The degradation process and mechanism of CuO_x photoelectrodes were investigated using several in situ characterization methods. Scanning electron microscope (SEM) images showed that the pristine Cu base surface was smooth and clean, free of any impurities (Fig. 2A and Fig. S2). By contrast, the CuO_x surface became rough due to the formation of Cu₂O/CuO crystal (Fig. 2B). Irregularly shaped Cu₂O/CuO particles, with an average size of ~500 nm, were evenly embedded on the Cu base surface (Fig. 2C and Fig. S3). When CuO_x was subjected to 15 h of long-term photoelectrocatalytic CO₂ reduction operation, the photocurrent decayed severely and exhibited 3 distinct stages of change (Fig. 2D). In the first 2 h, the photocurrent dropped from ~−3.2 to ~−4.0 mA/cm² (Stage 1). Then, the photocurrent reversed and rose to ~−2.1 mA/cm² after the intermediate 10 h (Stage 2) and finally stabilized at ~−1.7 mA/cm² after the last 3 h (Stage 3). Meanwhile, copper ions continuously leached from CuO_x during the operation, and the leaching rate was slow at first, became fast, and went slow again, which corresponded to the photocurrent changes (Fig. S4). The cumulative amount of copper ions leached reached up to ~212 μg/L at the end of the reaction (Fig. 2E).

The fresh CuO_x was black in appearance but turned rufous after use, indicating an obvious change in morphology and structure (Fig. S5). SEM images showed that the used CuO_x surface became relatively flat, and the Cu₂O/CuO crystal particles became obscured in contour and amorphous, confirming the degradation of morphology and structure (Fig. 2F and G and Fig. S6). In situ Raman spectra indicated that the peaks of CuO declined heavily in Stage 1 and almost disappeared by the end (Fig. 2H). In contrast, the peaks of Cu₂O were markedly enhanced in Stage 1, and then gradually weakened until they became blurred. The results demonstrated the successive degradation of CuO and Cu₂O. X-ray diffraction (XRD) patterns further proved that the CuO phase had completely degraded in used CuO_x (Fig. 2I). Based on the x-ray photoelectron spectroscopy (XPS) of the Cu 2p spectrum of used CuO_x, the peaks of Cu²⁺ disappeared, while the peaks of Cu⁺ were markedly enhanced (Fig. 2J). Meanwhile, the characteristic peaks of Cu⁰ appeared in used CuO_x. XPS O 1s spectra showed that the peak corresponding to Cu–O bonds was severely reduced in used CuO_x (Fig. S7). However, a new peak corresponding to Cu²⁺–OH species appeared [4].

The above analyses revealed that most of the CuO phase in CuO_x was reduced to Cu₂O phase by photoelectrons in Stage 1, leading to a remarkable increase in the Cu₂O proportion (Fig. S8). The intrinsic activity of Cu₂O was better than that of CuO; thus, the absolute value of photocurrent increased slightly. This phase transition released a small amount of copper ions due to the loss of half of lattice oxygen. In Stage 2, most of Cu₂O sharply decayed to Cu, resulting in rapid photocurrent attenuation. This phase transition released a large amount of copper ions due to the loss of almost all lattice oxygen. In Stage 3, the remaining CuO and Cu₂O further degraded, resulting in continuous but gradual photocurrent reduction and copper ion release. The degradation process can be summarized as: Cu₂O/CuO → Cu₂O → Cu. Such electron-mediated severe photocorrosion led to an obvious decrease in the light capture performance and photoelectric conversion efficiency of used CuO_x (Fig. 2K and Fig. S9). Although the formation of Cu reduced the impedance and increased conductivity, the disappearance of active components resulted in disastrous photocurrent decay (Fig. 2L and Fig. S10).

