Optimization Strategies for Selective CO2 Electroreduction to Fuels

Capturing CO2 from the atmosphere and converting it into fuels are an efficient strategy to stop the deteriorating greenhouse effect and alleviate the energy crisis. Among various CO2 conversion approaches, electrocatalytic CO2 reduction reaction (CO2RR) has received extensive attention because of its mild operating conditions. However, the high onset potential, low selectivity toward multi-carbon products and poor cruising ability of CO2RR impede its development. To regulate product distribution, previous studies performed electrocatalyst modification using several universal methods, including composition manipulation, morphology control, surface modification, and defect engineering. Recent studies have revealed that the cathode and electrolytes influence the selectivity of CO2RR via pH changes and ionic effects, or by directly participating in the reduction pathway as cocatalysts. This review summarizes the state-of-the-art optimization strategies to efficiently enhance CO2RR selectivity from two main aspects, namely the cathode electrocatalyst and the electrolyte.


Introduction
Global warming caused by massive greenhouse gas emissions has brought many environmental challenges, including sea level rise, species extinction, and climate change [1,2]. According to the State of the Climate in 2019, the average global temperature has increased by ca. 1.0 °C compared with the pre-industrial age because of the increased CO 2 concentration in the atmosphere [3]. Two strategies have been proposed to decrease the CO 2 concentration in the atmosphere: storing and utilizing CO 2 [4,5]. Compared with CO 2 storage, CO 2 utilization has received greater research attention in recent years because of its cost-effectiveness. CO 2 utilization can be divided into two methods: direct application of CO 2 and conversion of CO 2 into other chemicals, such as fuels. The direct application of CO 2 is employed on a small scale to improve oil recovery [6] and act as a supercritical liquid [7]. Therefore, the chemical conversion of CO 2 accounts for the majority of CO 2 utilization. Important chemical feedstocks, such as urea and salicylic acid, have been synthesized through thermal or catalytic processes using CO 2 as the raw material [8,9]. Moreover, renewable energy sources that are inherently hard to store, such as solar, wind, and geothermal, can be used to drive CO 2 conversion, providing an ideal way of sustainable development. Among numerous CO 2 conversion methods, electrocatalytic CO 2 reduction is the most promising one because of its ambient reaction conditions and simple device fabrication. However, the advancement of electrocatalytic CO 2 reduction reaction (CO 2 RR) still faces several challenges, including poor CO 2 solubility, high activation energy, low energy efficiency, and low selectivity [10]. In addition, the deactivation of electrodes in less than 100 h largely restricts the practical use and technological commercialization of CO 2 RR [11]. The poor stability of CO 2 RR has several causes, depending on the type of cathode. For example, noble metals suffer from CO poison during the reduction process [12]; metallic chalcogenides are easily reduced at the reduction environment [13]; and agglomeration of catalysts with high dispersion decreases active site concentration [14]. The most problematic one is the low selectivity, which results in costly separation steps and energy consumption at downstream. In general, the CO 2 RR products that have been reported so far include C 1 products (e.g., CO, HCOOH, CH 3 OH, HCHO, CH 4 ), C 2 products (e.g., C 2 H 4 , C 2 H 6 , C 2 H 5 OH, H 2 C 2 O 4 , CH 3 COOH), and C 2+ products (e.g., n-propanol). The number of transferred electrons required to form these products and their corresponding standard redox potentials in acid and base electrolytes are listed in Table 1.
Although hydrocarbons and alcohols are thermodynamically favored, CO and HCOOH/HCOO − are kinetically preferred because they require fewer electron transfers to complete the reaction (2e − products) [11]. As a result, CO and formic acid, as the most common CO 2 RR products, usually show high selectivity using either metal or metal-free cathodes [15][16][17][18]. Meanwhile, products involving > 2e − transfer are prone to desorbing from the electrode surface before deep reductions occur, leading to low selectivity. For example, the simplest C 2 product of C 2 H 4 (12e − ) has a maximum Faradic efficiency (FE) of 72% [19], whereas C 3 products such as n-propanol (18e − ) have an FE less than 20% [20]. Despite their low conversion efficiency, products with > 2e − transfer are still the preferred fuels owing to their higher energy density than CO and formic acid [21,22]. Furthermore, due to the similar thermodynamic potentials, another competitive cathodic reaction for CO 2 RR in the presence of water, i.e., hydrogen evolution reaction (HER), concurrently occurs and affects the overall CO 2 RR selectivity [23]. The selectivity of CO 2 RR is influenced not only by the electrocatalysts but also by the electrolytes. The two factors interplay with each other. The most important factor dominating CO 2 RR activity and selectivity is the cathode material. Given the difficult activation of CO 2 , a suitable electrocatalyst can efficiently reduce the activation energy barrier and thus decrease the overpotential. As to the electrolyte, a suitable solvent presents high CO 2 solubility; however, the ion species of the supporting electrolyte greatly decide the buffer capacity and solution pH. The cations adsorbed on the cathode-electrolyte interface via electrostatic interaction can modify the superficial environment and thus manipulate the adsorption affinity of CO 2 and intermediates, leading to varied product distribution [24]. This review focuses on the two aforementioned factors and discusses their mechanism on the reaction pathways. Hopefully, our perspective could serve as a reference for the regulation of CO 2 RR selectivity and for the future design of advanced electrolysis systems for CO 2 RR.

Electrocatalyst
Cathode materials largely determine the activity and selectivity of an electrocatalytic cell. CO 2 RR electrochemical catalysts can be categorized according to various perspectives, such as homogeneous and heterogeneous catalysts or metallic and non-metallic catalysts. Widely used in mechanistic exploration, homogeneous catalysts can exclude environmental disturbance, including supporting electrolytes, temperature, pressure, and disclose the intrinsic reactivity of the electrocatalyst. However, most homogeneous catalysts can hardly overperform heterogeneous catalysts because of their sparse and discrete active sites, low stability, high synthesis cost, and toxicity. Therefore, this section focuses on the design strategies, including composition manipulation, morphology control, surface modification, and defect engineering, of high-performance heterogeneous catalysts.
