How to make an efficient and robust molecular catalyst for water oxidation

Pablo Garrido-Barros ab, Carolina Gimbert-Suriñach a, Roc Matheu ab, Xavier Sala c and Antoni Llobet *ac
aInstitute of Chemical Research of Catalonia (ICIQ), Barcelona Institute of Science and Technology (BIST), Avinguda Països Catalans 16, 43007 Tarragona, Spain. E-mail: allobet@iciq.cat
bDepartament de Química Física i Inorgànica, Universitat Rovira i Virgili, Marcel·lí, Domingo s/n, 43007 Tarragona, Spain
cDepartament de Química, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, 08193 Barcelona, Spain

Received 6th April 2017

First published on 18th July 2017


Energy has been a central subject for human development from Homo erectus to date. The massive use of fossil fuels during the last 50 years has generated a large CO2 concentration in the atmosphere that has led to the so-called global warming. It is very urgent to come up with C-neutral energy schemes to be able to preserve Planet Earth for future generations to come and still preserve our modern societies' life style. One of the potential solutions is water splitting with sunlight (-WS) that is also associated with “artificial photosynthesis”, since its working mode consists of light capture followed by water oxidation and proton reduction processes. The hydrogen fuel generated in this way is named as “solar fuel”. For this set of reactions, the catalytic oxidation of water to dioxygen is one of the crucial processes that need to be understood and mastered in order to build up potential devices based on -WS. This tutorial describes the different important aspects that need to be considered to come up with efficient and oxidatively robust molecular water oxidation catalysts (Mol-WOCs). It is based on our own previous work and completed with essential contributions from other active groups in the field. We mainly aim at describing how the ligands can influence the properties of the Mol-WOCs and showing a few key examples that overall provide a complete view of today's understanding in this field.


image file: c7cs00248c-p1.tif

Pablo Garrido-Barros

Pablo Garrido Barros graduated in Chemical Engineering at the University of Granada in 2013. In 2014, he obtained a Masters Degree in Synthesis, Catalysis and Molecular Design from Rovira i Virgili University. Then he joined Prof. Antoni Llobet's group at the Institute of Chemical Research of Catalonia (ICIQ) as a graduate student where he has been working on the development of water oxidation catalysts.

image file: c7cs00248c-p2.tif

Carolina Gimbert-Suriñach

Carolina Gimbert Suriñach obtained her PhD from the Autonomous University of Barcelona (UAB) in 2008. After one year as an assistant professor at the same university, she moved to the University of New South Wales (UNSW) to undertake post-doctoral research. In 2012 she joined Prof. Antoni Llobet's team at the Institute of Chemical Research of Catalonia (ICIQ), where she works on hydrogen evolution and water oxidation photocatalysis for water splitting.

image file: c7cs00248c-p3.tif

Roc Matheu

Roc Matheu obtained his bachelor degree in Chemistry from the Autonomous University of Barcelona in 2012. The same year he joined Prof. Antoni Llobet's group at the Institute of Chemical Research of Catalonia (ICIQ) as a graduate student where he has been working on the development of molecular water oxidation catalysts.

image file: c7cs00248c-p4.tif

Xavier Sala

Xavier Sala obtained his PhD in Chemistry from the University of Girona in 2007. After his postdoctoral research with P. W. N. M. van Leeuwen and A. Llobet at ICIQ, he is currently Associate Professor of Chemistry at the Autonomous University of Barcelona (UAB) where he leads the Selective Oxidation Catalysis (SelOxCat) research group.

image file: c7cs00248c-p5.tif

Antoni Llobet

Antoni Llobet received a PhD from UAB. Then he was a Postdoctoral Fellow at the University of North Carolina with T. J. Meyer and at Texas A&M University with D. T. Sawyer and A. E. Martell. He is currently a Professor of Chemistry at UAB and a Group Leader at ICIQ.



Key learning points

1. Controlling redox properties of transition metal complexes

2. Designing complexes that can access reactive high oxidation states

3. Visualizing the potential mechanism of the water oxidation reactions

4. Essential tools to characterize catalytic water oxidation reactions

5. Description of deactivation paths that need to be avoided


1. Introduction

The ability to control fire (energy) by early humans triggered human evolution. Fire provided not only a source of heat but also protection and a method for cooking. As a whole, it generated a turning point in human culture and development. Evidence for the use of fire in a controlled manner by Homo erectus dates from 400[thin space (1/6-em)]000 years ago.1 Since then energy has played a major role in the development of societies as is the case of the steam engine in the industrial revolution powered by carbon and the conception of our modern societies fueled mainly by carbon, natural gas and oil.2 A common feature of the history of mankind is the use of biomass and fossil fuels to generate energy that in turn produces CO2, a greenhouse gas. The massive burning of fossil fuels during the last 50 years has resulted in a large CO2 concentration in the atmosphere that is responsible for today's global warming and the catastrophic consequences associated with the latter.3

In order to preserve Planet Earth it is imperative that our modern societies shift from fossil to C-neutral energy sources and thus new sustainable and environmentally respectful energy conversion schemes are needed.

Water splitting with sunlight (-WS) is one of these potential respectful solutions and fits within the concept of energy democracy.4 It consists of simultaneous oxidation and reduction of water to form dioxygen and dihydrogen, respectively, as indicated in eqn (1),

 
image file: c7cs00248c-t1.tif(1)
This reaction is thermodynamically uphill and thus ideally should be driven by sunlight. In this respect this implies storing solar energy in chemical bonds. This is analogous to the photosynthetic process carried out by green plants and algae where carbohydrates, C6H12O6,5 are obtained from CO2 and water as shown in eqn (2),
 
6CO2 + 6H2O → C6H12O6 + 6O2 ΔG° > 0(2)
To be able to carry out these reactions successfully both in nature and in -WS a number of conditions need to be met. First, neither H2O nor CO2 absorbs sufficient energy from sunlight to be able to break the needed bonds and thus an alternative mechanism will be needed. In nature chlorophylls are responsible for absorbing sunlight and produce a charge separation state that is followed by properly catalyzed oxidative and reductive processes. In the oxidation site for the water oxidation reaction (eqn (3)), nature uses the Oxygen Evolving Center (OEC) at Photosystem-II (PSII) as a catalyst, a Mn4O4Ca cluster whose structure is still a matter of debate today,5
 
H2O + H2O → O2 + 4e + 4H+E° = 1.23 V vs. SHE at pH = 0(3)
The water oxidation catalyst is thus one of the key elements of the whole process that provides a fast conversion of water to dioxygen at a relatively low overpotential. In a similar manner a successful water splitting device using sunlight will need an efficient and robust water oxidation catalyst.

