Anthropogenic CO₂ emissions from fossil fuels use and industry amounted to around 33 billion tonnes in 2019 (1, 2), mainly arising from the power generation sector. The use of carbon capture and storage (CCS) is considered a key contributor in the portfolio of options for achieving the CO₂ emission reduction targets agreed by the international community in 2015 during the 21st Conference of Parties on Climate Change (COP21). With the premise of keeping the Earth’s temperature rise below 2°C (and more ideally below 1.5°C) above the pre-industrial average (the ‘two degree scenario’), this will require that at least 94 billion tonnes of CO₂ are captured and stored until 2050 (3, 4). The extent of deployment will naturally depend on the economics together with government support and political will, but nevertheless a consequence is that CO₂ capture implementation will make large quantities of ‘pure’ CO₂ available within the next 10 years. Although the focus on CO₂ is being its contribution to climate change, value can be obtained from its use as a feedstock for a range of products and services. It can either be used directly (unconverted) such as for enhanced oil recovery (EOR) and for the food and beverage industry or indirectly through being converted to fuels, chemicals and building materials (Figure 1). At present, the annual consumption of CO₂ is approximately 230 million tonnes, with fertiliser production, namely urea, being the largest consumer (~130 million tonnes CO₂), followed by its use for EOR (~70–80 million tonnes CO₂) (5). With the growing interest for further development and investment in some of the newer (less technically mature) conversion routes then in the future these indirect CO₂ users could start contributing more towards the CO₂ consumption figures.

Fig. 1

The use of CO₂ as a raw material is viewed in the literature as becoming much more widespread than it is currently, possibly generating a ‘CO₂ economy’ (6). Carbon dioxide capture and utilisation (CCU) has even been proposed as an alternative to CCS as a climate change mitigation strategy, notably where CCS opportunities are limited, but the present scale of applications is too small to have a large impact in emission reductions, as pointed out by others (7). In the context of potential mitigation options this view is also relevant if considering a widespread electrochemical-based solution to CO₂ emissions (viz. CO₂ electroreduction) due to the seemingly inconceivable electricity demands and scale of marketplace that would be required for it to be considered a prime choice for addressing the climate change targets (8). However, directionally some benefit in emissions is obtained from using the waste CO₂ for the synthesis of fuels, powered by renewable electricity such as solar, wind, hydroelectric and geothermal (9). This approach comes under the umbrella of power-to-X schemes and has been reviewed in the literature (10–12). Many of the processes rely on water electrolysis to produce hydrogen, itself a gaseous fuel, followed by thermocatalytic reduction of captured CO₂ to produce synthesis gas (syngas), formic acid, methanol, methane (synthetic natural gas (SNG)) or higher hydrocarbons. Power-to-X schemes could produce carbon-based fuels as storage vectors at times when electricity production outweighs demand, which is a significant problem of the intermittent wind or solar power generation. One challenge in these schemes is the inevitable loss of efficiency when multiple processes are coupled together. From an electricity perspective it is more efficient to use electricity directly than to make chemicals; however, market demand for sustainable fuels and chemicals means that these concerns can be overcome, especially when electricity prices are low. It should be noted that compared with some of the more niche CO₂ uses, the market sector for fuels is vast and while this makes the opportunity for deriving them from CO₂ appealing it also brings about its own challenges. The potential scale of deployment is very large, with a choice of distributed or centralised production, which also needs to be supported by electrolyser manufacturing scale-up if water electrolysis is required. The processes are often energy intensive leading to high production costs to compete with their fossil fuel based equivalents. Regulatory requirements and technical specifications for some of the products (including whether product blends are acceptable rather than pure products), would also need to be met. Nevertheless, with process improvements gained from further development and project investment from governments with industrial support, it’s seen as an important area to target for CO₂ utilisation.

In addition to electrochemical and thermocatalytic routes other options include biological-based conversions that use CO₂ as the carbon source, and a co-reagent such as hydrogen for energy supply, with some technologies already reaching commercial status. Examples are LanzaTech’s gas fermentation technology platform for producing fuels and chemicals such as ethanol and 2,3-butanediol (13) and Electrochaea’s methanation process (14). Both the direct use of CO₂, or CO₂ conversion, should not be viewed as a mitigation strategy per se for reasons already mentioned, but instead as an enabler for increasing the value of a CO₂-capture project and helping the ‘business case’. By displacing the use of fossil fuels and promoting a more sustainable circular economy, CO₂ reuse can also help improve a country’s energy security. Furthermore, there is also the positive social-economic impact through creation of jobs around the construction and operation of such a facility.

From the outset, the conversion of CO₂ into fuels and chemicals by chemical reduction presents a thermodynamically uphill challenge, as is suggested firstly by its large C=O double‐bond energy (750 kJ mol^–1) and the higher, i.e. less negative, Gibbs free energies for some of the main carbon-based reduction products listed in Table I. The conversion chemistry is driven by the free energy differences between the reactants and products indicating their relative stabilities (16). With CO₂ being the most highly oxidised product in many carbon-based processes, including chemical and biological pathways, it exhibits the lowest energy state of all carbon-containing binary neutral species. The large energy required to reduce it is a significant obstacle and would be reflected in the energy consumption costs of any economic evaluation. This energy can either be supplied as physical energy, for example thermal or electrical, or indirectly via the use of reactive chemical species as reagents, such as hydrogen that exhibit a higher Gibbs free energy to promote the conversion of CO₂ from a thermodynamics perspective.

One result of the energetic challenges inherent in the CO₂ conversion chemistry is there being an exciting opportunity for catalyst development in this area, to generate the reaction rates and product selectivities required for an economic process (17).

