Efficient electroreduction of CO2 to C2+ products on CeO2 modified CuO

Electrocatalytic reduction of CO2 into multicarbon (C2+) products powered by renewable electricity offers one promising method for CO2 utilization and promotes the storage of renewable energy under an ambient environment. However, there is still a dilemma in the manufacture of valuable C2+ products between balancing selectivity and activity. In this work, cerium oxides were combined with CuO (CeO2/CuO) and showed an outstanding catalytic performance for C2+ products. The faradaic efficiency of the C2+ products could reach 75.2% with a current density of 1.21 A cm−2. In situ experiments and density functional theory (DFT) calculations demonstrated that the interface between CeO2 and Cu and the subsurface Cu2O coexisted in CeO2/CuO during CO2RR and two competing pathways for C–C coupling were promoted separately, of which hydrogenation of *CO to *CHO is energetically favoured. In addition, the introduction of CeO2 also enhanced water activation, which could accelerate the formation rate of *CHO. Thus, the selectivity and activity for C2+ products over CeO2/CuO can be improved simultaneously.

Cu-based catalysts are the most promising electrocatalysts for converting CO 2 into C 2+ products, [33][34][35][36][37][38][39][40][41][42][43][44][45] owing to their moderate adsorption capacity for the crucial intermediate (*CO). Based on previous reports, 1,18,46,47 the selectivity of C 2+ products over Cu-based catalysts can be notably improved by the introduction of another component but the understanding of the structure-selectivity relationship remains controversial because the valence state and the microstructure of copper may be inuenced simultaneously. What is more, complexity also exists in the production of C 2+ products during CO 2 RR due to the C-C coupling step involved, which not only contains multiple electron-transfer and protonation steps, 6 but also exhibits various potential coupling paths on heterogeneous catalysts. As a result, it is necessary to comprehensively reveal the role of another component in the promotion of the selectivity towards C 2+ products during CO 2 RR.
Given the neutral or basic electrolyte used in CO 2 RR, H 2 O can serve as the hydrogen source and the activity should be bound up with the activation of H 2 O in CO 2 RR. 47 According to the Sabatier principle, the energy barrier for the activation of water should be particularly controlled, which could provide enough hydrogen for the hydrogenation of intermediates but not cause excessive production of H 2 . Considering cerium oxide (CeO 2 ) has a high activity for water activation in CO 2 hydrogenation and shows poor activity for the hydrogen evolution reaction (HER), [48][49][50] we can assume that the activity for C 2+ products would be improved compared to the CeO 2 modi-ed Cu-based catalyst in CO 2 RR.
Herein, we used CeO 2 to modify CuO to obtain CeO 2 /CuO catalysts, and both a high current density and selectivity towards C 2+ products were achieved in CO 2 RR. A faradaic efficiency (FE) of 75.2% for the C 2+ products could be attained on the catalyst with a total current density of 1.21 A cm À2 in a owcell system. The experiments and density functional theory (DFT) calculations indicate the energy of generation of *CHO is thermodynamically reduced by the interfacial effect compared to CeO 2 modied CuO catalysts and the rapid activation of water around CeO 2 accelerates the formation of *CHO kinetically, thus the C-C coupling step is facilitated via the *CHO route, endowing the CeO 2 /CuO catalyst with an excellent catalytic performance towards C 2+ products.

Results and discussion
The Ce(OH) 2 /Cu(OH) 2 catalysts were rst prepared by the coprecipitation method, then the CeO 2 /CuO catalysts were gained by annealing at 600 C in air. As the amount of Ce in the catalysts increased from 0 to 30%, a set of peaks belonging to the CeO 2 phase gradually emerged on the base of the primary CuO phase in the X-ray diffraction patterns (Fig. 1a), indicating the coexistence of CeO 2 and CuO in the catalysts, and the CeO 2 / CuO catalysts were named CCX (X ¼ the molar ratio of Ce and Cu times 100). From scanning electron microscopy (SEM) and transmission electron microscopy (TEM), we can observe that CeO 2 nanoparticles below 5 nm were evenly loaded on the surface of CuO (Fig. 1b, c and S2 †). Two typical d-spacings of 0.31 nm and 0.23 nm were observed in the image of highresolution transmission electron microscopy (HR-TEM) for CC20 (Fig. 1d), corresponding to CeO 2 (111) and CuO(111). According to the distribution of the elements of Cu, Ce and O in the energy dispersive X-ray spectroscopy maps (Fig. 1e), the uniform element dispersion of Cu and Ce over the catalyst conrmed that CeO 2 was uniformly dispersed on the CuO.