Regarding man made water oxidation catalysts there are mainly two different types of compounds used for this purpose, namely molecular catalysts and oxides. The capacity of oxides to carry out water oxidation to dioxygen electrocatalytically can be traced as far back as 1902 in the work of Cohen and Glazer.6 On the other hand, well defined molecular water oxidation catalysts (Mol-WOCs) capable of oxidizing water to dioxygen needed the development of coordination chemistry and thus were not discovered until 1982 by TJ Meyer.7 The present review is solely devoted to Mol-WOCs.

Most of the knowledge we have today about Mol-WOCs comes mainly from detailed, electrochemical, mechanistic and reactivity studies that have been carried out using Ru complexes.8 However, this information extracted from Ru complexes can be carefully extrapolated to other transition metal catalysts. For this reason, the following discussion will be mainly based on Ru complexes but where appropriate we will also introduce other relevant transition metal complexes.

The objective of the present review is not to give a comprehensive description of all the molecular complexes that have been described in the literature capable of oxidizing water to dioxygen but to describe the key elements that give a consistent and rational understanding of the phenomena that contribute to water oxidation catalysis and thus allow us to come up with a unified description. Therefore, the following is an account of how ligands and transition metal geometry influence reactivity.

2. Access to high oxidation state species and redox potentials

The four-electron four-proton nature of the water oxidation catalysis (eqn (3)) implies that the catalyst will need to act as a reservoir of holes so that at a sufficiently high oxidation state it will be able to oxidize water to dioxygen by accepting 4 electrons. Subsequent removal of “electrons-only” from a transition metal complex by simple outer sphere electron transfer can generate high oxidation states. However, upon removal of electrons-only the Coulombic charge generated at the metal center will produce oxidized complexes of very high energy that will not be useful for subsequent electron removal. On the other hand, the removal of protons and electrons at the same time, by proton coupled electron transfer (PCET), allows highly reactive high oxidation states to be reached at a relatively low energy. An early example where this phenomenon occurs is the Ru–aquo polypyridyl complex, [RuII(L1)(H-L2)(H2O)]2+, 19 (see Charts 1 and 2 for ligand labeling), where a couple of PCET processes occur at pH = 7 as shown in eqn (4), where the auxiliary polypyridyl ligands are not shown.
 
image file: c7cs00248c-t2.tif(4)
A complete description of the dominant species depending on the degree of oxidation and protonation can be graphically represented in the corresponding Pourbaix diagram as depicted in Fig. 1 for 1.

image file: c7cs00248c-c1.tif
Chart 1 Ligands I.

image file: c7cs00248c-c2.tif
Chart 2 Ligands II.

image file: c7cs00248c-f1.tif
Fig. 1 Pourbaix diagram for complex [RuII(L1)(H-L2)(H2O)]2+, 1, showing the variation of the different redox potentials (vs. NHE) as a function of pH. The labels indicate the zone of the predominance of the main species. Only the Ru–OH2 group at different oxidation states and different proton content is shown (auxiliary ligands L1 and H-L2 are not shown). Vertical lines indicate pKa values. Data extracted from ref. 9.

In this particular case the removal of protons and electrons from the RuII–OH2 group allows the generation of high oxidation state species with a constant overall charge. In the specific case of RuIV[double bond, length as m-dash]O, this species is further stabilized by the σ- and π-donating ability of the oxo group. In order to easily track the electron content, all the metal oxidation states in this report are described as “formal” oxidation states. However, and obviously, for all the aquo transition metal complexes discussed here the oxygen atom of the M–O group will also be responsible for partial electron density removal upon oxidation, at formal oxidation states above III. Furthermore, all the redox potentials discussed in this review are reported versus the Normal Hydrogen Electrode (NHE). The same PCET phenomenon also occurs with other transition metal complexes and thus the principles described here are also valid for those complexes. Indeed nature also uses the same trick in a number of oxidative metalloproteins9 as well as in the OEC-PSII case where each of the four electron removals in the Kok cycle is accompanied by proton removal.5

In this respect it is vital to understand how electronic perturbations exerted by the ligands can influence the electron density at the metal center and thus its redox potentials.10,11 An enlightening example is described in a series of Ru–aquo complexes containing the trispyrazolyl tridentate facial ligand L4 and substituted 2,2′-bipyridine (bpy) of general formula [RuII(L4)(X-L2)(H2O)]2+ (X = H, 2; X = NO2, 3; X = NH2, 4), where the addition of electron donating groups produces an increase in electron density at the metal center whereas the electron withdrawing groups do just the opposite. Thus the former produce a decrease of the redox potential whereas the latter increase them. These trends are illustrated in Fig. 2 where a plot of the IV/III and III/II redox couples for complexes 2–4 is offered. Interestingly, the influence at the two different redox potentials is not of the same magnitude. Indeed, as can be seen in Fig. 2, while the III/II redox couple is strongly influenced, the IV/III redox couple is much less affected. As such this means that with sufficiently strong electron withdrawing groups attached to the bpy ligands we should be able to generate complexes where a single two-electron process should occur.


image file: c7cs00248c-f2.tif
Fig. 2 Graph of redox potentials for the IV/III and III/II couples in V vs. NHE at pH = 7.0 for complexes [RuII(L4)(H-L2)(H2O)]2+, 2; [RuII(L4)(NO2-L2)(H2O)]2+, 3 and [RuII(L4)(NH2-L2)(H2O)]2+, 4. The dark green number corresponds to ΔE1/2 = E1/2(IV/III) − E1/2(III/II) for the three complexes. Data extracted from ref. 11.