A question that often arises is which fuels or chemicals should industry target for scale-up and production and which conversion technology should be used, with a realisation that some technologies are still in early-stage development such as direct low temperature CO₂ electroreduction exhibiting performance challenges with respect to product selectivity and energy conversion efficiency. The answer to this will often be dictated by the process economics and market demand for the product and a technoeconomic assessment (TEA) comparing technologies and target products will be an important decision-making exercise. The economics depend on not only the capital cost of the technology equipment items (including reactant purification, product separation duties and ancillary equipment) but also on the availability, intermittency and costs of energy such as renewable electricity and heat, which vary depending on location and can greatly influence the operating costs. Furthermore, this might be supplemented by any government incentives that might exist, for example the latest Renewable Energy Directive (RED II) for transport fuels in the European Union (EU) (18), which itself could be revised as a result of the European Green Deal initiative (19). As a result, a credible business case for justifying deployment of a particular technology to produce a target fuel or chemical product would be dependent on the policies within the host country (or shared policies between countries as in the EU) and also site-specific providing accessibility to the CO₂ feedstock and a suitable energy supply. Such a lined-up scenario is highlighted by the commercial renewable methanol facility that is owned and operated by Carbon Recycling International (CRI). The George Olah plant is located in Iceland’s Svartsengi geothermal field near Grindavík on the Reykjanes peninsula and exploits captured CO₂ from the nearby Svartsengi geothermal power station and renewable electricity from the Icelandic grid to produce electrolytic hydrogen for use in methanol synthesis via the direct hydrogenation of CO₂. The methanol is used in gasoline blends with the George Olah plant effectively recycling 5500 tonnes of CO₂ a year. The renewable methanol product from this facility is sold under the brand name of Vulcanol^TM and is displacing a small amount of fossil fuels in the transport sector (20, 21).

Greenhouse gas (GHG) emission reduction often plays an integral part in the justification of a CO₂ utilisation plant and a full and robust life cycle assessment (LCA) is required to confirm CO₂ mitigation benefits compared to the conventional (often fossil fuel) route, taking into account the whole value chain from CO₂ origin to the final use of the product: a ‘cradle-to-grave’ analysis. Any climate change benefits will depend on the source of the CO₂ feedstock (fossil fuel, biomass or directly from air), the carbon footprint of the conventional product or process route that is being displaced, the nature of the energy used for the conversion, the scale of the CO₂-utilising process and the lifetime of the CO₂ in the final product. The analysis would reflect that the CO₂ avoided is not the same as CO₂ used.

In terms of the origin of the CO₂ feedstock then in the first instance it would invariably be from point sources, for example power stations or industrial cement or steel works, with CO₂ clean-up and concentration being required for the majority of utilisation processes. However, there has been much interest lately in direct air capture (DAC) of CO₂ with differing capture technologies being developed and showcased (22, 23). Although from a fundamental analysis (24) the separation energy would conceivably be significantly higher than that from an industrial point-source (0.04 vol% CO₂ in atmospheric air vs. for example a power station flue‐gas containing 8–15% CO₂ depending on the fossil fuel) and hence also the separation cost, several companies are working on this and have already attained both demonstration and commercial operating levels. There have been projections made for the CO₂ capture cost from DAC to come down to approximately US$100 tonne^–1 within a decade (22, 23), which although encouraging is still high compared to cost targets for the demonstrated and in some cases commercialised liquid amine-based capture systems. However, as well as delivering a high-purity CO₂ stream and reducing the costs and emissions associated with any transport of the CO₂ feedstock, an obvious advantage of DAC is that it is not limited to locating the conversion or direct utilisation plant close to a CO₂ point-source. This attribute lends itself towards smaller-scale distributed production rather than necessarily building a large centralised facility that takes advantage of the economies of scale. The use of DAC CO₂ would close the carbon loop and likely provide improved carbon mitigation benefits from either carbon neutral or carbon negative emissions (depending on final product), to be quantified by an LCA. Furthermore, a DAC-based source of the CO₂ feedstock could qualify for government incentives for fuels and chemicals in the future. Currently the EU’s RED II allows for DAC-derived CO₂ for producing synthetic fuels (renewable fuels of non-biological origin (RFNBO)), but also accommodates the use of lower cost CO₂ from point-sources such as power station flue-gas, so not exclusively incentivising the deployment of DAC (23). Like for many developing technologies within the CCU space, government support through grants and tax credits and robust CO₂ accounting frameworks will ultimately be needed to help drive the technology through the metaphorical ‘valley of death’ and into commercialisation with an investable package.

This paper provides a perspective on some of the different CO₂ conversion technologies and their applicability to target products that are being developed by both academia and industry. It covers a range of approaches, which are mainly catalytic in nature including homogeneous and heterogeneous (thermo)catalytic, electrocatalytic and biological methods, each of which are at different technology readiness levels (TRL) and exhibiting different strengths and research challenges.

In nature, the bulk of CO₂ reduction is carried out by plants and autotrophic microorganisms, that are capable of converting CO₂ and an external energy source into biomass and side products. Central to autotrophic metabolism are carboxylase enzymes, that incorporate CO₂ (or in some cases, HCO₃^–) to specific organic molecules. Autotrophs also possess energy harvesting systems, that take reducing potential from light or inorganic electron donors for CO₂ reduction.

Recent advances in genetic engineering have led to the use of these biological tools (autotrophic microorganisms, a variety of carboxylases, autotrophic energy-harvesting systems) and others as modular units to create synthetic biological routes from CO₂ to virtually any product of interest. This review discusses three such synthetic biology approaches for CO₂ reduction.