Next, we investigated the durability of CuO_x@C photoelectrodes using the same methods. SEM images showed that CuO_x was fully encapsulated in a carbon layer (Fig. 3A and B and Fig. S11). Transmission electron microscope (TEM) images illustrated that coating was extremely uniform, with an average thickness of ~6 nm, confirming that the carbon layer was formed by the self-assembly of single-row carbon nanodots (Fig. 3C and Figs. S12 and S13). When CuO_x@C was subjected to 15 h of long-term photoelectrocatalytic CO₂ reduction operation, the photocurrent remained almost constant (Fig. 3D). The absolute value of the photocurrent only decreased slightly from the initial ~4.8 mA/cm² to the final ~4.7 mA/cm². It is worth noting that the carbon layer encapsulation increased the onset photocurrent by ~30%. Additionally, there was extremely low copper ion leaching from CuO_x@C during the long-term operation, and the amount of copper ions leached did not exceed ~20 μg/L at the end of reaction (Fig. 3E and Fig. S14).

CuO_x@C was black in appearance, and its color hardly changed after use, indicating negligible changes in morphology and structure (Fig. S15). SEM and TEM images showed that used CuO_x@C surface was almost identical to fresh CuO_x@C, with the carbon layer (thickness ~6 nm) uniformly covering the CuO_x surface (Fig. 3F and G and Figs. S16 and S17). The atomic ratios of Cu, O, and C in CuO_x@C remained unchanged before and after use, further confirming the structural stability. In situ Raman spectra indicated that the peaks of CuO and Cu₂O in CuO_x@C remained unchanged throughout the long-term operation (Fig. 3H and Fig. S18). XRD patterns further proved that neither CuO phase nor Cu₂O phase degraded during use (Fig. 3I). In the XPS spectra of Cu 2p, O 1s, and C 1s for CuO_x@C, the positions and intensities of all peaks remained largely unchanged before and after use (Fig. 3J and Figs. S19 and S20).

The above analyses revealed that the carbon layer encapsulation effectively mitigated CuO_x degradation, endowing CuO_x@C with gratifying durability. The CuO phase and Cu₂O phase remained stable throughout all stages of the long-term operation, resulting in a steady photocurrent with only slight attenuation (Fig. S21). Meanwhile, there was extremely low copper ion release due to the absence of obvious phase transition and lattice oxygen loss. This outstanding stability resulted in negligible changes in the light capture performance, photoelectric conversion efficiency, and electrochemical impedance of used CuO_x@C (Fig. 3K and Figs. S22 and S23). Linear sweep voltammetry (LSV) curves of CuO_x exhibited 2 reduction peaks, corresponding to the processes of Cu²⁺ → Cu¹⁺ and Cu¹⁺ → Cu⁰, respectively. In contrast, these 2 peaks were almost undetectable in the LSV curves of CuO_x@C, further demonstrating its satisfactory stability (Fig. 3L).

The carbon layer acted as an electron reservoir and effectively extracted photoelectrons from CuO_x, resulting in spatial isolation and thus inhibiting photocorrosion (Fig. 4A and B). To accurately unveil the interfacial charge migration mechanism, the energy band structures of CuO_x, carbon layer, and CuO_x@C were systematically investigated (Fig. S24). Diffuse reflection spectroscopy (DRS) spectra revealed distinct optical absorption edges at ~605 nm for CuO_x and ~476 nm for carbon layer [30,31]. Tauc plots derived from the Kubelka–Munk transformation yielded an optical band gap (E_g) of ~2.06 eV for CuO_x and a molecular energy gap (ΔE) of ~2.60 eV for the carbon layer. Ultraviolet photoelectron spectroscopy (UPS) spectra positioned the valence band (VB) of CuO_x at ~−5.71 eV and the highest occupied molecular orbital (HOMO) of the carbon layer at ~−6.64 eV. The conduction band (CB) of CuO_x was derived as ~−3.65 eV (E_g + VB), and the lowest unoccupied molecular orbital (LUMO) of the carbon layer was found at ~−4.04 eV (ΔE + HOMO). These energy levels were calibrated relative to the reversible hydrogen electrode (RHE) scale using the standard conversion factor (−4.44 eV for 0 V) [32]. The resulting potentials positioned the CB and VB of CuO_x at ~−0.79 and ~1.27 V, respectively, while the LUMO and HOMO of carbon layer were located at ~−0.40 and ~2.20 V, respectively. CuO_x@C retained the precise band energy alignment of pristine CuO_x, indicating that a heterojunction was formed between CuO_x and the carbon layer.

