Artificial photosynthesis: understanding water splitting in nature

Abstract

In the context of a global artificial photosynthesis (GAP) project, we review our current work on nature's water splitting catalyst. In a recent report (Cox et al. 2014 Science 345, 804–808 (doi:10.1126/science.1254910)), we showed that the catalyst—a Mn₄O₅Ca cofactor—converts into an ‘activated’ form immediately prior to the O–O bond formation step. This activated state, which represents an all Mn^IV complex, is similar to the structure observed by X-ray crystallography but requires the coordination of an additional water molecule. Such a structure locates two oxygens, both derived from water, in close proximity, which probably come together to form the product O₂ molecule. We speculate that formation of the activated catalyst state requires inherent structural flexibility. These features represent new design criteria for the development of biomimetic and bioinspired model systems for water splitting catalysts using first-row transition metals with the aim of delivering globally deployable artificial photosynthesis technologies.

1. The need for a global project on artificial photosynthesis

The world's power supply is based predominantly on the use of non-renewable energy resources such as oil, coal and natural gas. These geological resources have been accumulated over millions of years from photosynthetic activity and thus constitute ‘stored solar energy’. The steep rise in energy consumption over the last century has led to an inevitable shortage of these valuable resources with the economic, social and political consequences already felt today. Human population and economic growth, particularly in fast-developing countries, will lead to a further increase in energy demand. Furthermore, the burning of carbon-rich fuels has increased CO₂ concentration in the atmosphere. This increase is associated with anthropogenic climate changes, with projected adverse effects for our planet and human society (http://www.ipcc.ch/report). It is therefore one of the great challenges of mankind to identify and develop alternative sustainable energy sources [1–5].

Society's energy needs come in many forms. The dominant fraction is related to liquid fuels used for example in transportation (approx. 80%); whereas electrical energy only accounts for a small part of our energy consumption (approx. 20%) (https://www.bdew.de/internet.nsf/id/DE_Energiedaten). This is particularly problematic as there currently exists no satisfactory method which would be able to supply our demand for renewable liquid fuels. One possible solution would be to generate electricity using alternative, clean physical energy conversion technologies such as photovoltaics, wind turbines and hydropower and subsequently convert it into a chemical storage form. Such technologies already represent an increasing share of electrical energy generation worldwide (in Germany in 2013 16%) (https://www.bdew.de/internet.nsf/id/DE_Energiedaten). The question that then arises is what represents the best feedstock for a synthetic, renewable fuel. Here, biology provides an answer: photosynthesis uses water, a cheap, non-toxic, abundant material. Thus, in an idealized scenario ‘clean’ electricity would be used in the electrolysis of water, generating hydrogen (H₂), as a fuel [6]. Hydrogen itself is a very important base chemical and can be converted to many other important energy-rich materials for storage and/or transport, e.g. with CO₂ to methane, methanol or formic acid, and with molecular nitrogen to ammonia (figure 1).

Figure 1.

Biomimetic device showing light-induced charge separation in a photo(electro)catalyst P, transport of redox equivalents to a water splitting catalyst (A, anode) and to a hydrogen-producing catalyst (C, cathode). The generated primary ‘solar fuel’ is dihydrogen. This can be used directly, e.g. in fuel cells to generate electricity or burned in combustion engines, or can be converted for storage, transport or other uses. The examples are reactions with carbon dioxide to yield formic acid, methanol or methane and ammonia production with nitrogen (e.g. for fertilizers, agriculture). These reactions also require efficient catalysts.

A simple calculation shows that the yearly energy needs of a single person can be met by decomposition of 12 000 l of water, i.e. the contents of a small water pool (4 × 3 × 1 m). In addition, such a scenario eliminates the need for an independent power supplier, allowing energy production at the site of use [7], reducing transportation losses. A bottleneck though for such a ‘hydrogen economy’ blueprint is the relatively high price of the hydrogen produced. The high cost is derived from the materials used currently in commercially viable electrolysers: noble metals such as platinum or precious metal oxides (Ru and Ir oxides). These need to be replaced by less expensive metal catalysts before this model for the hydrogen economy becomes viable.

