Exploration of Gas-Dependent Self-Adaptive Reconstruction Behavior of Cu₂O for Electrochemical CO₂ Conversion to Multi-Carbon Products

Highlights

• We revealed a universal self-adaptive structural reconstruction from Cu₂O to Cu@Cu_xO composites, ending with feeding gas-dependent microstructures and catalytic performances.

• We uncovered a CO₂-induced passivation behavior by identifying a reduction-resistant but catalytic active Cu(I)-rich amorphous layer.

• We designed and fabricated hollow Cu₂O nanospheres, demonstrating durable electrolysis at a partial current density of −200 mA cm⁻² in producing C₂H₄ with an FE of up to 61% at −0.6 V_RHE.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40820-024-01568-1.

Abstract

Structural reconstruction of electrocatalysts plays a pivotal role in catalytic performances for CO₂ reduction reaction (CO₂RR), whereas the behavior is by far superficially understood. Here, we report that CO₂ accessibility results in a universal self-adaptive structural reconstruction from Cu₂O to Cu@Cu_xO composites, ending with feeding gas-dependent microstructures and catalytic performances. The CO₂-rich atmosphere favors reconstruction for CO₂RR, whereas the CO₂-deficient one prefers that for hydrogen evolution reaction. With the assistance of spectroscopic analysis and theoretical calculations, we uncover a CO₂-induced passivation behavior by identifying a reduction-resistant but catalytic active Cu(I)-rich amorphous layer stabilized by *CO intermediates. Additionally, we find extra CO production is indispensable for the robust production of C₂H₄. An inverse correlation between durability and FE_CO/FE_C2H4 is disclosed, suggesting that the self-stabilization process involving the absorption of *CO intermediates on Cu(I) sites is essential for durable electrolysis. Guided by this insight, we design hollow Cu₂O nanospheres for durable and selective CO₂RR electrolysis in producing C₂H₄. Our work recognizes the previously overlooked passivation reconstruction and self-stabilizing behavior and highlights the critical role of the local atmosphere in modulating reconstruction and catalytic processes.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40820-024-01568-1.

Introduction

Electrocatalytic CO₂ reduction to chemical feedstocks and value-added products coupled with intermittent renewable energy is a promising approach to warrant carbon neutrality [1, 2]. Electrocatalysts play a pivotal role in energy-effective conversion. Among them, Cu-based materials have attracted the most research interest, as they can reduce CO₂ to a wide range of hydrocarbons and oxygenates (especially the C₂₊ products) [3–6]. Strategies including selective facet exposure [7], morphology design [8], electronic state modulation [9, 10], surface modification [11], valence modulation/oxide reconstruction [12], etc. have been utilized to optimize the catalytic performances. However, it remains challenging to prompt the C₂₊ products selectively and effectively, due to the sluggish CO₂ activation and coupling of carbon–carbon bonds (C–C) and the deactivation of catalysts [10, 13, 14].

In comparison with polycrystalline bulk-like Cu [15–17], their oxides and derivatives (Cu_xO) favor value-added multi-carbon products [18–20]. Cu(I) species are suggested to be of vital significance in this process [21–23], due to the favorable *CO adsorption and C–C coupling [17, 24, 25]. Unfortunately, Cu(I) species are thermodynamically unstable. It undergoes over-reduction to metallic Cu(0) (0.52 V vs. RHE), leading to the decay of activity and selectivity [26]. The transition has been ubiquitously disclosed in copper oxides (Cu₂O or CuO) and even metallic Cu(0), accompanied by structural reconstruction in most cases [27, 28]. Much effect has been devoted to elucidating the reconstruction process, aiming to stabilize Cu(I) by interfacing, doping, pre-oxidation, or reviving them by pulse electro-reduction technique [29–31]. Due to its dynamic nature, the behavior has yet been superficially understood, leading to controversial structure-performance relationships. The performances (activity, selectivity, and durability) vary from group to group for the same catalyst [32–34]. The active phases were recognized with different compositions and microstructures, even if catalysts were fabricated by the same method. In some cases, Cu(0) was only detected, questioning the attribution of high C₂ selectivity to the presence of Cu(I) for C–C coupling [35].

Besides catalysts themselves, microenvironments including gas accessibility, local pH, and triple-phase interface are equally important [36, 37]. The impact of microenvironments on practical electrolyzers has been well investigated but with much emphasis on linking them to catalytic processes. It should be noted that redox/chemical reactions are usually coupled with reconstruction and occur prior to or concurrently with catalytic reactions. In this sense, microenvironments could modulate redox/chemical reactions and the related reconstruction behavior in a similar way as catalytic reactions. However, this has yet been largely overlooked. Wang et al. showed that the CO₂ atmosphere preferred electrodeposition of Cu-based catalyst favorable for C₂H₄ [38]. A specific *CO binding was suggested to be vital in guiding the preferential facet growth for good selectivity. Alivisatos et al. revealed that feeding gases could drive diverse reconstruction of metallic Cu nanocrystals under catalysis [39]. Ren et al. observed nanoscale surface roughness on the Cu surface induced by a strong affinity of CO under CO₂-sufficient conditions [40]. Niu et al. demonstrated that the CO₂ atmosphere facilitates the Lanthanum leaching, thereby exposing active Cu species in orthorhombic-type perovskite La₂CuO₄ pre-catalyst [41]. These preliminary results indicate a strong link between atmosphere and structural reconstruction. However, how they interact with each other and further with the catalytic process is still vague.