Density functional theory (DFT) calculation was employed to simulate carrier dynamics [33,34]. Both CuO_x and the carbon layer had a relatively small work function (Φ), which facilitated smooth electron transfer from the bulk phase to the surface (Fig. 4C and D). The Fermi levels (E_F) of CuO_x and the carbon layer were −5.57 and −4.81 eV, respectively, indicating that the E_F of CuO_x was close to VB, while the E_F of the carbon layer was close to LUMO (Fig. 4E). When they came into contact with each other, electrons transferred from the carbon layer to CuO_x until an equilibrium state was reached, namely, the same E_F at interface. This created an internal electric field (IEF) that pointed from the carbon layer to CuO_x. Charge density difference and partial density of states (PDOS) further confirmed that there was obvious charge redistribution at the interface, with electrons accumulating around the CuO_x surface (Fig. 4F and Figs. S25 and S26). Meanwhile, the IEF bent the interfacial energy bands of CuO_x and the carbon layer, resulting in the formation of a Type II heterojunction between CuO_x and the carbon layer. During photoelectrocatalytic operation, photoelectrons continuously transferred from the CB of CuO_x to the LUMO of the carbon layer. Electron density distribution showed that the electrons mainly accumulated around oxygen atoms in CuO_x (Fig. 4G). By contrast, in CuO_x@C, electrons were extracted into carbon layer and evenly distributed around carbon atoms (Fig. 4H).

In situ Kelvin probe force microscopy (KPFM) was employed to map the spatial distribution of photocarriers in CuO_x@C (Fig. S27) [35–37]. The surface contact potential difference (CPD) of the carbon layer-coated edge zone and central CuO_x area was ~380 and ~420 mV, respectively. The CPD difference (∆CPD) was up to ~40 mV, indicating that photoelectrons rapidly and directionally transferred from CuO_x to the carbon layer. Such fast charge migration and spatial isolation effectively suppressed electron–hole recombination and extended the photocarriers’ lifetime, further evidenced by photoluminescence (PL) analyses. There was an obvious emission peak at ~364 nm in the steady-state PL spectrum of CuO_x, which was induced by the intrinsic radiative recombination in CuO_x (Fig. 4I). In contrast, the peak intensity dramatically decreased in CuO_x@C, implying the substantial inhibition of electron–hole recombination. Moreover, the recorded transient PL decay curves indicated that the average lifetime of carriers in CuO_x@C (~107.5 ns) was markedly longer than that in CuO_x (~67.1 ns) (Fig. 4J). Surface photovoltage (SPV) spectra showed that the SPV of CuO_x@C (~8.6 μV) was much smaller than that of CuO_x (~27.4 μV), further suggesting that electrons transferred and accumulated in the carbon layer in CuO_x@C (Fig. 4K). In short, the carbon layer effectively extracted photoelectrons from CuO_x, resulting in efficient inhibition of electron-induced reduction photocorrosion and suppression of electron–hole recombination, thereby enhancing the activity and durability of CuO_x@C (Fig. S28).

The CuO_x photoelectrodes have garnered much attention in the fields of energy conversion and environmental purification, with applications including CO₂ reduction, nitrate reduction, and dehalogenation, among others [38–40]. To enhance the activity and selectivity of the target reactions, it was typically necessary to modify the CuO_x surface with specific cocatalysts, such as Ag, Ru, and Pd (Fig. 5A and Fig. S29) [28,29,41]. However, their wide application has been consistently hindered by the poor stability of CuO_x, which was caused by severe photocorrosion. In this work, we innovatively designed and synthesized the CuO_x@C photoelectrodes, which exhibited outstanding activity and unprecedented stability. As a result, CuO_x@C can replace conventional CuO_x and be further modified with specific cocatalysts, including Ag, Ru, and Pd, to achieve large-scale applications (Fig. S30).