The generation of H₂ from water requires two catalytic steps: (i) the oxidation of two water molecules into O₂ and protons and (ii) the subsequent reduction of protons to molecular H₂. For proton reduction interesting approaches already exist (e.g. based on biomimetic catalysts from hydrogenase research) [8] that are close to enabling feasible commercial/industrial applications [9]. By contrast, there exists no completely satisfactory water oxidation catalyst. First-row transition metal based materials used for this reaction display typically a limited stability and durability (low turnover number, TON) or have an intrinsically low turnover frequency (TOF), precluding their wide range application. Another problem is often the excessive voltages needed to drive such catalysts (overpotential), which are due to the lack of suitable interfaces between the catalysts and the photo-active semiconductors.

Clearly renewed public investment should be targeted towards the scientific, technological and economic challenges needed to initiate the energy transition described above, e.g. the ‘Energiewende’ policy adopted by the current German government. The central discipline in this endeavour is chemical catalysis in all its facets, allowing energy to be efficiently used, stored, transported and interconverted. In this endeavour much can be learned from nature, which has developed many highly efficient metallo-enzymes [10] for the chemical conversion of small molecules with high TONs, long lifetimes and low overpotentials, and which even have built-in protection and repair mechanisms.

To solve the energy problems of the future requires initiatives such as the Global Artificial Photosynthesis (GAP) project [5], which bring together the efforts of many researchers across a wide range of scientific disciplines. Ideally, technologies that use nature as a blueprint, such as photosynthesis, should be considered part of the common heritage of mankind. We should aim that such technologies are affordable and accessible throughout the world with maximum benefit to all. We endorse the spirit of the GAP project: ‘Our goal is to work cooperatively and with respect for basic ethical principles to produce the scientific breakthroughs that allow development and deployment of an affordable, equitably accessed, economically and environmentally sustainable, non-polluting global energy and food system that also contributes positively to our biosphere’, as discussed at the Chicheley Hall conference in July 2014.

2. Design strategies for artificial photosynthesis, learning from nature

The term ‘artificial photosynthesis’ or equally ‘synthetic photosynthesis’ describes technologies that attempt to capture the energy of sunlight to make energy-rich molecules, i.e. ‘solar fuels’. This relies on the thesis that chemical bonds allow energy to be efficiently stored. This is essentially what nature does. It performs a photocatalytic water oxidation reaction which is coupled to carbon dioxide fixation (CO₂ reduction). In oxygenic photosynthesis, light-induced water oxidation occurs in a biological supercomplex, called photosystem II (PSII) that is found in all green plants, algae and cyanobacteria. It harbours a protein-embedded oxygen-bridged Mn₄Ca cluster with four bound water-derived ligands [11], as shown in figure 2. This cluster, called the water oxidizing complex (WOC) or oxygen evolving complex (OEC), is coupled to a pigment–protein reaction centre, which successively extracts electrons from it. After four such extractions, the complex splits water into protons and molecular oxygen, which is released as by-product, returning to its resting state (for reviews, see [13–16]). Protons generated via this process could, in principle, be used by another enzyme found in many green algae and bacteria, hydrogenase, which catalyses the conversion of protons to molecular hydrogen (for a recent review, see [8]). In figure 2 next to PSII, the structure of an [FeFe] hydrogenase is also depicted. The active site of this hydrogenase is a six-iron cluster, made up from a classical [4Fe–4S] cubane linked to a di-iron complex that carries uncommon ligands (3 CO, 2 CN⁻, azadithiolate) [8]. Under certain metabolic conditions, this enzyme is activated and produces hydrogen from excess protons. Other types of hydrogenases are known that harbour a different binuclear metal cluster ([NiFe] hydrogenases) that converts H₂ into protons and electrons, thereby generating energy for the organism [8,17,18]. In principle, both types of hydrogenases catalyse both reactions, dihydrogen splitting and production, albeit with different rates.

Figure 2.

Light-induced water splitting into molecular oxygen and hydrogen by the enzymes PSII and [FeFe] hydrogenase. This cascade reaction (water oxidation and subsequent proton reduction) is for example exploited by green algae, e.g. Chlamydomonas reinhardtii. The structure of the proteins ((a) PDB 3ARC [11], (b) PDB 1HFE [12]) and the structure of the active sites are shown (see text for details).

