An overview of Cu-based heterogeneous electrocatalysts for CO 2 reduction

The electrochemical (EC) reduction of CO 2 is a promising approach for value-added fuel or chemical production. Cu-based electrodes have been extensively used as a ‘ star ’ material for CO 2 reduction to hydrocarbons. This review mainly focuses on the recent progress of Cu-based heterogeneous electrocatalysts for CO 2 reduction from 2013 to 2019. Various morphologies of oxide-derived, bimetallic Cu species and their activity in EC CO 2 reduction are reviewed, providing insights for the standardization of Cu-based heterogeneous systems. We also present a tutorial manual to describe parameters for the EC CO 2 reduction process, especially for the pretreatment of the reaction system. This will o ﬀ er useful guidance for newcomers to the ﬁ eld. Aqueous and non-aqueous electrolyte e ﬀ ects based on Cu electrodes are discussed. Finally, an overview of reaction systems of EC/PEC CO 2 reduction and H 2 O oxidation for Cu-based heterogeneous catalysts is provided.


Introduction
According to the "World Energy Outlook 2015" from the International Energy Agency (IEA), global energy demand reached 18 TW in 2013 and will increase to 26 TW by 2040. 1 More than 80% of this energy is derived from fossil fuels, 1-3 resulting in a series of problems such as energy supply. Another aspect is environmental issues involving the continuous increase of CO 2 emissions from fossil fuels combustion, which will increase from 32 Gt up to 44 Gt per year. 1 In terms of alleviating the energy crisis and environmental problems, therefore, there is an increasing demand for recycling CO 2 to produce value-added fuels and chemicals.
In recent years, CO 2 reduction to fuels and chemical products has been investigated by various methods, such as biochemical approaches, 4 building blocks for organic synthesis, 5 thermal hydrogenation, 6-11 photocatalysis, 12-14 electrocatalysis [15][16][17][18] and dry reforming with methane. 19,20 Among them, electrochemical (EC) reduction has attracted much attention. [21][22][23] However, as researchers have acknowledged, electrochemical CO 2 reduction is strongly inuenced by the pH [24][25][26][27] and conductivity or concentration of electrolyte, 23,28 applied potentials and currents, 29 CO 2 concentration 30 and ow Jian Zhao is a professor at the School of Chemistry and Chemical Engineering, Tianjin University of Technology. Prior to this, she conducted postdoctoral research at Nanyang Technological University and Singapore University of Technology and Design aer receiving her PhD degree from Queensland University of Technology in 2013. Her research area mainly focuses on electrochemical CO 2 reduction and photo/thermal heterogeneous catalytic organic synthesis.

Song Xue is a professor at the School of Chemistry and Chemical
Engineering, Tianjin University of Technology. Before entering Tianjin University of Technology in 2007, he worked at the Institute of Chemistry, Chinese Academy of Sciences, Beijing. The focus of his research has been the design and synthesis of organic dyes for dyesensitized solar cells, holetransporting materials for perovskite solar cells, and nanomaterials for photochemical conversion of CO 2 . rate, 30,31 and temperature, 32,33 making a direct, quantitative comparison of data from different groups difficult. To help provide information and to push for a more standardized "benchmark" for EC CO 2 reduction, we provide a tutorial manual for researchers who intend to start EC CO 2 reduction and clearly state the experimental parameters when comparing different systems.
EC reduction of CO 2 is involved in a variety of products ranging from CO, HCOO À , HCHO, CH 4 , CH 3 OH, C 2+ hydrocarbons (e.g. C 2 H 4 , C 2 H 6 ) and oxygenates, to higher hydrocarbons. The standard potentials for selected CO 2 reduction reactions are listed in Table 1. Since Hori's work in 1989 (ref. 34) and 1994, 35 Cu electrodes have attracted much attention due to their unique advantages for hydrocarbon production compared with other pure metallic electrodes. [36][37][38] However, a large overpotential is required and low selectivity is observed due to the wide range of products and the competing hydrogen evolution reaction. Numerous publications on Cu-based electrodes have reported a lower overpotential and/or improved product selectivity, especially since 2013, as shown in Fig. 1. Different selectivity is sometimes reported with similar Cu-based materials. The difference may be caused by the experimental conditions and the material itself, such as the oxidation state of Cu, dimensional structure or surface roughness. Therefore, there is a need to compare these results under certain experimental conditions to further understand their differences and to provide a guide for the development of Cu-based electrocatalysts for CO 2 reduction.
In this review, we rst provide a tutorial guide for conducting EC CO 2 reduction experiments. Also, various conditions such as the catalyst state, ow rate and type of electrolyte are specied when comparing studies from different research groups. CO 2 electrocatalysts are usually divided into two categories: homogeneous and heterogeneous systems. Some reviews on molecular electrocatalysts have covered the advantages of homogeneous systems. [39][40][41][42][43] Readers interested in Cu-based complexes could refer to other work. [44][45][46][47][48][49][50][51] Cu-based heterogeneous photocatalysts have been reviewed for direct conversion into solar fuels. 52 In contrast to the reviews of Cu-based heterogeneous catalysts reported for EC CO 2 reduction (review James Barber (1940.7-2020.01) was the Emeritus Ernst Chain Professor of Biochemistry, Senior Research Fellow at Imperial College London. He was also the Visiting Canon Professor to Nanyang Technological University in Singapore and the Lee Kuan Yew Distinguished Visitor to Singapore in 2009. Much of his research focused on PSII and the watersplitting process that it catalyzes, and its crystal structure obtained in 2004. His research area also included investigating inorganic systems to mimic PSII in order to develop technology for non-polluting solar fuels.
Jie Meng received her B.S. from the School of Chemistry and Chemical Engineering, Tianjin University of Technology. She is currently under the supervision of Prof. Jian Zhao for her Master's degree. Her main area of research is plasmonic goldbased nanomaterials for photocatalytic organic synthesis.
of Cu-based nanocatalysts, 53 review of Cu-binary alloys, 54 review mainly focused on theoretical studies, nanostructured Cu and bimetallics 55 ), our review includes a tutorial guide for newcomers, lists the parameters specied by different research groups for comparison (Table 2), summarizes the benchmark activity for special products (Section 3) and covers more types of Cu-based heterogeneous electrocatalysts. Different types of Cubased heterogeneous electrocatalysts, including lm and power systems, will be discussed in ve main categories: (1) morphology. Morphology control allows us to improve the catalytic activity by tailoring the structure of active sites and increasing the surface area/number of active sites. (2) Oxidederived Cu. One recent method to enhance CO 2 reduction is the oxidation and subsequent reduction of Cu. Cu x O formed by annealing, electrodeposition, Cl À and plasma induction all exhibited improved performance for EC CO 2 reduction aer in situ or ex situ redox processes. (3) Bimetallic species. Alloys are known to be able to tune the geometric and electronic properties of their parent metals. A bifunctional interface (separated composite) without forming an alloy could also improve the performance of each metal while showing fewer changes to the intrinsic electronic properties. The synergistic effects between different metals could create novel catalytic properties. (4) Surface modication. Modication with inorganic species to enhance durability and organic ligands to capture key intermediates is another strategy to improve the performance of Cubased electrodes. (5) Supports. Supports or substrates are critical to uniformly deposit the catalyst and create novel catalytic features at the interface. Following these sections, as shown in Fig. 2, the electrolyte effects and the overall reaction systems coupling CO 2 reduction and H 2 O oxidation will be discussed for Cu-based heterogeneous electrocatalysts.

A note on conducting EC CO 2 reduction
Conducting EC CO 2 reduction may be difficult for many newcomers. A slight oversight will result in the failure of the experiment. Here, we introduce several considerations for conducting EC CO 2 reduction experiments, including pretreatment of the electrolyte, pre-treatment of the electrolysis cell, ow rate of CO 2 gas and the electrolyte, electrolysis cell type and product analysis.

Pre-treatment of electrolyte
The purpose of electrolyte pre-treatment is to remove any metallic contaminants that might be present. The deposition of metal ion impurities will poison the electrocatalytic activity and cause the deactivation of the Cu electrode. 43,56 This may be also the reason why some beginners observed that almost all the products are H 2 . Proper operation could suppress H 2 evolution, a competing reaction accompanying EC CO 2 reduction.
stops to avoid recontamination of the electrolyte caused by the dissolution of electrodeposited impurities on the electrode. Another method is treating the electrolyte with an ion exchange resin (e.g. Chelex). [61][62][63] Generally the ion exchange resin was stirred with the electrolyte for at least 24 h to minimize the concentration of transition metal impurities. During electrolysis, trace metal ions coordinated in situ with ethylenediaminetetraacetic acid (EDTA) or ex situ with solidsupported iminodiacetate resin should also be considered. 62

Pre-treatment of electrolysis cell
The purpose of cell pre-treatment is similar to that of electrolyte pre-treatment. Normally the electrochemical cell should rst be cleaned with strong acid and nally boiled with de-ionized water (18.2 MU cm). For example, it could be sonicated in 20 wt% HNO 3 for 1 h, 64 cleaned in a "nochromix" bath and concentrated HNO 3 for 1 h respectively, 65 cleaned overnight in HNO 3 , 62 or rst cleaned by boiling in a mixture of 1 : 1 concentrated HNO 3 and HSO 4 and then boiling in ultra clean water before each experiment. 66