Here, we investigate the influence of feeding gas on the structural reconstruction of CO₂RR catalysts using well-defined Cu(I)-rich Cu₂O as pre-catalysts and further link it to catalytic activity and durability. Though the catalytic process and structural evolution of Cu₂O have been well investigated [40, 42, 43], their dependence on feeding gas has been unexplored to the best of our knowledge. Revisiting the reconstruction by taking gas accessibility into account could provide additional information to rationalize previous debates [44, 45]. Cu₂O is electrochemically activated in the Ar atmosphere to monitor the reconstruction in CO₂-deficient regions and compared to that activated in the CO₂ atmosphere (Scheme 1). Microstructural and spectroscopic analysis is then conducted to identify the structural reconstruction and link it to catalytic processes and performances. A CO₂-induced passivation behavior is disclosed in the reconstruction process. Furthermore, a self-stabilization behavior involving the absorption of *CO intermediates is uncovered in the catalytic process. The strong affinity between metastable Cu(I) sites and *CO intermediates is highlighted and suggested to be essential in structural reconstruction, C–C bond coupling reaction, and durable electrocatalysis. Finally, hollow Cu₂O nanospheres favoring CO₂-rich reconstruction are fabricated purposely, which validates our new understanding of atmosphere driving self-adaptive reconstruction and catalytic process.

Scheme 1.

Illustration of sample synthesis, CO₂ accessibility-dependent reconstruction, and catalytic performances

Experimental Section

Reagents and Materials

All reagents and materials were used directly with no further purification. Copper sulfate pentahydrate (CuSO₄·5H₂O, 99.9%), Cupric chloride (CuCl₂·2H₂O, 99.99%), sodium hydroxide (NaOH, 98%–100.5%), potassium hydroxide (KOH, 99.99%), potassium bicarbonate (KHCO₃, 99.99%), L-Ascorbic acid (C₆H₈O₆, ACS) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Nafion^® (D520) dispersion was purchased from Dupont. Nickel foam, Sigracet 28BC carbon paper with a gas diffusion layer, and the anion exchange membranes (FAA-3-50, Fumapem, and Sustainion^® X37-50 Grade RT, Dioxide Material) were purchased from the Fuel Cell Store.

Synthesis of Cu₂O Nanocrystals

Cu₂O Cube

10 mL CuCl₂ solution (0.1 M) was added into 90 mL distilled water dropwise to form a homogenous solution. After the solution was heated to 55 °C for 30 min, 10 mL NaOH aqueous solution (2 M) was dropped slowly till the solution turned from blue to dark brown. After stirring for another 30 min, 10 mL of 0.6 M L-ascorbic acid was added. The above mixture was stirred for an extra 3 h before centrifugation. Then the as-prepared sample was rinsed with deionized water and ethanol. Finally, the powders were dried in a vacuum box at room temperature.

Cu₂O Hollow Spheres

0.05 g CuSO₄·5H₂O was dissolved in 100 mL of 0.1 M CTAB aqueous solution. Then 0.18 g L-ascorbic acid was added. The above mixture was heated to 60 °C and maintained for 20 min. 0.2 M NaOH was added dropwisely into the above solution to form a yellow precipitate. After stirring for another 10 min, the precipitate was centrifuged, washed sequentially with deionized water and ethanol several times, and then dried at 50 °C under vacuum.

Gas-Mediated Reconstruction

The atmosphere-mediated reconstructed catalyst was accomplished via applying potential by bubbling various gases into the solution for a certain time, and then the pretreated catalyst was used for further electrocatalytic tests. During the pretreatment, the flow rate of gases was controlled at 10 s.c.c.m. via a mass-flow controller.

Density Functional Theory Calculations

All density functional theory (DFT) calculations were performed using the Vienna ab initio simulation package (VASP). The generalized gradient approximation (GGA) with PBE functional was applied. The core electrons were approximated through projector augmented wave functions (PAW). We employed a kinetic cutoff energy of 500 eV for the plane-wave basis set. A smearing of 0.1 eV was added to facilitate the convergence of the wave function. Geometries were optimized until the energy was converged to 1.0 × 10^–6 eV atom^–1 and the force was converged to 0.05 eV Å^–1. Initial structures of Cu and Cu₂O were obtained from the Materials Project database based on our X-ray diffraction (XRD) and transmission electron microscopy (TEM) results. After building the slab models, a vacuum layer of 15 Å was added to avoid interactions between adjacent images. These slab models’ bottom two atomic layers were kept fixed during the simulations.

The Gibbs free energy change for each reaction step is calculated as:

$$\mathrm{\Delta }G=\mathrm{\Sigma }G\left(\text{products}\right)-\mathrm{\Sigma }G\left(\text{reactants}\right)$$

where G(i) is the Gibbs free energy of species i. Gibbs free energy of each species was calculated as G = E + ZPE − TS, where E is the total energy obtained from DFT calculations, ZPE is the zero-point energy, and S is the entropy. Temperature T was set to be 298.15 K.

Results and Discussion

Feeding Gas-Dependent Microstructure Evolution of Cu₂O Nanocubes

Cu₂O nanocubes were synthesized by a previously reported method [33] (see details in the Experimental Section). The scanning electron microscopy (SEM) images show that Cu₂O nanocrystals are uniform in size of ~ 180 nm (Fig. S1). The atomic ratio of Cu and O is around 73:27 for a single cube according to the STEM EDS mapping (Fig. S1 and Table S1), indicating the presence of some O vacancy. Each nanoparticle has the nature of a single crystal, as evidenced by TEM and corresponding high-resolution transmission electron microscopy (HRTEM) images (Figs. 1a and S2). They are enclosed by (100) facets for Cu₂O nanocubes. The clear lattice fringes in HRTEM images and the intense diffraction peaks in XRD patterns (Figs. 1a and S1) indicate the highly crystalline nature. The interplanar spacings of 2.135, 1.51, and 3.02 Å are ascribed to the (200), (220), and (110) planes of Cu₂O nanocrystals, respectively. The XRD peaks correspond well to the cubic phase of Cu₂O (PDF#05-0677, Figs. 1e and S1), indicating the absence of crystalline impurity.

Fig. 1.

