Integrated CO₂ capture and electrochemical conversion: coupled effects of transport, kinetics and thermodynamics in the direct reduction of captured-CO₂ adducts

Abstract

Upgrading anthropogenic CO₂ from concentrated point sources or directly from the atmosphere is a valuable approach in closing the carbon cycle. Existing processes capture the CO₂, concentrate it into pure gas streams, transport it, and then convert it into fuels and chemicals in a separate process plant. This sequential approach results in higher energy and operating costs which can be reduced by integrating the capture and conversion steps to directly reduce the captured CO₂-bound adduct to value-added products. The direct reduction of the captured CO₂-bound adduct is called the captured-CO₂ reduction reaction (c-CO₂RR). Understanding of c-CO₂RR has been obscured by the higher intrinsic complexity of the system. The CO₂ capture media is a complex space of several buffer reactions that allow the co-existence of different carbon species in solution depending on CO₂ loading, temperature, pressure, and pH. In order to design improved capture agents and catalysts for integrated CO₂ capture and conversion, it is essential to identify the carbon source and the primary factors influencing product formation on a c-CO₂RR catalyst. This review delineates the strategies to determine the active carbon species for integrated CO₂ capture and conversion systems. Furthermore, it summarizes the fundamental applications of mass transport, thermodynamics, and kinetics across various c-CO₂RR scenarios.

---

Avishek Banerjee

Avishek Banerjee received his BTech in Chemical Engineering from National Institute of Technology Durgapur, India. He then pursued his MSc at McGill University in Montreal, Canada, where he worked with non-thermal plasma for synthesis of valuable fuels and chemicals. In 2022, he joined UCLA as a PhD student in the Department of Chemical and Biomolecular Engineering. His research focuses on the electrocatalytic approaches of integrated CO2 capture and conversion, with an emphasis on understanding the roles of transport, thermodynamics and kinetic phenomena in these processes.

Carlos G. Morales-Guio

Carlos G. Morales-Guio is an Assistant Professor of Chemical and Biomolecular Engineering at UCLA. He received his BEng degree in Chemical Engineering from Osaka University and his MS and PhD in Chemistry and Chemical Engineering from EPFL. He then moved to California as a postdoctoral research fellow at Stanford University. His lab at UCLA is interested in understanding engineering principles relevant to the scale-up of electrocatalytic processes. The lab focuses on understanding the fundamentals of mass, charge and heat transport at the electrode/electrolyte interface, reactor design, process intensification and process automation to guide the development of sustainable energy applications.

---

Broader context

Existing processes for CO₂ capture and conversion start by separating the CO₂ from the effluent of a point source or the atmosphere, and then this concentrated CO₂ stream is used as the feedstock for a catalytic conversion step in a separate unit. The CO₂ capture step involves multiple thermal and pressure cycles that result in energy losses and increased operating costs. Recently, a one-step process called reactive capture of CO₂ has been proposed in which CO₂ in its captured form (c-CO₂) is directly transformed into value-added products while circumventing the CO₂ release step. The research on the electrochemical c-CO₂ reduction reaction (c-CO₂RR) is just starting, but it is already evident that the complexity of the system pushes the limits of what is known in science and engineering. Understanding individual phenomena of CO₂ capture and CO₂ conversion is not enough. When capture and conversion are integrated, new phenomena emerge that are unique to the performance of the capture step in the presence of a catalyst, and to the performance of catalysis in the presence of the capture agent. Thus, it is of critical urgency that all researchers entering this field learn the fundamentals of transport, thermodynamics, and kinetics that link CO₂ capture and conversion. This review covers the latest developments in c-CO₂RR research and our viewpoint on the challenges and future opportunities that exist in this space as the field matures.

1 Introduction

CO₂ contributes to almost 79% of total greenhouse gas emissions and is considered the major cause of global warming.¹ The development of technologies that contribute to the reduction of CO₂ emissions and its direct capture from the atmosphere is more urgent than ever before. Global emissions of CO₂ are estimated to be approximately 35 billion tons per year, with approximately 2 billion tons per year attributed to the chemical industry.^2,3 Despite the increasing adoption of electrification for heating in various chemical manufacturing processes, burning of fossil fuels for heating is just a fraction of the CO₂ emissions and there remains a pressing need to source carbon atoms for chemicals from sources different from fossil fuels. As the demand for chemicals continues to grow, technologies that can upgrade the anthropogenic CO₂ from concentrated point sources and the atmosphere are needed to close the carbon cycle. Although the use of renewable electricity to capture and convert CO₂ into value-added chemicals offers a pathway to decarbonization, significant challenges persist in terms of energy intensity and cost compared to conventional methods. Overcoming these hurdles is crucial to making these processes industrially viable.⁴

Existing processes capture CO₂, concentrate it in pure gas streams, compress CO₂, and then convert it into fuels and chemicals, all in separate reaction units (top panel in Fig. 1). This sequential multistep approach results in higher energy and operating cost, arising from the inefficiency of the CO₂ release and capture media regeneration step before CO₂ compression. CO₂ release from direct air capture (DAC) processes have an estimated energy cost between 3400 and 8810 kJ per kg CO₂ for solid and liquid sorbent, while the CO₂ compression step along with the transport of gas requires an energy of 324–432 kJ per kg of CO₂.^5–10 Recently, a one-step route has been proposed where CO₂ capture and conversion are integrated into one process that circumvents the inefficient CO₂ release step, by directly reducing the captured CO₂-bound adduct to value-added chemicals (bottom panel in Fig. 1). The elimination of the inefficient temperature and pressure swing cycles during CO₂ capture and conversion can potentially reduce 50% of the total energy cost. The gains, however, must occur on both processes: (1) the CO₂ capture, and (2) the electrochemical captured-CO₂ reduction reaction (c-CO₂RR) steps. This requires that the yield and energy efficiency of the c-CO₂RR electrolyzer is comparable to current state of the art CO₂ electrolyzers.¹¹ In principle, integrated CO₂ capture and conversion has the potential to be more efficient and less expensive than the conventional approach relying on CO₂ electrolyzer technologies for the CO₂ reduction reaction (CO₂RR). However, a long road remains ahead to bring c-CO₂RR electrolyzer technologies in par with today's CO₂RR electrolyzers.

Fig. 1 Illustration showing the conventional CO₂RR pathway (shown by the green path) and the c-CO₂RR pathway (shown by the orange path). The CO₂RR pathway is a multistep process comprising additional energy costs from the CO₂ release, compression, and transport steps before the CO₂ reduction reaction can take place, whereas the integrated c-CO₂RR is a one-step process that could potentially reduce energy and operating costs. Furthermore, both the pathways are interrelated to the fundamental principles of thermodynamics, transport, and kinetics which influences the product formation.

CO₂ capture technologies are much more mature than the reduction step in c-CO₂RR. Many studies have been conducted on solvent-based capture technologies with amines or carbonates. However, preliminary work suggest that amines might not be appropriate for c-CO₂RR due to their strong interactions with the catalyst surfaces and the enhanced hydrogen evolution reaction.¹² Alternative capture agents like phenoxides,¹³ alkoxides,¹² and amino acids¹⁴ could indeed result in renewed interest in the development of capture agents specifically designed for integration with c-CO₂RR, beyond those being commercialized today. Further optimization of the capture step is guaranteed once we find promising catalysts for c-CO₂RR and as we start to extract capture agent design rules for integrated carbon capture and conversion.

Integrated CO₂ capture and conversion systems are complex, with several carbon species co-existing in the reactors. Identifying the source of carbon while quantifying products, and elucidating the key factors that affect the activity and selectivity of a catalyst, is important. Although c-CO₂RR has been studied in the context of electrochemical,^15–19 thermal^20–23 and biological systems,^24–27 this review focuses solely on the electrochemical route. In recent years, researchers have reported evidence of electrochemically reducing the captured CO₂-bound adduct to products.²⁸ However, significant challenges still remain in understanding the transformation of the captured CO₂-bound adduct. In the electrolyte solution the vapor–liquid equilibrium, ion-speciation, and reaction equilibriums dynamically varies as a function of the temperature, pressure, and pH. These factors are further governed by the fundamental principles of mass transport, thermodynamics, and kinetics, which when not carefully investigated can convolute the way we interpret the data collected from c-CO₂RR.

In the field of electrocatalysis, there is a consensus within the scientific community regarding the importance of transport, thermodynamics, and kinetics; however their full integration into scientific studies remain somewhat limited. Thermodynamics determines the feasibility of a process by delineating the minimum energy requirements and equilibrium conditions. It also links variations in local concentration, temperature, pressure, and electric field of an electrochemical system to changes in the electrochemical potential. Thermodynamics further determine the direction of mass and energy fluxes in and out of a control volume, which is linked to spacial and temporal differences in local concentration, temperature, and pressure. Transport properties determine the scalability of a process by giving insights into technically feasible energy and mass fluxes, which in turn will be linked to reactions in multidimensional space and time coordinates. Transport phenomena are used to describe processes of diffusion, convection, and migration of species in electrochemical cells. Different reactors have different hydrodynamics, resulting in different transport properties. By and large, the electrocatalysis community has failed to recognize that we measure experimentally reactor kinetics and not reaction kinetics, slowing the progress of the field. Thus, understanding of mass, heat, and charge transport is necessary to decouple reaction kinetics from reactor kinetics, and this is also true for c-CO₂RR systems. Thermodynamics, transport, and kinetics cannot be disentangled because of their interdependence. In order to overcome the challenges of successfully scaling-up electrochemical systems, it is necessary to measure, model and control the coupling of thermodynamics, transport, and kinetics even when studying the most basic electrochemical system at a bench scale (Fig. 1).

Most reviews and perspectives focuses on the different amine capture agents and discusses scalability challenges that could arise from c-CO₂RR, while comparing the value proposition of c-CO₂RR with traditional CO₂RR.^11,29–34 In contrast, our review is meant to serve as an introduction to the field of c-CO₂RR for chemists, material scientists, and engineers who although might have learned the various concepts of transport, thermodynamics, and kinetics separately, have not seen how these three concepts come together at different length and time scales in c-CO₂RR. This review will further aim to lay out some guidelines on the processes and factors that influence product selectivity and that should be considered while studying and reporting data for c-CO₂RR systems. It also includes the latest breakthrough results and the wider implications for the scientific community, and provides the tools necessary for understanding the fundamentals of electrochemical c-CO₂RR.

2 Understanding vapor–liquid equilibrium and ion speciation under c-CO₂RR conditions

The different species present in a solution filled with a CO₂-bound capture agent can be probed by understanding the vapor–liquid equilibrium of the system. When a capture agent is dissolved in a solution, and then it is put in contact with a gas phase containing CO₂, free and dissolved CO₂ will always be present in solution due to the equilibrium reactions between the gas and liquid phases. The amount of dissolved CO₂ is furthermore a function of CO₂ loading, pH, temperature and pressure of the system. Industrially, amines are currently the leading CO₂ sorbents, and an entire field of science and engineering has been developed around the optimization of CO₂ absorption capacities of different amines.^35–39 This section dives deeper into CO₂ capture with amines used primarily in post-combustion processes along with insights into other groups of capture agents, including hydroxyls used for direct air capture (DAC),^40–44 and the untapped group of alkoxides and phenoxides that is recently gaining interest.^13,45

2.1 Chemical reaction processes

The CO₂ absorption process into different capture solutions is governed by different equilibrium reactions, allowing different species to coexist in the system.

For amine capture solutions, it is known that the species present are RNH₂, RNH₃⁺, RNHCOO⁻, H₃O⁺, OH⁻, HCO₃⁻, CO₃²⁻, CO₂, and H₂O. R represents an alkyl/alkanol group and can belong to a primary, secondary or tertiary family of amines. Primary and secondary amines have a higher basicity than water, and thus CO₂ being the Lewis acid attacks the amine to undergo a zwitterion mechanism to form carbamates. Initially, two amines are consumed for each CO₂ molecule absorbed, generating an ammonium carbamate. However, if the CO₂ loading is increased and the amine is almost depleted, then the ammonium carbamate decomposition reaction starts occurring to form ammonium bicarbonate and carbonate species.⁴⁶ Similarly in the presence of more sterically hindered amines, like tertiary amines, water comparatively becomes a stronger Lewis base and can directly assist in the formation of ammonium bicarbonate with the amines.⁴⁷ The following chemical reactions describe the formation of different species.^37,38

Water ionization:

CO₂ reaction with water:

dissociation of bicarbonate ion:

zwitterion formation from RNH₂ and CO₂ reaction:

carbamate formation by deprotonation of zwitterion:

carbamate hydrolysis:

carbamate consumption:

de-protonation of amine:

bicarbonate formation:

The equilibrium constants of these reaction, K₁–K₁₁, can subsequently be used to understand the speciation as a function of pH, temperature and CO₂ loading. In the case of primary and secondary amines, carbamate is the main species formed when the CO₂ is bound to the amine, while tertiary amines follow a base-catalyzed hydration mechanism producing only bicarbonate ions and no carbamate, which would essentially mean that reactions (4)–(9) would not occur in the presence of tertiary amines.

Similarly, hydroxyls commonly used for DAC, form bicarbonate and carbonate species in the solution. The different species that are known to co-exist are H₃O⁺, OH⁻, HCO₃⁻, CO₃²⁻, CO₂, and H₂O, i.e., reactions (1)–(3), and (11) would only occur in such systems, and it reduces the number of buffer reactions compared to amine capture agents. In CO₂RR, we are largely concerned only with the equilibrium reactions between CO₂ and bicarbonates. It should thus be evident for researchers working with CO₂RR that c-CO₂RR represents a step-up in complexity when amines and other capture agents are involved.

Other capture agents belonging to the alkoxide family have also been recently investigated for c-CO₂RR. Kowalski et al., recently investigated potassium methoxide as a CO₂ capture agent, and observed that the methoxide ions require alcohols (other non-aqueous medium) as the solvent medium to capture CO₂ and form methyl carbonate. In an aqueous medium, the OH⁻ present in the reaction primarily behave as the capture agent forming HCO₃⁻ and CO₃²⁻, limiting the methylcarbonate production pathway. In the presence of alkoxide ions, similar reactions were described to understand the capture mechanism.¹³

Alcohol ionization:

CO₂ reaction with alkoxides:

In the alkali metal alkoxide systems, the speciation looks fairly simple with only species being the ROH, RO⁻, ROCOO⁻, CO₂, and H⁺ in addition to the alkali metal cation. The R could represent a primary, secondary, tertiary or an aromatic alcohol group.

The equilibrium equations shown here only describe the vapor–liquid equilibrium, while a full description of the system needs to add the solid components and all the reactions that lead to precipitation of species as well as dissolution of the solids in equilibrium. For simplicity, solid phases will be not discussed here but these need to be considered for any practical system.

2.2 CO₂ loading effect on speciation

The concentration distribution in the bulk of the solution is a function of the CO₂ loading when the temperature and pressure is constant. From the different equilibrium constants of the reactions shown in Section 2.1, vapor–liquid equilibrium models have been developed to predict the CO₂ absorption in different amines. Fig. 2 shows a typical chemical speciation distribution at different CO₂ loadings in amines. The CO₂ loading (mol mol⁻¹) is defined as the ratio of mol of CO₂ present (dissolved + captured) to the mol of capture agent present in the solution. The concentration of CO₃²⁻ does not vary significantly, especially for primary and secondary amines, when the CO₂ loading is varied between 0–1 mol mol⁻¹. At low CO₂ loadings, it is typically observed that all the CO₂ dissolved in the solution gets converted to carbamate as denoted by reactions (5)–(7). CO₂ loadings of 0.5 mol mol⁻¹ usually give the highest carbamate concentration, which could be also rationalized from the stoichiometry of the reaction where 2 amine molecules are required to capture 1 CO₂ molecule. Through the predicted equilibrium concentration of different species as a function of CO₂ loading, it has been shown that the carbamate consumption, as denoted by reaction (8), and CO₂ hydrolysis starts to simultaneously occur at around 0.3 mol mol⁻¹. When a CO₂ loading of 1 mol mol⁻¹ is reached the speciation is mostly dominated by bicarbonate and free CO₂ present in the system, with little to no carbamate. Thus for electrochemical c-CO₂RR, it is necessary to probe the loading of the systems to understand the speciation and also it is beneficial to be at a loading <0.5 mol mol⁻¹ if the intention is to reduce the carbamate species within the system. However in case of higher loadings, if small amounts of carbamate present near the active site of the catalyst is reduced, it is possible that the bicarbonate can equilibrate back to carbamate. The speed of this equilibration will vary and would depend on the pK_a and the equilibrium constants of the different amines. For comparison, CO₂ loadings cycle typically between 0.3 and 0.7 during the thermal regeneration of amine capture solutions, where the low loadings belong to the CO₂-lean capture solution that is fed to the absorber unit and the high loading belong to the CO₂-rich amine streams that leave the bottom of the absorber.^38,48

Fig. 2 VLE models showing equilibrium concentration of different species in liquid-phase as a function of loading in (a) 2 M MEA at 301 K which is a primary amine (lines represent model and dots represent experimental data points from Liu et al.)⁴⁹ (b) 2 M DEA at 301 K which is a secondary amine (lines represent model and dots represent experimental data points from Liu et al.)⁴⁹ (c) 2 M MDEA at 298 K which is a tertiary amine (lines represent model and dots and hollow dots represent experimental data points from Derks et al. and Jakobsen et al. respectively)^50,51 Reproduced with permission from ref. 48, copyright 2024 American Institute of Chemical Engineers.