An additional important aspect to be understood is how ligand σ-perturbation and π-perturbation influence the redox potentials of the different redox couples. In this sense it is enlightening to observe the plot of ΔE1/2 = E1/2(IV/III) − E1/2(III/II) against the Lever parameter12 as indicated in Fig. 3. As can be observed here both the sigma-donating ligands and pi-accepting ligands align but at two different zones of the diagram, suggesting that the right combination of σ-perturbation and π-perturbation should lead to access to the ΔE1/2 < 0 zone. Indeed, the right combination of pyridyl and carbene ligands in [RuII(L5)(L6)(H2O)]2+, 14, allows us to obtain RuIV[double bond, length as m-dash]O complexes that behave as two-electron acceptors. This is nicely applied to the epoxidation of cis-alkenes to their corresponding cis-epoxides by 14, where the two-electron transfer favors the concerted mechanism and in turn favors the stereoselectivity of the reaction.11


image file: c7cs00248c-f3.tif
Fig. 3 Plot of ΔE1/2 = E1/2(IV/III) − E1/2(III/II) vs. ΣL for selected Ru–aquo complexes: [RuII(NH3)5(H2O)]2+, 5; [RuII(L1)(L9)(H2O)]+, 6; [RuII(L1)(L10)(H2O)]+, 7; [RuII(L1)(L11)(H2O)]2+, 8; [RuII(L1)(L7)(H2O)]2+, 9; [RuII(L4)(NO2-L2)(H2O)]2+, 10; [RuII(H-L2)2(L8)(H2O)]2+, 11; [RuII(L1)(L12)(H2O)]2+, 12; [RuII(L5)(H-L2)(H2O)]2+, 13; [RuII(L5)(L6)(H2O)]2+, 14. Data extracted from ref. 11 and 12. See the text for details.

For Ru–aquo complexes containing a neutral polypyridyl type of ligand such as in 1, there is an additional oxidation redox process that is a simple outer sphere electron transfer (OSET) process with no proton coupling that corresponds to the V/IV redox couple, as shown in the Pourbaix diagram in Fig. 1. In a number of Ru–aquo complexes containing pyridyl ligands, this high oxidation state is responsible for a suite of events that end up making dioxygen.13,14 It is thus very important to understand how electronic perturbation on ligands can influence the redox potential of the RuV/RuIV couple.

An example that illustrates this ligand effect is the comparison of the redox potentials of two Ru complexes containing the bpy-dicarboxylato ligand L132−, [RuII(L13)(MeO-L14)], 15, and [RuII(L13)(L15)]+, 16.15 For 15, containing an axial isoquinoline ligand with a strong electron donating group (MeO-L14), the potentials appear at: V/IV = 0.69 V, IV/III = 0.48 V and III/II = 0.01 V, whereas for 16, containing a methylated pyridyl group, L15, the potentials appear at: V/IV = 0.70 V, IV/III = 0.48 V and III/II = 0.18 V. Comparing these values clearly shows that the remote electronic perturbation through the π-system strongly affects the III/II potential but affects very little the IV/III and V/IV potentials. This is frustrating from a catalytic perspective because the latter couple is the most important one, associated with the overpotential of the electrocatalysis. This is due to the fact that the higher the oxidation state the lower the participation of the axial pyridyl auxiliary ligands in the HOMO/LUMO will be and thus this remote influence will be negligible. In other words, above the oxidation state III for Ru, the removal of electron density to generate high oxidation states will be predominantly occurring from the oxygen of the Ru–O group.

Another strategy to generate reactive high oxidation state species involves the use of sufficiently strong electron withdrawing groups so that the Ru(IV)[double bond, length as m-dash]O species has a IV/III redox potential higher than that of the water oxidation thermodynamic value and thus facilitating water oxidation. This is the case of complex [RuII(L1)(L16)(H2O)], 17, that contains a carboxylato moiety as the electron withdrawing group and whose IV/III couple is at 0.90 V at pH = 7. Indeed, this complex at oxidation state IV oxidizes water to dioxygen although it occurs very slowly.16 Another option that manifests the richness of the reactivity of Ru complexes is the capacity of Ru(IV) species to disproportionate into Ru(III) and its highly active Ru(V) counterparts. The latter, as indicated above, is a highly active species that can foster the formation of dioxygen. This has been proposed to occur in [Ru(L1)(H-L17)(H2O)]2+, 18.17

3. O–O bond formation paths

From a phenomenological perspective the formation of an oxygen–oxygen bond can occur via the so-called water nucleophilic attack mechanism (WNA) or by interaction of two M–O units (I2M) which are illustrated in Fig. 4 for the case of Ru–Hbpp, the complex containing the dinucleating ligand bpp (L17), in,in-{[RuII(L1)(H2O)]2(μ-L17)}3+, 19.18 In this particular case the bpp ligand acts as a scaffold that preorganizes and properly places the two RuII–OH2 groups so that after losing 4H+ and 4e, the corresponding RuIV[double bond, length as m-dash]O groups can form the O–O bond, which in this way is entropically favored. In this case, all the experimental evidence points toward the I2M mechanism. The mechanistic elucidation includes the use of spectroscopic techniques, kinetic analysis and reactivity tests. Particularly useful are the labeling experiments where the Ru–aquo complexes and the solvent H2O contain 18O with a different degree of labeling. This allows us to differentiate unambiguously between the two pathways by the use of mass spectrometry, just checking the molecular weight of the O2 generated in the first turnover as shown in Fig. 4. This strategy has been carried out in a number of occasions and produces extremely valuable information from a mechanistic perspective.19 The energy difference between those two mechanisms calculated using Density Functional Theory (DFT) is of 9.1 kcal mol−1 and is basically dictated by the reactivity of the Ru–O groups. In all cases studied the energy differences are large enough so that there is no mixing of the two mechanisms but there is no reason why a particular combination of auxiliary ligands could generate a reactive species that could go both ways.
image file: c7cs00248c-f4.tif
Fig. 4 Two potential water oxidation reaction mechanisms involving WNA and I2M for complex in,in-{[RuII(L1)(H2O)]2(μ-L17)}3+, 19, based on experimental evidence and supported using DFT calculations. Energy values in kcal mol−1. The L1 ligands are not drawn for clarity. Red oxygens correspond to the 18O initially labeled catalyst whereas blue oxygens refer to the initially non-labeled solvent H216O. Data extracted from ref. 17.