Compared with pure metals, alloys and composites undergo a more intricate catalytic process and show better catalytic properties. Bi-or multi-metal catalysts present a synergistic effect [27][28][29]. For instance, Au-deposited Cu displays a 100-fold enhanced catalytic activity for converting CO 2 into C 2+ products compared with bare Cu foils [30]. Benefiting from the bimetallic synergy, the catalytic process is divided into two regionalized steps, which greatly boost the reaction efficiency. First, CO 2 is reduced into *CO at Au active sites. Then, the *CO species transfer to nearby Cu active sites and undergo further reduction into multi-carbon products, such as C 2 H 5 OH and n-propanol. In this work, Au and Cu were delicately integrated into a novel bimetal catalyst, and both components participated in the catalytic process, achieving a high catalytic performance. A similar tandem catalysis was also observed on the CuZn catalyst, in which *CO was stabilized on the catalyst surface and transferred from Zn atoms to adjacent Cu sites for further dimerization and protonation to yield C 2 H 4 , with FE up to 33.3% at − 1.1 V versus RHE [31]. Our group achieved a tunable CO 2 /H 2 ratio of syngas (i.e., CO 2 and H 2 ) on partial Zn-substituted CdS-amine inorganic-organic hybrids at − 1.16 V versus RHE [32]. The amine provides the active sites for CO formation, whereas Zn substitution boosts HER. Therefore, different CO 2 /H 2 ratios can be obtained by altering the Zn concentration. Aside from the tandem catalytic effect, the introduction of a secondary component can also influence the surface electronic structure and thus change the binding strength of intermediates. Previous studies reported that the formation of C 2 H 4 is more favored than that of C 2 H 5 OH on Cu.
Intermediates of the C 2 H 4 pathway, such as *CCH, possess higher π electron density than those of the C 2 H 5 OH pathway (*HCCHOH) at the synchronous step, which requires dense available binding sites (Fig. 1a). Hence, Li et al. [33] selected Ag, which has a weaker bonding ability, to combine with Cu to reduce the number of strong binding sites. The Ag/Cu catalyst can successfully switch the selectivity from C 2 H 4 to ethanol, achieving a FE of 41% and a current density (j) up to 250 mA/cm 2 at − 0.67 V versus RHE. Such synergy can also be observed in metal oxide compounds. For example, Schreier et al. [34] reported that CuO nanowires (NWs) coated with an atomic layer of SnO 2 outperform pure CuO, showing a FE of ca. 90% toward CO. The enhanced selectivity is attributed to the decreased binding strength of *H and *CO: The former leads to the suppression of HER, and the latter promotes the fast release of *CO before further reduction.
In other cases, some components in a composite may not directly participate in the reaction but improve the physical or chemical properties of the active component to increase the overall reactivity. For example, carbon-based materials, as the most popular supporting materials, can improve the electrical conductivity, increase the specific surface area, enhance dispersity, and improve stability [35,36]. As an example, Luc et al. [27] prepared a core@shell nanostructured catalyst, AgSn@SnO x , in which the bimetallic core improves the electrical conductivity of the electrocatalyst.
Interestingly, the catalyst exhibits tunable FEs toward formic acid by varying the Ag/Sn ratio of the core, and a volcanolike relationship exists between Ag/Sn ratio and FE HCOOH . Two balance factors are responsible for the volcano-shaped correlation. On the one hand, the electrical conductivity of the electrocatalyst decreases with increasing Sn concentration, which reduces the overall catalytic activity. On the other hand, a lower Ag/Sn ratio means a thicker SnO surface layer availing the stabilization of *OCHO intermediates, resulting in a lower rate-determining step (RDS) (i.e., *CO 2 to *OCHO) reaction barrier for the HCOOH pathway. However, for the CO pathway, the composition effect on key intermediates *CO 2 and *COOH offsets with each other. Thus, the RDS energy barrier (i.e., *CO 2 to *COOH) remains the same.
Heteroatoms with electron-donating or electron-withdrawing ability can create abundant positively and negatively charged sites. For example, Zhou et al. [37] utilized B dopants to alter the local electronic structure of Cu, stabilizing the surface Cu δ+ sites. A FE of ~ 79% toward C 2 products and a durability of 40 h were obtained at − 1.1 V versus RHE. Moreover, as pure carbon materials have limited electrocatalytic CO 2 RR activity, heteroatoms, such as N, B, and S, are frequently used as dopants to optimize their surface electronic structure and achieve high catalytic properties. N-doped carbon (NC) has been extensively investigated as a material support and catalyst [23,[38][39][40]. Nitrogen in a carbon skeleton has three major forms, namely, pyridinic N, pyrrolic N, and graphitic N. The electrocatalytic CO 2 RR on NC greatly depends on the form and density of the N atoms. For example, Cui et al. [16] developed a steam etching  (111) surface. Binding illustration for b Cu and c Ag/Cu catalysts to produce ethylene and ethanol, respectively. Reproduced with permission from Ref. [33]. Copyright 2019 American Chemical Society method to partially wipe off pyridinic N and graphitic N but leave pyrrolic N on NC-wrapped CNT (CN-CNT). The as-prepared CN-CNT exhibits a FE of ca. 88% toward CO at − 0.5 V versus RHE with a ~ 20% increase compared with the untreated catalyst. Wu et al. [41] prepared N-doped graphene quantum dots, which exhibit a catalytic performance comparable to that of Cu with 45% C 2 selectivity (including C 2 H 4 and C 2 H 5 OH), instead of the CO selectivity of most non-metallic electrocatalysts [42][43][44]. The catalytic activity of the three N configurations can be arranged in the following order: pyridinic N > graphitic N > pyrrolic N [45][46][47]. Pyridinic N is the most active center. For instance, Liu et al. [43] synthesized a series of 3D N-doped graphene nanoribbon networks with different N site distributions. XPS spectra suggested that the FE CO is linearly related to the content of pyridinic N. Furthermore, j shows a conspicuous decline when phosphate is used as a poison agent to selectively adsorb onto the pyridinic N sites. These observations, combined with density functional theory (DFT) calculations, indicate that the Lewis-acidic pyridinic N is mainly responsible for the high catalytic performance. Our group [48] also showed that the pyridinic N in the Ar-plasma-treated N-doped CNT arrays is the active site for CO formation on the basis of the correlation between structure variation and FE CO . Similarly, various O defects on carbon materials were also investigated. Yang et al. [49] synthesized singlelayer graphene nanodisks with different dominating types of O-containing surface groups, such as hydroxyl (-OH), epoxide (C-O-C), carbonyl (C=O), and carboxyl (-COOH). The catalytic performance was determined by the synergistic effects among all O-containing groups on the basis of their DFT computation. Therefore, unlike NC, in which each single N form alone can catalyze CO 2 RR, increasing the amount of O species as much as possible is more crucial. In addition, the reduced O-containing groups under reduction potentials can be easily regenerated by electrochemical in situ oxidation.