Although most studies just look at speciation considering all amines behave similarly, but there lies differences in how these vary between primary, secondary and tertiary amines. As can be seen from Fig. 2, the VLE models of a primary, secondary and tertiary amine shows that the equilibrium concentration of the different species are different as a function of loading when the sterics of the amines change. Also, even between the same group of amines, the speciation curves could be different, which mainly arises from the different equilibrium constants of carbamate consumption and amine deprotonation in aqueous systems, K₉ and K₁₀. For instance, between monoethanolamine (MEA) and diethanolamine (DEA), at 0.5 mol mol⁻¹ of loading the carbamate concentration is slightly higher in MEA than in DEA (Fig. 2a and b). From the equilibrium constant values, it was determined that the carbamate consumption (reaction (9)) is slightly faster for DEA compared to MEA which is responsible for the difference. Also, at loading of 1 mol mol⁻¹ slightly lower concentrations of RNH₃⁺ and HCO₃⁻ was observed compared to MEA. The pK_a of DEA is lower than MEA and so it acts as a weaker Lewis base leading to weak binding of CO₂ to the amine. This leads to faster deprotonation of DEA allowing the carbamate consumption reaction to accelerate (reaction (9)). For tertiary amines like methyldiethanolamine (MDEA), the dominant species that forms is bicarbonate and thus the CO₂ loading is equivalent to bicarbonate formation with small amounts of CO₃²⁻. This small amount of CO₃²⁻ arises from the high pK_a of MDEA that accelerates the deprotonation of the bicarbonate (Fig. 2c).

When hydroxyls are used as capture agents, the main species present in the system is HCO₃⁻, CO₃²⁻ and dissolved CO₂. It is understood that at the beginning of CO₂ loading the main reaction is the formation of HCO₃⁻ with CO₃²⁻ forming at low CO₂ loadings, when the pH of the solution is highly alkaline.⁴⁵

Similarly alkoxide capture agents can be modeled to understand the speciation at different CO₂ loadings. Considering the stoichiometry of alkoxides, it requires 1 molecule of alkoxide to capture 1 molecule of CO₂, which means the highest concentration of alkylcarbonate will form at CO₂ loading of 1 mol mol⁻¹. Also, the bonding characteristics of CO₂ on alkoxides is similar to a deprotonated amino group. Recently, Kowalski et al. studied the loading of CO₂ in a methoxide capture agent in methanol solvent. Experimentally the mass change was measured while flowing pure CO₂ in 5 wt% methoxide and they observed that the change in mass was higher in the presence of methoxide in the methanol solvent compared to when no methoxide was present in the methanol. They further experimentally calculated the CO₂ loading to be 1.09 mol mol⁻¹ which is similar to the expected stoichiometric loading (Fig. 3).¹²

Fig. 3 Comparison of mass change from CO₂ loading in (a) mass change during CO₂ absorption in 100 mL of pure methanol. The net mass change is approx. 0.7 g of CO₂ after subtracting the CH₃OH evaporation amount measured under the flow of Ar and the mass change observed during CO₂ flowing into the methanol solvent. (b) Mass change during capture of CO₂ in 100 mL of 5 wt% CH₃OK in CH₃OH. CO₂ forms CH₃OCOO⁻ when reacted with the methoxide ions and the mass change for this process is about 3.5 g. This measurement is also an experimental verification of the fact that the solubility of CO₂ is different in water and in methanol. The CO₂ loading, defined as the ratio of mol CO₂ captured and dissolved to the mol CH₃OK present, was calculated to be 1.09 mol mol⁻¹ from (b).¹² Reproduced with permission from ref. 12, copyright 2024 American Chemical Society.

2.3 pH effects on chemical speciation

The pH of the solution affects the equilibrium species distribution in the capture solution. As CO₂ is an acidic gas, its capture is favored in more alkaline conditions. The concentration of the different species when plotted as a function of pH shows that at very low pH the species concentration is dominated by free CO₂ and HCO₃⁻, at intermediate pH the species concentration is dominated by the carbamate, and at very high pH the dominant species is the amine capture agent in its initial form. In MEA the highest carbamate concentration is achieved at a pH of 9.3 whereas for DEA it is at a pH of 8.85, which also corresponds closely to the pK_a of these amines which are 9.45 and 8.88 at 298 K respectively.⁵² The faster rates of carbamate consumption and CO₂ hydrolysis in DEA can also be realized from Fig. 4a and b, where it can be seen that the HCO₃⁻ and free CO₂ shoots up slightly quicker at lower pH compared to MEA. The pH is also a function of the CO₂ loading, and typically it decreases as more CO₂ is captured within the system, giving rise to the different speciation at lower pH. Also, the partial pressure of CO₂ in equilibrium with the solution is higher at higher temperatures. This translates into a higher release rate at higher temperatures resulting in a more alkaline solution, indicating temperature as an important parameter in regulating the pH. Also, the partial pressure of CO₂ increases by several orders of magnitude at a low pH, which is attributed to the presence of a high concentration of free CO₂ in the bulk of the solution. Thus in general it can be concluded that for aqueous systems a more alkaline pH is favorable for CO₂ capture, while a more acidic pH favors the release of CO₂ from the CO₂ captured adduct. Consequently, pH-swing for release of CO₂ from the captured adduct is currently being explored as an alternative to temperature-swing or pressure-swing cycles.^53–56 Similar conclusions can be drawn for alkaline CO₂ capture systems where hydroxyls are used as the capture agent. In such systems, a low pH is accompanied by the formation of carbonic acid H₂CO₃, a neutral pH consists of HCO₃⁻, while a high pH forms CO₃²⁻ in the system.⁵⁷

Fig. 4 Equilibrium speciation as a function of pH at 298 K for (a) 2 M MEA (primary amine), (b) 2 M DEA (secondary amine) and (c) 2 M MDEA (tertiary amine) obtained from VLE models. It can be seen that at a low pH, dissolved CO₂ has a very high concentration along with high concentration of HCO₃⁻. Reproduced with permission from ref. 48, copyright 2024 American Institute of Chemical Engineers.

Although the vapor–liquid equilibrium models can be leveraged to predict the speciation as a function of pH in the aqueous systems, for alkoxide/phenoxide capture agents, the solvent medium is primarily non-aqueous. Thus, it becomes difficult to define pH under such conditions and find equivalent speciation plots. However, Zhang et al. reported that the CO₂ binding strength to different alkoxides followed a linear scaling relation (LSR), and more interestingly it lied on the same LSR as the amines, which suggests that the binding of CO₂ is solely a function of the pK_a of the capture agents (Fig. 5). Furthermore, the findings of their investigation can be extended to conclude that the equilibrium constants which vary between different amines for developing the VLE models are mainly a characteristic of the changing pK_a of the amines.¹³

Fig. 5 DFT calculated values of CO₂ binding constant (pK_CO2) as a function of pK_a for different amines and alkoxides. The orange dashed line and the blue dotted line represent the data points of phenoxide and methoxide, respectively.¹³ Reproduced with permission from ref. 13, copyright 2022 National Academy of Sciences.

2.4 Temperature effects on equilibrium concentrations

The equilibrium species concentration of capture solutions as a function of temperature can be probed to investigate its influence on c-CO₂RR. The equilibrium constants of all the reactions stated in Section 2.1 are a function of the enthalpy (ΔH) and temperature (T) by eqn (14).

For amine capture solutions, at low CO₂ loading the RNHCOO⁻ concentration increases with temperature, and reaches a maximum value at 0.5 mol mol⁻¹ loading. As the loading is increased more, it has been observed that the RNHCOO⁻ concentration declines. This decline accelerates at higher temperatures which could be explained by Le Chatelier's principle, where higher temperatures shift the equilibrium to the left, which in amine capture solutions shift the equilibrium for reactions (5)–(7) towards the formation of the zwitterionic intermediate to ultimately form dissolved CO₂ and RNH₂, as shown in reaction (4). Furthermore, carbamate consumption reaction becomes more dominant at loadings >0.5 mol mol⁻¹, as K₉ is negatively related to temperature denoting the process is exothermic. In addition, the temperature was shown to have an effect on the concentration of dissolved CO₂. As the loading increased, the concentration of dissolved CO₂ was increased, and it increased further at higher temperatures. At higher loadings, it is rationalized that the amine capture agent is completely consumed and then the CO₂ undergoes hydrolysis to form bicarbonates in the system. Also, as CO₂ hydrolysis is an exothermic reaction, it facilitates the release of more CO₂ at higher temperatures. Consequently, when the partial pressures of CO₂ is compared to the CO₂ loading with the use of the temperature dependent Henry's constant, it is observed that higher temperature results in a higher partial pressure of CO₂, implying lower solubility in the solution (Fig. 6). When comparing primary and secondary amines, DEA shows a lower partial pressure of CO₂ compared to MEA, which again arises because of the faster carbamate consumption and CO₂ hydrolysis in secondary amines. Also for MDEA which is a tertiary amine, the effect of temperature on partial pressure of CO₂ had little effect compared to MEA and DEA (Fig. 6c).

Fig. 6 Partial pressure of CO₂ as a function of CO₂ loading and temperature for (a) 2.5 M MEA (primary amine) where solid lines are predicted from VLE models and hollow dots are experimental data points from Lee et al.,⁵⁸ (b) 3.33 M DEA (secondary amine) where solid lines are predicted from VLE models and hollow dots are experimental data points from Bottinger et al.,⁵⁹ and (c) 2 M MDEA (tertiary amine) where solid lines are predicted from VLE models and hollow dots and solid dots are experimental data points from Sulaiman et al. and Jou et al.^39,60 Reproduced from with permission ref. 48, copyright 2024 American Institute of Chemical Engineers.

For hydroxyl capture agents, the partial pressure of CO₂ as a function of loading is only governed by CO₂ hydrolysis and HCO₃⁻ formation, reactions (1)–(3). The partial pressures at similar conditions of pressure and temperature are usually higher for such systems implying more free dissolved CO₂ present in the solution.

For alkoxide capture agents, the partial pressure of CO₂ is one order of magnitude lower than amine/hydroxyl capture agents. This difference arises mainly because of the higher Henry's constant when using methanol as solvent and also a higher solubility of CO₂ in methanol.

Temperature has a significant effect on the chemical speciation of the system and can be captured in the vapor–liquid equilibrium model used while studying c-CO₂RR systems. At high temperature conditions, the CO₂-bound adduct can release free CO₂ and regenerate the capture agent. This is known as temperature-swing capture agent regeneration. In conventional carbon capture and utilization processes, the capture step is followed by a high-temperature capture agent regeneration process while simultaneously releasing a concentrated stream of CO₂ for further storage or upgrading. Similarly, in c-CO₂RR, the captured CO₂-adduct will decompose at higher temperatures leading to low c-CO₂RR performance, however, consequently higher temperatures will release more free CO₂ in the system that can compete for the active sites of the catalyst for further reduction. Banerjee et al., in their study showed that by increasing the temperature of the capture solution during electrolysis, they can achieve a higher partial current density of CO using silver catalyst. They concluded that this behavior is mainly due to the release of free CO₂ in the system as they increase the temperature from 20 °C to 40 °C.⁴⁸ In such cases, at higher temperatures the performance of c-CO₂RR can appear to be improved, but this is not to be confused with the direct reduction of the CO₂-bound adduct. Therefore, understanding the vapor–liquid equilibrium effects discussed throughout Section 2 will be important in identifying the active carbon species getting reduced as the field of c-CO₂RR evolves.

3 Identifying the source of carbon in c-CO₂RR

The VLE models can provide a clear understanding of the speciation and its variation with the pH, temperature and CO₂ loading. Leverick et al. quantified the speciation of amines with different pK_a through NMR analysis to verify the speciation with the VLE-models and identify the source of carbon. There have been three main approaches that have been undertaken to identify the active carbon species which are by leveraging characterization techniques to correlate with the VLE-models of different amines, continuum modeling, and correlating the two-film theory with reaction transport model. The last method provides a robust way for quick screening of different capture agents, especially if VLE models are not easily accessible. This section goes through each method in detail.

3.1 Characterization techniques coupled with VLE-models

Recently, the most detailed study to identify the source of carbon in c-CO₂RR systems with the VLE-models have been performed by Leverick et al. using amine capture agents. Using a H-cell reactor with an Ag catalyst, they investigated the faradaic efficiency (F.E.) of CO in the presence of amines with varying pK_a and different propensity to form carbamates. They reported that at pH > 10 the F.E. CO is <0.1% and the F.E. CO increases to >20% when the pH is decreased, and it is ubiquitous to all the amines tested (Fig. 7c). To rationalize this behaviour they correlated the experimental data to the VLE-models, and explained that the carbamate concentration decreases at lower pH but the concentration of HCO₃⁻ and dissolved CO₂ increases, which also increases the CO₂ partial pressure (Fig. 7a and b). The changes in carbamate concentration were further tracked with ¹³C NMR along with the different species present in the solution. Furthermore, they showed that the partial current density of CO (j_CO) had a first-order dependence with the partial pressure of CO₂ verifying that the source of carbon is the dissolved CO₂ in the solution and not the carbamate itself. In cases, where they did not see a linear increase of j_CO with decreasing pH, they discussed the possibility of the different amines having different activity towards HER, which could suppress the production of CO.⁶¹

Fig. 7 (a) Concentration of carbamate from NMR vs. pH for 2 M KCl + 2 M amine with DMAE (square), AMP (right-pointing triangle), 2A1P (left-pointing triangle), and MEA (circle). (b) Concentration of bicarbonate from NMR vs. pH. (c) FE of CO (filled symbols) and H₂ (open symbols) production vs. pH from potentiostatic measurements at −1.30 vs. SHE-iR.⁶¹ Reproduced with permission from ref. 61, copyright 2023 American Chemical Society. (d) j_CO at −1.1 V vs. normal hydrogen electrode (NHE) of Ni–N/C and cAg for the c-CO₂RR in CO₂-absorbed MEA (1–5 M) solutions with purging Ar.¹⁵ Reproduced with permission from ref. 15, copyright 2022 Royal Society of Chemistry.

Similar studies were performed by Kim et al. using a membrane electrode assembly with a single atom Ni catalyst and a Ag foil as catalyst. With both catalysts, they probed the source of carbon for the produced CO by changing the concentration of the carbamate in the system and also by changing the temperature of the reactor. They also quantified the carbamate concentration in the system using ¹³C NMR and observed a zero-order dependence on the F.E. CO with the carbamate concentration, meaning the carbamate did not act as the active species for reduction (Fig. 7d). They further observed an increase in the F.E. CO as the temperature increased, which they attributed to more dissolved CO₂ released in the system at higher temperature.¹⁵

Thus, VLE models can be an important source to identify the active species. Changing the CO₂ loading, temperature, pH, and pressure of the system affects the species distribution and thus the VLE models in conjunction with the multi-variable dataset can rationalize the results to provide a basis for identification. In c-CO₂RR studies, keeping track of these parameters can help in understanding how the amount of dissolved CO₂ changes in the system, and these changes can be correlated with the changing faradaic efficiency/partial current density of the products. If there were to be any direct reduction of the CO₂-bound capture agent, the total partial current density of the reduced product should have a dependence on the carbamate concentration of the solution.

3.2 Continuum models

Safipour et al. investigated CO₂RR with MEA solution in a flow-cell configuration to investigate the species getting reduced. By increasing the concentration of MEA they observed an increase in HER activity and reduction in CO production activity. To understand this effect, they developed a 1D continuum model to determine the concentration distribution of different species within the electrical double layer. Assuming concentration dependent Tafel kinetics, the electrode reactions were modelled. It was found that the CO₂ concentration at the surface of the electrode is lower in the presence of amines compared to conventional CO₂RR, which was responsible for the lower CO₂ conversion rates. The high activity for HER was related to the presence of high concentration of MEAH⁺ at the surface which can act as a proton source. While a higher concentration of carbamate was observed to be present at the surface, the lower partial current density of CO compared to amine free electrolytes ruled out the possibility of any carbamate reduction (Fig. 8).¹⁷

Fig. 8 Simulated boundary-layer concentration profiles at an applied potential of −1.5 V vs. SHE, where the zero coordinate corresponds to the Ag cathode surface. (a) Comparison of the dissolved CO₂ concentration profile in 0.1 M KHCO₃ electrolyte with and without 0.2 M MEA additive. (b) MEA, MEAH⁺, and carbamate concentration profiles in 0.1 M KHCO₃ + 0.2 M MEA electrolyte, showing the depletion of MEAH⁺ and accumulation of MEA and carbamate relative to their respective bulk values. Carbamate concentration is elevated due to the reaction of locally generated MEA with aqueous CO₂, causing the result in (a).¹⁷ Reproduced with permission from ref. 17, copyright 2023 American Chemical Society.

