Operando cathode activation with alkali metal cations for high current density operation of water-fed zero-gap carbon dioxide electrolyzers

Abstract

Continuous-flow electrolyzers allow CO₂ reduction at industrially relevant rates, but long-term operation is still challenging. One reason for this is the formation of precipitates in the porous cathode from the alkaline electrolyte and the CO₂ feed. Here we show that while precipitate formation is detrimental for the long-term stability, the presence of alkali metal cations at the cathode improves performance. To overcome this contradiction, we develop an operando activation and regeneration process, where the cathode of a zero-gap electrolyzer cell is periodically infused with alkali cation-containing solutions. This enables deionized water-fed electrolyzers to operate at a CO₂ reduction rate matching that of those using alkaline electrolytes (CO partial current density of 420 ± 50 mA cm⁻² for over 200 hours). We deconvolute the complex effects of activation and validate the concept with five different electrolytes and three different commercial membranes. Finally, we demonstrate the scalability of this approach on a multi-cell electrolyzer stack, with a 100 cm² / cell active area.

The capture, storage, and utilization of carbon dioxide (CO₂) has entered to the central stage of science in the past decade¹. Among other strategies, significant attention is devoted to its electrochemical valorization^2,3, allowing the simultaneous decrease of the emitted CO₂ amounts, and the production of raw chemicals, such as carbon monoxide (CO) or formic acid (HCOOH). Employing renewable electricity, the industrial implementation of this technology might be a step in closing the artificial carbon cycle^4,5. For industrialization, however, several requirements must be fulfilled simultaneously, such as operation at high current density (j), high energy efficiency (EE), high product selectivity (Faradaic efficiency, FE), and long-term stability^6,7. Besides the catalysts,^8,9 electrolyzer cell design is equally important in determining these parameters¹⁰.

To achieve high reaction rate, CO₂ reduction (CO₂R) must be performed in continuous-flow electrolyzers, encompassing gas diffusion electrodes (GDEs). In such devices, CO₂ gas is fed through a porous cathode support to the catalyst (together forming the GDE)^11,12. This leads to a thin hydrodynamic boundary layer (few 10s of nm) through which CO₂ must diffuse to reach the catalyst (in contrast to aqueous solutions), allowing high current density operation.¹² Two different types of low temperature electrolyzer cells are widely applied: fuel-cell-like (zero-gap) setups, operating without liquid catholyte, where catalyst layers are only separated by a membrane; and microfluidic cells, with continuous liquid electrolyte feed(s). In the latter case, either separate anolyte and catholyte are fed to membrane separated electrodes, or one solution flows between the anode and the cathode.²

In microfluidic cells, highly alkaline conditions allowed high current density (>1 A cm⁻²) production of CO¹³, ethylene^14,15, methane^16,17 and multi-carbon products¹⁸. While these results are very promising, the scale-up of microfluidic electrolyzers seems challenging.² In our opinion, the use of zero-gap electrolyzer cells is more promising for industrial applications. Such setups also offer a straightforward implementation of alternative anode reactions, such as glycerol oxidation, to form high-value products on both electrodes, with high EE^19–21.

Anion exchanging polyelectrolyte membranes (AEM) are typically used as separators in zero-gap electrolyzer cells to avoid high local H⁺ concentration and therefore excessive H₂ evolution on the cathode. The direction of ion conduction is from the cathode to the anode (through the membrane), maintained partially by the electrogenerated OH⁻ ions, but mostly by CO₃ ²⁻ and HCO₃ ⁻ ions, formed in the reaction of CO₂ and OH⁻.^22,23 Consequently, AEM-based operation shall be independent from whether an alkaline solution or pure water is fed to the anode (given that the anode catalyst functions in both media). Still, in contrast to the wealth of alkaline studies,^10,22–26 to the best of our knowledge, there is no study in the literature where a zero-gap electrolyzer cell with a commercial membrane was operated with water anolyte at high CO₂R current. In one single study, high current density was ensured using an experimental membrane.²⁷

In this study, our aim was to understand which parameters dictate high reaction rate and product selectivity in AEM-based zero-gap electrolyzers and why is it generally required to use concentrated alkaline anolytes.²⁸ Cation crossover from the anode to the cathode was identified as a major contributor to the high performance, but it also resulted in precipitate formation in the cathode GDE. To turn this challenge into an opportunity, we designed a scalable process and experimental setup to operate electrolyzer cells with different commercial membranes and pure water anolyte at high current density, by periodically infusing the cathode with different alkali cation containing solutions.

