Oxides and Carbonates Accelerate Copper Instability in CO₂ Electroreduction

Abstract

The electrochemical CO₂ reduction reaction (CO₂RR) is one of the key chemical transformations promoting the transition from fossil fuel-based energy systems to renewable systems. Copper (Cu)-based materials uniquely catalyze the production of multicarbon (C₂ ⁺) products from CO₂. Yet, copper operational instability limits long-term performance. Herein, we investigate the impact of the chemical nature of the initial Cu surface, particularly oxidation state and carbonate formation, on the structural and operational stability of Cu catalysts along with the reconstruction kinetics of the catalyst. We combine state-of-the-art well-defined catalysts with quasi-operando electrochemical liquid-phase transmission electron microscopy (ec-LPTEM) along with electrochemical characterization to learn about underlying differences. We demonstrate that catalysts with higher initial oxide content undergo faster structural reconstruction and suffer from faster operational deactivation. Interestingly, we find that Cu carbonates further exacerbate structural instability while also suppressing the CO₂RR activity. Our results highlight the critical role of oxides and carbonates in dictating the reconstruction pathways and durability of Cu under CO₂RR conditions, offering insights into tuning the Cu-based catalyst design for enhanced CO₂RR stability and efficiency.

Introduction

The electrochemical CO₂ reduction reaction (CO₂RR) into valuable chemical feedstocks is a promising lever to move from fossil to renewable energy while closing the carbon loop. , Cu-based materials remain the most promising catalysts to convert CO₂ into C₂₊ products, including ethylene, ethanol, and acetate. − Efforts have been primarily devoted to catalyst design and cell engineering to overcome the intrinsic lack of selectivity of Cu catalysts, leading to near unity faradaic efficiency for some of the major products at industrially relevant currents. The overarching hurdle is now the long-term operation stability of CO₂RR, which the structural reconstruction of Cu-based catalysts substantially contributes to along with electrochemical reactor failure (e.g., flooding). ,−

Different processes are involved in the reconstruction of Cu catalysts during CO₂RR. ,− First, the Cu surface oxidizes, when starting with metallic surfaces, and dissolves in contact with the aqueous electrolyte at open circuit potential, releasing Cu ions. , Then, the oxide reduces and the Cu ions are electrodeposited during the cathodic ramp to operational potential, triggering the first phase of structural reconstruction. − Finally, transient carbonyl intermediates drive the second phase of copper reconstruction through surface diffusion and dissolution-redeposition cycles during operational CO₂RR. ,,, These processes are intrinsic to Cu catalysts for CO₂RR and, thus, are expected to be independent of the electrochemical reactor used. Yet, discrepancy exists in the literature regarding the operational stability of Cu electrodes, with time scales ranging from less than 1 h to hundreds of hours. −

While often generically referred to as “Cu electrodes”, the chemical nature of the initial Cu catalyst surface might vary. Indeed, metallic Cu, Cu oxides, and Cu exposed to preconditioning treatments (e.g., exposure to open-circuit potential, air oxidation during electrode preparation) are used across the literature. ,− In addition, carbonates have been recently suggested to contribute to copper surface diffusion and associated with reactor failure. , The discrepancies in stability of Cu electrodes across the literature suggest that the chemical nature of the initial copper surface might influence reconstruction kinetics and pathways and, thus, operational stability. The impact of the copper oxide on selectivity has been studied. , However, studies addressing the impact of the chemical nature of the initial surface on copper reconstruction and stability have been limited so far. ,

Here, we combine well-defined Cu catalysts with quasi-operando electrochemical liquid-phase transmission electron microscopy (ec-LPTEM) along with electrocatalytic performance and electrochemical descriptors to correlate the initial Cu surface composition to the structural and operational stability of Cu catalysts during the CO₂RR. We demonstrate that oxide-rich catalysts undergo faster structural reconstruction and suffer from faster operational deactivation compared to metallic-rich catalysts. Interestingly, we find that Cu carbonates form at open-circuit potential and upon air exposure and further exacerbate the structural instability while suppressing the CO₂RR activity, thus providing additional insight into the critical role of carbonates for Cu catalyst deactivation.