Feeding gas-dependent structural evolution of Cu₂O nanocubes in the pre-electrolysis at a potential of −1.1 V versus RHE in an H-cell. TEM and HRTEM images and corresponding SAED patterns of a pristine Cu₂O, and those after pre-electrolysis in CO₂ atmosphere for b 1000 s and c 2000 s, and d after pre-electrolysis in Ar atmosphere for 2000 s; The a2 and a3 show the edge of nanocube in a1. The a4 is the SAED pattern of the nanocube in a1. The b2 shows the edge of the nanocube in b1. The d2 is the enlarged rectangle region in d1. The d3 is the enlarged square region in d2. The inset in d3 is the SAED pattern. The c3 and c4 are the enlarged square regions marked with numbers ① and ② in c2, respectively. e XRD patterns and f, g Cu LMM Auger spectra of Cu₂O-CO₂ and Cu₂O-Ar after pre-electrolysis for different periods

Based on previous electrochemical analysis of Cu₂O for CO₂RR catalysis [33, 34], a typical potential of −1.1 V versus RHE was applied to drive the surface reconstruction for 2000 s in CO₂ and Ar, which monitors the CO₂-rich and CO₂-deficient micro-environments, respectively. The samples are denoted as Cu₂O-CO₂ and Cu₂O-Ar, respectively. Techniques including TEM, XRD, and Auger spectra were employed to probe the reconstruction behavior (Figs. 1 and S3–S8). In particular, the identical location TEM technique was used to trace the structural evolution at the same location (Fig. 1a–d), which gives intact information on structure change. The pristine Cu₂O can be identified with a well-defined single-crystal structure (Figs. 1a and S3). It shows that the crystalline surface underwent amorphization in a CO₂-rich atmosphere. The amorphous layer of Cu_xO was ~ 10 nm in thickness in 1000 s’ pretreatment, and prone to expand toward the interior in prolonged electrolysis (Figs. 1b and S4). Meanwhile, the solid interior started to fragment into small nanoparticles and eventually became porous or/and hollow in 2000 s’ electrolysis (Figs. 1c and S4) [40]. Careful examination of the surface shows inner Cu₂O nanoparticles are encapsulated by an amorphous Cu_xO layer, where Cu(0) nanoclusters of ~ 5 nm in size are well dispersed (Figs. 1c2, c3 and S4). The elemental mapping analysis showed an atomic Cu:O ratio of 77:23 on the surface (Table S2), indicating a small amount of O atoms were extracted to form a passivating amorphous Cu_xO layer. The SAED pattern of the core–shell structure shows diffraction patterns of both Cu(0) and Cu₂O nanoclusters (inset in Fig. 1c2). The Cu(0) nanocluster can be distinguished by the clear lattice spacing of 2.088 Å, corresponding to (111) planes of metallic Cu. In contrast, only a rough surface was formed in the Ar atmosphere (CO₂-deficient) (Figs. 1d and S5-S6). The thickness is over 40 nm in the 2000s pretreatment. HRTEM images and SAED rings show that the surface is almost made of metallic Cu(0) nanoclusters, with the absence of any amorphous Cu_xO matrix. The lattice spaces of 1.808 Å correspond to the crystal plane of (200) of metallic Cu (inset in Fig. 1d3.). The elemental mapping analysis shows an Cu:O atomic ratio of 94:6 on the surface (Table S3), suggesting a high degree of transition from Cu₂O to Cu on the surface under the Ar atmosphere. The distinct microstructures and compositions offer strong evidence of the feeding gas-dependent structure evolution and phase/composition transition.

XRD analysis over a mass of nanocrystals confirmed the partial phase transition from Cu₂O to metallic Cu(0) in both cases (Fig. 1e). Moreover, even though the interiors of Cu₂O were fragmented in the CO₂ atmosphere, more Cu₂O phase retained in Cu₂O-CO₂ than in Cu₂O-Ar (as suggested by the higher peak intensity of Cu₂O phase in the former one). This indicates the amorphous layer might act as a passivation layer to protect inner Cu₂O nanoparticles from electrochemical reduction. Cu LMM Auger spectra were further collected from Cu₂O-CO₂ and Cu₂O-Ar in the increasing period of pre-electrolysis (from 500, 1000, to 1500 s, Figs. 1f, g and S7), to trace the reducing process on surface. The spectra were deconvoluted to peaks of Cu(0), Cu(I), and Cu(II). The Cu(II) was attributed to the oxidation in air. The peak ratio of Cu(0)/Cu(I) increases with prolonged pre-electrolysis, suggesting a reducing process in both Cu₂O-CO₂ and Cu₂O-Ar under the applied negative potential. Moreover, Cu₂O-CO₂ exhibits a much lower Cu(0)/Cu(I) ratio (0.36:1 for the 1500 s) than Cu₂O-Ar (0.90:1 for the 1500 s), validating the high resistance of the amorphous layer to the reducing potential in Cu₂O-CO₂. The results are consistent with the abundant Cu(I) and a few tiny Cu nanoclusters in the amorphous layer.

To probe the local valence state of copper in Cu₂O-CO₂ and Cu₂O-Ar, we collected the electron energy-loss spectroscopy (EELS) on the surface (Fig. 2). The spectra were recorded in the rectangle regions in Fig. 2a, d. To reveal the spatial distribution, the spectra were collected from the inner part to the utmost surface part, along the lines marked in their HETEM images in Fig. 2b, e. Region 1# and 2# in Fig. 2b were the inner crystalline Cu₂O that was not reduced yet in Cu₂O-CO₂. They exhibit peaks located at 935.9 eV (Cu-L₃) and 955.7 eV (Cu-L₂), in line with the reference Cu₂O sample. Region 3# and 4# were in the amorphous regions. The EELS peaks narrowed and shifted to 935.6 eV (Cu-L₃) and 955.4 eV (Cu-L₂). The peaks were between that of reference Cu(I) and Cu (0), indicating an in-between valence state (forming Cu_xO, 0 < x < 1). The peak intensities were also weakened, resulting from the extraction of oxygen from Cu₂O. Region 5# and 6# were Cu(0) clusters that were well dispersed in the amorphous layer. Their spectra exhibited wide and inconspicuous white lines, consistent with the metallic nature, where electrons were filled within the 3d band and no other oxygen hybridization contributed to white lines [46–49]. The peak positions at 934.5 and 954.3 eV also agree well with that of reference Cu. Differently, Cu₂O-Ar showed only metallic Cu components (Fig. 2f), in good agreement with the results in Figs. 1 and S4–S9.

