Rational Manipulation of Intermediates on Copper for CO2 Electroreduction Toward Multicarbon Products

Excess greenhouse gas emissions, primarily carbon dioxide (CO2), have caused major environmental concerns worldwide. The electroreduction of CO2 into valuable chemicals using renewable energy is an ecofriendly approach to achieve carbon neutrality. In this regard, copper (Cu) has attracted considerable attention as the only known metallic catalyst available for converting CO2 to high-value multicarbon (C2+) products. The production of C2+ involves complicated C–C coupling steps and thus imposes high demands on intermediate regulation. In this review, we discuss multiple strategies for modulating intermediates to facilitate C2+ formation on Cu-based catalysts. Furthermore, several sophisticated in situ characterization techniques are outlined for elucidating the mechanism of C–C coupling. Lastly, the challenges and future directions of CO2 electroreduction to C2+ are envisioned.


Introduction
Massive carbon dioxide (CO 2 ) emissions arising from industrial activities have caused a substantial increase in the atmospheric CO 2 concentration, and the ensuing global warming has resulted in increasingly frequent environmental disasters such as starvation, habitat loss, species extinction, and sealevel rise [1,2]. To tackle these problems, carbon capture, utilization, and storage have been conceived as pathways to peak carbon emissions and eventually reach a carbon-neutral society before 2060 [3]. Electrochemical CO 2 reduction reaction (CO 2 RR) powered by renewable electricity provides a sustainable solution for converting waste CO 2 emissions into useful feedstocks and thereby realizing a net negative carbon footprint and long-term storage of the intermittent renewable electricity in chemical bonds [4,5]. In the past three decades, significant progress has been achieved in using CO 2 RR to generate diverse products, including CO [6][7][8][9], HCOOH [10,11], CH 4 [12][13][14], C 2 H 4 [15][16][17], C 2 H 5 OH [18][19][20], and n-C 3 H 7 OH [21][22][23]. Compared to monocarbon (C 1 ) products, multicarbon (C 2+ ) products are more desirable because of their higher energy density, richer chemical structure, and more versatile applications [24].
Since CO 2 is a thermodynamically stable molecule with strong C=O bonds (750 kJ/ mol) [25], converting CO 2 to C 2+ products is extremely difficult. Copper (Cu) is the only known metallic electrocatalyst capable of generating C 2+ products in considerable amounts [26][27][28]. However, its selectivity and the activity of the C 2+ products are drastically limited by the sluggish kinetics of the C-C coupling step (i.e., the bifurcation for the generation of C 2+ and C 1 products), as well as because of competition with H-H formation [29]. In the multielectron and multiproton transfer process of CO 2 RR, the adsorption energy and coverage of key intermediates, such as *H, *COOH, *OCHO, and *CO, are commonly considered 1 3 to determine the reaction routes and activity of CO 2 RR [30,31]. For example, the first proton-coupled transfer reaction will generate *OOCH or *COOH intermediates. *OOCH is the intermediate for formate production, while *COOH is the intermediate for the formation of CO gas or *CO, which is a key intermediate for C-C coupling [32,33]. Theoretical calculations indicate that the binding energy of *H weakens with increasing *CO coverages, and this in turn inhibits the activity of the competing hydrogen evolution reaction (HER) [31]. Therefore, controlling the intermediate adsorption is critical for inhibiting the formation of H 2 and C 1 products but facilitating the formation of adsorbed CO dimers (e.g., *CO-CO or *CO-CHO). In addition, Cu moderately binds with most intermediates, resulting in multiple reaction pathways and thus yielding a mixture of numerous products [26]. To improve the selectivity and activity of C 2+ products, a variety of strategies such as grain boundary exposure [34][35][36], heteroatom doping [37,38], and local microenvironment regulation [39,40] were developed to manipulate the adsorption state of intermediates and facilitate their deep reduction.
This review focuses on integrated strategies for regulating the adsorption state of intermediates and their influence on C 2+ production for Cu-based electrocatalysts (Fig. 1). Moreover, we discuss the vital role of in situ characterization techniques for examining CO 2 RR intermediates. Finally, we highlight the challenges and perspectives associated with the electroreduction of CO 2 to C 2+ products. This review can provide a better understanding of the principles of catalyst design with the aim of achieving an overall improvement for the CO 2 -to-C 2+ catalytic systems.

Proposed Mechanisms for CO 2 RR Toward C 2+ Products
Since Hori et al. [41,42] first discovered that Cu is capable of electroreducing CO 2 to C 2+ products, Cu-based catalysts have been recognized as the best electrocatalysts for C 2+ production during the past three decades. To investigate the origin of the deep reduction activity of Cu, the binding energies of the adsorbed CO (*CO) and hydrogen (*H) were used as a descriptor to predict the CO 2 RR product distribution of different metal catalysts [26,43]. In the case of metals such as Au and Ag that bind weakly to *CO and *H, *CO prefers direct desorption to form CO. Metals such as Pt and Fe have high binding strength with *CO, but their favorable adsorption of *H facilitates HER. Unlike in the case of those metals, the adsorption energy of *CO and *H Fig. 1 Schematic illustration of the intermediate manipulation strategies for the electrochemical CO 2 reduction reaction (CO 2 RR) toward multicarbon products on Cu is neither too strong to poison its active sites nor too weak to be immediately desorbed. This moderate intermediate binding energy of Cu enables continuous C-C coupling and multistep hydrogenation, which is essential for the deep reduction of CO 2 beyond CO. Different reaction routes have been proposed for C 2+ formation (Fig. 2). *CO is generally recognized as a key reaction intermediate in the electroreduction of CO 2 to C 2+ products [29,44,45]. The dimerization of *CO [46,47] and coupling of *CO with protonated *CO (*CHO or *COH) [48,49] are widely accepted possible routes for C-C coupling. In the *CO dimerization mechanism, C-C coupling is generally regarded as the rate-determining step (RDS) for C 2+ production and has been widely applied in theoretical calculations [50,51]. In this case, two *CO intermediates on the Cu surface are coupled to form *OCCO, and this process is accompanied by an electron transfer [52]. However, density functional theory (DFT) simulations suggest that the RDS of C 2+ products may be changed from *CO dimerization to *CO-*CHO or *CO-*COH coupling with the increase in the local *CO coverage at higher negative potentials [53]. In the mechanism of *CO coupling with hydrogenated *CO, *CO hydrogenation is considered the RDS for C 2+ formation on Cu (111) and (100) surfaces [54]. The competition between *CO and proton sources (i.e., *H, H + , or adsorbed H 2 O) on the active sites of Cu will lead to a shift in product distribution [55,56]. After the C-C coupling step, the *COCHO or *COCOH is subsequently reduced to *CHCOH, which is a common precursor of C 2 H 4 and C 2 H 5 OH [57,58]. *CHCOH deoxygenation to *CCH or hydrogenation to *CHCHOH can generate C 2 H 4 and C 2 H 5 OH, respectively [20].
In addition, the formation of C 3 products on Cu has been observed. Some rationally designed Cu catalysts attained a Faradaic efficiency (FE) of up to 15.4% for n-propanol (the major C 3 product) in CO 2 RR (33% in CO reduction reaction (CORR)) [21,59]. However, the reaction pathway to C 3 products, which inevitably requires two successive C-C coupling steps, has not been well-studied yet because of the lack of sophisticated characterization techniques for *C 2+ intermediates. As an approximate solution, electroreducing C 2 /C 3 compounds toward those C 3 products that may share the reaction pathway as the electrochemical CO 2 RR toward the same products is employed for a preliminary mechanistic exploration. Hori et al. [60] found that the electroreduction of propionaldehyde produces large amounts of n-propanol, suggesting that propionaldehyde is a potential intermediate for the n-propanol product. Through electrochemical differential mass spectrometry, Bell's group [61] observed that the relative abundance of ethanol increased at the expense of propionaldehyde at more cathodic potentials. This observation indicates that ethanol and propionaldehyde share the same intermediate (*CH 3 CHO). To identify the intermediates for the second C-C coupling step, Zhang et al. [62] performed a co-electroreduction of isotopically labeled CO and acetaldehyde and found that only a minor fraction (up to 36%) of the n-propanol product originates from the cross-coupling between CO and acetaldehyde. The adsorbed methylcarbonyl (*OCH 2 CH 3 or *CHOHCH 3 ) is considered a possible intermediate for the second C-C coupling step, in which the reaction pathways bifurcate toward C 2 or C 3 products. Therefore, a reasonable speculation for the C 3 pathway is that a *C 2 intermediate undergoes intermolecular C-C coupling with a neighboring *CO intermediate, and this is followed by proton or electron transfer to form propionaldehyde. The key to C 3 production is to stabilize *C 2 intermediates and optimize the *CO coverage [22,63].
In conclusion, the C-C coupling step is essential for the conversion of CO 2 to C 2+ products. This step is regulated by the adsorption states of the intermediates on Cu. With regard to the initial intermediate for C 2+ generation, the surface coverage and adsorption duration of *CO affect the probability of C-C coupling. Thereafter, the adsorption states of the subsequent intermediates determine the energy barrier of C-C coupling and the followed reaction routes. Thus, manipulating intermediates by rational Cubased catalyst design is a promising way to produce highvalue C 2+ products efficiently.