Results and Discussion

Tuning and Characterizing the Initial Surface Composition of Copper Catalysts

We employed well-defined Cu-based nanocrystals (NCs) as our catalytic platform. We compare metallic Cu cubic NCs, Cu oxide (Cu₂O) cubic NCs, and pretreated Cu cubic NCs. We choose exposure to open-circuit potential (OCP) and air exposure as pretreatments; we refer to these samples as Cu_OCP and Cu_air NCs, respectively.

Metallic Cu cubic NCs have been critical for elucidating facet-dependent selectivity, , for enabling active site engineering, , and for investigating structural degradation during CO₂RR , while being compatible with industrially relevant conditions. , Cu₂O cubic NCs have been widely used to prove the impact of oxides on selectivity. , All Cu electrodes eventually undergo exposure to air and open-circuit potential, which motivated us to select these pretreatments. We choose long exposure times for both OCP (15 h) and air (1 week) to mimic day on/day off intermittent operation and an eventual sample handling during shipping/storage, respectively, along with aiming at maximizing the effect of these pretreatments on the sample composition (Figure S1).

We thoroughly characterized the samples by combining different techniques (Figure ). High-angle annular dark field scanning transmission electron microscopy (HAADF STEM) shows that all catalysts are cubic and have similar size (Figure a,e,i,m). More interestingly, the images provide information on the presence and uniformity of surface oxide layers for Cu, Cu_OCP, and Cu_air NCs. The Cu NCs have a lower contrast patchy shell on their surface (Figure a–d). High-resolution (HR)­STEM imaging identifies lattice fringes corresponding to Cu₂O {111} planes (Figure b). The Cu₂O NCs match what is expected for these materials, which is fully crystalline Cu₂O (Figure e–h). The Cu_OCP NCs and Cu_air NCs both possess a uniform Cu₂O surface layer (Figure i–l,m–p, respectively). The oxide layer is thinner for the Cu_OCP (≈1.5 nm) than for the Cu_air NCs (≈2.5 nm) based on HRSTEM (Figure j,n). Selected area electron diffraction (SAED) (Figure q) and electron energy loss spectroscopy (EELS) (Figure r) of all samples confirm the oxide identification made via HRSTEM.

1.

Characterization of Cu catalysts with tunable surface composition. (a–p) HAADF STEM and HRSTEM image along with 3D schematic representation of Cu (a–d), Cu₂O (e–h), Cu_OCP (i–j), and Cu_air (m–p) NCs. Scale bars, 50 nm, 10 nm, and 5 Å for HAADF STEM, inset, and HRSTEM images, respectively. (q) SAED radial profile of all studied samples along with references. (r) EEL spectra of the Cu L_3,2 edge for all studied samples. (s) Surface copper speciation retrieved from X-ray Auger electron spectroscopy (XAES) (left axis) along with the oxide thickness measured from the HRSTEM (b,j,n) (right axis).

Complementary to electron microscopy, Cu LMM X-ray Auger electron spectroscopy (XAES) provides insight into the surface composition (≈5 nm) of the different catalytic systems (Figure s and Figure S2). The Cu NCs are predominately metallic with 75% Cu(0), 12% Cu­(I) (Cu₂O), and 13% Cu­(II) (CuO/Cu­(OH)₂). Cu₂O NCs are composed of 60% Cu­(I) and 40% Cu­(II), which is in line with previous studies. , The Cu_OCP NCs consist of 40% Cu(0) and show a higher fraction of oxide compared with the metallic Cu NCs (Cu­(I) 15% and Cu­(II) oxide 20%). Interestingly, a significant contribution of Cu carbonates (i.e., CuCO₃ and Cu₂CO₃OH) emerges (c.a. 25%) for this sample. The Cu_air NCs have only traces left of metallic Cu (less than 1%) and consist of Cu oxide (Cu­(I) oxide 15% and Cu­(II) oxide 35%) and Cu carbonate (c.a. 50%). Studies on Cu atmospheric corrosion report on the formation of Cu carbonate. , The data reported explicitely correlate the formation of Cu carbonate and CO₂RR conditioning. Specifically, the formation of Cu carbonate depends on the presence of CO₂ and on the time of exposure to air (Figure S3).

