1. Mechanistic Details and Acid Stability of Platinum Group Metal Oxides

There is already a large body of literature on the mechanism of OER electrocatalysis, particularly over binary iridium oxides (2, 3, 4). The process is a complex heterogeneous reaction involving interaction of water with a solid surface under applied potential resulting in evolution of a gas. A number of models have been developed to propose how OER might be broken into fundamental steps to describe the overall phenomenon (5). The purpose of this section is to summarise briefly some of the key, general conclusions of the prior work and then consider the more recent results concerning mechanism that have emerged from the new multinary phases we have introduced in the sections above. As will be appreciated, one important aspect of this work is consideration of the stability of the oxide electrocatalysis, since this is a primary concern when any material is implemented in a real device, where longevity of performance must be a priority (6).

The four electron OER process over a solid surface can be written as a set of one-electron reactions (2, 7, 8). This mechanism has been termed the adsorbates evolution mechanism (AEM), as illustrated in Figure 1. The generation of adsorbed hydroperoxide, *OOH, is one of the steps in this process, and this species contains oxygen in oxidation state –1.

For pgm-oxides the involvement of lattice oxide ions in OER has been reported by many groups, with experimental studies of materials such as IrO₂ and RuO₂ providing evidence for the exchange of water between lattice oxide and water (4, 8). For example, Schweinar et al. used isotope labelling and atom probe tomography to show that crystalline IrO₂ surfaces release lattice oxide, resulting in structural change at the surface with dissolution of metal cations and their partial redeposition (9). This mechanism has been termed the lattice-oxygen participation mechanism (LOM) and may be represented as shown in Figure 2.

Computational work has revealed that the LOM pathway offers lower kinetic barriers than AEM so provides the possibility of more effective electrocatalysis (10). Other computational work has shown that for perfect crystalline surfaces, the AEM pathway is preferred, but with the introduction of defects the LOM becomes favoured (11). The AEM and LOM may not be distinct situations but may occur simultaneously, to a greater or lesser extent, depending on the surface chemistry of the particular material used as a catalyst. In each case, a change in pgm oxidation state is implied during the OER process, and for IrO₂-based materials this may involve oxidation to Ir⁵⁺ or Ir⁶⁺ (12–16), or reduction to Ir³⁺ (17). For less crystalline iridium oxides, there may be a proportion of Ir³⁺ present before electrochemistry is performed (18). In fact these observations are not consistently made in the literature and the role of iridium redox in OER is not universally established: this may be in part because the experimental methods to probe oxidation state change are sensitive to only either the bulk or surface of sample, or have not been implemented in operando , but equally it may be the case that not all materials react by the same universal mechanism. It is highly possible that metal oxide surfaces prepared by different routes, with different crystallinity and degree of hydration undergo different changes in response to applied potential in an aqueous acid electrolyte. Although is it clear that surface binding of water must occur during OER, followed by redox chemistry associated with the pgm, another consideration to make is that involvement of electrons from the conduction band of metallic pgm-oxide may play a role in the catalysis. That is, rather than just the surface pgm cations undergoing oxidation-state change, the whole particle responds by adjustment of the Fermi level of the conduction band. This ‘band model’ as distinct to the ‘bond model’ was discussed by Hillman et al. , with experimental evidence provided using X-ray absorption spectroscopy of iridium oxide films (19), and some evidence was also found for its role in the OER activity of the pyrochlores (Na,Ca)_2−x Ir₂O₆·H₂O (13).

As well as the redox chemistry of the pgm, the involvement of oxygen redox in OER has been proven experimentally, with evidence for the formation of electrophilic O^–1 as a reactive intermediate before the formation of molecular O₂ (20). It has been proposed that the dynamic nature of the iridium framework in amorphous IrO_x imparts the flexibility in iridium oxidation state required for the formation of this active electrophilic oxygen (21). In the case of the LOM, however, the presence of vacant oxide sites leads to weakened binding of the metals, which induces the dissolution of the catalyst (22, 23). Evidence for increased iridium-iridium interactions upon lattice oxide removal has been proposed to account for the relative stability of IrO₂ (24). For RuO₂, the loss of lattice oxygen may occur more readily, leading to greater instability compared to IrO₂ (11), although oxidation to soluble RuO₄ is also an important contributory factor (25).