The efficient coupling of these two biological processes, water oxidation and hydrogen production (shown in figure 2), is currently being investigated in many laboratories aiming at producing (bio)hydrogen [19–23]. An interesting feature of the enzymes is the high turnover rate, which is about 500 s⁻¹ for the WOC and up to 10 000 s⁻¹ for the hydrogenase [24]. Unfortunately, both enzymes suffer from limited stability: the WOC has a half-life of only 30 min in average light under working conditions [25] and the [FeFe] hydrogenase is very sensitive to oxygen [26,27]; in the living cell, protection and repair mechanisms exist that efficiently resynthesize and replace the damaged parts of the enzymes. Among the [NiFe] hydrogenases, there is also a class of oxygen-tolerant enzymes [28–32].

An understanding of the principles and mechanisms of water oxidation and hydrogen conversion in nature would allow us to better design artificial systems capable of (bio)hydrogen production, be they biological photosynthetic organisms or purely synthetic systems, such as the hypothetical device in figure 1. For this latter system, the ultimate aim is to synthesize new catalysts for large-scale water splitting, hydrogen production and energy storage in chemical compounds. Such compounds may also be of importance for a future sustainable energy economy. In this short review, the current knowledge about the enzyme water oxidase (PSII) is described as an example. It is not intended to describe in depth the biological system but instead focus on the active metal cluster, and the recent insight gained about its function by structural, spectroscopic and theoretical results. Design criteria for the construction of respective biomimetic and bioinspired model systems for water splitting are discussed. We note that the reversible conversion of protons to molecular hydrogen is described in recent reviews (e.g. [8]).

3. Nature's water splitting catalyst

Nature's water oxidizing catalyst represents a penta-oxygen tetramanganese–calcium cofactor (Mn₄O₅Ca) [11,13,16] (figure 2). Its catalytic cycle comprises five distinct redox intermediates, the S_n states, where the subscript indicates the number of stored oxidizing equivalents (n = 0–4) required to oxidize two water molecules and release dioxygen [33] (figure 3b). It is embedded in a unique pigment–protein supercomplex, PSII, which acts as a photosensitizer. Outer light harvesting complexes absorb visible light and funnel excitation energy to the reaction centre of PSII, where charge separation takes place. The reaction centre comprises chlorophyll, pheophytin and quinone molecules arranged about a twofold symmetry axis [35]. The radical cation generated upon charge separation, termed the primary oxidant (P680^•+), is predominately associated with the chlorophylls P_D1 and P_D2 (figure 3a). The photochemically generated electron is passed in succession to the pheophytin (Pheo_D1) and subsequently to the bound plastoquinone (Q_A), increasing the electron–hole distance, thus stabilizing the charge separated state. Reaction centre photochemistry occurs on a picosecond to nanosecond time scale [35].

Figure 3.

(a) Electron transport chain of PSII linking reaction centre photochemistry to the site of water splitting catalysis. (b) The catalytic (Kok) [33] cycle of the cofactor showing the sequence of electron and proton withdrawals and the time constant for each transition [34].

Subsequently, the electron is passed from Q_A to a second plastoquinone (Q_B), which acts as a mobile charge carrier. It accepts two electrons and two protons, before leaving its protein pocket to be exchanged with a new plastoquinone. In this way, electrons are passed to the next photosynthetic supercomplex, the cytochrome b₆f complex.

The photogenerated cation P680^•+ is coupled to the Mn₄O₅Ca cofactor via an intermediary redox-active tyrosine residue Y_Z, which acts as a single-electron gate (figure 3a). Initial electron (hole) transfer between P680^•+ and Y_Z occurs on a nanosecond time scale, whereas the subsequent electron (hole) transfer between Y_Z^• and the Mn₄O₅Ca cofactor occurs on a microsecond to millisecond time scale, depending on the intermediate (S-state). Fast reversible oxidation of Y_Z is a property of the protein pocket. A nearby histidine residue acts as an intermolecular base, stabilizing the deprotonated (Tyr-O^•) state, while structured waters (bound to the Ca²⁺ ion) act to tune its redox potential [11,36]. After four charge separation events, the transiently formed S₃Y_Z^• state [S₄] rapidly decays to the S₀ state with the concomitant release of molecular triplet dioxygen and probably the rebinding of one substrate water molecule [37,38].