The first approach is to use autotrophic microorganisms with genetic engineering and process optimisation to maximise yield and purity of their natural products (C1 to C4) or to add heterologous enzymatic pathways that lead to new products (Table II). A commercial example is LanzaTech’s acetogenic strain Clostridium autoethanogenum, that naturally converts CO, CO₂ and hydrogen to ethanol, acetate, 2,3-butanediol and other products. Strain selection and evolution combined with the development of the gas fermentation process have led to improved ethanol production. LanzaTech’s first commercial plant was established in China in 2018 and is located at a steel mill, using industrial waste gas as feedstock. Biomass and other organic waste can also be gasified, giving Clostridia and other acetogenic bacteria versatility in feedstock. The enzymatic pathway acetogens use, Wood-Ljungdahl, is the most energy-efficient natural autotrophic pathway. Depending on the species, acetate is either the main product or precursor and hydrogen is used as the source of reducing energy (25). Other species of Clostridia have been used to produce butanol (26).

Photoautotrophic microorganisms such as microalgae and cyanobacteria are another example, using light to split water to generate oxygen, as well as hydrogen and H⁺ gradient for energy. This is used for the Calvin cycle (also known as the Calvin-Benson-Bassham cycle or the reductive pentose pathway), an enzymatic pathway that requires a lot of energy to fix CO₂. Microalgae, typically grown in large scale in ponds and can use wastewater, convert CO₂ to biomass for animal feed and a range of lipids for biodiesel (27). Chemical production from CO₂ has seen recent advances due to the expansion of available genetic tools for cyanobacteria. Combined with improved photobioreactor design, this has enabled production of 1-butanol at 4.8 g l^–1 (302 mg day^–1) from Synechocystis, the highest production rate of 1-butanol from CO₂ (28). Phytonix Corporation is a producer of biobutanol and biooctanol using cyanobacteria (29).

The ‘Knallgas’ bacterium Cupriavidus necator also utilises the Calvin cycle, but takes reducing energy from hydrogen or formate. This bacterium naturally accumulates polyhydroxybutyrate, a bioplastic precursor. With genetic engineering, it has been exemplified to produce branched-chain alcohols and alkanes from CO₂ (25). C. necator has been used in microbial electrosynthesis (MES, discussed later) where formate or hydrogen is produced from the cathode.

The second approach is to take model heterotrophic (not able to grow on CO₂) microorganisms and introduce carboxylases or autotrophic pathways that enable CO₂ utilisation. These organisms, with well-established use in industrial fermentations, are capable of high growth rates (specific growth rates of 0.5 h^–1 to > 3 h^–1) and high productivities when grown on sugars or other carbon sources. In addition, their well-characterised metabolism and a wide range of available genetic tools make them ideal hosts for genetic engineering. Genetic modifications and the introduction of the Calvin cycle including ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) into bacterium Escherichia coli enabled the phospho-sugar synthesis from CO₂ (30). In yeast Saccharomyces cerevisiae, the same approach led to 10% increased ethanol production from glucose and galactose, at the expense of side product glycerol (31). More recently, the introduction of Calvin cycle components and the modification of a native metabolic pathway of the methylotropic yeast Pichia pastoris has enabled growth on CO₂ as a sole carbon source (specific growth rate of 0.018 h^–1). This is a significant demonstration of conversion of an industrial heterotroph to a synthetic autotroph. Pichia is widely used for the production of commercial enzymes and pharmaceutical chemicals and can use methanol as a carbon source. This work facilitates the use of Pichia for CO₂ conversion into commodity or specialty chemicals using methanol as an energy source (32).

The third approach is to integrate the first two: a simultaneous or sequential fermentation of autotroph (to reduce CO₂ to a C1 or C2 product, such as formate, ethanol or acetate) and heterotroph (to utilise the first product as a substrate for >C2). This combines the beneficial characteristics of both types of organisms. An example is the two-stage process using the acetogen Moorella thermoacetica to convert CO₂, CO and H₂ to acetate, which was then used in a separate bioreactor by the oleaginous yeast Yarrowia lipolytica to make triglycerides (33).

As genetic engineering and bioreactor designs continue to improve, certain bottlenecks become apparent. Limited energy and CO₂ uptake occurs due to inefficient microbial photosystems (in photosynthesis, only 1% of light energy is utilised for biomass and product conversion) and the low solubility of both CO₂ and hydrogen in aqueous solutions. CO₂ conversion is also rate-limited by the carboxylase enzymes in CO₂-fixation pathways, which evolved in environments where fast kinetics was not a requirement. These bottlenecks are being addressed through further engineering and targeted evolution of biological energy-harvesting systems and individual carboxylases and through studies of natural CO₂-concentrating mechanisms such as carboxysomes (subcellular compartments where RuBisCO is localised). Aided by advances in bioinformatics and high-throughput enzyme production, more efficient synthetic carboxylation pathways (featuring faster carboxylases and less energy-demanding steps) are also being designed, constructed and tested in either cell-free systems or microbial organisms.