It is very clear from the work on IrO₂ and RuO₂, in both crystalline and amorphous, hydrated and anhydrous forms, that the surface chemistry is dynamic under acid conditions. The LOM pathway implies stress on the oxide surface that can contribute to amorphisation, dissolution and redeposition of the electrocatalyst (26), i.e. with operating potentials of OER results in reconstruction of the crystal structure. It is highly likely that the active surface layer of the electrocatalyst is amorphous once subjected to electrochemical conditions, and also becomes increasingly hydrous. Relating to this surface restructuring is the loss of surface species into solution, and although some may be redeposited, this dissolution can ultimately lead to loss of the active pgms and eventually failure of the electrocatalysts. Another resulting issue is the interference of any dissolved metal cations with other components of the device: for example interaction with the fluorocarbon polymer electrolyte may inhibit its proton conductivity (27). It is observed that the most active materials for OER electrocatalysis are also the least stable with respect to dissolution (28), and resolving this ‘activity–stability conundrum’ is an important goal in developing resilient materials with properties optimal for applications (29).

2. Measurement of Oxygen Evolution Performance of Platinum Group Metal Oxides

Before we examine the new OER electrocatalyst materials discovered in the last few years, it is important to consider how performance is measured and assessed. One initial observation is that different conditions may have been applied in the measurement of OER performance: as well as the pH of the electrolyte (usually aqueous HClO₄ or H₂SO₄) and the temperature of measurement, the composition of the catalyst layer may not be consistent. In some studies a powdered sample is mixed with carbon as a conducting binder or support, while in other cases an oxide is deposited as a layer or film directly onto a substrate. The surface area and particle size of the catalyst must be considered, and the applied electrochemical conditions (range of potential, sweep rate, number of cycles) must be taken into account, when comparing data from what notionally is the same catalyst. For example, the rate of dissolution of RuO₂ in acid electrolytes depends on its synthesis method, and can vary over several orders of magnitude (30). Notwithstanding these experimental issues, the metrics used to assess performance are not necessarily the same between different reported studies: assignment of a consistent ‘figure of merit’ is not apparent in the literature on OER electrocatalysts. Parameters used in the literature to assess performance of OER electrocatalysts, include the following (31): (a) the onset potential, the applied potential at which OER is detected; (b) the Tafel slope, which is rate of change of current density, giving information about kinetics of OER; (c) the overpotential, the potential to achieve a specified geometric current density, typically 10 mA cm^–2, but which may not reflect intrinsic activity as it does not consider loading or surface area; (d) the mass activity, determined by normalising the catalytic current by the catalyst loading, and may be expressed as mass of catalyst or mass of pgm, to reflect the cost efficiency; (e) Faradaic efficiency, obtained by relating the amount of evolved oxygen to the theoretical amount.

The ‘activity’ of a catalyst is often the most sought-after property, and it can be seen that this may be quoted in a number of ways, each of which may be dependent on the experimental variables mentioned above (electrolyte pH, mass loading of catalyst, surface area or choice of support). In terms of stability, a chronopotentiometric hold at constant current density for a chosen period of time, or repeated cyclic voltammetry with specific scan rates for thousands of cycles may provide appropriate measures (31). But, as already discussed, the dissolution of catalyst is a crucial factor. Gieger et al. proposed a ‘stability number’ as a benchmark for assessing stability. Their S– number is defined as the ratio between the amount of evolved oxygen and the amount of dissolved pgm (extracted from ICP-MS data) and this describes how many oxygen molecules are formed per pgm atom lost into the electrolyte (22).

3. Oxygen Evolution Activity of Complex Iridium and Ruthenium Oxides

OER performance data presented for some of the complex ruthenium and iridium oxides have recently been tabulated and compared in some review articles (32, 33, 34). Table I summarises some of the activity and stability values reported for a variety of the recently discovered ruthenate and iridate materials. For comparison, similar values for crystalline RuO₂ and IrO₂ are recorded, but it should be noted that while these data are representative, even for these simple binary solids there exists a range of metrics reported in the literature (35). In the case of IrO₂, as noted earlier, amorphous forms of the oxide can be considerably more active than crystalline forms. Despite these difficulties in making quantitative comparisons, it can be observed that many of the recently discovered mixed-metal oxides offer enhanced properties over the binary oxides, with low overpotential required to generate significant current densities, higher specific activities, smaller Tafel slope values, and with more favourable stability. The materials Cr_0.6Ru_0.4O₂, 6H-SrIrO₃ and CaCu₃Ru₄O₁₂ stand out in this respect. The stability number of Gieger et al. has not yet been applied by many authors, but in the case of the proton-exchanged form of the Ruddlesden-Popper phase H_3.6IrO₄·3.7H₂O (not included in Table I) the S-value of ~10⁵ was higher than for 6H-SrIrO₃ (~10⁴) and Sr₂Ir_0.5Fe_0.5O₄ (~10³) when studied under the same conditions, and similar to crystalline (10⁶) and amorphous forms of IrO₂ (10⁶) (48).