Fig. 2.

Valence state of copper on the surface of Cu₂O-CO₂ and Cu₂O-Ar. a, d HAADF images and b, e HRTEM images of a, b Cu₂O-CO₂ and d, e Cu₂O-Ar. The points for EELS were marked in the HRTEM images in b and e; EELS L_2,3 edge spectra of c Cu₂O-CO₂ and f Cu₂O-Ar collected from the rectangle region marked in b and f, from the inner part to the utmost surface part along the lines marked in b and e

Feeding Gas-Dependent Electrocatalytic Performances Toward CO₂RR of Cu₂O Nanocubes

The electrochemical CO₂RR performances, as well as the products (Fig. S10), were then investigated to probe the catalytic response to feeding-gas-dependent reconstruction in an H-cell configuration (Fig. S11). The potential of −1.1 V versus RHE was maintained to drive CO₂RR electrolysis for more than 30 h. The current densities were recorded, and the products were traced in specific time intervals (Fig. 3a, b). After initial activation in the CO₂ atmosphere, the main product of Cu₂O-CO₂ is ethylene (C₂H₄) (Figs. 3a and S12). The detailed gas concentration sampled at each time point is provided in Fig. S13. The partial current density reaches the peak value of −24.1 mA cm⁻² at a faradic efficiency of 38% for C₂H₄, which is consistent with previously reported data (Table S4) [50, 51]. In sharp contrast, Cu₂O-Ar can deliver a partial current density of 8 times less (−3.0 mA cm⁻²) with the faradic efficiency reduced by half (16%) (Figs. 3b and S14). The sampling information in detail is shown in Fig. S15. The higher FE for C₂H₄ could be attributed to the Cu(I) sites that are stabilized in the amorphous layer of Cu₂O-CO₂. This is in line with previous reports showing that Cu(I) sites are favorable for C–C coupling reactions [52]. Moreover, we observed that it took a long time for Cu₂O-Ar (10,400 s) to reach the peak partial current density and Faradaic efficiency. This indicates CO₂ would penetrate slowly into the interiors of Cu₂O nanoparticles and drive some CO₂-rich reconstruction for improving C₂H₄ selectivity in Cu₂O-Ar. In contrast, Cu₂O-CO₂ reaches saturated current density and Faradaic efficiency at a rate of 5 times faster. No further CO₂-rich reconstruction could occur in a couple of hours, likely due to the passivating effect of the amorphous layer. The current density much higher than Cu₂O-Ar should be ascribed to the highly porous structures and the well-dispersion of the catalytic active amorphous layer through an in situ conversion process.

Fig. 3.

Electrochemical CO₂RR performances of Cu₂O nanocubes in an H-cell. Partial current density and Faradaic efficiency of the main CO₂RR products after activation in a CO₂ and b Ar atmosphere, respectively. The applied potential is −1.1 V versus RHE for electrolysis in CO₂-saturated 0.5 M KHCO₃ electrolyte; Total current density and Faradaic efficiency of the main CO₂RR products plotting against applied potentials after activation in c CO₂ and d Ar atmosphere, respectively. Red hollow squares are C₂H₄ partial current densities at each sampling time

To obtain more information on feeding gas-dependent catalytic performances, catalysts were analyzed in a wide potential range from −0.6 to −1.2 V versus RHE (Fig. 3c, d). For both catalysts, the C₁ compound (CO) is the dominating product below the potential of −1.0 V versus RHE, whereas C₂₊ compounds (C₂H₄) increase apparently above the potential. At each potential, Cu₂O-CO₂ exhibits better selectivity to CO₂RR, and the current density is at least ~ 2 times higher than Cu₂O-Ar. The FE trends for CO and C₂H₄ are similar in the whole potential range. Moreover, once C₂H₄ is produced at an appreciable rate, the CO product is much constrained. This indicates that CO intermediate could be the precursor of C₂H₄. Also, it might play a pivotal role in forming the passivating amorphous layer during the reconstruction process [53].