Furthermore, in this mechanism it is important to realize the importance of solvent H2O as a proton acceptor in the O–O bond formation step where a Ru–OOH intermediate is formed. Indeed, a DFT analysis shows how a gas phase study in the absence of solvent water changes the relative energies of the two pathways favoring the WNA mechanism.20 The influence of the base in the O–O bond formation step has also been clearly put forward in elegant studies using different external bases. In this way, depending on the strength of the base, the rate constant increases by 5 orders of magnitude from water to HO.21 In a recent work a spectacular effect on kinetics has also been shown when the base used was judiciously situated in the ligand framework so that it works in an intramolecular manner.22

A change from intramolecular I2M to intermolecular I2M occurs when the meridional trpy, L1 ligand in 19, is replaced by the facial tridentate ligands L4 or L18. This produces a change in the relative disposition of the two Ru–aquo groups within the complex from in,in to trans, as in trans-{[RuII(L18)(H2O)]2(μ-L17)}3+, 20, and thus constitutes an example of how ligand design can strongly influence reactivity.23

A remarkable catalyst whose O–O bond formation occurs by an intermolecular I2M mechanism is the Ru–bda complex, [RuII(L13)(Me-L3)], 21.24 This catalyst contains a flexible ligand framework that adapts to the electronic demands of the different oxidation states. Thus at oxidation state II the complex has a coordination number of 6 in a pseudo-octahedral geometry but at oxidation states IV and V it becomes 7-coordinate adopting a pentagonal bipyramidal geometry. Interestingly under restricted mobility conditions the complex cannot dimerize but still can carry out the oxidation of water to dioxygen by shifting to higher energy pathways based on the WNA mechanism. This, however, produces catastrophic consequences with regard to the ligand stability as will be discussed in the next section.25 Interestingly a recently reported cyclic trimer based on the Ru–bda complex, [RuII(L13)(L21)]3, 22, generates a very robust structure that also changes the mechanism from I2M to WNA and maintains the stability of the ligand framework.26 Similarly this is achieved with complex {[RuII(H-L3)2(H2O)]2(μ-L20)}3+, 23, that has a very similar arrangement as in in,in-{[RuII(L1)(H2O)]2(μ-L17)}3+, 19, but the extra pyridyl groups at L20 with regard to L17 place the two metal centers sufficiently apart from one another, such that the mechanism also changes from I2M to WNA.19

4. Redox non-innocent ligands

Transition metal complexes that undergo redox events at the ligand have been flourishing lately after the pioneering work of Wieghardt and others.27 For the water oxidation reaction that involves multiple electron and proton removal, the option of sharing the burden to reach high oxidation states between the metal and the ligand is certainly an attractive concept.

A Ru–aquo catechol complex reported in 2009, [RuII(L1)(L22)(H2O)], 24, has been shown to access a number of different oxidation states based on pulse radiolysis coupled to electrochemical experiments and supported by DFT calculations.28 It is shown that by progressive electron removal there is an interplay of metal and ligand oxidation, identifying the corresponding Ru semiquinone and quinone complexes as a function of pH. Recently an anionic Cu(II)–amidate complex, [Cu(L23)]2−, 25, has been reported where after the initial metal based III/II oxidation it suffers an additional oxidation at the ligand basically forming a phenyl radical cation. While the latter is a reversible process in dry organic solvents it is responsible for a large electrocatalytic wave in the presence of water, due to the oxidation of the latter to dioxygen. Remarkably this ligand oxidation event is chemically reversible thanks to the fast catalytic reaction.29 Furthermore, the use of electron donating groups at the aromatic ring allows us to strongly fine-tune the electronic properties of this radical cation. Using a dimethoxy substitute allows the achievement of a record low overpotential for first row transition metal based water oxidation catalysts of 170 mV at pH = 11.5.

These examples show how the right combination of metal and ligand based oxidation can exert extraordinary electronic control over the different redox processes involved in the water oxidation process and particularly and remarkably the most important process that is responsible for the rate determining step.

5. Intermediate characterization

The characterization of reaction intermediates in any catalytic cycle is one of the fundamental requirements for the understanding of the reaction mechanisms. For many Ru complexes electrochemical and spectroscopic experiments allow us to characterize several intermediates as is the case for instance of complex 1, whose Pourbaix diagram is shown in Fig. 1. However, the most interesting intermediate species, RuOOH, still remains ill characterized although some attempts have been carried out recently for [RuII(L19)(L24)(H2O)]2+, 26,30 and [(L1)(H-L2)RuIII–O–RuIII(L1)(O)(H2O)]4+, 27.31 In this respect the best technique for the characterization of a peroxide and/or a superoxide is undoubtedly vibrational spectroscopy through infrared and/or Raman spectroscopy using labeled oxygen atoms to identify the O–O bond vibration. For the case of the Ru complexes this vibration is difficult to characterize because it is very weak and shows unexpectedly low isotopic shifts with 18O. According to DFT the weakness of these signals is due to the low Raman activity of the vibrational modes involved and the small shifts suggest a distribution of the O–O stretching coordinates among several modes.30,31 In sharp contrast, for the Co–Hbpp water oxidation catalyst, {[CoII(L1)(H2O)]2(μ-L17)}3+, 28,32 both the peroxo and superoxo complexes before oxygen generation have been beautifully identified by resonance Raman (rR) spectroscopy (see Fig. 5) and also by a number of additional spectroscopic techniques including Nuclear Magnetic Resonance (NMR), Electron Paramagnetic Resonance (EPR), Cyclic Voltammetry (CV), and X-ray Absorption Spectroscopy (XAS) and further complemented using DFT calculations.
image file: c7cs00248c-f5.tif
Fig. 5 Normalized rR spectra for side-on peroxido [(L1)CoIII-(μ-OO)(μ-L17)-CoIII(L1)]3+, and end-on superoxido [(L1)(OO)CoIII-(μ-L17)-CoIII(L1)(H2O)]4+ complexes derived from 28, with labeled 16O, 17O and 18O oxygen atoms. Left, black line, peroxo-32 (839 cm−1); blue line, peroxo-34 (814 cm−1), red line, peroxo-36 (795 cm−1). Right, black line, superoxo-32 (1121 cm−1); blue line, superoxo-34 (1088 cm−1), red line, superoxo-36 (1055 cm−1). The inset in the graphs shows the drawn structure of the complexes omitting the trpy ligands L1. Data extracted from ref. 32; asterisks indicate vibrations from solvent.