3.1 C1 Molecules

3.2 Larger Molecules

Synthetic fuels from renewable energy sources (known as ‘e-fuels’) have become increasingly attractive to achieve GHG emission targets as discussed in the introduction. Increasingly abundant low-cost renewable electricity combined with site specific advantages and policies, has enabled electrochemical processes to compete with traditional thermocatalysis methods, for example the George Olah plant outlined in the introduction coupling hydrogen through electrolysis with thermocatalysis to produce methanol. Water electrolysis is a well-established technology with many reviews (66, 67), hence this section focuses on going beyond hydrogen to the direct electrochemical reduction of CO₂. This area, while much more embryonic, is pushing towards pilot scale with companies like Siemens and Evonik Industries (68), Avantium (69), Opus 12 (70), CERT Systems (71) and Skyre (72) developing commercial systems. High temperature electrolysis to produce CO and syngas (73, 74) using solid oxide electrolyser cell (SOEC) systems could be advantageous if coupled with thermochemical processes to reduce heating cycles. SOECs are not able to reduce CO₂ directly to other hydrocarbons and oxygenates, unlike low temperature electrolysis.

4.1 What are the Possible Products of Carbon Dioxide?

4.2 What are the State of the Art Systems?

4.3 Future Perspectives

Hydrogenations and hydrotreating reactions with molecular hydrogen are critical operations in chemical and related industries. Regardless of the nature of the hydrotreated reactants and products, as society walks away from fossil fuels, the need for hydrogen will remain indispensable. With the increasing access to surplus renewable electricity, the hydrogen production by water electrolysis may take its rightful place not only in transportation and energy storage, but also as a clean hydrogen supply for chemical and related industries (1), replacing greenhouse gas (GHG)-producing methane steam reforming. Among the commercialised and emerging technologies, acidic proton exchange membrane (PEM) electrolysers can operate at up to 20 A cm^–2 current density and deliver up to 700 bar hydrogen at high efficiencies (2). Active and durable electrocatalysts are required to reduce the power input, the bottleneck being the sluggish anodic OER. The acidic environment, however, demands corrosion-resistant materials at high potentials. The winner so far is the OER-active conductive and corrosion-resistant iridium at typical loadings of 1–2 mg cm^–2. Lower iridium requirements were demonstrated, for example, for a PEM electrolyser with the 3 M nanostructured thin film (NSTF) catalyst reaching 2 A cm^–2 current density at 1.86 V at 0.25 mg_Ir cm^–2 (3), which translates into ca. 100 tonnes iridium for the production of 1 TW hydrogen (4). It is obvious that with the current technologies, the terawatt-scale hydrogen production cannot be met with the annual supply of scarce iridium of less than 10 tonnes (4). The annual global demand of hydrogen was reported as 73.9 million tonnes in 2018; as it is almost entirely supplied from fossil fuels, its current production emits 830 million tonnes of carbon dioxide per year (5). The hydrogen production replacement in chemical industries with water electrolysis is relatively well-positioned as a target area for decarbonisation of the industrial sector.

Hundreds of research papers have been focussed on the development of active and durable iridium catalysts, deposition techniques and associated catalyst layer components, which may limit the performance of the most active iridium catalyst formulations. Recent reviews classified iridium-containing catalysts for acidic OER (6) and the variety of methods for the synthesis of iridium oxide (7). Commercial catalyst production methods must be scalable, preferably not require specialised equipment apart from what is available in the catalyst production industries, not produce significant waste and lack the need for large amounts of chemicals, especially those that are hazardous to the environment. With this in mind, the objective of the current minireview is to select a number of the most efficient state-of-the-art iridium catalysts for acidic OER within reported wet-chemistry synthesis methods, focussing on the practicality and scalability of the techniques. We address only wet-chemistry routes, as they are most frequently reported, being relatively accessible in a research environment. Figure 1 and Table I summarise the catalyst synthetic routes and selected catalysts, addressed in this review; this is not a comprehensive summary of all possible routes and catalysts, but rather a careful selection of studies demonstrating promising combination of activity and stability in acidic OER.

Fig. 1.

The catalyst layer preparation methods, such as deposition methods, are out of the scope of this work, although they significantly affect the catalyst performance. Gas-phase catalyst (layer) preparation techniques (13, 14) are omitted for the same reason, as they require specialised equipment and feature simultaneous catalyst formation and its deposition. The review is based only on published research works; we acknowledge that it may become obsolete due to the rapid developments in the field or may miss some critical proprietary information. The citations are chosen only to support our viewpoints; they cannot be considered as a comprehensive list of the relevant works. Herein, we aim to provide comprehensive insights in selected promising wet-synthesis methods. We hope that the review may help the reader in the selection of a state-of-art catalyst for benchmarking purposes, as well as to assist in further developments of potentially scalable synthesis of active and durable iridium catalysts.

Bernt (15, 16) estimated that, in order to decarbonise the transportation sector by transitioning to fuel cell vehicles fuelled by renewable hydrogen, the metal loadings on the anode of polymer electrolyte water electrolysers should be decreased to 0.05 mg_Ir cm^–2. This loading can meet the demand for approximately 150 GW year^–1 installed capacity while using only 50% of the annual iridium production. In order to meet this requirement and also use electrolysis for other needs, such as energy storage and chemical industry supply, the specific iridium activity must be increased substantially. An ideal OER catalyst would have negligible overpotential; a highly desirable catalyst would be one that requires 200–300 mV overpotential (1.43–1.53 V) at 10 mA cm^–2 current, with catalysts achieving this same current density at overpotentials of 300–400 mV (1.53–1.63 V) being acceptable (17). As a durability criterion, if the overpotential is maintained for 10 h, the catalyst may be suitable for real device fabrication, taking into consideration the catalyst nature and mass loading as well (17). The high activity would manifest itself in a low Tafel slope at real operating potentials. The catalyst specific activity (specific current for electrocatalysis) is the product of the following catalyst characteristics, assuming ideal kinetics with no transport limitations (Equation (i)):

(i)

This equation clarifies that it is not enough to develop highly dispersed iridium catalysts, which would be a relatively easy task to do by synthesising sub-2 nm iridium particles with >50% dispersion but that a proper type of iridium species must dominate the surface. Furthermore, the involved parameters must be stable over the device lifetime.