4. Key design features of the oxygen evolving complex

Nature's ‘photosynthetic cell’—the water splitting cofactor coupled to the pigment–protein reaction centre photosynthesizer—has a number of unique design features, including the following.

• (i) Light-induced single-electron transfer on a picosecond time scale coupled to a much slower millisecond four-electron process, the oxidation of two water molecules to molecular oxygen.

• (ii) The temporary storage of oxidation equivalents required for this reaction, i.e. transient oxidation of the Mn ions.

• (iii) Decoupling of O–O bond formation and proton release. Protons are released continuously throughout the S-state cycle [34], whereas the O–O bond forms in a single concerted step. Cofactor oxidation coupled with deprotonation (charge balancing/redox levelling) ensures that the redox potential of each of the four-electron withdrawal steps is similar, allowing all transitions to be driven by a single oxidant, P680, i.e. it lowers the overpotential needed to drive catalysis.

• (iv) Efficient interfacing of the catalyst and the photosensitizer via a redox-active tyrosine Yz.

• (v) Regulated substrate (water) delivery and product release (O₂, H⁺) via protein channels.

• (vi) Protection and repair mechanisms that limit oxidative damage of the protein pocket and replace the protein scaffold every 30 min under full sunlight [39]. It is noted that the water splitting catalyst binds to the protein and assembles spontaneously from free Mn²⁺ and Ca²⁺ in solution under visible light, a process called photoactivation [40].

5. The structure of the catalyst

The arrangement of the five metal ions and the connectivity of the cofactor (oxo bridges) have been resolved by conventional X-ray crystallography using synchrotron radiation (XRD) [11] supplemented by femtosecond X-ray crystallography using a free electron laser (XFEL) [41]. The structure represents a distorted chair where the back of the chair is defined by a plane containing Mn₁, Mn₃ and Mn₄. A second plane containing the Ca ion and Mn₂ defines the front of the chair. The Mn–Mn distances seen in the XRD structure are systematically longer than those seen in extended X-ray absorption fine structure (EXAFS) and in model complexes [42–46]. This is not the case for the XFEL structure, where the same Mn–Mn distances as seen in EXAFS are observed. The longer Mn–Mn distances in the XRD structure are thought to reflect reduction [47–49] of the Mn ions, and as such the XFEL structure is considered more reliable.

Our own work has concentrated on the development of methods for calculation of spectroscopic observables from magnetic resonance and X-ray spectroscopies, to further refine structural models of the cofactor and precisely describe the electronic structure of the catalyst in all catalytic states [50,51]. Our starting point was the structure of the cofactor in the S₂ state, for which extensive spectroscopic data exist, in which the net oxidation state of the cofactor is Mn^IIIMn^IV₃. Theoretical modelling has shown that the cofactor can adopt two interconvertible core topologies in this state [52–54], an open cubane (S₂^A) and a closed cubane motif (S₂^B) (figure 4). The S₂^A form is slightly lower in energy by ca 1 kcal mol⁻¹ and displays an electronic ground state of spin S_G = 1/2. In this structure, the only Mn^III ion of the cofactor, which is five-coordinate, is found within the cuboidal unit (Mn1, figure 5a). In the S₂^B form, the cofactor instead displays an electronic ground state of spin S_G = 5/2. In this structure, the five-coordinate Mn^III ion is now located at the terminal Mn4 (figure 5a) [52–54]. Movement of the Mn^III site is thus coupled to a reorganization of the core connectivity of the cofactor, involving the movement of a single μ-oxo bridge (O5) (figures 4 and 5a). In the open structure (S₂^A), O5 represents a bis-μ-oxo linkage between Mn3 and Mn4 (open form, S₂^A), whereas in the closed structure it becomes (S₂^A) a vertex of the heterobimetallic cubane unit. These two structures show a one-to-one correspondence with the two well-known electron paramagnetic resonance (EPR) signals for the S₂ state, the multiline (S = 1/2) and g4 (S = 5/2) signals (figure 5a) [52].

Figure 4.

Models for each S state of the Kok cycle, showing the optimized geometry, protonation pattern and Mn oxidation states of the inorganic core, modified from Krewald et al. [55]. The cofactor exists in two forms in the S₂ state, an open cubane (S₂^A, S_G = 1/2) and a closed cubane (S₂^B, S_G = 5/2) structure which differ in their core connectivity by the reorganization of O5 [52].