Morphology Control
The shape, size, and dimensionality of the nanostructures can easily change the intrinsic activity and selectivity of the catalyst toward CO 2 RR [50][51][52][53][54][55][56]. Many CO 2 RR catalysts show facet-dependent selectivity, i.e., the formation of target products can be controlled by synthesizing single crystals with specific exposed facets or by increasing the exposure of certain facets [57]. For example, the formation of C 2 H 4 is more favored on the (100) facets of Cu, whereas that of CH 4 is more favored on the (111) facets instead [58]. The Cu(711) facets, which consist of Cu (100) terraces and a small number of (111) steps, show the highest C 2 /C 1 ratio in product distribution, even superior to that of Cu(100) [59]. In addition, the relationship between products and operating potential suggests that C 2 H 4 forms at a lower overpotential than CH 4 [60]. In DFT calculations, the *CO dimerization on the Cu(100) surface has a lower energy barrier than those on (111) and (110), and high *CO coverage is beneficial for reducing the energy barrier. The binding strength of key intermediates, such as *CO, over the Cu electrode, must be strong enough to facilitate the subsequent proton-electron transfer steps but not too strong to poison the active sites, creating an optimal adsorption and desorption environment for *CO [61]. To be more specific, Luo et al. [62] proposed two protonation pathways of *CO that lead to selectivity difference on Cu (111) and Cu(100) facets on the basis of DFT simulation. Hydrogenation on the Cu(111) facet occurs at the O end, forming *COH, whereas that on the Cu(100) facet occurs at the C atom instead, forming *CHO. Their proposal for the CH 4 pathway on the Cu(111) facet has been adopted by other theoretical and experimental researchers [63,64]. However, the reaction mechanism on Cu(100) is still under debate. As stated by Cheng et al. [65], two adjacent *CO will dimerize into *OCCO, which is the RDS of C 2 H 4 formation, followed by a protonation process to form *OCCOH (Fig. 2). This mechanism was also verified by Koper et al. [66], who directly detected *OCCOH intermediates on the Cu(100) facet for the first time at low overpotentials ranging from + 0.1 V to − 0.15 V versus RHE using in situ Fouriertransform infrared spectroscopy. The facet dependency of Cu was also used to achieve exclusive C 2 selectivity via tailoring surface morphology [54]. For instance, the nanoporous Cu prepared via the alloying-dealloying method can suppress CH 4 generation to less than 1% and show 35% FE C2H4 at − 1.3 V versus RHE because of the extensive exposure of specific facets, especially (100) facets, together with the high concentration of catalytically active step and edge [67]. Nevertheless, further evidence is needed to attest current conjectures on the facet-dependent catalytic properties of Cu.
Ma et al. [68] found that the morphology differences can alter the local pH value near the electrode surface, thus changing product distribution based on their pH sensitivity, which will be discussed later. In their work, Cu NW arrays with diverse lengths and densities were prepared by simply controlling the reaction time. The longest and densest Cu NWs had the best C 2 H 4 selectivity (Fig. 3a). According to the authors, the morphology-related catalytic performance stems from the sluggish superficial mass transfer with the increase in length and density, which impedes not only the entry of bicarbonate as a buffering agent but also suppresses the diffusion of OH − by-products, finally leading to a significant local pH rise to improve C 2 H 4 selectivity (Fig. 3b). Developing materials with ultrahigh specific surface area can realize high reactivity because of the increased active sites, especially edges and steps. For example, Dutta et al. [69] synthesized a bifunctional Ag nanofoam cathode to selectively catalyze the formation of CO and CH 4 at low and high overpotentials, respectively. Notably, this Cu-like selectivity at a high overpotential is different from the commonly observed CO selectivity on Ag electrodes. The abundant coordination-unsaturated sites in the nanofoam could enhance the CO binding strength, which greatly increases the surface coverage and the resident time of *CO intermediates, favoring the further reduction to CH 4 . Under proper morphology control, catalysts with similar compositions but different shapes and sizes present pronounced CO 2 RR activity and selectivity. For instance, Kortlever et al. [70] synthesized a two-dimensional PdAu catalyst with a Pd-excessive surface alloy structure. The unique surface structure of PdAu alloy enables the formation of C 1 and C 2 products and C 3 -C 5 hydrocarbons and oxygenates at the onset potential of − 0.8 V versus RHE. Unfortunately, the detailed mechanism was not fully explored. By contrast, a different selectivity was observed on one-dimensional twisted PdAu NWs by Zhu et al. [53]. Results of transmission electron microscopy (TEM) and high-angle annular dark-field (HAADF) STEM ( Fig. 4a-p) showed that the nanowires are metastable core-shell structures composed of pure Pd monolayer shell and PdAu core with abundant grain boundaries (GBs). Benefiting from the GB-rich surface, PdAu NWs present a FE CO of ca. 94% at − 0.6 V versus RHE (Fig. 4q). In situ spectroscopic investigation and DFT calculations revealed that the sluggish activation of CO 2 to *COOH is accelerated by the GBs, leading to a 0.23 eV decrease in the *COOH binding energy at the Pd/PdAu (110) facet compared with Pd (111). The weakened linear adsorption of *CO by incorporating Au into Pd also accounts for the high CO selectivity at a low overpotential (Fig. 4r).
Downsizing the metal bulk to the atomic-level site is a promising approach for material optimization [71,72]. As we mentioned before, bulk Zn was classified as a metal mainly yielding CO, while in the case of Zn single-atom catalyst, unexpectedly high CH 4 selectivity with a FE of 85% and a j CH4 of − 31.8 mA/cm 2 were achieved at − 1.8 V versus saturated calomel electrode [73]. The remarkable catalytic performance can be well maintained after undergoing a 35-h-long CO 2 electrolysis. DFT simulations revealed that the first activation step is the formation of *OCHO (∆G 0.46 eV) rather than regular *COOH (∆G 1.2 eV) (the key intermediate of producing CO), where the O atom bonds to Zn instead of C atom. It was further confirmed by in situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS), which detected *OCH 3 and *OCH 2 species. Atomically dispersed transition metals supported on NC have been exploited as promising nonprecious electrocatalysts for converting CO 2 to CO, in which nitrogen-coordinated single-atom metal (M − N x ) moieties serve as the active sites [35,36,[74][75][76]. Zheng et al. [74] developed a facile method to prepare Ni SAC catalysts on  [65]. Copyright 2020 National Academy of Sciences commercial carbon black with FE CO close to 100% and j CO above 100 mA/cm 2 . Pan et al. [75] reported that engineering the local support structure of SAC can significantly boost the catalytic performance. The Fe-N 4 moieties supported at the holes of the graphene basal plane (edge-hosted Fe-N 4 ) are more catalytic toward CO 2 -to-CO conversion than the pore-deficient graphene-supported counterparts (bulk-hosted Fe-N 4 ). DFT and density-of-state studies suggested that the incorporation of pore edges can downshift the d-band center of Fe-N 4 moieties, which will weaken the binding strength between Fe and *CO and thus facilitate CO desorption.