Strategies of Intermediate Manipulation for Promoting C 2+ Production
In the following section, we will summarize the principal strategies of manipulating intermediates, including the surface structural effects, introduction of additional elements, chemical state effects, electric field effects, substrate effects, confinement effects, and local microenvironment effects for promoting the deep conversion of CO 2 to C 2+ products.

Crystal Facets
The crystal facet is one of the critical structural parameters for Cu. The adsorption strength of intermediates has a high sensitivity to the crystal facets. Hence, the distributions of CO 2 RR products on various crystal facets are different. In general, Cu (111) facets have an appropriate ratio of *CO to *H, and thus, these facets favor CH 4 production. Unlike Cu (111), Cu (100) facets are more favorable for *CO adsorption, thereby promoting C-C coupling to generate C 2 H 4 ( Fig. 3a, b) [64]. DFT calculations demonstrate that *CO dimerization on Cu (100) exhibits the lowest energy among the low-index Cu facets [49]. Therefore, the selective exposure of Cu (100) facet is an important strategy to improve C 2+ selectivity [65]. Wang's group [23] selectively exposed Cu (100) facets through the metal-ion battery cycling method, achieving a sixfold improvement in the ratio of C 2+ and C 1 products compared with the polished Cu foil. Furthermore, the C 2+ / CH 4 value was plotted as a function of crystal orientation, illustrating that high-index facets can further improve the selectivity for C 2+ products (Fig. 3c) [42]. For example, Cu (711) is thermodynamically favored for C-C coupling via cross-coupling of the *CO and *COH intermediates based on DFT calculations [66]. Similarly, Cu (751) has higher activity and selectivity for C-C coupling compared to low-index facets [67]. Changes in the Cu facets also affect the activity and selectivity of CO 2 RR. High-purity 4H Cu and heterogeneous 4H/ face-centered cubic (fcc) Cu synthesized on the template of 4H and 4H/fcc Au exhibit higher overall activity and catalytic selectivity than fcc Cu, indicating the high dependence of electrocatalytic behaviors on crystal facets (Fig. 3d) [68].
Although the C 2+ activity and selectivity can be effectively improved by controlling the crystal facet, Cu usually suffers from dynamical reconstruction during the CO 2 RR process. This reconstruction causes changes in its original crystal facets and catalytic performance [69,70]. Besides, note that the active Cu surface may also undergo an irreversible evolution even after the removal of the applied potential [71,72]. This susceptibility of Cu reconstruction makes it difficult to identify the active sites under realistic conditions [73]. To probe the actual active sites on Cu and to obtain fundamental evidence on structure-activity relationships, numerous in situ characterization techniques have been developed under controlled conditions during CO 2 RR [74][75][76]. Through operando electrochemical scanning tunneling microscopy (EC-STM), Kim et al. [77,78] found that a polycrystalline Cu electrode held at a fixed negative potential undergoes stepwise surface reconstruction in both alkaline and neutral electrolytes, first transforming to Cu (111) and then to Cu (100). The resulting Cu (100) surface remains stable, without further surface transformations, during the subsequent tests. In situ grazing incidence X-ray diffraction (GI-XRD) also revealed the reconstruction of polycrystalline Cu toward the (100) facet in the presence of CO 2 [71]. The degree of reconstruction increases as the applied potential becomes more negative, and the reconstructed facets are partially preserved in the subsequent anodic scanning step. The in situ characterizations described earlier indicate that the dynamic reconstruction of Cu catalysts is mainly driven by the cathode potential, but the influence of adsorbates cannot be ruled out.
Recently, the adsorption of intermediates was shown to affect the formation of preferred crystal facets with high C 2+ selectivity during the CO 2 RR. Wang et al. [79] reported a self-selective method to stabilize the crystal surface with the strongest binding to the target intermediates because the adsorption of reactants tends to reduce the relative surface energies of these surfaces. Sargent's group [80] performed Cu electrodeposition in the presence of CO 2 RR intermediates and achieved a 70% increase in the ratio of Cu (100) facets to the total surface area compared to Cu electrodeposited in the presence of HER intermediates (Fig. 3e). This Cu catalyst has a high FE of 90% for the total C 2+ products at current densities higher than 580 mA/cm 2 , and the FE of C 2 H 4 remains constant over 65 h of electrolysis. In addition to the reaction intermediates, electrolyte additives also affect the crystal facet reconstruction on Cu catalysts. Our recent work demonstrated that ethylenediamine tetramethylenephosphonic acid (EDTMPA) molecules preferentially adsorb on Cu (110) during the CO 2 RR, inducing the selective generation of Cu (110) facets with an intrinsically high *CO binding strength (Fig. 3f, g) [39]. These studies demonstrate that the adsorption of intermediates or electrolyte additives on the catalyst surface can induce the formation of specific crystal facets with high activity; these facets are beneficial for the adsorption and conversion of intermediates.

