Co-electrolysis of CO2 and H2O: From electrode reactions to cell-level development

Abstract The electroreduction of CO2 into value-added products (e.g. CO) constitutes an excellent means of decreasing this greenhouse gas emissions, but limited efforts have been devoted to the implementation of this reaction within the so-called co-electrolysis cells operating at process-relevant currents >> 100 mA·cmgeom−2. Reaching such performances shall require a combination of gas-fed reactants and the corresponding diffusion electrodes, along with ion-exchange membranes and ionomers that set the operative pH at the cells' cathode and anode. The latter constitutes a key design parameter that must be combined with the need to minimize the crossover of reaction products and/or (bi)carbonate anions from the cathode to the anode, whereby their reoxidation to carbon dioxide leads to a decrease in the device's net CO2 consumption.


Introduction
Attaining the global temperature increase target of 2 C set by the Paris Climate Agreement requires a careful combination of decarbonization policies (e.g. carbon taxation) and sound business cases for novel, Cfootprint reduction technologies [1]. The transformation of anthropogenic CO 2 emissions into valueadded chemicals and fuels is increasingly regarded as a key technological enabler to reach this objective, and electrochemical CO 2 -to-chemicals pathways are considered particularly promising to this end [2,3]. With this motivation, research on CO 2 electroreduction over the last z3 decades has preponderantly focused on understanding and improving the sluggish kinetics of this reaction on a large variety of catalytic systems (from model, extended metal surfaces to atomically dispersed sites) [4,5]. However, such studies are predominantly conducted in quasi-stagnant electrochemical cells filled with aqueous electrolytes with an xintrinsically low conductivity and limited CO 2 solubility (z0.1 S$cm À1 and z30 mM, respectively), thus limiting the attainable currents to values of z1 to z 10 mA$cm À2 that remain z1e2 orders of magnitude below the !200 mA$cm À2 threshold for amenability within industrially relevant processes [6].
In an effort to bridge this gap, an unprecedentedly large number of studies published over the last z5 years have explored the implementation of CO 2 electroreduction catalysts in electrochemical setups in which currents in the hundreds of milliamps per centimeter square range can be drawn. Specifically, herein we focus on those works in which these current regimes have been attained by feeding gaseous CO 2 to a gas diffusion electrode combined with a membrane electrolyte, thus following the membrane electrode assembly approach well established for proton-and anion-exchange membrane (PEM, AEM) fuel cells and electrolyzers. As schematized in Figure 1a, the resulting, so-called coelectrolysis cell is constituted by a cathode and an anode generally fed with this CO 2 (g) and H 2 O, and on which the corresponding reduction and oxidation reactions lead to the evolution of value-added C-products (vide infra) and O 2 , respectively.

Electrochemical reactions and acid-base equilibria in co-electrolysis cells
The extensive work devoted to the study of CO 2 electroreduction kinetics discussed previously has served to establish the large variety of C-products that can be derived from this reaction (e.g. CH 4 , C 2 H 4 ) [7,8], but recent technoeconomic analyses have concluded that the electrochemical production of CO has the greatest chance to become cost competitive with regard to current chemical routes (while the business case for formate/formic acid production remains elusive owing to its small market size) [6,9e11]. As it will be discussed in the following section, the pH at which this and all other CO 2 electroreduction reactions are undertaken plays a determining role on their selectivity and on the corresponding device's features (from membrane choice to expected by-products), and thus, in Figure 2, we have plotted the effect of this key variable on the reversible potentials (E rev s) of this and all other co-electrolyzer reactions [12,13]. Notably, the E rev vs. pH line for CO 2 reduction to CO (formulated in Eq. 1 of Figure 2) almost overlaps with that of the H 2 oxidation/evolution equilibrium (cf. Eq. 2) and, as a result, hydrogen is a common co-electrolyzer cathode by-product. Chiefly, the cell's cathode cannot be operated under acidic reaction conditions to minimize this H 2 yield because the enhancement of the H 2 evolution kinetics at low vs. high pHs on most metal surfaces [5] would lead to the exclusive evolution of H 2 , as opposed to the desired reduction of CO 2 [13].