Surface Modification
The introduction of functional groups, such as organic ligands, to manipulate the catalytic reactivity is a common method for cathode design [77]. These surfactant-like functionalized species can tune the polarity and acidity of the interfacial environment, which is directly related to the adsorption affinity of Lewis-acidic CO 2 reactants and intermediates. For example, Fang and Flake [78] reported that Au electrodes modified by three ligands exhibit disparate product distribution. In specific, the 4-pyridinylethanemercaptan (4-PEM)-modified Au electrode allows the formation of CO and H 2 . By contrast, the 2-mercaptopropionic acid (2-MPA)-modified Au electrode only yields H 2 . The high selectivity of 4-PEM-modified Au can be explained by a proton-induced desorption mechanism (Fig. 5a). First, a proton from the electrolyte is reduced to *H on the Au surface, which is attacked by CO 2 to form *OCHO. Then, the protonated pyridine of 4-PEM donates a proton to the adjacent O of *HCO 2 , together with an electron delivered from the cathode to release the formic acid. Therefore, the acidity of the surface ligand, as an essential H donor in CO 2 reduction, largely determines the catalytic performance, which is consistent with the pK a order of the three ligands, CYS > 4-PEM > 2-MPA. The moderate proton-donating ability of 4-PEM generates a delicate balance: The surface proton concentration is neither too high to favor the HER nor too low to affect the hydrogenation of *OCHO. Similarly, Wu et al. [79] reported the switchable production of CO and HCOO − at different potentials over cetyltrimethylammonium bromide (CTAB)-capped Au/SnO 2 . Interestingly, only H 2 is generated on capping-free Au (CF-Au) and citrate-capped Au (Cit-Au). Comparing the FE H 2 on Au/SnO 2 with the FEs measured on Au and SnO 2 ( Fig. 5b-g), CTAB-Au presents a more obvious deviation, indicating the existence of the strongest ligand-metal-oxide interactions to suppress HER.
Cu electrodes can also be modified with polymer or organic molecules to manipulate CO 2 RR activity and selectivity. For example, Wei et al. [80] reported that the surface functionalization of Cu by polyaniline (PANI) can accelerate the reaction rate and substantially increase the C 2+ hydrocarbon FE from ~ 15% to 60% at − 1.1 V versus RHE in the KHCO 3 electrolyte and up to 80% C 2+ hydrocarbon FE at − 1.2 V versus RHE, 40% of which comes from C 2 H 4 . Unlike the aforementioned CATB-Au/SnO 2 [79], the PANI used in this work does not directly participate in CO 2 RR. According to earlier work from Zheng et al. [81], the N atoms in PANI show a strong Lewis acid-base interaction with CO 2 , which greatly enhances CO 2 adsorption, improving the catalytic properties of Cu, Pd, and Pt. Wei et al. [80] further confirmed the role of PANI via ATR-SEIRAS. Their results showed that the PANI-modified Cu presents higher coverage of linear adsorbed CO species (CO L ) than pristine Cu foil, which is more prone to dimerize into *OCCO than bridge-adsorbed CO (CO B ). This modification strategy is also applicable for Cu nanoparticles, which not only enhances their reactivity but also prolongs their lifetime. Very recently, Li et al. [19] have conducted a systematic investigation on the surface modification of Cu with a series of functional organic molecules. A series of Cu electrodes coated with an organic film was prepared via electrodeposition of N-arylpyridinium precursors. Interestingly, the relationship between Bader charge (BC), which indicates the electron-donating capacity of tetrahydro-bipyridines, and FE C 2 H 4 conforms to a volcano-shaped correlation (Fig. 6a). The FE C 2 H 4 and CO a /CO b ratio (i.e., top-/bridge-adsorbed CO) also fit in the volcano relationship, which is consistent with the trend of BC-FE correlation (Fig. 6b). The two correlations suggest that organic additives with strong electrondonating ability have a high CO a /CO b ratio. According to DFT calculations, the dimerization of CO a and CO b to produce C 2 H 4 has the lowest energy barrier of 0.72 eV, which is superior to that of a pair of CO b and CO a . Therefore, neither CO b nor CO a will lead to unfavored homogeneous coupling, rationalizing the existence of the optimal CO a /CO b ratio. Furthermore, the authors proposed that the tetrahydro-bipyridine film delivers its electron density to *CO to enhance CO adsorption and increase the possibility of dimerization. Such mechanism was verified by the excellent catalytic performance of N-aryl-dihydropyridine-based oligomer-modified Cu, which shows an unprecedented high FE of 72% toward C 2 H 4 and a high j of 232 mA/cm 2 at − 0.84 V versus RHE in KHCO 3 solution.
Aside from the organic ligands, some superficial inorganic small molecules are also efficient in boosting CO 2 catalytic activity and selectivity. For instance, the pore surface of metal organic frameworks modified by hydroxyl presents a stronger adsorption affinity toward CO 2 because of the electron-donor character of the oxygen atom [82]. As for the hydrophobic SiC quantum dot (SiC QD) cathode, surface hydroxylation (SiC-OH QDs) increases hydrophilicity, thereby promoting the superficial mass transfer [83]. Interestingly, DFT calculations showed that the surface hydroxyl on SiC-OH QDs directly participates in the CO 2 initial activation via transferring its H atom to the C atom in CO 2 forming *HCOO intermediates, instead of first generating CO 2 δ− on SiC QDs. The *HCOO pathway has a lower energy barrier than CO 2 δ− , contributing to the improvement in CO 2 hydrogenation performance. Deng et al. [84] found that the *OH coverage on hydroxyl functionalized Sn branches (Sn-OH) is influential to CO 2 RR catalytic properties because excess *OH will occupy active sites and cause the reduction of Sn-OH to Sn. With careful manipulation, the best performed Sn-OH with optimal *OH concentration

Vacancies
The introduction of vacancies can cause lattice deformation and thus alter surface electron density, contributing to the varied binding strength of intermediates which leads to different production preference. It was confirmed in the course of Zhao et al. [85], in which Bi cathode with abundant Bi atom vacancies (V-Bi) presented p states closer to the Fermi level than the vacancy-free counterpart owing to the electron-rich surface. The V-Bi cathode therefore delivered accelerated initial activation from CO 2 to CO 2 ·− and increased stability of *OCHO intermediate via p orbitals hybridization between O atoms, giving a FE HCOOH of 90% at an overpotential of 420 mV and remarkable longevity over 100 h. Similar optimization strategy toward metals was also applied to promoted the CO 2 -to-CO conversion on AuCu alloys [86]. DFT results indicated that the construction of vacancies enhanced the *COOH adsorption and Reproduced with permission from Ref. [79]. Copyright 2019 The Royal Society of Chemistry slightly weaken the *CO binding strength, accounting for the elevated CO yields and lower onset potential. Ji et al. [87] implemented a theoretical investigation on the