We have successfully prepared ternary CuO_x@C/Ag hybrid by incorporating Ag onto the CuO_x@C through electrodeposition and used it for CO₂ reduction to C₂H₄ (Fig. 5B). SEM and EDX-mapping images revealed a uniform and dense distribution of Ag nanoparticles (~80 nm in diameter) firmly anchored on the CuO_x@C surface in CuOx@C/Ag (Fig. 5C and Fig. S31). TEM image resolved distinct Ag (111) lattice fringes with a spacing of ~0.24 nm, confirming the excellent crystallinity of Ag nanoparticles, which was further corroborated by XRD patterns (Fig. S32) [41]. XPS spectra demonstrated that Ag nanoparticles remained predominantly in the metallic state both before and after prolonged photoelectrocatalytic CO₂ reduction operation, establishing metallic Ag⁰ as the durable active phase in reaction, which was consistent with some previous reports (Fig. S33) [41,42]. DRS spectra showed that CuO_x@C/Ag exhibited markedly broadened and enhanced light absorption compared to CuO_x@C, which was mainly caused by the strong localized surface plasmon resonance of Ag nanoparticles (Fig. S34) [43]. KPFM measurements revealed a substantially enhanced ∆CPD (~60 mV) between edge zone and central area in CuO_x@C/Ag, indicating that the Ag modification further facilitated the charge migration and separation (Fig. S35). This was further validated by the PL spectra of CuO_x@C/Ag (Figs. S36 and S37). A substantially quenched steady-state PL emission at ~364 nm was observed, which was complemented by a markedly extended carrier lifetime of 126.4 ns in the transient PL decay. SPV spectra showed that the SPV value of CuO_x@C/Ag was as low as ~7.7 μV, indicating that the Ag nanoparticles served as electron sinks that effectively extracted electrons from the carbon layer (Fig. S38). The aggregation of electrons on the Ag surface was highly beneficial for subsequent CO₂ adsorption and activation. Electrochemical active surface area (ECSA) analysis indicates that carbon layer coating and Ag modification, the latter in particular, markedly increase the ECSA, thereby offering more active sites available for the subsequent CO₂-to-ethylene reduction reaction (Fig. S39).

The photoelectrocatalytic activity, selectivity, and stability of CuO_x@C/Ag for CO₂ reduction to C₂H₄ was systematically evaluated (Fig. 5D). The CO₂ temperature-programmed desorption tests demonstrated that Ag decoration markedly enhanced CO₂ uptake, as evidenced by a distinct peak around 300 °C (Fig. S40). Such moderate-intensity adsorption not only facilitated efficient CO₂ activation but also promoted the timely desorption of products, thus preventing active-site blocking. LSV measurements showed that the modification with the carbon layer, Ag, and carbon layer/Ag increased the photocurrent density of CuO_x by ~2.1 times (CuO_x@C), ~1.7 times (CuO_x/Ag), and ~2.8 times (CuO_x@C/Ag), respectively, indicating greater electron accumulation on the CuO_x@C/Ag surface, which was conducive to enhancing the CO₂ reduction activity (Fig. 5E). Moreover, CuO_x/Ag and CuO_x@C/Ag exhibited a stronger response to CO₂ compared to CuO_x and CuO_x@C, demonstrating that Ag was a more effective active site for CO₂ activation. The C₂H₄ FE of CuO_x@C/Ag reached as high as ~66.4% at −0.1 V vs. RHE, obviously higher than that of CuO_x (~26.3%), CuO_x@C (~29.1%), and CuO_x/Ag (~54.7%), confirming the optimal selectivity of CuO_x@C/Ag for CO₂ reduction to C₂H₄ (Fig. 5F). The CuO_x@C/Ag maintained high C₂H₄ FE (>60%) across a wide potential window from 0.5 to −0.1 V vs. RHE, demonstrating its operational flexibility. At potentials below −0.1 V vs. RHE, the growing contribution of electrocatalysis promoted competing reduction channels, thereby leading to a moderate decline in C₂H₄ FE.

