Dynamic Relocation of Copper Catalysts in Gas Diffusion Electrodes during CO₂ Electroreduction

Abstract

Developing technologies that convert CO₂ into valuable carbon products using renewable energy is of growing significance. Copper (Cu) is a unique electrocatalyst capable of reducing CO₂ to value-added multicarbon (C₂₊) compounds. While recent in situ studies have elucidated the dynamic evolution of Cu catalysts during electrochemical CO₂ reduction reactions (CO₂RR), the relationship between catalyst behavior in gas diffusion electrodes (GDEs) and C₂₊ product selectivity at industrially relevant current densities remains insufficiently understood. In this study, we examined the correlation between the structure, chemical state, and C₂₊ selectivity of Cu catalysts in Cu-GDEs during CO₂RR operation at current densities exceeding 200 mA/cm². Ex situ and in situ scanning X-ray fluorescence microscopy revealed significant relocation of Cu within the GDE after CO₂RR. In situ X-ray absorption spectroscopy identified the presence of Cu¹⁺ species during operation, indicating that Cu relocation proceeds via a dissolution-redeposition. The dissolution-redeposition behavior was found to be pH-dependent and more pronounced at high pH. Online gas chromatography demonstrated that the decrease in C₂₊ selectivity over time was primarily due to flooding, overshadowing the impact of Cu relocation on C₂₊ selectivity. These findings provide important insights for designing stable and highly selective Cu-based GDEs for practical CO₂ electrolyzers.

1. Introduction

The conversion of CO₂ into valuable chemicals and fuels through electrochemical CO₂ reduction reactions (CO₂RR) using renewable energy sources offers a promising path toward a sustainable, carbon-neutral society. − Among the various CO₂ reduction products, multicarbon (C₂₊) compounds such as ethylene and ethanol have attracted considerable attention due to their higher energy density and economic value. , Copper (Cu) has been extensively investigated as an electrocatalyst for CO₂RR due to its unique ability to facilitate the formation of C₂₊ products. ,,, Nevertheless, the efficiency, selectivity, and reaction rates of Cu-based catalysts remain insufficient for practical implementation. To achieve the economic feasibility required for industrial applications, CO₂ electrolyzers must operate at current densities exceeding 200 mA/cm². , This has been made possible using gas diffusion electrodes (GDEs), in which catalysts are deposited on gas diffusion layers (GDLs). GDEs overcome the mass transport limitation of CO₂ in aqueous solutions by supplying CO₂ directly to the catalyst in gaseous form, thereby enhancing reaction rates at high current densities. − For developing commercially relevant CO₂ electrolyzers using GDEs, a comprehensive understanding of the catalyst structure and the local reaction environment under practical operating conditions is essential. In this context, in situ characterization techniques are expected to be critical in elucidating catalyst behavior and guiding rational electrode design.

Recent in situ characterization techniques, such as transmission electron microscopy (TEM), , scanning probe microscopy (SPM), − X-ray absorption spectroscopy (XAS), − infrared spectroscopy (IR), , Raman spectroscopy, , and their complementary analyses, − have shown that Cu catalysts undergo significant changes in chemical state, structure, and interfacial properties during CO₂RR. However, most of these in situ studies have been conducted using batch cell configurations rather than GDE-type cells. In batch cells, the low solubility of CO₂ in aqueous electrolytes (typically CO₂-saturated KHCO₃ solutions) severely limits the achievable CO₂ reduction current density, often restricting it to below a few tens of mA/cm². This constraint hinders the investigation of catalytic behavior under industrially relevant conditions. In contrast, practical CO₂ electrolyzers using GDEs, such as flow cells and membrane electrode assembly (MEA) cells, operate at much higher current densities. Despite their importance, the structural evolution and chemical dynamics of Cu catalysts in GDEs under such conditions remain poorly understood. This lack of knowledge makes it difficult to fully elucidate the formation mechanisms of C₂₊ products and hinders the rational design of high-performance CO₂RR systems.

This study aims to elucidate the structure and chemical state of Cu catalysts within Cu-GDEs during CO₂RR at practical current densities, as well as the influence of electrolyte penetration, which can lead to flooding. First, the Faradaic efficiency (FE) of each CO₂RR product is quantified in a neutral electrolyte (1 M KCl as catholyte) using a flow cell system with Cu-GDE as cathode. The spatial distributions of Cu catalysts and K (as an indicator of electrolyte penetration) in the Cu-GDE cross-section before and after CO₂RR are systematically analyzed by ex situ scanning X-ray fluorescence microscopy (SXFM) and X-ray computed tomography (CT). Furthermore, the temporal changes in the chemical state of the Cu catalysts during electrolysis at practical current densities are studied by in situ XAS. Subsequently, the time-dependent variations in the FE of each gaseous product were investigated under continuous constant current operation in neutral (1 M KCl) and alkaline (1 M KOH) catholytes. Finally, the implications of Cu relocation on C₂₊ product selectivity and flooding are discussed based on in situ SXFM observations of Cu structural changes during CO₂RR.