Flow rate of CO 2 gas and electrolyte
Before the electrolysis experiments, the electrolyte should rst be saturated with CO 2 gas by bubbling for at least 0.5 h. Newcomers to the eld should pay attention to the pH value before and aer CO 2 bubbling as well as the cation size (e.g. NaHCO 3 , KHCO 3 , etc.), since both these parameters have effects on the selectivity of products. The pH of 0.1 M, 0.2 M, 0.3 M and 0.5 M KHCO 3 saturated with CO 2 is 6.8, 6.9, 7.0 and 7.2 respectively. More information on the electrolyte effect is discussed in Section 5.
During electrolysis, there are three types of CO 2 gas and electrolyte ow: (1) continuously bubbling CO 2 during the process; 31,60,62,63,[67][68][69][70][71][72][73][74][75][76][77][78][79][80][81][82][83][84] (2) CO 2 bubbling plus ow electrolyte 58,85-87 (the CH 4 /C 2 H 4 ratio could be tuned by varying the CO 2 gas and KHCO 3 electrolyte ow rates); 58 (3) no report of any form of convection at all. 33,88 The unit of ow rate is standard cubic centimetres per minute (sccm) or ml min À1 . Certain ow rates of CO 2 gas were chosen by different research groups, such as 5 sccm, 63,67-72 10 sccm, 73-76 20 sccm (ref. 60,[77][78][79][80][81][82][83][84] and 30 sccm. The cell design and catalyst itself should also be considered when choosing the ow rate. The Takanabe group 73 chose 10 sccm to ensure sufficient CO 2 supply to the electrode surface while preventing the catalyst dropping off the electrode by gas bubbles. The Koper group 31 investigated the inuence of CO 2 ow rate on the activity of 3D porous hollow bre Cu and observed a maximum FE of 75% CO at À0.4 V vs. RHE when ow rate >30 sccm. In a ow setup, the ow rate of the CO 2 gas was set to 50 sccm and that of the electrolyte was 100 sccm. 85 To suit a specic system, one could also choose to adjust the ow of electrolyte with the applied potential. 86 With CO 2 gas bubbling, the catholyte could be stirred. 62,89,90 The rotating rate also has an inuence on the catalyst activity. 30 H 2 evolution increased and the product selectivity switched from CH 4 to CO when the rotating rate was increased, although there was increased availability of CO 2 at the electrode surface. This was caused by the enhanced mass transfer of dissolved CO away from the electrode surface and then less adsorbed CO was le for further reduction. 90

Electrolysis cell types
A variety of cells have been reported in the literature, but the most commonly used is a H-type cell. The total number of publications from 2007 to 2017 on selected metal-based electrocatalysts for CO 2 reduction in H-cell experiments was 1083, and 21 in continuous ow reactors. 91 In a typical H-type cell, two compartments are separated by an activated ion exchange membrane such as a Naon membrane (e.g. Naon@117 with 0.180 mm thickness and >0.90 meq g À1 exchange capacity). The Naon membrane should be activated rst, usually by boiling in 3-5 wt% H 2 O 2 , DI water, 0.5 M H 2 SO 4 and DI water at 80 C, respectively, for 0.5-1 h. The working electrode and reference electrode are in the cathode compartment with a CO 2 gas inlet and outlet. The counter electrode is in the anode compartment with or without a gas inlet and outlet.

Product analysis
Gas chromatography (GC) with a thermal conductivity detector (TCD) and a ame ionization detector (FID) is a universal method for gas product analysis. FID with a methanizer is normally used to quantify CO, CH 4 , C 2 H 4 and C 2 H 6 , and TCD is used to quantify H 2 . It is also able to detect a mixture of 100 ppm CO and 100 ppm H 2 with TCD, and 50 ppm CH 4 , 50 ppm C 2 H 4 and 50 ppm C 2 H 6 with FID, as shown in Fig. 3. High performance liquid chromatography (HPLC) and nuclear magnetic resonance (NMR) are used for detecting liquid products. For example, methanol, ethanol, formate and acetic acid products could be quantied by 1D 1 H NMR with DMSO as an internal standard (Fig. 4). More detailed information on gas and liquid product detection can also be found in other reviews. 92,93 Faradaic efficiency (FE) is the ratio between the amount of product actually detected by an analysis technique such as GC, HPLC or NMR, and the amount of product theoretically formed based on the charge passed during electrolysis. The faradaic efficiency or selectivity for each product in EC CO 2 reduction could be calculated according to the following equation: Without CO 2 gas bubbling during electrolysis:

Faradaic efficiencyðFEÞ
n is the amount of product detected (number of moles, mol); Q is the total charge passed through the system, recorded during electrolysis (coulombs, C); F is the Faraday constant (96 485 C mol À1 ); Z is the number of electrons required to obtain 1 molecule of the product. As shown in Table 1, the number of electrons required to form 1 molecule of CO, CH 3 OH, CH 4 , C 2 H 4 and C 2 H 6 is 2, 6, 8, 12 and 14, respectively. CO 2 gas was continuously bubbled during electrolysis (the rst and second measurements are not used to calculate faradaic efficiency to ensure that the data is from a system under equilibrium conditions): I is the recorded current (A); t is the time required to ll the sampling loop (s); V is the volume of the sampling loop (cm 3 ); n is the recorded ow rate (ml s À1 ). As stated, therefore, the most important thing for newcomers before conducting EC CO 2 reduction in aqueous electrolyte is the pre-treatment of the reaction system. Otherwise, H 2 may be the sole product rather than CO 2 reduction products, as the potential needed for water reduction to H 2 is less negative than that for CO 2 reduction. In addition, attention should be paid to the ow and convection state of CO 2 gas and the electrolyte to make sure that all the experiments are conducted under certain conditions for comparison.
3. Significant progress in the study of Cu-based heterogeneous electrocatalysts for EC CO 2 reduction Before a detailed summary and comparison between the work of different research groups, we provide here the benchmark activity of Cu-based heterogeneous electrocatalysts for each EC CO 2 reduction product (Fig. 5). Signicant progress has been made in C 1 and C 2 products, whereas the selectivity for C 3 and C 4 products is relatively low. New insights into the mechanistic study of Cu are also given in Fig. 5 (orange). According to the reported studies, C 3+ products are rarely formed and we will focus on the mechanisms for C 1 and C 2 products. Most mechanisms agree that the rst step involves the adsorption/ activation of CO 2 . 55,[110][111][112][113] Various adsorption or activation geometries have been proposed, which are reduced to CO and HCOO À , respectively. The adsorption intermediate on Cu via a carbon or oxygen atom is the key distinction governing the selectivity of CO or HCOO À . Therefore, altering the adsorption site and/or the stability of the adsorption intermediate is crucial for the formation of either CO or HCOO À . On the Cu surface, CO is adsorbed long enough to react further, forming HCHO, CH 3 OH, CH 4 and C 2 products. Fig. 6 shows the three most likely pathways from adsorbed *CO. Two pathways have been iden-tied for the formation of C 2 products, *CO dimerization for *C(O)(O)C* intermediates at low overpotentials and *CO hydrogenation for *CHO intermediates at high overpotentials.
High pH values will favour the major C 2 H 4 pathway (plain red arrows) and low pH value will favour the CH 4 pathway (plain blue arrows). Additionally, these intermediates are sensitive to the Cu structure and composition, supports and electrolyte, which will be discussed in detail in the next section.