Thus in different c-CO₂RR settings, a similar approach can be taken to estimate the captured CO₂-bound adduct concentration at the surface of the electrode and compare it with the partial current density of the reduced product to estimate if there is any CO₂-bound capture agent reduction. Physics-based continuum models is therefore a key tool in establishing connections between device level performance metrics (like current density and product selectivity), electrode characteristics, and operational parameters. In c-CO₂RR systems, continuum models can provide information regarding polarization behaviour, transport of products and reactants, and limitations in operational lifespan, which when complemented with experimental investigations can expedite the development of c-CO₂RR technologies.⁶²

3.3 Two-film theory combined with reaction transport model

Recent investigations by Shen et al. and Kowalski et al. used the rotating cylinder electrode (RCE) setup to study the activity for c-CO₂RR.¹⁶ Shen et al. looked into different mass transport regimes by changing the rotation speed and observed that the partial current density of CO is linearly dependent with the partial pressure of CO₂. They further developed a model for the partial current density of CO under limited mass transport regime of dissolved CO₂ to correlate j_CO to the amount of dissolved CO₂ in the system. Kowalski et al. used the same model to study the activity of c-CO₂RR and compared methoxide capture agents to hydroxyls and amines. They also reported dissolved CO₂ as the primary species being reduced in c-CO₂RR. This model is robust and unique where it measures the partial pressure of CO₂ from the headspace in operando to identify the active species getting reduced and allows for quick screening of different activity descriptors. The two main parameters that are relevant to understand the model are (i) the mass film transfer coefficient at the gas–liquid interface and (ii) the mass film transfer coefficient at the liquid–electrode interface.¹⁶

3.3.1 Gas–liquid interface. CO₂ absorption is a reversible process characterized by an equilibrium constant. The equilibrium between hydroxyl ions and dissolved CO₂ is given by reaction (11) for alkaline solutions. The absorption of CO₂ in alkaline amine solutions leads to the additional formation of the corresponding carbamate and ammonium ions (reactions (4)–(7)). The experimental challenge in c-CO₂RR systems is measuring the concentration of dissolved CO₂ which could be a source of carbon for the reduced products and to decouple the dissolved CO₂ from the CO₂-bound capture agent reduction. The CO₂ present in the solution will also be in equilibrium with the partial pressure of CO₂ (g) in the electrochemical cell. When a fast equilibrium is reached at the gas/liquid interface (eqn (15)), the partial pressure of CO₂ can be used to estimate the concentration of CO₂ dissolved in the solution using Henry's Law, given by eqn (16)

CO₂(g) ⇌ CO₂(dissolved) (15)

The Henry's constant, H_CO2 is a function of the temperature, pressure and also the solvent that is used in the system. Kowalski et al. in their study also pointed out that when looking at c-CO₂RR systems under different capture agents such as amines and alkoxides, the solubility of CO₂ is different in the different solvents and this changes the Henry's constant value to be used in eqn (16). The difference in solubilities in different media must be taken into account while investigating c-CO₂RR. This can be achieved through the use of VLE models and multi-scale models.

3.3.2 Liquid–electrode interface. The liquid–electrode interface is important in determining the flux of dissolved CO₂ to the electrode and realizing the local reaction conditions. Jang et al. developed a gas-tight rotating cylinder electrode (RCE) reactor which has well-defined mass transport properties and can quantify both gas and liquid products. In their work they characterized the mass transport properties of the RCE reactor and developed a Sherwood number relationship (eqn (17)) that universally holds true regardless of the experimental conditions including the effects of reactants, their concentration, electrode rotation, and temperature.⁶³ Using this relationship, both Shen et al. and Kowalski et al. determined the flux of CO₂ to their electrodes from the amount of dissolved CO₂ in their system. The equation was used as a reference to determine the convective transport-limited rates of reactions that cannot be readily obtained from experiments due to the complexity of c-CO₂RR systems including multiple electrocatalytic and buffering reactions. For instance, the limiting current density (j_lim) of CO in electrochemical reactors cannot be measured explicitly with the dominating hydrogen evolution and is complicated to estimate with equilibrium reactions between CO₂ and absorber adducts. However, the maximum flux of CO₂ (J_CO2,max) that can reach the electrode surface can be determined from eqn (18), assuming that the local concentration of CO₂ at the electrode is zero. This calculated maximum flux of CO₂ or its corresponding maximum partial current density to CO (j_CO,max) was sufficient to explain the relationship between the observed partial pressure of CO₂ and j_CO to a large extent (Fig. 9) considering the low concentration of dissolved CO₂ in this system (eqn (19)). The potential impact of dissolved CO₂ reacting directly with hydroxyls generated at the cathode electrode during electrochemical reduction reactions in aqueous electrolytes was disregarded. These chemical reactions can lead to lower maximum current densities compared to what was observed in the absence of such reactions. However it was rationalized that within the boundary layer as dissolved CO₂ becomes depleted, it is probable that bicarbonate and carbamate will decompose (reversing reactions (5) and (6), and (7)) to generate additional CO₂, offsetting the CO₂ consumed at the electrode surface. Local heating induced by ohmic resistance and the heating of the vicinity of the electrode due to the overpotential required to drive the reaction can facilitate further desorption of CO₂ from the amine absorber near the cathode electrode. Nonetheless, under steady-state conditions, the maximum flux of dissolved CO₂ from the electrolyte bulk would not exceed that given by eqn (18) and (19).

Sh = 0.204Re^0.59Sc^0.33 (17)

J_CO2,max = k_m,CO2C_CO2,bulk (18)

j_COmax = 2FJ_CO2,max (19)

Fig. 9 Left panel: Experimental and transport model-based maximum partial current density for the reduction of dissolved CO₂ in a solution at equilibrium with a partial pressure of CO₂ determined using a gas chromatogram. Experiments containing amines were conducted using 0.099 M KClO₄ + 0.001 M KOH. All the experiments were conducted using an electrodeposited Ag catalyst as the working electrode, a Pt counter electrode, and a Ag/AgCl reference electrode. Right panel: Applied potential versus partial CO current density for experimental data in the left panel.¹⁶ Reproduced with permission from ref. 16, copyright 2023 Elsevier.

4 Reactor and catalyst design for c-CO₂RR

Reactor and catalyst design have been extensively studied in the past for CO₂RR, but there is not much consensus on how these parameters play a role in c-CO₂RR. Indeed, many strategies have been suggested to improve CO₂RR performance. However, the community has now started to recognize that electronic modifications (d-band center tuning), alloying, and interface engineering alone are not sufficient to deliver a scalable CO₂RR technology. Jang et al. has recently shown that transport alone has a larger impact in product selectivity during CO₂RR on Cu than interface engineering (cation effects Cs⁺, K⁺, Na⁺etc.) or the predominant surface termination (i.e. comparison of pc-Cu, Cu(111), Cu(100), and Cu(751)).⁶⁴ Much of what has been claimed as catalytic enhancements are not maintained when the electrode areas are increased, indicating that the catalyst has a limited effect on the performance of the CO₂RR electrolyzer.^65,66 The explanations based on morphology control, d-band center tuning, and interface engineering often emerge as an explanation to an experimental observation in a restricted set of experimental conditions. Reproducibility across labs working on CO₂RR is an issue.⁶⁷ We note that this is not limited to CO₂RR, but that this is indeed a widespread issue in electrocatalysis.⁶⁸ Our purposeful reduced discussion of catalyst optimization strategies in this review originates from our interest in discussing at depth what we consider has been phenomena ignored in the field, that is, the role of transport in electrocatalysis for c-CO₂RR applications. Nevertheless, in order to provide a balanced discussion, other optimization strategies that do not focus on the role of transport are also discussed in Section 4.2. Additionally, in this section, we reflect on the effect that reactor geometry has on product selectivity where often different reactors result into different product distribution when using the same catalyst, suggesting reaction kinetics and reactor kinetics are convoluted in most electrocatalytic studies.

In this section, we will dive deeper into these discussions to understand how much knowledge from CO₂RR can be translated to c-CO₂RR, if at all.

4.1 Role of catalyst morphology and geometry

In the past, a bulk of research for CO₂RR has mostly focused on selecting catalyst based on their electronic properties and DFT calculations of binding energies of different intermediates. Two main descriptors for CO₂RR has been the binding energies of *CO and *H,^69,70 but complexities arising from the presence of capture agents raises questions on whether CO₂RR descriptors could be translated to c-CO₂RR. Recently, Choi et al., in their work showed that catalyst in the presence of amine capture agents can undergo corrosion, so binding energies of the amine to the surface is an important parameter when looking into c-CO₂RR.⁷¹

Early works by Hori et al. was conducted in an H-cell configuration where the CO₂ dissolved in the solution is reduced to products.⁷² This happens in the two-phase solid–liquid interface. Hori broadly divided metals into two categories; (i) CO producing metals like Au, Ag, Cu, and (ii) HCOO⁻ producing metals like Sn, Pb, In, Cd, Hg. To operate at higher current densities reactor configurations were later modified to gas-diffusion electrolyzers which operate at the three-phase boundary of solid–gas–liquid. However, with new reactor geometries researchers did not observe changes in selectivity of the catalysts, meaning the intrinsic activity of these catalysts at the two-phase boundary and the three-phase boundary remained the same. Although early reports suggest that CO₂RR catalysts do not work well in c-CO₂RR systems, there is still value in working with these catalysts to see if there is a suitable catalyst that can activate the CO₂-bound captured complex to make reduction products. The idea is that if a suitable catalyst is identified then there is value in developing a new reactor design like the gas-diffusion electrodes to operate at higher current densities, however it is still likely that in the case of c-CO₂RR the reaction will still occur at the two-phase boundary where the catalyst will be in contact with the liquid as in the H-cells.

Although most studies have focused on catalyst activity, recent studies have shown that catalyst morphology and geometry also plays an important role in determining product distribution and tuning them could increase the efficiency of electrochemical systems. For CO₂RR, several studies have already demonstrated that morphology of the catalyst plays a role to increase faradaic efficiency however for direct reduction of the captured CO₂-bound adduct, there are additional buffer reactions occurring at the interface which makes it harder to delineate the mechanisms. In this section, we will look into the direct reduction of the CO₂-bound capture agents from a catalyst perspective, where the two most important things is determining the active species at the surface of the electrode and investigating the effect of the geometry of catalyst on the electroreduction of these CO₂-bound capture agents.

The transport of species from the bulk of the electrode to the catalyst surface (external mass transport) and from the catalyst surface to inside the pores (internal mass transport) both influences the product distribution in CO₂RR and simultaneously can influence c-CO₂RR. Studies that have used porous catalysts to increase the CO₂ conversion argued that a porous structure by means of internal mass transport can increase the OH⁻ concentration within the pores and can suitably create a local CO₂ environment which increases the CO₂ utilization. The pH also plays an important role as a high pH reduces HER by reducing the concentration of protons.^73,74 In 1989 Hori et al. showed that a locally high pH at the electrode–electrolyte interface can facilitate C–C coupling to form multicarbon products.⁷⁵ A highly porous electrode have been argued to enhance this effect creating a very high pH within the pores by means of internal mass transport. Thus studies have looked into different mesoporous, microporous and nanoporous structure for CO₂RR. Macropores can shorted the ion diffusion distance by increasing the buffer-volume of the reactions, while mesopores facilitate mass transport properties and micropores help in increasing the active surface area.⁷⁶

Tuning the morphology of the catalyst can thus affect the local electronic structure by altering the grain boundaries, edges, facets, and corners. This can modify the pH, mass transport, adsorption/desorption kinetics, and local CO₂ concentration, at different scales. Different morphologies including nanocavities, mesopores, nanocages, crystalline porous frameworks and porous thin films have been studies in the past for CO₂RR.⁷⁶

Ma et al. using Cu nanowires showed that the structure of the catalyst can regulate the pH causing the changes in the internal transport of the HCO₃⁻ and OH⁻ species in and out of the pores, which altered the selectivity leading to more preferred CO dimerization pathway (Fig. 10a).⁷⁷ Zhuang et al. showed that a Cu catalyst structure with nanocavity can promote the C–C coupling between the C₂ and C₁ intermediates to form C₃ products.⁷⁸ They observed that a nanocavity could concentrate C₂ species inside the cavities through steric confinement, increasing the C₃ production inside the cavities (Fig. 10b). Mesoporous structures, mainly carbon-based, were also investigated for CO₂RR, as larger surface areas could potentially increase the active sites.^79–81 Although pristine carbon is known to be electrochemically neutral, chemical doping using N, B, F, and S atoms have been used to break the neutrality to generate active sites for CO₂RR.^82,83 Hursan et al. showed that morphological attributes like porosity and pore size can have an affect on the performance of N-doped catalysts on CO₂RR.⁸⁴ They observed that the nonporous N-doped catalyst was the least hydrophobic and produced the highest HER, compared to the porous films which enhanced CO₂RR performance. To describe this behavior, they correlated wettability with surface energy. Changes in wetting properties influences the dynamics of gas bubble formation on the electrodes. The non-porous samples can result in large bubbles that can block the active sites of the catalyst for CO₂ reduction. They highlight that the release of gas bubbles which are dependent on the wetting properties of the catalyst can bring about changes in mass transport leading to changes in the catalytic current densities. Hydrophobic surfaces enhance gas bubble diffusion and release. Other reports also suggest that the surface hydrophobicity can diminish the catalyst-water contact inducing a high concentration of OH⁻ near the electrode, increasing the local pH and decreasing HER⁸⁵ (Fig. 10c). The wetting properties of a material are usually determined by the interfacial interaction forces between a solid and the fluid. These van der Waals forces, which are dependent on polarity, dipole moment, structures, and charging states, develop at the solid–liquid interface. The net surface energy then determines the wetting properties of the material.^85,86 It has thus been observed that higher surface energy generally leads to more hydrophilic surfaces. However, during electrochemistry, these forces will act under an electric field which can further modify the interaction of these forces. Different studies have reported that the electrowetting properties of electrodes can be adjusted by tuning (i) the dielectric properties of the electrode and electrolyte and (ii) choosing a catalyst PZC such that there is no excess-charge at the solid–liquid interface.⁸⁵ Therefore, wetting properties have potential to further enhance the performance of c-CO₂RR and must be explored during the development of c-CO₂RR technologies. Hierarchical porosity have also been investigated by Ni et al. where they used an F-doped hierarchical nanocage structure and compared it to F-doped solid carbon and F-doped carbon with mesopores. They observed that the hierarchical structure possessed the largest surface area among all and also the highest F.E and current density for CO, which was rationalized by increased mass transport and low diffusion resistance from the hierarchical structure (Fig. 10d).⁸⁷ Dutta et al. showed that porous thin films, like Cu foam, have a difference in product selectivity where they make more C₂H₄ and C₂H₆ compared to CO in flat Cu. The catalytic difference was attributed to the presence of (100) facet in porous electrodes which have an intrinsic activity for C–C coupling (Fig. 10e).⁸⁸ Existing knowledge on CO₂RR can thus provide a basis on the activity of catalysts with different geometries for c-CO₂RR and should be further investigated to understand how the presence of amines or other capture agents might alter the product distribution.

Fig. 10 (a) Schematic of the diffusion of electrolytes into Cu nanowire arrays.⁷⁷ Reproduced with permission from ref. 77, copyright 2016 Wiley (b) energy profile for the formation of the C3 intermediate. Red, grey, and orange balls represent oxygen, carbon, and copper, respectively and schematic showing that the cavity confinement effect boosts C₂ species binding and further conversion to C₃ and the corresponding faradaic efficiency.⁷⁸ Reproduced with permission from ref. 78, copyright 2018 Nature (c) schematic showing the activity difference on flat and porous NC and TEM image of NC-27.⁸⁴ Reproduced with permission from ref. 84, copyright 2019 Elsevier (d) TEM image of F-CPC. The insert shows the schematic structure and CO₂RR performance of various catalysts at −1.0 V.⁸⁷ Reproduced with permission from ref. 87, copyright 2020 American Chemical Society (e) SEM image of an Ag-IO film with CO FEs and ECSA-normalized CO current of Ag-IO films with varied RF.⁸⁸ Reproduced with permission from ref. 88, copyright 2016 American Chemical Society.

A few studies have been conducted while flowing CO₂ in the presence of amines. Although they are not analogous to c-CO₂RR conditions, they can help understand the role of amines in electrochemical reduction reactions. Chen et al. reported an increase in formate production in porous In catalyst and an increase in CO production in porous Ag and Zn catalyst in a 30 wt% MEA solution.⁸⁹ Hossain et al. prepared nano-dendrites of Cu, Au, and Ag on a glassy carbon electrode and investigated CO₂RR in a 0.05 M MEA solution, where they observed an enhancement in current densities with the nanostructured catalyst.⁹⁰

The only study that reports the effect of porosity under c-CO₂RR environment is by Kowalski et al. where they investigated the current densities and faradaic efficiencies between sputtered Ag electrodes and electrochemically roughened Ag electrodes (Fig. 11). They found that although c-CO₂RR predominantly produces HER, increasing the porosity had an effect in slightly enhancing the CO production. They looked into amines, hydroxyls and methoxides as capture agents and in all of them the faradaic efficiency and j_CO of CO increased with porosity. Maximum F.E. CO of ≈1.5% and ≈1% were reported for the electro reduction of methyl carbonate CH₃OCOO⁻ and bicarbonate HOCOO⁻, respectively (Fig. 11a and b). The F.E. CO for ammonium carbamate although increased, the maximum remained very low at ≈0.03% (Fig. 11c). They further normalized their j_CO values to the mass transport limited j_CO from the flux of dissolved CO₂ and observed that the normalized j_CO values crossed the limiting transport line for reduction of dissolved CO₂ (Fig. 11d and e). They hypothesize that with porous electrodes the local pH within the pores becomes higher than at the entrance of the pores. This high pH suppresses HER resulting in more available sites for CO₂ reduction to produce CO. Although the increase in j_CO was higher than the calculated limiting value they still reported it as dissolved CO₂ reduction. Furthermore, from the partial current density obtained from the different capture agents it was concluded that H₂ was mostly produced from outside the pores as no changes in j_H2 was observed with the as-sputtered electrodes and the electrochemically roughened electrodes. Whereas the increase in CO was the result of more CO₂ being reduced within the pores. Thus internal mass transport and local reaction environment within and around the pores also play a role along with external mass transport in c-CO₂RR systems.