A note of caution should also be made around reported data from ‘wet cell’ electrochemical measurements, such as in classical rotating disc electrode (RDE) experiments, in that they do not necessarily replicate the activity measured under the real conditions in a device. Frydendal et al. pointed out that stability cannot be assessed by electrochemical measurements alone and that complementary methods must be used, in their case monitoring the mass of the electrocatalyst and the release of cations into solution (49). Alia et al. found that in tests on iridium oxides that RDEs kinetically underperformed MEAs by at least an order of magnitude and that durability of the catalyst in an MEA could not be estimated from the RDE performance (50). In our own work we observed similar conclusions from a study of substituted RuO₂ materials (51). Others have found that inaccuracies in RDE testing may be due to accumulation of gas bubbles, that lead to inconclusive results regarding stability of electrocatalysts (52, 53). Some authors have proposed standard protocols to allow benchmarking and also to assess suitability for use in real devices (54–56), but for the new materials we have focused on in this review, these rigorous testing protocols are yet to be applied.

Some general observations about the mechanistic information that has emerged in the literature on oxide OER catalysts are also appropriate to make. Early work by Trasatti correlated the overpotential for OER with the enthalpy associated with the transition from MO_x to MO_{x +1} (the ‘lower to higher oxide transition’) for a set of transition metal oxides and found a volcano plot with RuO₂ and IrO₂ close to the apex (i.e. with lowest overpotential for OER) (57). It was proposed that OER most readily takes place with low coverage of adsorbed intermediates, and so for oxides that are easily oxidised, there is high coverage of adsorbed intermediate and thus high overpotential. It can be concluded from this, and others’ work, that the metal-oxygen bond strength is an important factor in determining OER activity.

Although similar considerations have not yet been systematically made for the recently discovered novel complex iridium and ruthenium oxides, in multinary metal oxides the presence of partner metals, as well as the distinctive crystal structures, provides a means of regulating electronic structure. For example, in the pyrochlores Ln₂Ir₂O₇ (Ln = praseodymium, neodymium, gadolinium, terbium, holmium), Shang et al. found that an increased ionic radius of the lanthanide resulted in enhanced covelancy of the Ir–O bonds via broadening of the iridium 5d band and increased interaction with the O2p band, leading to improved OER activity; this was accompanied by increased conductivity owing to an insulator to metal transition (45). Hubert et al. found that among the A₂Ru₂O₇ a longer Ru–O bond and a weaker interaction of the Ru 4d and O2p orbitals compared with RuO₂ resulted in enhanced initial activity (58). On the other hand Abbot et al. noted that the OER activity of each pyrochlore series (i.e., iridate or ruthenate) generally improves as the size of the A-site cation decreases (59). These conflicting observations suggest that structure property relationships are more complex than merely a consideration of bulk crystal structure, and understanding of surface chemistry, and how this arises from preparative method also needs a greater consideration.

Considering the LOM mechanism for OER, where lattice oxide is removed to create a vacant site, the presence of oxide defects can also be considered a principle for the design of effective electrocatalysts. Thus preparation of materials that purposefully contain oxide defects is a strategy for optimisation of activity: this was demonstrated in the case of the pyrochore Y₂[Ru_1.6Y_0.4]O_7−δ , where replacement of some B-site Ru⁴⁺ by Y³⁺ results in oxide-ion deficiency, albeit accompanied by enhanced covalency of the Ru–O bonds owing to some oxidation of Ru⁴⁺ to Ru⁵⁺, giving higher activity than the parent Y₂Ru₂O₇ (46). Likewise, replacing Y³⁺ by Ba²⁺ in Y₂Ru₂O₇ generates oxygen vacancies as well as oxidation of ruthenium, thus boosting the OER performance (60). A similar effect was seen in Y_1.7Sr_0.3Ru₂O₇ with the additional effect of increased Ru–O covalency decreasing the charge transfer energy to accelerate interfacial charge transfer kinetics (61). For the materials Y_1.8M_0.2Ru₂O_7−δ (M = copper, cobalt, nickel, iron, yttrium) the substituent cations were used to purposely tailor the concentration of oxide vacancies and to enhance OER activity (62). Recently, fluorination of the pyrochlore Y₂Ru₂O_7−δ has been proposed as a means of increasing the concentration of oxide-ion defects (63). For the double perovskite La₂LiIrO₆ a unique mechanism for OER activity has been proposed in which the material is activated in acid conditions, by initial extraction of Li⁺, to yield Ir⁶⁺, leading to migration of cations to the surface and activation of surface oxide, with further steps involving lattice oxide participation, Figure 3 (43).

