthenium catalysts

Although an early work pointed out that oxygen evolution from water is possible using transition metal complexes as catalysts (Elizarova et al. 1981), the first thoroughly characterized molecular catalysts for water oxidation were Ru-polypyridyl complexes, starting with the blue dimer cis-[(H2O)RuIII(2,2′-bipyridine)2(μ-O)RuIII(2,2′-bipyridine)2(OH2)]4+ (Gersten et al. 1982) (Fig. 6). Ruthenium complexes are robust, inert toward ligand substitution and their higher oxidation states required for water oxidation are indeed accessible (Matheu et al. 2019). Moreover, typical Ru-based catalysts do not require redox-active ligands for efficient operation.

Fig. 6
figure 6

On the top: the blue dimer, cis-[(H2O)RuIII(2,2′-bipyridine)2(μ-O)RuIII(2,2′-bipyridine)2(OH2)]4+ and three highlighted states in the oxygen evolution catalysis. The Ru2V form that carries four electron vacancies relative to the Ru2III resting state is thermodynamically capable of water nucleophilic attack, instead of the coupling of the two oxo ligands. On the bottom: the dinuclear [RuII2(BTPYAN)(Q)2(OH)2]2+ complex along with the proposed catalytic pathways. In this case, electron transfer from the redox-active quinone ligands facilitates the intramolecular formation of the O–O bond. Abbreviations: bpy is 2,2′-bipyridine, BTPYAN is 1,8-bis(2,2′-terpyridyl)anthracene, Q is 3,6-di-tert-butyl-1,2-benzoquinone and SQ is 3,6-di-tert-butyl-1,2-benzosemiquinonate

The water oxidation mechanism by the blue dimer has been subject to a fruitful debate (Liu et al. 2008; Moonshiram et al. 2012) incorporating electrochemical and chemical activation using CeIV in highly acidic media, and photochemical studies near pH 7 using photosensitizing ruthenium complexes in the S2O82− and blue dimer molecular system. In the latter case the sequential addition of two H2O molecules has been suggested to the bipyridine ligands that may couple to generate O2 (Yamada and Hurst 2000; Cape and Hurst 2008). However, the pathway following the 2e oxidation of the OH adducts results in decomposition of bipyridine (Liu et al. 2008). The decomposition ultimately gives CO2 as a product after several redox cycles, since the partly oxidized rings are kinetically susceptible to further oxidation.

Upon electrochemical activation of Ru catalysts, generally the Ru centers and the water-derived ligands eject electrons rather than the ancillary ligands. The blue dimer is oxidized by a 1e plus 3e sequence. The produced [(O)RuV(2,2′-bipyridine)2(μ-O)RuV(2,2′-bipyridine)2(O)]4+ complex is thermodynamically competent to form a Ru–O2H intermediate upon water nucleophilic attack involving one RuV=O unit (Fig. 6). Insight into this mechanism allowed for a strategy for utilizing different N,N-bidentate ligands in ruthenium complexes as kinetically facile, external electron-transfer mediators demonstrating that such mediators can enhance the performance of catalytic water oxidation (Concepcion et al. 2008). Undoubtedly, these achievements inspired several new complexes contributing to the development of mechanistic insights into the molecular catalysis of the oxygen evolution reaction (Kamdar and Grotjahn 2019). Although ruthenium does not necessitate redox-active ligands for catalytic function, quinone ligands were found capable of assisting in Ru-based catalysis (Wada et al. 2001; Kobayashi et al. 2003; Tanaka et al. 2012) (Table 1).

Table 1 Ruthenium-containing molecular catalysts of the water oxidation reaction that contain redox-active ligand with or without protic sites, reaction conditions and catalytic performance, n.d. means not determined

A prominent example is the dinuclear [RuII2(BTPYAN)(Q)2(OH)2]2+ complex containing two redox-active quinone ligands (Fig. 6, where Q stands for 3,6-di-tert-butyl-1,2-benzoquinone and BTPYAN is 1,8-bis(2,2′-terpyridyl)anthracene). The immobilized complex on indium tin oxide electrode was capable of electrocatalytic oxygen evolution at + 1.70 V versus the normal hydrogen electrode in water at pH 4, producing O2 by 95% charge efficiency and reaching a turnover number of 6730 (Wada et al. 2000) (Table 1). Upon the O–O coupling the quinones act as oxidants via intramolecular electron transfer. In this reaction step two protons dissociate from the complex and (SQ)RuII(O–O)RuII(SQ) is formed, where SQ is 3,6-di-tert-butyl-1,2-benzosemiquinonate. Two features were found indispensable to a successful catalysis: the two hydroxide ligands located close to each other due to the structural constraint by the two terpyridine moieties, and the two quinone ligands helping intramolecular coupling of the OH ligands to form the O–O bond.

The semiquinone ligands undergo oxidation to quinone at + 0.4 V versus the normal hydrogen electrode, furnishing (Q)RuII(O–O)RuII(Q)2+ that is followed by metal oxidation to RuIII at + 1.2 V versus the normal hydrogen electrode. Valence tautomerism of the produced (Q)RuIII(O–O)RuIII(Q)4+ complex generates (Q)RuII(O=O)RuII(Q)4+ from which O2 liberates upon replacement by two H2O molecules. Note that a theoretical study identified [Ru2(O2)(Q−1.5)2(BTPYAN)] as a key intermediate and the most reduced catalyst species that is formed by removal of all four protons before 4e oxidation takes place (Muckerman et al. 2008). The studies on Ru-quinone catalysts signified that the O–O bond formation is attainable at a low oxidation state of the metal and this observation has very important consequences on first row transition metal catalysts.

3.2.2. First row transition metal catalysts

Ruthenium centers can undergo multiple oxidation steps in a relatively narrow potential range due to effective proton-coupled electron transfers. The situation can be very different when water oxidation involves first row transition metals that often exhibit high propensity toward ligand substitution and the access to higher oxidation states demands high potentials. The rapid ligand substitution is an intrinsic property that stands as major challenge in the design of catalysts and require careful selection of catalytic conditions. Moreover, the release of four protons in one catalytic cycle may easily cause ligand protonation and subsequent dissociation from the metal center as a competing process.

Stabilization of a catalytically active, highly oxidized first row transition metal center demands strong electron-donating ligands. However, this requirement can be circumvented if redox-active ligands are utilized and reversibly oxidized to cooperate with the metal center in storing electron vacancies needed for the water oxidation reaction. Ligand redox activity has been documented in complexes bearing rigid, planar macrocyclic or semi-macrocyclic ligands with extended π-conjugation capable of carrying unpaired electrons. Different families of tetra-amidate ligands were used in Fe (Ellis et al. 2010), Co (Du et al. 2018), Ni (Lee et al. 2020) and Cu (Garrido-Barros et al. 2015, 2017, 2020) complexes. The main experimental parameters and the catalytic performance of the selected complexes discussed herein are listed in Table 2, where the different compounds are listed according to the metal content.

Table 2 Molecular catalysts of the water oxidation reaction that contain first row transition metals and redox-active ancillary ligands, reported reaction conditions and catalytic performance

The complexes with tetra-amido macrocyclic ligands have been studied in detail (Fig. 7). The first report concerned Fe-tetra-amido macrocyclic complexes that could promote O2-evolution to a highly variable extent, by mixing with excess CeIV in water at pH 0.7 (Ellis et al. 2010). Loss of activity was associated with oxidative and hydrolytic inactivation pathways, also confirmed by control reactions using other Fe-ligand combinations. The rate of the biphasic O2 evolution correlated with the addition of electron-withdrawing substituents to the ligand.

Fig. 7
figure 7

Exemplary first row transition metal complexes with redox-active ligands. Metal complexes formed with tetra-amido macrocyclic ligands (TAML complexes) possess four amidic donor groups capable of stabilizing highly oxidized metal centers. The ligand redox-activity stems from the delocalization of the spin over the aromatic moiety. Metal complexes formed with N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) ligands are PBOA complexes. The oxalamide groups, each carrying two negative charges, can undergo oxidation. The oxidized molecule is stabilized by extended π-delocalization. N-Methylation eliminates NH protons that could potentially lead to ligand degradation. Metal complexes formed with tetra-amido macrocycles that are closed by two 1,2-phenylene moieties are called dibenzo-TAML complexes. The ring-closing alkylene group can adjust the size of the macrocycle, and the ligand can be oxidized by two electrons

The highest measured turnover frequency was 1.3 s−1 for the ligand with R1=R2=Cl and R3=F that was sufficiently large to be limited by the kinetics of bubble formation and release (Table 2). Based on theoretical calculations, the formation of a key intermediate, (tetra-amidate macrocyclic radical)FeV=O was suggested (Ertem et al. 2012; Liao et al. 2014). In the latter study, this species was proposed to undergo water nucleophilic attack, or nucleophilic attack by a nitrate ion, the water attack being more favored. According to the results, nitrate may function as a co-catalyst by first donating an oxygen atom to the oxo group to form O2 and a nitrite ion, which can then be re-oxidized to regenerate a nitrate ion. Competing pathways were suggested to modify the ligand such as water and nitrate attack as well as ligand amide oxidation, leading to the opening of the benzene ring and fast catalyst degradation.