Unsaturated Coordination Atoms
After obtaining an in-depth understanding of the crystallographic dependence of the product distribution for Cu, the high proportion of unsaturated coordination atoms was also considered to contribute to the high catalytic activity of the high-index crystalline facets for C 2+ production [67,81]. CO adsorption energies on low-index Cu facets (i.e., Cu (111), Cu (100), and Cu (110)) and several regular stepped and kinked facets (i.e., Cu (211), Cu (221), and Cu (532)) were Improvement of C 2+ selectivity by crystal facets. a X-ray diffraction patterns of Cu octahedra (Cu oh ), Cu cubes (Cu cub ), and Cu spheres (Cu sph ). b Selectivity of CO 2 RR products on Cu with different facet exposures. Reproduced with permission from Ref. [64]. Copyright 2020 American Chemical Society. c Variation in C 2+ /CH 4 with the angle of the crystal orientation with reference to Cu (100). Reproduced with permission from Ref. [42]. Copyright 2003 Elsevier. d Faradic efficiencies (FEs) for producing C 2 H 4 on different catalysts. Reproduced with permission from Ref. [68]. Copyright 2020 American Chemical Society. e Wulff construction clusters of Cu with the adsorption of CO 2 RR or hydrogen evolution reaction (HER) inter-mediates. Reproduced with permission from Ref. [80]. Copyright 2020 Springer Nature. f Transmission electron microscopy (TEM) image of an electrodeposited Cu TEM grid after electrocatalysis in the electrolyte to which ethylenediamine tetramethylenephosphonic acid (EDTMPA) was added. Inset shows the corresponding electron diffraction pattern of the selected area. g Grazing incidence X-ray diffraction (GI-XRD) patterns of polycrystalline Cu electrodes before and after electrocatalysis in the electrolytes with and without EDT-MPA. Reproduced with permission from Ref. [39]. Copyright 2022 Springer Nature 1 3 examined by thermal desorption spectroscopy. The results reveal that the high-index crystalline surfaces with a lower coordination number (CN) of surface atoms have higher CO adsorption energies than the low-index ones [82]. In addition to the sites on step edges and joints, the adatoms on the surface also tend to have a higher degree of unsaturation and bind strongly with CO. Theoretical calculations are usually used to reveal the role of adparticles with low CN and surface clusters on the pristine Cu surface in concentrating *CO and facilitating the formation of CO dimers (Fig. 4a-c) [83]. In the presence of Cu adparticles, the reaction barriers between *CO and *C 2 intermediates (*OCCOH or *CCH 2 ) are dramatically reduced, enabling the selective electrosynthesis of n-propanol. DFT calculations revealed that the highly undercoordinated sites (CN < 5.9) promote C 2 H 5 OH production; moderate coordination sites (5.9 < CN < 7.5) are beneficial for C 2 H 4 production; and high coordination sites (CN > 7.5) have a strong hydrogen adsorption energy [84]. Therefore, the CN of surface catalytic active sites is closely correlated with CO 2 RR product selectivity.
The CN of the atoms on the catalyst surface can be adjusted not only by crystal surface control, but also by manipulating the macroscopic morphology. Nanostructuring of the catalyst allows for increasing the specific surface area and exposing more unsaturated coordination atoms, thereby resulting in a low CN catalyst. For example, a high density of undercoordinated Cu was formed on the surface of the Cu foil treated by anodic halogenation and subsequent electroreduction processes, resulting in a selective conversion of CO 2 to C 2+ products with a FE of 72% [85]. Machine learning was applied to predict the *CO binding energy of 10,433 surface atoms on a rough Cu model, and The adsorption energy of CO. c The reaction energies of *CO dimerization. Reproduced with permission from Ref. [83]. Copyright 2018 Springer Nature. d Images of the computationally produced electropolished Cu surface (left) and the surface after Ar plasma bombardment (right). e Predicted distribution of CO adsorption energies. Reproduced with permission from Ref. [86]. Copyright 2020 American Chemical Society. f A correlation plot between the FE of C 2 H 4 (FE C2H4 ) and partial current density of C 2 H 4 (j C2H4 ) values with the crystallite sizes. g A correlation plot between the charges of the CO adsorption peaks with FE C2H4 and j C2H4 . Reproduced with permission from Ref. [88]. Copyright 2016 American Chemical Society. h Population of surface atoms with a specific CN as a function of particle diameter. Reproduced with permission from Ref. [89]. Copyright 2014 American Chemical Society the results show that a high percentage of undercoordinated Cu sites preferentially bind to *CO (Fig. 4d, e) [86]. These results illustrate that the undercoordinated sites formed by the nanostructuring of catalysts greatly increase the activity and selectivity of C 2+ products.
In addition, the size of nanocatalysts also influences the concentration of unsaturated atoms. Consider the example of a Cu nanocrystal cube, for which the percentage of low coordination atoms on corners and edges decreases with the increasing size of the cube; therefore, the surface structure of the nanocrystal cube is closer to that of a single crystal [87]. Cu 2 O-derived Cu particles were selected to investigate the correlation between the statistical microcrystal size and selectivity of CO 2 electroreduction to C 2 H 4 [88]. With a decrease in the microcrystal size from 41 to 18 nm, the selectivity of C 2 H 4 increased linearly from 10 to 43% (Fig. 4f). The cyclic voltammetric analysis of Cu particles was performed in CO-and N 2 -saturated electrolytes, and the results revealed a linear correlation between the adsorption charge and the selectivity of C 2 H 4 . This implies that smaller particles have more sites available for CO adsorption (Fig. 4g). However, hydrocarbon selectivity is sharply inhibited when the nanoparticles are smaller than 5 nm [89]. This is because the stronger binding of *CO and *H on atoms with CN < 8 largely reduces the surface mobility of intermediates, which results in a lower probability of subsequent CO hydrogenation to form hydrocarbons (Fig. 4h). Rong et al. [90] synthesized Cu catalysts with a size gradient from single atoms to nanoclusters on a graphdiyne substrate by an alkyne-bond-directed site-trapping method. Surprisingly, the increased size remarkably improved the selectivity of CORR, showing a high C 2+ FE of 91.2% at 312 mA/cm 2 for 1-1.5 nm Cu nanoclusters with a large number of low CN atoms. These results indicate that Cu nanoparticles of size between 5 and 18 nm with moderately unsaturated coordination can exhibit suitable *CO and *H binding energies for C 2+ production. Thus, this result highlights the practicality of unsaturated coordination atoms.

Grain Boundaries
Li et al. [91] first discovered that the reduction of Cu oxides greatly increases the catalytic performance compared to the performance of pure Cu. This increase is attributed to the abundant grain boundary (GB) structure on the oxidederived Cu (OD-Cu). Electroreduction and H 2 annealing reduction were performed to reduce Cu 2 O to Cu, and similar networks of interconnected nanocrystals with distinct GBs between the nanocrystals were obtained [92]. To investigate the effect of numerous GB structures in OD-Cu on CO 2 RR and CORR performances, a series of intensive mechanistic explorations and discussions were performed by Kanan's group [34][35][36]93]. The CO temperature-programmed desorption results reveal that the binding of CO on OD-Cu is stronger than that on a Cu foil, hence improving the catalytic performance of CORR for OD-Cu (Fig. 5a, b) [35]. The direct correlation between CORR activity and GB density was further quantified to establish a design principle for solid catalysts [93]. The bulk electrolysis of Cu nanoparticles with different GB densities reveals that the specific activity of CO reduction depends linearly on the ratio of GB surface terminations (Fig. 5c). Finally, GB terminations on the electrode surface are more active for CO 2 RR than for HER; this finding was confirmed by scanning electrochemical cell microscopy and electron backscatter diffraction studies (Fig. 5d, e) [36]. The surface-terminating dislocations accumulated at the GBs modify the density of undercoordination sites, selectively increasing the activity of CO 2 RR [34]. In addition, some research groups reported that GB structures have an excellent electrocatalytic performance for CO 2 reduction to C 2+ products [94][95][96][97][98][99][100]. This performance is attributed to the facilitated CO 2 activation [94,95], increased *CO adsorption energy [96,97], and reduced C-C coupling barriers [98,99].
Because *CO with a low vibrational frequency has been observed on the surface, fragmented Cu is considered to be active for rapid CO dimerization (Fig. 5f) [101,102]. Moreover, the highly fragmented Cu also assists in clustering the binding sites of the C 1 and C 2 intermediates, thereby facilitating further coupling of these intermediates (Fig. 5g) [94,103]. Adjusting the atomic-level spacing (atomic-d S ) between Cu particles is another efficient approach to achieve the highly active and selective generation of C 2+ products [48]. Metallic Cu-based catalysts with different particle spacings were constructed by lithiation, delithiation, and electroreduction of CuO x particles. The spacing range was confirmed by examining three-dimensional tomographs obtained using a Cs-corrected scanning transmission electron microscope. Theoretical and experimental results show that a spacing of 5-6 Å maximizes the binding energy of the intermediates involved in C-C bond formation, achieving a FE of ~ 80% for C 2+ products (Fig. 5h). These results further confirm the essential effect of GBs on C 2+ activity and selectivity.