The electrocatalytic performance of the catalysts was evaluated in the ow cell and 1 M KOH was used as the electrolyte, as reported in our previous work. 51 Before the CO 2 RR, the catalysts were rstly reduced around À0.3 V vs. RHE, which is more negative than the transformation of CuO to Cu (Fig. S3 †). The polytetrauoroethylene (PTFE) membrane (average pore size of 0.22 mm) was used as the gas diffusion electrode, and gaseous and liquid products were analyzed by gas chromatography (GC) and nuclear magnetic resonance (NMR) spectroscopy, respectively (Fig. S4 †). The 13 C labelled CO 2 was used as the source of the reactant gas and the results veried that CO 2 was the only carbon source in CO 2 RR (Fig. S5 †).
Based on the performance of the CCX catalysts in CO 2 RR, a typical volcano plot between FE C 2+ and Ce content was observed at À1.02 V (vs. RHE) and CC20 exhibited the best performance at various applied potentials ( Fig. 2a and S6-S8 †). From the TEM (Fig. S2 †), we can observe that the interfaces were produced with the increase of the Ce amount, thus we hypothesized that the selectivity of C 2+ was related to the interfaces. However, for the CC30, the selectivity of C 2+ products showed a signicant decrease because too many Cu sites were covered by the CeO 2 . Thus, CC20 was chosen for further comparison with CC0. It can be clearly observed that CC20 showed outstanding efficiency for C 2+ products in the CO 2 reduction (Fig. 2b). The FE of C 2+ products for CC20 could reach 75.2% at À1.12 V (vs. RHE), while that over CC0 was only 48.3% at the same condition. Moreover, the evolution of H 2 was suppressed over CC20 and the FE of n-propanol was notably improved on CC20 compared to CC0 (Fig. S7-S9 †), which might correlate with the escalation of C 2 intermediates over CC20 (Fig. S9 †). In the meantime, a signicant increase was also achieved on the current density on CC20. It is very impressive that the partial current density of C 2+ products (j C 2+ ) over CC20 could reach as high as 0.91 A cm À2 at À1.12 V (vs. RHE), which is   (Table S1 †). about 10 times higher than that on CC0 (Fig. 2c). Compared with the state-of-the-art catalysts, the activity and selectivity for C 2+ products over CC20 are among the highest values ( Fig. 2d and Table S1 †). The above results indicate that the introduction of CeO 2 could signicantly improve both the selectivity and activity for C 2+ products. Moreover, the performance of CeO 2 was also characterized ( Fig. S12 †), and only trace CO was detected at À0.87 V and À0.97 V (vs. RHE), while the current density was below 20 mA cm À2 at the applied potentials, indicating that pure CeO 2 showed poor activity for CO 2 RR. Besides, the catalysts were characterized aer the reaction and no obvious change was observed in the TEM images and XRD patterns (Fig. S14-S17 †). As a result, the proper content of CeO 2 would obviously benet the catalytic performance of the CuO catalyst towards C 2+ products in CO 2 RR.
To reveal the reasons for the superior catalytic performance of CC20 in the CO 2 RR, the electrochemical active surface areas (ECSAs) and electrochemical impedance spectroscopy (EIS) of the catalysts were studied. We can observe that similar ECSAs were obtained over the CCX catalysts with different CeO 2 contents (Fig. S18 †), indicating the similar surface area of the catalysts at the electrochemical conditions. Moreover, the charge transfer resistance (R ct ) for the catalysts was also similar (Fig. S19 †), suggesting that the discrepancy of the efficiency for C 2+ products did not mainly originate from the slight difference of the ECSAs and electronic conductivity.