Performance fading due to precipitate formation

After longer operation of a zero-gap electrolyzer cell (Fig. 1a, Supplementary Figure 1) with alkaline anolyte, salt precipitates at the cathode, gradually decreasing its performance^24,25,29. We also experienced this phenomenon (Supplementary Note 1), in the form of continuous product formation current decrease (Supplementary Figure 2a) and/or pressure build-up. The formation of KHCO₃ and K₄H₂(CO₃)₃×1.5 H₂O precipitate (Supplementary Figure 2b) was confirmed by XRD analysis and subsequent Rietveld-refinement (Supplementary Figure 2c). The second compound typically forms through the CO₂ sorption of K₂CO₃, when large excess of CO₂ is available.^30,31 This precipitate forms because of the unintended cation crossover through the AEM from the anolyte to the cathode (as verified by ion chromatography (Fig. 1b)), where it reacts with the CO₂ feed and the electrogenerated OH⁻.

Fig. 1.

Unintended cation crossover and precipitate formation in alkaline anolyte fed zero-gap CO₂ electrolyzers. (a) Schematic illustration of the operation of an AEM separated zero-gap CO₂ electrolyzer cell with alkaline anolyte. (b) Ion chromatographic quantification of K⁺ and Cs⁺ crossover through different commercially available AEMs during CO₂ electrolysis in a zero-gap electrolyzer cell. (c) Cross-section SEM-EDX and (d) micro-CT images of a GDE after continuous CO₂ electrolysis in a zero-gap cell (T = 50 °C 1 M KOH anolyte, ΔU = 3.0 V). The red and green colors in the SEM-EDX and Micro-CT images represent Ag and K atoms, respectively. Such experiments were repeated on separate cell assemblies independently 3 times, with similar results.

Scanning electron microscopy-X-ray microanalysis (SEM-EDX) and computed micro-tomography (micro-CT) studies revealed that precipitation occurs not only on the backside of the GDE (i.e., in the gas-flow channels), but also in its deeper regions (Fig. 1c). A compact layer of the precipitate was observed in the microporous layer and at the junction of the micro- and macroporous layers (see Supplementary Figures 3-5, and discussion therein). This hampers CO₂ from reaching the catalyst surface explaining the continuously decreasing product formation rate (Supplementary Figure 6).

The state-of-the-art method to overcome this issue is to rinse the cathode with water^24,32,33. The performance of the cell, however, cannot be fully restored by this method, as water can only wash the deposits from the backside of the GDE, and cannot penetrate the hydrophobic gas diffusion layer (GDL), unless excessive force (pressure) is applied. Such pressure would in turn damage the GDE structure, causing electrode flooding^24,34. Decreasing the anolyte concentration extends the operation to prolonged timescales. The crossover of K⁺ or Cs⁺ ions (Fig. 1b), and the consequent precipitate formation and performance fading still occur (see Supplementary Figure 6 and discussion therein). Increasing the humidity of the CO₂ gas stream is another alternative mitigation strategy, but it changes the overall operation of the cell^24,29. An elegant solution to this problem would be to operate such electrolyzer cells with pure water anolyte, inherently preventing precipitate formation in the cathode GDE.

Operation of CO₂ electrolyzers with pure water anolyte

Using pure water as anolyte instead of 0.1 M KOH the current density drops to its third (100 mA cm⁻² vs. 300 mA cm⁻², Supplementary Figure 7a). This large difference appears also in the low frequency limit of the impedance spectra (Supplementary Figure 7b) – which is a measure of the derivate of the current-voltage curve. In contrast, the high frequency limits do not differ much; it is around 1 Ω cm² for both; comparable to what was reported for similar systems.^10,27 This value is related to the total cell resistance, including the membrane, of which contribution to the total impedance spectra was found Ohmic in our control experiments (Supplementary Note 2, Supplementary Figure 8). The similarity of that value is not surprising, since the membrane conductance is mostly affected by the operational conditions and the charge transporting ions.^35,36 Analyzing the anode gas composition and the complete mass balance of the process, CO₃ ²⁻ was identified as the majority charge carrier through the AEM, irrespective of the used anolyte (Supplementary Note 3, Supplementary Figure 9). The transference number of the cations remain in the 10⁻⁴-10⁻⁵ range for both Cs⁺ and K⁺, which practically excludes their contribution to the membrane conductivity (see Fig. 1b and related discussion in Supplementary Note 4).

As both the membrane resistance and the charge carriers are the same in the two cases, the big difference in the currents with and without KOH in the anolyte, must be attributed to changes at the cathode and/or anode interfaces. With pure water anolyte, two arcs can be observed on the impedance spectra (Supplementary Figure 7b). In general, these are due to interfacial capacitances and charge transfer resistances on any of the electrodes; their exact origin is beyond our present scope. Notably, the sum of the spans (“diameters”) of these arcs is about three times larger when the cell is operated with pure water anolyte instead of 0.1 M KOH, indicating a higher total charge transfer resistance of the electrode processes. Notably, in a zero-gap cell the effect of the membrane and the catalyst/ionomer layer cannot be fully separated. In all our analysis, we consider the ionomer and the catalyst/ionomer/membrane interfaces to belong to the “catalyst” and not the membrane.