Vacancies
Vacancy engineering enables the alteration of the surface electron structure of catalysts, thereby facilitating CO 2 activation and intermediate adsorption to generate C 2+ products. For example, Cu surface vacancies with a Cu 2 S core increase the energy barrier of the ethylene pathway but leave the ethanol pathway virtually uninfluenced; thus, a selective conversion of CO 2 to polyalcohol is achieved (Fig. 6a-e) [104]. To accurately regulate the percentage of vacancies on Cu-based catalysts, the lithium electrochemical tuning method was proposed to remove anions while preserving the nanostructure of the electrode. In this method, double S vacancies are formed on the hexagonal CuS (100) surface, and the density of the S vacancies can be regulated by controlling the number of charging-discharging cycles (Fig. 6f) [21]. The unique double S vacancy structure provides efficient electrocatalytic active sites to stabilize *CO and *OCCO dimers simultaneously, and these sites facilitate the CO-OCCO coupling to form C 3 species with an FE of 15.4% toward n-propanol in H-cells (Fig. 6g). In conclusion, the local electron-rich environment at the vacancy sites favors electron transfer to the CO 2 RR intermediates, thereby Improvement of C 2+ selectivity by grain boundaries. a, b CO temperature-programmed desorption profiles of a polycrystalline Cu and b oxide-derived Cu (OD-Cu) under air oxidation at 500 °C (i.e., OD-Cu-500). Reproduced with permission from Ref. [35]. Copyright 2015 American Chemical Society. c Specific activity for CO reduction versus the grain boundary (GB) surface density at − 0.5 V vs. a reversible hydrogen electrode (RHE). Reproduced with permission from Ref. [93]. Copyright 2016 American Chemical Society. d Electron backscatter diffraction orientation map of the tested sample with GBs. Inset text and paths indicate the locations where line scans were collected. e Line scan generated from individual constant potential electrolysis across the GB at 1 atm Ar or CO 2 . Reproduced with permission from Ref. [36]. Copyright 2017 the American Association for the Advancement of Science. f CO vibrational frequency (νCO) observed during chronoamperometric scans. Reproduced with permission from Ref. [102]. Copyright 2020 the Royal Society of Chemistry. g Physical proximity of the optimal C 1 and C 2 sites that facilitate the coupling of C 1 -C 2 into C 3 products. Reproduced with permission from Ref. [103]. Copyright 2019 Springer Nature. h The FE versus the size of the atomic-scale interspace (atomic-d S ) measured at − 0.9 V versus RHE. Reproduced with permission from Ref. [48]. Copyright 2020 Wiley-VCH facilitating the coupling of the intermediates to generate C 2+ products. However, the stability of the S vacancies at cathodic potentials needs to be determined in future studies by in situ characterization.

Heteroatom Doping
Heteroatom doping has been demonstrated as an efficient way to increase active site exposure. Because of the lattice . g FE n-PrOH and FE n-PrOH /FE C1+C2+C3 of various catalysts. Reproduced with permission from Ref. [21]. Copyright 2021 Springer Nature mismatch between the dopant and Cu, the strain generated at the interface adjusts the electronic structure and the intermediate adsorption strength for CO 2 RR. Doping by p-block elements with higher electronegativity than Cu will induce the presence of oxidation states without a phase change [105][106][107]. In addition, the dopants with strong oxygen affinity facilitate the breaking of C-O bonds in *OCHCH 2 ; this condition thermodynamically favors the generation of ethylene and ethane but inhibits the formation of ethanol [108]. Zhou et al. [37] found that B doping can tune the local electronic structure of Cu via the transfer of electrons from Cu to B; thus, B doping can regulate the active site ratios (Cu δ+ to Cu 0 ). B doping also persistently stabilizes Cu δ+ to facilitate the adsorption and dimerization of CO, and thus, there is a high tendency for C 2 formation (Fig. 7a-c). To further improve the long-term stability of C 2+ production, Zn atoms were introduced in the B-doped Cu [109,110]. Recently, F-doped Cu was demonstrated to promote water activation, and the resulting local enrichment of *H facilitates the hydrogenation of *CO to *CHO, which in turn lowers the C-C coupling energy barrier and promotes the formation of C 2+ products [111]. Similarly, Cu-based catalysts derived from metal-organic framework (MOF) with different organic linkers were used to reveal the mechanism for C 2+ generation [112]. The C 2 /C 1 product ratios can be adjusted from 0.6 to 3.8 following the order NH 2− < OH − < bare < F − < 2F − , suggesting that the  (Cu(B)). c The partial current density of C 2 at different potentials on Cu(B)-2, oxidized nano-Cu (Cu(C)), and pristine Cu (Cu(H)). Reproduced with permission from Ref. [37]. Copyright 2018 Springer Nature. d Reaction barriers for C 1 -C 1 and C 1 -C 2 coupling on M-doped Cu systems calculated based on the density functional theory (DFT). Reproduced with permission from Ref. [59]. Copyright 2019 Springer Nature. e Calculated water dissociation reaction energies and hydrogen adsorption energies on various surfaces. f Product distributions of various hydroxides/oxide-modified Cu/PTFE electrodes, along with the corresponding C 2 H 5 OH/C 2 H 4 ratio. Reproduced with permission from Ref. [19]. Copyright 2019 Springer Nature increased dissociation of H 2 O by strongly electronegative groups in organic linkers favors C 2 production.
Since the electrosynthesis of C 3 products depends greatly on the presence of both C 1 and C 2 intermediates, metal doping strategies were proposed to achieve the simultaneous stabilization of these intermediates. Among metal-doped Cu candidates, Ag-doped Cu favors both C 1 −C 1 coupling and C 1 −C 2 coupling. This Ag-doped Cu has the highest FE (33%) for the reduction of CO to n-propanol [59]. The facilitation of multiple C−C coupling is attributed to the strain and ligand effects induced by Ag doping, which result in an energetic asymmetry of the adjacent Cu atoms (Fig. 7d). Compressive strain induced by doping with Ag atoms in the Cu lattice also shifts the valence band density of Cu to a deeper level [113]. This electronic structure lowers the binding energies of H and O compared to the binding energy of CO, and thus, the HER is selectively inhibited, and the generation of carbonyl-containing C 2+ products, accompanied by decreased generation of hydrocarbons, is facilitated.
Given the important role of hydrogenation reactions in multicarbon alcohol production [114,115], heteroatom doping is also proposed to modulate the *H on the surface of catalysts for facilitating the hydrogenation of intermediates after C−C coupling. Theoretical calculations revealed that Pt-or Pd-doped Cu exhibits the optimal H binding energy, which promotes the hydrogenation of C 2 intermediates to generate C 2 H 5 OH [116]. The results show that the CuPd and CuPt catalysts have an alcohol-to-ethylene FE ratio that is twice that of bare Cu. In addition, hydroxide-and oxidedoped Cu catalysts stabilized at reduction potential can also activate water and tune the surface *H coverage, thereby regulating the ethanol and ethylene reaction pathways (Fig. 7e) [19]. The increased *H is only involved in the branched reaction toward ethanol, accelerating the conversion of *HCCOH to *HCCHOH (a key intermediate for ethanol generation). Among all the hydroxide-and oxide-doped Cu catalysts, Ce(OH) x /Cu/PTFE exhibits the maximum FE of 43% for ethanol generation at a current operating density of 300 mA/ cm 2 (Fig. 7f). In conclusion, doping affects the adsorption of different intermediates, and thus, it is thus beneficial to break the linear adsorption relationship of intermediates for the selective generation of certain C 2+ products.