Cu with various morphologies
In order to improve the activity of Cu-based electrodes, diverse morphologies have been investigated and developed,  including nanoparticles, nanocubes, nanoneedles, and threedimensional (3D) structures. The relevant surface roughness, size effects, interparticle spacing, nanoparticle loading level and crystal-facets are explored for these various morphologies.
Nanoparticles. Compared with smooth Cu, its roughened counterpart could provide a high electrochemical surface area to enhance the current density for CO 2 reduction. By pretreating polycrystalline Cu foil via electropolishing, electrochemical cycling (50-100 nm Cu nanoparticles) and argon sputtering, I. The benchmark activity and progress of mechanism studies on Cu-based heterogeneous catalysts for EC CO 2 reduction.
Chorkendorff and co-workers 114 found that high surface roughness showed higher activity for hydrocarbon formation (CH 4 and C 2 H 4 ) in KClO 4 electrolyte. The enhanced activity was ascribed to the greater abundance of undercoordinated sites on the roughened surfaces. The high activity of the roughened surface of Cu nanoparticles was also reported by H. M. Zhang, X. F. Li and co-workers. 115 In their work, Cu nanoparticles with a diameter of 100 nm (thickness 47 nm) were coated on carbon paper via pulse electrodeposition.
Compared with constant potential electrodeposition, pulse electrodeposition created a more roughened surface to provide abundant active sites, leading to 85% CH 4 formation at À2.1 V vs. RHE (0.5 M NaHCO 3 , 1 sccm CO 2 ). According to these studies, the promoting effect of the roughened surface on C 2 H 4 and/or CH 4 formation is independent of the catalyst preparation method and the electrolyte used for CO 2 reduction. Fig. 6 Most likely reaction pathways from adsorbed *CO on Cu surface for C 1 and C 2 products. Plain red, plain blue and dashed blue routes are for major, minor and trace C 2 products. Particle size is a critical parameter in tuning the activity and selectivity of Cu nanoparticle catalysts. It may be difficult to control smaller particle sizes and interparticle distances directly via an electrochemical method. Therefore, although powder systems have to be assembled into electrodes for further application in CO 2 reduction, many studies have been reported based on Cu powder systems with a controlled small particle size. $5 nm Cu nanoparticles embedded in a thin lm of metal organic framework (MOF, used to restrict the particle size) on uorine-doped SnO 2 (FTO) exhibited 31% (major HCOO À , minor CO) CO 2 reduction selectivity at À0.82 V vs. RHE (0.1 M NaClO 4 , pH ¼ 4.6). 116 B. R. Cuenya, P. Strasser and co-workers 65 prepared Cu nanoparticles with a mean size range of 2-15 nm on glassy carbon. The hydrocarbon production decreased as the particle size decreased and vanished for sizes #2 nm, as shown in Fig. 7a. Cu nanoparticles (12 nm, 19 nm, 24 nm, 37 nm)/four types of carbon support also showed higher C 2 H 4 /CH 4 production than smooth copper lm, and smaller particle sizes were more favorable for C 2 H 4 formation (pH ¼ 6.8). 117 These individual studies give different trends for product distribution, probably caused by different preparation methods, supports and pH of the electrolytes.
Another critical parameter for Cu nanoparticle catalysts is the interparticle spacing. 118 B. R. Cuenya and co-workers 119 designed 1.5 nm, 4.7 nm and 7.4 nm Cu nanoparticles with interparticle distances of 10-22 nm, 24-53 nm and 41-92 nm, respectively. Smaller interparticle spacing was favorable for readsorption of the CO intermediate and its further reduction to hydrocarbons, and with the increase of interparticle spacing the CO 2 ux increased, as shown in Fig. 7b and c. P. Strasser and co-workers 120 also showed that C 2 H 4 selectivity could be tuned by particle density caused by diffusional interparticle coupling that controlled CO desorption/re-adsorption and in turn the effective CO ad coverage. Recently, it was also reported that stacked small Cu nanoparticles could be formed by in situ electrochemical fragmentation during the CO 2 reduction, promoting C-C coupling reaction. 121 The Cu nanoparticles loading level also has a signicant inuence on the morphology evolution and product selectivity. 77 P. D. Yang and co-workers 122 assembled densely packed Cu nanoparticles (6.7 nm) on carbon paper electrode. These densely packed nanoparticles changed to cube-like structures intermixed with smaller nanoparticles (Fig. 8a). Compared with spatially separated nanoparticles and ex situ prepared nanocubes, the in situ formation of cube-like particles from densely packed nanoparticles suppressed C 1 products and improved C 2 -C 3 formation (C 2 H 4 , C 2 H 5 OH, and n-C 3 H 7 OH 50% at À0.75 V vs. RHE, Fig. 8b).
Therefore, small interparticle distances and densely packed nanoparticles would lead to higher hydrocarbon formation. The above Cu nanoparticles are crystalline in form. Compared with the crystalline form, the amorphous form seems to be more favourable for C 2 products. J. M. Yan and co-workers 78 synthesized amorphous Cu nanoparticles (average size 3.3 nm), and achieved 37% HCOOH and 22% C 2 H 5 OH at À1.0 V vs. RHE (0.1 M KHCO 3 , 20 sccm CO 2 ). The crystalline Cu nanoparticles (average size 3.4 nm) only showed 26% HCOOH and no C 2 H 5 OH at the same potential. They ascribed the enhanced activity to the high electrochemically active surface area (ECSA) and CO 2 adsorption on the amorphous surface.
Similar to Cu nanoparticles formed on lm in situ as an electrode, the morphology of powder nanoparticles could also have an effect on product selectivity, and a surface morphology with more defects and boundaries promotes C 2 products. Star decahedron Cu nanoparticles 123 and branched CuO nanoparticles 106 are reported to achieve high faradaic efficiency of ethylene (C 2 H 4 ) up to 52% and 70% at À1.0 V vs. RHE, respectively.  Morphology evolution. The inuence of other morphologies on CO 2 reduction activity may be caused by many aspects. Although the enhancement effects and mechanisms of certain products are different for different morphologies, morphologies with more edges, corners or sharp tips seem to promote C 2 and even C 3 products. B. R. Cuenya and co-workers 124 electrodeposited prism-shaped Cu catalysts which exhibited higher C 2 H 4 current density than planar Cu foil. A total FE of $73% for C 2 and C 3 products ($45% C 2 H 4 , 22% C 2 H 5 OH, 9% C 3 H 7 OH) at À1.0 V vs. RHE was obtained. They attributed the enhanced selectivity to the increased local pH and high abundance of defect sites on the roughed prism Cu surface. Cu pillar structure 125 achieved much higher HCOO À selectivity than planar Cu foil at À0.5 V vs. RHE (0.1 M KHCO 3 ). Through electroredeposition, dissolution and redeposition of Cu from Cu 2 (-OH) 3 Cl sol-gel, E. H. Sargent and co-workers 82 prepared Cu nanoneedles with sharp tips (Fig. 9). These Cu nanoneedles with sharp tips could produce high local electric elds that concentrate electrolyte cations and CO 2 molecules at the catalyst surface, resulting in a high partial C 2 H 4 current density (160 mA cm À2 at À1.0 V vs. RHE) and C 2 H 4 /CH 4 ratio of up to 200 (ow cell). The effect of sharp tips was also explained from a kinetic point of view instead of a reaction barrier. 126 However, enhanced C 1 products were observed instead of C 2 products. N. F. Zheng and co-workers 99 reported a simple strategy to prepare ultrathin Cu/Ni(OH) 2 nanosheets with atomically thick ultrastable Cu nanosheets in the presence of sodium formate (HCOONa), which achieved 92% CO at À0.5 V vs. RHE. The presence of surface formate inhibited the oxidation of Cu(0) and the hybrid structure probably had an effect on the promoted CO production. Hybrid structure-enhanced C 1 product formation has been observed by other groups. 127,128 Similar trends in the role of morphology are observed for nanocube structures in powder systems, especially for C 2 H 4 production. R. Buonsanti and co-workers 129 fabricated different sizes of Cu nanocrystal spheres (7.5 nm and 27 nm) and Cu nanocrystal cubes (24 nm, 44 nm, and 63 nm). There was a monotonic size-dependent trend for both shapesthe smaller, the more active. Cube-shaped copper nanocrystals showed better performance than spheres. The overall CO 2 reduction activity changed from 50% to 80% and 63% for 24 nm, 44 nm, and 63 nm, respectively, and the highest faradaic efficiency of C 2 H 4 was 41% for 44 nm nanocubes at À1.1 V vs. RHE (Fig. 10). Edges and Cu(100) were responsible for maximizing C 2 H 4 selectivity. However, edges were also reported to promote CH 4 selectivity in nanowire structures. P. D. Yang and co-workers 130 prepared ultrathin ($20 nm) 5-fold twinned Cu nanowire on carbon black/glassy carbon plates. The catalyst could achieve 55% CH 4 at À1.25 V vs. RHE with <5% other CO 2 reduction products, likely due to the high density of edge sites. With the evolving morphology, CH 4 decreased and C 2 H 4 increased. Wrapping Cu nanowires with graphene oxide could prevent morphology changes and in turn prevent the decrease of CH 4 selectivity.
3D structure. The importance of the inuence of threedimensional (3D) structures on local pH, retention time of intermediates, gas permeability or liquid diffusion has been demonstrated by many researchers. G. Mul, M. T. Koper and coworkers 31 designed a 3D porous hollow bre Cu electrode, and the rate of formation of CO was one order of magnitude larger  than for nanocrystalline Cu. Using hydrogen bubbles as a template, G. T. R. Palmore and co-workers 88 fabricated Cu nanofoams with connected pores of 20-50 mm (Fig. 11b). Compared with an electropolished Cu electrode, they observed increased selectivity for HCOOH, decreased selectivity for CO, CH 4 and C 2 H 4 , and novel production of C 2 H 6 and C 3 H 6 . A maximum of 37% HCOOH was obtained at À0.9 V vs. RHE (0.1 M KHCO 3 ). They attributed these differences to the high surface roughness, hierarchical porosity, and connement of reactive species. For the 3D nanoporous structure, the authors also showed that the inner surface area of the nanopores only becomes accessible above a critical electrolyte concentration of 0.5 M KHCO 3 due to the overlapping electrical double layer (EDL). In another study, P. Broekmann and co-workers 131 prepared oxide-derived Cu nanofoams. The faradaic efficiency of C 2 (C 2 H 4 and C 2 H 6 ) could reach 55% at À0.8 V vs. RHE (0.5 M NaHCO 3 ). Compared with the copper foam prepared by the Palmore group, there was a signicant difference in the production distribution, which may be caused by the formation of Cu 2 O and different pore sizes. C 2 reached a maximum value for the surface pore size ranging from 50 to 100 mm, while it decreased signicantly below 50 mm (Fig. 11c). The Broekmann group 132 also found more efficient trapping of reaction intermediates (e.g. C 2 H 4 ) in the presence of mm-sized pores within the Cu foam on a 3D skeleton structure, favoring fully reduced C 2 products. Using a similar preparation method to the Palmore and Broekmann groups, 88,131 E. H. Sargent, D. Sinton and coworkers 79 prepared a Cu nanofoam with pore sizes in the range of microns and then oxidized it in a mixed solution of 60 mM HCl and 60 mM H 2 O 2 . They again proved that higher surface roughness and porosity favored C 2 H 4 over CH 4 (0.1 M KHCO 3 , 20 sccm CO 2 ).
In addition to pore sizes on the micron scale, nano-porous structures could also change the local pH and retention time of key intermediates 133,135 and, in turn, the product selectivity. Using a sputtering method on anodized aluminium oxide, as shown in Fig. 12a, the pore widths and depths of the Cu mesopore electrode could be precisely controlled. Compared with an electrode of 300 nm (width)/40 nm (depth), C 2 H 4 formation was enhanced from 8% to 38% when the pore width was narrowed to 30 nm, whereas the major C 2 product changed to C 2 H 6 with a faradaic efficiency of 46% when the depth was increased to 70 nm at À1.3 V vs. RHE (Fig. 12b). A pH change with 3D morphology was also reported by other groups. Using Cu foil with a mixed solution of (NH 3 ) 2 S 2 O 8 and NaOH for different times, W. A. Smith and co-workers 134,136 obtained Cu(OH) 2 nanowires with various lengths and densities. The nanowires with high lengths and densities had an inuence on the diffusion of HCO 3 À , leading to a high local pH, since HCO 3 À can neutralize OH À (HCO 3 Crystal facets. The crystal facet dependence of CO 2 reduction for Cu foil and Cu single crystals has been widely explored by experimental and theoretical methods, 137 especially for (111) and (100). 138 On a single-crystal copper electrode, M. T. Koper and co-workers 26,66,139,140 observed that one pathway for the formation of CH 4 preferentially occurs on (111) facets, while the other pathway leads to C 2 H 4 formation on (100) facets. 66 The Koper group also distinguished the reactivity of (100) terraces versus (100) steps, where selective reduction of CO to C 2 H 4 at low overpotentials occurs on terrace sites. 139 The theoretical calculation stated that the coupling of two CO molecules  mediated by electron transfer to form a *C 2 O 2 dimer is a ratedetermining step involved on Cu(100) for C 2 (C 2 H 4 and C 2 H 5 OH) formation. 140 Experiments on CO and CO 2 reduction in electrolytes with various pH values also showed a pHdependent pathway for CH 4 mainly on Cu(111) and a pHindependent pathway for C 2 H 4 on Cu(100). 26 Results from other groups suggest that Cu(100) favors CHO* intermediates and follows C 2 H 4 formation at relatively low overpotentials (À0.4 to À0.6 V vs. RHE), while Cu(111) favors COH* intermediates and CH 4 /C 2 H 4 formation at high overpotentials (<À0.8 V vs. RHE), as shown in Fig. 13. 138 B. S. Yeo and co-workers 141 studied Cu 2 O (hydrothermally prepared)-derived Cu and Cu single-crystal surfaces. Of the three single-crystal surfaces (100), (111) and (110), Cu(100) exhibited the lowest energy barrier for the dimerization of CO*. A. Nilsson and co-workers 157 investigated single crystal copper (100), (111), and (211) for comparison with Cu nanocubes. The (100) surface was the most comparable to the Cu nanocube surface in terms of C 2 H 4 production, whereas CH 4 was not suppressed. One possibility is that it has the ideal terrace length or active sites for C 2 H 4 formation. K. Chan, H. T. Wang and coworkers 59 galvanostatically cycled Cu foil in Cu(NO 3 ) 2 to obtain a Cu 2 O nanocube layer with smooth (100) facets on the surface. They also suggested that Cu(100) and stepped (211) facets favored C 2+ products over Cu (111). From the present results, Cu(100) crystal facets favor C 2+ products compared with other facets for most reported systems.
As stated above, understanding the effects of Cu morphology on the selectivity is highly complex since there is a combined effect of properties, such as low-coordinated sites, catalyst density or dispersion, CO 2 ux and electrical double layers in the electrolyte, on the activity of CO 2 electrochemical reduction. Although there is wide variation in the activity trends for Cu with various morphologies, the selectivity for possible products may be adjusted by tuning the particle size and interparticle spacing/particle density of Cu nanoparticles/nanocubes, tuning the pore size and depth/length of 3D structures, or tuning the energy facets of Cu crystals. In addition, attention should be paid to morphology evolution during the electrochemical CO 2 reduction process.