Fig. 11 The faradaic efficiency of CO obtained from as-sputtered Ag electrode compared to roughened Ag electrode at different applied potentials (vs. Ag/AgCl) for (a) KCH₂OCO₂ (methyl carbonate). (b) KHOCO₂ (bicarbonate) (c) NH₄–NH₂CO₂ (ammonium carbamate). Comparison of (d) j_CO and (e) j_H2 between the sputtered Ag electrodes (solid dots) and the roughened Ag electrodes (hollow triangles). (f) The j_CO is further normalized against the maximum flux of CO₂ and it is observed that the rough electrodes are able to cross the transport-based limiting j_CO which was rationalized from the local release of CO₂ within the pores.¹² Reproduced with permission from ref. 12, copyright 2024 American Chemical Society.

In electrocatalysis, recent investigations have deployed the use of the Thiele modulus to understand the relative balance between kinetic and mass transport resistances inside catalyst pores. The Thiele modulus in catalysis is used to describe the relation between reaction rates and diffusion rates. For large values of Thiele modulus, the overall reaction rate is constrained by diffusion resistance, and the reaction predominantly occurs within the outer catalyst layers. Conversely, when Thiele modulus values are small, the overall reaction rate is restricted by the catalyst volume, specifically the total accessible pore surface area. Traditionally, this concept has been mostly used for fuel cell reactor development across multiple length scales, however its use in other electrochemical systems is somewhat limited. Recently, Thiele modulus analysis have been explored by a few researchers to understand the effectiveness factor of their catalyst. Kim et al. and Wang et al. used Thiele modulus modeling in gas diffusion electrodes for CO₂ reduction and also developed analytical models to describe catalyst effectiveness.^91,92 Similarly Wan et al., investigated the potential dependent Thiele modulus models to probe the utilization of pores as a function of catalyst particles and reactant properties. Using a 1D reaction diffusion model through the pores they developed the Thiele modulus model and delineated external and internal mass transport to effectively outline design principles for the electrocatalysts.⁹³ The Thiele modulus can also be defined for c-CO₂RR to understand and quantify the effectiveness of morphology and catalyst geometry. It will further be effective in understanding the relative timescales of reaction and diffusion within the pore of the catalyst to rationalize the increased CO₂ reduction products observed within the pores under c-CO₂RR conditions.

Since all studies point towards the free CO₂ as the active species for reduction, the knowledge existing for CO₂RR on the effect of porosity can further be used to understand c-CO₂RR mechanisms. However, one study by Ma et al. reports the direct reduction of carbonate to formate which is discussed in detail in Section 5. The local OH⁻ concentration dynamically influences the buffer reactions, and when imagined in the context of c-CO₂RR there are additional buffer reactions that make these systems more complex. Suter et al. using a finite element model investigated different porosity and species concentration within the pores to predict CO and H₂ current densities. The model predicts that OH⁻ and CO₃²⁻ concentration within the pores are higher than at the surface resulting in a high alkaline pH, which is more obvious in a thicker film, as can be seen from Fig. 12. Predicted CO FEs and current density increased with increasing RF factor which also aligned with experimental findings.⁹⁴ Thus, a high alkaline pH is understood to suppress HER within the pores allowing more available sites for CO₂ reduction. The same study also coupled the effect of pore diameter and film thickness to investigate the changes in CO FE and observed a linear dependency of roughness factor (RF) with film thickness and non-linear dependency of RF with pore diameter. They concluded that multiple combinations of pore diameter and film thickness can give similar faradaic efficiency. For instance, a FE CO of 90% can be achieved from a pore diameter of 200 nm and film thickness of 10 μm or a pore diameter of 100 nm and film thickness of 7 μm. Thus it was concluded that roughness factor alone is not enough to characterize a porous film.⁹⁵ Experimentally, it is difficult to determine the local OH⁻ or CO₂ concentration inside the pores. A recent study by Bohme et al. developed a pOH sensor where a confocal laser scanning microscopy technique was used to map the pOH around the electrode during CO₂RR. They were able to experimentally verify that the OH⁻ concentration is high within the pores compared to outside the pores as suggested by many other studies mentioned before.⁹⁶ When extending it to c-CO₂RR, the high alkalinity of the pores can also suppress HER and promote CO₂ reduction, as has been observed by Kowalski et al. through their experimental findings. Furthermore, the additional buffer reactions of amine deprotonation and carbamate consumption is also affected with changes in local pH that contribute to altering the selectivity of the products formed. Thus, it is important to be critical about all these factors while investigating different activity and stability descriptors for c-CO₂RR.

Fig. 12 (a) Concentration profiles along the z-axis, where z = 0 is the innermost point inside the pore and z = 1 is the entrance of the pore, of OH⁻, CO₃²⁻, HCO₃⁻ and CO₂versus their bulk concentration at a potential of −0.8 V vs. RHE. (b) Partial CO current density and (c) CO faradaic efficiency for Ag-IO with pore diameters of 150–400 nm and film thickness of 3, 6 and 15 μm.⁹⁵ Reproduced with permission from ref. 95, copyright 2019 Royal Society of Chemistry.

4.2 Tuning catalyst properties for c-CO₂RR

Designing catalyst for c-CO₂RR that can tune (i) electronic properties, (ii) binding properties, and (iii) intermediate concentration at the electrode–liquid interface will be necessary as the field evolves. Electronic properties in particular has been shown to play an important role in modifying the binding properties of key intermediates in CO₂RR and can have the same effect on c-CO₂RR. The upper shift of d-band energy level results in increased binding strength of the adsorption intermediates. This understanding has been used at different scales to tune the selectivity of reduction products, from single atom catalysts (SAC), to nanoparticles, and surfaces with different terminations.⁹⁷

Different SACs have been designed with different metal centers like Ni, Fe, Co, In, Bi, and Sn for producing CO and HCOO⁻.^97–99 These dispersed metal sites along with their coordination environment create M–N₄ moieties of superior performance over their metal counterparts. In this unique assembly the electronic structure of the metal center is modified by the adjacent coordination environment and this have been shown to modify the catalytic activity, selectivity and stability. Further optimization of electronic structure has been studied by looking into vacancy defects to generate the Ni–N₃–V moiety or through heteroatom doping to generate Ni–N₃P. Both of these approaches have been suggested to have superior performnace for c-CO₂RR compared to the Ni–N₄.^100,101

For nanoparticles it is generally observed that the electrocatalytic performance is determined by the amount of undercoordinated sites where the CO₂RR intermediates preferentially bind and is dependent on the ratio of surface to bulk atoms. Gu et al. observed that the Ni-based nanoparticles had lower faradaic efficiencies compared to Ni SAC.¹⁰² However, this behavior varies depending on the metal that is being used as the catalyst. Huang et al. investigated the production of HCOO⁻ on In based catalysts and found that In₂O₃ nanocrystals which are 15 nm in diameter were more active than the nanocrystals which are 200 nm or In foil, which they attributed to the increased uncoordinated sites with lower particle size. On reducing the nanoparticle size to 5 nm they observed that the partial current densities for HCOO⁻ decreased. This shows that smaller may not be always better in catalyst design and that a deep understanding of structure–property relationship is important when designing catalysts.¹⁰³

Strategies like alloying have also been explored in numerous CO₂RR studies in the past to tune the spatial distribution of reduction intermediates and enhance the selectivity towards certain products.^104–106 For instance, Gao et al. designed a CuAg catalyst to improve the selectivity of CO₂ reduction to ethylene. Using operando Raman spectroscopy they showed that the bimetallic catalyst produces more *CO binding sites that can then hydrogenate at neighboring sites and couple to make C₂ products. This is often coupled with interface engineering strategies, as discussed in Section 5, to modify the local concentration of the active species and its intermediates.¹⁰⁷

Although, these strategies are useful to design catalyst for improved activity and selectivity, however it is often argued for scalable technologies transport plays a more significant role than the catalyst itself. Hursan et al. in their study highlighted on this effect. Using N-doped carbon they designed electrodes of different porosity and depending on the pore size they observed different selectivity towards CO and H₂ formation in CO₂RR. They concluded that porosity modifies mass transport and consequently the CO₂ adsorption properties that primarily governs CO₂RR. Therefore, morphological attributes that determines transport affect the catalytic performance and needs to be given more importance while designing large scale electrolyzers.⁸⁴

4.3 Engineering reactor design to tune mass transport and kinetics

The electrolyzer design and cell configuration are important considerations in c-CO₂RR. The electrolyzer architecture and the operating conditions (e.g. temperature, pressure, potential, current) influence the transport of the active species and products between the bulk of the solvent and the electrode surface, as well as the temporal speciation of the capture agent and CO₂ adducts. The reactor design will also determine the rate of equilibration between the species in the gas phase and those dissolved in the liquid. Recent studies on c-CO₂RR have been carried out in different reactor designs which have different mass transport characteristics. The commonly used electrochemical reactors, the H-cell and the flow cell configuration, have also been used extensively for CO₂RR in the past and recent work has shown that hydrodynamics in the cell indeed determine the product distribution obtained.^63,108 To date, very few works have been reported for c-CO₂RR but it can be anticipated that the type of reactors used will also impact the product distribution in this reaction. In some of the c-CO₂RR systems discussed in this section, the membrane separating the cathode and anode compartments is used to control the generation and transport of protons that shift the equilibrium and release CO₂ in the proximity of the catalyst (reverse direction of eqn (5)–(7)). Electrolyzer design is thus tightly tied to the performance of c-CO₂RR systems and needs to be carefully studied.

This section discusses four types of electrochemical cells (H-cell, compression, zero gap, and rotating electrode) that have been used for activity benchmarking in c-CO₂RR, as shown in Fig. 13. Leverick et al., used an H-cell to investigate amines of varying pK_a and sterics and reported the differences in faradaic efficiencies for producing CO as a function of pH and temperature.⁶¹ They observed that the partial current density of CO had a weak amine dependence but had a first order dependence with the partial pressure of CO₂ in the system, which suggested that the active species undergoing reduction is the dissolved CO₂ and not the amine. Safipour et al., investigated CO₂RR in the presence of mono-ethanol amine (MEA) in a sandwich-type compression cell configuration (Fig. 13b) and observed a maximum partial current density for CO of 5 mA cm⁻², which decreased with the increasing concentration of MEA. With the help of a 1D continumm model they report that the addition of MEA increases the concentration of MEAH⁺ which acts as a proton source for H₂ and reduces the local concentration of CO₂, which in turn reduces the partial current density of CO, with no evidence of any carbamate getting reduced.¹⁷ To increase the amount of reduced products from CO₂ capture solutions, the Berlinguette group designed the bicarbonate electrolyzer as shown in Fig. 13c, where they used a bi-polar membrane or a proton exchange membrane in a membrane electrode assembly for supplying protons and locally release CO₂ from bicarbonate in the solution and promote CO₂RR.^18,19 This increased the flux of CO₂ to the surface of the electrode and high faradaic efficiencies for CO was achieved, but at a penalty of high input energy. The thermodynamic minimum energy for converting CO₂ to CO is −1.33 V whereas the input voltage required to reach a current density of −100 mA cm⁻² was −4.4 V in their electrolyzer, while lab-scale CO₂RR electrolyzers have already demonstrated to achieve −200 mA cm⁻² at cell voltages below −3 V.^109,110 A similar setup was also used by Kim et al. to study the reduction of carbamates using single atom Ni catalyst and a high CO selectivity was achieved with FE of ≈69%.¹⁵ Shen et al., using the RCE cell investigated different mass transport regimes by changing the rotation speed and observed different faradaic efficiencies and partial current densities of CO, and further concluded that the produced CO is from the dissolved CO₂ within the system. The RCE cell system is unique in that it measures the partial pressure of CO₂ in the headspace of the cell, which is in a pseudo-equilibrium with the solution. The partial pressure of CO₂ can be used to estimate the amount of free and dissolved CO₂ in the bulk of the electrolyte and facilitates the discrimination between different carbon sources during c-CO₂RR.¹⁶

Fig. 13 (a) Conventional H-type cell configuration used by Leverick et al. to study the j_CO, j_H2 and F.E. CO and FE H₂ of c-CO₂RR at different pH and applied potentials for amines of different pK_a and sterics.⁶¹ Reproduced with permission from ref. 61, copyright 2023 American Chemical Society (b) a schematic of the compression cell type configuration used by Safipour et al. to investigate the species distribution near the surface of the electrode while studying CO₂RR in the presence of MEA.^17,111 Reproduced with permission from ref. 17 and 111, copyright 2023 American Chemical Society and 2012 Royal Society of Chemistry (c) the zero-gap membrane electrode assembly used as the bicarbonate electrolyzer by Li et al. and Lees et al. to release CO₂ in the bulk of the solution by supplying protons to increase the FE CO.^18,19 Reproduced with permission from ref. 18 and 19, copyright 2020 American Chemical Society and 2019 Elsevier (d) a schematic of the gas-tight rotating cylinder electrode cell used by Shen et al. to investigate mass transport limited conditions for c-CO₂RR.^16,63 Reproduced with permission from ref. 16 and 63, copyright 2023 Elsevier and 2022 Wiley.

Transport of mass, heat, and charge affect the kinetics in c-CO₂RR and therefore must be considered while designing electrolyzers for this reaction. A cell design with a thick hydrodynamic layer, like the H-cell, could slow down the local mass transfer of the species resulting in lower faradaic efficiency.¹¹² Also, as the dominant product of c-CO₂RR is H₂, the design must include considerations for ways to suppress HER. For instance, a cell design with very good mass transport, like the RCE cell, promotes HER as along with CO₂ and carbamate they also increase the flux of the bicarbonate and the ammonium cations to the electrode which act as a proton source.¹² Furthermore, cell resistance and uniform current distribution are some other factors that can potentially affect product distributions in electrochemical systems, especially for organic solvents with lower conductivity.

The availability of the active species at the surface of the electrode determines the production efficiencies. Thus designing the catalyst geometry to efficiently increase the contact of the active species can enhance c-CO₂RR. For instance, fundamental mass transport studies have mostly been carried out using the RDE/RRDE (rotating disk electrode/rotating ring disk electrode),¹¹³ however RDE setups present challenges of having low surface area, with poor hydrodynamics resulting in bubble formation at the tip of the electrode and concentration gradient along the radial direction of the disk.¹¹⁴ In comparison, the rotating cylinder electrode proposed by Jang et al. had a higher surface area, which leads to the availability of a higher concentration of CO₂ near the surface of the electrode, which was eventually used for mass transport studies.⁶³

CO₂RR suffers from slow kinetics and is rapidly limited by mass transport of CO₂ in most conventional cells used for fundamental studies. Similar difficulties can also be anticipated for c-CO₂RR if species relevant to the catalytic process are present in small amounts, and thus it is important to maintain a high flux of the active species to the electrode. Changing the residence time of reactants and reaction intermediates can alter the product selectivity without altering the intrinsic kinetics of the electrode itself, and distinguishing the relative timescales for transport and reaction at the catalyst scale is increasingly necessary.⁶³

5 Heterogeneous reaction mechanism

Although reaction mechanisms and reaction intermediates are known for CO₂RR, little is known about c-CO₂RR. It is unclear why hydrogen is the major product and what happens within the double layer. It is a common consensus that protonated amines act as the extra source of proton to facilitate hydrogen production, but how the carbamate is transported to the electrode–electrolyte interface is not known.⁶¹ Also, how the local pH or the buffer reactions play a role in such systems is not well-understood. To date c-CO₂RR studies have mostly looked into the formation of CO and HCOO⁻ as products and thus, delineating the different possible pathways at the surface of the electrode is important to gain deeper insights into the mechanisms relevant to c-CO₂RR.