Cu-Based Alloys
Alloying is another effective strategy to modulate the electronic and geometrical structures of catalysts for inhibiting HER competition and promoting CO 2 RR activity [117]. The electronic structure of the catalyst is directly correlated with the binding strength of intermediates, and the geometrical structure affects the local distribution of certain intermediates at active sites. Several Cu-based alloys were extensively studied from the viewpoint of C 2+ production [118,119].
For example, He's group [120] used E-beam evaporation to fabricate thin films of CuAg to precisely regulate the stoichiometric ratio of the two elements. Thus, they built a good model to reveal the real reaction mechanisms during the CO 2 RR process. The operando synchrotron radiation-Fourier transform infrared spectroscopy (SR-FTIR) demonstrated that the CuAg bimetal greatly inhibits the formation of O−C−O intermediates and increases the coverage of *CO and *OCCO intermediates, thus promoting the C 2+ production (Fig. 8a). Moreover, Cu supported on amorphous CuTi alloys (a-CuTi@Cu) can electroreduce CO 2 to C 2−4 products, such as ethanol, propanol, and n-butanol (Fig. 8b) [121]. Theoretical simulations and in situ characterization demonstrate that the subsurface Ti atoms increase the electron density of the surficial Cu sites with improved adsorption ability of *CO intermediates. Thus, the energy barriers for the dimerization or trimerization of *CO are reduced. The function of the interface in Cu alloys in promoting C−C coupling and C 2+ formation was also confirmed by theoretical calculations [122].
To expedite the discovery of Cu-based catalysts with high C 2+ selectivity, Sargent's group [123] developed a machine learning-accelerated high-throughput DFT framework for material selection. They selected 228,969 adsorption sites from 244 different Cu-containing intermetallic compound crystals to train a machine learning model. The framework was subjected to approximately 4000 DFT simulations for CO adsorption energy calculations, and the Cu-Al alloy exhibited the highest abundance of adsorption sites and site types with near-optimal CO adsorption energies. These properties created a favorable Cu coordination environment for C-C dimerization (Fig. 8c, d). The rationally designed Cu-Al electrocatalysts show a high C 2 H 4 FE of 80% at a current density of 400 mA/cm 2 . Thus, Sargent's group [123] demonstrated the essential role of high-throughput screening based on the adsorption energy of key intermediates in the development of Cu-based bimetallic catalysts, and they accelerated the targeted design of catalysts with high C 2+ selectivity. Weitzner et al. [119] calculated the *CO dimerization energies on different Cu-based alloys, and they found that Cu-Al alloys favor the dimerization reaction of *CO. In general, alloying is a promising approach to tuning the adsorption energy of intermediates by manipulating the electronic structure of Cu catalysts. Thus, alloying is extremely helpful in facilitating the efficient generation of C 2+ products.

Chemical State Effects
The adjustment of the valence state of Cu can modify the electronic structure of Cu-based catalysts, thus improving the catalytic performance. Some studies suggest the presence of oxidized Cu plays a crucial role in converting CO 2 to C 2+ products via ex situ characterizations [95,124]. However, Cu is highly susceptible to reoxidation at the open circuit potential [71], and the real valence state of Cu under operating conditions remains controversial, limiting the mechanistic understanding of Cu-based catalysts. To clarify whether the oxidized Cu exists at an extremely negative potential during the CO 2 RR, isotopic labeling was used in the preparation process of Cu oxides, and only < 1% 18 O remanence in the 18 OD-Cu was discovered after a 5 h CO 2 RR electrolysis (Fig. 9a) [125]. Lei et al. [126] investigated the distribution of various Cu species on electrodes by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and electron energy loss spectroscopy (EELS) and found that oxidized Cu was completely reduced to metallic Cu during CO 2 RR, regardless of the initial state (Fig. 9b). Further, based on DFT modeling, Mandal et al. [127] suggested that the reduction of Cu 2 O is kinetically and energetically more favorable than that of CO 2 RR, implying that oxidized Cu should be reduced to metallic Cu before forming CO 2 RR products.
Conversely, using in situ characterization techniques, some researchers found that Cu + and residual subsurface oxygen can exist even at negative potentials [128,129]. Cuenya and coworkers [129] provided experimental evidence for the survival of Cu + species during CO 2 RR via operando X-ray absorption spectroscopy (Fig. 9c). They revealed that Cu + significantly inhibits CH 4 production and thus increases the FE of C 2 H 4 (Fig. 9d). Yang et al. [130] synthesized Cu nanocavity catalysts to confine carbonaceous intermediates, which in turn cover the catalyst surface to protect the Cu + species, and they achieved a C 2+ FE of 75.2%. In addition, the coexistence of Cu 0 and Cu + on the surface affects the CO 2 RR selectivity by changing the CO adsorption configuration. The top-adsorbed CO (CO atop )

Fig. 8 Modulation of intermediate adsorption by Cu-based alloys.
a Operando synchrotron radiation-Fourier transform infrared (SR-FTIR) spectroscopy during CO 2 RR. Reproduced with permission from Ref. [120]. Copyright 2022 American Chemical Society. b FE for various electrocatalysts at − 0.8 V versus RHE. Reproduced with permission from Ref. [121]. Copyright 2021 Wiley-VCH. c A twodimensional activity volcano plot for CO 2 RR. d A two-dimensional selectivity volcano plot for CO 2 RR. Reproduced with permission from Ref. [123]. Copyright 2020 Springer Nature intermediates are mainly observed on Cu + sites, while the bridge-adsorbed CO (CO bridge ) intermediates are mainly observed on Cu 0 sites [131]. The adjacent-adsorbed CO bridge and CO atop are negatively and positively charged, respectively, because of which they are beneficial for CO dimerization (Fig. 9e) [132]. Despite these promising results, the catalytic role of oxidized Cu still remains a debated issue, and further advances in operando techniques are required to resolve this issue.