2. Results and Discussion

2.1. Morphological Changes of Cu-GDEs during Electrolysis in Neutral Electrolyte under CO₂ or Ar Supply

Cu­(x)-GDEs (thickness of Cu: x nm) were prepared by magnetron sputtering of Cu onto a commercial carbon-based gas diffusion layer (GDL) with a microporous layer (MPL) (Figure S1). The typical thicknesses of the MPL and GDL were approximately 70 and 100 μm, respectively (Figure S2). CO₂RR experiments were conducted in a three-chamber flow cell system employing a Cu(300)-GDE as the working electrode, platinum (Pt) mesh as the counter electrode, and Ag/AgCl as the reference electrode (Figure S3). The FEs of CO₂RR products were determined by continuous operation of chronopotentiometry (CP) for 20 min in a neutral electrolyte (1 M KCl) serving as the catholyte. Figure a presents the CO₂RR product distributions obtained for varying total current densities (J _total) in the Cu(300)-GDE system (see Supporting Note S1 for details on FE calculations). The FE for C₂₊ products (FE_C2+), comprising C₂H₄, C₂H₅OH, CH₃COOH, and C₃H₇OH, was found to be approximately 80% across a broad range of current densities (−400 to −800 mA/cm²). The maximum FE_C2+ reached 82% at −400 mA/cm², with FE_C2H4 = 41% and FE_C2H5OH = 36%. Given that the partial current density of C₂₊ (J _C2+) exceeds −300 mA/cm² (Figure S4), the catalytic behavior of CO₂RR under practical current densities can be effectively analyzed. However, upon increasing the current density to −1000 mA/cm², a decline in FE_C2+ and a corresponding increase in FE_H2 was observed. The reduction in FE_C2+ at elevated current densities is attributable to flooding, as discussed in subsequent sections. Figure b illustrates ex situ scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDX) images of Cu(300)-GDE cross sections before (pristine) and after CO₂RR. The pristine Cu(300)-GDE exhibited a uniformly deposited Cu catalyst layer (CL) on the MPL surface. In contrast, following CP at −400 mA/cm² for 20 min under CO₂ supply, Cu catalysts were observed not only near the MPL surface but also within the MPL. Notably, after CP at −800 mA/cm² for 20 min under CO₂ supply, Cu was no longer observed near the MPL surface but instead diffused further into the MPL. These SEM-EDX results indicate that the Cu-CL morphology undergoes significant structural transformations during CO₂RR.

1.

(a) Current density dependence of FEs of CO₂RR products with Cu(300)-GDE as the cathode in neutral electrolyte (1 M KCl as catholyte). The operation time for each CP was fixed at 20 min. (b) SEM-EDX images (×5000) of Cu(300)-GDE cross-sectional samples: as-prepared (pristine), after CP at −400 mA/cm² for 20 min and after CP at −800 mA/cm² for 20 min. SE image: secondary electron image.

Figure a presents ex situ SXFM images of Cu–Kα (Cu map) and K–Kα (K map) for the same samples shown in Figure b. See Figure S5 for details of the ex situ SXFM measurement system. It should be noted that the Cu distribution width observed in the SXFM image of the pristine sample (Figure a) appears broader than in the corresponding SEM-EDX image (Figure b). This discrepancy arises due to differences in detection depth: X-ray fluorescence analysis has a significantly greater detection depth compared to SEM (Figure S6c). Additionally, the GDE surface exhibits irregularities on the scale of tens of microns (see X-ray CT images in Figure S6b). Consequently, a Cu film with uniform submicron thickness may appear wider in SXFM due to the surface topography of the GDE, as illustrated in Figure S6d. Cu maps of postelectrolysis samples (CP at −400 and −800 mA/cm² for 20 min under CO₂ supply) show that Cu was distributed within the MPL, whereas the pristine sample remains Cu on the MPL surface (Figure a). Furthermore, in both postelectrolysis samples, K distribution in the K map closely resembles the Cu distribution in the Cu map (Figure a). Intensity profiles of Cu and K along the z-axis (Figure c) indicate that the penetration depths of both elements into the MPL increase with increasing charge passed. The splitting observed in Cu and K intensity profiles after CP at −800 mA/cm² (denoted by red arrows in Figure c) suggests that Cu is concentrated at the apical electrolyte region, implying that CO₂RR reaction sites dynamically migrate toward the electrolyte interface along with dissolved Cu. Figure b displays ex situ SXFM images of Cu(300)-GDE subjected to CP at −400 mA/cm² for 20 min under argon (Ar) supply. Under Ar supply, Cu remained localized on the MPL surface, whereas K penetrated deeper into the MPL. Comparative analysis of Cu and K intensity profiles for Cu(300)-GDE operated under CO₂ and Ar supply is shown in Figure d. Under Ar supply, Cu does not relocate within the MPL, while under CO₂ supply, Cu relocation into the MPL occurs. Ex situ spectral X-CT further confirms that Cu relocation occurs exclusively under CO₂ supply (Supporting Note S2 and Figure S7). The deeper penetration of K into the MPL observed under Ar supply (Figure d) indicates that electrolyte flooding progresses more rapidly under Ar supply than under CO₂ supply despite the identical electrolysis charges. The relatively negative potential observed under Ar supply at the same current (Figure b) suggests that electrowetting effects of the carbon MPL are more pronounced under Ar supply, consistent with previous findings by Yang et al. Additionally, the dynamic Cu relocation observed in the Cu(300)-GDE cross-section during electrolysis under CO₂ supply in a neutral electrolyte was also confirmed by in situ SXFM (Supporting Note S3 and Figures S8–S12).