Efficient electrocatalytic oxygen evolution in neutral aqueous solution was reported by stable CoIII-tetra-amidate macrocyclic complexes (Fig. 5) that were confirmed as molecular catalysts under the working conditions (Du et al. 2018). The catalytic cycle was examined in detail by electrochemical methods and density functional theory calculations. As the working potential was increased, the triplet complex was first oxidized via ligand-centered 2e + H+ transfer in the presence of water. The oxidation of the ligand was clearly indicated by the changes in the C−C bond length of the benzene ring and the N-aryl ring bond lengths, corresponding to the structural change expected in the oxidation of an o-phenylenedicarboxamido ligand to benzoquinonedicarboxamido form.

The CoIII−OH intermediate was further oxidized to the CoIV=O form, which could react with water to form an O−O bond in a buffer-assisted water nucleophilic attack. The influence of H2PO4 on the process revealed by theoretical calculation was in good agreement with the observed buffer effect. Analysis of the calculated structure revealed that both the oxo moiety and the cobalt center bear significant spin densities corresponding to a more correct CoIII−O assignment. The water nucleophilic attack afforded the quartet hydroperoxo complex, CoII−OOH that was oxidized to the superoxo complex CoIII(O2·−) through proton-coupled electron transfer, to finally release an oxygen molecule and regenerate the resting state (Fig. 8). The turnover frequency values calculated for the different complexes were very similar, between 7.53 and 8.81 s−1, but no catalytic activity could be observed for a complex with a non-redox-active ligand homolog (the benzene ring was substituted by an aliphatic bridge), indicating that the redox-active ligand played a critical role in this multi-electron catalytic cycle. This work underlined the interplay of ligand- and metal-centered redox activity could be a benefit for water oxidation catalysts.

Fig. 8
figure 8

On the top: proposed catalytic cycle for water oxidation by the NiII-tetra-amidate macrocyclic complex, [NiII(TAML4−)]2− as shown in Fig. 7 (Lee et al. 2020). The key to the catalysis is the one-electron oxidation of the metal followed by a two-electron oxidation of the ligand. The third oxidation step generates the highest oxidation state of the complex that can react with a water molecule. On the bottom: the proposed catalytic mechanism of the CoIII-tetra-amidate macrocyclic complexes. The structure of [CoIII(TAML4−)] is shown in Fig. 7. In the first stage the complexes suffer a ligand-centered two-electron oxidation and deprotonation of the aquo ligand. The highest oxidation state of the complex is reached in the next proton-coupled electron transfer step. This form reacts in a rate-limiting water nucleophilic attack. TAML: tetra-amido macrocyclic ligand

A NiII-tetra-amidate macrocyclic complex was characterized as electrocatalyst of the oxygen evolution reaction at neutral pH in phosphate buffer (Lee et al. 2020) (Fig. 7, R3 = CH3). The HPO42− anion served as proton acceptor to accelerate the formation of the O–O bond following atom-proton transfer pathway. The electrochemical activation of the complex started with a reversible, pH-independent NiII to NiIII oxidation at + 0.68 V versus normal hydrogen electrode, followed by two irreversible oxidation peaks at +1.03 V and +1.51 V versus the normal hydrogen electrode. The first irreversible and pH-dependent oxidation was associated with a 2e + H+ transfer to produce an intermediate NiIII–OH species, while the second one was assigned to produce NiIV=O. Oxygen evolution took place at 93% Faraday efficiency and the calculated turnover frequency was 0.32 s−1. The proposed catalytic cycle is shown in Fig. 8. The first-order dependence of the catalytic current on catalyst concentration clearly indicated a single-site mechanism.

The above catalysts could possibly exploit the one- and two-electron oxidation of the applied tetra-amido macrocycle ligand according to Fig. 9 thanks to the available formal o-phenylene-, benzosemiquinone- and benzoquinone-dicarboxamido redox forms. A similar amidate group has been suggested to facilitate the redox activation of a mononuclear CoIII-complex, [Co(2-[bis(pyridin-2-ylmethyl)]amino-N-quinolin-8-yl-acetamidate)(Cl)]Cl (Biswas et al. 2020). Efficient molecular electrocatalysis was reported to occur at a 500 mV overpotential and pH of 8.0. The redox-active amidate ligand facilitated catalysis through the formation of a reactive oxo-metal species (Costentin et al. 2012a, b) (Fig. 10). According to the mechanism proposal, water nucleophilic attack on this putative intermediate formed the O–O bond through a base-assisted proton transfer reaction.

Fig. 9
figure 9

Possible redox states of a phenylenediamide moiety. This structural feature is found in all tetra-amidate macrocycles and (1,2-phenylene)bis(oxalamide) ligands. When the closed-shell, dianionic form loses one electron a radical anion is formed. This radical is stabilized due to spin delocalization over the benzene ring and the two amidate functions. Upon losing a second electron, the molecule becomes closed shell again by adopting a neutral, benzoquinone-dicarboxamido structure. The two oxidation steps are typically separated by ~ 0.5 V

Fig. 10
figure 10

Redox activation of [Co(2-[bis(pyridin-2-ylmethyl)]amino-N-quinolin-8-yl-acetamidate)(Cl)]Cl. The catalytic cycle for water oxidation is initiated by a chloride-to-aquo ligand exchange. Three one-electron oxidation steps at gradually increasing potentials generate the reactive oxo-metal species, highlighted in the red frame. In this mechanism, the oxidation of the redox-active amidate arm is the last step before water nucleophilic attack could take place. After this chemical step, at the applied potential of electrocatalysis the hydroperoxide and the side-on peroxide intermediates are rapidly oxidized to provide the O2 product

In the case of copper, planar 4 N donor coordination sites like in tetra-amido macrocycles are well suited for stabilizing an oxidized CuIII metal center and perhaps this metal has been studied the most thoroughly with respect to the utilization of redox-active ligands. The N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) ligand family shown in Fig. 7 as PBOA demonstrated opportunities in using redox-active ligands to advance Cu-based water oxidation (Garrido-Barros et al. 2015). Introducing electron-donating R1 and R2 substituents at the benzene ring of PBOA allowed reducing the overpotential in electrocatalytic water oxidation by 530 mV.

The energy of the highest occupied molecular orbital in the ligand correlated with the observed overpotential, indicating that the electron-donating substituents increase the overall energy of the highest occupied molecular orbital and favor the oxidation of the ligand. Density functional theory analysis revealed a new mechanism progressing toward the rate-limiting O−O bond formation in single-electron transformations and generating a peroxide intermediate with no formal M−O bond (Funes-Ardoiz et al. 2017) (Fig. 11). The theoretical and experimental results led to a new, general mechanistic proposal, the so-called single-electron transfer water nucleophilic attack.

Fig. 11
figure 11

Proposed mechanism of the electrocatalytic water oxidation by the Cu complex formed with N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) abbreviated as PBOA for clarity. The structure is shown in Fig. 7. The resting state on the top of the cycle undergoes a pH-independent one-electron oxidation and CuIII is stabilized by the strong donor 4 N equatorial ligand environment. In the next step, the PBOA ligand is oxidized that requires higher potential, and hydroxide coordinates to the metal center. This step is followed by a rate-limiting chemical reaction that is a single-electron transfer water nucleophilic attack to result in the reduction of the PBOA ligand and an H2O2 species with a formal − 1.5 oxidation state of each oxygen atom. The further oxidation steps are fast at the applied potential. The homolog ligand 4-pyrenyl-N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) is abbreviated as py-PBOA. The structure of the (py-PBOA4−)CuII complex is found in Fig. 7 with the pyrenyl group at the R2 position. The differences in the oxidation potentials and the structure of the key intermediate are highlighted in red

A derived complex utilizing the ligand 4-pyrenyl-N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) (py-PBOA) was designed to extend the π-conjugation (Garrido-Barros et al. 2017). Both catalysts have been studied in the homogeneous phase and immobilized by π-stacking on graphene-based electrodes. In the homogeneous phase the electronic perturbation provided by the pyrene substituent reduced the overpotential by 150 mV and increased the catalytic rate by more than 20 times to achieve a turnover frequency of 128 s−1. Spectroscopic investigations on the one-electron oxidized form in aqueous solution confirmed that the oxidation process concerns the pyrene moiety in contrast to the parent complex, in which the oxidation remained metal-centered. (The differences are highlighted in red in Fig. 11.) The better performance of the pyrenyl derivative was associated with the lower oxidation potential of the ligand due to the pyrene group together with a large stabilization of the putatively charged oxidized species in the aqueous environment.