effects of anion vacancies on the catalytic property of transition metal dichalcogenides (Fig. 7). With the increase in vacancy concentration, the binding free energy of *COOH and *CHO got lower. Notably, excessive vacancies can result in the poison of active sites by intermediates or *H and *OH due to ultra-strong binding, suggesting the existence of optimal vacancy density. Constructing oxygen vacancies (OV) has been suggested as an efficient method to boost the catalytic performance of metal oxides. The enriched electrons derived by OV serve as Lewis base sites and increase CO 2 adsorption affinity. Gu and coworkers [88] developed a partially reduced cooper oxide with numerous OV (CuO x -Vo), which exhibited ethylene FE up to 63% at − 1.4 V versus RHE, Fig. 6 a Trend for ethylene FE and calculated Bader charge for the nitrogen atom of the N-arylsubstituted tetrahydro-bipyridines. b Relationship between the ethylene FE and the ratio of atop CO to bridge CO on Cu-x electrodes. c Relationship between the ratio of atop CO to bridge CO on Cu-x and the Bader charge for the nitrogen atom of the N-aryl-substituted tetrahydro-bipyridine formed from additive x. Reproduced with permission from Ref. [19]. Copyright 2019 Springer Nature   Fig. 7 a, b Scaling relations of group X TMDs with various densities of basal plane vacancies. The labels show the number of vacancies removed in 4 × 4 cells. c Change in the binding free energies of H and OH with the density of vacancies. Reproduced with permission from Ref. [87]. Copyright 2019 American Chemical Society outperforming that of OV-free and OV-deficient counterparts. Further mechanistic exploration showed that the OV can not only stabilize *CO and *CHO intermediates, but also facilitate the *CH 2 adsorption and thus promote the dimerization of *CH 2 to produce C 2 H 4 . Similarly, the Bi 2 WO 6 nanosheets delivered 11% higher maximal FE CO and higher j after introducing OV into the structure [89,90]. The significance of OV was even more pronounced in amorphous metal oxide. Han et al. [91] recently reported that OV enriched amorphous MnO x was capable of converting CO 2 to CO with a FE of 94.8% and j of 10.4 mA/cm 2 at an overpotential of 520 mV, consistent with the boosted CO 2 -to-HCOOH conversion on OV enriched amorphous InO x [92].

Interfaces and Boundaries
In most cases, defects are not uniformly distributed in the catalyst. Simulations showed that defects in pure metal and composites near the interfaces have low migration energy barriers. Such results agree with the experimental observations that interfaces are rich in defects [93]. On the one hand, defects can stabilize the interface; on the other hand, the interfacial environment favors the formation of defects. Gao et al. [94] reported that Au-CeO x with numerous interfaces displays greater CO FE than Au alone. The hydroxyl groups dissociated from water and interface can recover the Ce 3+ ions that were oxidized after bonding with CO 2 , thus enhancing CO 2 adsorption. The binding strength of the key intermediate of the RDS, *COOH, is reinforced by the reduced Ce sites due to the interaction between Ce 3+ and terminal oxygen atom in *COOH. Subsequently, the decomposition of *COOH to *CO and the desorption of *CO are also facilitated at the interface. In other words, the holistic promoting effects of the interface, including activation, adsorption, and desorption, basically come from the Ce 3+ stabilized at the Au-CeO x interface, which can be applied to the optimization of materials other than Au. The application of the transferable method was further expanded in the Ag-CeO x /C catalyst, which presents a geometric j CO over four times that of Ag/C at − 0.89 V versus RHE and a FE CO of 92.6%. In a previous work of our group, In/In 2 O 3−x catalyst with an abundant interface was prepared by the in situ reduction of In 2 O 3 [95]. The In/In 2 O 3−x shows a FE of ca. 89.2% toward formate, outperforming that of In or indium oxide alone (ca. 67.5%). The high performance can be well interpreted by the fast electron transfer near the heterostructure interface owing to the Schottky effects.
The fabrication of boundaries is another way to improve cathode performance. For instance, Chen et al. [96] prepared a high-performance monometallic Cu electrode with rich GBs via electrodeposition and found that the electrode presents up to 70% selectivity toward C 2 H 4 and C 2 H 5 OH at potentials ranging from − 1 to − 1.3 V versus RHE. In situ ATR-SEIRAS and DFT indicated that the GBs avail the *CO adsorption and increase the possibility of C-C coupling, thus increasing C 2 yields. Feng et al. [97] disclosed a linear correlation between CO selectivity and GB density in Au/ CNTs, indicating that the GBs are responsible for the high catalytic properties of Au/CNTs. DFT simulations were then conducted by Kim et al. [98] to gain mechanistic insights on this quantitative correlation. As revealed, the broken local spatial symmetry at the GB can tune the adsorbate-to-metal σ-bonding and metal-to-adsorbate π-back bonding strengths. The adsorption of key intermediates *COOH is therefore enhanced, which switched the RDS from the *COOH formation to the subsequent proton-coupled electron transfer (PCET) step. Likewise, Liu et al. [99] developed sub-2 nm SnO 2 quantum wires (QWs) with numerous GBs. GB-rich SnO 2 QWs present higher j than SnO 2 nanoparticles, with a FE HCOOH of more than 80% and an energy efficiency of over 50% in a wide potential window, further verifying the feasibility and flexibility of introducing GBs to boost CO 2 RR activity.
The emerging oxide-derived Cu (OD-Cu) catalysts, synthesized via incomplete reduction of copper oxides or O 2 plasma treatment of Cu, display overpotentials lower than 400 mV and high selectivity toward C 2 products with a greater proportion of C 2 H 5 OH than regular Cu catalysts. At high overpotentials, the catalytic activity of OD-Cu toward C 2 H 4 exceeds that of CH 4 , whereas more CO is generated at low overpotentials [100]. The origins of the improved catalytic activity remain elusive. GBs were proposed to be one of the active sites [101]. Other oxide/sulfide-derived metals also show notable CO 2 RR catalytic performance, making it a common material design routine. For instance, the bulk Co and bulk cobalt oxide are almost catalytically inert for CO 2 RR with a j HCOOH lower than − 0.5 mA/cm 2 . However, the four-atom-thick pure Co and OD-Co prepared by Gao et al. [102] enable the CO 2 -to-formate conversion, and the latter delivers higher activity (as low as 240 mV over potential) and selectivity (ca. 90%) than the former. This work turns inactive composition into active CO 2 RR catalysts, highlighting the significance of morphology and defects. Similarly, owing to the lattice vacancies created after the release of S, the sulfide-derived Bi (SD-Bi) obtained from the electrochemical reduction of BiS 3 shows a HCOO − selectivity of ca. 40%, which is higher than that of Bi foil, further confirming that OD/SD-metal is a suitable candidate for CO 2 RR electrocatalysts [103].