Alternatively, co-electrolyzer cathodes are operated at alkaline to quasi-neutral pHs at which the fed CO 2 readily equilibrates with OH À to yield (bi)carbonate anions (see the discussion in the following sections and the chemical equilibria lines in Eqs. 6 and 7 of Figure 2). Moreover, the production of hydroxyl groups (OH À ) associated with the cathodic reactions (see the alkaline medium formulations of Eqs. 1 and 2) implies an additional, current-driven increase in the local pH that can affect the interfacial concentration of CO 2 and the corresponding reaction products' distribution [14,15]. Complementarily, in alkaline and quasi-neutral media, the anodic O 2 evolution reaction (OER) is associated with a consumption of OH À groups that causes a concomitant drop in the interfacial pH (see Eq. 3); nevertheless, if that pH remains sufficiently high to grant the presence of (bi)carbonate anions (i.e. !6.4 e see Eq. 6), the large OER overpotentials will likely cause the additional oxidation of these HCO 3 À /CO 3 2À back to CO 2 (cf. Eqs. 4 and 5) [13].
Cell configurations as a function of the electrolyte/electrode pH To discuss the operational challenges faced by these gasphase devices, we first discuss their architecture and design, which can generally be categorized into two types [16e19]: one in which the two electrodes are in direct contact with the polymer electrolyte membrane (sometimes referred to as 'zero-gap' configuration) [20] and a second one featuring an additional liquid electrolyte compartment between the cathode catalyst layer (CL) and the membrane [21]. In the latter, 'flow cell' configuration, a highly concentrated hydroxide or bicarbonate aqueous solution is typically pumped through this compartment, and this forced electrolyte Schematic representation of an MEA-based co-electrolysis cell and of the three different catalyst layer -membrane interfaces discussed herein. (a) Co-electrolysis cell constituted by a cathode and an anode compartment fed with gaseous CO 2 vs. water reactants that get reduced/oxidized to CO and O 2 , respectively. (b) Cell configuration implementing an anion-exchange membrane (AEM) that transports anions from cathode to anode, where they get oxidized back to CO 2 ., (c) Cell with a reverse bias bipolar membrane (RB-BPM) in which water is split into protons and hydroxyl groups at the interface between cation-and anion-exchange layers (CEL, AELs). (d) Co-electrolyzer using a forward-bias bipolar membrane (FB-BPM) in which (bi)carbonate anions and protons recombine at the AEL-CEL membrane yielding CO 2 and H 2 O. Note that the species between parentheses in (b) and (c) represent the accessory addition of concentrated alkaline or (bi)carbonate solutions (e.g. KOH or KHCO 3 ) in the cell's reactant stream. convection helps to overcome mass transport limitations and favors high current density performance [22,23].
In the zero-gap case, the exclusive presence of a membrane solid electrolyte implies that the latter (along with the ionomers implemented in the cathodic and anodic CLs) determines the pH of the electrodes' reaction environment. Because of the limitations in the choice of acidic reaction media discussed previously, numerous studies have instead implemented AEMs and anionexchange ionomers used in alkaline fuel cells, whereby these operate in their OH À -exchange form. However, the latter state cannot be sustained upon contact with the gaseous CO2 feed in a co-electrolysis cell, which leads to a well-documented carbonation of the membrane that in terms causes a~2-to~10-fold decrease in its ionic conductivity with respect to its hydroxide form [24]. Despite this higher resistivity, AEM coelectrolysis cells display very appealing performances, especially when implementing imidazolium-based membranes and ionomers for which CO 2 -to-CO faradaic efficiencies >90% at a current density of 200 mA$cm À2 and an overall cell potential of 3 V have been reported [23e29]. However, as schematized in Figure 1b, the cathode-to-anode transport of HCO3/ CO3 species across the membrane (i.e. 'crossover') has been identified as a major limitation of AEM-based devices [29e32]. More precisely, these anionic species have been claimed to be neutralized by protons [31] and/or (more likely) to be electrooxidized [30] back to CO2 upon reaching the cell's anode, which leads to a low overall CO 2 utilization of the device in which only less than half of all the fed carbon dioxide is actually devoted to the production of the desired C products. In addition, when such AEM-based devices are co-fed with concentrated KOH and/or potassium (bi)carbonate solutions, the high consumption of water at their cathodes upon high current density operation reportedly leads to an oversaturation with K 2 CO 3 that precipitates and blocks the access of the reactants to the catalyst's surface, thus leading to cell failure [20,33].