The activity of the catalyst is usually dependant on the OER pathway, which has been shown to depend on the treatment of the iridium catalyst. In general, two mechanisms have been proposed, with the main difference being the predominant involvement of electron-deficient electrophilic oxygen (denoted as O^I–) and as activated lattice oxygen in the reaction. The OER on rutile-type IrO₂ proceeds by means of the classical oxide, electrochemical oxide or electrochemical peroxide pathways involving M–O, M–OH and M–OOH intermediates (18, 19) (Equations (ii)–(vii)):

(ii)

(iii)

(iv)

(v)

(vi)

(vii)

where M represents the metal oxide IrO₂; the oxide pathway involves reactions (ii), (iii) and (v), the electrochemical pathway involves reactions (ii), (iv) and (v), and the peroxide pathway involves reactions (ii), (iv), (vi) and (vii). The peroxide pathway has recently been shown to provide trends that are in agreement with experimental observations by Schuler et al. (19). The electrochemical oxide path is highlighted in red in Figure 2.

Fig. 2.

Catalyst featuring an electrochemically grown porous hydrous oxide layer, also known as oxy-hydroxide layer or amorphous IrO_x catalysts, exhibit an OER mechanism that involves an electrophilic O^I– species (21) and an activated lattice oxygen pathway (22, 23). According to Geiger et al. (22), a simplified pathway highlighting the need for the outer layer of the catalyst to be involved in the reaction could be Equations (viii)–(x):

(viii)

(ix)

(x)

where x is a vacancy in the porous hydrous oxide layer. Other references have attributed the increased activity to electrophilic O^I– species (21). Ir-O-Ir would play a similar role to the proposed highly reactive, electrophilic oxygen O^I– species (21, 24). The second step would have a similar function to the preliminary reaction proposed by Pfeifer et al. (24) (Equation (xi)):

(xi)

where IrO_x O represents the IrO₂ matrix with an adsorbed oxygen. The pathway involving the electrophilic O^I– species is also highlighted in Figure 2 in green, where HIrO₂ would loosely represent the O^I– intermediate.

The OER has been shown to be more active for the pathway involving the electrophilic O^I– species and activated lattice oxygen. Unfortunately, this pathway tends to be deactivated due to the lattice oxygen evolution leading to iridium dissolution, and to the transformation of the oxo-hydroxide to less active anhydrous species (22, 25–27). Stabilisation of the iridium atoms in the pathway involving the electrophilic O^I– intermediate via enhanced crystallinity (26) or the use of mixed oxides could minimise stability issues. Crystalline IrO₂ has a lower intrinsic activity but is more stable due to strong Ir–O bonds between IrO₆ clusters, with only topmost layers of the rutile contributing to both processes (25). The superior stability of thermal IrO₂ is explained by slower kinetics of IrO₃ hydrolysis as compared to its decomposition (20). It has been suggested that both pathways occur during the OER with the activated lattice oxygen pathway being dominant at low potential, due to its high activity, and the classical pathway being dominant at higher potential (20). At potentials relevant to the OER, it is possible that the oxo-hydroxide layer slowly transforms to anhydrous oxide with a subsequent loss in activity and enhancement in stability as recently shown by atomic probe tomography (27).

Speaking of which iridium phase is to be synthesised and introduced into the electrochemical device, the literature features metallic iridium with a variety of predominant crystallographic orientations, amorphous hydrous IrO₂, crystalline rutile IrO₂, their mixtures, as well as multimetallic iridium composites. The metallic iridium may be oxidised by calcination in air before the catalyst layer assembly (28) or electrochemically in situ (26, 29). Thermal iridium oxidation to IrO₂ occurs between 200°C and 500°C (30), the higher the temperature, the higher the crystallinity and electrochemical stability, but the surface area and the activity decrease (31). The 400–500°C region was recommended to strike a balance between activity, stability and conductivity (28).

Many state-of-the-art catalysts, as shown below, use electrochemical in situ oxidation. The iridium (110) surface evolves into two chemically different iridium species, with an active accessible oxide-metal interface (32). The most dense (111) iridium surface is more resistant to the oxidation, and once the oxide is formed, the metallic interface is buried. Although the kinetics of oxide formation and redox properties of the two surfaces are different, their final reached OER activities are rather similar. The same work (32) recommends that for the formation of a porous hydrous IrO₂, the in situ Ir(0) activation should include oxidising-reducing cycles, instead of conventionally used electrooxidation, although another study argues that the repetitive electrochemical oxidation and reduction unavoidably leads to dissolution (33). The electrochemical oxidation proceeds via hydroxide to the irreversible Ir(IV) oxide formation in the nanoparticles, while bulk iridium preserves its metallic subsurface with porous Ir(IV) surface layers (34). Thus, one must be mindful of the iridium dissolution during electrochemical oxidation via hydrous IrO₂ growth (33). When 20 nm iridium films are used for the acidic OER, their lifetime is similar to the lifetime of the hydrous IrO₂ and is significantly lower than for crystalline IrO₂ (22).

To produce highly crystalline IrO₂, which is more stable but less active than hydrous IrO₂, preliminary annealing in air may be recommended, whenever possible. The exceptions, of course, include unsupported polymer-stabilised nanoparticles (26), where annealing would result in particle agglomeration, as well as metal carbides, where it would lead to oxidation and loss of conductivity (35). In such cases, the electrochemical oxidation procedure must be optimised as it affects the catalyst stability.