Figure 5.

(a) CW-EPR data of the two forms of the S₂ state (S₂^A and S₂^B) modified from Cox et al. [13]. (b) High field (W-band) data of the S₃ state consistent with an open S₃^A type structure, modified from Cox et al. [56]. (c) ¹⁷O-EDNMR data of exchangeable oxygen sites of the Mn₄O₅Ca cofactor, modified from Lohmiller et al. [57]. The splitting of the largest doublet (green, O5) is sensitive to both Ca²⁺/Sr²⁺ replacement and W1/NH₃ exchange. (d) Replacement of W1 by NH₃. Also shown are the three different sites where ¹⁷O labelled water is taken up by the cofactor: (i) as a ligand to the Ca²⁺ ion, (ii) as a ligand to the outer Mn4 ion and (iii) as a μ-oxo bridge (O5) [58].

Recent multifrequency EPR measurements from our laboratory have now resolved the net oxidation state of the S₃ state (figure 5b). We find the net oxidation state of the cofactor is Mn^IV₄ [56], categorically ruling out a ligand centred oxidation during the S₂–S₃ transition as has been proposed in the literature. Concomitant double resonance data suggest that the cofactor adopts a single configuration (unlike the S₂ state), which is best described as an open cubane (S₃^A) structure (figures 4 and 5b). Importantly, these results require that all four Mn ions are six coordinate in the S₃ state. This constraint necessitates the binding of a new water molecule to the open coordination site of the cofactor in the form of a hydroxide (figures 4 and 5b) [56,59]. Further evidence for water binding during the S₂–S₃ transition comes from work of the Noguchi laboratory [60]; their Fourier transform infrared spectroscopic data suggest that the solvation of the cluster increases during the S₂–S₃ transition. Interpretations of X-ray absorption near-edge structure data of the S₂–S₃ state are also consistent with a water binding event, where the absence of a large edge shift may reflect the five-coordinate Mn^III ion of the S₂ state becoming six coordinate in the S₃ state [61].

From this starting point, i.e. an understanding of the atomic structure of the S₂ and S₃ states that includes assignments of the energetically most favourable and spectroscopically consistent protonation states of all titratable ligands [55], we recently extended our spectroscopy-oriented theoretical modelling to the S₀ and S₁ states. The guiding principle for selecting the best models was again maximal consistency with experimental observables, including spin states (S_G = 1/2 for S₀ and S_G = 0 for S₁) [56,62,63], Mn hyperfine couplings, metal–metal distances from EXAFS and profiles of Mn K-pre-edge X-ray absorption spectra [55]. Figure 4 summarizes the results of this effort, showing the cofactor geometry, protonation states of titratable groups and individual Mn oxidation states for all spectroscopically consistent S-states models of the Kok cycle.

6. The substrate (water) molecules of the oxygen evolving complex

Experiments performed with isotopically labelled water (H₂¹⁷O/ H₂¹⁸O) provide a means to identify the two substrate binding sites of the water oxidizing cofactor. Earlier membrane-inlet mass spectrometry (MIMS) results [64], which monitor the uptake of H₂¹⁸O into the product O₂ molecule, have demonstrated that the complex contains two chemically different substrate sites: an early (W_s) and late (W_f) binding substrate site, both of which exchange with bulk water in all catalytic states (S-states) [65–67]. Owing to the relatively slow rate of exchange of W_s (≈1 s⁻¹), and its dependence on the oxidation state of the Mn tetramer, W_s is usually considered to be an oxygen ligand of one of the Mn ions [38,67]. By using water labelled with the magnetic isotope (¹⁷O, I = 5/2), the same substrate binding site (W_s) can be characterized spectroscopically, using the EPR technique ELDOR-detected NMR (EDNMR) [13,58,68]. These measurements identify a unique, exchangeable μ-oxo bridge as a potential candidate for W_s and the identity of the exchangeable μ-oxo bridge has been inferred from small site perturbations of the cofactor (figure 5c).