Oxide-derived Cu electrocatalysts
Recently, oxide-derived Cu has drawn much attention in electrocatalytic CO 2 reduction. Various oxide-derived Cu electrocatalysts have been designed and the mechanisms involved have been discussed widely. Some groups [68][69][70][71]142,143 suggest that grain boundaries are the active sites. Some groups 80,144,145 believe that low-coordinated atoms act as active sites. There are also many groups 85,114,[146][147][148] which believe that the active phase is metallic Cu 0 since there is a signicant driving force for Cu 2 O reduction under CO 2 reduction conditions. Many groups 82,95,149,150 proved that the Cu + site is key for enhanced activity and remained on the catalyst surface during the reaction. Some groups 96,151,152 found that sub-surface oxygen stabilized in reduced oxide-derived Cu plays an important role and there is synergy between surface Cu + and surface Cu 0 sub-oxide species.
Although the true active site is still under debate, oxidederived Cu has shown excellent performance in decreasing the potential required and enhancing selectivity for specic products. Most recent reports also conrmed that two (Cu d+ and Cu 0 ) were better than one. 100,154 In this section, we will discuss oxide-derived Cu in detail based on the fabrication process, including annealed/oxide-derived Cu, electrodeposited/oxidederived Cu, Cl À /oxide-derived Cu, plasma/oxide-derived Cu, and in situ/oxide-derived Cu. Since the wide variation in experimental conditions results in various results for similar materials, detailed experimental conditions are included for different groups.
Annealed/oxide-derived Cu. Annealing is a simple and effective strategy to enhance the activity of Cu. During this process, the annealing temperature and the following redox process all have an inuence on the CO 2 reduction activity. M. W. Kanan and co-workers 69 annealed Cu foil at different temperatures in air, forming Cu 2 O layers with different thicknesses. Thick Cu 2 O layers formed at 500 C ($$3 mm) exhibited higher selectivity and lower overpotential to CO (45% at À0.3 to À0.5 V vs. RHE) and HCOO À (33% at À0.45 to À0.65 V vs. RHE) compared with the Cu foil counterpart (Fig. 14a). Later, using oxide-derived Cu (electrochemical reduction and thermal reduction with H 2 ), they 68 investigated CO electroreduction to liquid fuels in CO-saturated 0.1 M KOH (pH ¼ 13). Engineering the grain boundaries by altering the redox process could tune the product distribution. 57% carbon oxygenated species (C 2 H 5 OH, C 2 H 5 COO À and n-C 3 H 7 OH) were obtained at potentials ranging from À0.25 V to À0.5 V vs. RHE. Temperature-programmed desorption (TPD) experiments 70 showed that the active sites for CO reduction on oxide-derived Cu surfaces were strong CO binding sites supported by grain boundaries. The Kanan group 71 also prepared electrodes of Cu nanoparticles on carbon nanotubes (Cu/CNT) with different average grain boundaries via e-beam evaporation and subsequent annealing. The CO reduction activity was directly correlated to the density of grain boundaries in Cu nanoparticles, exhibiting a linear relationship at potentials ranging from À0.3 V to À0.5 V vs. RHE (Fig. 14b). A maximum faradaic efficiency of >70% C 2 H 5 OH and C 2 H 5 COO À was obtained at À0.3 V vs. RHE. I. E. L. Stephens, I. Chorkendorff and co-workers 101 investigated CO electroreduction on oxide-derived Cu prepared by the Kanan group. They showed that CH 3 CHO was a key intermediate in the electroreduction of CO to C 2 H 5 OH and formed at a low steady-state concentration.
Higher temperature annealing forms both Cu 2 O and CuO nanowires and the corresponding derived Cu also exhibits superior performance for CO 2 reduction. C. Wang and coworkers 80 annealed Cu gauze in air at 600 C to get CuO nanowires (with Cu 2 O present inside the nanowires) with a diameter of 50-100 nm and length of 10-50 mm. Then they obtained high density Cu nanowires by electrochemical reduction (ECR) or forming gas reduction (FGR). The Cu nanowires obtained via the ECR method exhibited a higher surface roughness and a thin surface layer of Cu 2 O nanocrystals less than 10 nm, while FGR led to larger Cu crystal segments (>100 nm size) without Cu 2 O present on the surface. At low overpotentials (À0.3 V to À0.5 V vs. RHE), the total faradaic efficiency of CO and HCOOH was as high as 70-80% for ECR nanowires, whereas H 2 was the dominant product for Cu gauze and FGR nanowires (0.1 M KCHCO 3 , 20 sccm CO 2 ). They ascribed the high activity to the high-density grain boundaries and low-coordinated surface sites associated with the small crystalline features, as well as more open facets (e.g. (100) and (211)) on the surface. Later studies by C. Wang, T. Mueller and co-workers 144,145 further suggested that the high activity and selectivity of the ECR nanowires could be ascribed to the (110) surface, high-angle grain boundaries, or some closely related metastable surface feature. Pre-reduction of annealed Cu in different solutions led to different activities, as also reported by J. J. Zhang's group. 155 Reduction of annealed Cu in 1 M NaOH formed only a layer of nanobers with 30-100 nm diameters, whereas in 1 M H 3 PO 4 the nanobers were surrounded by kernels and achieved 43% HCOO À , which was much higher than for the pre-reduced annealed Cu in NaOH (1% HCOO À ) in 0.5 M KHCO 3 .
Summarizing the above reports, the annealed/oxide-redox method could enhance the selectivity and lower the overpotential required for HCOO À and CO during CO 2 reduction, while promoting the selectivity of C 2+ products during CO reduction.
With related mechanism studies on enhanced HCOO À and CO formation, using ambient pressure X-ray photoelectron spectroscopy interpreted with quantum mechanical predictions of the structures and free energies, W. A. Goddard III, J. Yano, E. J. Crumlin and co-workers 96 showed that the presence of suboxide species below the Cu surface played a crucial role in the adsorption and activation of CO 2 on annealed/oxide-derived Cu. This thin layer of sub-oxide was essential for converting chemisorbed CO 2 in the presence of H 2 O as the rst step toward CO 2 reduction products such as HCOO À and CO. Quantum mechanical calculations and experimental results also showed that there was an optimized amount of subsurface oxide. More or none at all would destabilize the bent (chemisorbed) CO 2 . (1) In the case of Cu(111) without subsurface oxide (Fig. 15a), the C atom of CO 2 was chemically bonded to a Cu 0 . One of the O atoms was stabilized by hydrogen bonding to H 2 O ad on another Cu 0 . (2) When the subsurface oxide was increased to 1/4 ML (Fig. 15b), the C atom was chemically bonded to two surface Cu 0 . One O atom was chemically bonded to one Cu 0 , and the other O atom was stabilized by the Cu + pulled up by H 2 O ad . (3) When the subsurface oxide was increased to 1/2 ML (Fig. 15c), the C atom was chemically bonded to a Cu + that shares one O atom stabilized by hydrogen bonding to H 2 O ad on another Cu + . Later they 153 found that only surface Cu + itself could not improve the performance of CO 2 reduction, and proposed Cu metal embedded in an oxidized matrix model as a partially oxidized Cu surface, where the synergy between the surface Cu + and surface Cu 0 was responsible for CO 2 activation.
Electrodeposited/oxide-derived Cu. In contrast to the annealed-redox process above, the electrodeposited/annealed and electrodeposited Cu 2 O lm could also promote C 2 products besides HCOO À or CO in CO 2 reduction compared with pure Cu. P. Broekmann and co-workers 132 compared two types of oxide-derived catalyst, annealed skeleton (300 C, 12 h) and electrodeposited Cu nanofoam on 3D Cu skeleton. Both the annealed and electrodeposited skeleton catalysts showed preferential (100) texturing and profound activities toward C 2 product formation (C 2 H 4 and C 2 H 6 ) in 0.5 M NaHCO 3 . Later, they 156 electrodeposited dendritic Cu on Cu mesh via the mass control of Cu(II) ions, followed by thermal annealing at 300 C in air. Electrodeposited dendritic Cu showed high HCOO À and C 2 H 4 production, while the electrodeposited/annealed sample directed product selectivity toward C 2 and C 3 alcohols (detailed data given in Table 1) and high resistance against degradation (Fig. 16a). An identical location SEM study showed that Cu nanoparticles and nm-sized cavities and cracks on large dendritic structures were present for the annealed sample. They assigned the difference in stability to the change in reaction mechanism; namely, the electrodeposited sample relied on a coupled C 1 /C 2 pathway (catalyst poisoning/blocking effect predominantly caused by C 1 hydrocarbon pathway), while the annealed sample relied on a coupled C 2 hydrocarbon/alcohol pathway (Fig. 16b).
There is a dependence of CO 2 activity on lm thickness for electrodeposited Cu 2 O lm. J. Baltrusaitis, G. Mul and coworkers 72 electrodeposited Cu 2 O lm on a Cu plate. A C 2 H 4 /CH 4 ratio of $8-12 was observed for thin lm at À1.1 V vs. RHE, with a larger amount of CH 4 for thicker lm (0.1 M KHCO 3 , 5 sccm CO 2 ). B. S. Yeo and co-workers 81 electrodeposited Cu 2 O lm with different lm thicknesses on Cu discs. A maximum of 34-39% C 2 H 4 with a ratio of C 2 H 4 /CH 4 up to 100, and 9-16% C 2 H 5 OH was obtained at À1.0 V vs. RHE, as shown in Fig. 17a. Then they 146 prepared Cu 2 O-and CuO-derived Cu with different thicknesses via a hydrothermal method (1.3 and 11.5 mm). In contrast to other works proposing the protonation of C 2 H 4 , they showed that C 2 H 6 and C 2 H 5 OH were likely to form via the dimerization of -CH 3 intermediates on thick oxide-derived Cu (Fig. 17b). They observed that Cu 2 O was rapidly reduced to metallic Cu during CO 2 reduction by using in situ Raman spectroscopy. The surface reoxidized in tens of seconds aer the cathodic potential was removed. This is in contrast to the Lee group's 150 work on electrodeposited Cu 2 O/GDE, where they found that Cu 2 O was only partially reduced on the basis of ex situ XRD and Auger electron spectroscopy. Later again, the Yeo group only observed signals belonging to CO adsorbed on Cu metallic sites rather than oxide sites for electrodeposited Cu x Zn. 84 Cl À /oxide-derived Cu. Although the formation mechanism may still be unclear, cycling a Cu precursor in the presence of Cl À will lead to Cu nanocubes, which will in turn favour C 2 H 4 production. A. Nilsson and co-workers 157,158 reported a simple in situ method to fabricate nanocube-covered Cu by its successive oxidative-reductive cycling in the presence of KCl. Using online electrochemical mass spectrometry (OLEMS), there was an earlier onset potential and relatively high selectivity for C 2 H 4 over CH 4 (0.1 M KHCO 3 ). The Yeo group 159 also observed an enhanced C 2 H 4 /CH 4 ratio with Cl À /oxide-derived Cu. The Nilsson group ascribed the enhanced C 2 H 4 formation to the large number of exposed (100) facets and the rise in the local pH for the roughed surface of Cu nanocubes. The Ager group and the  Bell group did more work to understand the effect of pH on product formation for Cl À /oxide-derived Cu nanocubes. Using in situ X-ray absorption spectroscopy (XAS), A. Nilsson and coworkers 147 investigated the formation mechanism of the Cu nanocubes. Since no CuCl was observed in the Cu K-edge XAS spectra, they believed that the precursor for nanocube formation was Cu 2 O, not CuCl as previously assumed (A. T. Bell and J. W. Ager's work is shown in Fig. 18a). CuCO 3 /Cu(OH) 2 was also prepared via cycling in the absence of KCl. The results of OLEMS during CO 2 reduction showed that there were no signicant differences between CuCO 3 /Cu(OH) 2 -derived Cu and Cl À /Cu 2 O-derived Cu. Therefore, they pointed out that the inuence of the precursor oxidation state on the selectivity toward C 2 H 4 formation was not important. They also believed that the active species was metallic Cu, since no signicant concentration of residual oxide was detected on the order of a few nanometers in the thin XAS model samples. Their further study 151 used in situ ambient pressure X-ray photoelectron spectroscopy (APXPS) and quasi-in situ electron energy loss spectra (EELS), which showed that there was a small amount of subsurface oxygen but no residual copper oxide. Combined with DFT simulations, they proposed that the interaction of subsurface oxygen with metal causes higher CO binding energy, resulting in higher CO coverage. Higher CO coverage kinetically favored C-C bond formation.
In contrast to the Nilsson group's study 157 and the Yeo group's study, 159 J. Lee and co-workers 160 observed the preferential formation of multicarbon fuels, especially n-C 3 H 7 OH (the rst report over 10% C 3 -C 4 products, 0.1 M KCl), using in situ prepared Cl À -induced bi-phasic Cu 2 O-Cu. They also found that the catalyst exhibited relatively higher Cu + coverage than oxidederived Cu and the use of a KCl electrolyte could prolong the preservation of the Cu 2 O phase compared to a KHCO 3 electrolyte.
A. T. Bell, J. W. Ager and co-workers 142 electrochemically cycled copper foil in the presence of halide anions KF, KCl, KBr, and KI. They observed an enhanced faradaic efficiency for C 2 H 4 and C 2 H 5 OH (excluding KF). Their observation of C 2 H 5 OH was in contrast to the Nilsson group and Yeo group studies. Without electrochemical cycling in halide anions, the product distribution was not signicantly changed even with the addition of halide anions in electrolyte for CO 2 reduction. In situ Raman spectroscopy and SEM showed that during the oxidation sweep, anodic corrosion formed a Cu 2 O layer, which consisted of cubic crystals $150 nm. CuCl formed cubes when precipitated in the solution at pH > 4. In neutral and basic solutions with a low Cl À concentration, CuCl could convert to Cu 2 O. During the reduction sweep, irregular Cu nanoparticles (ca. 20 nm in diameter) with rounded edges were formed (Fig. 18a). They ascribed the enhancement in C 2 H 4 formation to a large number of (100) facets (similar to the Nilsson group) and the formation of grain boundaries and defects (similar to the Kanan group). Later they 64 also investigated four types of oxide-derived Cu electrocatalysts: "oxide-derived nanocrystalline Cu" developed by the Kanan group, "Cu nanowire arrays" developed by the Smith group, "electrodeposited Cu 2 O lms" developed by the Yeo group, and "electrochemical oxidation-reduction cycled Cu" developed by the Nilsson group. There was an optimal roughness factor for the oxide-derived layers to have a high local pH and maintain a high concentration of dissolved CO 2 . More recently, J. W. Ager and co-workers 143 prepared 18 O-enriched oxide-derived Cu by cycling Cu foil in the presence of KCl in H 2 18 O. The selectivity of C 2 and C 3 products in 0.1 M KHCO 3 was up to 60%. By analysis with ex situ secondary-ion mass spectrometry (SIMS), they found that <1% of the original 18 O content remained in the sample aer the CO 2 reduction reaction and believed that the grain boundary was more likely responsible for the high activity, as proposed by the Kanan group. Similarly to the Yeo group, they also observed the rapid re-oxidation of oxide-derived Cu with in situ Raman spectroscopy. They believed that this rapid re-oxidation was possibly due to the grain boundaries (Fig. 18b), since Cu lm with fewer grain boundaries did not oxidize quickly. E. H. Sargent and co-workers 82 applied constant potential in CO 2 -saturated 0.1 M KHCO 3 to reduce Cu 2 (OH) 3 Cl precursor on carbon paper. This dissolution and electro-redeposition process could form Cu 2 O nanoneedles, which exhibited excellent activity for C 2 H 4 formation (partial current density 160 mA cm À2 ) with a ratio of C 2 H 4 /CH 4 of up to 200 at À1.0 V vs. RHE (20 sccm CO 2 ). For the rst time, they used in situ Cu L-edge XAS to demonstrate that the Cu + surface species could direct C 2+ product selectivity. The process of Cu 2+ to Cu + was quick (within 5 min), while Cu + to Cu 0 was much slower, and 23% Cu + surface species still remained aer 1 h electrolysis at À1.2 V vs. RHE.
In summary, Cl À /oxide-derived Cu shows excellent activity for C 2 and even C 3 -C 4 product formation, regardless of whether it is caused by the morphology evolution to nanocubes or the presence of Cu + .
Plasma/oxide-derived Cu. Plasma-induced CuO and Cu 2 O as a precursor of Cu-based electrocatalysts for efficient CO 2 reduction was recently reported by B. R. Cuenya and co-workers. 95 The plasma treatment could suppress CH 4 formation while enhancing the formation of CO, HCOO À and C 2 H 4 . The onset potential towards CO and HCOO À was shied to lower overpotentials, similar to the result observed by the Kanan group. However, the presence of strongly binding defect sites such as grain boundaries with intermediates proposed by the Kanan group could explain the early onset potential for CO 2 reduction, but could not explain the suppression of C 2 H 4 due to the H 2 plasma treatment following O 2 plasma treatment (Fig. 19a). Therefore, using a combination of characterization techniques, including operando X-ray absorption ne-structure spectroscopy (XAFS) and scanning transmission electron microscopy (STEM) equipped with energy dispersive X-ray spectroscopy (EDS), the authors found that the oxides in the surface layer were resistant to reduction and Cu + species remained on the surface during the reaction. By controlling experiments with the same surface roughness (O 2 plasma plus H 2 and O 2 plasma-treated Cu foils), they demonstrated that the roughness of oxide-derived Cu catalysts played only a partial role in determining the catalytic performance, while the presence of Cu + was key for lowering the onset potential and enhancing C 2 H 4 selectivity. Later, the Cuenya group 149 used the same method to activate Cu nanocube catalysts obtained by electrochemical cycling of Cu foil in 0.1 M KCl. $73% C 2 and C 3 products were achieved on Cl À -induced Cu nanocubes and O 2 plasma-treated samples. High C 2 H 5 OH ($22%) and n-C 3 H 7 OH ($9%) were also obtained for the O 2 plasma-treated sample at À1.0 V vs. RHE. In their former work, 95 they observed that Cu + is the key. Through controlled experiments in this work (Fig. 19b), they believed that the oxygen content (oxygen ions associated with Cu + species) played a more important role in C 2 H 4 formation than Cu(100) facets.
In situ/oxide-derived Cu. In situ formation of oxide during CO 2 electrolysis is a promising method to activate Cu electrocatalysts, although the promoted products are uncertain. A. Engelbrecht and co-workers 161 oxidized Cu in situ by using a CO 2 /O 2 gas mixture instead of pure CO 2 gas and found that the formation of CH 4 was largely suppressed, while C 2 H 4 was favored (0.1 M KHCO 3 , 100 sccm). J. M. Spurgeon and coworkers 83 used a pulsed-bias technique for CO 2 reduction on Cu foil. Compared with the conventional potentiostatic technique, there was a major shi in the selectivity (Fig. 20). The syngas H 2 : CO ratio ranged from $32 : 1 to 9 : 16 for pulse times between 10 ms and 80 ms, respectively (0.1 M KHCO 3 , 20 sccm). J. Y. Lee and co-workers 162 also indicated that Cu 2 O formed under the anodic cycle in the pulsed electroreduction of CO 2 and this process also prevented the poisoning of carbon on the Cu surface. Using pulsed voltammetry, P. Rodriguez and coworkers 163 observed that oxygenated products associated to the coverage of OH species on single crystal Cu(100) and Cu (111). In contrast to Spurgeon and co-workers' work, 83 however, the selectivity trend shied to CH 4 and C 2 H 4 .
Early last year, however, I. Chorkendorff, I. E. L. Stephens and co-workers 164 investigated CO electroreduction on polycrystalline Cu foil in 0.1 M KOH at low overpotentials from À0.4 to À0.6 V vs. RHE. Compared with oxide-derived nanostructured Cu in the literature, the polycrystalline Cu foil exhibited higher selectivity for C 2 and C 3 aldehydes and ethylene. This indicated that future studies should focus on the intrinsic activity of Cu.
In summary, for oxide-derived Cu, various methods have been developed for the preparation of a Cu-oxide precursor (Cu 2+ , Cu + ) and for its subsequent redox process (in situ and ex situ, electrochemical and H 2 reduction). As stated at the beginning of this part, the true active site of Cu-based electrocatalysts fabricated from oxidation-reduction processes is still under discussion. Using Cu-based electrocatalysts with  oxidation-reduction pretreatment, the activity and product selectivity were indeed improved, especially for the ratio of C 2 H 4 and CH 4 , which has been proven by many research groups as above. Further mechanism investigations may focus on in situ and operando studies to gain more insight into the subsurface oxygen, chemical state or morphology of the Cu x O catalyst under CO 2 reduction conditions.