Shen et al. proposed the competitive mechanism of electrochemical carbamate reduction on Ag as shown in Fig. 14. They proposed that cations present in the system help in bringing the carbamate to the electrode and can facilitate the electron transfer from the electrode to the carbamate for its further reduction. The carbamate can either decompose through a C–N bond cleavage to form CO₂ and the amine RNH₂ (blue pathway) or a direct proton–electron transfer on the carbamate to form the *RNHCOOH intermediate (orange pathway, *indicates a chemisorbed species on the Ag electrode). Once CO₂ is formed, it can chemisorb on the catalyst surface as shown on the blue pathway and reduced subsequently to *CO via two proton–electron transfer steps, which is the typical CO₂RR pathway on Ag.¹¹⁵ In contrast, the orange pathway directly reduces the carbamate, initially keeping the C–N bond formed. After forming *RNHCOOH, the C–N bond can break to generate *COOH on the surface (green path) or it can further reduce by proton–electron transfer to *RNHCO and water (orange path). Consequently, *RNHCO undergoes an additional proton–electron transfer on N and C–N bond cleaves to form *CO and restore the amine simultaneously. In short, along the pathway, C–N bond cleavage and further reduction of CO₂ compete with the direct proton–electron transfer to O in the –CO₂ group of the carbamate and C–N cleavage at a later step, thus creating three possible bifurcating pathways.¹⁶

Fig. 14 Proposed mechanism for the competition between electrochemical carbamate decomposition combined with CO₂ electroreduction (blue pathway) and carbamate direct reduction (orange pathway). The C–N bond cleavage can occur at three possible steps in the mechanism. K is the alkali metal potassium.¹⁶ Reproduced with permission from ref. 16, copyright 2023 Elsevier.

Kowalski et al. in their work describes a generic reaction network with different possible elementary pathways that could occur in c-CO₂RR systems. They use KRCO₂ as the captured CO₂ adduct and study the different pathways that can occur when the R is CH₃O (methyl carbonate), NH₂ (carbamate) or OH (bicarbonate) as shown in Fig. 15.¹² They show that after the adsorption of the KRCO₂ on the Ag electrode four different pathways can occur. The first route is the cleavage of R–C bond by protonation from the proton source XH (XH is either the solvent or the capture agent) producing adsorbed CO₂, RH and KX. This pathway then leads to the usual CO₂RR pathway with the first proton-coupled electron transfer (PCET) producing *COOH and the second producing *CO and water. The second route modifies the order of PCET and protonation, starting from a PCET assisted cleavage of R–C bond followed by a second PCET to form *COOH. The final C–O bond cleavage in this pathway is chemically assisted by the proton transfer from XH. The third route starts with a PCET at an O atom forming KRCOOH from where it can either go through a PCET step to cleave C–OH bond to form RCO and then the R–C bond is cleaved in a final proton transfer step to make CO, water, and KX, or it can cleave the R–C bond first by PCET or proton transfer followed by C–O bond cleavage. The fourth route treats the first step as chemical protonation using the XH proton source, and the second and third steps can be electrochemical PCET. Thus, these different pathways provide an overview of the mechanisms that could exist for different capture agents in different c-CO₂RR setups.

Fig. 15 (a) Reaction scheme for the reduction of the captured CO₂ with a general capture agent (R–H, examples being NH₂–H, CH₃O–H, and HO–H). R represents the deprotonated capture agent (NH₂ for NH₃, OCH₃ for methanol, or OH for H₂O), XH represents a general proton source. In methanol solvent, the proton sources in our study include methanol and NH₄⁺, written formally as NH₄ClO₄ to be in the XH form. In the water–amine solvent, this includes H₂O, KHCO₃, and NH₄⁺. * represents the catalytic Ag site. All species containing * are absorbed on the site. Chemicals in red text represent the products evolving in the solvent at each step. The two most favored pathways from our calculations both initiate with a proton-coupled electron transfer and are highlighted in blue (initial R–C cleavage in the CO₂ complex) or in red (final R–C cleavage). Reaction schemes for (b) CO₂RR, and (c) HER are also recalled with similar notations.¹² Reproduced with permission from ref. 12, copyright 2024 American Chemical Society.

Recently Ma et al. reported the direct electroreduction of carbonate to formate using a Cu catalyst. Through DFT calculations they discussed four possible reaction pathways that can occur at the electrode. In Fig. 16, path 1 shows the adsorption of the CO₂ through the O sites in the *OCO⁻ bidentate form, in path 2 a molecular CO₂ is attacked by a hydride to make formate, path 3 is a typical CO₂RR pathway where the adsorbed CO₂ undergoes a hydration to produce formate, and path 4 shows the direct carbonate reduction mechanism. They further discuss that the dominant pathway is a function of potential. At lower potentials between 0.0 and −0.4 V vs. RHE, the dominant species is *CO₃²⁻ with only *H competiting for adsorption, while *CO₂⁻ and *OCO⁻* adsorb at potentials more negative than −0.4 and −0.8 V vs. RHE, respectively. At potentials more negative than −0.8 V vs. RHE, it is difficult to adsorb *OCO⁻*, and also for path 2 the *H⁻ intermediate is unstable and mostly dominated by HER. Thus the only relevant path for formate production is path 3 in medium potential ranges. At larger potentials it will converge to CO, which can further undergo chemical hydration to make formate. But, at potentials more negative than −0.8 V vs. RHE the authors discuss that the *CO is more likely to reduce to C₁ and C₂₊ products limiting the further reduction to formate. However, in their study formate production pathway corresponds to path 4, which is through the direct CO₃²⁻ reduction. The first step in this pathway is to hydrogenate the adsorbed *CO₃²⁻ into *HCO₃⁻, which further reacts with H to give formate, while OH⁻ is eliminated. In path 4, their proposed mechanism does not form CO which eliminates the possibility of forming any C₂ products, as these products are primarily formed via CO–CO coupling. It provides a comprehensive insight into the heterogeneous mechanisms that can occur in c-CO₂RR systems, especially for the formate producing pathway.¹¹⁶

Fig. 16 (a) Reaction network of eCO₂RR and carbonate reduction mechanisms leading to HCOOH. The consensual eCO₂RR mechanisms to formate start by η_O,O *OCO*⁻ (path 1) or by *H⁻ (path 2) adsorptions. Additionally, path 3 leads to *CO which can further evolve to HCOOH by hydration (path 3.1). *COOH can also evolve to HCOOH (path 3.2). Path 4 is the newly proposed carbonate mechanism that can have contributions from path 3 through the conversion of the *COOH intermediate (path 3.3). (b) Range of potential stability of different intermediates.¹¹⁶ Reproduced with permission from ref. 116, copyright 2023 American Chemical Society.

The DFT studies can be further supported by operando analysis, including Raman, FTIR, UV-vis. For example, to experimentally verify the reduction of carbonate, Ma et al. mounted a carbonate complex, Cu₂(CO₃)(OH)₂, on a glassy carbon working electrode and studied the reduction with in situ Raman spectra. They observed that copper hydroxide bands of Cu₂(CO₃)(OH)₂ gradually disappear while the carbonate peaks become sharper. They concluded that the Cu₂(CO₃)(OH)₂ was first converted to carbonate. After 900 s of electrolysis, they observed that the carbonate band disappeared, and they attributed this to the direct reduction of carbonate on Cu electrode. They also observed that after electrolysis the color of the electrode changed from green to brown which is the color of metallic Cu. They also used the same Cu₂(CO₃)(OH)₂ mounted working electrode in a H-cell with a solution of 0.1 M KOH and they observed 1.01% of formate faradaic efficiency. Using this data, they solidify their claim of direct carbonate reduction. They describe the low faradaic efficiencies of formate as a result of the carbonate species detaching from the electrode driven by high HER activity. They also acknowledge that by introducing KOH and flowing N₂ in their system they create conditions where the solution is free of dissolved CO₂ which allows the study of carbonate reduction. In an analogous experiment with PbCO₃ they observed a faradaic efficiency of 0.89% for formate. Their study concluded that PbCO₃ can also be reduced in a similar mechanism to formate.¹¹⁶

Choi et al. investigated different transition metals, Ag, Au, Cu, and Sn, for direct reduction of ammonium carbamate. The binding energies of the protonated ammonium cation (ΔE_NH4⁺, solid lines) and CO₂-captured carbamate anion (ΔE_H2NCO2⁻, dotted lines) were plotted as a function of potential for Ag, Au, Cu, and Sn as shown in Fig. 17. They investigated the potential of zero charge (PZC) of all the metals, which influences the electrosorption properties in electrocatalysis, and found that all the PZCs are more positive than the experimental applied potentials (black triangles). Therefore, they hypothesize that the solvated NH₄⁺ are preferentially accumulated on the surface of the electrode while the H₂NCO₂⁻ are repelled due to its negative charge. They further discussed that the binding energies ΔE_NH4⁺ is stronger than ΔE_H2NCO2⁻ at the experimentally applied potentials, which limits the production of carbon based products.⁷¹ Thus, NH₄⁺ will not only block the active sites but also serve as a proton source to promote HER, leading to the predominant production of hydrogen on all metals. Bringing the captured CO₂ to the electrode surface while limiting HER activity from the capture agent is a major roadblock in c-CO₂RR that needs to be overcome.

Fig. 17 The binding energies of ammonium cation NH₄⁺ (ΔE_NH4⁺, solid lines) and carbamate anion H₂NCO₂⁻ (ΔE_H2NCO2⁻, dotted lines) as a function of potential for Ag, Au, Cu, and Sn. The potential of zero charge (PZC) for each metal is plotted with a vertical dash-dotted line. The applied potential to reach −12 mA cm⁻² in the c-CO₂RR experiments conducted by Choi et al. is marked with a triangle on the respective ΔE_NH4⁺ curve.⁷¹ Reproduced with permission from ref. 71, copyright 2024 American Chemical Society.

Thus, tuning the local reaction environment becomes particularly important for c-CO₂RR, which can provide a sufficient supply of the active species while suppressing the unwanted HER. Along with the bulk speciation, the local speciation can also vary with CO₂ loading, temperature, pH, electrochemical operating conditions and the cations, as shown in Fig. 18. The common ways deployed until now to alter the reaction environment in the presence of amine capture agents is by (i) changing the types and concentration of the amines, (ii) addition of alkali cations, (iii) increasing the temperature, or (iv) incorporating surfactants to suppress HER.

Fig. 18 Conceptual local reaction environment for CO₂ loaded amine solutions. Speciation at the gas–liquid interface, electrolyte bulk and liquid–electrode interface.¹⁶ Reproduced with permission from ref. 16, copyright 2023 Elsevier.

Abdinejad et al. used MEA, ethylenediamine (EDA) and decylamine (DCA) with a Cu electrocatalyst in 0.1 M NaClO₄ solution for CO₂RR and observed that EDA gives the highest faradaic efficiency of CO which they attributed to the presence of two amine molecules that increases the capture efficiency of the system (Fig. 19a).¹¹⁷ Similarly Shen et al. studied the activity for c-CO₂RR with ammonium carbamate (AC) and MEA using a electrodeposited Ag catalyst and observed similar faradaic efficiency of CO for both capture agents.¹⁶

Fig. 19 (a) Faradaic efficiency of CO and H₂ over Cu catalyst in 0.1 M NaClO₄ solution with MEA, ethylenediamine (EDA), and decylamine (DCA) at −0.78 V vs. RHE.¹¹⁷ Reproduced with permission from ref. 117, copyright 2020 American Chemical Society (b) comparison of equilibrium population and concentration of alkali and tetrabutylammonium (TBA⁺) cations in CO₂-loaded 0.1 M EEA in DMSO.¹¹⁸ Reproduced with permission from ref. 118, copyright 2022 Royal Society of Chemistry (c) faradaic efficiency and applied potential during the electrochemical reduction of a 0.7 M MEA electrolyte with an Ag electrocatalyst at a fixed current density of 4 mA cm⁻² at an electrode rotation speed of 800 rpm.¹⁶ Reproduced with permission from ref. 16, copyright 2023 Elsevier (d) product distributions of CO₂ reduction over indium catalyst at −0.8 V vs. RHE in CO₂-saturated 30 wt% MEA aqueous solution with different concentrations of CTAB surfactant.⁸⁹ Reproduced with permission from ref. 89, copyright 2017 Wiley.

Few studies have investigated the addition of alkali cations as a way to increase selectivity towards CO₂ conversion. Monteiro et al. investigated CO₂RR on Cu, Ag, and Au and showed that the presence of large alkali cations (K⁺ or Cs⁺) in particular can undergo weak hydration spheres that can be concentrated at the electrode surface and stabilise the CO₂⁻ intermediate via local electric field effect.¹¹⁹ In an amine containing solution it can further enhance charge transfer and destabilise the formation of carbamate to enhance the C–O bond cleavage. Khurram et al. reported a high CO current density with alkali cations in a solution of 2-ethoxyethylamine (EEA) in dimethylsulphoxide (DMSO) for CO₂RR which depended on the size of the cations with K⁺ < Na⁺ < Li⁺, and attributed the behaviour to the unstable carbamate formation in the presence of large cations with easier desolvation and rapid pairing kinetics (Fig. 19b).¹¹⁸ As observed by Khurram et al. the size of the cations has an effect on the carbamate formation. With ¹H NMR studies they showed that the proportion of carbamate formed in their solution was a function of the alkali cation present in their electrolyte. The Lewis acidity of the cations was found to dictate the amount of ammonium cation that will be formed. For alkali cations, the Lewis acidity decreases in the order Li⁺ > Na⁺ > K⁺, and thus it was found that the stronger Lewis acid, Li⁺, associates more strongly with the initial carbamic acid driving the reaction faster to the formation of carbamate. The interaction of the alkali cation with carbamate thus played a role in the binding of the cations to the carbamate anion. DFT studies also showed that the reaction of the carbamate anion with the alkali cations is spontaneous in decreasing order of Li⁺ > Na⁺ > K⁺. The type of anion present in the electrolyte did not play any significant role in the formation of carbamate. Larger cations like K⁺ and Cs⁺ have been reported to be able to increase the availability of CO₂ near the surface of the electrode in CO₂RR studies. The measurement of interfacial CO₂ concentration for different alkali metal cations through in situ ATR-SEIRAS have also shown that Li⁺ has the highest CO₂ concentration and K⁺ has the lowest CO₂ concentration.¹²⁰ This is primarily because the hydration sphere of cations inversely scales with the size of the cations where Li⁺ > Na⁺ > K⁺ > Cs⁺ which changes the local buffering capacity by changing the interfacial pK_a.¹²¹ In contrast, for c-CO₂RR larger cations can potentially help in reducing the local pH of the system near the electrode, enabling the local release of CO₂ near the electrode from the captured-CO₂ adduct. This can help in reducing HER and promote c-CO₂RR activity. In current c-CO₂RR studies the local pH is alkaline and thus tuning the cations to maintain pH close to the pK_a of the capture agents can help prevent the deprotonation of the amines and the shifting of equilibrium reactions locally.

Temperature studies were also investigated for c-CO₂RR and it was seen that higher temperatures increases the faradaic efficiency of CO formation. Shen et al. and Kim et al., investigated the temperature dependency and reached faradaic efficiency of ≈20–40% at higher temperatures which they attributed to the shift in equilibrium to have more free CO₂ in the system (Fig. 19c). The effect of surfactants on CO₂RR has also been studied in 30 wt% MEA by Chen et al., as they are known to reduce HER activity by blocking the active sites. Using CTAB they show that they can boost CO₂ conversion to formate and CO, primarily through HER suppression. They further show that a high concentration of CTAB (>0.1 wt%) helps in producing CO while a low concentration (0.01 wt%) was sufficient to promote CO in indium electrodes. The highest formate faradaic efficiency of 45.4% and CO faradaic efficiency of 17.0% was achieved when combined with porous In electrodes (Fig. 19d).⁸⁹

Although these techniques have been mostly investigated under CO₂RR conditions, the same ideas can be translated to c-CO₂RR conditions. The interactions at the liquid–electrode interface can be tuned to increase the selectivity towards CO. The trimethylammonium alkyl tail in surfactants like CTAB can align itself at the surface of the electrode creating hydrophobic interactions that suppresses HER in c-CO₂RR studies.^122,123 Similarly different sizes of alkali cations can also be combined with surfactants to alter the selectivity.¹²⁴ Larger cations like Cs⁺ have been reported to be able to maintain a near neutral pH at the surface of the electrode increasing the availability of CO₂ in CO₂RR studies.^125–127 In current c-CO₂RR studies the local pH is alkaline and thus tuning the cations to maintain pH close to the pK_a of the capture agents can help prevent the deprotonation of the amines and the shifting of equilibrium reactions locally. The use of ionomer coatings have also been studied to tailor microenvironments by controlling water and ion transport through the polymers. Ionomers like Nafion and Sustanion has shown that they can maintain a high local pH and CO₂ concentration to increase C₂₊ products with Cu catalyst and CO with Ag catalyst.^128–130 In addition, pulsed electrolysis can also be investigated to increase CO selectivity.¹³¹ The challenge in c-CO₂RR systems is to bring the negatively charged carbamate species close to the cathode, which could be overcome by the pulsing technique. It can also reduce the local pH during the anodic cycle to shift the equilibrium and generate more carbamate species near the electrode that can potentially undergo reduction reactions.