Activating the cathode catalyst by infusion with alkali cation containing solutions

Considering the results of previous studies on the promoting effect of alkali cations in electrochemical CO₂R,^37–44 cation permeation through the membrane from the anolyte might contribute to achieve high reaction rates (which is absent in the case of water-based operation). In microfluidic electrolyzers operating with liquid catholyte, alkali ions are inherently present around the cathode catalyst, which might (partly) explain why these systems outperform their zero-gap counterparts in terms of the achievable current densities^13,45. Furthermore, a long activation period (~60 min) was observed for zero-gap cells operated with dilute alkaline anolytes (Supplementary Figure 10), indicating that cations, slowly crossing the AEM (note the small transference number in Fig. 1b) and reaching the cathode surface, have a boosting effect on the operation.

To validate this notion, we designed a process and an experimental setup allowing the injection of electrolyte solutions into the CO₂ feed, which is in turn pushed through the cathode GDE by the gas stream (Fig. 2). This solution infusion method (referred hereafter as regeneration) allows the efficient removal of the precipitate potentially formed in the GDE (Fig. 1c), and also introduces alkali metal cations to the cathode (referred hereafter as activation). Repeatedly performing the activation, each new portion of the solution dissolves any precipitate from the GDE, incidentally formed in/after the previous activations. During the solution infusion, the CO₂ gas feed pushes the “solution plug” through the cell, and therefore there is only a limited contact area between the solution and the gas where precipitate might form in situ. This leads to a constant presence of the activating ions at the cathode GDE but avoids accumulation.

Fig. 2.

Schematic piping and instrumentation diagram of the test framework employed. In the inset “1” shows the default positions of the manual valves, forming a continuous gas path to the cell, bypassing the activation loop. Turning the valves into position “2” the gas is driven through the activation loop, carrying the activation fluid into the cell.

To enable the electrolyte solution to reach the catalyst layer, the composition of the solvent mixture must be tailored to the wetting properties of the GDE. Contact angles were determined for a series of solvent mixtures (Fig. 3a). As the GDL in our case is a hydrophobic carbon paper, a solvent mixture of 25 V/V% isopropanol in water was employed for further studies. This wets the GDE properly, without dissolving its polytetrafluoroethylene (PTFE) content or damaging the AEM (at least within the studied timeframe), and it dissolves the studied salts.

Fig. 3.

Cathode activation using different commercially available AEMs. (a) Contact angles of different water/isopropanol solvent mixtures on the microporous side of a Sigracet 39BC GDL. (b-d) Chronoamperometric curves and CO-formation partial current densities (T_cathode = 60 °C, 12.5 cm³ cm⁻² min⁻¹ CO₂ feed rate, pure water anolyte)measured using (b) Sustainion X37-50 (ΔU = 3.1 V), (c) PTFE-reinforced 15 μm thick PiperION TP-85 (ΔU = 3.2 V), (s) Fumasep FAB-PK-130 (ΔU = 3.1 V)AEMs. 10 cm³ 0.5 M KOH (for (b) and (d)) or 0.5 M CsOH (c) solutions in 1:3 isopropanol/water mixture were used to activate the cathode at the times marked with asterisks in the figures. Such experiments were repeated on separate cell assemblies independently at least 3 times, with similar results. The values in (a) are the mean value of 3 independent measurements, together with the calculated standard deviations.

Upon activating the cathode GDE with 10 cm³ (ca. 50 times the free volume of the cathode compartment) 0.5 M KOH solution, meanwhile operating the electrolyzer cell with pure water anolyte, a three-fold increase in the CO formation rate (j_CO) was observed (Fig. 3b-d). For the PiperION membrane (Fig. 3c), the j_co=500-550 mA cm⁻² represents higher current density than the state-of-the-art water anolyte operated CO₂ electrolyzer cell.²⁷ The current does not drop after the activating solution leaves the cell, but remains relatively stable, with a slow decay (to be discussed later). Importantly, this effect was very similar for three commercial membranes (Sustainion X37-50, PiperION TP-85, Fumasep FAB-PK-130), of very diverse chemical composition, thickness etc. Except a short transient period, the CO formation selectivity does not drop upon excess liquid injection in the cathode compartment, but it even increases for Sustainion (Fig. 3b, from 70 to 85 % FE) and PiperION (Fig. 3c, from 55 to 85 % FE) membrane containing cells.

For comparison, we performed similar activation experiments with pure aqueous solutions (i.e., no isopropanol content). A less pronounced activation was seen, albeit with very large variance (Supplementary Figure 11). This is rooted in the hydrophobicity of the GDE, which only allows the aqueous activation solution to reach the catalyst surface if numerous imperfections (e.g., cracks) are present in the structure. Again, tailoring the solvent mixture is crucial for wetting the deeper regions of the GDE without damaging its structure.