Labelling experiments also give crucial evidence especially with regard to the O–O bond formation mechanism as has been described above for Ru complexes. One of the key aspects to be able to extract meaningful information from these types of experiments is the substitution kinetics of the bonded Ru aquo or hydroxido groups by solvent water. For Ru complexes this process has been successfully slowed down upon reaching high oxidation states.18 For first row transition metal complexes there is only one example, Ru2Co, {[(L1)(H2O)Ru]2[μ-Co(H2O)2(L17)2]}4+, 29,33 where these types of experiments have been performed. In this case the labeling experiments carried out together with the rest of the electrochemical and spectroscopic properties are in agreement with the presence of a water nucleophilic WNA mechanism occurring at the Co site of the trinuclear complex, with cooperative interaction of the two Ru sites via electronic coupling through the bpp (L17) bridging ligand and via neighboring hydrogen bonding.

6. Main deactivation pathways

As indicated in eqn (3), the thermodynamic potential at which water oxidation occurs is 1.23 V, which is very high. Thus any catalyst that can oxidize water will oxidize a whole variety of organic substrates. Therefore, it is imperative to use complexes containing ligands that are not easily oxidized or that are oxidized at a sufficiently slow rate so that its degradation is negligible. It is only under these conditions that meaningful information can be extracted, otherwise the main information extracted is related to the ligand oxidation and/or oxide nanoparticle formation, rather than water oxidation by the initial molecular complex. Under these unavoidable conditions it is extremely important to check the nature of the molecular catalyst during and after the catalytic reaction. This is particularly important since the corresponding metal oxides, of the used transition metal complexes, are also active water oxidation catalysts.34

A few examples of deactivation pathways that occur in Mol-WOCs are summarized in Scheme 1 and will be discussed below.


image file: c7cs00248c-s1.tif
Scheme 1 Main deactivation pathways of water oxidation catalysts.

Among the most easily oxidizable organic substrates are those containing methylenic groups in benzyls, benzyl alcohol and benzyl pyridyls (e.g. 1 in Scheme 1). Some complexes have been prepared using these groups but a careful analysis of their behavior clearly shows their degradation. Some of the many examples reported include: {[RuII(L1)(H2O)]2(μ-D1)}3+, 30,35 {[RuII(H2O)]2(μ-D2)}3+, 31,36 and [RuII(D3)(H2O)]2+, 32.37 In these complexes a careful analysis simultaneously measuring the amount of O2 and CO2 generated shows the formation of the latter from the very beginning. This implies that massive ligand degradation occurs together with the formation of O2 and thus points to the formation of RuO2 as the active species rather than the initial molecular complex. Importantly, the [CO2] should be measured simultaneously with the generation of [O2], since otherwise a simple measurement at the end of the reaction will contain a very low [CO2] as compared to that of the [O2] generated, especially if the corresponding oxide is relatively active.

Another paradigmatic error that appears constantly in the literature is to show a totally chemically irreversible redox wave in the voltammogram for a particular complex and claim that the initial complex is intact after bulk electrolysis is carried out above the irreversible wave's potential. This was reported for the dinuclear Co complex {[CoIII(D4)]2(μ-OH)(μ-OO)}3+, 33,38 and it was later shown that the only active species was CoOx.39 Other related contradictory examples include [MnII(D5)(H2O)2]2+, 34,40 and recently [FeIII(D6)(H2O)], 35.41 In the latter case the chemical irreversibility of the waves is most likely due to ligand dissociation and oxide formation rather than ligand degradation since amides are relatively stable with regard to oxidative degradation.

In most of the Ru–aquo polypyridyl complexes described as water oxidation catalysts, the ligands bonded to the metal center are kinetically inert except for the aquo group where most of the chemistry is occurring. On the other hand, first row transition metal complexes based on Mn, Fe, Co or Ni contain ligands that are kinetically labile, and depending on the pH there will always be significant competition between the ligand and solvent water for the coordination site of the metal center, as exemplified in the following equations for [FeII(D8)(H2O)2], 36,42

 
[Fe(D8-κ-N4)(H2O)2]2+ + H2O → [Fe(D8-κ-N3)(H2O)3]2+(5)
 
[Fe(D8-κ-N3)(H2O)3]2+ + H2O → [Fe(D8-κ-N2)(H2O)4]2+(6)
 
[Fe(D8-κ-N2)(H2O)4]2+ + H2O → [Fe(D8-κ-N1)(H2O)5]2+(7)
 
[Fe(D8-κ-N1)(H2O)5]2+ + H2O → [Fe(H2O)6]2+ + D8(8)
This partial decoordination can also be a source of potential ligand oxidation leading to catalyst degradation and accelerate oxide formation, as may occur in Ni complexes containing a tetraazamacrocyclic type of ligand such as [NiII(D7)]2+, 37,43 and in Fe complexes like [FeII(D9)(H2O)2]2+, 38.44–46 A recent example that most likely involves ligand substitution has been reported for a Co porphyrin complex [CoII(D10)], 39 (2, Scheme 1). This complex was initially described to act in a molecular manner but recently it has been shown by Synchrotron-based Soft X-ray Photoelectron Spectroscopy (SOXPES) and Hard X-ray Photoelectron Spectroscopy (HAXPES) that the active species is again CoOx.47