The feeding gas-dependent reconstruction and catalytic performances were further probed in a flow cell with 1 M aqueous KOH (Fig. S16), aiming to assess the feasibility in catalytic environments close to practical application. Cu₂O nanoparticles were loaded on gas diffusion electrodes (GDE, 1 cm² with catalyst loading of 1 mg cm⁻²) to facilitate gas transportation for improved CO₂ accessibility. Prior to electrochemical characterization, Cu₂O nanocubes were activated at the potential of −0.6 V vs. RHE. As shown in Fig. 4a, b, the onset potential for CO₂RR was negatively shifted by 200 mV and the maximum current density was increased by 2.5 times in comparison with those in H-cell for both catalysts (Fig. 3a–d). The apparent improvement in catalytic performances is ascribed to the change in cell configuration and electrolyte (from KHCO₃ to KOH) [54]. More importantly, the feed gas-dependent catalytic performances are translated from H-cell to flow cell. In detail, Cu₂O-CO₂ exhibits a much better selectivity to CO₂RR than Cu₂O-Ar in the whole testing potential range. Once C₂H₄ is produced at an appreciable rate, the CO product becomes much more constrained. Akin to in H-cell, this indicates that CO intermediate (acting as the precursors of C₂H₄) plays a pivotal role in forming the passivating amorphous layer in the reconstruction process. At the optimal potential of −0.6 V, Cu₂O-CO₂ manifests an FE of 71% to C₂H₄ at the partial current density of −147.3 mA cm⁻², whereas Cu₂O-Ar shows an FE of only 17% at a much less partial current density of −10.5 mA cm⁻². Moreover, we find that the selectivity to CO₂RR is inferior at a lower loading of 0.5 mg cm⁻² (Fig. 4c). For instance, the faradaic efficiency of C₂H₄ decreased by 24% at −0.6 V vs RHE. The inferior selectivity can be well rationalized by the *CO intermediate modulated reconstruction as proposed above. Catalysts with a decreasing loading lower the local intermediate concentration of *CO, thereby resulting in a reconstruction processed as in a CO₂-deficient atmosphere (Ar) and leading to the production of HER active sites to inferior their selectivity toward C₂H₄. Similar catalyst density-dependent selectivity has been observed before, though it is attributed to the decreasing re-absorption of CO intermediates for CO₂RR rather than for reconstruction [55, 56]. The result strongly supports our conclusion that the absorption of CO intermediate is crucial for reconstruction toward the amorphous layer that is favorable for the C–C bond coupling reaction. We further examined the influence of cell configuration and electrolytes on reconstruction. We conducted the pre-electrolysis and CO₂RR electrolysis in a flow cell, using the same electrolyte as in the H-cell (0.5 M KHCO₃). The resultant catalyst shows FE of C₂H₄ similar to that of Cu₂O-CO₂ in H-cell, ruling out the contribution to FE from cell configuration. It is also noteworthy that the difference in current density and potential should be ascribed to the cell configuration [54]. Following this, we did the pre-electrolysis in 0.5 M KHCO₃ and then tested the CO₂RR electrolysis in KOH electrolytes. The catalytic performances of the resultant catalyst are suppressed to some extent over that of Cu₂O-CO₂, but still much higher than that of Cu₂O-Ar (Fig. S17). The result again confirms the dominating role of feeding gas rather than electrolytes in reconstruction.

Fig. 4.

Electrochemical CO₂RR performances and phase information of Cu₂O nanocubes in a flow cell. Total current density and Faradaic efficiency of the main CO₂RR products after activation in a CO₂ and b Ar; c Faradaic efficiency of the main CO₂RR products after activation with different loading; d XRD peak intensity ratio of Cu (111) to Cu₂O (200), inset is typical XRD patterns after activation at different potentials of Cu₂O-CO₂ and Cu₂O-Ar in a flow cell

To further evidence the passivation behavior, we investigated the potential dependent phase transition. XRD patterns of Cu₂O-CO₂ and Cu₂O-Ar were recorded after being activated in the potential range of −0.4 ~ −0.7 V versus RHE with a decreasing interval of 0.1 or 0.05 V in Figs. 4d and S18. The peak intensity ratio of Cu-(111)/Cu₂O-(200) is used to measure the amount of residual Cu₂O. Besides, the typical XRD patterns of catalysts after pre-electrolysis at −0.4 and −0.7 V versus RHE were exhibited in the inset of Fig. 4d. The intensity ratio plotting against applied potentials shows that much more Cu₂O was retained in Cu₂O-CO₂ than in Cu₂O-Ar at potential higher than −0.65 V versus RHE. At the potential range of −0.4 to −0.6 V, the amount of metallic Cu is 7%–36% less in Cu₂O-CO₂. It suggests a much higher resistance of Cu₂O-CO₂ to reducing potential, agreeing well with the formation of a passivating amorphous layer under a CO₂ atmosphere. When potentials are more negative than −0.6 V versus RHE, they exhibit similar conversion rates while distinct catalytic activity and selectivity. To reveal the underlying reason, we characterized the microstructures of Cu₂O-CO₂ and Cu₂O-Ar after activating at the potential of −0.7 V (Fig. S18c-h). Cu₂O-Ar retained the compact cubic shape, with Cu coated on inner Cu₂O. In contrast, Cu₂O-CO₂ increased the transition of the amorphous part to Cu(0), leading to a quick increase in I_Cu(111)/I_Cu2O(200). Importantly, we could still observe some residue amorphous part in Cu₂O-CO₂, which contributes to the higher catalytic performances (FE and current density) than that of Cu₂O-Ar at the potential of −0.7 eV.

Though the feeding gas played a pivotal role in leading the selective CO₂RR, we noted that it did not guarantee stable electrolysis for a period over 10 h in H-cell. To find the reason for the decay of performances, we trace the microstructure on the surface of Cu₂O-CO₂ in 3, 10, 20, and 30 h’ electrolysis (Fig. S9). (HR)TEM images reveal the progressive transition of the amorphous layer to Cu(0) nanoclusters. The particle size was getting larger with the prolonged electrolysis (Fig. S9c to S9f to S9i to S9l). Eventually, the surface transformed into a metallic state like Cu₂O-Ar, exhibiting low catalytic activity in producing C₂H₄. This structure-performance correlation in the durability test underscores the essential role of the amorphous layer for selective CO₂RR electrolysis.

Mechanism of Feeding-Gas-Mediated Structural Evolution

The microstructural and electrochemical analysis above has shown the feeding gas plays a pivotal role in mediating the structure reconstruction and based on this we link the final structures to catalytic performances by identifying the catalytic active amorphous layer favorable for C–C bond coupling reaction. Given that *CO is regarded as an important intermediate for CO₂RR and acts as the precursor for C₂H₄, we propose that *CO intermediate produced in the CO₂ activation process could play a determining role in reconstruction, as well as the consequent high selectivity to C₂₊.