Cu-M bimetallic species
Combining Cu with other metals (M) to form alloys or separated Cu-M composites is another efficient approach to enhance the activity and selectivity for CO 2 reduction. In this part, we summarize Cu-M alloys, including pure metallic alloys and oxide-derived alloys, as well as Cu-M bifunctional interfaces (separated Cu-M composites). For pure metallic alloys, the catalytic activity is affected by the nature of the secondary metal atom. Generally, CO selectivity is enhanced for Cu-M (M ¼ In, Zn, Ag, Au) and HCOO À for Cu-M (M ¼ Sn, Pd), while hydrocarbon selectivity increases with increasing Cu atoms in the alloy. When the composition of ordered or separated arrangements is precisely controlled, different behavior will be observed and C 2 products could be promoted. When the alloys are formed from oxide species, different behavior could also be observed, especially for oxide-derived blended and separated composites for C 2 H 5 OH production. There are electronic and geometric effects for each individual component of the alloy, while the geometric effects play important roles for separated Cu-M composites. Therefore, we may expect distinct performances of Cu-M bimetallic species via precise control of their atom and phase arrangements.
Cu-M alloy. Cu-based bimetallic electrocatalysts for CO 2 reduction have been utilized since 1991, as reported by Watanabe and co-workers. 165 Cu alloys such as Cu-Ni, Cu-Sn, Cu-Pb, Cu-Zn, Cu-Cd, and Cu-Ag exhibited different catalytic activities compared to their respective elemental metals. Coupling Cu (d metal) with more oxyphilic materials (sp metals) such as Sn, In, Bi, and Sb might inhibit H 2 evolution while enhancing the adsorption of CO and CHO to facilitate subsequent H addition. 166 G. Zangari and co-workers 166 prepared Cu x In y electrodes with dendritic morphology by electrodeposition. At À1.0 V vs. RHE (0.1 M KHCO 3 ), the selectivity of HCOO À could be up to 62% with 80 atom% In alloy, and the CO/H 2 ratio could be tuned to 2.6 : 1 with 40 atom% In alloy. K. Takanabe and coworkers 73 thermally oxidized a Cu metal sheet to form a hairy CuO nanowire structure on Cu 2 O-Cu layers. The CuIn electrode was then prepared through in situ electrochemical reduction of the oxide-derived Cu in a mixed solution of InSO 4 and citric acid. High efficiency of CO was obtained (>70%) at low overpotential (À0.4 V to À0.7 V vs. RHE), with a maximum of 95% CO at À0.7 V vs. RHE (0.1 M KHCO 3 , 10 sccm CO 2 ). Similar trends were also observed when using Sn and Zn as the second metals. DFT calculations suggested that In was preferentially located at the edges of the Cu surface (H 2 evolution sites are presumably edges) and caused weakened adsorption of H over CO (high-overpotential metal for H 2 formation), while the intact Cu corners might be still responsible for CO production. More importantly, there was only a slight improvement in CO selectivity for the Cu-In electrode without initial oxidation treatment of the Cu sheet. They 74 also fabricated CuSn alloys using a similar electrodeposition method with initial oxidation. High CO selectivity with FE >90% over a wide potential range of À0.4 V to À0.8 V vs. RHE was achieved. According to their results, to improve the CO selectivity with CuIn or CuSn alloys, initial oxidation could be adopted. C. P. Berlinguette and coworkers 67 investigated ternary alloys of Cu-Zn-Sn for CO 2 reduction to CO and HCOO À . With an optimized ratio, >80% CO and HCOO À could be achieved with a partial current density of 3 mA cm À2 at an overpotential of 0.2 V (0.5 M NaHCO 3 , 5 sccm CO 2 ).
E. H. Sargent, P. D. Yang and co-workers 89 prepared a Cuenriched Au nanoneedle electrode via an underpotential deposition (UPD) method with various Cu content. Designed syngas ratios were obtained (0.5 M KHCO 3 , 20 sccm CO 2 ). In situ SERS and DFT calculations illustrated how the surface electronic structure could be tuned by Cu enrichment to inuence CO binding. Tuning the composition of CuAu alloys from Au-rich to Cu-rich resulted in a selectivity change from CO to hydrocarbon, which was also reported in other studies. 167 For Cu-rich alloys in another study, Au addition led to the suppression of CH 4 while increasing CO production. 168 A. T. Bell and co-workers 63 prepared strained CuAg surface alloys via melting physical mixtures of Cu and Ag under argon in a vacuum arc furnace. The incorporation of Ag atoms onto the Cu surface modied the Cu electronic structure, where the valence band density states shied to deeper levels. As a result, the binding energies of H and O relative to CO decreased and the ratio of CO to H 2 products increased (0.05 M Cs 2 CO 3 , 5 sccm CO 2 ). An AgCu dendritic catalyst was also electrodeposited on Cu foil and Ag 57 Cu 43 exhibited 2.2 times higher CO production than pure Ag in terms of Ag mass activity at À0.83 V vs. RHE (0.5 M KHCO 3 , 10 sccm CO 2 ). 76 These results indicate that by alloying Cu with Au or Ag, the ratio of CO to H 2 could be tuned. Additionally, additives during the electrodeposition process and the supports applied could affect the morphology of the deposited alloy and its corresponding activity for CH 4 or C 2+ products. Through the addition of 3,5-diamino-1,2,4-triazole into the electroplating solution, A. A. Gewirth and co-workers 102 obtained homogeneous CuAg wire samples, which exhibited higher selectivity than their counterparts of up to 60% C 2 H 4 and 25% C 2 H 5 OH at À0.7 V vs. RHE (1.0 M KOH). T. Meyer and co- workers 169 electrodeposited $6.6 nm CuPd nanoalloy on a polymeric lm. 51% CH 4 was obtained with Cu 2 Pd in organic electrolyte at an overpotential of À0.86 V. They believed that the enhanced CH 4 formation was due to the synergistic interplay between Pd-H sites and Cu-CO sites with the polymer as a basis for local CO 2 concentration. Later, they 170 electrodeposited $6 nm CuAg nanoalloy on this polymer on glassy carbon. At 0 C, 21% C 2 H 3 OO À was achieved with Cu 2 Ag 3 at À1.33 V vs. RHE in 0.5 M KHCO 3 with 8 ppm benzotriazole.
As stated before, we separately summarized the electrode used aer fabrication without post-treatment and the powder materials used aer assembly or drop casting to form lms/ electrodes. The following catalysts are Cu-M powder alloys, which sometimes have to be mixed with a Naon binder and nally cast on a conductive substrate. Based on some groups' work, C. P. Berlinguette and co-workers 171 provided a general relationship between the primary product formed and the metal-CO bonding strength (Fig. 21). The best CO formation catalysts oen had a CO heat of adsorption (DH) of 10 kcal mol À1 . A lower value of DH is more favorable for HCOO À formation, whereas a higher value of DH favors H 2 and hydrocarbon formation. A series of In-M on titanium substrate was prepared and followed the trend FeIn < CoIn < ZnIn < NiIn < CuIn for CO production.
Aer preparing different ratios of Au x Cu y (x ¼ 3, y ¼ 1; x ¼ 1, y ¼ 1; x ¼ 1, y ¼ 3) alloy nanoparticles, P. D. Yang and coworkers 172 obtained corresponding monolayer samples on various substrates via self-assembly by the Langmuir-Schaefer method. The monolayer samples showed great mass activity, achieving 67% CO with a partial current of À230 mA mg À1 at À0.73 V vs. RHE for Au 3 Cu. The activity for CO 2 reduction exhibited a volcano shape, where Au 3 Cu represented the peak, as determined by the electronic and geometric effects (Fig. 22). These effects were associated with the local atomic arrangements at the active sites. In contrast to the AuCu alloys, alloying Pd with Cu could enhance CO and/or C 2 products. M. Yamauchi, P. J. A. Kenis and co-workers 86 designed CuPd nanoalloys with ordered, disordered, and phase-separated elemental arrangements (Fig. 23). With the same atomic ratio, phase-separated arrangements (more sites with neighbouring Cu atoms) favored the production of C 2 products compared to the other two arrangements, with >60% (48% C 2 H 4 and 15% C 2 H 5 OH) at À0.7 V vs. RHE. CuPd with ordered arrangements gave 75% CO at À0.7 V vs. RHE. Surface valence bond spectra suggested that geometric effects were key in determining the selectivity compared to electronic effects. N. Umezawa, J. H. Ye and co-workers 173 electrodeposited CuPd on glassy carbon and optimized the ratio between Pd and Cu. 80% faradaic efficiency of CO was obtained with optimal Pd 7 Cu 3 at À0.8 V vs. RHE. DFT calculations suggested that synergistic geometric and electronic  effects were responsible for the high selectivity. D. Ma, G. X. Wang and co-workers 174 loaded CuPd nanoparticles (3.3 nm) on carbon black, and Pd 85 Cu 15 /C achieved 86% CO at À0.89 V vs.