Immobilization on a graphene surface provided additional delocalization that improved the catalytic performance of both catalysts; however, the py-PBOA complex was more active. The overpotential of 538 mV, turnover frequency of 540 s−1 and turnover number over 5300 demonstrated that rational design and a redox-active ligand can dramatically increase the stability of first row transition metal catalysts.

The same group underlined in a later work that elucidating the actual sites affected by consecutive oxidations and subtle changes perturbing the delocalization of electron density are of high importance. They developed two new ligands with 13- and 14-membered rings that can be regarded as dibenzo-tetra-amido macrocycles, or shortly dibenzo-TAML as shown in Fig. 7 (Garrido-Barros et al. 2020) in addition to the above-discussed semi-macrocyclic N1,N1-(1,2-phenylene)bis(N2-methyloxalamide). They found that a mere one-atom change in the ring size and a change in solvent polarity were enough to shift the delocalization of electron density from the metal over the ligand π-system.

Valence tautomerism exerted by the solvent interaction evidenced the energetic near degeneration in the frontier orbitals and the subsequent easy access to the different oxidation states, corresponding to metal-centered, d-orbital oxidation or ligand centered oxidation. For example, the coordinated 13-membered dibenzo-tetra-amido macrocycle was involved in two 1e oxidations due to the high delocalization of spins (Fig. 12) resulting in two energetically levelled oxidation steps within an 80 mV potential window. In contrast, this metal–ligand redox cooperativity was missing from the complex containing the 14-membered dibenzo-tetra-amido macrocycle, where the π-conjugation between both phenylene moieties was broken due to the presence of two dimethylmalonamide bridges with sp3 carbons and the saddle distortion of the ligand (Fig. 7, R=CH(CH3)2).

Fig. 12
figure 12

The oxidation sequence of the CuII complex containing the 13-member dibenzo-tetra-amido macrocyclic ligand during electrocatalytic water oxidation and the related water oxidation mechanism at two different pH values (Garrido-Barros et al. 2020). Schematic representation of the electronic states correspond to each oxidation state of the complex with the calculated redox potential and the energy difference between the different multiplicities according to the authors. The molecular moieties concerned by the gradual 1, 2 and 3e oxidations are color-coded. The key to the efficient and robust water oxidation catalysis is the extensive spin-delocalization at the ligand that enables redox metal–ligand cooperativity and energy leveling of the oxidation steps. If the π-conjugation between the phenylene moieties is broken by inserted sp3 carbons, the complex decomposes under catalytic conditions

The addition of only one member group to the macrocycle structure fundamentally determined the fate of the catalysts upon electrolytic water oxidation, since the stability of the spin multiplicities at the different oxidation states relied strongly on the extent of π-delocalization. The complex bearing the larger macrocycle ligand decomposed during catalysis, whereas the one containing the 13-membered macrocycle (Fig. 12) remained stable and highly active thanks to the evidenced redox cooperativity between the metal and the ligand. The highly oxidized reactive form thus becomes more stable due to the delocalization of the accumulated charges. The proposed mechanism and charge delocalization in the different oxidation states are shown in Fig. 12.

Importantly, this way the high-energy oxidized states that are located only on either the ligand or the metal center have been avoided. This strategy neglects both the ligand oxidative degradation and high-valent metal centers that exist only at large potentials and frequently result in the formation of metal oxides. In addition, the two ligand-based oxidations in a narrow potential window were essential to enable water oxidation at pH 7. As a result, catalysis proceeded with a turnover frequency of 140 s−1 at an overpotential of only 200 mV.

A comparison was also made in this context with a catalyst featuring electronic delocalization limited to a single phenylene moiety, formed with the N1,N1-(1,2-phenylene)bis(N2-methyloxalamide) ligand (Fig. 7, R1 = R2 = H) (Garrido-Barros et al. 2015). In this case, the catalytic cycle at pH 12 proceeded via metal oxidation followed by ligand oxidation (Fig. 12). However, the acyclic ligand was prone to acidic demetallation below pH 10 that prevented application in water oxidation at neutral pH, in sharp contrast to the macrocyclic complex. This observation underlined the advantage of the cooperative metal–ligand design approach to find more stable and efficient molecular catalysts for redox reactions. The authors also emphasized that irreversible pre-catalytic electrochemical oxidation features may hinder substantial changes in the pre-catalysts upon oxidation and may indicate degradation.

Following simple logic, we can conclude that shutting down the redox activity of the metal center in parallel with triggering that of the ligand is possible by appropriate ligand design. Indeed, recently a new family of dianionic [2,2′-bipyridine]-6,6′-dicarboxamide ligands (Fig. 13), substituted with redox-active phenyl or naphthyl moieties have been described (Gil‐Sepulcre et al. 2021). The electrocatalytic oxygen evolution process involving only ligand-based redox events was a consequence of two features of the ligand. First, a CuIII/II redox couple at relatively low potentials necessitates highly anionic, strongly σ- and π-donating ligand environment. In the case of the bipyridyl-dicarboxamide ligands, the two neutral N donors of the bipyridine and the two anionic N donors of the amide groups had insufficient electron-donating capacity to stabilize the oxidized metal center below or at the onset potential of catalysis. Second, the two redox-responsive phenyl or naphthyl groups attached to the ligand and conjugated with the amide donors were able to accommodate two distal electron vacancies. Upon oxidation, one of the aryl-amide groups could be detached from the metal, opening a quasi-equatorial coordination site for an OH ligand.

Fig. 13
figure 13

Catalytic cycle for water oxidation including two intramolecular single-electron transfer steps that appear as ISET in the figure. This mechanism has been suggested for CuII complexes containing [2,2′-bipyridine]-6,6′-dicarboxamide ligands, shown here by the example of the 4,4′-(([2,2′-bipyridine]-6,6′-dicarbonyl)bis(azanediyl))dibenzenesulfonate) ligand (Gil‐Sepulcre et al. 2021). Upon consecutive one-electron oxidations, distal aromatic substituents shown in red lose electrons. The metal center oxidation is disfavored in the applied potential window due to the donor environment. The first oxidation of the ligand induces the dissociation of an aryl-amidate donor that opens a quasi-equatorial site for hydroxide coordination. This orientation is necessary to promote the O–O bond formation that occurs upon the follow-up ISET steps. The presence of p-sulfonate groups is crucial to water solubility

A detailed electrochemical analysis at pH 11.6 revealed a large electrocatalytic current associated with O2-evolution at an overpotential of 830 mV for the complex shown in Fig. 13. The experimental and computational results supported a catalytic cycle progressing toward the active form through single-electron transfer steps, including intramolecular single-electron transfer steps (Fig. 13, shown as ISET steps). Upon the first oxidation, a triplet CuII complex is formed with a radical cation mainly centered on one of the aryl-amidate moieties. Evidence for this possibly rate-limiting ligand-based electron transfer was provided by the correlation between the substituent effect on the aryl groups and the observed onset potential of catalysis.

Accordingly, the turnover frequency values were in the range of 5–35 s−1 and the Faraday efficiencies were between 40 and 76%, both related to the structure of the ligand and the stability of the oxidized moiety. The CuII center served as a scaffold for the two bipyridyl-amide groups bonded to the metal center. At the same time OH coordination and activation could take place.

According to the proposed mechanism, the hydroxyl group is attacked by an external hydroxide via a radical-nucleophilic pathway, generating a Cu-(HOOH) species. Since this process involves an intramolecular single-electron transfer event between the aryl radical cation and the hydroxyl group, the metal center is responsible for placing the two groups sufficiently close so that fast electron transfer can occur (Gil‐Sepulcre et al. 2021). The role for the aryl substituent was clearly discussed from a thermodynamic perspective, because the substitution tuned the overpotential for oxygen evolution reaction. The kinetic perspective was apparent in the low reorganization energy of extended polyacenes upon electron transfer ensuring fast kinetic processes. Considering the high versatility in ligand design, the low efficiency obtained for the above complexes that is due to competitive deactivation pathways can be solved in the future.