Twin boundaries (TBs) are stable and well-defined defects in most metallic CO 2 RR catalysts, such as Au and Cu [104]. Tang et al. [104] reported a transferable routine to synthesize metal electrodes with quantitative TBs (TB-M) and measured the CO 2 catalytic performances of a series of TM-Cu with TM densities ranging from 0 to 10 15 cm −1 . The CH 4 FE continuously increases with increasing TM content and reaches a maximum of 59% at − 1.6 V versus RHE. Comparing CO reduction (COR) with CO 2 R, the *CO protonation on TB atoms is 10 4 faster than that on plane atoms, indicating the ultrahigh catalytic activity of TB sites. A similar precise investigation was also conducted in TB-rich Ag nanostructure [105]. The atom-specific activity, namely average current generated from one atom, of TB edges (TBEs) on Ag (~ 16 × 10 −18 A/atom) is much higher than that of Ag(111) facets (~ 0.1 × 10 −18 A/atom), indicating the spectacular intrinsic CO 2 reduction activity of TBEs. DFT calculation suggested that the unique bridge-type adsorption configuration of *COOH on TBEs is thermodynamically favored, which decreases the energy barrier of RDS (protonation of CO 2 to form *COOH), accounting for the high catalytic activity.

Electrolyte
In an electrolysis system, the electrolyte conducts current by the directional movement of the charged ions within the bulk phase and by the transfer of protons and electrons at the electrode surface. Notably, the type and the ionic concentration of the electrolyte can influence the catalytic activity and selectivity. The electrolytes used in CO 2 RR studies can be classified into four categories: aqueous, organic, ionic liquids (ILs), and molten ionic compounds. Aqueous electrolytes are preferred when the syngas is the target product because of the HER at the cathode. However, organic electrolytes and ILs with high CO 2 solubilities are also frequently used because of the poor dissolution performance of CO 2 in water. ILs are molten organic salts consisting of organic cations and inorganic anions, with a melting temperature below 100 °C, sometimes even as low as room temperature [106]. Unlike the volatile and flammable organic solvents, ILs are chemically and thermally stable and nontoxic; thus, they are deemed as the most promising alternatives for conventional CO 2 RR electrolytes [107]. However, given their high viscosity and the need of protons for CO 2 hydrogenation, ILs are usually mixed with water or organic solvents to enhance mass transfer. Meanwhile, molten ionic compounds usually require high operating temperatures, which limit their application. In this section, we will introduce electrolyte-related factors affecting the activity and selectivity of CO 2 RR.

pH and Buffer Capacity
For an aqueous electrolyte, the pH effects on CO 2 RR are fairly complex. First, based on the pH dependence of the CO 2 hydrolysis equilibrium, CO 2 molecules dissolved in aqueous electrolytes can appear in many phases, such as carbonate, bicarbonate, and carbonic acid. In specific, bicarbonate dominates at pH 6-10.3, whereas carbonic acid and carbonate dominate in a more acidic and more alkaline environment, respectively. Increasing CO 2 pressure can construct a CO 3 2− /HCO 3 − buffer system that benefits CO 2 solubility and pH stability. Notably, CO 3 2− and HCO 3 − are not the actual reactants of CO 2 RR [108]. The competitive HER at the cathode has two well-established pathways: proton reduction (i.e., H + + 2e − → H 2 ) and water reduction (i.e., H 2 O + 2e − → H 2 + 2OH − ). The proton reduction pathway shows strong pH sensitivity, leading to a boosted hydrogen evolution rate at a high proton concentration. By contrast, the pH exerts minimal on the water reduction pathway. Therefore, under alkaline conditions, the water reduction pathway is more favorable than the proton reduction pathway because of the suppression of the latter [109,110]. Owing to the mass transfer limitation, the available protons around the cathode are gradually depleted, leading to a shift of the HER pathway from the proton reduction to the water reduction even in electrolytes with bulk pH 2.5 [109]. In specific, the proton consumption and the OH − generation from water electrolysis lead to a conspicuous pH rise in the vicinity of the cathode, forming a huge pH gradient between the local and bulk phases. In an electrochemical flow cell, the forced mass transfer effect by high flow rates can reduce the thickness of the Nernst diffusion layer and thus mitigate the pH gradient [111]. The pH level is a significant factor that affects the selectivity of CO 2 RR. For instance, consistent with the observations in the flow cell, the generation of C 2 H 4 and alcohols (ethanol and n-propanol) over Cu is promoted in non-buffering solutions with high local pH contrary to the CH 4 , which prevailed over other dominant products in buffer solutions, such as KHCO 3 [59,[112][113][114]. Agreed with the Cu cathode, a similar pH dependency of the above-mentioned products was also found on other materials [111,115]. Whether a reaction is pH-dependent lies on the reaction path it takes. Reaction pathways with PCET as their RDS are sensitive to pH variation, and a relatively low pH is more favorable for such reactions to proceed. Other pathways with decoupled proton-electron transfer (DPET) as their RDS are usually immune to pH changes [116,117].
Considering the unique catalytic property of copper, the pH effects on Cu cathode have attracted considerable attention [118]. Given that surface-adsorbed *CO is easy to form on copper, COR was studied as an analog of CO 2 RR to obtain mechanistic insights [64,119,120]. On the basis of the modeling and experimental results of Liu et al. [121], the desired C 2 reduction products can be selectively formed on Cu(211) facets by exploiting the differences in pH sensitivity between the C 1 and C 2(+) pathways (Fig. 8). In general, both of them begin with the formation of *CO intermediates. For the C 1 pathway at pH 13, the initial *CO species undergo two steps of proton-electron transfer to form *CHOH. For the C 2(+) pathway, the initial *CO species undergo a dimerization of *COCO and then protonate at the O atom (not C) to produce *COCOH. The RDS of C 1 production at a low overpotential is the second proton-electron transfer from *CHO to *CHOH, whereas that of C 2(+) production is the protonation of *COCO. At a high overpotential, the RDS of the C 1 path is the first protonation of *CO, whereas the generation of C 2(+) is largely limited by the declining *CO coverage due to the sluggish formation of *CO. This discrepancy in RDS explains the different pH dependencies of C 1 and C 2+ products. When switching the pH from 7 to 13, an overpotential decrease of ca. 360 mV occurs for the generation of C 2(+) products, together with a significant selectivity enhancement of 1-2 orders of magnitude over the generation of C 1 products.