RHE.
Cu-M bifunctional interface. Compared with Cu alloys, separated Cu-M composites have shown their own advantages in tuning product selectivity. Via theoretical calculations, Y. S. Jung, Y. T. Kim and co-workers 175 found that a Au-Cu bifunctional interface was more favorable for the stabilizing *COOH intermediate (Fig. 24a) and its intrinsic electronic properties were less affected compared to the bulk alloy. T. Takashima and co-workers 176 synthesized atomic layers of Cu on Pd particles (Cu-Pd) without the formation of an alloy by underpotential deposition. They ascribed the improved CO tolerance and HCOO À production to the charge transfer from Pd to Cu and a downward shi of the d band centre to the Fermi level (0.5 M NaHCO 3 , 16 sccm CO 2 ). Additionally, C 2 H 5 OH selectivity could be promoted on a Cu-M bifunctional surface via a two-site mechanism. B. S. Yeo and co-workers 84 electrodeposited Cu-Zn oxides with various amounts of Zn dopants, which exhibited different C 2 H 5 OH selectivities. XRD did not show any alloy signals except separated Cu and Zn. The maximum faradaic efficiency of 29% C 2 H 5 OH was obtained on Cu 4 -Zn at À1.05 V vs. RHE (0.1 M KHCO 3 , 20 sccm CO 2 ). J. Y. Lee and co-workers 177 incorporated Ag in Cu 2 O by electrodeposition and found that phase blended Ag-Cu 2 O exhibited higher C 2 H 5 OH selectivity (34%) than its phase separated counterpart (20%), with 3 times the selectivity of Cu 2 O (11%) at À1.2 V vs. RHE, as illustrated in Fig. 24b. This was because of the role of the Ag dopant and the closer distance between Ag and Cu was efficient for the insertion of CO on Ag sites to other C 1 intermediates on Cu sites, as illustrated in Fig. 24c.
To design Cu-M bimetallic species, therefore, the rst aspect to be considered is the group that the parent metals belong to. The preparation method also has an inuence on the activity.
The most important thing for mechanism investigation in this system is precise control of the composition, morphology and position of the bimetallic species.