The catalytic performance of Cu-based catalysts was closely related to the oxidation state and local structure, which could alter the adsorption of intermediates, 30,41,42,52 thus the operando X-ray absorption spectroscopy (XAS) was used to track the evolution of the oxidation state and local structure of Cu and Ce over CC0 and CC20 during CO 2 RR. At the open circuit potential (OCP), near the Cu K-edge, both the X-ray absorption near edge structure (XANES) and the k 3 -weighted Fourier-transformed (FT) extended X-ray absorption ne structure (EXAFS) spectra of CC0 and CC20 showed the typical features of CuO ( Fig. 3a and b, S20 and S21 †), indicating that CuO was dominant in CC0 and CC20 before the reaction. As the potential was applied as À0.62 V (vs. RHE), there was no obvious change in either the XANES or FT-EXAFS spectra. Meanwhile, CC0 and CC20 showed a low FE C 2+ and these results could be due to the large proportion of Cu(II). When the applied potential decreased to À0.82 V (vs. RHE), features of Cu with low oxidation states emerged in the XANES spectra and EXAFS analysis also displayed that the Cu rst shell coordination switched to the mixture of different Cu species over CC0 and CC20. It can be found that the FEs of C 2+ at À0.82 V (vs. RHE) also showed a signicant increase compared to that at À0.62 V (vs. RHE), suggesting the potential correlation between the Cu oxidation state and the FE C 2+ in CO 2 RR. Furthermore, according to the XANES spectra, the oxidation state of Cu in the catalysts continued to decrease and the results in the EXAFS data were different from the initial CuO-like state, demonstrating that Cu with a low oxidation state became the main phase. Interestingly, the FE of C 2+ in CO 2 RR still slightly increased from À0.82 V (vs. RHE) to À1.02 V (vs. RHE) on CC0 and CC20, which supported the conclusion that the low-valent Cu species on catalysts were the active phase in CO 2 RR. Moreover, according to the distribution of various Cu species for CC20 and CC0 (Fig. S22 and S23 †), we can observe that the Cu 2 O species occupied the higher proportion over CC20 than on CC0. These results indicated that the introduction of CeO 2 could stabilize the Cu 2 O, which could be attributed to the interaction between Ce and Cu, and the role of Cu 2 O in CO 2 RR will be discussed in the later section. In addition, the operando XANES data at the Ce L 3 -edge of CC20 showed a negligible change during CO 2 RR (Fig. 3c), indicating that CeO 2 remained stable during CO 2 RR.
Generally, the activity and selectivity of C 2+ products are closely related to the surface species on the catalysts during the reduction. So, an in situ surface enhanced Raman spectroscopy (SERS) study was carried out to explore the surface species over CC0 and CC20 (Fig. 3d and e, S24 †). Aer the pre-electrolysis at N 2 atmosphere, only two weak bands at 524 and 610 cm À1 were observed, which belonged to Cu 2 O, 53 and then disappeared in CO 2 electrolysis. Instead, bands at 390 and 536 cm À1 emerged at negative potentials in CO 2 electrolysis, which were attributed to the chemisorption of CO 2 on the surface Cu. 57,58 Furthermore, we can observe that no Cu 2 O could be found on both CC0 and CC20 during CO 2 RR from the Raman spectra. Combined with the results of the operando XAFS, we can assume that the Cu 2 O species exists on the subsurface of the catalysts, due to the Raman spectroscopy being sensitive to the surface species of the catalyst, 56,57 which is consistent with previous reports. 21,59 In addition, the signals of CeO 2 cannot be found on CC20 in the Raman spectroscopy, this may be due to the signals of CeO 2 being too weak under the existence of the electrolyte in the in situ experiments. 60 As the applied potential negatively moved, both on CC0 and CC20, peaks at 285, 365, 1800-1860 and 2000-2100 cm À1 became cognizable, corresponding to the restricted rotation of adsorbed *CO on Cu, Cu-CO stretching, and bridge and top C^O stretching, respectively. [54][55][56][57][58]61 It is interesting to note that there was a distinct disparity in the performance of the above *CO related peaks over CC0 and CC20. For CC0, at À0.37 V (vs. RHE), *CO related peaks began to be observed and the peak around 1820 cm À1 was weak. On the contrary, those peaks were clearly present over CC20 aer À0.17 V (vs. RHE), and the peak between 1800-1860 cm À1 even showed a red shi while the peak at 2000-2100 cm À1 became strong. The difference between those two catalysts supported the conclusion that CO 2 could be transformed into CO at lower applied potentials on CC20 than on CC0, indicating the superior activity of CC20 towards CO in CO 2 RR, which is consistent with the results in the electrochemical tests (Fig. S11 †). Moreover, the excellent catalytic capability for the CO product in CO 2 RR should be favourable for the following steps in CO 2 RR.