To clarify the critical role of *CO intermediate, we performed in situ Raman analysis to directly probe the intermediate in the self-adaptive reconstruction process (Figs. 5a–c and S19-S22). The Raman spectra of Cu₂O-CO₂ were recorded every 5 min under electrolysis conditions in a CO₂-saturated 0.5 M KHCO₃ electrolyte. By contrast, the Cu₂O-Ar catalyst was pretreated in an Ar atmosphere every 5 min and then transferred back to a CO₂-saturated electrolyte to collect Raman spectra under electrolysis conditions. The peaks at 1026, 1353, and 1548 cm⁻¹ are ascribed to the monodentate *CO₃²⁻ (ν₁CO₃²⁻ of η₁-CO₃²⁻), bicarbonate, and bidentate carbonate (ν_asCO₂⁻ of η₂(C,O)-CO₂⁻), respectively (Figs. S19 and S20) [57]. Cu₂O exhibits characteristic peaks at 142, 215, 405, 529, and 623 cm⁻¹, corresponding to the 2Γ^–₁₂, 4Γ^–₁₂, Γ⁺₂₅, and Γ^–₁₂ plus Γ⁺₂₅ phonon modes. The peaks were weakened in prolonged pretreatment, in line with the transition from Cu₂O to Cu(0) as revealed above [58]. To be noted, the transition is much slower in Cu₂O-CO₂. For instance, Cu₂O is discernable even in 30 min of electrolysis for Cu₂O-CO₂, whereas it became insignificant in 10 min activation for Cu₂O-Ar (Figs. 5a, b and S21). This observation is consistent with the passivation effect of the amorphous layer in protecting inner crystalline Cu₂O nanoparticles from reduction. More interestingly, we observed an intermediate of *CO at a wavelength of around 2050 cm⁻¹ (corresponding to νCO bands of *CO on atop sites) in only 5 min electrolysis in Cu₂O-CO₂, whereas it took more than 20 min to probe a distinguishable amount of *CO in Cu₂O-Ar (Fig. 5a, b). This result validates the essential role of intermediate of *CO played in the formation of a catalytic active amorphous layer. Moreover, a much higher signal of *CO intermediate is present in Cu₂O-CO₂, corresponding well with the larger amount of Cu(I) in the amorphous layer.

Fig. 5.

Mechanism of feeding-gas-mediated structural evolution. In situ Raman spectra evolution in a time scale with an interval of 5 min for a Cu₂O-CO₂ and b Cu₂O-Ar in the pre-electrolysis process for activation. The applied potential is −1.1 V versus RHE; c In situ Raman spectra evolution with decreasing potential from OCP to −1.2 V versus RHE. The typical peaks of Cu₂O are labeled with pink rhombus. The *CO intermediate is highlighted with an orange background. d Partial current density and Faradaic efficiency of the main CO₂RR products after activation in O₂. Red hollow squares are C₂H₄ partial current densities at each sampling time. e The reaction energy diagram for CO₂RR to C₂H₄ on Cu (110) slab, Cu₂O(110) slab, and Cu/Cu₂O interface. (Color figure online)

To confirm the role of the intermediate of *CO in reconstruction, the driving force was modulated by varying the applied potential in the range from OCP to –1.2 V versus RHE with an interval of 0.2 V. The Raman spectra were recorded at each potential to analyze the phase and intermediate transition (Fig. 5c). The increasing driving force pushes forward the phase transition at more negative potentials (Fig. S22). Notably, the transition degree is slower in Cu₂O-CO₂ than in Cu₂O-Ar, in line with the passivation effect of the amorphous layer as revealed above. Moreover, Cu₂O-CO₂ started to showcase the signal of *CO intermediate at a potential of −0.6 V versus RHE, which is around 0.2 V more positive than Cu₂O-Ar. Meanwhile, the signals were more evident in Cu₂O-CO₂ over in Cu₂O-Ar at each potential. The higher onset potential and higher intensity of the Raman signal indicate that *CO intermediate is responsible for the formation of Cu(I)-rich amorphous layer, which in turn contributes to the better selectivity for value-added C₂H₄.

To validate the role of *CO intermediate, we changed the feeding gas to pure CO for activation in the pre-electrolysis (Fig. S23a-d). The product (denoted as Cu₂O-CO) shares the same catalytic performances (total current density and Faradaic efficiency) as that of Cu₂O-CO₂ (Fig. S23a-d), confirming the essential role of *CO intermediate in producing an amorphous layer. This is further validated by the similar FE of C₂H₄ in electrocatalytic CORR (Fig. S23e, f). We further changed the feeding gas to a mixture of CO/CO₂ (volume ratio: 1:1). It again exhibits the same catalytic performance. These data suggest a few CO are enough to drive the transition from crystalline Cu₂O to the amorphous Cu_xO layer. The in situ formed *CO could be favorable, as CO is released near the active sites and results in a higher local concentration than the feeding CO externally.

To confirm the mediating process of reaction intermediate in structural evolution, we shift the gas from CO₂ to O₂, because oxygen-related intermediates (for instance, *OH, *O, and/or *OOH) could absorb on Cu-based catalysts in electrocatalytic oxygen reduction reaction (ORR) or cocatalysis of ORR and CO₂RR [59, 60]. Akin to the above protocol in Ar/CO₂, Cu₂O Cubes were imposed to a potential of −1.1 V versus RHE in pure O₂ atmosphere for 2000 s for reconstruction, and then subjected to electrocatalytic CO₂RR. It turned out that the resulting catalysts exhibited in-between total current densities and partial current densities and FE of C₂H₄ (−59.6 and −15.4 mA cm⁻², and 26%, Figs. 5d and S24). The potential dependent catalytic activity and selectivity were further evaluated for Cu₂O-O₂ (Fig. S25). The trend is the same as that of Cu₂O-Ar and Cu₂O-CO₂. The FE of C₂H₄ increased at the potential where the FE of CO decreased, and the more negative of potential, the larger the FE ratio of C₂H₄ to CO. This result supports our conclusion that the reaction intermediate mediates the structural evolution and selectivity. Also, there are some inconsistencies between Cu₂O-O₂ and Cu₂O-CO₂/Cu₂O-Ar. Firstly, Cu₂O-O₂ seems more stable than Cu₂O-CO₂, despite the lower catalytic activity (Figs. S24 and S26). The better durability should be due to the production of extra CO, which could supply abundant *CO intermediates to stabilize Cu(I) active sites for the C–C bond coupling reaction. This attribution agrees well with the fact that weak durability corresponds to the depletion of CO product at the potential applied for durability test. Secondly, the optimal FE of C₂H₄ shifts to a more positive potential of −1.0 V versus RHE (Fig. S25). The change could be due to the distinct absorption of intermediates and the different surface electronic structures, which are under further investigation.