Surface modication of Cu-based electrocatalysts
In recent years, surface modication has also been investigated for Cu-based electrocatalysts, including inorganic and organic outlayer species. Inorganic outlayers could protect the Cu surface and alter the direct contact between the electrolyte and Cu surface. Then the stability and activity could be altered compared to bare Cu-based electrocatalysts. Y. J. Liang, H. L. Wang and co-workers 60 decorated Pd atoms on Cu mesh by soaking it in PdCl 2 + HCl solution. During CO 2 reduction, the foreign Pd atoms induced continuous restructuring of the Cu surface, in turn preventing the deactivation of catalysts by incorporating or desorbing accumulated carbonaceous species (Fig. 25). The deactivation of electrodes caused by the adsorption of carbon or intermediates has been reported by many researchers. 24,178,179 According to their reports, the Pd content should be optimized to inhibit the deactivation of the Cu electrode and maintain the CO 2 reduction activity, since less Pd is not sufficient to improve catalyst durability, while more Pd will change the product selectivity and lead to more H 2 production. The selectivity of CH 4 and C 2 H 4 remained above 50% for 4 h continuous electrolysis at À0.96 V vs. RHE (0.5 M KHCO 3 , 20 sccm). J. P. Ramírez and co-workers 180 found that Cu-In nanoalloys evolved to a separated core-shell (Cu-In(OH) 3 ) structure aer successive electrochemical cycles. The separated catalysts with an In(OH) 3 shell showed high selectivity for CO production, and In(OH) 3 modication plays an important role in this enhanced selectivity. Later, they also observed >80% HCOO À at À0.8 V vs. RHE (0.1 M KHCO 3 ) with submicron S-modied Cu. 181 J. S. Luo, M. Grätzel and co-workers 61 modied the surface of CuO nanowire electrodes with SnO 2 via atomic layer deposition (ALD) and as high as 97% CO was achieved at À0.7 V vs. RHE (10 sccm CO 2 ). Gas phase absorption measurements conrmed the signicantly decreased binding strength of both CO and adsorbed H* aer SnO 2 modication. However, whether the residual oxides (mainly Sn 2+ with some Cu + ) were active catalysts remained uncertain.
Besides modication with inorganic materials, organic ligands are coated on the Cu surface to enhance CO 2 reduction, especially for CH 4   ligands. By properly modifying Cu(OH) 2 nanowires with amino acid, 182 H 2 formation could be suppressed and CO 2 reduction was promoted since the interaction between the key intermediate CHO* and -NH 3 + of the adsorbed zwitterionic glycine stabilized this key intermediate during CO 2 reduction. E. Andreoli and co-workers 183 modied the Cu foam surface with polyamide and obtained enhanced C 2 H 4 production with unaffected CH 4 (0.1 M NaHCO 3 ). They ascribed this enhancement to the charge transfer to the Cu surface, stabilization of the CO dimer by the -NH 2 group, and the adsorption of CO near the polymer.