Electric Field Effects
The stabilizing effect of electric fields on intermediates (e.g., *CO or *COCO), as demonstrated by theoretical calculations, makes it a promising method to manipulate the reaction pathways by regulating the applied electric field [47]. Sargent's group [133] investigated the effect of an intensified electric field on localized CO 2 enrichment. They used a metallic nanotip electrode to generate a locally high electric field at low overpotentials. In their work, the cations were concentrated around the sharp tip, leading to an increased local concentration of CO 2 near the active surface for CO 2 RR. The electric field-induced effect on the concentration of reagents around the tips was also analyzed using kinetic simulations, indicating that the electric field-induced aggregation effect can provide additional CO 2 for improving the CO 2 RR performance [134,135]. Recently, Liu's group [136,137] also reported that the high electric field created locally by high-curvature Cu nanoneedle (Cu NN) arrays can optimize *CO adsorption and reduce C−C coupling energy barriers (Fig. 10a). Besides, encapsulating the Cu NNs in polytetrafluoroethylene (PTFE) conformal coatings was proposed as a typical strategy to enhance the local electric and thermal fields simultaneously (Fig. 10b) [138,139]. O percentage of OD-Cu catalysts at different CO 2 reduction catalytic times. Reproduced with permission from Ref. [125]. Copyright 2018 Wiley-VCH. b High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images and electron energy loss spectroscopy (EELS) mapping of the cross section of HQ-Cu after 1 h of CO 2 RR. Reproduced with permission from Ref. [126]. Copyright 2020 American Chemical Society. c Extended X-ray absorption fine structure spectra of different catalysts under operando conditions. d Summary of the activity of plasma-treated Cu electrodes. Reproduced with permission from Ref. [129]. Copyright 2016 Springer Nature. e The DFT calculation of CO dimerization and CO hydrogenation toward CHO over mixed Cu 0 /Cu + catalyst. Reproduced with permission from Ref. [132]. Copyright 2017 National Academy of Sciences DFT calculations indicated that the enhanced electric field reduces the Gibbs free energy of the C−C coupling, while increasing the thermal field intensity boosts the reaction rate of C−C coupling; thus, a CO 2 -to-C 2 FE of over 86% can be achieved (Fig. 10c). To summarize, a locally enhanced electric field effectively enhances the adsorption strength of CO 2 and intermediates and reduces the C−C coupling energy barriers, thus enabling the highly selective production of C 2+ at the sharp tip catalytic hotspots.

Substrate Effects
Although catalyst substrates are primarily used to load catalysts and provide electron-conductive channels, their structures also affect CO 2 RR performance, resulting in the so-called "substrate effects." Specifically, the substrates can not only affect the reconstruction of Cu nanoparticles, but also have electronic interactions with active Cu sites [140]. Because of the outstanding chemical and electrochemical stability of carbon materials, they are the most widely used substrates. Since N has strong CO 2 adsorption capability, N-doped substrates can increase the local CO 2 concentration and enrich the key intermediates of C−C coupling [141]. For example, N-doped nanodiamonds (N-ND) were used as substrates for sputtered Cu nanoparticles, and an FE of 63% was realized for C 2 oxygenates at − 0.5 V vs. RHE (Fig. 11a) [142]. DFT calculations indicate that the increased *CO binding at the interface between the Cu nanoparticles and an N-doped nanodiamond substrate suppresses the *CO desorption and reduces the barriers to *CO dimerization, thus promoting C 2 production (Fig. 11b, c). The similar synergy between N-doped carbon nanospikes and loaded Cu nanoparticles was demonstrated for the highly selective conversion of CO 2 to C 2 H 5 OH [143].
The substrate also affects the reconstruction process of Cu during CO 2 RR, thereby altering product selectivity. Cu cubes on carbon paper were observed to undergo drastic morphological and compositional changes during the CO 2 RR with selective inhibition of the C 2+ products relative to CH 4 [144]. This is caused by the physical separation of the active sites on the carbon substrates with high specific surface areas. However, as the electrolysis duration increases, the C-supported Cu nanocubes also gradually agglomerate, leading to the transition of CO 2 RR products from C 1 to C 2 again (Fig. 11d) [145]. Hence, a reasonable selection of substrates and the dispersion of active sites are crucial to be able to make full use of the substrate effects.

Confinement Effects
Spatial confinement effects can change the mass transfer process and optimize the local distribution of intermediates [146]. Constructing porous Cu electrodes is a typical method of altering the mass transfer based on the specific confinement effect of pores. Increasing the pore depth results in locally higher pH and lower CO 2 concentration inside the pore, and this effect can be intensified by increasing the cathodic potential [147,148]. Changes in the local environment caused by mass transfer further affect the product distribution of the CO 2 RR. For example, C 1 products strongly depend on the local concentration of CO 2 , while C 2 products tend to be formed in a locally high pH environment [63]. Besides, confinement effects can also restrict the diffusion of intermediates via the formation of closed spaces. Cu nanocavities can limit the efflux of C 2 intermediates, thereby increasing the coverage and retention time of the intermediates required for C 3 production. Finite element simulations revealed that the cavity-opening angles play critical roles in the C 3 /C 2 selectivity ratio (Fig. 12a). An opening of ~ 60° is Fig. 10 Improvement of C 2+ selectivity by electric field effects. a Schematic diagram of C 2 formation process on the single tip of Cu nanoneedle (NN) arrays. Reproduced with permission from Ref. [137]. Copyright 2022 American Chemical Society. b The electric field distribution on a pristine Cu NN (left) and on a Cu NN with 99% PTFE coverage (Cu-PTFE NN) (right). c Product distribution and corresponding FEs. Reproduced with permission from Ref. [138]. Copyright 2022 American Chemical Society the most favorable for C 3 production, where the enrichment of C 2 intermediates is not limited by the reduced CO availability (Fig. 12b, c) [149]. Hollow porous Cu nanospheres, hollow multishelled Cu, and multishelled CuO microboxes were also reported to show similar confinement effects. The high C 2+ selectivity is attributed to the C−C coupling facilitated by the high coverage of carbonaceous intermediates in the cavity [150][151][152].
Confinement effects can also be exploited to design tandem catalysts to reutilize the products of CO-selective where FE is shown as a function of time for unsupported Cu nanocubes (U-NC) and for C-supported nanocubes (S-NC). Reproduced with permission from Ref. [145]. Copyright 2020 Wiley-VCH catalysts. For example, nanoparticles with Ag cores and porous Cu shells can simultaneously use the CO overflow from the Ag cores and confinement effects of the porous Cu channels (Fig. 12d) [153]. In this system, CO 2 is selectively reduced to CO on the Ag core at the bottom of the porous Cu channel, while a prolonged diffusion time provides more opportunities for the porous Cu shell to reduce the substrateenriched CO to C 2+ products. The C 2+ /C 1 product selectivity can be further increased by optimizing the diameter of the channels (Fig. 12e) [154]. Therefore, the confinement effects that can alter the electrolyte environment and locally enriched intermediates are a highly effective strategy to improve the utilization of carbonaceous intermediates for C 2+ products.