DFT calculations were then performed to elucidate the mechanism of the crucial C-C coupling step and to gain insight into the excellent performance of CC20 in CO 2 RR. According to the above results, the introduction of CeO 2 can not only form the interface between CeO 2 and Cu, but can also stabilize the subsurface Cu 2 O. Although both the interface and subsurface Cu 2 O can promote the CO 2 RR, 62,63 they have been studied separately in previous reports. 64,65 Thus, the role of interface and subsurface Cu 2 O on enhancing the C 2+ products should be studied simultaneously and three specic models were used to study the effect of subsurface Cu 2 O and CeO 2 on promoting C-C coupling (Fig. S27 †). First, a model with more metallic Cu on the surface and less Cu 2 O on the subsurface (Cu-M) was built to represent the CC0 (Fig. 4a). Then, a model with less metallic Cu on the surface and more Cu 2 O on the subsurface (Cu-L) was built to represent the CC20 without CeO 2 (Fig. 4b). Last, Cu-L with CeO 2 on the surface (CeO 2 /Cu-L) was built to represent CC20 (Fig. 4c). The Cu(111) and CeO 2 (111) were chosen as the basic models according to the results of XRD (Fig. S16 †), and the ratio of Cu and Cu 2 O was set according to the results of in situ XAS (Fig. S22 and S23 †).
Generally, CO 2 can be rst reduced into CO through the *COOH pathway, 28 and the adsorbed CO is regarded as the common intermediate for the C 2+ products in CO 2 RR. 66 In this condition, four potential reaction pathways are taken into account in the C-C coupling step and all of them are generated from the vital intermediate *CO (Fig. S28-S33 †). On Cu-M, the energy of 1.51 eV is required for the dimerization of *CO (path 1), higher than that on Cu-L (0.93 eV), indicating that more subsurface Cu 2 O are benecial for the C-C coupling through the *CO-*CO route (Fig. 4d and e), which is consistent with previous reports. 21,63 Further addition of CeO 2 on Cu-L barely alters the energy for the dimerization of *CO (0.92 eV) compared with Cu-L (Fig. 4f). Consequently, we can assume that the energy for dimerization of *CO can be decreased by subsurface Cu 2 O, however, the energy for the formation of the *O*CCO intermediate was still very high, indicating that C-C coupling through *CO dimerization is difficult.
We notice that both path 2 and path 3 suffer from the endothermic protonation of adsorbed *CO and the subsequent exothermic coupling step in all the models. In terms of the lower energy needed for the generation of *CHO compared to *COH, we can assume that the C-C coupling step would prefer the *CHO route rather than the *COH route. However, the formation of *CHO in each model is different in energy. 0.59 eV is required for the hydrogenation of *CO into *CHO on Cu-M, while a higher energy of 0.87 eV is needed on Cu-L, suggesting that more subsurface Cu 2 O were not advantageous for the formation of *CHO. This may be due to the fact that the adsorption of *CHO can be affected by the subsurface Cu 2 O, and the intrinsic reason should be further studied. Surprisingly, the energy for the hydrogenation of *CO into *CHO dramatically declined to À0.11 eV and became exothermic near the interface of CeO 2 and Cu-L ( Fig. S32 and S33 †). The above results convincingly demonstrate that *CHO is easily formed from *CO on CC20 and this should be attributed to the introduction of CeO 2 and the formed interface, rather than more subsurface Cu 2 O. Furthermore, for the following C-C coupling step related to *CHO, the coupling of *CO and *CHO (path 4) is also possible in theory except for the dimerization of *CHO (path 3). Nevertheless, the coupling of *CO and *CHO is endoenergic over all surfaces, suggesting that the exoenergic dimerization of *CHO would be favourable to the coupling process. On the whole, the coupling of *CHO into *OHCCHO* is most favourable in the C-C coupling step among the above possible pathways in the three models and the process even becomes spontaneous in the presence of CeO 2 . Ma and coworkers also found that the coupling between *CHO showed lower barriers on the Cu(111) surfaces. 28 In consequence, the *CHO route (path 3) is favoured on the three models in the C-C coupling step and becomes exothermal on CC20 due to the formed interface, elucidating the high FE for C 2+ products on CC20. In addition, we can observe that all the intermediates were mainly adsorbed on the exposed Cu sites, so we can assume that Cu was the active site.