Due to cyclic π-delocalization porphyrins and corroles are capable of stabilizing electron vacancies (Han et al. 2014). In addition, the two or three negative charges at the N donor groups, respectively, can stabilize the coordinated metal center in higher oxidation states (Nakazono et al. 2013, 2015). Thus, porphyrins and corroles represent an exciting redox non-innocent and potentially redox-active ligand platform with rich possibilities in derivatization as indicated in Fig. 14. The advantages of porhyrins and corroles have been discussed earlier (Zhang et al. 2017b). In context with the present overview, we emphasize that metalloporphyrin catalysts for the oxygen evolution reaction are generally proposed to undergo first a 1e oxidation concerning the ligand that is followed by 1e oxidation of the metal center.

Fig. 14
figure 14

Macrocyclic ligands and a semi-macrocyclic analog to phthalocyanines. All these rigid and robust ligands are capable of planar coordination to a broad scope of transition metal ions, depending on the ionic radius. Porphyrins, corroles and phthalocyanines can stabilize oxidized metal centers due to a variable number of negatively charged donors. The shown ligands are also potentially redox-active platforms available in many substituted forms. The typical positions for introducing electron donating, or withdrawing groups to perturb the electronic structure, or the solubility are indicated by red circles. The substrate water molecule coordination is confined to the axial position. In contrast, bis(arylimino)isoindolines exhibit a more variable substrate coordination, depending on the aryl moieties

Metallocorroles can behave in a similar fashion upon oxidative activation (Zhang et al. 2017b). In metallocorroles the stabilization of high-valent metal ions by corroles is viable through a combination of short metal-nitrogen bonds and large metal out-of-plane displacements (Gross 2001). A key difference from metalloporphyrins is the wider prevalence of non-innocent electronic structures and full-fledged corrole•2– radicals among metallocorroles (Ghosh 2017). There is gathering evidence that the non-innocent or redox-active behavior of this ligand family is prevalent (Dogutan et al. 2011; Lemon et al. 2016; Schöfberger et al. 2016; Sinha et al. 2018, 2020; Garai et al. 2018; Mondal et al. 2020). Moreover, the high adsorptivity of the planar molecules to various surfaces (Schöfberger et al. 2016) makes their utilization in (photo)electrodes directly available (Zhang et al. 2017b). Generation of hydrogen peroxide from water by using metalloporphyrins with redox silent metal centers like Al, or Sn has been also addressed as viable alternative strategy for artificial photosynthesis systems (Kuttassery et al. 2018) as well as light-driven oxidation of halide anions by metallocorroles (Mahammed and Gross 2015). Importantly, corrole and porphyrin ligand architectures are suitable for surface immobilization through covalent linkers or axial coordination of the metal center by tether ligands grafted onto the surface.

Phthalocyanines are expected to exhibit advantages like those of porphyrins and corroles in the catalysis of the oxygen evolution reaction (Fig. 14). For example, photocatalytic water oxidation in the presence of a water soluble CuII-phthalocyanine complex was proposed to follow the interaction of two metal–oxygen units mechanism (Fig. 5) in borate buffer at pH 9.5 (Terao et al. 2016). According to computations, upon 1e oxidation of the complex a phthalocyanine radical is generated instead of an oxidized metal center, thus the CuIII oxidation state generally leading to complex photodegradation does not occur. Importantly, competitive ligand exchange between chloride ions and water molecules for the fifth coordination site inhibits the reaction, calling attention to the potential detrimental effect of impurities.

Copper-phthalocyanine was reported to act as efficient redox mediator in TiO2 nanorod-based photoelectrochemical water splitting (Li et al. 2019). Moreover, conjugation of Cu-phthalocyanine to carbon quantum dots by π–stacking and then coupling with BiVO4 to construct a hybrid water oxidation photocatalyst also seems to be a viable strategy (Xu et al. 2021). A somewhat similar system with Co-phthalocyanine deployed on BiVO4 was also very recently reported to enhance the photocurrent in photoelectrochemical oxygen evolution (Shen et al. 2021). In conclusion, recent findings on the possible application of metallo-phthalocyanines are encouraging.

Polymeric copper and cobalt phthalocyanines were prepared as thin films on fluorine-doped tin oxide from evaporated metal films via chemical vapor deposition process with 1,2,4,5-tetracyanobenzene. The system was successfully operated in 0.1 M KOH electrolyte to generate O2 electrocatalytically (Geis et al. 2016). The phthalocyanines and other types of polymers, like metallated azo-naphthalene diimide-based redox-active porous organic polymer frameworks (Bhat et al. 2018), are encouraging to envision future applications of such materials. On the other hand, the example of a MnII-phthalocyanine complex results in manganese oxide formation upon electrocatalytic water oxidation in acetonitrile–water mixture (Mousazade et al. 2019). The reported decomposition shows that organic ligand platforms may not be suitable under harsh conditions and may undergo hydrolytic degradation.

3.3. Ligands possessing protic sites to facilitate proton-coupled electron transfer

Ancillary ligands with sites occupied by labile protons can facilitate changes in both the number of protons and the electrons during water oxidation catalysis. This feature allows uncommon mechanistic pathways that resembles to the function of tyrosine residues in photosystem II. Some emblematic examples for different ligand design strategies and some recent findings are discussed below.

The 3,5-bis(4-carboxy-1H-benzimidazol-2-yl)-1H-pyrazole ligand framework was reported to form a mixed valence RuIIRuIII complex with 4-picolines occupying the remaining sites (Laine et al. 2015b, a) (Fig. 15). This complex could carry out water oxidation catalysis by both chemical and photochemical activation and O2 evolution was observed in high turnover number of circa 800–890 (Table 1). The strongly electron-donating 3,5-bis(4-carboxy-1H-benzimidazol-2-yl)-1H-pyrazole ligand not only decreased the redox potentials and enabled the higher metal oxidation states, but also allowed proton transfer and exhibited a non-innocent character. The proposed mechanism based on the calculated lowest energy isomers is shown in Fig. 15. A similar ligand design was also successful in the case of mononuclear Ru complexes (Kärkäs et al. 2016) (see Table 1 for data). Importantly, the mononuclear complexes exhibited catalytic capabilities similar to that of the diruthenium complex.

Fig. 15
figure 15

Structure and proposed water oxidation mechanism for the diruthenium complex formed with 3,5-bis(4-carboxy-1H-benzimidazol-2-yl)-1H-pyrazole and picoline that are abbreviated as H5bim2pz and pic herein, respectively (Laine et al. 2015b, a). This complex could be activated by chemical or photochemical methods. The H5bim2pz ligand enabled the higher metal oxidation states. The proton dissociation from the benzimidazolyl group makes this ligand non-innocent, since the deprotonated form lowers the oxidation potential of the RuIII centers. The nature of the O–O bond formation step involving the Ru2IV,V oxidation state remained unclear. Single-site ruthenium complexes were also designed with different ligand substituents (Kärkäs et al. 2016; Abdel-Magied et al. 2016). Those complexes also followed a similar mechanistic pathway, in which the ligand redox non-innocence was a crucial feature

A relatively simple ligand, 6,6′-dihydroxy-2,2′-bipyridine with pendant hydroxyls to mimic the function of tyrosine-Z in photosystem II was applied in a Cu-based electrocatalyst (Zhang et al. 2014). As an effect of the hydroxyl groups, the oxidation potential of the Cu center was reduced by circa 200 mV according to a direct comparison made by the same authors, but the turnover frequency of 0.4 s−1 was well below that of circa 100 s−1 for the parent 2,2′-bipyridine complex at pH ~ 12.5 (Barnett et al. 2012) (Table 3). According to density functional theory calculations, two oxidations preceded the water nucleophilic attack step (Fig. 16). The second oxidation concerned the redox-active ligand to yield a radical anion at a calculated potential of + 1.4 V versus normal hydrogen electrode. Spin delocalization and stabilization of the corresponding oxidized state were claimed to lower the onset potential but also to reduce the electrophilic character disfavoring water nucleophilic attack.