Ion Effects
Alkali metal cations with anions, such as HCO 3 − and Cl − , are the most common aqueous electrolyte in an electrochemical CO 2 reduction cell. The nature of these ions can affect CO 2 RR reactivity. For instance, large alkaline cations are conducive to the formation of C 2 products [26,122]. Thorson et al. [123] reported the hindering effects of large cations, such as Cs + and Rb + , to HER on Ag electrodes, leading to a high CO FE. They attributed the high CO FE to the size-induced differences in cation hydration degree and adsorption ability. The enhanced adsorption of large cations means higher outer Helmholtz plane (OHP) potential, which will increase the kinetic overpotential of HER and thereby vary the reaction preference [124,125]. When Cu is used as the cathode, electrolytes with large cations accelerate the generation of HCOO − and C 2 products, such as C 2 H 4 and C 2 H 5 OH, but exert negligible effects on the formation of H 2 , CO, and CH 4 [126][127][128]. The selectivity can be well explained by the field stabilization effect. The dipole field created by the hydrated cations in OHP can stabilize adsorbed intermediates with large dipole moments, such as *CO 2 and C-C coupling moieties [127,128]. However, Singh et al. [126] deemed that the elevated activity and the shifted product distribution result from the preferential hydrolysis of cations with different radii near the cathode surface. Large cations possess low pK a [126]. Owing to the polarization of the solvated water molecules between the positive charges of cations and negative charges on the cathode, the pK a of interfacial cations further decreases. When it falls below the local pH, the solvated water dissociates a proton. The hydrolysis becomes more arresting as the cation Fig. 8 Reaction schemes of major pathways considered for CO reduction toward C 1 and C 2(+) products. Green path denotes C 2 production via OC-CHO coupling; blue and red paths represent C 2 production via protonation of OCCO to form OCCHO and OCCOH, respec-tively; yellow path represents C 2 production via OC-CHOH coupling. Black path denotes C 1 production via CHOH and the dashed CH 2 O. Reproduced with permission from Ref. [121]. Copyright 2019 Springer Nature moves toward the cathode, and the electrostatic interaction becomes stronger. This phenomenon was experimentally verified via in situ ATR-SEIRAS combined with the accurate measurement of interfacial pH by a rotating ring-disk electrode. The buffer ability can be arranged in the following order: Cs + > K + > Na + > Li + [126]. The formation of CO is immune to the cation effects [129]. Ringe et al. [130] used a combined ab initio/continuum model to simulate cation and electric double layer field and concluded that CO 2 RR is easily affected by the cation-induced interface charge density and electric field. This theoretical model matches the prior experimental observation that the selectivity varies with increasing cation radius on Ag, Cu, Pt, and Au (111) single-crystal electrodes.
Sato et al. [131] reported a synergistic mechanism between the cations and the electrocatalysts. In their work, surface-adsorbed K + cations and multi-walled carbon nanotubes (MWCNTs) support the catalytically inert Mn complex into a highly efficient CO 2 -to-CO electrocatalyst (Fig. 9a). The absence of either K + or WMCNTs would not produce such catalytic activity. DFT calculation revealed that the surface-enriched electrons on the MWCNTs attract K + via electrostatic attraction, thereby enhancing the CO 2 adsorption and decreasing the activation energy of the formation of complex-CO 2 adducts (Fig. 9b). This synergy is also efficacious in modifying the inert Co and Re complexes. For inorganic anions, aside from the pH and buffering effect described above, the specific adsorption of anions on the cathode surface is also an important factor for CO 2 RR selectivity. Ogura et al. [132] proposed that the halide anions in electrolytes are prone to bind with Cu, forming Cu-X species, which was also later demonstrated by another theoretical study [133]. In the presence of strong Cu-X bonds, the binding strengths of proton and reaction intermediates change, leading to the suppression of HER and differences in activity and selectivity. Varela et al. [134] conducted an in-depth investigation about the halide effects and found that the presence of Cl − and Br − increases the CO FEs and that I − shows a sixfold enhancement of CH 4 selectivity compared with the counterpart without halides. The surface-bonded halides donate negative charges to the Cu and modify the surface electronic environment, which enhances the interaction between the electrode and the intermediates with positively charged carbon atoms and thus increase the possibility of protonation (Fig. 10f). The electron-donating capability of the three halide anions can be arranged in the following order: I − > Br − > Cl − . In addition to the effect on the surface electronic structure, the halide anions also affect the morphology of cathode materials. For example, Br − and I − induce the formation of nanocubes on Cu, whereas Cl − does not (Fig. 10a-e) [134]. Furthermore, the surfaceadsorbed Cl − helps stabilize the Cu + active sites in Cu 2 O/ Cu electrodes, leading to a long-term stability toward C 2+ production [135].

Ionic Liquids
ILs have been widely used as the electrolyte for electrochemical CO 2 RR because of their superior CO 2 solubility, high intrinsic ionic conductivity, and wide potential window. ILs can directly participate in CO 2 RR as a cocatalyst. For instance, Rosen et al. [136,137] reported that the surface-adsorbed cations of 1-ethyl-3-methylimidazolium tetrafluoroborate (EMIM + ) act as a cocatalyst toward CO 2 RR via complexing with CO 2 molecules, forming EMIM-CO 2 to lower the initial activation barrier MWCNT electrode during electrocatalytic CO 2 reduction. Reproduced with permission from Ref. [131]. Copyright 2018 American Chemical Society and thereby boost the yield of CO on the Ag electrode. This result was verified in the study by Sun et al. [107], in which the carboxylate group covalently binds to the C2-site of the imidazolium ring in EMIM + cations as confirmed by nuclear magnetic resonance and electrospray ionization mass spectroscopy (EI-MS) on the Pb electrode. On the one hand, the EMIM + adsorbate layer traps CO 2 and decreases the overpotential at Pb by 0.18 V; on the other hand, it serves as a surfactant and restrains the CO 2 ·− dimerization yielding oxalate, thus increasing the CO FE. DFT simulations and sum frequency generation experiments conducted by Wang et al. [138] revealed an exhaustive mechanism of imidazolium: CO 2 is first hydrogenated at the C2-site of the imidazolium ring, forming the key intermediate [EMIM-COOH] − , followed by sequential proton-electron transfer (SPET) steps to yield CO (via C-O breakage) or formic acid (via C-C breakage). Kemna et al. [139] provided spectroscopic evidence of the existence of imidazolium carboxylic acid species and the SPET mechanism. Recent studies have found that the overpotential diminution of ILs stems from the stability effects of electrostatic interaction between IL cations and negatively charged intermediates of the re-centered molecular catalyst Re(bpy)-(CO) 3 Cl (where bpy is 2,2′-bipyridine) [140]. In addition, imidazolium-based ILs prevail over pyrrolidinium-based ILs, presenting higher activity and selectivity. The authors ascribed this finding to the unique π-π stacking interaction between imidazolium cations and bpy − p orbitals. The impressive catalytic performance inherently abided by the optimization strategy of "solid ionic liquid layer" firstly proposed in the field of heterogeneous catalysis [141]. Inspired by this concept, Zhang et al. [142] prepared an IL-modified Cu catalyst (Cu-IL) via immobilized bis(trifluoromethylsulfonyl)imide (BMIm-NTf 2 ) onto Cu foam. After the modification, the formation of ethylene, ethanol, and n-propanol on Cu-IL is completely blocked with the FE and j unchanged. This result can be ascribed to the fact that the suppressed products share a mutual intermediate, carbene (pathway I), which can be quenched by the IL cations, whereas those products formed via CO dimerization (pathway II) are immune to the IL effects. On the basis of this phenomenon, the elusive reduction pathways of various products are clarified as follows (Fig. 11): (1) The formation of ethylene, ethanol, and n-propanol The formation of glycol is slightly suppressed by IL, suggesting that it occurs via the dimerization of two adjacent *CO 2 ·− species separated by surface-adsorbed IL cations, accounting for the decline in FE; (3) ethane is immune to IL and is produced through pathway II involving ethoxy intermediates instead of the re-adsorption and reduction of pre-formed ethylene (otherwise ethane will present ILdependent FE); and (4) methane is formed through *CO hydrogenation to *CHO. This unprecedented mechanistic investigation method via IL modification exhibits the possibility of analyzing complicated CO 2 RR pathways without the need to identify surface intermediates.