Fine tuning of the oxide crystallinity, crystallographic orientation, number of oxygen defects and length and strength of Ir–Ir and Ir–O bonds via thoughtful synthetic approaches may diminish the gap between the active but unstable and stable but less active phases. Some such examples, leading to state-of-the-art catalysts, are given below.

Thus, the treasure hunt for the most efficient iridium OER catalyst is a simultaneous optimisation of activity vs. stability. In addition, in order to achieve high iridium utilisation, the catalyst must be easy to integrate into a catalyst layer in order to yield a layer that has both low loadings, and excellent charge and mass transport. Charge transport should include both electronic and protonic in-plane and through-plane conductivities. Proton transport can be optimised by controlling the amount of electrolyte in the layer (36, 37), while improving electron conductivity can be achieved with the use of a conductive support and interconnected IrO_x network. Similarly, good in-plane conductivity could be maximised by using a porous transport layer (PTL) with a microporous layer (38).

The catalyst lifetime depends on its operating current density and overpotential. It may be estimated using the so-called activity-stability factor (39) or stability number (S-number) (22) i.e. the ratio of evolved oxygen to dissolved iridium at constant overpotential (39). Geiger et al. reported that IrO₂ powder features a one year lifetime at 200 A g_Ir^–1 vs. one month for hydrous oxide (22). The numbers are specific to the used electrochemical cell (Figure 3); for example, oxygen bubble accumulation, a common occurrence in rotating disk electrode (RDE), could make the stability studies not a reliable predictor for the catalyst lifetime in a PEM electrolyser long-term behaviour (16). In a PEM electrolyser, iridium dissolution might result in both iridium ions in the water feed, and migration and redeposition at the anode/membrane interface, membrane (40–42) and possible deposition on the cathode leading to platinum deactivation, especially at high current densities and overpotentials (42).

Fig. 3.

Last but not least, even if the most ideal active and stable iridium phase is developed, the benefits will only manifest if other occurring phenomena are not rate-limiting. The high intrinsic activity and stability may be masked by the limitations in the characterisation technique used, or by the electrode fabrication methodology (43, 44). For example, the commonly used RDE technique might be subject to inert backing passivation (44), and catalyst coated membrane (CCM) fabrication techniques, such as spray-coating, doctor blade and inkjet printing method, could result in very different electrode structures for full-cell testing. Inadequate RDE or CCM fabrication, or non-optimal electrolyte loading can result in excessive charge and mass transport limitations. Further, possible causes of loss of activity are poisoning of the catalyst surface by Nafion^TM (45), impurities in the electrolyte in RDE. In the presence of non-iridium components, cations place-exchange with sulfonic acid groups in the polymer electrolyte resulting in decreased proton conductivity (46). All these complicate not only the scale-up, but also the assessment of the intrinsic catalyst performance in laboratory-scale devices. It is possible that efficient catalysts have not been identified because of these limitations. Some benchmarking procedures for the OER evaluation have been proposed in RDE (26, 47) and alternative liquid-flow (44, 48, 49) or vapour-fed cells (19) and are urgently needed to be followed.

3.1 Adams’ Fusion Method

3.2 Iridium Nanoparticles Stabilised by a Capping Agent

3.3 Stabiliser-Free Wet Chemical Synthesis Methods

In heterogeneous catalysis, supports, typically with a high specific surface area, are used to stabilise highly dispersed active metal nanoparticles, both during catalyst synthesis and to ensure their stability against agglomeration during a reaction. In the case of electrolysers, the supports can also be used to enhance the electrode electrical conductivity as they will reduce the contact resistance between particles. The acidic OER environment dictates specific requirements to the type of support: it must be resistant to chemical and electrochemical dissolution, and preferably must have a high electronic conductivity. The latter need is mandatory if iridium loading is low; however, if IrO₂ covers most of the support, it may provide sufficient percolative transport for the electrons (3). The film, however, must be as thin as possible to provide advantages over unsupported IrO₂ nanoparticles. In this subsection, we focus on the wet synthesis of iridium catalysts on powdered supports, which could be mixed with a Nafion^TM solution for catalyst layer preparation.

In recent years, carbon and metal carbides have been receiving less and less attention because of carbon oxidation and volatilisation to CO₂ at high applied potentials (58). As a notable exception, one of the most active iridium catalysts reported so far features a carbon-based support (10). A 40 wt% iridium catalyst was prepared by impregnation of graphitic carbon nitride (GCN) nanosheets with the metal precursor followed by annealing in air at 350°C. Thus embedded IrO₂ possesses compressed Ir–Ir bonds and decreased coordination numbers of Ir–O and Ir–Ir, which was suggested to weaken the adsorption of oxygen intermediates leading to increased OER activity. The reported specific currents are 580 A g_Ir^–1 at 1.55 V and 1493 A g_Ir^–1 at 1.6 V. The catalyst required the overpotential of 278 mV at 10 mA cm^–2 at 0.07 mg_Ir cm^–2 loading (in RDE in 0.5 M H₂SO₄). Authors demonstrated only 35 mV potential increase at 20 mA cm^–2 for ca. 4 h in an RDE; and 78.5% current retention in a laboratory water splitting device after a 24 h operation at 1.6 V. The fate of GCN was assessed by holding 2.2 V_RHE for 2 h with intermediate cyclic voltammetry (CV) measurements between 0.4 V_RHE and 0.6 V_RHE; the double-layer capacitance decreased by 10% in the first 0.5 h and remained stable up to 2 h (10); apparently, the graphitic support nature with nitrogen heteroatom provides its stability in acidic electrolysis. Given the high catalyst activity and a rather easy and potentially scalable preparation of GCN and Ir/GCN, the studies of the catalyst durability and performance in a PEM electrolyser are warranted.