2.

Ex situ SXFM images of Cu(300)-GDE cross-section. (a) Cu and K maps of pristine, after CP at −400 mA/cm² for 20 min and after CP at −800 mA/cm² for 20 min under CO₂ supply in neutral catholyte (1 M KCl). (b) Cu and K maps of after CP at −400 mA/cm² for 20 min under Ar supply in neutral catholyte (1 M KCl). (c) Intensity profiles of Cu and K along z-axis averaged on x-axis of (a). (d) Intensity profiles of Cu and K along z-axis averaged on x-axis of (b). The results of CP at −400 mA/cm² under CO₂ supply are also shown for comparison.

3.

(a) In situ normalized Cu–K edge XANES first derivative spectra of the Cu(70)-GDE during CP at −400 mA/cm² and OCP under CO₂ or Ar supply in neutral electrolyte (1 M KCl), together with the standard samples. (b) Plot of [integrated intensity of Cu¹⁺]/[integrated intensity of Cu⁰] calculated from multipeak-fitting of (a) and corresponding CP-OCP curves. The horizonal axis is the time elapsed from the start of gas supply to the cell.

2.2. Chemical States of Cu during Electrolysis in Neutral Electrolyte under CO₂ or Ar Supply

The above-mentioned ex situ and in situ SXFM results indicate that Cu migration into the MPL during CO₂RR is closely linked to the chemical state of Cu, including its dissolved ionic form (Cu⁺). To further investigate the chemical state of Cu under electrolysis conditions, in situ Cu–K edge XAS measurements were conducted using the flow cell modified to allow X-ray measurements (Figure S13). It should be noted that the Cu–K fluorescence X-ray signal represents the average XAS spectrum of all Cu present within the X-ray irradiation area. Due to the predominant presence of Cu⁰ in Cu(300)-GDE, where Cu⁺ species were buried and could not be directly quantified, in situ XAS measurements were instead conducted on a Cu(70)-GDE system, which features a thinner CL thickness to facilitate the detection of Cu⁺ embedded within the bulk Cu⁰ structure. Figure a presents the in situ normalized first derivative spectra of Cu–K edge X-ray absorption near edge structure (XANES) for Cu(70)-GDE during constant current electrolysis (CP at −400 mA/cm², approximately −1.1 V_RHE) and at open circuit potential (OCP, approximately +0.4 V_RHE) under CO₂ and Ar supply, along with reference standard samples. While the overall spectral profiles of the first derivative XANES remain similar, a distinct variation in peak intensity at the Cu¹⁺ energy position is observed between CO₂ and Ar supply conditions (Figure a). A quantitative methodology for determining the ratio of Cu⁰ and Cu¹⁺ from first derivative XANES spectra, as previously reported, was employed for this analysis. The complete XANES data set, including the multiple peak fitting of the first derivative XANES spectra based on characteristic energy positions of Cu⁰ and Cu¹⁺, is provided in Figures S14 and S15. The Cu¹⁺/Cu⁰ ratio, defined as the integrated intensity of Cu¹⁺ relative to Cu⁰, is plotted in Figure b. During CP at −400 mA/cm² under CO₂ supply conditions, the Cu¹⁺/Cu⁰ ratio remained approximately 0.2 (Figure b, red), while it was nearly 0 under Ar supply (Figure b, blue). This suggests that a persistent concentration of Cu¹⁺ species was maintained under CO₂ supply despite cathodic current application, whereas under Ar supply, Cu was predominantly present as Cu⁰ with no detectable Cu¹⁺ species. Furthermore, under the experimental conditions of in situ XAS, Cu²⁺ was not observed during electrolysis, regardless of CO₂ or Ar supply.