Furthermore, we examined the effect of pre-electrolysis in mixed CO₂/O₂ with a volume ratio varying from 3:1, 1:1, to 1:3 (Fig. S27). When the ratio is equal to or larger than 1:1, the electrocatalysts behave like Cu₂O-CO₂, showing higher FE and partial current density in producing C₂H₄. When it is less than 1:1, the electrocatalysts behave like Cu₂O-O₂, exhibiting suppressed FE but improved durability (Fig. S27). This is in line with the stronger absorption of *CO than oxygen-related intermediates (*OH, *O, and/or *OOH), further confirming the essential role of the intermediate of *CO in the resulting catalyst surface favorable for selective C₂H₄ production. Previous studies have also shown that *CO intermediate has a strong absorption ability to Cu-based materials, and O-related intermediates have a moderate absorption ability, whereas Ar is relatively inert to absorption on any substrates. The absorption ability of gas molecules and/or the corresponding intermediates correlates well with the surface reconstruction and the resulting catalytic performances, supporting the critical role of the reaction intermediate in structural reconstruction. More importantly, our finding could imply a new dimension to activate precatalysts for CO₂RR that is to control the components of feeding gases for activation. Indeed, previous work has shown the unique selectivity in co-feeding gases of O₂ and CO₂ [59]. Though they rationalized it by the concept of co-electrolysis involving intermediates of *OH, the feeding-gas resultant reconstruction cannot be ruled out.

We performed density functional (DFT) theory calculations to further specify the essential role of *CO in the resulting catalyst surface that is selective for C₂H₄ production. Since the reconstruction is along with CO₂RR, we calculated the change of Gibbs energy for intermediates adsorption on Cu(110) slab, Cu₂O(110) slab, and the interface of Cu(110)/Cu₂O(110) (Figs. 5e and S28-S31). Cu(110) and Cu₂O(110) share the same rate-determining step (RDS) of *CO + *CO → *CO + *COH, whereas the Cu(110)/Cu₂O(110) has RDS of CO₂ activation (* + CO₂ → *CO₂). The energy barrier in RDS is only 0.69 eV for the Cu(110)/Cu₂O(110) interface, significantly lower than that for Cu(110) (1.24 eV) and Cu₂O (110) (1.82 eV). This is consistent with the outstanding CO₂RR performance of Cu₂O-CO₂, possessing a large amount of Cu/Cu₂O interface. The high energy barrier of Cu(110) and Cu₂O(110) indicates the surface is mostly covered by *CO, in line with our in situ Raman spectra. Furthermore, we calculated the absorption enthalpy of *CO. Cu₂O(110) exhibits the lowest absorption enthalpy of −1.09 eV. The low absorption enthalpy of Cu₂O(110) is in line with the strong and stable absorption of *CO in Cu₂O-CO₂. Once Cu₂O was reduced to Cu(0), the absorption enthalpy increased to −0.13 eV on surfaces and to −0.20 eV on Cu(110)/Cu₂O(110) interfaces. We noted that the absorption enthalpy of Cu(110)/Cu₂O(110) interfaces is lower than Cu(0), suggesting that Cu(110)/Cu₂O(110) interfaces mitigate the further reduction of the catalysts to metallic copper.

In the previous literature, Cu₂O has been well investigated, given the superb catalytic performances in producing C₂H₄ [40, 42, 43]. The structural evolution was revealed and the driving force was ascribed to the applied negative potential solely. Though CO₂ has been found to work for modulating the crystal growth, cation leaching, and morphology evolution of copper [38–41], it has rarely been reported to regulate the reconstruction of Cu₂O. Different from those focusing on the catalyst itself, this work took the gas accessibility into account. The new findings provide additional information to elucidate previous controversial results and rationalize the debates on catalytic mechanisms [44, 45]. This work also highlights the importance of the microstructure design of catalysts. The modulation of microenvironments including gas accessibility, local pH, and triple-phase interface not only promotes catalytic reaction as most works suggested before [36, 37] but also guides the catalyst reconstruction toward favorable active structures as found here.

Knowledge Guiding Design of Selective and Durable CO₂RR Catalysts for Value-Added Products of C₂H₄