In some cases the ligand degradation can partially occur as is the case of the Ir complex [Ir(D11)(D12)Cl], 40,48 where after some controversy now the active species is proposed to be a dinuclear Ir–(O)2–Ir where the D11 ligand is gone but D12 is still coordinated (3, Scheme 1). Ligand degradation as indicated previously can occur all the way to CO2, thus generating the corresponding metal oxide as the active species. This active species could be part of a large aggregate typical of metal oxides generating large nanoparticles that can be detected by Dynamic Light Scattering (DLS).46 However, they can also generate the so-called “molecular oxides”, as depicted in Fig. 6, which could be a combination of relatively low molecular weight oxides as described in the case of CoOx (also labeled as CoPi; this label is unfortunate since CoPi does not contain P atoms in its structure) that will not be detectable by DLS.49 Additionally in acidic solutions discrete types of species like ferrates [H3FeVIO4]+ and diferrates [H4FeVI2O7]2+ are also known to oxidize water to dioxygen and will not be detected by DLS.50


image file: c7cs00248c-f6.tif
Fig. 6 Drawn structures of molecular cobalt oxides obtained from the XAS data for CoPi. Data extracted from ref. 49. Color codes: blue, cobalt; red, bridging oxygen; pink, terminal oxygen.

A particularly interesting example of ligand degradation is the case of the Ru–bda complex. [RuII(L13)(D13)2]2+, 41, containing the D13 axial ligands that upon reduction liberates N2 gas and anchors at graphitic surfaces making a C–C bond.25 This is an extremely robust water oxidation catalyst in the homogeneous phase that forms O–O bonds in a bimolecular manner via an I2M mechanism. Attached to the surface of a graphitic electrode with restricted mobility it cannot undergo dimerization. This results in the need for accessing high energy pathways that in parallel facilitate ligand degradation. As a result of this after a few catalytic cycles the only Ru species left at the electrode is RuO2 (4, Scheme 1).25

Finally, during the catalytic process side reactions can occur that can lead to the suppression of the catalyst activity due to the generation of inactive species but also could generate new robust species that might be more active than the initial catalyst. This is the case of [Ru(L1)(H-L2)(H2O)]2+, 1,51 that slowly but irreversibly loses one of the bpy ligands generating a dinuclear Ru–O–Ru complex, [(L1)(H-L2)RuIII–O–RuIII(L1)(O)(H2O)]4+, 27, that is actually a better catalyst and that shows no signs of degradation after a turnover number (TON) of more than 14[thin space (1/6-em)]900 cycles (5, Scheme 1).52

7. Conclusions

Molecular water oxidation catalysis is nowadays one of the most exciting challenges in coordination chemistry that if successful can provide the world with environmentally respectful energy conversion schemes based on -WS.

The design of Mol-WOCs encompasses the preparation of transition metal complexes that contain aquo ligands bonded to the metal center allowing reactive high oxidation states to be reached at relatively low potential by PCET. Importantly this can also be achieved by the complementary involvement of redox active ligands. Once an M–O reactive species is reached the next crucial step is the O–O bond formation that can occur via WNA or I2M. We have shown in the previous sections the important consequences this can have when the catalysts are anchored on solid supports that in turn are necessary to simplify potential water splitting device engineering. This mechanistic understanding, based on multiple spectroscopic, electrochemical and reactivity studies, is of paramount importance in order to come up with a bottom up approach for the design of efficient WOCs.

For Mol-WOCs one of the main challenges is the intrinsically high thermodynamic redox potential that needs to be reached for the water oxidation reaction. Therefore, a Mol-WOC capable of oxidizing water will oxidize a large number of organic groups. It is thus extremely important that the ligands used to bind the metal center are kinetically robust with regard to oxidation, otherwise the ligand will degrade. This will lead to oxides being responsible for water oxidation catalysis. When this happens no meaningful information can be extracted from the system and thus it becomes useless. It is surprising and excruciating to see the large amount of wrong examples published recently in the literature. Fortunately, if a judicious and careful choice of ligands is made for the preparation of WOCs, then useful information can be obtained that allows us to follow a bottom up approach. In this manner spectacular catalysts have been prepared over the last ten years that have significantly improved our knowledge in this field. For instance, the turnover frequencies (TOFs) reported over the last ten years have increased by more than 7 orders of magnitude. Indeed, we now have Mol-WOCs that operate at TOF-max that are two orders of magnitude better than that of nature in the OEC-PSII. Also the oxidative ruggedness of the ligands has been extraordinarily improved with systems achieving turnover numbers in the range of millions.

For practical applications, system performance still needs to be improved. However, with the present knowledge, succinctly described in this review, we have most of the elements needed that when judiciously used should drive us to the desired improvement in the near future, which will lead to the engineering of useful devices.

Acknowledgements

Sustained support over the last decade from MINECO and FEDER is gratefully acknowledged. The latest grants awarded are: CTQ2016-80058-R, CTQ2015-73028-EXP, CTQ2015-64261-R, SEV 2013-0319, ENE2016-82025-REDT (FOTOFUEL), and CTQ2016-81923-REDC (INTECAT). R. M. and P. G.-B. acknowledge the “La Caixa” foundation for the PhD grant.