The horizontal axis in Figure b represents the time elapsed from the start of CO₂ or Ar gas supply to the cell. Notably, approximately 10 min transpired before the current was applied. In the case of CO₂ supply, the pH of the KCl catholyte (initial pH ∼ 7) decreased to 5.5 within this period, indicating that the initial XANES measurement at OCP (#0) was conducted under mildly acidic conditions. According to the Pourbaix diagram, Cu dissolution is influenced by both acidic and highly alkaline environments: Cu near OCP (0 V_SHE) exists as Cu⁰(s) or Cu₂O(s) at neutral pH, as Cu⁺ at lower pH, and as Cu­(OH)₂ ^– at higher pH. − The substantial increase in the Cu¹⁺/Cu⁰ ratio at OCP under Ar supply (Figure b, blue) suggests that Cu underwent oxidation to Cu₂O. In contrast, under CO₂ supply, the Cu¹⁺/Cu⁰ ratio remained stable regardless of cathodic current application, indicating the continuous presence of Cu¹⁺ species throughout electrolysis (Figure b, red). The existence of Cu¹⁺ under CO₂RR conditions has also been reported in prior studies on Cu-based catalysts. − However, the presence of Cu¹⁺ in CO₂RR conditions contradicts predictions based on the Pourbaix diagram, implying that reaction kinetics and local microenvironmental effects play a significant role in stabilizing thermodynamically unfavorable phases during CO₂RR. Vavra et al. proposed that adsorbed *CO intermediates form Cu-adsorbate complexes, which can persist in solution under CO₂RR operating conditions. We speculate that dissolved Cu¹⁺ interacts with CO₂ to form a stable complex ion, potentially [Cu-*CO]⁺, which maintains a detectable concentration within the MPL throughout CO₂RR. To evaluate whether carbonate ions or CO-derived anions play a role in Cu migration, the Cu distribution within Cu(300)-GDE postelectrolysis in KCl and KHCO₃ catholytes was examined via scanning transmission electron microscopy (STEM)-EDX (Figure S16). Under CO₂ supply, Cu relocation was observed regardless of whether KCl or KHCO₃ was used as the electrolyte (Figure S16a,b). In contrast, under Ar supply, Cu relocation was observed in KHCO₃ but absent in KCl (Figure S16c,d). These results suggest that the presence of carbonate ions is essential for facilitating Cu migration.

2.3. Cu Solubility and CO₂RR Selectivity as a Function of Initial Electrolyte pH

Regardless of the gas type (CO₂ or Ar) supplied to the flow cell depicted in Figure S3, the presence of dissolved Cu species in the catholyte after OCP and CP was confirmed by inductively coupled plasma mass spectroscopy (ICP-MS) (Table S1). The concentration of dissolved Cu in the KCl catholyte (initial pH ∼ 7) after 30 min of OCP corresponded to approximately 15% of the initial Cu(300)-GDE loading, yielding a dissolution rate of ∼ 0.5%/min. This dissolution rate aligns with findings reported by Kok et al., who observed Cu dissolution under OCP conditions using Cu-PTFE GDE cathodes in a neutral catholyte (1 M KHCO₃, pH = 8.3), further corroborating the dependence of Cu dissolution on the initial catholyte pH. Notably, the amount of dissolved Cu in the KOH catholyte (initial pH ∼ 14) after 30 min of OCP exceeded that in KCl catholyte by more than 3-fold (Table S1). The migration of dissolved Cu complexes is enhanced under locally alkaline conditions, where the presence of OH^– ions induces anomalous oxidation-redeposition cycles of Cu. , Since Cu under OCP is in a relatively high oxidation state, the formation of Cu–OH complexes may be promoted at elevated pH levels, potentially increasing dissolution regardless of the supplied gas species (CO₂ or Ar). In contrast, the concentration of dissolved Cu in the catholyte following CP electrolysis was lower than that observed under OCP conditions (Table S1). This decrease in dissolved Cu levels can be attributed to the processes where Cu species transported into the MPL along with the catholyte are redeposited within the MPL during CP, minimizing Cu dissolution into the bulk electrolyte. These results suggest that the pH and composition of the electrolyte influence Cu dissolution and redeposition mechanisms.