Local Microenvironment Effects
Regulation of the local microenvironment near the catalyst surface is another promising strategy to facilitate the electroreduction of CO 2 to C 2+ products. However, many influencing factors have complex effects on the local microenvironment [155]. In this section, the effects of electrolyte composition and concentration on the local microenvironment properties, such as local pH and electrostatic interactions, are mainly discussed.
Although KHCO 3 solution is widely used in CO 2 RR experiments to maintain neutral pH, a pH gradient exists near the electrode surface because of the OH − formed during the reduction reaction [156,157]. The local pH near the electrode plays an important role in the *CO/*H coverage, which determines whether *CO subsequently undergoes dimerization or hydrogenation. At lower pH values, *CO prefers to form CH 4 in the presence of abundant *H. At higher pH, where *H is scarce and *CO coverage is high, ethylene formation dominates over ethane formation [158]. As the rate of OH − production depends on the applied potential, it is necessary to probe the local pH near the electrode under operational conditions. Lu et al. [159] designed a flow Raman electrochemical cell to measure the local pH near the gas diffusion electrode during CO 2 RR (Fig. 13a). By analyzing the CO 3 2− and HCO 3 − concentrations in different electrolyte regions near the electrode via in situ microarea Raman spectroscopy, the corresponding pH values can be deduced from the equilibrium between HCO 3 − and CO 3 2− . A much higher local pH (11.9) than that of the electrolyte bulk was observed near the cathode surface (Fig. 13b). This accurate estimation of the local pH facilitated the further investigation of the pH effect during the CO 2 RR. In addition, anions with buffering capacity can also alter the local pH, thus affecting the pH-sensitive reactions at the catalyst Fig. 12 Improvement of C 2+ selectivity by confinement effect. a CO (left), C 2 (middle), and C 3 (right) concentrations (color scale, in millimoles) and flux distributions (arrows) on the cavity confinement structure. b Representative scanning electron microscopy images of different catalysts. c C 3 /C 2 product selectivity on different catalysts obtained from experiments and from finite element simulations. Reproduced with permission from Ref. [149]. Copyright 2018 Springer Nature. d Schematic representation and compositional characterization of the Ag core/porous Cu shell particle. Reproduced with permission from Ref. [153]. Copyright 2021 Wiley-VCH. e The C 2+ / C 1 product selectivity on the Ag@Cu catalysts with different average pore diameters at 400 mA/cm 2 . Reproduced with permission from Ref. [154]. Copyright 2022 American Chemical Society , CO 2 (aq), and OH − and b pH profile in 1 mol/L KHCO 3 with respect to the distance from the GDE surface. Reproduced with permission from Ref. [159]. Copyright 2020 American Chemical Society. c Kinetic energy diagrams of an H atom transferred from *MPA to Cu (110) and an H atom compensated from H 2 O to *MPA − H. Reproduced with permission from Ref. [39]. Copyright 2022 Springer Nature. d Partial current densities of ethylene at different potentials as a function of the electrolyte metal cation on Cu (100). Reproduced with permission from Ref. [162]. Copyright 2017 American Chemical Society. e Cyclic voltammetric curves of Cu electrodes in Ar and CO 2 atmosphere after CO 2 RR in 1 mmol/L H 2 SO 4 with or without Cs + . f Schematic representation of the interaction of the cation with the negatively charged CO 2 − intermediate. Reproduced with permission from Ref. [40]. Copyright 2021 Springer Nature surface [160]. Buffering anions increase the selectivity of H 2 and CH 4 in the following order of decreasing pK a : HCO 3 − (10.33) > H 3 BO 3 (9.23) > HPO 4 2− (7.21) [161]. This finding suggests that buffer anions with pK a lower than water can provide hydrogen directly to the electrode surface. Notably, special adsorption additives can also modulate the local proton feeding microenvironment. Our group [39] has explored boosting proton feeding toward Cu surfaces using EDTMPA additives. The adsorbed EDTMPA serves as a proton-delivering medium that accelerates the dissociation of water, providing abundant *H to assist *CO protonation toward *CHO. Typically, one H atom is transferred from the adsorbed methanephosphonic acid (*MPA, a fragment of EDTMPA) to Cu (110), and the *MPA that loses one H (*MPA − H) subsequently captures one H atom from the adjacent H 2 O molecule to become an *MPA again (Fig. 13c).
In addition to local pH, the electrostatic interactions between cations and negatively charged CO 2 intermediates in the local microenvironment also have a considerable impact on CO 2 RR. Bell's group [162] measured the CO 2 RR performance of Cu catalysts in bicarbonate electrolytes with different alkali metal cations. The activity of ethylene formation on Cu follows the trend of atomic radius increasing in order: Li + < Na + < K + < Rb + < Cs + (Fig. 13d). Theoretical calculations suggest that the presence of solvated cations in the outer Helmholtz plane stabilizes the negatively charged reaction intermediates such as *CO 2 − and *OCCO -. The difference in activity when using various cations is attributed to the broader coverage of cations as the cation size increases. Recently, Koper's group [40] found that CO 2 RR can only proceed on Cu when metal cations are present in the electrolyte (Fig. 13e). DFT simulations confirmed that the partially desolated metal cations stabilized the CO 2 − intermediates by short-range electrostatic interactions, thus allowing its reduction (Fig. 13f). These works highlight the necessity of cations and water in the electrochemical activation of CO 2 .

In situ Characterization Techniques
To deeply understand the CO 2 RR mechanisms, various in situ characterization techniques were developed to capture real-time information under actual operating conditions. To date, in situ infrared (IR) spectroscopy, in situ Raman spectroscopy, in situ mass spectrometry (MS), and in situ isotope tracer technology have been widely used to characterize the reaction intermediates to propose possible reaction pathways.
In situ IR spectroscopy was extensively used to reveal the CO 2 RR mechanisms by detecting the chemisorbed species on the electrode surface [163]. Then, the adsorbed species with unique vibrational modes such as *CO, adsorbed CO 3 2− , and *CO dimers can be analyzed in real time during the CO 2 RR. Hori and coworkers [164] first identified *CO at a Cu electrode during CO 2 RR by performing in situ IR spectroscopy. Heyes et al. [56] detected the vibration bands of *CO and *H on Cu at 2060 and 2090 cm −1 , respectively, by performing surface-enhanced infrared absorption spectroscopy (SEIRA). Further, peak deconvolution revealed that *H can partially replace *CO, while *CO cannot replace *H (Fig. 14a). Moradzaman et al. [165] assigned the band at ~ 1610 cm −1 to the CO 2 -dimer radical anion, which subsequently decomposed into CO 3 2− and *CO. The adsorbed carbonate was distinguished from the dissolved carbonate because of its strongly potential-dependent peak position in the range of 1510-1570 cm −1 . Koper's group [166] found that the vibrational bands at 1191 and 1584 cm −1 corresponding to the C-O-H and C=O stretching modes of the hydrogenated dimer (*COCOH), respectively, suggesting that the stable adsorption of *COCOH is responsible for the high C 2 H 4 selectivity on Cu (100).
In situ Raman spectroscopy can effectively complement IR spectroscopy because of its relatively low response to water, which is usually used to detect varying rotational or vibrational states of different molecules on the catalyst surface [167]. The high-speed data acquisition in the case of Raman spectroscopy helps distinguish the real-time reaction intermediates. *CO is the most common CO 2 RR reaction intermediate observed by in situ Raman spectroscopy, and a low-intensity carbonate signal can also be detected [168,169]. Zhan et al. [170] found that the ratio of the intensity of the Cu-CO stretching band to that of the CO rotational band follows a volcano-shaped trend, which depends on the potential-dependent surface coverage of CO. At high CO surface coverage, the mixed adsorption conformation of CO at the top and bridge sites kinetically and thermodynamically favors C-C coupling, thereby effectively improving the C 2+ selectivity. An et al. [101] used subsecond in situ time-resolved surface-enhanced Raman spectroscopy to reveal the dynamics of CO intermediates during the CO 2 RR on Cu. A highly dynamic *CO with characteristic vibration below 2060 cm −1 is associated with C-C coupling and C 2 H 4 production, while the isolated and stationary *CO with a distinct vibration at 2092 cm −1 favors the generation of gaseous CO (Fig. 14b).
Quantitative isotope measurements are indispensable for understanding the CO 2 RR mechanisms at the molecular and atomic levels. Because of the different kinetic radii of different isotopes, the kinetic isotope effect (KIE) is commonly used to investigate the effect of isotopic substitution on the reaction rate to verify the potential RDS. In addition, the isotopic labeling method is usually used to reveal the origin of the intermediates or final products or both by detecting the location of the isotopically labeled atoms, thus providing suggestions of potential reaction pathways [171].
For the precise quantification of different isotopic atoms, nuclear magnetic resonance and MS are usually applied in the isotopic analysis. Ma et al. [111] reported that fluorinemodified Cu promotes the activation of water supported by a drastically reduced KIE of H/D (H/D is defined as the ratio of the rate of ethylene formation in H 2 O and D 2 O), and a KIE value close to 1 indicates that dissociation of H 2 O is no longer involved in the RDS. Ager and coworkers [172] used 12 CO 2 / 13 CO co-feed experiments to identify three productspecific active sites on OD-Cu electrocatalysts; these three sites lead to the production of ethylene, ethanol or acetate, and n-propanol (Fig. 14c). Chang et al. [62] elucidated the  [56]. Copyright 2016 American Chemical Society. b Comparison of the steady-state Raman spectra during reduction at various potentials. Reproduced with permission from Ref. [101]. Copyright 2021 Wiley-VCH. c Hypothetical scenario for the reduction in a mixture of 13 CO and 12 CO 2 . Reproduced with permission from Ref. [172]. Copyright 2019 Springer Nature. d, e MS signal of the relative abundance of the liquid phase products generated on Cu. The solid lines represent the relative abundances of the liquid phase products in a traditional H-cell when analyzing the bulk electrolyte. Reproduced with permission from Ref. [61]. Copyright 2018 American Chemical Society formation mechanism of the C 3 products by combining isotopic labeling with in situ spectroscopy. Since only 36% of n-propanol is derived from the cross-coupling of CO with acetaldehyde, the adsorbed methylcarbonyl intermediate is proposed to be the bifurcation point of the reaction pathway for the C 2 and C 3 products.
In situ MS is typically used to achieve the real-time detection and semiquantification of gaseous and evaporable liquid products based on mass-to-charge ratios. By improving the time resolution of in situ MS, the formation of intermediates and products under transient operation can be detected, and the potential correlations between them can be established [173,174]. Koper and coworkers [54] proposed two possible reaction mechanisms for ethylene formation using online electrochemical MS. In one pathway, C 2 H 4 shares an intermediate with CH 4 on the Cu (111) and Cu (100) surfaces, while in the other pathway, the selective reduction of CO to C 2 H 4 occurs on Cu (100) at a relatively low overpotential, presumably via the formation of a CO dimer. Clark et al. [61] found that the ratio of aldehydes to alcohols detected in the region close to the Cu electrode was much higher than that in the bulk electrolyte. With increasing overpotential, the relative abundance of ethanol rises along with a decrease in propionaldehyde, suggesting that acetaldehyde is a bifurcated intermediate for ethanol and propionaldehyde formation, while n-propanol is formed via propionaldehyde (Fig. 14d, e).