Ammonia, a substrate water analogue, is known to bind to the water oxidizing cofactor in the S₂ state [69,70]. Its binding mode was recently determined as a terminal ligand to the Mn4 ion, replacing the water ligand W1 [71,72]. Such a binding motif readily explains the observed hyperfine and quadrupole couplings; the asymmetric quadrupole coupling is a property of the H-bonding network of the W1/NH₃ site (Asp61) [57,72]. W1 displacement as opposed to other binding modes (μ-oxo bridge displacement) is also preferred based on energetic considerations and an analysis of available vibrational data [71–73]. In samples where NH₃ is added, the hyperfine coupling of the exchangeable μ-oxo bridge is strongly perturbed (ca 30%) [57,72]. This result is readily understood as due to a trans effect: the W1 ligand is trans to the O5 bridge. Displacement of W1 by NH₃ elongates the O5–Mn4 distance perturbing the observed hyperfine coupling (figure 5c,d) [71].

The assignment that the exchangeable μ-oxo bridge is O5 is further supported by measurements on PSII, where the Ca²⁺ ion is biosynthetically replaced with Sr²⁺ [57]. In such samples, the hyperfine coupling of the exchangeable μ-oxo bridge is perturbed (ca 5%, figure 5c). As O5 is a ligand to the Ca²⁺/Sr²⁺ site this result is again easily understood. Furthermore, in samples where the Ca²⁺ is replaced by Sr²⁺ and NH₃ is bound to the cofactor, an additive effect is seen on the exchangeable μ-oxo bridge hyperfine coupling (ca 35%) [57], again consistent with the notion that O5 represents the exchangeable μ-oxo bridge (figure 5c).

The above results on Ca²⁺/Sr²⁺ exchange point to O5 representing W_s. This hypothesis comes from correlation to MIMS results which show that Ca²⁺/Sr²⁺ replacement perturbs the rate at which W_s exchanges with bulk water [38,67,74]. As Ca²⁺/Sr²⁺ replacement similarly affects the electronic properties of O5 (but no other exchangeable oxygen ligand) [57], this result can again be easily understood. We note that the fast rate of exchange of O5 when compared with that of μ-oxo bridges in simple synthetic model systems is likely due to the unusual flexibility of its metal coordination (figures 4–6) [52,53].

Figure 6.

The reaction cycle of the Mn₄O₅Ca cofactor. In the resting states (S₀, S₁, S₂^A), the cofactor represents an ‘open’ cubane structure with an open coordination site for substrate binding. In the activated states (S₃ and possibly S₄), the structure also resembles an ‘open’ cubane but now the open coordination site is occupied by a hydroxide ligand. Structural flexibility in S₂ (S₂^A, S₂^B) is hypothesized to be critical for the insertion of this water-derived ligand and formation of an activated S₃ state. The O–O bond may then presumably form between the flexible µ-oxo bridge and the inserted (proximal) water-derived ligand.

Assignment of the second substrate site W_f is less clear. The above description of the structure of the S₃ state suggests that the second substrate would represent the newly bound terminal hydroxide ligand, in close proximity to O5/W_s. This result though appears at odds with MIMS data that show that W_f's exchange kinetics are almost identical for the S₂ and S₃ states [13,37,75], suggesting that it is associated with the Mn₄O₅Ca cofactor in the S₂ and S₃ states in the same way [38,75]. A possible solution is that W_f represents a pre-existing water ligand of the cluster (W2 bound to the Mn4 ion or W3 bound to Ca²⁺ ion) which is relocated in the S₃ state, suggesting a concerted mechanism for substrate delivery [38,71,75,76].

7. The reaction cycle of the catalyst

As described above, the water oxidizing cofactor takes two discrete conformations during its catalytic cycle and as such, its reaction cycle can be divided into two (figure 6). In the ‘resting’ S-states (S₀, S₁), the cofactor adopts a low-spin configuration (S = 1/2 and 0) [62,63]. Calculations demonstrate that this is a property of the incomplete cuboidal unit [52,55], i.e. a manganese unit with a vacant coordination site. In this form, the catalyst is inactivated. In the ‘activated’ S states, the cofactor instead adopts a high-spin configuration (S = 3) [56], concomitant with the insertion of a water ligand into the cubodial unit. S₂ displays both low- and high-spin forms (S = 1/2 for the open cubane form S₂^A, S = 5/2 for the closed cubane form S₂^B) suggesting that it may represent a switching point in the catalytic cycle (figure 6). Spin-state interconversion, which involves the opening of the cubodial unit may facilitate water insertion/relocation into the heterobimetallic cubane. It is noted that EPR data suggest that Ca²⁺ is necessary to proceed past the low-spin S₂Y_Z^• state, and thus Ca²⁺ may play a role either in allowing efficient S₂^A/S₂^B interconversion, or at least in allowing spin-state crossover [77–79]. It is then the high-spin topology that allows low barrier single-product formation to proceed, for example, via an oxo–oxyl coupling mechanism as proposed by Siegbahn [56,59,80]. Late second substrate binding may avoid deleterious two-electron chemistry (slow peroxide formation) in the resting states of the catalytic cycle [81].