6 Effect of different capture agents on c-CO₂RR

Different capture agents behave differently under c-CO₂RR conditions. So far only the changing binding energies and the CO₂ loading capacity have been investigated,^16,61 however it needs to be understood that the mass transport conditions, thermodynamics of the captured-CO₂ adduct species and vapor–liquid equilibrium in different capture agents change at different conditions which needs due consideration. In theory, the best capture agent will be the one that has the minimum thermodynamic barrier to get reduced to products, but species with different equilibrium constants will have different rates of making CO₂ in the bulk of the solution that will further influence the products that we see. For instance, as shown in Fig. 20, from a thermodynamic standpoint it can be observed that CO₂ binding to the amines are the strongest which is around −60–80 kJ mol⁻¹ followed by alcohol based capture agents −20 to −25 kJ mol⁻¹, followed by bicarbonates −13 to −18 kJ mol⁻¹.^31,45 Thus when we talk about c-CO₂RR, a larger thermodynamic barrier needs to be overcome while going from the carbamates to reduced products compared to the alkoxide based capture agents or the hydroxyl based capture agents. Furthermore, Appel et al. in their work showed that the partial pressure of CO₂ also determines the free energy of the CO₂-bound adduct and dictates the thermodynamic potential required to make reduced products.¹³² They further discuss that if the capture agent has the ability to bind CO₂ stronger than what is required from a specific dilute stream then the process could lead to inefficiencies from the excess binding energy. Thus the influence of these different binding energies on c-CO₂RR is important to understand for future research in c-CO₂RR.

Fig. 20 Conceptual thermodynamic free energy change comparing c-CO₂RR and CO₂RR of amine, hydroxyl and alkoxide capture agents. The free energy of capture of a dilute stream of flue gas in amine is −60 to −80 kJ mol⁻¹ (shown in red), in alkoxide is −20 to −25 kJ mol⁻¹ (shown in green), and in alkali hydroxyls is −13 to −18 kJ mol⁻¹ (shown in grey).^31,45 The extra steps of release and compression in CO₂RR (shown in blue lines) are responsible for the inefficiencies in the conventional CO₂RR system. Although the barriers to go from captured-CO₂ to reduced products are the same through the CO₂RR and the c-CO₂RR paths, the c-CO₂RR process (shown in orange line) can facilitate the conversion in a one step process reducing the inefficiencies in the capture and conversion process.

To investigate these changes, Kowalski et al. used methoxide as a capture agent to determine its activity for c-CO₂RR and compared it to hydroxyls and amine capture agents using a silver electrocatalyst. Through DFT calculations they show that the nature of proton source and its pK_a has a significant impact on the onset potentials for c-CO₂RR. They hypothesized that a proton source with a low pK_a and use of a solvent like methanol which has a low dielectric constant can improve c-CO₂RR while penalizing the CO₂RR. However, on Ag(111) surfaces they found that HER has a lower onset potential than CO₂RR or c-CO₂RR Fig. 21a–f, which is the dominant product also seen experimentally. As can be seen from (Fig. 21g and h), they measured the partial current density of CO and H₂ as a function of potential for all the three capture agents. In all cases, the partial current density of CO observed was very low with H₂ being the dominant product. For the ammonium carbamate, negiligible amounts of CO were observed as it was more selective for H₂ evolution. The authors further normalized the partial current densities of CO observed to the limiting partial current density of CO to quantify the activity for c-CO₂RR and determine the active species undergoing reduction. Using the method described in Section 3.3, they found that in all cases it was the free dissolved CO₂ that is getting reduced (Fig. 21i). They also discussed that alcohol based capture agents are promising as the only proton source in such systems is the alcohol itself compared to more Bronsted acidic HCO₃⁻ or NH₄⁺. Thus it can suppress HER which is the bottleneck for c-CO₂RR systems (Fig. 21b).¹² The work then needs to be combined with new catalyst design, moving away from Ag or other transition metals that are typically known for its good activity in CO₂RR. Catalysts that are known to suppress HER to improve the activity for c-CO₂RR are thus needed along with the ability to bring the negatively charged carbamate species to the surface of the electrode.

Fig. 21 Calculated onset potential (blue, vs. SHE) and over-potential (brown) for the electroreduction of (a) methyl carbonate KCH₃OCO₂, (b) bicarbonate KHOCO₂, (c) carbamate KNH₂CO₂ on Ag(111) using the various solvent-proton source combinations of CH₃OH and H₂O. The compound being electrochemically reduced is labeled in the top right of each panel, together with the solvent (notation: compound@solvent). The x axis denotes the proton source. Values for (d) HER, (e) CO₂RR in CH₃OH solvent, and (f) CO₂RR in H₂O solvent are also given for comparison. Onset potential is defined as the least negative potential lower than the equilibrium potential at which the ES becomes equal to 0.75 eV, while the overpotential is equal to the difference between the equilibrium potential and the onset potential. The pathway that has the smallest onset potential is displayed. Partial current densities of (g) CO and (h) H₂ and (i) CO with respect to the maximum CO₂ flux as a function of applied potential. Three CO₂-captured complexes are considered: KHOCO₂ (bicarbonate), NH₄–NH₂CO₂ (ammonium carbamate) and KCH₃OCO₂ (methyl carbonate).¹² Reproduced with permission from ref. 12, copyright 2024 American Chemical Society.

Alkanolamine solutions for CO₂ capture faces limitations with regard to high degradation rates, toxicity and high regeneration energy requirements. More recently amino acids have been proposed as an alternative to circumvent these challenges. Amino acids have advantages of low toxicity, low corrosivity, fast capture kinetics with CO₂, and good resistance to oxidative and thermal degradation. Ramezani et al. compared different amino acid salts at different conditions to position their effectiveness with respect to alkanolamines, MEA in particular.¹³³ In their comparison, they reported that amino acid salts like K-Lys (lysine), K-Pro (proline), and K-Sar (sarcosinate) have higher pK_a and higher CO₂ loading capacity than MEA. The high pK_a of these salts makes the carbamate unstable and facilitates the formation of carbonates and bicarbonates which in turn reduces the heat of absorption in these systems. They further report that the CO₂ absorption rate increased as the temperature and concentration of the amino acid salt increased. Thus, the advantages of amino acids makes them favorable for c-CO₂RR. The challenge in c-CO₂RR systems with amines is the strong binding of the CO₂ in the form of stable carbamates that makes it hard to directly reduce the CO₂-bound adduct. From the investigation of Ramezani et al., it was observed that the absorption rate and the desorption rate can be tuned by using smaller chains of amino acids that reduces the distance between the amino and the carboxyl groups. This along with sterically hindered amino acids can slow down the absorption rate and enhance the desorption rate due to the bulkier substituent group (Fig. 22a).¹³³ For instance, K-Asp (asparagine), K-Tau (taurate), K-Ser (serine), K-Arg (arginine), K-Cys (cysteine), and K-Glu (glutamine) showed faster desorption rate than MEA and can be promising candidates for c-CO₂RR (Fig. 22b). Thus for c-CO₂RR it is important to achieve a balance between absorption and desorption kinetics to optimize the conversion.

Fig. 22 (a) The overall rate constant for CO₂ absorption as a function of concentration for different amino acid salts at 298.15 K. (b) The desorption rate of 1 M potassium salts of different amino acid at 353.15 K. K-Lys: potassium lysine, K-Pro: potassium proline, K-Arg: potassium arginine, K-His: potassium histidine, K-Gly: potassium glycine, K-Ala: potassium alanine, K-Thr: potassium threonine, K-Tau: potassium taurate, K-Ser: potassium serine, K-Sar: potassium sarcosinate, and MEA: monoethanolamine.¹³³ Reproduced with permission from ref. 133, copyright 2022 De Gruyter.

Although amine capture agents are the state of the art for postcombustion capture processes due to its high capture efficiency, most c-CO₂RR studies only show dissolved CO₂ as the active species getting reduced. Selecting capture agents for c-CO₂RR thus becomes more complicated. Amine capture agents binds to CO₂via chemical absorption, however if dissolved CO₂ is the main species getting reduced then other solvents, for instance the ones used in precombustion CO₂ capture via physical absorption, can also be explored. Physical solvents exhibit characteristics such as low vapor pressure, high reactivity or absorptivity, facile regeneration, high thermal and mechanical resilience, minimal corrosiveness, and minimal environmental impact which proves ideal for CO₂ capture.^134,135 Such solvents also possess exceptional thermal stability, mitigating the risk of solvent loss and contamination while regenerating at elevated temperatures. Commercially available physical processes for CCS encompass selexol (dimethyl ether of polyethylene glycol (DMPEG)), rectisol (methanol), purisol (N-methyl-2-pyrolidone (NMP)), morphysorb, sulfolane (tetrahydrothiophenedioxide), and fluor (propylene carbonate) solvent.^136,137 Selexol (a liquid glycol solvent) stands out among these processes, having been widely utilized in CO₂ capture for decades. Its advantages over other physical solvents include lower vapor pressure, non-corrosive properties, reduced heat requirements, and inertness towards select gases.¹³⁸

c-CO₂RR can pose challenges on stability depending on the capture agent that is being used. CO₂ capture solutions consist of several species that coexist in the solution which can trigger the interaction between the catalyst and the capture solution resulting into side reactions. There have been reports of morphological changes and corrosion of metal electrodes under c-CO₂RR conditions.⁷¹ Thus, understanding the vapor–liquid equilibrium models and linking them with electrochemical c-CO₂RR is important to determine the concentration of different species present in the capture solution and identify the species getting reduced or causing the instability of the catalyst. Recent studies have reported that direct reduction of CO₂-capture solutions undergoes high HER.^12,61,71 For instance, in amine capture solution the protonated amines contribute to the high HER. This reduces the generation of carbon products favoring the side reaction of HER. Therefore, according to the reported studies, the reduction of CO₂-bound complex is not favored at the electrode surface. In addition, catalysts like Cu which is known to produce multi carbon products in CO₂RR have been reported to undergo corrosion in the presence of amine capture agents, whereas catalysts like Au and Ag have been reported to be relatively stable and inert to corrosion.⁷¹ In the presence of ammonia, Cu was reported to be able to form ligands with the free amines in the solution to form copper–ammonia complex. This is possible even at reductive potential, making Cu not suitable for c-CO₂RR with amine capture agents.¹³⁹ Many other mechanisms have been proposed for cathodic corrosion, including cathodic etching in both aqueous and non-aqueous solutions, formation of (meta)stable metal hydrides at the cathode as possible intermediates that eventually are released from the cathode, and leaching due to the interaction of alkali metals with the cathode. These would also need to be investigated as new capture agents are developed to determine the stability of the catalysts (Chem. Rev., 2021, 121(17), 10241–10270). This requires a deep understanding of the degradation and restructuring mechanisms of the catalyst along with an understanding of the transport of different species during electrocatalysis to be able to develop strategies to mitigate these processes.

7 Effect of impurities on electrochemical c-CO₂RR

Industrial flue gas contains contaminants like NOx, SOx and O₂ that should be considered while developing c-CO₂RR technologies. Several challenges have been identified by researchers on the impact of impurities for CO₂RR that includes identifying catalysts that can maintain high selectivity and activity and designing suitable reactor configurations for viable operation. The presence of O₂ which is 25 times less soluble in water than CO₂ at STP conditions can result in significant loss of current density to oxygen reduction reaction (ORR). The lower standard reduction potential due to lower thermodynamic requirements and higher kinetic favorability promotes ORR compared to CO₂RR. Furthermore, the higher diffusion rate of O₂ compared to CO₂ also plays a role in changes observed in selectivity. O₂ can also oxidize the catalyst leading to changes in the oxidation state of the catalyst during catalytic turnover or lead to corrosion. To prevent catalyst oxidation and maintain CO₂RR selectivity, strategies like deploying ionomer coatings which are hydrophilic in nature have been suggested.¹⁴⁰ Hydrophilic nanopores have been argued to reduce the mass flux of O₂ in to the electrode as the O₂ needs to diffuse in the electrolyte-wet form, while the CO₂ mass flux have been observed to not have any significant difference from different ionomer coatings. Hydrophilic support of TiO₂ has also been proposed by Xu et al. that has shown to impede the effects of O₂ an undergo stable CO₂RR.¹⁴¹

NO_x and SO_x impurities in flue gas have high solubility in water which dissolve to form acids in the solution as described in eqn (20)–(23). The formation of acids lowers the pH of the solution which then leads to increased HER, lowering the current densities for CO₂RR. Both NO₂ and SO₂ also can adversely react with the catalyst to poison its surface and reduce the activity of the catalyst.^142,143 NO₂ concentrations above 1600 ppm have been shown to reduce the CO₂RR faradaic efficiency on Cu catalyst. Similar results have been observed for SO₂ on Cu catalysts. However, exception to this poisoning effect have been observed when using Ag or Sn catalysts. This has been attributed to the fact that they form Ag₂S and SnS₂ during CO₂RR which are both thermodynamically unstable.¹⁴⁴

SOx dissolution:

SO₂(g) + H₂O(l) → SO₃²⁻(aq) + 2H⁺(aq) (20)

SO₃(g) + H₂O(l) → SO₄²⁻(aq) + 2H⁺(aq) (21)

NOx dissolution:

2NO₂(g) + H₂O(l) → NO₂⁻(aq) + NO₃⁻(aq) + 2H⁺(aq) (22)

NO₂(g) + NO(g) + H₂O(l) → 2NO₂⁻(aq) + 2H⁺(aq) (23)

Therefore, similar effects could be observed in c-CO₂RR. In the presence of O₂, c-CO₂RR which has a higher thermodynamic barrier than CO₂RR, will result in even higher thermodynamic unfavorability while competing with ORR compared to CO₂RR, unless the oxidized catalyst surface is a better catalyst for the absorption and activation of the CO₂-bound adduct. Similarly, as alkaline environments are preferred for CO₂ capture, unreacted NO_x and SO_x could accumulate in the capture agent and reduce the capture efficiency by lowering the pH of the capture solution resulting in lower CO₂ loadings. Therefore, the system will have lower concentrations of CO₂-bound adducts in the presence of SO_x and NO_x that can eventually be electrochemically reduced. However, there are instances where Ag and Sn have been reported to be stable in SO₂ impurities. Therefore, for c-CO₂RR catalysts which have stable activity and selectivity in the presence of impurities will be important to identify.

8 Energy comparison of CO₂RR vs. c-CO₂RR

The energy benefits of c-CO₂RR can be extrapolated by comparing them with state-of-the-art CO₂ electrolyzers. Most high-performance CO₂ electrolyzer designs are inspired from the fuel cell community and are developed as zero-gap gas-fed electrolyzers to maximize the CO₂ transport to the electrode and thus enable their operation at high current densities. These electrolyzers can typically operate at current densities higher than 100 mA cm⁻² with a cell voltage of 3–3.5 V to make CO with a faradaic efficiency as high as 90%. Thus, these typically require 600–700 kJ mol⁻¹ of energy to operate at standard temperature and pressure conditions. However, energy penalties lie at the anode where the bicarbonates or carbonate salts can be converted to CO₂ which requires at least 254 kJ mol⁻¹ and therefore the CO₂ utilization rate is typically low in CO₂ electrolyzers (Fig. 23).¹⁴⁵

Fig. 23 Energy comparison between the sequential route (conventional CO₂RR route) vs. the integrated route (the c-CO₂RR route).³³ The optimistic scenario assumes that c-CO₂RR performs at the same efficiency as the current state-of-the-art CO₂ electrolyzers with a 90% F.E. CO at cell potential of 3 V, the baseline scenario uses Lee et al.'s report of 72% F.E. CO at −0.8 V vs. RHE,²⁸ and the pessimistic scenario assumes a 40% F.E. CO at a large cell potential of 5 V. Reproduced with permission from ref. 33, copyright 2022 Nature.

Compared to the CO₂ electrolyzers, although the current c-CO₂RR system requires significantly higher energy (800–10 000 kJ mol⁻¹), primarily because of the low CO production and high HER. Therefore, as discussed earlier, improving the CO production rates remains a challenge and addressing which can improve the energy requirements of c-CO₂RR systems. This requires fundamental insights on reaction environments, catalysts activity and stability, and influence of different capture agents to optimize the c-CO₂RR system. However, as reported by Li et al., considering that the c-CO₂RR system can achieve similar faradaic efficiencies and current densities at similar cell voltage as the CO₂ gas-fed electrolyzers, there is a potential of saving 44% of the total energy required (Fig. 23). This saving is due to the fact that c-CO₂RR can save energy costs arising from capture agent regeneration, CO₂ compression, and product purification needed in conventional CO₂ reduction process.³³

9 Conclusion and future outlook

The long term development of c-CO₂RR technologies posseses many challenges that needs to be overcome to make this technology industrially viable. Along with the coupled effects of transport, kinetics, and thermodynamics that plays a role in c-CO₂RR development, recent studies have reported corrosion and restructuring of the catalyst in the presence of amines as a crucial obstruction to further technology development.⁷¹ The activity and stability of the transition metal catalyst that were known for CO₂RR thus do not apply to c-CO₂RR anymore. Furthermore, alkoxide capture agents showed promise in suppressing HER, but the reaction when looked from the anodic half would probably oxidize the methyl carbonate to give back CO₂, thus eliminating the purpose of c-CO₂RR. Thus finding a combination of suitable capture agent, catalyst and operating condition is needed to advance the development of c-CO₂RR. Although, the use of VLE models, effect of thermodynamic barriers, high HER, catalyst stability are discussed for selected capture agents, similar concepts and understanding are relevant for all capture agents that are used for c-CO₂RR.