Among oxidation-resistant conductive metal oxide supports, antimony tin oxide (ATO), indium tin oxide (ITO) and fluorine tin oxide (FTO) have been the focus of the most research because of their relatively high conductivity. Unfortunately, the dopant’s corrosion brings down the conductivity, increasing ohmic losses and decreasing the energy efficiency (59). Dissolved cations may also ion-exchange with the membrane and lead to its degradation (46). A recent study observed neither activity nor stability benefits from the dopant addition (60). Although FTO possess the lowest conductivity, it was found to be the most stable material between –0.34 V_RHE and 2.7 V_RHE, followed by ITO and ATO. The stability is assigned to the oxygen atom exchange in SnO₂ with fluorine, instead of cation exchange in the case of ATO and ITO synthesis (59). ATO, in turn, was suggested to mitigate iridium dissolution by preserving it in lower oxidation states (61). Commercially available samples usually feature low surface areas; several synthetic techniques were suggested in literature for the preparation of mesoporous doped SnO₂ with relatively high areas (for example, between 125 m² g^–1 and 263 m² g^–1) (62). To deal with toxic NH₄F for the FTO synthesis, safety measures must be in place, as in any chemical and engineering process. A number of iridium catalysts supported on doped SnO₂ with high activities and low overpotentials at low iridium loadings were recently reported.

A popular method for the preparation of supported iridium OER catalysts is a colloidal precipitation of IrO₂ from an iridium molecular precursor by means of NaOH; the synthesis may proceed in ethylene glycol (29), which serves both as a solvent and a reducing agent, or, for example, in a hydrothermal microwave reactor (63). Small 2–3 nm particles may be obtained (61), or even smaller (1.5 ± 0.2 nm) if a stabiliser is added (64). The support may be added to the colloidal dispersion either during synthesis, or after the nanoparticle formation. The use of high-boiling ethylene glycol, though, complicates the potential process scale-up because solvent removal under vacuum is usually used (57), instead of centrifugation or filtration of highly diluted suspensions, as practiced in laboratories. As an example of such preparation method, when SnO₂ was doped with tantalum to produce tantalum tin oxide (TaTO) and used to deposit preformed 1.7 nm IrO_x nanoparticles at 11–18 wt% iridium loading, the OER activity of the fresh catalysts after electrochemical conditioning approached 250 A g_Ir^–1 at overpotentials of 280 mV and 370 mV at 0.020 mg_Ir cm^–2 loading (25°C, 0.05 M H₂SO₄) (11). Although the electronic conductivity of TaTO was two orders of magnitude lower than that of ATO, its use did not result in decreased activity, which was ascribed to the conducting role of well dispersed IrO_x nanoparticles. In accelerated ageing tests at 1.2–1.6 V potential steps, the IrO_x /TaTO catalysts demonstrated between 70% and 90% activity retention vs. 60% for the ATO-supported catalyst. The loss of the dopant was one or two orders of magnitude lower for tantalum as compared to antimony, while the loss of tin was not affected. The iridium dissolution was found dependent on the tantalum loading: the higher loading decreased iridium oxidation state contributing to its dissolution, while at lower loadings tantalum shell suppressed IrO_x nanoparticle detachment. This study (11) also demonstrates that the use of a support contributes to enhanced iridium leaching, as compared to commercial IrO₂. As a result, although the activity of unsupported IrO₂ is significantly lower than that of the developed catalysts, its stability to dissolution contributes to high S-numbers (ratio of evolved oxygen to dissolved iridium) (22). For example the S-number for IrO₂ was twice as high as that for selected IrO_x /TaTO catalysts at 1.6 V and similar at 1.5 V (11). When hydrous IrO_x was supported on ATO, its S-number was also lower than the one for unsupported IrO_x , but the calcined supported samples demonstrated up to two orders of magnitude higher S-number as compared to IrO₂ (60).

When the above-mentioned (Section 3.2) surfactant-stabilised iridium nanodendrites (39 m² g^–1 area) were deposited on high-surface-area ATO (235 m² g^–1), an initial overpotential of 260 mV at 10 mA cm^–2 at only 0.0102 mg_Ir cm^–2 loading (RDE, 0.05 M H₂SO₄) was observed while accelerated durability test showed a minor overpotential increase by 30 mV over 15 h compared to an abrupt increase for other tested catalysts at earlier times (29). The specific current at 280 mV overpotential was reported at 70 A g_Ir^{– 1} vs. 8 A g_Ir^–1 for Ir black. In a PEM electrolyser, the catalyst demonstrated the current density of 1.5 A cm^–2 at 1.8 V and 1 mg cm^–2 loading compared to 0.8 A cm^–2 for Ir black.

An atomically dispersed iridium on ITO with ultimate iridium dispersion was developed by grafting 0.86 wt% iridium as an organometallic iridium complex followed by calcination in air at 400°C (65). The specific current of 156 A g^–1 iridium was reached at 280 mV overpotential and 0.021 mg_Ir cm^–2 loading (0.1 M HClO₄). An overpotential of 350 ± 20 mV was required to drive the 10 mA cm^–2 current at such low metal loadings over the course of 2 h; some iridium agglomeration was observed in the used catalyst and its consequences on the long-term performance requires further analysis.