References

  1. F. Berna, P. Goldberg, L. K. Horwitz, J. Brink, S. Holt, M. Bamford and M. Chazan, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, E1215–E1220 CrossRef CAS PubMed .
  2. World Energy Outlook 2016, International Energy Agency, 2016.
  3. “A Milestone for Carbon Dioxide in the Atmosphere”, Schwartz, J. The New York Times, 2016. http://www.nytimes.com/2016/10/04/science/atmospheric-carbon-dioxide-400-ppm.html.
  4. A. Llobet and F. Meyer, Angew. Chem., Int. Ed., 2011, 50, A30–A33 CrossRef PubMed .
  5. I. D. Young, M. Ibrahim, R. Chatterjee, S. Gul, F. D. Fuller, S. Koroidov, A. S. Brewster, R. Tran, R. Alonso-Mori, T. Kroll, T. Michels-Clark, H. Laksmono, R. G. Sierra, C. A. Stan, R. Hussein, M. Zhang, L. Douthit, M. Kubin, C. de Lichtenberg, L. Vo Pham, H. Nilsson, M. H. Cheah, D. Shevela, C. Saracini, M. A. Bean, I. Seuffert, D. Sokaras, T.-C. Weng, E. Pastor, C. Weninger, T. Fransson, L. Lassalle, P. Brauer, P. Aller, P. T. Docker, B. Andi, A. M. Orville, J. M. Glownia, S. Nelson, M. Sikorski, D. Zhu, M. S. Hunter, T. J. Lane, A. Aquila, J. E. Koglin, J. Robinson, M. Liang, S. Boutet, A. Y. Lyubimov, M. Uervirojnangkoorn, N. W. Moriarty, D. Liebschner, P. V. Afonine, D. G. Waterman, G. Evans, P. Wernet, H. Dobbek, W. I. Weis, A. T. Brunger, P. H. Zwart, P. D. Adams, A. Zouni, J. Messinger, U. Bergmann, N. K. Sauter, J. Kern, V. K. Yachandra and J. Yano, Nature, 2016, 540, 453–457 CrossRef CAS PubMed .
  6. A. Coehn and M. Gläser, Z. Anorg. Chem., 1902, 33, 9–24 CrossRef CAS .
  7. S. W. Gersten, G. J. Samuels and T. J. Meyer, J. Am. Chem. Soc., 1982, 104, 4029 CrossRef CAS .
  8. X. Sala, S. Maji, R. Bofill, J. Garcia-Anton, L. Escriche and A. Llobet, Acc. Chem. Res., 2014, 47, 504–516 CrossRef CAS PubMed  , and references therein.
  9. D. R. Weinberg, C. J. Gagliardi, J. F. Hull, C. F. Murphy, C. A. Kent, B. C. Westlake, A. Paul, D. H. Ess, D. G. McCafferty and T. J. Meyer, Chem. Rev., 2012, 112, 4016–4093 CrossRef CAS PubMed .
  10. A. Llobet, Inorg. Chim. Acta, 1994, 221, 125–131 CrossRef CAS .
  11. E. Masllorens, M. Rodriguez, I. Romero, A. Roglans, T. Parella, J. Benet-Buchholz, M. Poyatos and A. Llobet, J. Am. Chem. Soc., 2006, 128, 5306–5307 CrossRef CAS PubMed .
  12. A. Dovletoglou, S. A. Adeyemi and T. J. Meyer, Inorg. Chem., 1996, 35, 4120–4127 CrossRef CAS PubMed .
  13. L. Vigara, M. Z. Ertem, N. Planas, F. Bozoglian, N. Leidel, H. Dau, M. Haumann, L. Gagliardi, C. J. Cramer and A. Llobet, Chem. Sci., 2012, 3, 2576–2586 RSC  , and references therein.
  14. M. Murakami, D. Hong, T. Suenobu, S. Yamaguchi, T. Ogura and S. Fukuzumi, J. Am. Chem. Soc., 2011, 133, 11605–11613 CrossRef CAS PubMed .
  15. C. J. Richmond, R. Matheu, A. Poater, L. Falivene, J. Benet-Buchholz, X. Sala, L. Cavallo and A. Llobet, Chem. – Eur. J., 2014, 20, 17282–17286 CrossRef CAS PubMed .
  16. L. Mognon, J. Benet-Buchholz and A. Llobet, Inorg. Chem., 2015, 54, 11948–11957 CrossRef CAS PubMed .
  17. S. Roeser, P. Farras, F. Bozoglian, M. Martínez-Belmonte, J. Benet-Buchholz and A. Llobet, ChemSusChem, 2011, 4, 153 CrossRef .
  18. S. Romain, F. Bozoglian, X. Sala and A. Llobet, J. Am. Chem. Soc., 2009, 131, 2768–2769 CrossRef CAS PubMed .
  19. S. Neudeck, S. Maji, I. Lopez, S. Meyer, F. Meyer and A. Llobet, J. Am. Chem. Soc., 2014, 136, 24–27 CrossRef CAS PubMed .
  20. X. Yang and M.-H. Baik, J. Am. Chem. Soc., 2008, 130, 16231–16240 CrossRef CAS PubMed .
  21. N. Song, J. J. Concepcion, R. A. Binstead, J. A. Rudd, A. K. Vannucci, C. J. Dares, M. K. Coggins and T. J. Meyer, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 4935–4940 CrossRef CAS PubMed .
  22. R. Matheu, M. Z. Ertem, J. Benet-Buchholz, E. Coronado, V. S. Batista, X. Sala and A. Llobet, J. Am. Chem. Soc., 2015, 137, 10786–10795 CrossRef CAS PubMed .
  23. S. Maji, L. Vigara, F. Cottone, F. Bozoglian, J. Benet-Buchholz and A. Llobet, Angew. Chem., Int. Ed., 2012, 51, 5967–5970 CrossRef CAS PubMed .
  24. L. Duan, F. Bozoglian, S. Mandal, B. Stewart, T. Privalov, A. Llobet and L. Sun, Nat. Chem., 2012, 4, 418–423 CrossRef CAS PubMed .
  25. R. Matheu, L. Francàs, P. Chernev, M. Z. Ertem, V. Batista, M. Haumann, X. Sala and A. Llobet, ACS Catal., 2015, 5, 3422–3429 CrossRef CAS .