To evaluate the dependence of C₂₊ selectivity on the initial electrolyte pH, time-resolved FEs of gaseous products were monitored via online gas chromatography (GC) during continuous CP electrolysis at −400 mA/cm² (Figure ). FE_C2H4 was employed as an indicator of sustained C₂₊ selectivity. In the neutral catholyte (1 M KCl, initial pH ∼ 7), FE_C2H4 was initially high at 39% and remained above 30% for approximately 90 min (Figure a). Subsequently, FE_C2H4 declined, reaching 0% after 160 min. The maximum FE_C2H4 observed (∼ 40%) is consistent with typical values reported for Cu-based GDE, and the trend of decreasing FE_C2H4 coupled with increasing FE_H2 over extended operation is in agreement with previous studies. Electrolysis was halted upon reaching 0% FE_C2H4, after which the cell was disassembled, and the GDE was subjected to solvent-based washing and drying to remove accumulated water and salts (denoted by red arrows of “washing” in Figure ). Following reassembly with fresh electrolyte, CP electrolysis resumed, and FE_C2H4 partially recovered to 28%. A similar trend was observed in the alkaline catholyte (1 M KOH, initial pH ∼ 14): the initial FE_C2H4 was 41%, decreased to 0% after 160 min of continuous CP, and recovered to 40% upon washing (Figure b). The bulk pH of the KCl catholyte measured at 40 min intervals revealed an increase over operating time (Figure a, top). At practical current densities (≥200 mA/cm²), the local pH at the reaction interface of neutral catholyte rises, resembling alkaline catholyte conditions. , Given that CO₂RR in Figure was performed at 400 mA/cm², comparable local pH conditions and analogous product selectivity were achieved in both neutral and alkaline systems. Figure also demonstrates that the onset of FE_H2 increase occurred earlier in KOH (∼ 80 min) than in KCl (∼ 100 min), a phenomenon further clarified in Figure S17. Carbonate formation is promoted wherever dissolved CO₂ and OH^– ions coexist and is further enhanced under increasing alkalinity. Consequently, alkaline conditions expedite salt accumulation within the GDE, which may contribute to operational flooding. These results indicate that the decrease in C₂₊ selectivity over operating time is primarily due to flooding effects, where salt accumulation impedes gas diffusion within the GDE, restricts CO₂ supply, and promotes the hydrogen evolution reaction (HER).

4.

FE curves of gas products (H₂, CO, CH₄, and C₂H₄) as a function of electrolysis time at −400 mA/cm² under CO₂ supply. A Cu(300)-GDE was used as cathode, with (a) neutral catholyte (1 M KCl) and (b) alkaline catholyte (1 M KOH). Washing of the cathode was examined at the indicated red arrow position, where the GDE was removed, solvent-washed, and dried before being reused to reassemble the cell with flesh electrolyte. The bulk catholyte pH values recorded every 40 min are shown at the top.

2.4. Effect of Cu Relocation on C₂₊ Selectivity and CO₂RR Stability

Although significant Cu relocation within the GDE was observed within 20 min of CP in 1 M KCl catholyte (Figures , S7 and S11), FE_C2H4 remained consistently high both before and after 20 min (Figure a). These results indicate that Cu relocation does not directly govern C₂₊ selectivity. To further investigate the effect of Cu relocation on electrochemically active sites, we measured the electrochemically active surface area (ECSA), a proxy for the electrolyte-wetted electrode surface. , An efficient triple-phase boundary (TPB; gas–liquid–solid interface) prevents excessive electrode wetting and results in a relatively low electrochemical double layer capacitance (C _dl). Figure S20 presents the C _dl values of Cu-GDE in 1 M KCl and 1 M KOH before and after 10 min of CP at −400 mA/cm², derived from cyclic voltammetry (CV) measurements in 1 M KCl (Figure S18) and 1M KOH (Figure S19). In both cases, C_dl was initially low but increased after CP, indicating progressive electrode wetting. Notably, washing the GDE reduced C _dl to its initial or lower values (Figure S20c), with most C₂₊ selectivity restored (Figure ). These results suggest that the TPB can be reconstructed by removing accumulated salts, even after Cu relocation. Initially, we anticipated that Cu relocation would have a significant effect on C₂₊ selectivity. However, these findings reveal that the decline in C₂₊ selectivity over time is primarily attributed to flooding rather than Cu relocation. However, this is likely because the dominant effect of flooding masks the influence of Cu relocation on C₂₊ selectivity. Under conditions where flooding is effectively suppressed, Cu relocation may still exert a significant influence on C₂₊ selectivity. In such cases, the formation of highly dispersed Cu nanoparticles via dissolution and redeposition could further enhance C₂₊ selectivity. Nevertheless, if Cu relocation progresses further and Cu catalyst is completely lost from the GDE via dissolution, the electrode would become a simple carbon electrode, resulting in irreversible loss of CO₂RR activity.