Catalytic Performance for CO₂RR

Next, we evaluated the impact of the initial surface composition of the Cu catalysts on the catalytic performance for the CO₂RR (Figure and Table S1 and Figure S4). We focused on −1.1 V vs RHE as one representative potential comparable with previous studies. − ,−

2.

CO₂RR performance and long-term operational stability. (a) Total FEs for all gaseous products (that is, H₂, CO, CH₄, and C₂H₄) and the main liquid products (that is, HCOO^–, acetate, ethylene glycol, C₂H₅OH, and 1-propanol) at −1.1 V vs RHE for 1h. (b–e) Evolution of the FEs for all gaseous products (that is, H₂, CO, CH₄, and C₂H₄) during 10 h of CO₂RR for Cu (b), Cu₂O (c), Cu_OCP (d), and Cu_air (e) NCs. The data are the average of three independent experiments, and the error bars are the calculated standard deviation.

First, we comment on the faradaic efficiency (FE) of each sample, which was calculated based on the first hour of operation (Figure a). The Cu and Cu₂O NCs are mostly selective for ethylene with around 30% faradaic efficiency (FE). The results are in line with previous studies. − ,, Interestingly, the FE for methane of Cu_OCP NCs is nearly double compared to Cu NCs (from around 10% to 20%) while the FE of ethylene drops to 20%. Finally, Cu_air NCs are mostly selective for hydrogen with an FE above 60%.

Second, differences between the catalysts become more pronounced when looking at the FE over the 10 h CO₂RR (Figure b–e, Table S1 and Figure S4). The product distribution of the Cu NCs remains mostly stable compared with the other samples (Figure b). A slight decay in ethylene FE occurs starting at around 6 h, matching our previous results. The Cu₂O NCs show signs of earlier CO₂RR deactivation (Figure c); here, ethylene FE decreases after 2–3 h. This observation is consistent with previous studies. ,, Interestingly, the Cu NCs oxidized to match the surface composition of Cu₂O NCs also show an earlier sign of CO₂RR deactivation (Figure S5). This observation strengthens the connection between the higher oxide content and accelerated CO₂RR deactivation. A more pronounced HER increase at the expense of the CO₂RR occurs for Cu_OCP (Figure d). This faster loss in ethylene selectivity for Cu_OCP compared to Cu₂O and Cu_H2O2 indicates that the Cu speciation strongly influences the catalyst deactivation rates, where the presence of Cu carbonate seems to accelerate those effects. Cu_air remains almost inactive for CO₂RR over the 10 h with more than 60% hydrogen FE (Figure e).

Altogether, these observations indicate a clear difference in the stability of the product distribution of the investigated systems, which was herein compared under the same conditions for the first time. We observed comparable, if not poorer, C₂₊ selectivity between Cu and the oxide-rich catalysts. Our data highlight that oxide-derived catalysts do not necessarily enhance C₂₊ selectivity but rather greatly impact the operational stability of the catalyst (Supplementary Note 1). All the oxide-rich catalysts undergo faster operational deactivation for ethylene selectivity compared to the metallic-rich catalyst while high content of surface carbonates generate a HER-active Cu catalyst.