Another important aspect is the necessity of the electric field during the activation: if we inject the activation solution without polarization and start the electrolysis subsequently, there is only a minor current increase. This indicates that beyond simple physical interactions, electro-sorption of the cations on the negatively polarized Ag catalyst particles might be a major contributor.

Mechanism of the activation process

Performing the cathode activation on the water-fed cell increases the current to a level almost identical with those measured with 0.1 M KOH anolyte (Fig. 4a). While the high frequency impedance of the water-fed cell remains unchanged, the gross charge transfer resistance decreases to a similar value detected with alkaline anolyte (Fig. 4b, Supplementary Figure 12, with further discussion in Supplementary Note 5). As the same anode catalyst (IrO_x) is routinely employed in PEM water electrolyzers at high current densities (1–3 A cm⁻²) ⁴⁶, it is reasonable to assume that this difference is mostly related to the cathode catalyst. To safely exclude the contribution of the anode process (Supplementary Note 6), we have recorded polarization curves of IrO_x/PiperION ionomer layers in different electrolyte solutions (i.e., NaOH, KOH, CsOH, Supplementary Figure 13a). The polarization curves overlap, showing that the electroactivity of IrO_x is not affected by these cations significantly. This confirms that the eventual penetration of small amounts of the activation solution to the anode side cannot be responsible for the enhanced performance.

Fig. 4.

Mechanism and reversibility of cathode activation. (a) Chronoamperometric curves and (b) EIS spectra recorded before and after activating the cathode GDE with 10 cm³ 0.5 M KOH solution in 1:3 isopropanol/water mixture (Sustainion membrane. ΔU = 3.1 V, T_cathode = 60 °C, 12.5 cm³ cm⁻² min⁻¹ CO₂ feed rate,). (c) Chronoamperometric curve and CO formation partial current density measured under identical conditions as in (a). The cathode GDE was activated at the marked times with 10 cm³ 0.5 M KOH solution in 1:3 isopropanol/water mixture, while it was rinsed with 10 cm³ 1:3 isopropanol/water mixture to de-activate it. (d) Time-resolved current density and product stream composition change during, and immediately after activating the electrolyzer cell (with 3 cm³ 1 M CsOH solution in 1:3 isopropanol/water mixture, ΔU = 3.2 V, T_cathode = 60 °C, 12.5 cm³ cm⁻² min⁻¹ CO₂, PiperION membrane). In the upper panel, the pressure in the CO₂ inlet pipe during activation is indicated. Such experiments were repeated on separate cell assemblies independently at least 3 times, with similar results.

The activation is fully reversible: rinsing the cathode GDE with an isopropanol/water mixture (without any dissolved electrolyte) in the absence of polarization restores the current to its initial low value within 1 min, which can be increased again by repeating the activation (Fig. 4c). This repeatable instantaneous activation-deactivation further suggests that the performance enhancement is mostly related to the change in the cathode reaction kinetics. If the electrolyte penetration to the anode side (upon activation) would be a major contributor, a single rinsing of the cathode side could not have such a rapid effect. Additional control experiments were carried out without anolyte recirculation, to avoid the possible accumulation of cations or change in the pH. The activation phenomenon occurred in this case as well (Supplementary Figure 13b). All these observations together imply that the activation mostly affects the cathode interface.

To take a closer look on processes occurring during activation, we followed the product-stream composition quasi-real-time via mass spectrometry, while also monitoring the pressure in the CO₂ inlet piping (Fig. 4d). The pressure trace shows that it takes about 50 seconds for the activating solution to leave the cell (indicated by a rapid pressure drop after the initial increase). While the cathode compartment is filled with the activation fluid it is deprived of CO₂, resulting in high H₂ formation rate and decreased CO₂ and CO partial pressures in the product gas stream (lower panel of Fig. 4d). When the liquid leaves the cell, the CO partial pressure (hence j_CO) immediately increases, meanwhile the H₂ concentration decreases. Both stabilize in 2 minutes after the activation, demonstrating that the activation is instantaneous (and therefore related to the cathode GDE and its interface with the AEM). Finally, the activated state does not diminish rapidly, which would be expected if entrapped cations leaving the AEM would be mostly responsible for the activation, as their amount could only be finite in the thin membranes.

Factors determining the efficiency of the cathode activation

The activation process influences the pH and ionic strength at the cathode (at least temporarily), and the infused ions can also adsorb on the catalyst surface. To separate the effects of these parameters, we performed two sets of experiments: one in which different alkali metal hydroxides (i.e., different cations) were employed (Fig. 5a and Supplementary Figure 14a,c), and another in which solutions with constant K⁺ concentration, but with different anions were injected (Fig. 5b and Supplementary Figure 14b,d). The activation becomes more pronounced in the NaOH<<KOH<CsOH order, whether considering the j_CO values (Fig. 5a), the onset voltage, or the slope of the LSV curves (Supplementary Figure 14a) or the decrease in the arc diameter in the impedance spectra (Supplementary Figure 14c). As the pH and ionic strength of these solutions are identical, these measurements directly prove the promoting effect of K⁺ and Cs⁺ ions, and the less considerable effect of Na⁺ ions in the electrochemical CO₂R, in accordance with theoretical^37,41,42 and experimental results.^{38–40,43,44} The same trend was found when operating the same cell with NaOH, KOH and CsOH anolytes (Supplementary Figure 15).