Table 3 Performance and reaction conditions of molecular electrocatalysts for the water oxidation reaction that contain redox-active ancillary ligands with protic sites to facilitate proton-coupled electron transfer step in the catalytic process
Fig. 16
figure 16

Suggested water oxidation mechanism for the CuII-complex containing the redox-active 6,6′-dihydroxy-2,2′-bipyridine ligand (Zhang et al. 2014). The 6,6′-dihydroxyl functions participate in a hydrogen-bonding network involving the two coordinated water molecules as seen on the top of the cycle. A proton-coupled electron transfer facilitates the oxidation to CuIII. The next oxidation step is pH-independent and concerns the ligand. The ligand radical anion is stabilized by π-delocalization. However, the same stabilization reduces the catalytic rate by lowering the electrophilic character of the coordinated OH ligands. The presence of the hydroxyl groups also induces in situ polymerization of the complex on the electrode surface. Degradation was reported pointing out limited stability dependent on catalytic conditions (Gerlach et al. 2014)

Importantly, a partially oxidized complex polymer was found on the electrode surface upon electrolysis that could be re-dissolved in pure electrolyte. Crystallographic studies suggested that the film was built up by a one-dimensional coordination polymer with μ–OH bridges and an H-bonding network. Importantly, in cyclic voltammograms the catalytic peak current showed a curving-over behavior at high complex concentrations that has been associated with the film formation due to the local pH drop. Although such electrochemical features are often associated with catalyst decomposition and concomitant metal oxide film formation, coordination polymers may also generate catalytic films that operate at lower overpotential than the parent complex (Cui et al. 2016; Mishra et al. 2017; Kuwamura et al. 2018). These three studies highlighted that the pH-dependent behavior and redox transformations of molecular catalysts may change their surface affinity thus promote immobilization.

Finally, 6,6′-dihydroxy-2,2′-bipyridine in 2:1 stoichiometry to Cu2+ formed the aqua-coordinated [(6,6′-(O)2-(2,2′-bipyridine)2CuII(H2O)]2− like the homolog 2,2′-bipyridine compounds (Gerlach et al. 2014). Beside the catalytic water oxidation, a competing non-catalytic degradation has been identified. The 4,4′-disubstituted analogs were inactive in water oxidation. Thus, the neighboring O or OH groups indeed facilitate proton-coupled electron transfer and H-bonding interactions to promote water oxidation, in addition to the redox non-innocence of 6,6′-dihydroxy-2,2′-bipyridine (Burks et al. 2018) in line with findings on the corresponding IrII-complex (DePasquale et al. 2013).

The tridentate bis(arylimino)isoindoline pincer ligands can bind metal ions in neutral or anionic form owing to exocyclic imine functions that can act as internal proton acceptor site (Csonka et al. 2015; Saha et al. 2021) (Figs. 14 and 17). The rigid isoindolines occupy three co-planar sites around metal centers and often support both stability and reactivity, because variable coordination sites are available for substrate binding and activation.

Fig. 17
figure 17

Metal binding by bis(arylimino)isoindoline ligands in the radical, anionic and neutral, protonated forms. Several complexes have been characterized structurally, in which the ligand is anionic. In this case the two imine –C=N– groups exhibit symmetrical bond distance pattern that contributes to the rigid, planar structure and intense π − π* intra-ligand charge transfer absorption bands in the visible spectral region. Only a few complexes were characterized with neutral ligands, where one of the imine groups is protonated. The presence of this proton causes a characteristic change in the bonding pattern and shifts the intra-ligand charge transfer bands in the electronic spectrum. The radical form of the ligand is expected to be reactive and no structurally characterized complex is known. The protonation site and the redox-activity may create a suitable platform to facilitate proton-coupled electron transfer by this ligand family

Recently, FeIII- and CuII-bis(arylimino)isoindoline complexes were reported to catalyze water oxidation (Al-Zuraiji et al. 2020b, 2021; Benkó et al. 2021). The [FeIIICl2(1,3-bis(2′-thiazolylimino)isoindolinate)] complex (Váradi et al. 2013) was pre-catalyst to efficient water oxidation when immobilized on indium tin oxide electrodes and applied in borate buffer at pH 8.3 (Al-Zuraiji et al. 2020b). Product analysis indicated more than 80% Faraday efficiency and a turnover number of 193 (Table 3). Surface analysis before and after electrolysis and re-dissolution tests suggested that an immobilized molecular catalyst was responsible for catalysis and deactivation occurred by depletion of the metal.

The catalytically active form was claimed to arise by the exchange of chloride ligands to water molecules, while the pincer ligand rendered water-insolubility to the Fe(1,3-bis(2′-thiazolylimino)isoindolinate) assembly. Electrochemical and spectroscopic investigations in homogeneous water–acetone mixtures indicated a single-site molecular catalysis. Both ligand oxidation and intramolecular hydrogen-atom transfer were suggested to occur, as all three forms in Fig. 17 should be capable of metal binding. The latter conclusion relied on electrochemical detection of an intermediate in re-dissolved and chloride-free samples from drop-casted indium tin oxide that was previously exposed to long-term electrolysis. Interestingly, the ligand exchange occurring in water–organic mixtures could be exploited to develop a material saving electrodeposition method to form chloride-depleted catalyst ad-layers for efficient water oxidation electrocatalysis (Al-Zuraiji et al. 2021). Changing the thiazolyl arms to benzothiazolyl increased the durability of the catalytic films (Table 3), highlighting that the ancillary ligands play a crucial role in the stabilization of surface-adsorbates.

A more insightful, theoretical and experimental study on the possible combined redox-active and proton acceptor role of isoindoline-based pincer ligands was published very recently (Benkó et al. 2021). A pre-catalyst complex, [CuII(1,3-bis(2′-pyridyl)imino-isoindoline)(OClO3)(NCCH3)]+ was immobilized on indium tin oxide to generate O2 electrocatalytically for over 20 h at pH 10 in carbonate buffer, reaching a turnover number of 139 with no signs of CuOx or Cu(OH)2 formation (Table 3). The experimental results in turn indicated that a [CuII(1,3-bis(2′-pyridyl)iminoisoindolinate)(OH)] complex form dissolved in the aqueous phase might be responsible for catalysis. In order to identify the actual catalyst, [CuII(1,3-bis(2′-pyridyl)iminoisoindolinate)(OClO3)(OH2)] was isolated and structurally characterized in the solid state. Spectroscopy and density functional theory calculations were conducted to reveal a distorted equatorial coordination plane with three nitrogen atoms and one oxygen atom around copper in solution in close resemblance to that suggested by others for a similar redox-active ligand containing system (Gil‐Sepulcre et al. 2021). Each catalytic intermediate was tested with different spin multiplicities and the most stable states were identified. The proposed mechanism is shown in Fig. 18. In acetonitrile–water solutions, experimental findings supported by theoretical calculations suggested that the 1,3-bis(2′-pyridyl)-iminoisoindolinate ligand was oxidized while the CuII center remained redox-silent in the proposed catalytic cycle.

Fig. 18
figure 18

Proposed catalytic cycle of electrocatalytic water oxidation starting from [CuII(1,3-bis(2′-pyridyl)iminoisoindolinate)(OH2)]+ in acetonitrile–water mixture (Benkó et al. 2021). The cycle proceeds counterclockwise through proton-coupled electron transfer steps. The first oxidation generates a ligand radical stabilized by spin-delocalization over the central NCNCN structural motif. According to density functional theory calculations the next step generates a CuII-oxyl species that can undergo water nucleophilic attack. Hydrogen atom transfer to the ligand radical generates the neutral form 1,3-bis(2′-pyridyl)iminoisoindoline along with the O–O bond formation. Following this step, the cycle proceeds through rapid oxidations and closed by O2 evolution. Crossing between the quartet and doublet spin states plays an important role in the O–O bond formation that is a similar feature to the S2A, S2B spin multiplicity change in the oxygen-evolving complex of photosystem II

According to the suggested mechanism, the first proton-coupled electron transfer step would oxidize the isoindolinate ligand to produce [CuII(1,3-bis(2′-pyridyl)iminoisoindolyl)(OH)]+ that is followed by a second oxidation to generate the oxyl radical, instead of CuIII. The separation of the two oxidized sites explained the appearance of the second event only after adding H2O. The redox silence of CuII could be attributed to the spatial configuration of the complex, since the steric hindrance made difficult the formation of a square planar field around copper and stabilize the d8 CuIII configuration. The oxyl radical was found capable of single electron transfer water nucleophilic attack resulting in [CuII(1,3-bis(2′-pyridyl)iminoisoindoline)(OOH)]+. In this step, the electron from the attacking water molecule reduces the oxidized ligand and the ancillary ligand could then act as internal proton acceptor before an external base caught the released proton and restored again the anionic ligand form.