From the catalyst synthesis viewpoint, one must be mindful that the support’s chemical composition may change during the synthesis, affecting the electrochemical performance. For example, if Adams’ method involving high-temperature calcination is used to produce IrO₂ on a carbide support, the support oxidation leads to the loss of conductivity (for example, TaC lost its conductivity from 120 S cm^–1 to 10^-8 S cm^–1 at such circumstances) (35). A similar carbide oxidation to a less-conductive oxide was reported for iridium nanoparticle synthesis in the presence of a support by a polyol method (heating in a reducing ethylene glycol with precipitating NaOH) (58). Moreover, iridium, tin and indium oxides may form mixed oxides; the lattice vacancies are thus produced upon tin and indium in situ dissolution, improving the initial activity but jeopardising the durability (66) due to enhanced iridium dissolution.

Development of multimetallic iridium-containing catalysts has recently attracted considerable attention as a means of enhancing the OER catalyst performance, as had proven beneficial for catalyst development for fuel cells and alkaline water electrolysis. However, with iridium being the most corrosion resistant metal and still dissolving under the acidic OER conditions, any other metal would have an even higher dissolution rate. A rather popular combination of iridium and ruthenium features high activities due to the higher OER activity of ruthenium than that of iridium, but is not practical because of low corrosion resistance of ruthenium, which is also a scarce and expensive metal. The studies of the mixed oxide OER catalysts do typically address (and inevitably show) the dissolution of the catalyst components, but there is a lack of studies on the effect of the leached ions on the PEM system level. It is likely that the non-iridium cations may not only ion-exchange on Nafion^TM changing its properties (67), but may also travel to the cathode side and poison the platinum cathode as was shown for iridium (42). Membrane degradation may also occur due to the attack of HO^• and other radicals, whose formation is catalysed by transition metal cations (46). For example, iron and copper ions were shown to dramatically enhance membrane degradation (68).

Thus, a practical mixed oxide catalyst for an acidic OER application may be envisioned as one of the following composites (Figure 6): (a) an IrO_x shell fully covering the core with an earth-abundant metal increasing the iridium dispersion. In this case, electrolyte contamination with the second metal may be delayed as compared to the mixed alloys until the iridium shell atoms leach exposing the core atoms; (b) iridium nanostructures produced by the preliminary removal of a sacrificial second component from a bimetallic composite, either by potential cycling or chemically. The selective leaching of the second component leads to surface restructuring (69), porosity enhancement (39), formation of lattice vacancies (70) and ECSA increase (71). Lattice vacancies formation via secondary metal leaching is a unique opportunity to modify the electrophilicity and Ir–O bond length leading to the enhanced OER activity as compared to IrO_x synthesised only from an iridium precursor (70, 72). Some works report, though, that iridium leaching from the composites may even be increased due to the created lattice vacancies, as compared to monometallic iridium catalysts (39).

Fig. 6.

Below, we provide some examples for such catalyst synthesis and performance. We note that the direct deposition methods, such as reactive sputtering and physical vapour deposition, are very frequently used for the mixed oxide studies (39, 71–75), because they ensure structures with well-controlled composition and stoichiometry, thus, enabling the fundamental understanding of the composite’s behaviour. As this review does not cover the catalyst layer preparation methods, we only focus on the production of metal oxide powders or colloidal dispersions which were or could be mixed with the Nafion^TM solution.

5.1 Core-Shell Bimetallic Nanoparticles

5.2 Selective Leaching of the Sacrificial Component (Hard Templating)

The fast-growing pool of published studies on the iridium-catalysed acidic OER continually contributes to the mechanistic understanding of iridium activity and durability, the role of the second metal in alloys and metal-support interactions. As a myriad of more or less sophisticated methods and structures emerges, one must keep in mind the method practicality, safety, ease and scalability, and that it must use as little resources and produce as little waste as possible. This, most likely, precludes the use of surfactant-assisted routes of wet catalyst synthesis and might jeopardise stabiliser-free colloidal synthesis in vast amounts of organic solvents.

Despite the many advances on catalyst synthesis, and recently proposed metrics to assess catalyst activity and stability, such as the S-number and the activity-stability factor, a lack of consensus on the metrics used to report catalyst performance, as well as the electrochemical testing benchmarking procedures, make the comparison of different catalysts challenging. Activity studies reviewed for the fresh catalysts are not always accompanied by meaningful stability assessment, and high reported activities should be regarded with care, as the most active (hydrous) amorphous IrO_x is less stable than the less active crystalline IrO₂. With such considerations, it appears, to our subjective opinion, that Adams’ fusion method (or other thermal methods) producing relatively stable rutile, deserve closer attention and further modifications. To take advantage of the enhanced activity of electrophilic oxygen sites, as follows from the multioxide studies, modification of the electrophilicity and Ir–O bond length via secondary metal leaching is a possible way to bridge the gap between the activity and stability. The use of supports may increase the stability of the calcined catalysts, however, the addition of dopants to improve electrical conductivity must be used with care.

One must keep in mind that the catalyst layer fabrication can have a tremendous effect on the activity observed in a real electrode. The ionomer-to-catalyst ratio, particle size, the use of a support, catalyst-ionomer-solvent interactions in the dispersion, and deposition technique, can have a significant impact on the electrode electrochemical surface area, ion and electron conductivity and its ability to transport reactant and byproducts in and out of the cell. Gas-phase catalyst synthesis and deposition methods, for example (13), which were not reviewed here, may provide a promising alternative to the wet-chemistry techniques. Moving forward, not only the catalyst synthesis, but also development of common characterisation techniques and metrics to assess activity and stability in RDE, and optimal electrode fabrication methods must become paramount in our collective effort to develop the most efficient OER catalyst, and OER electrode.