Fig. 5.

Deconvolution of the complex effect of the activating electrolyte. Partial current densities for CO and H₂ production during constant voltage electrolysis with water anolyte, after cathode activation using 10 cm³ solution (in 1:3 isopropanol/water mixture) of (a) different alkali metal hydroxides (c = 0.5 M) (b) different potassium salts (c(K⁺) = 0.5 M). The cell was operated at ΔU = 3.1 V, T_cathode = 60 °C with 12.5 cm³ cm⁻² min⁻¹ CO₂ feed rate, using a Sustainion membrane in the cell. The plotted values are the mean from 3 gas composition measurements (GC analysis), together with the calculated standard deviations. All experiments were repeated independently on separate cell assemblies 3 times, with similar results.

To investigate the effect of the local pH change caused by the activation, solutions of different potassium salts were infused. Although some differences can be seen, they are less striking compared to the case of different cations (Fig. 5b and Supplementary Figure 14b,d). The current density increase followed the KNO₃ <K₂CO₃ <KOH order, suggesting that the effect of the local pH during the activation cannot be fully ruled out. The same trend was found for the double layer capacitance of the catalyst/ionomer layer, studied in the solution of these salts (Supplementary Figure 16). These differences might stem from the pH-dependent structure of the ionomer (it is a polyelectrolyte) or from the differences in the double-layer structure (and therefore the specific capacitance) in the different solutions. In any case, this can affect the cation adsorption process during the transient period of infusion, and in turn the amount of adsorbed alkali metal cations.

The activation efficiency scales with the concentration (up to 0.5 M, Supplementary Figure 17) and volume (up to 2 cm³, Supplementary Figure 18) of the electrolyte solution. Finally, the membrane and gross interfacial charge transfer resistance continuously decreases with the increasing temperature (Supplementary Figure 19), indicating the room for further performance enhancement through careful process engineering.

Long-term operation

Long-term electrolysis with pure water feed at the anode was performed with a PiperION membrane containing cell for 224 hours (Fig. 6a and b), meanwhile the cathode was activated with 5 cm³ 1.0 M CsOH solution after every 12 hours. The initially (before activation) measured j_CO ≈ 120 mA cm⁻² increased to j_CO ≈ 350 mA cm⁻² after the first activation. This increased further when the activation was repeated after 24 and 36 hours, and subsequently it remained stable at j_CO = 420±50 mA cm⁻² (with ~90% FE_CO) for over 200 hours (eight days) of continuous operation. The regular spikes on the chronoamperometric curve (every 12 hours) belong to the activation processes (Fig. 6a), while the initial ones (in the first 25 hours) represent irregular water release from the cell. Clearly, there is an initial period when the water management of the cell stabilizes. The FE_CO first decreases during the activation steps, as the large current transient is associated with the increased H₂ evolution (see also Fig. 4d). FE_CO is then restored (or even increased), while j_CO increases compared to its value before activation. The single pass conversion was also calculated and the achieved 23% (see calculations in Supplementary Note 7) is about 11 times larger compared to the state-of-the-art water-based CO₂ electrolyzer report.²⁷ A similar experiment was performed with Sustainion X37-50 membrane, where the electrolyzer was operated for 10 hours. After activation, the j_CO increased from 120 mA cm⁻² to 220 mA cm⁻², which has not changed significantly in the first five hours, followed by a slight decrease to j_CO ≈ 200 mA cm⁻² (Supplementary Figure 20).

Fig. 6.

Long-term operation of a CO₂ electrolyzer with water anolyte and periodic activation. (a) Total current density and (b) CO and H₂ partial current densities during constant voltage electrolysis, using a PTFE-reinforced 15 μm thick PiperION TP-85 membrane separated cell (ΔU = 3.2 V, T = 60 °C water anolyte, 12.5 cm³ cm⁻² min⁻¹ CO₂ feed rate). The cathode was activated with 5 cm³ 1 M CsOH solution in 1:3 isopropanol/water mixture after every 12 hours of the electrolysis. Long term experiments (over 100 h) were repeated on separate cell assemblies independently 5 times, with similar results.