The spatial aspect of this proton capture was also considered. Since the N...H(Cu–OOH) distance was 4.82 Å in the [CuII(1,3-bis(2′-pyridyl)iminoisoindolinate)(HOOH)]+ intermediate, the assistance of a carrier is necessitated to transport a proton to the ligand N, asynchronous way with the formation of the O–O bond. Note that the [CuII(1,3-bis(2′-pyridyl)iminoisoindolyl)(O)]+ form contains three unpaired electrons, one at the ancillary ligand, one at the d9 CuII center and one at the oxyl ligand giving rise to different available spin multiplicities. Importantly, the calculated energy diagram of the system during the O–O bond formation clearly displayed crossing between the quartet and doublet states with the O–O distance near 2.0 Å. Note that spin multiplicity changes in the S state cycle of the natural oxygen-evolving complex of photosystem II is crucial to regulate the relative rates in the reaction sequence. Therefore, the utilization of redox-active ancillary ligands bearing labile protons may have prominent role in the advancement of future molecular catalysts for water oxidation.

3.4. Ligands participating in the oxygen–oxygen bond formation step

So far, the participation of ligands in the key O–O bond formation step has been limited to those oxyanions, which may in principle form peroxo-metallo-cycles with transition metals. These ligands are carbonate and borate, both having rich literature on reactivity as perhydrate and persalt, respectively (Pizer and Tihal 1987; Tsoler 1999; Durrant et al. 2011; Deary et al. 2013; Liu et al. 2021). The presence of carbonate or borate anions by a catalyst in a sufficiently high concentration is naturally expected only in electrolytes; thus, this section will include electrocatalytic examples.

In a catalytic cycle producing O2 from two H2O molecules, involving a single-site catalyst and inner-sphere coordination of oxyanions, one can imagine two general pathways. In Fig. 19, the cycle on the top starts from the aquo-metal assembly highlighted with a frame in the middle of the figure. This aquo-metal complex is oxidized and lose an electron and a proton, while the assisting LA ligand seen in blue remains coordinated through an oxygen donor. The next oxidation step concerns LA, in which a reactive oxyl radical is formed that in turn attacks the hydroxyl function, resulting in a peroxo-metallo-cycle and eventually O2 evolution.

Fig. 19
figure 19

Two possible mechanistic routes leading to O2 formation. The coordinating ligand O–LA–OH in blue stands for the assisting ligand that may participate in the O–O bond formation step in different ways, L in green represents an ancillary ligand, while the oxygen atoms in red are derived from bulk water molecules and the Mox, Mred and Mred2 stand for the oxidized, reduced and doubly reduced metal center. Two oxyanions were documented to function as the O–LA–OH in water oxidation catalysis, bicarbonate and carbonate and tetrahydroxyborate. The cycle on the top has been proposed for a carbonate-containing system. The reactivity relies on carbonate oxidation that initiates intramolecular peroxo-metallo-cycle formation (Winikoff and Cramer 2014). Two alternative 2e oxidation pathways for carbonate shortcutting the cycle to produce hydrogen peroxide are highlighted by gray background (Mizrahi et al. 2018; Mizrahi and Meyerstein 2019). The cycle on the top involves carbon-dioxide elimination shown as LA=O in the figure, and a bimolecular O–O bond formation step requires two catalyst molecules. The second highlighted pathway may apply when external carbonate is available to generate O2COO–CO22−. The cycle on the bottom applies to tetrahydroxoborate-containing systems differing in the location of OH protons and the Mox and Mred sequence from the cycle on the top (Huang et al. 2017; Lukács et al. 2019; Ruan et al. 2021)

The second option is the generation of higher metal oxidation state with oxyl radical Mox–O as shown in the cycle at the bottom of Fig. 19. In this case, the oxyl radical attacks a hydroxyl of LA forming peroxo intermediates. Stabilization of the Mox state generally requires suitable ancillary ligands (L in green, Fig. 19). The ancillary ligand can be also an oxyanion, as carbonate is known to stabilize higher oxidation states of metal ions in solution. The following examples may fit with one of the two general scenarios; however, as will be seen, there are ongoing debates about the dominant mechanisms.

3.4.1. Participation of carbonate in the oxygen–oxygen bond formation

An early report on copper based, electrocatalytic water oxidation introduced an astonishingly simple solution equilibrium system. Different copper salts were added to a NaHCO3 buffer at pH 8.2 or at pH 6.7 saturated with CO2. When high anodic potentials were applied to an inert indium tin oxide, fluorine-doped tin oxide or glassy carbon working electrode immersed into the solutions the electrocatalytic process yielded O2 with high charge efficiency (Chen and Meyer 2013). Complex formation between the metal and bicarbonate–carbonate in high excess was clear from the characteristic shifts of the d-d transitions in the electronic spectra.

Later, density functional theory calculations indicated that carbonate is a non-innocent, reactive ligand forming a peroxocarbonate metallo-cycle (Winikoff and Cramer 2014). This mechanism would be compatible with the cycle on the top of Fig. 19. The peroxocarbonate is then oxidized rapidly to yield O2 and CO2. The high reactivity was explained by means of a low energy singlet–triplet splitting in the 3e oxidized state facilitating the evolution of triplet molecular oxygen. An easy access by carbonate to both mono- and bidentate coordination modes allowing rearrangements that favor the O–O bond formation, and spin delocalization through the Cu–O bonding were pointed out as key features to reach a high catalytic rate (Table 4).

Table 4 Performance of molecular water oxidation electrocatalysts binding buffer anions that participate in the O–O bond formation step and the required experimental conditions

Alternative proposals argued in favor of reaction steps leading to short circuiting the catalytic cycle at the stage of peroxide formation that corresponds to a 2e + 2e mechanism. According to a combined experimental and theoretical study, O2COO–CO22− is produced from two CuIII-carbonate complex units (Mizrahi et al. 2018). This means the electrochemical oxidation of CuII to CuIII upon which no CO3·− radical anion should occur. Instead, the hydrolysis of C2O62− that is O2COO–CO22− gives H2O2 that reacts with the CuIII complexes producing O2. However, below pH 8.2 the electrocatalytic reaction was found to be first order in copper. This is not in line with the above conclusions that are rather compatible with a second order in copper. Note, on the other hand that second order kinetics in copper has been found in concentrated CO32− or HPO42− and PO43− buffers at pH 10.8 on the analogy to the known copper peroxide chemistry (Lewis and Tolman 2004; Elwell et al. 2017), but the rate-liming step was the formation of a CuO–OCu intermediate.

There are gathering arguments on the viability of similar mechanistic pathways for other first row transition metals like cobalt and nickel (Mizrahi and Meyerstein 2019; Patra et al. 2020a, b). In the case of planar NiII(1,4,8,11-tetraazacyclotetradecane) complexes with an aliphatic 4N donor environment two axially coordinated carbonate anions are required to stabilize the NiIII oxidation state (Ariela et al. 2017). Consequently, the concomitant oxidation of [NiIII(1,4,8,11-tetraazacyclotetradecane)(CO3)2] to a catalytically active NiIV species is either preceded by a slow exothermic carbonate-to-H2O ligand exchange to provide [NiIV(1,4,8,11-tetraazacyclo-tetradecane)(CO3)(OH)]+, or a bond breaking produces [NiIV(1,4,8,11-tetraazacyclo-tetradecane)(CO3)2(O)] + CO2 from [NiIV(1,4,8,11-tetraazacyclotetradecane)(CO3)2]. In the upcoming steps, the [NiIV(1,4,8,11-tetraazacyclotetradecane)(CO3)2(O)] + [NiIV(1,4,8,11-tetraazacyclotetradecane)(CO3)2(OH)] coupling would produce a NiIII-O–O-NiIII species that in turn would release H2O2 (Table 4).

If [NiIV(1,4,8,11-tetraazacyclotetradecane)(CO3)2(O)] reacts with an external HCO3, then O2COO–CO22− or HCO4 is produced, which leads again to H2O2 release. All mechanistic scenarios producing peroxide may be conceived as a short circuiting of the 4e process due to a disfavored metal promotion of cyclization as shown by the reaction steps with grey background in Fig. 19. The rest of the oxidation steps take place away from the inner sphere of the original complex.

Whichever is the case, the results underline that carbonate ions are not only ligands that stabilize NiIII, but also active, non-innocent participants in the redox process. Note that the occurrence of a η2-peroxo-monocarbonate metallo-cycle has been considered for manganese in catalytic epoxidation (Lane et al. 2002) and importantly, the X-ray structure of an L2FeIII-peroxomonocarbonate complex (Hashimoto et al. 2002) illustrated that such entities should not be eliminated from mechanistic speculations. The redox non-innocent scenarios for carbonate are especially fascinating since L–M compounds involved in water oxidation are known to bind carbonate (Chen et al. 2017; Benkó et al. 2021), and efficient material systems are also known with this anion (Ji et al. 2018).