In addition to C₂H₄, some other products, including H₂, CO, CH₄, HCOOH, C₂H₅OH, and CH₃OOH, were also detected on CuO_x@C/Ag (Fig. 5G). Especially at low potentials, as the proportion of electrocatalysis increased, the FE of C₂₊ products markedly improved, which was consistent with the existing reports [44,45]. The effective photocurrent density of CuO_x@C/Ag dedicated to C₂H₄ production was derived by correcting the total current for dark contributions and scaling by the C₂H₄ FE (Fig. 5H). This value increased gradually as the potential decreases and then stabilized at potentials more negative than −0.3 V vs. RHE, reaching a maximum of ~−3.66 mA/cm², suggesting full extraction of available photogenerated carriers under these conditions. Correspondingly, the solar-to-chemical (STC) conversion efficiency for C₂H₄ production of CuO_x@C/Ag rose with decreasing potential before stabilizing, attaining a remarkable peak value of ~1.36% at −0.3 V vs. RHE (Fig. 5I). It was worth noting that even under a diluted CO₂ atmosphere (~20%), CuO_x@C/Ag retained a C₂H₄ FE exceeding 30%, highlighting its exceptional potential for practical deployment in real-world scenarios like industrial exhaust upgrading (Fig. 5J). Furthermore, the photoelectrocatalytic performance of CuO_x@C/Ag toward CO₂ reduction to ethylene was optimized by tuning the ratio of CuO to Cu₂O (Fig. S41). The results reveal that a CuO-to-Cu₂O ratio approaching 1:1 delivers the highest activity. This optimal performance is attributed to the formation of a more complete Type II heterostructure, which maximizes the density of Cu²⁺–O–Cu⁺ bond bridges, thereby facilitating the migration and separation of photogenerated charge carriers. The C₂H₄ FE of CuO_x and CuO_x/Ag sharply decreased from ~26.3% to ~2.9% and from ~54.7% to ~7.6%, respectively, after 14 h of continuous use (Fig. 5K). In contrast, CuO_x@C and CuO_x@C/Ag maintained >95% of their initial performance after prolonged operation. The accompanying time–current curves showed that CuO_x@C/Ag exhibited a remarkably stable current density with minimal decay, further underscoring its robust structural durability and resistance to deactivation (Fig. S42). Strikingly, the activity, selectivity, and durability of CuO_x@C/Ag for CO₂ reduction to C₂H₄ were outstanding and markedly superior to most existing studies (Fig. S43 and Table S1), demonstrating the effectiveness of the synergistic carbon encapsulation and Ag decoration strategy.

A comprehensive investigation of the reaction mechanism for CO₂ reduction to C₂H₄ on CuO_x@C/Ag was conducted. The ¹³CO₂ isotopic labeling experiment verified that the C₂H₄ production originated from CO₂ reduction rather than from other potential carbon sources (Fig. 6A). Meanwhile, the detection of ¹³CO species definitively identified CO as a pivotal intermediate in the C₂H₄ formation pathway [46]. The differential electrochemical mass spectrometry (DEMS) results further revealed the simultaneous generation of substantial CO (m/z = 28) and C₂H₄ (m/z = 28, verified by its characteristic fragment C₂H₃ [m/z = 27]), confirming the conversion sequence of CO₂ → CO → C₂H₄ (Fig. 6B) [47]. The absence of DEMS signals of CH₄ (m/z = 16) and HCOOH (m/z = 46) further underscored the remarkable selectivity toward C₂H₄ production on CuO_x@C/Ag. The in situ attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy collected on CuO_x@C/Ag showed that the distinct vibrational signatures of *CO (~2,053.1 cm⁻¹) and *COCHO (~1,112.2 and 1,637.3 cm⁻¹) were observed, while the characteristic *CHO band (~1,698.6 cm⁻¹) remained absent (Fig. 6C) [48,49]. This provided compelling evidence that the crucial C–C coupling step proceeded predominantly through symmetric dimerization of 2 *CO intermediates, rather than the alternative asymmetric *CO–*CHO pathway. The reaction cascade was reconstructed through further identification of additional intermediates, including *COOH (~1,216.2 and 1,480.3 cm⁻¹) and *CHOCH₂ (~1,567.2 cm⁻¹), collectively establishing the general reaction sequence of *CO₂ → *COOH → *CO → *COCO → *COCHO → *CHOCH₂ → *C₂H₄. In contrast, all these intermediate signals were markedly attenuated on the Ag-free CuO_x@C, consistent with its inferior activity and C₂H₄ selectivity, and highlighting the essential role of Ag in promoting intermediate stabilization and the decisive C–C coupling step (Fig. S44).