Quasi-Operando Liquid-Phase TEM Monitoring Catalyst Restructuring

We chose electrochemical liquid-phase transmission electron microscopy (ec-LPTEM) to follow the eventual restructuring of the NCs triggered by the different surface composition. ec-LPTEM has been demonstrated to provide valuable insight into Cu catalyst reconstruction during the initial stages of CO₂RR. ,,,,, While ec-LPTEM operates on a short-time scale, the structural changes monitored by this technique have been essential to better understand CO₂RR catalyst behavior on longer time scales. ,,,,, In contrast, insights gained from post-mortem TEM have been less useful because of the structural and compositional Cu changes occurring upon exposure to OCP. ,,−

We monitored the morphological and structural changes during the potential ramp from OCP to −0.8 V vs RHE by linear sweep voltammetry (LSV) (i.e., start-up) and subsequent chronoamperometry (CA) at −0.8 V vs RHE for 5 min via synchronized image sequences, particle analysis data, and selected area electron diffraction (SAED) (Figure ) along with performing several control experiments, including the exclusion of beam effects (Figures S6–S12).

3.

Quasi-operando liquid-phase TEM monitoring catalyst restructuring at CO₂RR conditions. (a,c,e,g) Relative area loss of Cu (a, red), Cu₂O (c, green), Cu_OCP (e, blue), and Cu_air (g, yellow). (b,d,f,h) Time-lapse images of Cu (b), Cu₂O (d), Cu_OCP (f), and Cu_air (h) NCs during LSV-CA. Arrows of corresponding color indicate redeposited particles in panels (b,d,f). Arrows in panel (h) indicate hollowing events of primary Cu NCs. Scale bar: 100 nm. In panel (h), insets depict hollowing of one single NC. Scale bar, 20 nm. (i) Average area of redeposited particles and the onset time of redeposition for all samples. (j) In situ SAED of NCs before and after LSV-CA measurement.

The Cu NCs experience a progressive decrease in the projected area during the LSV, which continues during the CA up to 250 s before stabilizing at a total relative area loss of around 15% (Figure a). The synchronized image sequence shows that the formation of secondary particles (at ca. 70s) accompanies the size decrease of the Cu NCs (Figure b, Figures S6 and S7). The behavior of the Cu NCs agrees with previous ex situ observations on the same NCs. The size decrease and formation of the secondary particles is indicative of the dissolution-redeposition process driving Cu reconstruction. ,,

The Cu₂O NCs undergo a more pronounced and steeper decrease in the projected area during LSV compared with the Cu NCs. The change continues during the first 30 s of CA before stabilizing at a total area loss of around 30% (Figure b). Concomitantly, an earlier formation onset of secondary particles (at ca. 10 s) and significant sintering of the Cu₂O NCs occurs (Figure c,d and Figure S8). These observations are in line with previous ex situ and in situ results on the same NCs. ,

The Cu_OCP NCs exhibit a similar behavior to the Cu NCs, although with most change occurring during LSV and less change occurring during CA with fewer secondary particles forming (Figure e,f and Figure S9).

The Cu_air NCs differ significantly in their restructuring behavior from the rest of the samples with their unique facet-selective core etching (hollowing) and a shell (edge and corner) remaining intact while secondary particles still form (Figure g,h).

Plotting the average area of redeposited secondary particles together with the onset time of their observation indicates a faster and more drastic reconstruction occurring for the Cu₂O NCs among all samples and a larger extent of reconstruction for the Cu_OCP/Cu_air compared to the Cu NCs (Figure i).

Complementary, in situ SAED of the samples confirms that all the Cu catalysts consist of mostly a metallic phase after LSV-CA (Figure j).

Strikingly, stabilizing the Cu surface by means of an inert oxide shell (here amorphous ZrOx) following our previous studies results in no change in the ec-LPTEM (Figure S10–S12), which confirm the importance of redox surface changes in the Cu stability during CO₂RR. ,,

Overall, the quasi-operando electron microscopy evidence different kinetics among the CO₂RR active Cu samples (i.e., Cu, Cu₂O, and Cu_OCP) while a different mechanism emerges for the Cu_air, which is HER-active.

Electrochemical Characterization Monitoring Catalyst Active Area and Selectivity

Additionally, we examined in more depth the partial current densities, the electrochemically active surface area (ECSA), and the selectivity trends from the benchtop measurements (Figure ).

4.