The crossover problem discussed previously can be overcome by substituting the AEM with a bipolar membrane (BPM), in which an anion-exchange layer (AEL) and a cation-exchange layer (CEL) are laminated against each other [34,35]. This configuration has mostly been implemented with the BPM operated in a so-called 'reverse bias' (RB) mode (i.e. the RB-BPM mode schematized in Figure 2c), in which the AEL and CEL are located at the cell's anode and cathode, respectively [36]. With this orientation, the dissociation of water at the AELeCEL interface leads to a sustained supply of H þ and OH À ions toward the cathode and anode CLs, which helps to maintain constant pHs at both electrodes [36À38] and to sustain 24 h of continuous operation at 100 mA$cm À2 and z3.5 V with a CO faradaic efficiency of z70% [37]. However, this configuration has systematically been implemented in flow cells including a supply of concentrated bicarbonate (or water) vs. hydroxide solutions to the cell's cathode vs. anode, respectively, practically complicating Pourbaix diagram displaying the effect of the electrolyte's pH on the reversible potential (E rev , at 298 K and assuming 1 M or 1 bar partial concentrations/ pressures for liquid vs. gaseous species, respectively, and reported in volts vs. the reversible hydrogen electrode (VSHE)) of the cathodic and anodic electrochemical reactions (Eqs. 1-5, including their standard potentials [E 0 ] and, when applicable, formulated in the acid and base forms) and acid-base equilibria (Eqs. 6 and 7) at play in co-electrolysis cells [12,13].
the device's operation [36e38]. Most importantly, we are not aware of any studies in which the possible cathode-to-anode crossover of CO 2 (g) and its carbonation and likely oxidation at the latter electrode (following Eqs. 4 and 5 in Figure. 2) have been quantified, and thus, this configuration's net CO 2 consumption remains unclear.
Alternatively, (bi)carbonate crossover can be completely suppressed using the forward bias (FB) BPM configuration, schematized in Figure 1d and first implemented in the study by Patru et al [39], whereby HCO 3 À /CO 3 2À ions produced at the anode CL and transported across the membrane's AEL react with the protons in the acidic CEL forming CO 2 and H 2 O. Chiefly, the same study [39] demonstrated that this FB-BPM setup also allows reaching significantly larger current densities than in the RB-BPM configuration because in the latter case this key operational parameter is limited by the splitting of H2O taking place at the interface between charge exchange layers. Moreover, the optimization of the AEL-CEL interface to avoid the layers' delamination due to this evolution of H 2 O and CO 2 at their junction led to a CO Faradaic efficiency of = 10 % at 100 mA cme2 and an overall cell potential of = 2.7 V, with H 2 production becoming preponderant at larger current densities / cell potentials. Thus, even if this FB-BPM configuration constitutes the only cell layout for which the suppression of (bi)carbonate crossover and reoxidation have unambiguously been proven, further improvements of its operation are required to achieve a CO 2 reduction product selectivity on pair to that reported with AEM and RB-BPM cells, as well as to improve the evacuation of produced water (and corresponding electrode flowing) reported in the study by Patru et al [39].