Importantly, no physical precipitate formation was observed in the cells, which were activated repeatedly (see an example for a cell operated for over 100 hours in Supplementary Note 8, Supplementary Figure 21). The lack of precipitate formation is attributed to the inherent nature of the activation process, where the applied solvent mixture can dissolve and remove the previously formed precipitate crystals. Therefore, while small amount of alkali cations is present after the activation, no accumulation occurs.

The presence of cations in the cathode GDE can also be ensured by other methods. For example, controlled crossover from the anolyte can provide the proper amount of alkali cations to the cathode, especially if combined with periodic regeneration of the cathode with the proper solvent mixture (Supplementary Figure 6). As a further example, the Ag catalyst layer can be deposited from alkali metal ion containing suspensions. This way, the promoter ions are inherently present on the catalyst surface and within the ionomer layer. The electrochemical performance and impedance spectrum of the electrolyzer cell assembled with this catalyst layer, and operated with water anolyte, are similar to those of the activated cells, presented before. The performance, however, decreased with time, as the K⁺ ions gradually leached out from the catalyst layer (Supplementary Figure 22.).

Operating larger cells and cell stacks with water anolyte

The scalability of the cathode activation approach was demonstrated on a cell having 100 cm² active surface area, and on a 3-layer electrolyzer stack formed thereof (Fig. 7, Supplementary Figure 23). The cathode activation boosted the j_CO in all experiments, and j_CO also increased gradually with the applied cell voltage (Fig.7a-c). Importantly, j_CO = 450 mA cm⁻² was measured at ΔU = 3.2 V (Fig. 7c), which rivals the results recorded on our smaller cell (Fig. 3c, Fig 6). This value further increased to j_CO = 650 mA cm⁻² at ΔU = 3.5 V (Fig. 7b), which is close to the best value measured in the same electrolyzer cell using 0.1 M CsOH anolyte.¹⁰ Importantly, the rate of the current decay after the activation decreased with the increasing cell voltage (i.e., the effect of activation lasts longer at higher cell voltages (Fig. 7c)). This again confirms that the activation occurs via the electrosorption of the cations at the negatively polarized cathode.

Fig. 7.

Cathode activation experiments in larger electrolyzer cells and stack. Chronoamperometric measurements on an A = 100 cm² single layer electrolyzer cell (T = 60 °C, water anolyte, 12.5 cm³ cm⁻² min⁻¹ CO₂ feed) at (a) ΔU = 3.3 V and (b) ΔU = 3.5 V. (c) CO partial current densities 10 minutes after performing the cathode activation (20 cm³ 1 M CsOH solution in 1:3 isopropanol/water mixture) and the ratio of CO formation partial current densities 90 and 10 minutes after performing the cathode activation, recorded at different cell voltages. (d) Chronoamperometric measurement on a A = 100 cm² 3-layer electrolyzer cell stack (T = 60 °C, water anolyte, 12.5 cm³ cm⁻² min⁻¹ CO₂ feed, cathode activation with 60 cm³ 1 M CsOH solution in 1:3 isopropanol/water mixture) at ΔU = 9.9 V. Such experiments were repeated on separate cell assemblies independently 3 times, with similar results.

The activation was also performed in a 3-cell stack at ΔU = 9.9 V stack voltage (Fig. 7d), translating to ΔU = 3.3 V cell voltage for comparability with the single-cell. The activation increased j_CO to over 500 mA cm⁻², a current density even slightly higher than what was measured in the single layer cell (Fig 7a). One key challenge during scale-up is the increased resistance of the cell (stack) against the flow of gas and the activation fluid during infusion. Therefore, we have monitored the pressure build-up on the cell (cell-stack). It was indeed higher than in the case of the small cell (0.8 bar for the small cell, ~2 bar for the large cell, and ~3 bar for the cell stack, see Supplementary Figure 22), but not detrimental at all. Finally, the cell design was not optimized for this activation process, leaving a lot of room for improvement. In this regard, scale-up and optimization of such electrolyzer stacks are ongoing (see Supplementary Figure 23 for a 1000 cm²/cell size stack).

Conclusions

We uncovered some microscopic/mechanistic reasons behind the performance fading in zero-gap CO₂ electrolyzer cells. SEM-EDX and micro-CT analysis confirmed the formation of alkali metal bicarbonate and mixed carbonate-bicarbonate precipitate plaques within the cathode GDE of cells operating with alkaline anolyte, hindering CO₂ gas from reaching the catalyst surface. In an apparent contradiction to this, we identified the presence of alkali metal ions on the cathode catalyst surface to be a major contributor for high current density CO₂ reduction. To overcome this ambiguity, we developed an operando activation method, where the cathode of pure water anolyte fed CO₂ electrolyzer cell is infused periodically with different alkali-cation containing solutions with proper wetting properties. We have shown that the activation predominantly affects the cathode chemistry and has negligible effect on the anode process and the AEM properties.