Dissolution of the new iridate and ruthenate oxides has been an important part of their study, and just as with the earlier work on IrO₂ and RuO₂, it is becoming apparent that significant surface reconstruction plays an important role in their activity as OER electrocatalysts. For IrO₂ there is now a body of work on degradation mechanism under acid conditions, including both experimental and computational studies (23, 64–66). During the degradation of IrO₂, it is concluded that redox chemistry of iridium takes place, and this is intimately connected with the chemistry of the electrocatalysis process (64). Kasian proposed the possible pathways of iridium dissolution, Figure 4, where release of soluble iridium species, such as [IrO₄]^2– and Ir³⁺ conceivably occurs at different points on the OER cycle (16). While it is likely that similar degradation pathways may occur during the use of the new mixed-metal oxide electrocatalysts that have recently been discovered, at present there are no experimental data available to propose mechanism to the same level of detail. It may also be the case that there is not a common mechanism at play for all materials, given the variety of crystal chemistries.

In the case of pyrochlores, Hubert et al. observed that the OER activity of A₂Ru₂O₇ catalysts changed over time and is accompanied by both A-site and ruthenium dissolution at different relative rates depending on the identity of the A-site (58). In our own work on (Na,Ca)_2−x Ir₂O₆·n H₂O pyrochlores, we found that upon immersion in concentrated H₂SO₄ almost complete removal of the A-site sodium and calcium takes place to leave a poorly crystalline material that still resembles the pyrochlore structure and with charge balance provided by protons that are accommodated as bridging OH groups in the pyrochlore lattice (67). Thus proton exchange provides a mechanism to counter the leaching of cations and may indeed allow the effective operation of the electrocatalyst, although this idea needs further investigation. Exchange of cations for protons, prior to use in electrocatalysis, has been applied to form active materials, in the case of H_3.6IrO₄·3.7H₂O prepared from Ruddlesden-Popper SrIr₂O₄ (48).

For the perovskite SrIrO₃, grown as thin films as a pseudo cubic phase, Seitz et al. showed that the active electrocatalyst after exposure to electrochemical potential in acid was actually IrO_x /SrIrO₃ formed from strontium leaching from the surface layers of SrIrO₃, and this outperformed binary oxides of iridium or ruthenium (41). Leaching of praseodymium and barium from the double perovskite Ba₂PrIrO₆ was accompanied by only negligible leaching of iridium, and gave a surface enriched in Ir⁵⁺ (42), while leaching of Y³⁺ from the pyrochlore Y₂Ir₂O₇ gave an active iridium-oxide layer (68). In other complex iridium oxides the importance of the leaching of partner ions is being recognised in the formation of the most active catalysts (69–74). These collective results, from iridates with a variety of crystal structures, illustrate how the partner metal cations in multinary oxides can be preferentially lost into the acid electrolyte to leave behind a stabilised iridium oxide surface, or amorphous particles of IrO_x anchored on a support. Surface restructuring is a key part of the nature of the iridate electrocatalysts, and recent studies have begun to investigate the atomic-scale mechanism of these processes (75). It is likely that ruthenates behave similarly, and electrochemical leaching of strontium from SrRuO₃ produced RuO₂ clusters with favourable OER activity (76).

While leaching of metal cations from mixed-metal oxides may simply be expected to leave active IrO_x and RuO_x particles, as could be prepared by other routes, it can be seen that in fact the multinary oxides offer distinctive properties. For example, as noted above for La₂LiIrO₆, delithiation increases iridium oxidation state to activate the surface for OER, while for Ruddlesden-Popper and pyrochlore materials the ability to accommodate protons yields robust materials with acid stability.

In terms of relating crystal structure to OER performance, Song et al. studied 11 different iridates, representative of the materials we surveyed in Part I (1), including pyrochlores, perovskites and Ruddlesden-Popper phases, and compared their OER activity to IrO₂ (77). Despite having different crystal chemistry, all materials show similar OER activity, however their stability upon repeated cycling showed a remarkable relationship to crystal structure. Those with crystal structures with predominance of face- and edge-sharing connectivity of IrO₆ octahedra offered highest stability, such as hexagonal perovskites, whereas the materials consisting of corner-shared or isolated octahedra, such as Sr₄IrO₆, gave rapid structural collapse and poor electrochemical cyclability. Even the most stable phases showed leaching of the base metal cations and formation of an iridium-rich surface. Song et al. classified the structures into three types, Figure 5, to relate their stability to crystal structure.