DFT calculations provided atomic-scale validation for the proposed reaction pathway (Fig. 6D and E and Fig. S45). The introduction of Ag markedly enhanced CO₂ adsorption, with adsorption energies strengthening from −0.25 eV (CuO_x@C) to −0.75 eV (CuO_x@C/Ag). The hydrogenation barriers of *CO₂ to *COOH dramatically reduced from 1.95 eV on CuO_x@C to 0.5 eV on CuO_x@C/Ag. This enhancement was further corroborated by differential charge density analysis, revealing substantial electron transfer between *CO₂ and CuO_x@C/Ag (Bader charge: 1.253 |e|) compared to only 0.744 |e| with CuO_x@C. The strengthened interaction facilitated C–O bond activation (from 1.171 to 1.183 Å) and enabled efficient *CO₂ hydrogenation. The resulting *CO intermediate was effectively stabilized on CuO_x@C/Ag, with adsorption energies of −0.3 eV, in contrast to positive values (0.4 eV) on CuO_x@C, where *CO tended to desorb. The comparison of formation barriers for *CO (−0.1 eV) and *HCOOH (0.1 eV) provided a theoretical foundation for the experimentally observed minimal HCOOH production. On CuO_x@C/Ag, the C–C coupling barrier via *CO dimerization (~0.7 eV) was markedly lower than the barrier for *CO hydrogenation to *CHO (~0.9 eV), establishing the thermodynamic preference for symmetric *CO dimerization over asymmetric *CO–*CHO pathways. Meanwhile, the unfavorable *CHO formation fundamentally explained the suppressed productions of CH₄ and CH₃OH observed experimentally. In stark contrast, CuO_x@C exhibited prohibitively high C–C coupling barrier (~2.2 eV). Further evidence emerged from the substantial charge transfer between *CO and CuO_x@C/Ag (~1.247 |e| versus ~0.741 |e| for CuO_x@C) and strengthened C–O bond activation (from 1.142 to 1.151 Å). The complete reaction pathway on CuO_x@C/Ag for CO₂ reduction to C₂H₄ proceeded through an optimized sequence of *CO₂ → *COOH → *CO → *COCO → *COCHO → *COCHOH → *CHOCHOH → *CHOCH₂OH → *CHOCH₂ → *CHOHCH₂ → *CH₂OHCH₂ → *C₂H₄, demonstrating how Ag incorporation simultaneously enhanced CO₂ adsorption, stabilized critical intermediates, and promoted selective C–C coupling toward C₂H₄ production.

To establish the general applicability of the coupled carbon-encapsulation and metal-decoration strategy, we further engineered a series of functional photoelectrodes, including CuO_x@C/Ru, CuO_x@C/Pd, and BiVO₄@C/Pt (Fig. 6F to H). These systems were respectively deployed for nitrate reduction to NH₃, chlorophenol dechlorination, and the oxygen evolution reaction, respectively (Fig. S30). A particularly noteworthy configuration was realized in the BiVO₄@C/Pt, where the carbon layer served as an efficient hole-extraction mediator. This led to spatial isolation of photoholes, thereby inhibiting hole-induced oxidation photocorrosion and suppressing electron–hole recombination. The FE for nitrate reduction to NH₃ on CuO_x and CuO_x/Ru sharply reduced from ~31.4% to ~4.1% and from ~89.1% to ~21.3%, respectively, after 14 h of continuous use (Fig. 6I). By contrast, CuO_x@C and CuO_x@C/Ru maintained ~95% of their initial FE for nitrate reduction to NH₃ after prolonged use. As shown in Fig. 6J, the dechlorination rate of p-chlorophenol on CuO_x and CuO_x/Pd sharply reduced from ~39.2% to ~7.4% and from ~92.3% to ~20.7%, respectively, after 9 h of continuous use. By contrast, CuO_x@C and CuO_x@C/Pd maintained ~97% of their initial dechlorination rate after prolonged use. The photocurrent densities of BiVO₄ and BiVO₄/Pt decreased over time, remaining at only ~57% of their initial values after 40 h of chronoamperometry (Fig. 6K). By contrast, BiVO₄@C and BiVO₄@C/Pt exhibited stable photocurrent during continuous operation, maintaining ~98% of their initial photocurrent density at the end of the test. It was worth mentioning that the activity and durability of CuO_x@C/Ru, CuO_x@C/Pd, and BiVO₄@C/Pt were outstanding and markedly superior to most existing studies (Fig. S43 and Table S1), demonstrating the synergistic carbon encapsulation and cocatalyst decoration strategy.