To advance research in c-CO₂RR the integration of capture agent, operando characterization, catalyst discovery, and reactor and catalyst design needs to be simultaneously investigated for achieving c-CO₂RR and for potential scale-ups (Fig. 24). The selection of capture agents will determine the binding energy of CO₂ to the capture agent and the thermodynamic energetic barriers relevant for direct reduction of the captured CO₂-bound adduct. Catalyst discovery will involve investigating different catalysts for activity and stability under c-CO₂RR by leveraging the use of both DFT calculations and experiments. These findings will further need to be linked with operando characterization to gain insights into the morphology, composition, crystallographic structural information to correlate experimental and theoretical results. Reactor and catalyst design also needs to be investigated in parallel to tune external and internal mass transport properties of the system and facilitate selectivity towards c-CO₂RR. This screening loop of finding activity, stability and design descriptors for c-CO₂RR is also beneficial to determine and address system-level challenges and accelerate the development of c-CO₂RR processes (Fig. 24).

Fig. 24 Illustration showing the screening loop between capture agent discovery, catalyst discovery, reactor and catalyst design principles, and operando characterization that needs to be investigated for the eventual scale-up of c-CO₂RR systems.

Most literature on c-CO₂RR reported C₁ products, more specifically, CO. Table 1 summarizes the recent works on c-CO₂RR. As the field evolves strategies for producing C₂₊ products should be explored. Currently, several challenges lie in c-CO₂RR which should be investigated to gain more fundamental insight into the process. These include the higher thermodynamic barrier to reduce captured CO₂-adduct, high HER, and identifying the carbon source getting reduced during electrolysis. Furthermore, catalysts like Cu which are known for producing multi carbon products undergo corrosion and restructuring when amines are used as capture agents.⁶⁵ However, as reported by Choi et al., transition metals like Au and Ag do not undergo corrosion in the presence of amines.⁷¹ These metals are known for producing CO with faradaic efficiencies as high as 90% which perhaps explain why only C₁ products have been observed. The understanding of corrosion mechanisms, which is one of the key bottlenecks in using catalysts like Cu in c-CO₂RR needs to be addressed. Inhibiting cathodic corrosion can be achieved with the use of additives to tailor the electrical double layer or by alloying to alter the physico-chemical properties of the cathode. The addition of (poly)cationic ammonium salts can prevent the formation of parasitic hydrogen and corroding compounds in the double layer. Alloying can change the electronic band structure which can result in lower interaction between the cathode and the corrosive species or increase the energy barrier for reaction with the corrosion intermediates. The ternary alloy of CuSn₇Pb₁₅ has been shown to be physically and chemically stable towards the formation of formate in electroreduction of CO₂ and can be employed in amine-based solvents in c-CO₂RR.¹⁴⁶ Therefore, a combination of design strategies is necessary for eventually extending the space of c-CO₂RR towards C₂₊ products.

Table 1 Summary of electrolyzer design, capture agent, catalyst, and c-CO₂RR performance in heterogeneous electrocatalysis. (All the chemical names are provided under the acronym section)

Reactor	CO2 capture agent	Catalyst	Supporting electrolyte	Gas flow	c-CO2RR performance	Ref.
Zero-gap membrane electrode assembly	5 M MEA	Ni–N/C	—	Ar	F.E. CO: 78.3%	Kim et al.^15
					j CO: 2.6 mA cm^−2
	5 M MEA	Ag	—	Ar	F.E. CO: 38.3	Kim et al.^15
					j CO: 0.79 mA cm^−2
Compression cell	0.1 M MEA	Ag	0.1 M KHCO3	CO2	j CO: 6 mA cm^−2	Safipour et al.^17
	0.2 M MEA	Ag	0.1 M KHCO3	CO2	j CO: 3 mA cm^−2
H-cell	2 M MEA, 2 M BAPN, 2 M n-BA, 2 M 2A1P, 2 M AMP, 2 M DMAE	Ag	2 M KCl	N2	F.E. CO: 1–20%	Leverick et al.^61
					j CO: 0.01–1 mA cm^−2
RCE reactor	0.2 M and 0.7 M AC, 0.2 and 0.7 M MEA, 0.2 M and 0.7 M KHCO3	Ag	0.099 M KClO4 + 0.001 M KOH	Ar	F.E. CO: 1–9%	Shen et al.^16
					j CO: 0.01–0.4 mA cm^−2
RCE reactor	0.5 M AC, 0.5 M KHCO3, 5 wt% methoxide	Ag	0.099 M KClO4 + 0.001 M KOH	Ar	F.E. CO: 0.01–1.68%	Kowalski et al.^12
					j CO: 0.001–0.28 mA cm^−2
RCE reactor	0.2 M AC	Cu, Ag, Au, Sn, Ti	0.099 M KClO4 + 0.001 M KOH	Ar	F.E. CO: ≈0.01%	Choi et al.^71
					j CO: 12 mA cm^−2
Three electrode configuration	2 M MEA	Ag	2 M KCl	N2	F.E. CO: 10%	Lee et al.^28
					j CO: 0.6 mA cm^−2
Three electrode configuration	2 M MEA	Ag	2 M KCl	N2	F.E. CO: 10%	Lee et al.^28
					j CO: 0.6 mA cm^−2
H-cell	0.05 M K2CO3	Cu	—	N2	F.E. HCOO^−: 0.61%	Ma et al.^116

---

In addition to the following, developing anodes for the captured CO₂ media will guide the development of c-CO₂RR technologies. Water oxidation reaction at the anode is kinetically sluggish and can lead to lower efficiency of the electrochemical reactor. Thus, choosing a catalyst that is thermodynamically favorable to water oxidation can improve the overall system efficiency. Furthermore, the loss of CO₂ from the anode during c-CO₂RR should be at similar rates as CO₂RR considering a similar acidic microenvironment is generated at the anode during water oxidation. However, there could be added complexities to the development of catalyst materials for the anode for c-CO₂RR. Cathodic corrosion that has been observed in c-CO₂RR conditions, especially with Cu electrode in amine solutions, can occur at the anode too. Therefore, catalysts that are compatible with the CO₂-captured complex are needed to be identified to prevent any anodic corrosion. Pt so far has been used for different c-CO₂RR studies as the anode, but their stability in these capture solutions has not been investigated. Additionally, cheaper alternatives are required and needs to be tested for industrial scale electrolyzers that can operate at overpotentials similar to Pt.

Integrated CO₂ capture and conversion posses benefits in terms of energy and capital cost, however there maybe trade-offs when trying to implement the capture and conversion in one unit. In current carbon capture and utilization processes, the CO₂ is fed in a compressed form which increases the conversion efficiency of CO₂ compared to if the whole process takes place at room temperature and atmospheric pressure. Thus the current technologies still outcompetes the conversion efficiency of c-CO₂RR. Until now c-CO₂RR technologies have mostly focused on batch mode of operation, and thus continuous mode of operation will simultaneously have to be investigated and compared with the current state-of-the-art conversion efficiency.¹⁴⁷ Also, point sources of CO₂ has a wide range of gas temperature, CO₂ concentration, and contaminants which must also be explored during the development phase. A typical c-CO₂RR system should be stable for thousand of hours at high current densities and operating in a continuous mode of operation.¹¹

In addition to performance, the development of c-CO₂RR can further be coupled with life-cycle and techno-economic analysis (TEA) analysis in parallel keeping in mind future viability. Li et al. in their work compared the TEA of a conventional CCU process to the integrated route of c-CO₂RR. They found that at the current level of technology development there is no foreseen advantage of using the integrated route as the energy cost of converting captured CO₂ to products is higher. This offsets the cost reduced from process intensification. Also due to higher cost of electricity than heat, the integrated route further uses up more costs. However in an optimistic scenario, where c-CO₂RR reaches the same performance in terms of current density and faradaic efficiency, the overall energy can be reduced by 44% with a 22% savings in energy cost.³³ Thus the development of c-CO₂RR technologies is an attractive option for closing the carbon cycle using renewable electricity.

Acronym

MEA	Monoethanolamine
AC	Ammonium carbamate
BAPN	β-Aminopropionitrile
n-BA	n-Butylamine
2A1P	2-Amino-1-propanol
AMP	Aminomethyl propanol
DMAE	Dimethylethanolamine
DEA	Diethanolamine
MDEA	Methyldiethanolamine
KHCO3	Potassium bicarbonate
K2CO3	Potassium carbonate
KClO4	Potassium perchlorate

Data availability

The review article consists of information and research findings which are already published in peer-reviewed journals. All the necessary citations are included in the main article and permission for reproducing the figures are also obtained.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We thank the University of California Office of the President and the National Laboratory Research Fees Program for supporting this work with award L22CR4468, which supports the Center for Direct Conversion of Captured CO₂ into Chemicals and Fuels.