On the importance of OER electrocatalysis at quasi-neutral pHs While the latter FB-BPM configuration implies the operation of the cell's anode under strongly acidic conditions that limit the choice of OER catalyst materials to the scarce Ir oxides customarily implemented in PEM electrolyzers [5,40], the neutral to high pH at the cell's  anode concomitant to the two other operational modes (AEM, RB-BPM) allows implementing inexpensive, oxide-based OER catalysts. Nevertheless, although highly active and relatively stable OER catalysts have been reported for the alkaline environments intrinsic to alkalinedand AEMdelectrolyzers [5], significantly less is known concerning materials displaying a comparably high OER activity and stability under the quasi-neutral pHs concomitant to co-electrolyzers' anodes.
In this regard, the relatively limited number of studies devoted to such OER catalysts in neutral media has focused on Coephosphate [41e43], Coeborate [44,45], and Ni based hydroxides [46], borates [47], or carbonates [48] that display a relatively promising OER activity compared with Ir-based catalysts. Moreover, a recent study dealing with NiCoFeP oxyhydroxides postulated that transition metals in the highest possible valence state should present the enhanced OER activities in the neutral environment [49]. Based on this, density functional theory calculations of Ni-based catalysts suggested that the formation energy of Ni 4þ could be modulated by incorporating Co, Fe, and P to the material. This led to investigations of the oxidation state of NiP, NiCoP, and NiCoFeP oxyhydroxides by in situ soft X-ray absorption spectroscopy in neutral pH (cf. Figure 3a), whereby a correlation between the presence of Ni 4þ under operative conditions and the corresponding materials' OER activity was experimentally validated (see Figure 3b). As shown in Figure 3c, NiCoFeP featured the highest Ni 4þ :Ni 2þ ratio and correspondingly highest OER activity among these catalysts in 0.5 M KHCO 3 , outperforming an IrO 2 benchmark and retaining this activity over 100 h of operation [49].
Alternatively, Sr 2 GaCoO 5 brownmillerite oxide (see Figure 3d for the crystallographic structure) has recently been reported to outperform NiCoFeP and IrO 2 as an OER catalyst at pH 7 (cf. Figure 3e) [50], and this catalytic performance appears to be insensitive to the electrolyte's pH when compared with that of two other materials (Figure 3f) This suggests that a different reaction mechanism is at play as a function of the electrolyte's pH, and operando X-ray absorption spectroscopy measurements indicate that while in alkaline environment the oxidation of lattice oxygen [52] can take place in parallel to the conventional OER mechanism, the latter becomes predominant at quasi-neutral pHs.

Conclusions and outlook
In summary, the aforementioned discussed studies demonstrate the potential of co-electrolysis cells to reduce CO 2 into value-added products at applicationrelevant current densities, while highlighting the importance of numerous cell design aspects that were overlooked in previous electrocatalysis works but that are extremely relevant to the operation of such devices (e.g. membrane configuration, water management). Despite significant advancements, the optimum cell configuration for scale-up remains unclear and may be dictated by the preferred reaction product(s) and subsequent integration within industrially relevant processes. Notably, this ideal configuration should only be adopted after a rigorous assessment of its net CO 2 consumption, including a quantification of the possible evolution of carbon dioxide at the cell's anode. Finally, more research efforts should be devoted to the latter electrode, whereby highly active and stable OER catalysts based on earth-abundant elements and capable of operating at quasi-neutral pHs are urgently needed to assure rapid co-electrolyzer development. We postulate that this OER catalyst design shall be greatly accelerated by combining synthetic efforts with fundamental studies including operando characterization, as to elucidate the parameters that determine these materials' OER activity and stability in such quasi-neutral reaction environments.

Conflict of interest statement
Nothing declared.  Manuscript featuring the best RB-BPM co-electrolysis performance reported to date, and showcasing the importance of membrane hydration (either through gas humidification or by co-feeding water / buffer solution) to attain stable cell operation.