Electrochemical characterization monitoring catalyst active area and selectivity. (a) CO₂ partial current density normalized by the geometric area as a function of time. (b) ECSAs from Pb-UPD of the different catalysts before CO₂RR and after 1 and 10 h of CO₂RR 5 min at −1.1 V vs RHE in CO₂RR conditions were applied on Cu₂O prior Pb-UPD for “before”. (c) Relative change in selectivity for H₂, CO and C₂H₄ between 1 and 10 h.

The change in the CO₂ partial current density normalized by the geometric surface area (J_CO2‑geo) as a function of time reflects the evolution of the total number of CO₂RR active sites during the 10 h electrolysis (Figure a). The data indicate that Cu₂O and Cu_OCP NCs undergo the most drastic change. Indeed, the J_CO2‑geo values of the Cu NCs remain stable. On the contrary, the J_CO2‑geo of the Cu₂O NCs and Cu_OCP NCs decreases over time, with a final loss of 20 and 45%, respectively. Cu_OCP NCs deactivate faster than the Cu₂O NCs. The Cu_air NCs remain overall inactive for the CO₂RR during the 10 h. The total current density also decreases for the Cu₂O NCs and Cu_OCP NCs without any catalyst detachment being observed (Figure S13). Additionally, the J_CO2‑geo of surface-oxidized Cu NCs, which possess similar Cu(0) content of Cu_OCP NCs yet no carbonate (Figure S5), decreases similarly to Cu₂O NCs (Figure S14), which points to carbonates playing a key role in deactivation kinetics.

The observed evolution over time in current density, while knowing that all catalysts reconstruct, indicate that the number of active sites and/or their selectivity (e.g., more sites that are less active or active for different products) must change for all.

We used ECSA to qualitatively examine the change in the total number of active sites, whether CO₂RR or HER-active sites (Figure b and Figure S15). The ECSA of the Cu NCs continuously increases from before the CO₂RR to 10 h after CO2RR. The ECSA of the Cu₂O NCs increases from before to 1 h to reach the same value of the Cu NCs. The ECSA of Cu_OCP and Cu_air NCs only slightly fluctuates around a value that is lower compared to the Cu NCs and Cu₂O NCs. Pb-UPD confirms the slower transformation of the Cu NCs from low-coordinated atoms (i.e., defects, corners (111), and edges (110)) to more stable coordination geometry (100) compared to the Cu₂O NCs (Figure S16,17).

Then, we looked at the relative change in selectivity for hydrogen, CO, and ethylene between 1 and 10 h of the CO₂RR (Figure c). In addition to hydrogen and ethylene, we chose CO because of its role as key intermediate in the CO₂RR mechanism and in driving copper reconstruction via Cu-CO intermediate formation. ,,,,, The ethylene remains constant for Cu NCs while it decreases for Cu₂O and Cu_OCP NCs. Hydrogen increases for Cu₂O and Cu_OCP NCs. The CO selectivity increases only for Cu NCs. Cu_air does not undergo major change.

Discussion

The evolution of the catalytic performance and the catalyst restructuring indicate a clear difference between the Cu, Cu₂O, Cu_OCP, and Cu_air catalysts, which possess a different initial surface composition. The presence of oxides and carbonates correlates with faster operational CO₂RR deactivation and structural changes to the extreme point of having a catalyst not active at all for CO₂RR (i.e., Cu_air). Interestingly, previous studies in flow cells and MEA also point at a higher stability of metal Cu compared to Cu oxide-derived catalysts. ,,,,,

Figure provides a summary of the findings. The catalysts with higher initial oxide content (i.e., Cu₂O followed by Cu_OCP) undergo faster and/or more drastic structural changes compared to catalysts with higher initial metallic content (i.e., Cu NCs). The metal-rich Cu exhibits a more stable CO₂RR and, particularly, C₂₊ selectivity compared to oxide-rich Cu. Eventually, CO selectivity increases for the metal-rich Cu while hydrogen selectivity increases for oxide-rich Cu. Remarkably, the CO₂RR deactivation effect becomes even more pronounced in the presence of Cu carbonates and HER dominates along with a uniquely different reconstruction pathway (i.e., Cu_air NCs).