Summary and Perspectives
This review focuses on strategies to improve the selectivity of CO 2 RR to C 2+ products through manipulation of the intermediates on the surface of the Cu-based catalysts. A systematic discussion of C 2+ formation mechanisms revealed that the binding energies and competitive adsorption of intermediates on the catalyst surface are crucial for C 2+ selectivity as these factors can affect the C-C coupling and hydrodeoxygenation processes. Subsequently, we provided deep insights into the manipulating strategies and factors influencing intermediate adsorption, covering topics such as surface structural effects, introduction of additional elements, chemical state effects, electric field effects, substrate effects, confinement effects, and local microenvironment effects. Table 1 summarizes the electrochemical CO 2 -to-C 2+ performance of Cu-based electrocatalysts according to the intermediate manipulation strategies. Through an optimal combination of these strategies, C 2+ production can be further improved in practical applications. For example, the simultaneous utilization of doping and confinement strategies improves the selectivity of C 2+ products without any catalytic activity degradation [107]. Specifically, quasigraphitic C shells can stabilize the crystal size of Cu nanoparticles based on the confinement effects, while doping with p-block elements can drastically improve the C 2+ selectivity and cathodic power conversion efficiency via the modulation of the binding properties of the Cu catalysts.
As the CO 2 RR involves complex reaction steps and numerous intermediates, the use of advanced in situ characterization techniques is important for elucidating the relationship among structures, intermediates, and performance. In particular, the operando characterization of intermediates is the key to accurately revealing the reaction pathways and RDS of the CO 2 RR, and thus, such studies can provide useful guidance for optimizing catalysts. Therefore, we comprehensively discussed various in situ characterization techniques to identify the intermediates and their adsorption states under operating conditions. These techniques included IR spectroscopy, Raman spectroscopy, MS, and quantitative isotope measurements. In addition, theoretical calculations and simulations that closely match the actual operating conditions are also irreplaceable critical tools for understanding the catalytic mechanisms in detail. By considering the adsorption intermediate as a bridge, the correlation between the catalyst structure and catalytic performance was established to promote the practical application of CO 2 RR.
Although the activity and selectivity of CO 2 reduction to C 2+ products on Cu-based catalysts have greatly improved in recent years, their in-depth mechanisms are still unclear, and the industrial application of the CO 2 RR is still challenging. On the one hand, most reaction mechanisms for the conversion of CO 2 to C 2+ products are presumed based on the final product distribution. Such an inference from the effect to the cause makes it difficult to objectively and accurately understand the actual reaction process. On the other hand, for meeting the industrial requirements for the CO 2 RR, the focus is on greatly reducing the cost of the catalytic systems rather than further improving catalyst selectivity, activity, and stability. Note that cofeeding of CO 2 with other feedstocks provides a promising pathway for the generation of high-value-added products, and this pathway should be actively explored. To this end, efforts should be focused on the following aspects in future research: tion characterization techniques to accurately detect the intermediates [177,178].
2) Rational design of electrolyzers. Recently, it was established that high-efficiency gas diffusion electrodes and membrane electrode assembly are more promising electrolytic devices for the industrial application of the CO 2 RR [179]. The optimization of electrolyzers can not only reduce the working voltage, but also significantly increase the current density (by up to 1.6 A/cm 2 ), which is close to the technoeconomic requirements for the industrial application of CO 2 RR [111,180]. However, the problems of flooding and salt precipitation that occur at gas diffusion electrodes over long-term electrolysis should be addressed to avoid a rapid decrease in selectivity and current density. Moreover, the formation and crossover of carbonates in alkaline electrolytes also affects the single-pass conversion of CO 2 and contaminates the O 2 stream released from anodes, thereby increasing the cost of feedstock and separation. Impressively, the solid-electrolyte reactor, which decouples the ion conduction and product collection, successfully recovers the substantial carbon loss that occurs during CO 2 RR electrolysis and avoids the requirement for cost-intensive downstream product-separation processes [181][182][183]. Besides, acid electrolytes [184] and pulsed electrolysis [185] were also used to avoid carbonate accumulation, offering promising ways to realize longterm electrolysis. 3) Cofeeding of CO 2 and other reactants. Currently, highpurity CO 2 , which requires considerable energy to capture and purify, is the main feedstock used in experimental research. However, because of its high cost, high-purity CO 2 is impractical for industrial applications of CO 2 RRs. In this situation, cofeeding CO 2 with other chemicals can decrease the cost of CO 2 purification and/or create high-value products, in turn improving the technical economy of the catalytic systems. For example, flue gas is a potential source of CO 2 and can be adopted to reduce the purification cost; however, more deliberate catalyst design and pressurization are required to overcome the competition reaction caused by the oxidizing gases [186,187]. Besides, the formation of C-N compounds by the coelectrolysis of CO 2 and nitrogencontaining reactants also holds great promise [188][189][190]. For example, the coactivation of N 2 and CO 2 enables the electrosynthesis of desirable urea [191].
In conclusion, the CO 2 RR provides a promising solution for converting CO 2 emissions into valuable feedstocks. Encouragingly, much progress has been achieved toward the electroreduction of CO 2 to C 2+ products using Cu-based catalysts during the last decades. However, many problems still persist for industrial applications. Together with a mechanistic exploration supported by advanced in situ characterization techniques and theoretical simulations, the optimization of electrolyzers and CO 2 feeding modes is expected to further advance the industrial applications of the CO 2 RR.

Declarations
Conflict of interest The authors declare that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.