  26. M. Schulze, V. Kunz, P. D. Frischmann and F. Würthner, Nat. Chem., 2016, 8, 576–583 CrossRef CAS PubMed .
  27. See for instance, papers contained in the “Forum on the Redox-Active Ligand” in, P. J. Chirik, Inorg. Chem., 2011, 50, 9737–9740 CrossRef CAS PubMed .
  28. M.-K. Tsai, J. Rochford, D. E. Polyansky, T. Wada, K. Tanaka, E. Fujita and J. T. Muckerman, Inorg. Chem., 2009, 48, 4372–4383 CrossRef CAS PubMed .
  29. P. Garrido-Barros, I. Funes-Ardoiz, S. Drouet, J. Benet-Buchholz, F. Maseras and A. Llobet, J. Am. Chem. Soc., 2015, 137, 6758–6761 CrossRef CAS PubMed .
  30. J. J. Concepcion, M.-K. Tsai, J. T. Muckerman and T. J. Meyer, J. Am. Chem. Soc., 2010, 132, 1545–1557 CrossRef CAS PubMed .
  31. A. Keidel, I. López, J. Staffa, U. Kuhlmann, F. Bozoglian, C. Gimbert-Suriñach, J. Benet-Buchholz, P. Hildebrandt and A. Llobet, ChemSusChem, 2017, 10, 551–561 CrossRef CAS PubMed .
  32. C. Gimbert-Suriñach, D. Moonshiram, L. Francas, N. Planas, V. Bernales, F. Bozoglian, A. Guda, L. Mognon, I. López, Md. A. Hoque, L. Gagliardi, C. J. Cramer and A. Llobet, J. Am. Chem. Soc., 2016, 138, 15291–15294 CrossRef PubMed .
  33. L. Mognon, S. Mandal, C. E. Castillo, J. Fortage, F. Molton, G. Aromi, J. Benet-Buchholz, M.-N. Collomb and A. Llobet, Chem. Sci., 2016, 7, 3304–3312 RSC .
  34. A. Harriman, I. J. Pickering, J. M. Thomas and P. A. Christensen, J. Chem. Soc., Faraday Trans., 1988, 84, 2795–2806 RSC .
  35. L. Francàs, X. Sala, J. Benet-Buchholz, L. Escriche and A. Llobet, ChemSusChem, 2009, 2, 321–329 CrossRef PubMed .
  36. A. C. Sander, A. Schober, S. Dechert and F. Meyer, Eur. J. Inorg. Chem., 2015, 4348–4353 CrossRef CAS .
  37. B. Radaram, J. A. Ivie, W. M. Singh, R. M. Grudzient, J. H. Reibenspies, C. E. Webster and X. Zhao, Inorg. Chem., 2011, 50, 10564–10571 CrossRef CAS PubMed .
  38. H.-Y. Wang, E. Mijangos, S. Ott and A. Thapper, Angew. Chem., Int. Ed., 2014, 53, 14499–14502 CrossRef CAS PubMed .
  39. J.-W. Wang, P. Sahoo and T.-B. Lu, ACS Catal., 2016, 6, 5062–5068 CrossRef CAS .
  40. W.-T. Lee, S. B. Muñoz III, D. A. Dickie and J. M. Smith, Angew. Chem., Int. Ed., 2014, 53, 9856–9859 CrossRef CAS PubMed .
  41. S. Pattanayak, D. R. Chowdhury, B. Garai, K. K. Singh, A. Paul, B. B. Dhar and S. S. Gupta, Chem. – Eur. J., 2017, 23, 3414–3424 CrossRef CAS PubMed .
  42. D. Hong, S. Mandal, Y. Yamada, Y.-M. Lee, W. Nam, A. Llobet and S. Fukuzumi, Inorg. Chem., 2013, 52, 9522–9531 CrossRef CAS PubMed .
  43. M. Zhang, M.-T. Zhang, C. Hou, Z.-F. Ke and T.-B. Lu, Angew. Chem., Int. Ed., 2014, 53, 13042–13048 CrossRef CAS PubMed .
  44. J. Lloret-Fillol, Z. Codola, I. Garcia-Bosch, L. Gomez, J. J. Pla and M. Costas, Nat. Chem., 2011, 3, 807–813 CrossRef PubMed .
  45. W. A. Hoffert, M. T. Mock, A. M. Appel and J. Y. Yang, Eur. J. Inorg. Chem., 2013, 3846–3857 CrossRef CAS .
  46. G. Chen, L. Chen, S.-M. Ng, W.-L. Man and T.-C. Lau, Angew. Chem., Int. Ed., 2013, 52, 1789–1791 CrossRef CAS PubMed .
  47. Q. Daniel, R. B. Ambre, B. Zhang, B. Philippe, H. Chen, F. Li, K. Fan, S. Ahmadi, H. Rensmo and L. Sun, ACS Catal., 2017, 7, 1143–1149 CrossRef CAS .
  48. S. W. Sheelan, J. M. Thomsen, U. Hintermair, R. H. Crabtree, G. W. Brudvig and C. A. Schmuttenmaer, Nat. Commun., 2015, 6, 6469 CrossRef PubMed .
  49. M. W. Kanan, J. Yano, Y. Surendranath, M. Dinca, V. K. Yachandra and D. G. Nocera, J. Am. Chem. Soc., 2010, 132, 13692–13701 CrossRef CAS PubMed .
  50. R. Sarma, A. M. Angeles-Boza, D. W. Brinkley and J. P. Roth, J. Am. Chem. Soc., 2012, 134, 15371–15386 CrossRef CAS PubMed .
  51. I. López, M. Z. Ertem, S. Maji, J. Benet-Buchholz, A. Keidel, U. Kuhlmann, P. Hildebrandt, C. J. Cramer, V. S. Batista and A. Llobet, Angew. Chem., Int. Ed., 2014, 53, 205–209 CrossRef PubMed .
  52. Catalyst performances are usually characterized in terms of turnover numbers (TONs) and turnover frequencies (TOFs). Please see any general textbook in catalysis such as “Masel, Richard I. (2001) Chemical Kinetics and Catalysis. Wiley-Interscience, New York. ISBN 0-471-24197-0”.

This journal is © The Royal Society of Chemistry 2017