Figure schematically illustrates the Cu relocation process within Cu-GDE cathodes during CO₂RR. Regardless of electrolyte pH, Cu dissolves into electrolyte under OCP conditions (Figure a). Cu dissolution is more pronounced in alkaline electrolytes (Table S1); therefore, minimizing time spent at OCP is critical for ensuring catalyst longevity, irrespective of electrolyte pH. Upon cathodic current application, Cu dissolves as Cu complex ions and undergoes dynamic migration within the GDE through cyclic dissolution and redeposition, ultimately resulting in Cu relocation (Figure b). This Cu relocation process is more pronounced in alkaline electrolytes and was confirmed by in situ SXFM analysis of Cu distribution during CP-OCP cycles under CO₂ supply (Figure S21). Indeed, the spatial distribution of Cu maps after CP-OCP cycles reveal more extensive Cu dispersion within the MPL in alkaline electrolytes than in neutral electrolyte (Figure S22). We speculate that the pronounced Cu relocation observed in alkaline conditions may result from cathodic corrosion under highly reducing potentials. ,

5.

Schematic depiction of the Cu relocation process. (a) Cu dissolves in the electrolyte at OCP, with dissolution being more pronounced in an alkaline catholyte (e.g., KOH) than in a neutral catholyte (e.g., KCl). (b) The reaction site shifts as electrolyte penetration progresses. Cu ions migrate into the MPL/GDL and relocate during CO₂RR operation. The extent of liquid penetration (Cu migration) is greater in an alkaline electrolyte than in a neutral electrolyte. (c) A magnified view of the local environment at the reaction site. CO₂RR proceeds at the triple-phase boundary (TPB) at the electrolyte penetration tip, accompanied by the reposition of dissolved Cu species.

Previous studies have highlighted the importance of residual Cu⁺ species in CO₂RR for improving C₂₊ selectivity. − Additionally, nanoparticle catalyst morphology has been reported to influence C–C bond formation in CO₂RR. Although Cu morphology and chemical state play a role in determining CO₂RR performance, their effect is likely substantial only under nonflooded conditions. We propose that the efficient formation of TPB within the local reaction environment is more critical for achieving high C₂₊ selectivity in commercially viable CO₂RR systems utilizing GDEs. Traditionally, CO₂RR in GDEs has been conceptualized as a TPB-based mechanism, where catalyst particles remain stationary at the GDE while the gas–liquid interface dynamically moves in within the GDE. , Recently, the double-phase boundary (DPB; liquid–solid interface) has been suggested as an alternative model, treating CO₂ in GDEs as dissolved CO₂. Our findings present a novel perspective: Cu catalyst dynamically migrates within GDEs along with the electrolyte and function as a dynamic TPB within the GDE interior (Figure c). This interpretation aligns with prior reports suggesting that the hydrophobic microenvironment within GDEs enhances TPB efficiency and improves CO₂RR selectivity. ,− Moreover, recent operando X-ray scattering investigations of GDE-type CO₂RR cells have underscored the detrimental effects of salt precipitation in the GDE cathode under high current densities (≥100 mA/cm²), leading to decreased CO₂RR selectivity. − Our results demonstrate that C₂₊ selectivity can be partially restored via salt removal through washing; however, once flooding occurs, complete recovery of initial peak performance is challenging (Figures and S17). This suggests that deeply embedded salt precipitates may be difficult to remove by the current washing methods, or that salt deposition may irreversibly damage the porosity of the GDE, thereby increasing permeability and reducing hydrophobicity. Periodic restoration of GDEs via ex situ and in situ washing protocols are necessary to prolong CO₂RR selectivity. , As the strategies to mitigate flooding, developing high-activity catalysts with low onset potentials, or the modification and/or substitution of GDL materials to optimize wettability has also been proposed. Ultimately, technical solutions are required to minimize flooding and improve performance in GDE-type CO₂RR electrolyzers operated under high current densities.