The presented proof-of-concept demonstrates scalable water-fed electrolyzer cells operating at CO₂ reduction rates matching that of using alkaline electrolytes (j_CO of 420 ± 50 mA cm⁻²) over prolonged times (over 200 h). The activation is repeatable, implying that the observed slow current decrease over time is not caused by any degradation mechanism, but by the continuous desorption and leaching of the alkali cations from the cathode catalyst layer during operation. The rate of performance decay depends on the cathode GDE composition (e.g., ionomer compositionand content, thickness, and porosity) and on operational parameters such as humidification of the CO₂ stream, water crossover through the membrane, cell voltage, temperature, current density etc. Optimizing these parameters is a major engineering task, which in turn might bring a big reward: the ability to operate zero-gap CO₂ electrolyzer cells with pure water anolyte at high current density, with negligible maintenance requirements. Such studies are in progress in our laboratory, together with scaling-up the technology further.

Methods

Materials

All chemicals were purchased from commercial suppliers (Sigma-Aldrich, VWR International), and were of at least analytical grade and were used without further purification. MilliQ grade (ρ = 18.2 MΩ cm) ultrapure deionized water was used to prepare all the solutions.

Electrode preparation

Ag nanoparticles (d_avg < 100 nm, Sigma-Aldrich) were dispersed in a 1:1 isopropanol/water mixture at a concentration of 25 mg cm⁻³, together with 15 wt% ionomer, matching the used membrane (Sustainion XA-9, Fumion FAA-3, PiperION I-46). As for the anode catalyst, IrO_x nanoparticles (Fuel-Cell Store) were dispersed in an identical solvent mixture and ionomer concentration, but with a concentration of 20 mg cm⁻³. The IrO_x dispersion was homogenized in a regular ultrasonic bath for 20 minutes (keeping the bath temperature below 35 °C), while a high-power immersion sonotrode was used to disperse the Ag nanoparticles.

The Ag dispersion was spray-coated on Sigracet 39BC carbon gas diffusion layers (GDLs), preheated on a hotplate at 100 °C, using a hand-held airbrush and compressed air carrier gas (~100 cm³ min⁻¹). The anode catalyst was spray-coated similarly, on a porous titanium frit. The cathode catalyst loading was 1.0 ± 0.1 mg cm⁻² for the measurements with the PiperION membrane, and it was 3.0 ± 0.1 mg cm⁻² for the other systems, while the anode catalyst amount was 1.0 ± 0.1 mg cm⁻².

Commercially available AEMs were used to separate the anode and the cathode chambers (Class T X37-50 Sustainion (50 μm thickness) from Dioxide Materials, Fumasep FAB-PK-130 (130 μm thickness) from FUMATECH BWT GmbH, PTFE-reinforced 15 μm thick PiperION TP-85 from W7energy LLC). The membranes were ion-exchanged before use for 24 hours in the respective 1 M alkaline solution (NaOH/KOH/CsOH) which was exchanged to a fresh solution after the first 5 hours.

Zero-gap electrolyzer cells

All experiments presented in the paper were performed in zero-gap electrolyzer cells. This consist of (from bottom to top in Supplementary Figure 1): an anode current collector on which an anolyte flow pattern was formed, a porous Ti frit with catalyst layer on its side in direct contact with the AEM, held in place by a spacer element, an AEM (different membranes were used in our measurements, as described in the manuscript), a GDE with its catalyst layer facing the membrane, a spacer element to set the compression ratio of the GDE and a cathode current collector with a concentric flow-pattern with a central inlet, and an outlet on the perimeter of the flow-pattern. The electrolyzer elements were assembled directly on top of each other and were held together by 6 bolt screws (3 Nm torque was applied). The active surface is circular, of 3.2 cm diameter, resulting a geometrical surface of A = 8 cm², which value was used to normalize the current values reported here. A larger (A = 100 cm²) cell was also used to demonstrate the scalability of our approach. The design of this cell is very similar to its smaller counterpart, but instead being spherical, it is rectangular in shape. A stack containing three cells connected in series electrically and parallel with regards of the gas distribution.^10,24

Test framework

The test framework (Fig. 2) was based on our previous experimental setup, reported earlier.²⁴ Briefly, the CO₂ feed rate was controlled with a Bronkhorst F-201C type mass-flow controller, while an Agilent ADM flow meter was used to measure the flow rate of the gas outlet. The accuracy of this measurement was confirmed by repeating these measurements with the classical bubble-soap method periodically. The gas flow rate was normalized with the surface area of the cell (hence the cm³ cm⁻² min⁻¹ unit). The CO₂ gas was passed through a temperature-controlled humidifier before entering the cell. An extra loop was inserted before the electrolyzer to be able to inject electrolyte solutions in the gas stream. This activation loop can be bypassed by turning two three-way valves. After filling the loop with the respective electrolyte solution, it can be injected in the gas stream by turning these two valves. The pressure of the CO₂ in the electrolyzer was measured using a digital pressure gauge. The anolyte was circulated in the anode compartment using a peristaltic pump. The electrochemical measurements were performed using a Biologic VMP-300 type instrument, equipped with EIS- and high current (up to 10 A) booster options. All the experiments presented here (including the EIS measurements) were performed using the potentiostat with the booster connected. For the measurements with the larger electrolyzer cell and stack, a programmable power supply (TDK-Lambda GEN-15-220) was used.