Notes and references

1. P. R. Shukla, J. Skea, E. Calvo Buendia, V. Masson-Delmotte, H. O. Pörtner, D. Roberts, P. Zhai, R. Slade, S. Connors and R. Van Diemenet al., Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems, IPCC, 2019.
2. R. Lindsey, Climate change: atmospheric carbon dioxide, 2020, https://www.climate.gov/news-features/understanding-climate/climate-change-atmospheric-carbon-dioxide.
3. A. Kätelhön, R. Meys, S. Deutz, S. Suh and A. Bardow, Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 11187–11194.
4. R. G. Grim, Z. Huang, M. T. Guarnieri, J. R. Ferrell, L. Tao and J. A. Schaidle, Energy Environ. Sci., 2020, 13, 472–494.
5. R. Socolow, M. Desmond, R. Aines, J. Blackstock, O. Bolland, T. Kaarsberg, N. Lewis, M. Mazzotti, A. Pfeffer, K. Sawyeret al., Direct air capture of CO₂with chemicals: a technology assessment for the APS Panel on Public Affairs, American Physical Society technical report, 2011.
6. D. W. Keith, G. Holmes, D. S. Angelo and K. Heidel, Joule, 2018, 2, 1573–1594.
7. J. Valentine, A. Zoelle, S. Homsy, H. Mantripragada, M. Woods, N. Roy, A. Kilstofte, M. Sturdivan, M. Steutermann and T. Fout, Direct air capture case studies: sorbent system, National Energy Technology Laboratory (NETL), Pittsburgh, Pa, Morgantown, WV, technical report, 2022.
8. National Academies of Sciences, Division on Earth, Life Studies, Ocean Studies Board, Board on Chemical Sciences, Board on Earth Sciences, Board on Energy, Environmental Systems, Board on Atmospheric Sciences, Committee on Developing a Research Agenda for Carbon Dioxide Removal, Reliable Sequestration, Negative emissions technologies and reliable sequestration: a research agenda, National Academies Press, 2019.
9. S. Deutz and A. Bardow, Nat. Energy, 2021, 6, 203–213.
10. A. Aspelund and K. Jordal, Int. J. Greenhouse Gas Control, 2007, 1, 343–354.
11. M. C. Freyman, Z. Huang, D. Ravikumar, E. B. Duoss, Y. Li, S. E. Baker, S. H. Pang and J. A. Schaidle, Joule, 2023, 7, 631–651.
12. R. M. Kowalski, A. Banerjee, C. Yue, S. G. Gracia, D. Cheng, C. G. Morales-Guio and P. Sautet, J. Am. Chem. Soc., 2024, 146, 20728–20741.
13. Z. Zhang, A. L. Kummeth, J. Y. Yang and A. N. Alexandrova, Proc. Natl. Acad. Sci. U. S. A., 2022, 119, e2123496119.
14. Y. C. Xiao, S. S. Sun, Y. Zhao, R. K. Miao, M. Fan, G. Lee, Y. Chen, C. M. Gabardo, Y. Yu and C. Qiu, et al. , Nat. Commun., 2024, 15, 7849.
15. J. H. Kim, H. Jang, G. Bak, W. Choi, H. Yun, E. Lee, D. Kim, J. Kim, S. Y. Lee and Y. J. Hwang, Energy Environ. Sci., 2022, 15, 4301–4312.
16. K. Shen, D. Cheng, E. Reyes-Lopez, J. Jang, P. Sautet and C. G. Morales-Guio, Joule, 2023, 7, 1260–1276.
17. J. Safipour, A. Z. Weber and A. T. Bell, ACS Energy Lett., 2023, 8, 5012–5017.
18. E. W. Lees, M. Goldman, A. G. Fink, D. J. Dvorak, D. A. Salvatore, Z. Zhang, N. W. Loo and C. P. Berlinguette, ACS Energy Lett., 2020, 5, 2165–2173.
19. T. Li, E. W. Lees, M. Goldman, D. A. Salvatore, D. M. Weekes and C. P. Berlinguette, Joule, 2019, 3, 1487–1497.
20. S. Kar, A. Goeppert, V. Galvan, R. Chowdhury, J. Olah and G. S. Prakash, J. Am. Chem. Soc., 2018, 140, 16873–16876.
21. J. Kothandaraman, A. Goeppert, M. Czaun, G. A. Olah and G. S. Prakash, Green Chem., 2016, 18, 5831–5838.
22. S. Kar, R. Sen, A. Goeppert and G. S. Prakash, J. Am. Chem. Soc., 2018, 140, 1580–1583.
23. R. Sen, A. Goeppert, S. Kar and G. S. Prakash, J. Am. Chem. Soc., 2020, 142, 4544–4549.
24. Q. Fu, Y. Kuramochi, N. Fukushima, H. Maeda, K. Sato and H. Kobayashi, Environ. Sci. Technol., 2015, 49, 1225–1232.
25. R. Sleat, R. A. Mah and R. Robinson, Int. J. Syst. Evol. Microbiol., 1985, 35, 10–15.
26. S. Xu, B. Fu, L. Zhang and H. Liu, World J. Microbiol. Biotechnol., 2015, 31, 941–950.
27. N. Monnerie, P. G. Gan, M. Roeb and C. Sattler, Int. J. Hydrogen Energy, 2020, 45, 26117–26125.
28. G. Lee, Y. C. Li, J.-Y. Kim, T. Peng, D.-H. Nam, A. Sedighian Rasouli, F. Li, M. Luo, A. H. Ip and Y.-C. Joo, et al. , Nat. Energy, 2021, 6, 46–53.
29. R. E. Siegel, S. Pattanayak and L. A. Berben, ACS Catal., 2022, 13, 766–784.
30. M. Li, K. Yang, M. Abdinejad, C. Zhao and T. Burdyny, Nanoscale, 2022, 14, 11892–11908.
31. S. E. Jerng and B. M. Gallant, iScience, 2022, 25, 104558.
32. I. Sullivan, A. Goryachev, I. A. Digdaya, X. Li, H. A. Atwater, D. A. Vermaas and C. Xiang, Nat. Catal., 2021, 4, 952–958.
33. M. Li, E. Irtem, H.-P. Iglesias van Montfort, M. Abdinejad and T. Burdyny, Nat. Commun., 2022, 13, 5398.
34. D. J. Heldebrant, J. Kothandaraman, N. Mac Dowell and L. Brickett, Chem. Sci., 2022, 13, 6445–6456.
35. S. H. Park, K. B. Lee, J. C. Hyun and S. H. Kim, Ind. Eng. Chem. Res., 2002, 41, 1658–1665.
36. R. L. Kent and B. Eisenberg, Hydrocarbon Processing, 1976, 55, 87–90.
37. A. Aboudheir, P. Tontiwachwuthikul, A. Chakma and R. Idem, Chem. Eng. Sci., 2003, 58, 5195–5210.
38. K. Zhu, C. Yue, Z. Wei, J. Huang, M. Hu, Y. Ji, H. Liu, H. Zhu, W. Guo and F. Zhou, et al. , Adv. Mater. Sci. Eng., 2022, 2022, 6278342.
39. M. Haji-Sulaiman, M. K. Aroua and A. Benamor, Chem. Eng. Res. Des., 1998, 76, 961–968.
40. A. Bandi, M. Specht, T. Weimer and K. Schaber, Energy Convers. Manage., 1995, 36, 899–902.
41. F. Zeman, AIChE J., 2008, 54, 1396–1399.
42. J. K. Stolaroff, D. W. Keith and G. V. Lowry, Environ. Sci. Technol., 2008, 42, 2728–2735.
43. D. W. Keith, M. Ha-Duong and J. K. Stolaroff, Clim. Change, 2006, 74, 17–45.
44. E. S. Sanz-Pérez, C. R. Murdock, S. A. Didas and C. W. Jones, Chem. Rev., 2016, 116, 11840–11876.
45. P. C. De Carvalho Pinto, T. V. Batista, G. De Rezende Ferreira, G. P. Voga, L. C. Oliveira, H. S. Oliveira, L. A. De Souza and J. C. Belchior, ChemistrySelect, 2022, 7, e202202731.
46. R. B. Said, J. M. Kolle, K. Essalah, B. Tangour and A. Sayari, ACS Omega, 2020, 5, 26125–26133.
47. P. V. Kortunov, M. Siskin, M. Paccagnini and H. Thomann, Energy Fuels, 2016, 30, 1223–1236.
48. A. Banerjee, C. Yue, J. Choi and C. G. Morales-Guio, AIChE J., 2024, e18560.
49. H. Liu, M. Li, X. Luo, Z. Liang, R. Idem and P. Tontiwachwuthikul, AIChE J., 2018, 64, 2515–2525.
50. P. W. Derks, P. J. Huttenhuis, C. van Aken, J.-H. Marsman and G. F. Versteeg, Energy Procedia, 2011, 4, 599–605.
51. J. P. Jakobsen, J. Krane and H. F. Svendsen, Ind. Eng. Chem. Res., 2005, 44, 9894–9903.
52. I. M. Bernhardsen and H. K. Knuutila, Int. J. Greenhouse Gas Control, 2017, 61, 27–48.
53. H. Seo, M. P. Nitzsche and T. A. Hatton, Acc. Chem. Res., 2023, 56, 3153–3164.
54. S. Jin, M. Wu, R. G. Gordon, M. J. Aziz and D. G. Kwabi, Energy Environ. Sci., 2020, 13, 3706–3722.
55. S. Jin, M. Wu, Y. Jing, R. G. Gordon and M. J. Aziz, Nat. Commun., 2022, 13, 2140.
56. W. Jiang, W. Liu, Y. Wang, Z. Zhao, Q. Li, Y. Wu, T. Liu and H. Xie, Ind. Eng. Chem. Res., 2022, 61, 13578–13588.
57. A. E. Al-Rawajfeh and M. N. Al-Amaireh, Desalin. Water Treat., 2009, 7, 191–197.
58. J. I. Lee, F. D. Otto and A. E. Mather, J. Appl. Chem. Biotechnol., 1976, 26, 541–549.
59. W. Böttinger, M. Maiwald and H. Hasse, Fluid Phase Equilib., 2008, 263, 131–143.
60. F. Y. Jou, A. E. Mather and F. D. Otto, Ind. Eng. Chem. Process Des. Dev., 1982, 21, 539–544.
61. G. Leverick, E. M. Bernhardt, A. I. Ismail, J. H. Law, A. Arifutzzaman, M. K. Aroua and B. M. Gallant, ACS Catal., 2023, 13, 12322–12337.
62. J. C. Bui, E. W. Lees, L. M. Pant, I. V. Zenyuk, A. T. Bell and A. Z. Weber, Chem. Rev., 2022, 122, 11022–11084.
63. J. Jang, M. Rüscher, M. Winzely and C. G. Morales-Guio, AIChE J., 2022, 68, e17605.
64. C. Morales-Guio, J. Jang, M. Ruscher, M. Winzely, D. Rodriguez, E. Reyes-Lopez, S. Srivastava, P. Christofides and P. Sautet, Research Square, 2024 DOI:10.21203/rs.3.rs-4189647/v1.
65. L. Yuan, S. Zeng, X. Zhang, X. Ji and S. Zhang, Mater. Rep.: Energy, 2023, 3, 100177.
66. V. E. Nelson, C. P. O’Brien, J. P. Edwards, S. Liu, C. M. Gabardo, E. H. Sargent and D. Sinton, ACS Appl. Mater. Interfaces, 2024, 16, 50818–50825.
67. D. Segets, C. Andronescu and U.-P. Apfel, Nat. Commun., 2023, 14, 7950.
68. S. Appelhaus, L. Ritz, S.-V. Pape, F. Lohmann-Richters, M. R. Kraglund, J. O. Jensen, F. Massari, M. Boroomandnia, M. Romanò and J. Albers, et al. , Int. J. Hydrogen Energy, 2024, 95, 1004–1010.
69. T. K. Todorova, M. W. Schreiber and M. Fontecave, ACS Catal., 2019, 10, 1754–1768.
70. J. T. Feaster, C. Shi, E. R. Cave, T. Hatsukade, D. N. Abram, K. P. Kuhl, C. Hahn, J. K. Nørskov and T. F. Jaramillo, ACS Catal., 2017, 7, 4822–4827.
71. J. Choi, S. Chiu, A. Banerjee, R. L. Sacci, G. M. Veith, C. Stieber, C. Hahn, A. N. Alexandrova and C. G. Morales-Guio, J. Phys. Chem. Lett., 2024, 15, 8007–8017.
72. Y. Hori, H. Wakebe, T. Tsukamoto and O. Koga, Electrochim. Acta, 1994, 39, 1833–1839.
73. M. Ma, B. J. Trześniewski, J. Xie and W. A. Smith, Angew. Chem., 2016, 128, 9900–9904.
74. Y. Lum, B. Yue, P. Lobaccaro, A. T. Bell and J. W. Ager, J. Phys. Chem. C, 2017, 121, 14191–14203.
75. Y. Hori, A. Murata and R. Takahashi, J. Chem. Soc., Faraday Trans. 1, 1989, 85, 2309–2326.
76. F. Pan and Y. Yang, Energy Environ. Sci., 2020, 13, 2275–2309.
77. M. Ma, K. Djanashvili and W. A. Smith, Angew. Chem., Int. Ed., 2016, 55, 6680–6684.
78. T.-T. Zhuang, Y. Pang, Z.-Q. Liang, Z. Wang, Y. Li, C.-S. Tan, J. Li, C. T. Dinh, P. De Luna and P.-L. Hsieh, et al. , Nat. Catal., 2018, 1, 946–951.
79. F. Pan, A. Liang, Y. Duan, Q. Liu, J. Zhang and Y. Li, J. Mater. Chem. A, 2017, 5, 13104–13111.
80. J. Liu, N. P. Wickramaratne, S. Z. Qiao and M. Jaroniec, Nat. Mater., 2015, 14, 763–774.
81. S. Chen and S.-Z. Qiao, ACS Nano, 2013, 7, 10190–10196.
82. P. P. Sharma, J. Wu, R. M. Yadav, M. Liu, C. J. Wright, C. S. Tiwary, B. I. Yakobson, J. Lou, P. M. Ajayan and X.-D. Zhou, Angew. Chem., 2015, 127, 13905–13909.
83. X. Cui, Z. Pan, L. Zhang, H. Peng and G. Zheng, Adv. Energy Mater., 2017, 7, 1701456.
84. D. Hursán, A. A. Samu, L. Janovák, K. Artyushkova, T. Asset, P. Atanassov and C. Janáky, Joule, 2019, 3, 1719–1733.
85. M. Li, M. N. Idros, Y. Wu, T. Burdyny, S. Garg, X. S. Zhao, G. Wang and T. E. Rufford, J. Mater. Chem. A, 2021, 9, 19369–19409.
86. K. Shi, Z. Ren, Z. Meng and X. Feng, ChemCatChem, 2024, 16, e202301308.
87. W. Ni, Y. Xue, X. Zang, C. Li, H. Wang, Z. Yang and Y.-M. Yan, ACS Nano, 2020, 14, 2014–2023.
88. A. Dutta, M. Rahaman, N. C. Luedi, M. Mohos and P. Broekmann, ACS Catal., 2016, 6, 3804–3814.
89. L. Chen, F. Li, Y. Zhang, C. L. Bentley, M. Horne, A. M. Bond and J. Zhang, ChemSusChem, 2017, 10, 4109–4118.
90. M. N. Hossain, S. Ahmad, I. S. da Silva and H.-B. Kraatz, Chem. – Eur. J., 2021, 27, 1346–1355.
91. J. H. Kim and T. J. Kang, ACS Appl. Mater. Interfaces, 2019, 11, 28894–28899.
92. J. Blake, J. Padding and J. Haverkort, Electrochim. Acta, 2021, 393, 138987.
93. C. T.-C. Wan, K. V. Greco, A. Alazmi, R. M. Darling, Y.-M. Chiang and F. R. Brushett, J. Electrochem. Soc., 2021, 168, 123503.
94. Y. Yoon, A. S. Hall and Y. Surendranath, Angew. Chem., Int. Ed., 2016, 55, 15282–15286.
95. S. Suter and S. Haussener, Energy Environ. Sci., 2019, 12, 1668–1678.
96. A. Böhme, J. C. Bui, A. Q. Fenwick, R. Bhide, C. N. Feltenberger, A. J. Welch, A. J. King, A. T. Bell, A. Z. Weber and S. Ardo, et al. , Energy Environ. Sci., 2023, 16, 1783–1795.
97. Z. Wang, Y. Zhou, P. Qiu, C. Xia, W. Fang, J. Jin, L. Huang, P. Deng, Y. Su and R. Crespo-Otero, et al. , Adv. Mater., 2023, 35, 2303052.
98. J. Zhang, W. Cai, F. X. Hu, H. Yang and B. Liu, Chem. Sci., 2021, 12, 6800–6819.
99. K. A. Gandionco, J. Kim, L. Bekaert, A. Hubin and J. Lim, Carbon Energy, 2024, 6, e410.
100. Y. Zhang, L. Jiao, W. Yang, C. Xie and H.-L. Jiang, Angew. Chem., Int. Ed., 2021, 60, 7607–7611.
101. X. Yang, J. Cheng, X. Yang, Y. Xu, W. Sun, N. Liu and J. Zhou, Nanoscale, 2022, 14, 6846–6853.
102. Z. Li, D. He, X. Yan, S. Dai, S. Younan, Z. Ke, X. Pan, X. Xiao, H. Wu and J. Gu, Angew. Chem., 2020, 132, 18731–18736.
103. Y. Huang, X. Mao, G. Yuan, D. Zhang, B. Pan, J. Deng, Y. Shi, N. Han, C. Li and L. Zhang, et al. , Angew. Chem., 2021, 133, 15978–15982.
104. V. Okatenko, A. Loiudice, M. A. Newton, D. C. Stoian, A. Blokhina, A. N. Chen, K. Rossi and R. Buonsanti, J. Am. Chem. Soc., 2023, 145, 5370–5383.
105. Y. Xu, C. Li, Y. Xiao, C. Wu, Y. Li, Y. Li, J. Han, Q. Liu and J. He, ACS Appl. Mater. Interfaces, 2022, 14, 11567–11574.
106. M. Ding, Z. Chen, C. Liu, Y. Wang, C. Li, X. Li, T. Zheng, Q. Jiang and C. Xia, Mater. Rep.: Energy, 2023, 3, 100175.
107. J. Gao, H. Zhang, X. Guo, J. Luo, S. M. Zakeeruddin, D. Ren and M. Gratzel, J. Am. Chem. Soc., 2019, 141, 18704–18714.
108. N. B. Watkins, Z. J. Schiffer, Y. Lai, C. B. Musgrave III, H. A. Atwater, W. A. Goddard III, T. Agapie, J. C. Peters and J. M. Gregoire, ACS Energy Lett., 2023, 8, 2185–2192.
109. F. P. Garca de Arquer, C.-T. Dinh, A. Ozden, J. Wicks, C. McCallum, A. R. Kirmani, D.-H. Nam, C. Gabardo, A. Seifitokaldani and X. Wang, et al. , Science, 2020, 367, 661–666.
110. X. Chen, J. Chen, N. M. Alghoraibi, D. A. Henckel, R. Zhang, U. O. Nwabara, K. E. Madsen, P. J. Kenis, S. C. Zimmerman and A. A. Gewirth, Nat. Catal., 2021, 4, 20–27.
111. K. P. Kuhl, E. R. Cave, D. N. Abram and T. F. Jaramillo, Energy Environ. Sci., 2012, 5, 7050–7059.
112. L.-C. Weng, A. T. Bell and A. Z. Weber, Phys. Chem. Chem. Phys., 2018, 20, 16973–16984.
113. A. Goyal, C. J. Bondue, M. Graf and M. T. Koper, Chem. Sci., 2022, 13, 3288–3298.
114. J. Newman, J. Electrochem. Soc., 1966, 113, 1235.
115. E. L. Clark, S. Ringe, M. Tang, A. Walton, C. Hahn, T. F. Jaramillo, K. Chan and A. T. Bell, ACS Catal., 2019, 9, 4006–4014.
116. H. Ma, E. Ibáñez-Alé, R. Ganganahalli, J. Pérez-Ramrez, N. López and B. S. Yeo, J. Am. Chem. Soc., 2023, 145, 24707–24716.
117. M. Abdinejad, Z. Mirza, X.-A. Zhang and H.-B. Kraatz, ACS Sustainable Chem. Eng., 2020, 8, 1715–1720.
118. A. Khurram, L. Yan, Y. Yin, L. Zhao and B. M. Gallant, J. Phys. Chem. C, 2019, 123, 18222–18231.
119. M. C. Monteiro, F. Dattila, B. Hagedoorn, R. Garca-Muelas, N. López and M. T. Koper, Nat. Catal., 2021, 4, 654–662.
120. J. Wu, W. Li, K. Liu, A. Kucernak, H. Liu, L. Chai and M. Liu, Next Energy, 2023, 1, 100032.
121. S. Ringe, E. L. Clark, J. Resasco, A. Walton, B. Seger, A. T. Bell and K. Chan, Energy Environ. Sci., 2019, 12, 3001–3014.
122. S. Banerjee, X. Han and V. S. Thoi, ACS Catal., 2019, 9, 5631–5637.
123. S. Sarkar, A. Maitra, S. Banerjee, V. S. Thoi and J. M. Dawlaty, J. Phys. Chem. B, 2020, 124, 1311–1321.
124. S. Banerjee, Z.-Q. Zhang, A. S. Hall and V. S. Thoi, ACS Catal., 2020, 10, 9907–9914.
125. J. Resasco, L. D. Chen, E. Clark, C. Tsai, C. Hahn, T. F. Jaramillo, K. Chan and A. T. Bell, J. Am. Chem. Soc., 2017, 139, 11277–11287.
126. M. R. Singh, Y. Kwon, Y. Lum, J. W. Ager III and A. T. Bell, J. Am. Chem. Soc., 2016, 138, 13006–13012.
127. O. Ayemoba and A. Cuesta, ACS Appl. Mater. Interfaces, 2017, 9, 27377–27382.
128. M. Sadeghpour, R. Yusoff and M. K. Aroua, Rev. Chem. Eng., 2017, 33, 183–200.
129. X. Ren, T. D. Myles, K. N. Grew and W. K. Chiu, J. Electrochem. Soc., 2015, 162, F1221.
130. E. W. Lees, B. A. Mowbray, D. A. Salvatore, G. L. Simpson, D. J. Dvorak, S. Ren, J. Chau, K. L. Milton and C. P. Berlinguette, J. Mater. Chem. A, 2020, 8, 19493–19501.
131. J. C. Bui, C. Kim, A. J. King, O. Romiluyi, A. Kusoglu, A. Z. Weber and A. T. Bell, Acc. Chem. Res., 2022, 55, 484–494.
132. A. M. Appel and J. Y. Yang, ACS Energy Lett., 2024, 9, 768–770.
133. R. Ramezani, S. Mazinani and R. Di Felice, Rev. Chem. Eng., 2022, 38, 273–299.
134. C. Song, Q. Liu, S. Deng, H. Li and Y. Kitamura, Renewable Sustainable Energy Rev., 2019, 101, 265–278.
135. E. S. Rubin, H. Mantripragada, A. Marks, P. Versteeg and J. Kitchin, Prog. Energy Combust. Sci., 2012, 38, 630–671.
136. T. Wilberforce, A. Baroutaji, B. Soudan, A. H. Al-Alami and A. G. Olabi, Sci. Total Environ., 2019, 657, 56–72.
137. W. L. Theo, J. S. Lim, H. Hashim, A. A. Mustaffa and W. S. Ho, Appl. Energy, 2016, 183, 1633–1663.
138. A. Gautam and M. K. Mondal, Fuel, 2023, 334, 126616.
139. Y. Guan, J. Kümper, S. D. Mürtz, S. Kumari, P. J. Hausoul, R. Palkovits and P. Sautet, Chem. Sci., 2024, 15, 14485–14496.
140. S. Mondal and S. C. Peter, Adv. Mater., 2024, 36, 2407124.
141. Y. Xu, J. P. Edwards, J. Zhong, C. P. O’Brien, C. M. Gabardo, C. McCallum, J. Li, C.-T. Dinh, E. H. Sargent and D. Sinton, Energy Environ. Sci., 2020, 13, 554–561.
142. D. J. Pimlott, A. Jewlal, B. A. Mowbray and C. P. Berlinguette, ACS Energy Lett., 2023, 8, 1779–1784.
143. M. Gautam, D. T. Hofsommer, S. S. Uttarwar, N. Theaker, W. F. Paxton, C. A. Grapperhaus and J. M. Spurgeon, Chem Catal., 2022, 2, 2364–2378.
144. Y. Kwon, B. Wu, N. Zhang, D. Hand, T. Mou, X. Han and Q. Chang, MRS Commun., 2024, 1–13.
145. R. I. Masel, Z. Liu, H. Yang, J. J. Kaczur, D. Carrillo, S. Ren, D. Salvatore and C. P. Berlinguette, Nat. Nanotechnol., 2021, 16, 118–128.
146. M. D. J. Gálvez-Vázquez, P. Moreno-Garca, H. Guo, Y. Hou, A. Dutta, S. R. Waldvogel and P. Broekmann, ChemElectroChem, 2019, 6, 2324–2330.
147. Y. Kim, E. W. Lees, C. Donde, A. M. Jewlal, C. E. Waizenegger, B. M. de Hepcée, G. L. Simpson, A. Valji and C. P. Berlinguette, Joule, 2024, 6, 3106–3125.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ey00285g

---