3. Conclusions

This study systematically investigated the correlation between the Cu relocation process and C₂₊ selectivity of Cu-GDE cathodes in CO₂RR operated at practical current densities. Ex situ and in situ SXFM results showed that the Cu catalyst undergoes dynamic structural changes during CO₂RR, migrating and dispersing within the GDE along with electrolyte penetration. In situ XAS results showed that Cu migration to the MPL is governed by the dissolution and redeposition of Cu¹⁺ species. These Cu¹⁺ species may exist as stable dissolved ions, such as [Cu-CO]⁺ complex ions, in the presence of CO₂ or carbonate ions. Based on these findings, we proposed a mechanism in which the Cu catalyst undergoes continuous relocation within the GDE in conjunction with electrolyte permeation in the CO₂RR. We also investigated the time-dependent changes in Cu solubility and C₂₊ selectivity as a function of electrolyte pH. It was found that Cu dissolution occurs under OCP conditions, and that the concentration of dissolved Cu increases at higher electrolyte pH. Analysis of the time-dependent variations in the FEs of gaseous products measured by online GC during continuous CO₂RR revealed that the observed temporal decrease in C₂₊ selectivity was primarily attributed to flooding rather than Cu relocation. Nevertheless, during the period before significant flooding began, the dynamically shifting electrolyte penetration front functioned as an effective TPB, maintaining high C₂₊ selectivity in the CO₂RR. These results highlight the importance of considering not only the optimization of the initial catalyst morphology but also the structural changes and flooding behavior of the catalyst during CO₂RR to improve catalyst selectivity and long-term stability at practical current densities.

4. Experimental Methods

4.1. Materials

All electrolytes were made by dissolving appropriate amounts of chemicals in Milli-Q water (Millipore, resistivity >18.2 MΩ cm). All chemicals were used without any further purification: KCl (99.5+ %), KHCO₃ (99.5%), and KOH (85.0%) were purchased from FUJIFILM Wako Pure Chemical Corporation. AvCarb GDS2130 was used for a commercial carbon-based gas diffusion layer (GDL) with a microporous layer (MPL). A cation exchange membrane (CEM, Nafion 117) and an anion exchange membrane (AEM, Sustanion X37–50) were purchased from Fuel Cell Earth.

4.2. Preparation of Cu-GDEs

Cu­(x)-GDEs (thickness of Cu: x = 70, 300 nm), were prepared by RF magnetron sputtering of Cu onto a carbon-based MPL/GDL. In the case of flow cells, cathodes were prepared by affixing a PTFE seat with a 0.5 cm² aperture onto the MPL/GDL through hot pressing (250 °C, 0.5 ton for 1 min) prior to Cu sputtering. This ensured that the electrode area was fixed. In the case of cross-sectional spectro-electrochemical cells for in situ SXFM, the cathodes were prepared by sputtering Cu over the entire surface of the MPL/GDL and then cutting them into pieces of 0.5 cm × 1.0 cm (0.5 cm²).

4.3. Ex Situ SEM-EDX

SEM images were obtained using a Hitachi S-4800 microscope. EDX elemental mappings were obtained utilizing an Oxford Instruments Aztec Ultim Max spectrometer.

4.4. Electrochemical Measurements for CO₂RR

Electrochemical CO₂RR experiments were performed in a flow cell configuration comprised of three distinct chambers: an anolyte chamber, a catholyte chamber, and a gas flow chamber (Figure S3). The cathode GDE was fixed between the catholyte chamber and the gas flow chamber, with the reverse side oriented toward the gas flow chamber and the catalyst side oriented toward the catholyte chamber. A Pt mesh was used as the counter electrode. The catholyte and anolyte chambers were separated by a CEM or AEM. The Nafion membrane (CEM) was completely hydrated by boiling it in deionized water, 5% hydrogen peroxide (H₂O₂) and 1 M sulfuric acid (H₂SO₄) at 80 °C for an hour each before use. The Sustanion membrane (AEM) was activated by immersion in 1 M KOH for 24 h before use. The catholyte chamber was equipped with an Ag/AgCl electrode (3 M KCl), which served as the reference electrode. The catholyte and anolyte were supplied through separate silicon tubes, each connected to a peristaltic pump. In the case of CO₂RR with neutral electrolyte, a CEM was used and the catholyte was a 1 M KCl solution and the anolyte was a saturated KHCO₃ solution. In the case of CO₂RR with alkaline electrolyte, an AEM was used with a 1 M KOH solution as both catholyte and anolyte. The catholyte was pumped at a rate of 10 mL/min using a peristaltic pump and passed once through the chamber without circulation, while the anolyte was circulated through the chamber. A mass flow controller (MFC) was connected to a CO₂ or an Ar gas cylinder to regulate the flow rate of 30 sccm into the gas flow chamber. All the electrochemical tests were conducted using a Biologic VSP potentiostat/galvanostat, which was equipped with a ± 2 A internal booster. The potential obtained in the flow cell were referenced to a reversible hydrogen electrode (RHE), and 85% iR compensation was performed based on the following equation: E (V vs RHE) = E (V vs Ag/AgCl) + 0.222 + 0.059 × pH + 0.85 × iR. where i is applying current, R is the resistance measured by electrochemical impedance spectroscopy (EIS) at OCP, and pH is the measured value of the bulk electrolyte before use. The current densities were calculated based on the geometric surface area.