The measurements were conducted in a two-electrode setup, and the total cell voltage is given as the voltage difference between the anode and the cathode (hence the positive values). The absolute value of the current/current density is shown (positive values) in all figures. No IR-correction was applied on the voltage values presented throughout the manuscript. EIS spectra have been taken at constant cell voltage with added 10 mV rms perturbation in the 100 kHz down to 0.1 Hz frequency range, with 10 points per frequency decade. The self-consistency of the EIS data was checked by performing Kramers–Kronig tests. Impedance spectra are presented in the form of complex-plane plots (“Nyquist plots”) where the points of f = 10^k Hz frequencies (where k is an integer, -1 ≤ k ≤ 5) are marked. To help in distinguishing spectra they are shifted along the ordinate.

The composition of the cathode product stream was analyzed using a Shimadzu GC-2010 Plus type instrument, equipped with a barrier discharge ionization (BID) detector. A Restek ShinCarbon ST column was employed for the separation with 6.0 grade Helium carrier gas. An automatized 6-port valve was used to take samples in regular time intervals. In certain experiments the product stream composition was simultaneously monitored using an m/z analyzer (SRS UGA200), equipped with an atmospheric sampling capillary. This technique allowed the quasi realtime determination of the product composition, hence providing time-resolved information on the processes occurring during and immediately after cathode activation. The anode gas composition was analyzed with a BGA-244 type Binary Gas Analyzer (Stanford Research Systems), to monitor the CO₂/O₂ ratio.

Membrane impedance measurements

A Zahner Im6e type instrument was used to characterize the membrane impedance in a four-electrodes cell (Supplementary Figure 8a). The EIS spectra were recorded in galvanostatic mode (I = 0), with 1 mA perturbation amplitude, from 100 kHz to 1 Hz, recording twice 8 points/decade.

Contact angle measurements

An EasyDrop (Krüss GmbH) type instrument was used to measure the wetting properties (i.e., contact angles) of different solvent mixtures on the microporous side of the used Sigracet 39BC GDL. A drop of the solvent mixture was formed on the plate with the use of a syringe. Using the CCD camera of the goniometer, the drop contour of the captured photos was analyzed.

X-ray micro-computed tomography (micro-CT) analysis

A Bruker SkyScan 2211 Multiscale X-ray Nanotomograph (Skyscan, Bruker, Belgium) instrument was used to record micro-CT images. The 3-dimensional structure of the samples was scanned using 11 Mp CCD detector by applying a source voltage of 70 kV and current of 400 μA (in microfocus mode, with a resolution of 1 μm/pixel). NRecon reconstruction software (Skyscan, Bruker, Belgium) was used to reconstruct the projected images, while the CTAn and CTVox software (Skyscan, Bruker, Belgium) were applied to carry out the image segmentation and visualizing the 3D-rendered objects, respectively.

SEM-EDX analysis

A Hitachi S-4700 scanning electron microscope (SEM) coupled with a Röntec EDX detector was used to take images of the GDEs. The microscope was operated at 10 kV acceleration voltage.

XRD measurements

X-ray diffraction (XRD) patterns were obtained using a Bruker D8 ADVANCE X-ray diffractometer, applying Cu K_α (λ = 1.5418 Å) radiation. Rietveld refinement was carried out to quantify the phase composition of the obtained materials.

Ion-chromatography measurements

To gain direct proof on the K⁺ and Cs⁺ crossover through the AEM, electrolysis experiments were performed with 0.01 M KOH and 0.01 M CsOH anolytes at 60 °C, and the humid gas leaving the cell was collected in a mechanical liquid/gas separator, initially filled with 10 cm³ pure water. The K⁺ or Cs⁺ concentration of the fluid in this water separator was then monitored during the electrolysis by ion chromatography (IC, Fig. 1b). All IC measurements were performed using a Shimadzu ion-chromatographic system, consisting of a high pressure chromatographic module (LC-20AD SP), an eluent degas module (DGU-20A5R), an autosampler (SIL-20AC), a conductivity detector for non-suppressed conductivity measurements (CDD-10A VP), a column oven to maintain constant 40 °C temperature (CTO-20AC) and a Shodex IC YS-50 column. 4 mM methanesulfonic acid was used as eluent with a flow rate of 1.0 cm³ min⁻¹, and the sample volume was 50 μl.

Data availability

The authors declare that all data supporting the findings of this study are available within the paper and Supplementary Information files. Source data are provided with this paper.