3.4.2. Participation of borate in the oxygen–oxygen bond formation

With respect to borate only a few examples are known to date, in which this anion was suggested to participate in the O–O bond formation step, including an electrocatalytic system under neutral conditions using copper (Huang et al. 2017). The in situ formation of a stable Cu-borate catalytic film has been described earlier with a catalytic performance comparable to that of Ni-borate, but somewhat lower than that of Co-borate (Yu et al. 2015). In the study by Huang et al., the addition of borate to a 1 M solution of sulfate resulted in a soluble complex and UV–visible spectrophotometric evidence confirmed a ternary sulfate-Cu-borate catalyst species, in which sulfate acted as ancillary ligand. Electrocatalytic oxygen evolution at pH 7 was stable and no catalytically active deposit could be detected on the surface of the working electrode after several hours of electrolysis. The observed first-order dependences in both borate and copper suggested single-site catalysis relying on proton-coupled electron transfer steps.

The lack of a protium over deuterium kinetic isotope effect ruled out the participation of the inner sphere tetrahydroxyborate ligand in the rate-limiting step as a proton acceptor. Instead, the enhanced rate of electrocatalysis (Table 4) was associated with the participation of tetrahydroxyborate in the rate-limiting O–O bond formation step as oxygen donor, on the grounds of density functional theory calculations. This mechanism may be conceived as that shown in Fig. 19 (cycle on the bottom). In the key step, the 2e oxidized form, CuIII–O attacks a hydroxyl group on the coordinated (HO)3BO ligand providing an energetically favorable pathway for the O–O bond formation.

In the presence of different N-donor ligands, for example 2,2′-bypyridine, various tripeptides and a dinucleating peptidomimetic ligand, borate was also reported to coordinate to copper (Huang et al. 2017; Lukács et al. 2019; Ruan et al. 2021). However, the outcome of the electrochemical oxidation seems to depend strongly on the ancillary ligand. In the case of mononucleating tripeptides the formation of [LCuIII–OOB(OH)3]2− has been proposed upon electrochemical oxidation, leading to complex degradation and in situ CuO deposition (Lukács et al. 2019). As for the di-copper complex formed with the peptidomimetic ligand results implied that borate coordination occurred upon CuII→CuIII oxidation (Ruan et al. 2021). In the rate-limiting step the CuIII–O attacks the neighboring (HO)3B(O)CuIII tethered by the peptidomimetic ligand (for data see Table 4). In essence, considering the rate-limiting chemical step this mechanism corresponds to that shown in the bottom cycle of Fig. 19 noting that the order of the proton and electron transfers may be shuffled.

Nickel complexes have been scarcely studied in the presence of borate to our knowledge, and even in those cases the complex was unstable under catalytic conditions, and thus, the role of borate has not been discussed (Aligholivand et al. 2019). Recent studies conducted under alkaline conditions eventually concluded that nickel complexes undergo degradation and NiOx is formed under catalytic conditions no matter of what buffer was chosen (Feizi et al. 2018; Garrido-Barros et al. 2019). If one regards this general tendency of the nickel complexes, a similar mechanism to that in Fig. 19 is rather unlikely with this metal. On the other hand, one might wonder about the role of borate anions in water oxidation catalysis by nickel oxide surfaces.

The growth of nickel oxide films from borate electrolyte have been known for long (MacDougall and Graham 1981) and their utilization as efficient heterogeneous water oxidation catalyst was an important milestone in this field (Dinca et al. 2010). The role of borate has been studied in particular (Bediako et al. 2013). Bediako’s and Dinca’s studies elucidated that in borate-buffered electrolyte borate anions occupy the active surface sites and prior to commence the catalytic cycle their reversible dissociation from the domain edge of NiO6 octahedron is required.

A later study spectroscopically confirmed that an increase in the number of NiO6 domains is necessary to achieve an efficient water oxidation catalyst (Yoshida et al. 2015). Lately, a Ni3(BO3)2-Ni3S2 amorphous-crystalline heterostructure has been identified as an efficient bifunctional catalyst for both the hydrogen and the oxygen evolution reactions (Sun et al. 2020). Although the authors constructed schematic models of –OH, –O, and –OOH surface-bound intermediates on varied sites of Ni3(BO3)2, Ni3S2, and Ni3(BO3)2-Ni3S2 in order to calculate the free energy for each elementary step, the possible role of a peroxyborate intermediate was not discussed explicitly.

With respect to other first row transition metals, a few cobalt (Shaghaghi et al. 2021) and iron (Al-Zuraiji et al. 2020a, b, 2021) complexes have been studied in borate electrolyte. Moreover, Mn2OBO3 (manganese-oxoborate) was an efficient surface catalyst under neutral conditions (Elmaci et al. 2021). However, the possible role of borate in the O–O bond formation step has not been fully elucidated. The above examples illustrate that due to a similar, supportive function in the O–O bond formation step the use of carbonate or borate could be beneficial in advanced electro-, photoelectro- or photocatalytic water splitting and integrated CO2 reduction systems (Fu et al. 2019, 2020a, b; Liu et al. 2019a; An et al. 2019; Cao et al. 2020, 2021; Hamza et al. 2020; Feng et al. 2020; Chen et al. 2020; Li et al. 2020) to further improve efficiency.

3.5. Ligand redox transformations leading to surface catalysts

3.5.1. General degradation pathways of molecular catalysts and related analysis methods

Potential molecular catalysts of the oxygen evolution reaction can undergo degradation in rather general ways that are shown in Fig. 20. First row transition metal complexes can be especially prone to spontaneous hydrolytic degradation, or sensitive to the acidification of the solution that occurs at the anode upon catalytic oxygen evolution. In such cases, the hydrolytic dissociation of the complex does not require redox events, and in principle, the produced aqua or hydroxide complexes can serve as precursors to metal oxide or hydroxide surface catalysts (Fig. 20). However, this scenario is normally unwanted, since the solubility of metal oxides and hydroxides is limited and the bulk precipitation of the metal content makes this approach inefficient and less controllable. Therefore, spontaneous hydrolytic degradation should be avoided by proper ligand design. Other complexes act as precursors to catalytic films only when the oxidation state of either the metal center or the ligand is changed (Fig. 20). The use of redox-active precursors results in a much more controlled strategy for surface catalyst fabrication because the redox event produces the precursor species right at the target electrode surface.

Fig. 20
figure 20

Pathways for the degradation of an M(L) catalyst under conditions used in water oxidation. M is the metal center, L is the ancillary ligand, the Mox and Lox in red represent the oxidized forms, and MOx stands for the produced metal oxides in general. The first possibility is the spontaneous hydrolysis of the metal–ligand assembly that is an undesired reaction resulting from poor ligand choice. Electrocatalytic water oxidation often leads to a local acidification at the electrode surface. As a result, ligand protonation may induce ligand dissociation and in situ metal oxide formation. The third and fourth process starts with the oxidation of the metal or the ligand that changes the strength of the metal–ligand interaction. Upon dissociation, the aquo complex undergoes in situ oxidation at the electrode. The two pathways involving redox events are the most controllable and may be useful in the fabrication of highly efficient, nanostructured oxide coatings by electrodeposition

Under catalytic conditions, degradative side-reactions may affect any of the oxidized molecular catalysts to some extent that has been subject of several studies using advanced in situ techniques to reveal the true nature of water oxidation catalysis (Li et al. 2017). The technical advancements gave impetus to a gradual shift from the concept of traditional homogeneous and heterogeneous classification of catalysts to an evolutionary concept of dynamic catalyst transformations (Eremin and Ananikov 2017), ranging from molecules over nanoparticles and leaching-derived catalysts to complex catalyst mixtures.

A recent manuscript underlined that the combination of some main critical tests is crucial to elucidate the molecular nature of an electrocatalyst by the example of a known pentanuclear iron complex (Okamura et al. 2016). Rigorous analysis of several catalytic cycles in voltammetry and in parallel, X-ray absorption spectroscopy of the initial complex in frozen acetonitrile and that of the glassy carbon electrode surface used in voltammetry clearly demonstrated the in situ formation of iron oxides from the complex (Pelosin et al. 2020). Obviously, tracking such possible conversions of metal complexes to the corresponding oxides is of major concern from the perspective of drawing structure–activity relationships with respect to the molecular catalyst. However, the ex situ and in situ methods can be just as useful from the perspective of efficient precursor design; therefore, we listed typical techniques along with the experimental considerations and expected results in Table 5. Laboratory-scale application of classical spectroscopic, electroanalytical and separation-based analytical methods can be very helpful for basic distinctions between molecular and heterogeneous water oxidation processes. Further, direct tracking of the reaction mechanisms and intermediate species requires advanced in situ techniques that are pronouncedly resource intensive and only possible with the involvement of special areas of expertise. Finally, the ex situ analysis of the electrode surface is a classical approach that remains invaluable to provide a more complete understanding of a catalytic system.