Separately, Yang et al. have noted that structures consisting of face-shared IrO₆ octahedra exhibit superior catalytic activity to those that contain corner-shared IrO₆ octahedra due to the weakened surface Ir–O bonding, which lowers the energy barrier of the potential-determining step (39). It has also been observed that highly distorted IrO₆ octahedra tend to be more active than regular IrO₆ octahedra (36), and similar observations about local structural distortion have been made by others, as noted above (78, 79). Taken together, these conclusions suggest some design principles for targeted preparation of materials that are stable in acid solution.

Finally, it is important to consider that as well as the inherent activity and stability of the oxide electrocatalyst, the nature of the catalyst support is important to consider, since this may modify the activity, ideally to provide further stability without loss in activity. This has been recently explored in the case of IrO₂: for example, in the case of titanium metal supports, if the temperature of calcination was too high then migration of titanium into the iridium oxide resulted in loss in electrocatalytic performance (80), the support may allow the oxidation state of iridium to be adjusted in the case of amorphous IrO_x (81), and the choice of support can provide stability to IrO_x nanoparticles (82). Similar considerations are likely to apply to the new mixed-metal oxides produced, but this is yet to be studied.

4. Conclusions and Outlook

In the past few years, a large set of complex oxides of ruthenium and of iridium have been prepared for study as electrocatalysts in acid electrolytes for the OER. Many of these materials have activities that apparently match or surpass that of IrO₂, although as we have highlighted comparison of reported activity data between literature studies is not always straightforward. The issue of long-term stability needs also to be considered, and several recent works have highlighted the need for proper benchmarking of OER catalysts. We also note that many of the emerging materials have not yet been tested in real devices, and initial findings are based on simple solution electrochemical measurements. There is clearly much work to be done to prove the value of the new materials in realistic conditions, where stability in the presence of other device components requires careful investigation towards scale-up to produce working devices (83). The synthetic route to materials is also important to emphasise, not only in targeting new materials in a rational way, but also when considering that materials must be prepared by a method applicable for manufacture at a suitable scale in a form for processing into devices. Our review highlights some of the diverse synthetic chemistry methods already employed, which suggests that there is opportunity still for discovery of new phases, and that some of the solution-mediated synthesis may allow control of crystal morphology and be suited for scale-up.

Mechanistic understanding is challenging for heterogeneous catalysis in general, but particularly for an electrocatalyst under conditions of device operation, and computation may play a role here, as highlighted for the case of binary IrO₂ and RuO₂ materials. Some interesting structure-property relationships are emerging now that a set of materials is available for consideration, particulary among the iridates that have been studied. Here, a relationship between local atomic connectivity and catalytic behaviour is becoming apparent, with suggestions that certain structural motifs allow the desired balance between activity and durability to be reached. Leaching of the partner metal cation in multinary phases is a common emerging theme, and while this may lead to active iridium-rich surfaces, the fate of the released cations in a real electrochemical device is yet to be established. The role of protons in this leaching is an aspect that warrants further investigation, since this might provide a dynamic mechanism of charge balance. Another structural aspect that has emerged is the role of oxide-ion vacancies and how creation of high-levels of these defects by elemental substitution can lead to enhanced electrocatalytic activity: there are possibilities in extending this work to explore a much wider susbtitutional chemistry, for example by anion replacement to give oxide-substituted materials, as shown recently for sulfur incorporation in SrIrO₃ (84).

Another area of research direction for consideration is the exploitation of interfacial effects, where judicious choice of partner materials may lead to a cooperative enhancement of electrocatalysis (8). For example, it has been recently been shown that a Ru/IrO₂ composite could be tuned by addition of small amounts of manganese to moderate the oxidation states of the pgm, and obtain high OER activity in an MEA device, assigned to interfacial enhancement (85). A related opportunity lies in using conducting oxides of iridium and ruthenium, such as pyrochlores, as supports for metallic nanoparticles (86). The role of the substrate and catalyst-support interactions is likely to be of importance in optimising the activity stability-balance in future work.

Finally, we emphasise that quantitative benchmarking of electrocatalyst performance must be considered in future work. A number of different metrics have so far been used to report activity and stability, notwithstanding that catalysts may not have been tested under comparative conditions. Future work must also consider the mechanism of degradation of the materials in order to optimise properties. At present, although it is clear that materials with highly desirable properties for OER electrocatalysis have been discovered in the past few years, the most appropriate materials for practical application have not yet been unambiguously identified.