Zhan et al. [143] fabricated an imidazolium-Pt catalytic system and found that CO 2 RR in ILs is affected by temperature (Fig. 12). The amount of formed CO 2 ·− increases with increasing temperature. At subzero temperatures, the formation of imidazolium carboxylate (EMIM + -CO 2 − ) is dominant. At 0 °C, CO 2 ·− radicals couple with EMIM radicals, forming EMIM-CO 2 − . Some EMIM radicals Fig. 11 a Probing CO 2 reduction pathways for copper catalysis using an IL as a chemical trapping agent. b A simplified diagram summarizing the proposed CO 2 RR pathways on Cu. IL-suppressed pathways and products are highlighted in yellow. Reproduced with permission from Ref. [142]. Copyright 2020 WILEY-VCH can receive an electron and be reduced to EMIM carbene, which consequently bonds with CO 2 to form EMIM + -CO 2 − . At 40 °C, sufficient CO 2 ·− bonds to almost all EMIM radicals. This work disclosed a new parameter influencing catalytic performance for further research on IL-metal catalyst systems.
Low Bronsted and Lewis acidities of imidazoliumbased electrolytes, especially aprotic non-aqueous solvents such as acetonitrile (AN), weaken the intermolecular forces, thus reducing reaction activity [144,145]. Based on this, Atifi et al. [146] introduced aprotic IL based on hydrogenated N-heterocycle 1,8-diazabicyclo [5.4.0] undec-7-ene cations ([DBU-H] + ) to cocatalyze CO 2 RR on Bi electrode in AN. Interestingly, the Bi/IL synergetic CO 2 reduction system presents switchable product selectivity between CO (imidazolium-based IL, BMIM + ) and HCOOH ([DBU-H] + ) via simply modulating the electrolyte composition. In the presence of H 2 O, [DBU-H] + can be generated in situ via directly extracting protons from water molecules, thereby saving ex situ operation and enhancing overall j. H 2 O serves as a H donor for CO 2 hydrogenation in ILs; thus, its concentration should be determined. Ratschmeier et al. [147] reported that the onset potential of CO formation shifts as the water concentration increases on the Pt electrode. In a 0.1 M tetraethylammonium H 2 O/AN mixture electrolyte over the Pt cathode, oxalic acid is the dominant product at low H 2 O concentrations, whereas formic acid and H 2 are the main products at high H 2 O concentrations [148].

Summary and Outlook
Capturing CO 2 from the atmosphere and converting it into value-added chemicals, such as fuels, by using electrochemical methods not only mitigate the greenhouse effect but also relieve the global energy crisis. Although promising, the development of electrochemical CO 2 RR systems still faces several obstacles, among which the most pivotal one is the limited selectivity, especially toward multicarbon hydrocarbon and oxygenates. To solve this issue, previous studies have developed several optimization strategies for the electrolysis system, including cathode materials modification and electrolytes regulation. The catalyst composition largely determines catalytic performance. Except for physical properties, such as conductivity, adsorption strength toward reactants and intermediates can be adjusted by constructing composite materials, thereby presenting controllable products preference. Increasing the specific area or altering the shape to introduce more active sites, downsizing particles into a single atom to motivate performance, and selectively inducing the formation of exclusive facets to obtain facet-dependent high selectivity are efficient methods to optimize electrocatalysts morphologically. Suitable surface modification with organic ligands or inorganic groups can affect the superficial electronic environment and modulate the acidity, thus changing the binding affinity. Catalytic activity and selectivity can increase after fabricating abundant  [143]. Copyright 2020 American Chemical Society defects, which are usually active centers in CO 2 RR. As to the electrolyte effects, ILs and organic solvents avail the solubility of CO 2 molecules, which facilitate the reduction rate. Anions often present buffer capacity and influence the pH near the cathode. Reduction products involving PCET, such as formic acid and by-products H 2 , are sensitive to the proton concentration, based on which their production can be controlled via simply tuning interfacial pH. Electrolyte cations, such as imidazolium, adsorbed onto the catalyst surface via electrostatic interaction serve as surfactants altering the superficial environment or as H donors directly participating in CO 2 hydrogenation. In brief, the cathode-electrolyte interface bears the mass transfer and provides the reaction platform, making it the core of future investigations on CO 2 RR mechanism. However, most CO 2 RR electrocatalysts reported by now still suffer from poor stability, which is far from the industrial standard. How to improve the lifetime of electrocatalysts without changing the catalytic performance will be a major task in the future design of cathode materials. Furthermore, our current understanding of the reaction mechanism requires deeper exploration, calling for further in-depth studies to establish a clear guidance toward future research on CO 2 RR. In terms of methodologies, theoretical calculations are highly needed to verify the mechanism hypothesis, and the in situ characteristic techniques are important to track the variations in the electronic structure of active sites and detect the metastable intermediates during the reactions. Therefore, future research on electrocatalytic CO 2 RR should not be limited to the development of excellent electrocatalysts but also focus on the entire electrolysis system and study the synergy among various factors.
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/.