In consideration of the 1 M KOH used in CO 2 RR, H 2 O should be considered as the hydrogen donor. As a result, we introduced the water activation process to further study the kinetic process for the formation of *CHO. For Cu-M and Cu-L, water is spontaneously adsorbed on the surface Cu and then the high energy barriers of 1.20 eV and 1.21 eV are needed to form the transient state (TS) for the following formation of *CHO (Fig. 4g, h and S34 †), respectively. For the CeO 2 /Cu-L, H 2 O would like to be adsorbed around the CeO 2 and undergo dissolution to offer active hydrogen. Due to the sufficient active hydrogen, the barrier for TS decreases to only 0.53 eV (Fig. 4i), making the formation of *CHO more kinetically feasible on CC20. In conclusion, the formation of *CHO is faster on CeO 2 /Cu-L than that on other surfaces without CeO 2 due to the rapid water activation around CeO 2 , which agrees with the high current density for CC20 during CO 2 RR.
In addition, the DFT calculations were also carried out at the bias of À0.5 V and À1.12 V (Fig. S35 and S36 †), respectively, which are the requirement to overcome the C-C coupling step and are consistent with the reaction potential. At the selected potentials, we can observe that hydrogenation of *CO to *CHO and then coupling of *CHO into *CHO-*CHO still remain the favourable path for C-C coupling on each surface during CO 2 RR. More importantly, both the thermodynamic process and kinetic process for the formation of *CHO on CeO 2 /Cu-L are more feasible than that on Cu-M or Cu-L. These results elucidate the motivation for the simultaneously enhanced selectivity and activity for C 2+ products by the introduction of CeO 2 .
According to the DFT calculation results, hydrogenation of *CO to *CHO played a crucial role for enhancing the C 2+ products, especially incorporated with the activation of H 2 O. In consequence, the kinetic isotopic effects (KIEs) of H/D over CC0 and CC20 were measured to further ensure the role of water activation in CO 2 RR (Fig. S37 †). As the H 2 O was replaced by D 2 O as the solvent in 1 M KOH solution, the formation rate of ethylene signicantly decreased on CC0, and the KIE (the ratio of ethylene formation rates in H 2 O and D 2 O) was about 2.0, which suggests that dissolution of H 2 O should be involved in the rate-determining step (RDS) for the ethylene formation. On the contrary, the KIE value on CC20 was nearly 1, suggesting that hydrogen was not related to the rate-determining step over CC20. The above conrmed the results of the DFT calculations that the *CHO route was endothermic on CC0 and exothermic on CC20. In addition, CC20 yielded 312 mA cm À2 at À1.12 V (vs. RHE) for HER under N 2 atmosphere, about 2.5 times higher than that on CC0 (Fig. S37 †), supporting the argument that CC20 had a superior capability for water activation. Thus, it can be concluded that the existence of CeO 2 accelerated the dissolution of H 2 O to offer enough active hydrogen and thus beneted the generation of *CHO, which enhances the C-C coupling step through the dimerization of *CHO.

Conclusions
In conclusion, the introduction of CeO 2 on the surface of CuO signicantly enhanced the selectivity and activity towards C 2+ products in CO 2 RR. Experimental and in situ SERS results conrmed the generation of the important intermediate CO was notably enhanced on CC20, which offered abundant precursors for the following steps. More importantly, DFT calculations revealed that the C-C coupling step followed the *CHO route and was facilitated both thermodynamically and kinetically on CC20 by the interfacial effects and the rapid water activation, respectively, ndings which were also supported by the KIE experiments. Consequently, the FE of the C 2+ products could reach up to 75.2% with the current density of 1.21 A cm À2 at À1.12 V (vs. RHE) in 1 M KOH. We believe that the ndings in this work contribute to understanding the role of the introduced component and could help to design efficient catalysts towards C 2+ products in CO 2 RR.

Conflicts of interest
There are no conicts to declare. Science & Technology Commission (Z191100007219009), and Chinese Academy of Sciences (QYZDY-SSW-SLH013). The operando X-ray adsorption spectroscopy (XAS) measurements were performed using a ow cell at the 1W1B, 1W2B beamline at Beijing Synchrotron Radiation Facility (BSRF), China.