8. Conclusion and outlook

The results described above place new, important restrictions on the structure of nature's water oxidizing cofactor, its catalytic cycle and the mechanism of water splitting and O₂ evolution. These are listed below with reference to design criteria for the development of biomimetic and bioinspired water splitting catalysts using manganese.

• (i) The oxidation states of all individual Mn sites in all S-states are now resolved: S₀ (Mn^III)₃Mn^IV, S₁ (Mn^III)₂(Mn^IV)₂, S₂ Mn^III(Mn^IV)₃, S₃ (Mn^IV)₄ and ligand oxidation can be excluded prior to formation of the S₄ state. This assignment is in agreement with the average Mn oxidation state inferred for heterogeneous manganese oxide water oxidation catalysts [82].

• (ii) An exchangeable μ-oxo bridge (O5) is likely the first substrate of the water splitting reaction. This result, coupled with the observation that the cofactor gains a new hydroxide ligand in the S₃ state supports the notion that the O–O bond forms between oxygen ligands of neighbouring redox-active Mn ions (Mn₁ and Mn₄).

• (iii) Structural flexibility of the cofactor appears to be critical for function, namely catalyst activation, i.e. the formation of the S₃ state, which is coupled to second substrate inclusion. This flexibility is only readily apparent in the S₂ state, with the other S-states adopting a single preferred configuration. This property seems to be conferred in part by the Ca²⁺ ion, which also may act to tune the redox/magnetic properties of the cuboidal (Mn₃Ca) unit [83], or of the tyrosine radical [36] or both. The inclusion of calcium has been correlated with enhanced water splitting activity of manganese oxide water oxidation catalysts [82,84].

The heterobimetallic nature of the cluster and the choice of Mn as the redox active metal for biological water splitting must therefore reflect a confluence of effects. Key properties are

• (i) that it can stabilize multiple oxidation states to be accessed over a narrow range of redox potentials. This property is derived from protein/cofactor interactions being able to tune the pK_a of cofactor bound water (substrate) molecules allowing sequential H⁺ release upon each Mn oxidation event;

• (ii) the possibility that the heterobimetallic cubane allows the formation of an energetically accessible Mn^V-oxo or Mn^IV-oxyl radical in the transition state [85–88]; and

• (iii) that the Mn ions represent trapped valence states of local high-spin states, which magnetically couple together in a precise way in each S-state.

While the points above pose interesting theoretical questions, from a more practical perspective, the static properties of heterobimetallic cubane appear to be universal and can be reproduced in simpler model compound systems [89–91]. Similarly, heterogeneous Mn/Ca materials already show reasonable, if somewhat slow water splitting capacity normalized to Mn content [82,92,93]. Although it cannot be excluded that such Mn/Ca materials adopt lower efficiency pathways for O–O bond formation, the alternative explanation, that such materials only contain an intrinsically small number of ‘active’ heterobimetallic cubane units, seems equally reasonable. If, as figure 6 suggests, nature's catalyst essentially toggles between an ‘inactive’ and ‘active’ form, the latter having a substrate bound within the heterobimetallic cubane, the challenge in designing a better catalyst amounts to introducing this property (structural flexibility) or stabilizing a solvent accessible cuboidal structure that is stable/robust. These key design criteria should form the basis for new synthetic complex development—akin to the progress recently made in H₂ production using hydrogenase mimics [9]—for water splitting catalysts using manganese.

Funding statement

This work was supported by the MPG, the Cluster of Excellence RESOLV (EXC 1069) and the DIP (Lu 315/17-1), both funded by the Deutsche Forschungsgemeinschaft (DFG).