The actual outlet gas flow rate of the flow cell was measured using a precision membrane flowmeter (SF-2U, Horiba). The gaseous products (H₂, CH₄, CO, C₂H₄) were quantified utilizing gas chromatography (dual column 990 Micro GC system, Agilent) with a thermal conductivity detector (TCD). The detection of H₂, CH₄, and CO was conducted using a Molsieve 5A (MS5A) column with an Ar as carrier gas. A PorraPlot Q (PPQ) column with a He carrier gas was employed for the detection of C₂H₄. The liquid products were evaluated using gas chromatography (GC2030, Shimadzu) with a flame ionization detector (FID) for alcohols (C₂H₅OH and C₃H₇OH) and high-performance liquid chromatography (HPLC, Nexcera Organic Acid System, Shimadzu) with a conductivity detector (CD) for organic acids (HCOOH and CH₃COOH). The details of Faradaic efficiency calculations are given in Supporting Note S1 in the Supporting Information.

4.5. Ex Situ SXFM

SXFM experiments were conducted at the BL16XU of the SPring-8, Japan. The monochromatized X-ray beam, generated by an undulate and Si(111) double-crystal monochromator (DCM), was focused to a pinhole (virtual X-ray source) by a bend-cylindrical total reflection mirror. A focused micro-X-ray beam was formed by two elliptical total-reflection mirrors of Kirkpatrick-Baez configuration (KB mirrors) with a beam size of 1 μm square or less at the focal position (Figure S5b). The sample was mounted on the XYZ piezoelectric tables (PZT), which have a maximum travel of 250 μm and contraction at 10 V application. Subsequently, the X-ray microbeam was irradiated at the sample, and the X-ray fluorescence generated from the sample was detected by the Silicon Drift Detector (FAST SDD, Amptek), which was positioned diagonally upstream of the sample (Figure S5a). To prevent deformation of the cross-sectional Cu(300)-GDE samples during SXFM measurements, they were placed in a rigid plastic case. The ex situ SXFM images of the Cu(300)-GDE cross-section were obtained with an incident X-ray energy of 10.0 keV (over the Cu–K edge), with a scan area of 250 × 250 μm, with a pitch of 1 μm and 251 × 251 points, the scan rate of 10 Hz (resulting one image acquired in 1.74 h). The ex-situ SXFM images of the X-ray fluorescence intensity of Cu–Kα (Cu map) and K–Kα (K map corresponding to the KCl electrolyte) were subjected to analysis. The SXFM images were subsequently normalized by the incident X-ray intensity (I ₀). Due to the limitations of the scan area, it was necessary to obtain multiple SXFM images in the longitudinal direction, with some overlap in the scan area. Subsequently, the multiple images were superimposed and analyzed as a single longitudinal SXFM image.

4.6. In Situ XAS

XAS measurements were conducted at BL16B2 at SPring-8, Japan. Cu–K edge XAS was measured using a Si(111) monochromator. The beam size was 2.0 mm (H) × 1.0 mm (V). All data were recorded in fluorescence using the quick-XAS (QXAS) mode with a 25-element Ge solid-state detector (Canberra, 25SSD). A schematic and photograph of the in situ XAS measurement was shown in Figure S13. The configuration of the cell and the conditions for CO₂RR were identical to those of the flow cell (Figure S3), but the outer wall of the gas chamber was replaced with a Kapton window to permit X-ray transmission. The X-rays were incident from the rear of the Cu(70)-GDE at a horizontal tilt of 45̊, and the Cu–K fluorescence X-rays were detected at the rear of the GDL. The in situ fluorescence XAS measurements were acquired on a repetitive and continuous basis, with each spectrum recorded over a period of 300 s. The Cu–K edge XANES data were processed using the LabView custom software and the Athena program of the Demeter data analysis package. To elucidate the role of Cu¹⁺ during electrolysis, multipeak fitting of the first derivative XANES spectra was conducted based on the energy positions of Cu⁰ (8978 eV) and Cu¹⁺ (8980 eV).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c07944.

• Quantification of CO₂RR products, ex situ X-ray CT procedure and images, schematic diagram of the flow cell for CO₂RR, in situ SXFM experimental setup and results, in situ XAS experimental setup, multipeak-fitting of the first derivative XANES spectra, ex situ STEM-EDX images, ICP-MS results of the electrolytes, time-dependence of FEs during CO₂RR, ECSA measurements (PDF)

The manuscript was written through the contributions of all authors.

The authors declare no competing financial interest.