Table 5 Various in situ and ex situ analysis methods that can elucidate the degradation of molecular catalysts

3.5.2. Possible ways of ligand-based redox transformations leading to surface catalysts

Herein, we would like to show a few examples on redox transformation of the ancillary ligand that can lead to an active surface catalyst in order to point out further perspectives. Importantly, the utilization of metal–ligand assemblies may be justified by special requirements, such as solubility limit of the precursor, narrow redox stability of the substrate material, need for size-control of the deposited catalyst material or surface morphology preserving method.

Electrocatalytic CoOx films that can be directly deposited from a number of cobalt complexes in situ are revealing examples. The organic ancillary ligands include oximes (Han et al. 2013), salen derivatives (Chen et al. 2015) and porphyrins (Daniel et al. 2017). The precise structure of the coordinating salen ligand had an especially strong influence on the catalytic performance of the resulted film. This established a strong link between molecular catalyst construction and precursor design for heterogeneous catalysts. Note that the oxide films produced from the above Co-complexes showed superior performance over those resulted from simple cobalt salts that clearly underlines the advantage of using organic ancillary ligands (Table 6). The above studies did not discuss the possible role of ligand oxidation in the catalytic film formation; however, oximates, salen and porphyrin ligands are all capable of chemical redox state changes in complexes (Wang and Groves 2013; Fu et al. 2014; Dinda et al. 2022). Therefore, the option of utilizing redox-active ligands in precursors should not be ruled out and worthy of further investigations.

Table 6 Examples of molecular precursors with suspected or proven redox-active ligand types that undergo in situ degradation to produce a surface catalyst film

Ligand degradation through redox transformation is obvious, if the identification of any resulting by-product is possible (Lu et al. 2016), but in most of the cases the initiating event cannot be revealed perfectly, because only a minute part of the molecular catalyst undergoes degradation and acts as precursor. Copper-diglycylglycine, which is a copper-tripeptide complex known since 1955 (Dobbie and Kermack 1955) is a good example. In fact, this water-soluble and stable complex can be precursor to a heterogeneous oxide catalyst in water oxidation (Lukács et al. 2019). In sodium borate solution at pH 9.2 different equilibrium species could be identified by a complementary spectroscopic and electrochemical analysis revealing axial coordination of borate, which allowed a new kinetic route of oxidation involving a CuIII-peroxo-borate intermediate.

At pH 12 and suitable absolute concentration copper-diglycylglycine exhibited a purely homogeneous water oxidation catalysis, but at lower pH, heterogeneous features appeared. Post-catalytic surface analysis revealed that a black amorphous Cu2O, CuO and Cu(OH)2 deposit was responsible for a smaller overpotential and efficient O2-evolution (Table 6). The produced spherical surface nanoparticles as well as their in situ generation from bio-degradable complexes are known in the literature, but the surface instability to overpolarization has remained unmentioned.

Operando spectroelectrochemical investigations revealed that anodic overpolarization on indium tin oxide may induce excessive oxidation of the nanoparticles ending up in the total destruction of the nanostructure and re-dissolution of the copper-oxide catalyst, possibly due to anodic current density limited by a chemically destabilizing density of surface CuIII-oxyl functions. The dynamics of the deposition and re-dissolution processes, the dependences on pH, buffer anion and type of the electrode all highlight the immensely prospective application of molecular precursors involving redox-active components in catalyst film fabrication.

Another example involves nickel complexes with substituted 1,2-phenylenebis(oxalamidate) (PBOA) ligands that were investigated as molecular water oxidation catalysts at basic pH (Fig. 7, R1=R2=H, R1=R2=CH3 and R1=H, R2=OCH3) (Lin et al. 2017; Garrido-Barros et al. 2019). Electrochemical studies revealed that the first oxidation wave corresponded to the NiII/NiIII redox couple. At pH 10–13 a second oxidation wave was observed due to the generation of a phenyl radical cation associated with the coordination of a hydroxide anion to form [NiIII(PBOA·)(OH)], which was responsible for the O–O bond formation and O2 release. The dimethyl and methoxy substituted 1,2-phenylenebis(oxalamidate) complexes also exhibited two oxidation processes. The electronic perturbation by the more electron-donating ligands resulted in a decreasing overpotential for water oxidation, 0.5 V for the parent 1,2-phenylenebis(oxalamidate) complex, 0.17 V and 0.22 V for the substituted ones, respectively.

While the oxalamidate complexes were stable in basic aqueous solutions from pH 10 to 13, below this pH the protonation of the amidate group caused metal ion release (Garrido-Barros et al. 2019). The 1e oxidized [NiIII(PBOA)] complex was fully stable within pH 10−12 on the electrochemical time scale, but at pH 13 the complex underwent progressive substitution in four steps leading to [NiIII(OH)4]. The doubly oxidized phenyl radical cation [NiIII(PBOA·)(OH)] also suffered from hydroxide substitution processes. Additional degradation occurred through the hydroxylation of the aromatic ring with the OH present in the medium that in turn was competing with the O−O bond formation.

Based on stability measurements as a function of pH, the authors concluded two possible pathways competing with the homogeneous catalytic process: first, the formation of the hydroxylated aromatic species with no water oxidation activity and second the formation of a catalytically active NiOx, that is a nickel oxide film of general formula NiXOyHz, attached to the surface of the electrode with higher overpotential for water oxidation than the molecular catalyst (Table 6). The dominant process was pH dependent, raising the pH from 10 to 12 promoted NiOx formation. At pH 13, the amount of the initially active NiOx decreased, indicating that the material dissolved during a further, slower process. The impact of the applied potential also proved that slight modifications in the experimental conditions might influence the stability and the catalytically active form of nickel complexes. Importantly, the parent complex could be used effectively as a molecular precursor for the formation of NiFeOx, that is nickel and iron oxides of general formula NixFe1-xOyHz that behave as extremely powerful water oxidation anodes (Table 6).

During controlled potential electrolysis at +1.05 V versus the normal hydrogen electrode, at pH 11 no NiOx could be detected, instead oxygen evolved at 94% Faraday efficiency with turnover number of 3.81. Similar experiments with the dimethylated 1,2-phenylenebis(oxalamidate) complex showed a higher stability and lack of NiOx emphasizing the opportunities in fine-tuning the redox properties of the ligand in order to control metal oxide film formation.

The pH-dependent operando formation of nickel oxide from a [N,N′-bis(salicylidene)ethylenediamino]nickel(II) complex was also investigated (Feizi et al. 2018). Here, the applied ligand is often found to play an active role in redox catalysis. However, the complex suffered degradation and a surface deposit containing Ni, P and O in a carbon matrix was formed. The complicated structure of the matrix clearly showed activity that was different from a simple nickel oxide highlighting the unique role of the ligand (Table 6).

No doubt, tunable and well-understood metal oxide formation originated from molecular catalysis should be considered as a practical option, when size- or morphology control of the deposit is important, when a metal oxide component has to be introduced into a system under controlled conditions, or the regeneration of a surface catalyst is required without dismantling the cell. Redox-active ligands should not be omitted from the line of options in such cases.

4. Conclusion

Molecular complexes are often regarded as the chemists’ warm little pond to allow insight into unexplored details of water oxidation and thus into artificial photosynthesis. Indeed, molecular compounds are intrinsically sensitive to degradation; thermodynamics dictates that the harsh oxidative conditions demanded by the endergonic oxygen evolving reaction most often are enough to eventually break up aliphatic bonds or open aromatic rings. Yet, there are examples of ligands accumulating electron vacancies, serving as internal base to facilitate coupling between electron and proton transfer, participating in the O–O bond formation, or allowing true catalyst formation upon redox transformation efficiently. Realization of such phenomena and the concept of self-healing catalytic systems may be guidance for new discoveries and a timely paradigm shift. From the viewpoint of direct practical applications, cost-efficient precursor complexes to surface catalyst films are of utmost importance. The diversity of first row transition metal complexes containing redox-active ancillary ligands is very promising, if refined methods are sought in order to carry out stringent, nanoscale surface modifications on conducting and semiconducting materials.

This overview is merely a short glimpse at the various catalyst types of an ever-evolving field – artificial photosynthesis. On the timescale of human research, the evolution of the natural system might seem static, but let us not forget that evolution has been in motion since the appearance of the first living organism and our species is only a milestone on the path. How many futile exercises shaped the image of biosphere known today, and how many left a mark that is waiting for a prepared explorer; these provoking questions should inspire us to step forward and explore new, advanced artificial systems.