5.

Schematic representation summarizing the impact of the initial Cu catalyst composition on catalyst reconstruction and the CO₂RR selectivity.

Metallic Cu dominates in all catalysts at cathodic potential, corroborated here by the quasi-operando ec-LPTEM and in agreement with previous studies. ,,,,, The data suggest that the initial surface Cu composition impacts the reconstruction kinetics and the type of generated active sites responsible for selectivity.

Studies evidence that the greater the initial oxide content on copper is, the more the formation of grain boundaries, defects, and undercoordinated atoms during CO₂RR is favored. ,,,,,,− A few studies indicate that these low-coordinated sites interact more weakly with CO. We suggest that the oxide-rich Cu transforms more rapidly into these low-coordinated sites, shifting the selectivity from the CO₂RR to the HER-active sites. The observation of continuously increasing CO selectivity for the metal-rich Cu might indicate that different low-coordinated active sites are generated from the slower reconstruction or that the slow generation of the same weakly CO-interacting low-coordinated active sites enables the CO detection. HER might then still take over at longer operation time than those studied in this work.

A few reports hint at poor CO₂RR activity associated with Cu carbonates. , Postelectrolysis Cu oxide cubes formed at OCP in carbonate electrolyte showed irreversible structural transformation and suppressed C₂₊ selectivity compared to the initial Cu spheres, hinting at a possible detrimental role of Cu carbonate. One study demonstrated that Cu carbonate is a poor CO₂RR catalyst compared to Cu oxide-derived catalysts. Complementary in situ surface-enhanced Raman spectroscopy (SERS) indicates persistence of the Cu-carbonate signal during the CO₂RR (Figure S18andTable S2). Thus, we suggest that the Cu carbonate might render the surface inaccessible for CO₂RR or that the reduction of Cu carbonates generates active sites with poor binding affinity for CO₂RR intermediates binding affinity.

As for the observed hollowing for the carbonate-rich Cu, competing reconstruction mechanisms might occur in the absence of significant CO₂RR activity, which are different than the CO-driven reconstruction. In particular, cathodic corrosion has been previously identified to drive reconstruction of HER-active catalysts and has been recently suggested to cause hollowing of Cu cubes, although not yet connected to carbonate presence and HER activity. − One alternative or contributing pathway might involve continuous reduction–oxidation cycles of surface Cu wherein the reduction is voltage-driven while the reoxidation is caused by OH* radicals generated from oxygen transfer between H₂O and HCO₃ ^–. This cycle might ultimately lead to the dissolution of the internal metallic copper of carbonate-derived Cu NCs through hollowing.

Conclusions

In conclusion, this study clarifies how the initial copper surface composition critically dictates structural catalyst evolution and operational stability during the CO₂RR. Our results underscore that greater initial oxide content triggers faster reconstruction and CO₂RR performance degradation. Interestingly, the formation of Cu carbonates emerges as particularly detrimental, potentially persisting under cathodic conditions and promoting irreversible surface changes promoting HER and new reconstruction pathways. These findings might help to explain previously conflicting literature and emphasize the importance of precise control over surface composition and preconditioning in designing robust Cu catalysts. The comparison of Cu catalysts with different initial copper surface composition under the same conditions also raises an interesting question on the differences in their product selectivity, which might open up follow up studies.

Overall, this work inspires future strategies that consider both chemical and structural factors in catalyst preparation to achieve selective and durable CO₂RR performance across various reactor platforms.

Experimental data are openly available in Zenodo at https://zenodo.org/uploads/18235845

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c21287.

• Supplementary methods; additional experimental considerations; supplementary notes; additional TEM, XPS, Pb-UPD, and CO₂RR (PDF)

+.

Petru P. Albertini and Saltanat Toleukhanova contributed equally.

The authors declare no competing financial interest.