Besides catalytic activity and selectivity, it appears that the dynamic reconstruction also affects the durability. The partial current density for C₂H₄ started to decay for Cu₂O-CO₂ once it achieved the highest FE and partial current density of C₂H₄ in both an H-cell and a flow cell (Figs. 3a and S32). The FE of C₂H₄ decreases by 35.6% in 31 h in the H-cell and 28.5% in 18 h in the flow cell (Fig. S32). The above analysis has shown that the absorption of *CO intermediate is crucial for the formation and stability of the catalytic active amorphous layer. In particular, it stabilizes Cu(I) sites favorable for C–C bond coupling reaction. Given that *CO intermediate is the precursor of C₂H₄ and rare CO product was released at the potential for stability assessment (as revealed in Figs. 3a, b and 5d), we suggest that the decay could be attributed to the fact that active Cu(I) sites were destabilized by the depletion of *CO intermediates for the production of C₂H₄. This is supported by the much better durability of Cu₂O-O₂, which exhibited a CO FE of up to 12% (Fig. 5d). To further validate it, we excavated Cu₂O nanoparticles to form hollow nanospheres (Figs. S33 and S34), aiming to increase the accessibility of CO₂ to produce more CO intermediates and confine them inside to maintain a CO-rich environment to stabilize Cu(I). Activation in a CO₂-sufficient atmosphere did not change the hollow morphology but resulted in amorphous layers encapsulating on crystalline Cu₂O nanoparticles (Fig. 6a–d). Analogous to Cu₂O-CO₂, well-dispersed tiny Cu(0) nanoparticles of 10 nm in size were present in the amorphous layer. The catalytic performances were then assessed in both H-cell (in the range from −0.6 to −1.2 V versus RHE, Fig. 6e) and flow-cell (in the range from −0.4 to −0.7 V versus RHE, Fig. 6f). Cu₂O hollow nanospheres exhibited the similar catalytic performances (including current densities and FE of various products) as Cu₂O-CO₂. The significant difference between them is that more CO products were released for hollow nanospheres at the optimal potential (the one for the stability test). The FE of CO is 14.5% in H-cell and 4.8% in flow-cell for Cu₂O hollow nanospheres, 2–6 times higher than that for Cu₂O-CO₂ (2.4%). As expected, Cu₂O hollow nanospheres exhibited improved durability than Cu₂O-CO₂. In a H-cell, Cu₂O hollow nanospheres maintained a C₂H₄ partial current density of around −26 mA cm⁻² with a FE of 47% in > 30 h electrolysis, whereas Cu₂O-CO₂ lost the FE by 15.3% (Fig. 6g) and partial current density by 9% (Fig. 6h). In a flow cell, the durability was measured at a practical current density of 200 mA cm⁻² (Figs. 6i and S35-S36). The FE of Cu₂O hollow nanospheres decreased at a decay rate of 0.53% h⁻¹ in 38 h of electrolysis, one-third of that of Cu₂O-CO₂ (1.58% h⁻¹ in 18 h of electrolysis). The superb activity, selectivity, and durability outperform Cu-based electrocatalysts reported previously (Table S4). More interestingly, we find that durability described by the decay rate correlates with the FE ratio of CO/C₂H₄ (Figs. 6j and S37-S38). It shows that a larger proportion of CO corresponds to better stability, agreeing well with the role of CO intermediates in stabilizing Cu(I) for a favorable C–C bond coupling reaction. Our finding also rationalizes the durable and selective production of C₂H₄ for low-dimensional Cu₂O catalysts [27].

Fig. 6.

Microstructural evolution and electrochemical CO₂RR performances of Cu₂O hollow spheres in an H-cell and a flow cell. a TEM images and b-d HRTEM images of Cu₂O hollow spheres after activation in CO₂ atmosphere for 2000 s. The dashed line in b is the border between crystalline Cu₂O and amorphous Cu_xO. The c is the enlarged region of ① in b and the d is the enlarged region of ② in b; e Total current density and Faradaic efficiency of the main CO₂RR products plotting against applied potentials after activation in an H-cell; f Total current density and Faradaic efficiency of the main CO₂RR products plotting against applied potentials after activation in a flow-cell; g Faradaic efficiency and h partial current density of the main CO₂RR products of Cu₂O hollow spheres and solid Cu₂O nanocubes after activation in CO₂. The applied potential is −1.1 V versus RHE for electrolysis in an H-cell; i Stability of hollow Cu₂O-CO₂, Cu₂O-CO_2, and Cu₂O-Ar at the current density of −200 mA cm⁻² in a flow-cell; j Correlation between FE decay rate and FE_CO/FE_C2H4 ratio during the stability tests

Conclusion

In conclusion, our work uncovered the essential role of feeding gas (CO₂ accessibility) on structural reconstruction and further elucidated how it affected the catalytic performances, by recognizing a CO₂-induced passivation process in structural reconstruction and a self-stabilizing process in catalytic production of C₂H₄. The CO₂-rich atmosphere led to a catalyst surface favorable for CO₂RR, whereas the CO₂-deficient one (Ar) preferred that for hydrogen evolution reaction (HER), exhibiting a ~ 4 times difference in ethylene (C₂H₄) Faradaic efficiency (FE) and ~ 8 times difference in current densities. The *CO intermediates played a pivotal role in stabilizing Cu(I) sites, leading to the formation of a reduction-resistant but catalytic active amorphous layer in the reconstruction process. Furthermore, we found extra CO production was indispensable for the robust production of C₂H₄. An inverse correlation between durability and FE_CO/FE_C2H4 was disclosed. We attributed it to the self-stabilization of Cu(I) sites in the CO atmosphere. Taking advantage of this knowledge, we fabricated hollow Cu₂O nanospheres and demonstrated durable electrolysis for over 48 h at a current density of −200 mA cm⁻² in producing C₂H₄ with an FE of up to 61% at −0.6 V_RHE in a flow cell. Our work recognizes the previously overlooked passivation reconstruction and self-stabilizing behavior and highlights the critical role of the local atmosphere in modulating the reconstruction behavior and catalytic process, serving as an important supplement in understanding and designing high-performance CO₂RR (pre)catalysts. A hybrid catalyst is therefore suggested, with one component producing *CO to increase the local CO concentration for favorable reconstruction and durable catalysis. Additionally, the strong link between atmosphere and structural reconstruction has an implication for the design of precatalysts and the revival of degraded catalysts.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

F. S. conceived the idea and led the project. C. Z. prepared the samples, did the structural characterization, and tested the electrochemical activity, with the assistance of Y. G., Q. J., Z. S., R. F., S. W., H. Z., Q. X. and Z. Y. Q. J. and W. S. assisted to perform the in situ Raman analysis and helped to analyze the Raman spectra. F. S. and C. Z. contributed to the data analysis. F.S. and C. Z. wrote the paper, with input from all the other authors.

Declarations

Conflict of Interest

The authors declare no interest conflict. They have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.