Supports for Cu-based electrocatalysts
We list all the substrates and/or supports in Table 2, since supports also have an inuence on activity. Supports can maintain good dispersion and stabilization of active sites, as well as creating synergistic interactions or active interfaces between supports and copper catalysts. CuO on CO 2 capture material exhibited higher C 2 H 4 faradaic efficiency than CuO on carbon black or without support. 184 Supports with sufficient surface area were also critical for the high C 2 -C 3 products for densely packed Cu nanoparticles. 122 Electrodeposited Cu gave higher activity on graphene oxide and pure graphite than on glassy carbon, which was attributed to the preferential deposition of Cu nanoparticles at defects present on the graphene layers of the former supports. 57 Generally, Cu foil, Cu plate, Cu mesh, or even FTO is used as a substrate for Cu-based electrocatalysts for EC CO 2 reduction, while glassy carbon is used for Cu powder electrocatalysts. Gas diffusion layers (GDL) or gas diffusion electrodes (GDE) were also chosen as substrates to enhance the performance of the corresponding loaded Cu-based electrocatalysts due to their gas/electrolyte penetrability. 86,150 Polymer on FTO 169 or glassy carbon 170 was used as a substrate to enhance the local concentration of CO 2 . Polymer-based diffusion layers or electrodes have also been fabricated recently to enhance the activity and stability of Cu electrocatalysts, where as high as 76% C 2 H 4 was obtained at À0.55 V vs. RHE. 103,185 Therefore, in order to improve the performance of Cu-based electrocatalysts, supports or substrates with high surface area and high gas and liquid penetrability should be considered. More recently, it was found that Cu 3 N support could act as an underlying stable Cu + species to reduce the CO dimerization energy barrier. 186 This might be another consideration when choosing supports for copper electrocatalysts.

Electrolyte effect on CO 2 reduction with Cu-based heterogeneous electrocatalysts
In EC CO 2 reduction, aqueous electrolytes are generally selected by researchers due to their environmentally friendly properties, low cost and potential for coupling with water oxidation. Nonaqueous electrolytes are also studied by many researchers due to their large electrochemical window, high CO 2 solubility and low proton availability.

Aqueous electrolytes
In a pioneering study, Hori and co-workers 34 investigated CO 2 reduction on Cu sheet electrodes in various electrolytes at 5 mA cm À2 . The major product was H 2 with FE >70% in KH 2 PO 4 / K 2 HPO 4 , which decreased to 10% in KClO 4 . C 2 H 4 and alcohols were favored in KCl, K 2 SO 4 , KClO 4 and diluted KHCO 3 electrolytes, whereas CH 4 was favored in concentrated KHCO 3 and KH 2 PO 4 /K 2 HPO 4 .
The concentration of bicarbonate and cation size both play important roles during CO 2 reduction. A high concentration of bicarbonate leads to a relatively high pH, which in turn favors CH 4 formation, while a big cation size promotes C 2 H 4 and other C 2 products. G. Mul and co-workers 24  H. Sargent, D. Sinton and co-workers 79 predicted the pH and CO 2 concentration at the electrode surface using a diffusionbased model (Fig. 26). CO 2 limitation occurred under high local pH conditions. In 1991, Hori and co-workers found that cationic species (Li + , Na + , K + and Cs + ) in bicarbonate solution affected the product selectivity (C 2 H 4 /CH 4 ). 187 A. T. Bell and coworkers 188,189 recently also reported that there was a decrease in faradaic efficiency for H 2 and CH 4 , and an increase in faradaic efficiency for C 2 H 4 and C 2 H 5 OH for Cu cathodes with increasing cation size (Li + , Na + , K + , Rb + , and Cs + ) (Fig. 27a). They ascribed the effect of the electrolyte cation size on CO 2 reduction to cation hydrolysis in the vicinity of the cathode. The pK a for hydrolysed cations decreased and they served as buffer agents to lower the local pH, leading to an increase in the local concentration of dissolved CO 2 (Fig. 27b and c). The concentration of molecular CO 2 decreased with increasing pH due to its rapid consumption by hydroxyl anions to form HCO 3 À and CO 3 2À . This process occurred at much higher rates than the rate of CO 2 reduction. In contrast to CO 2 reduction, by using singlecrystal Cu(100), single-crystal Cu (111), and polycrystalline Cu electrodes, M. T. Koper, F. C. Vallejo and co-workers 190 found that the cation effects were potential-and structure-dependent in CO reduction (Fig. 27d); the onset potential for C 2 H 4 was lower for the single crystals (À0.3 V and À0.4 V vs. RHE for Cu (100) and Cu (111)) than for the polycrystalline electrode, for which the overpotential increased with increasing cation size. The onset potential for CH 4 (À0.65 V vs. RHE) was independent of both cation size and surface structure. When the applied potential was more negative than À0.65 V vs. RHE, larger cations enhanced CH 4 formation. When the applied potential was from À0.65 V to À0.3 V vs. RHE, larger cations increased C 2 H 4 selectivity.
Halides are sometimes directly added into aqueous electrolytes to enhance the CO 2 reduction and suppress the competing H 2 evolution. P. Strasser and co-workers 191 added various concentrations of halides into KHCO 3 electrolytes and observed that the addition of Cl À and Br À resulted in increased CO selectivity. The adsorbed I À exhibited a larger effect on CH 4 production than C 2 H 4 . The probable reason was the induced negative charge possessed a remarkably positive effect favoring the protonation of CO. KCl was also used as a catholyte because it resulted in a higher local pH and the formation of bi-phasic Cu 2 O-Cu, favored for multicarbon fuel production. 177 As indicated above, high concentrations, large cation sizes and halide additives could be considered for CO 2 reduction in bicarbonate electrolytes in order to obtain more hydrocarbon products. Moreover, regarding the role of bicarbonate aqueous electrolytes, recently M. H. Shao's group 192 directly observed that CO 2 molecules were mediated to the Cu surface via their equilibrium with bicarbonate anions rather than direct adsorption from the solution (Fig. 28), using attenuated total reection surface-enhanced infrared absorption spectroscopy, isotopic labelling, and potential stepping techniques.

Non-aqueous solvents
Organic electrolytes. Organic solvents have been studied in CO 2 reduction since the early 1980s. 193,194 Although organic solvents have environmental and safety issues, these solvents possess unique advantages in EC CO 2 reduction, such as (1) large electrochemical window. For example, with 0.65 M supporting electrolyte, the reduction/oxidation potential window is À2.8 V to +3.3 V vs. SCE for acetonitrile (MeCN), À3.0 V to +1.6 V vs. SCE for dimethylformamide (DMF), and À2.9 V to +1.5 V vs. SCE for dimethyl sulfoxide (DMSO). (2) Low proton availability: thus, the application of organic solvent could suppress the competing H 2 evolution reaction  and improve the total faradaic efficiency for CO 2 reduction.
(3) High CO 2 solubility: in 2000, Hori's group also studied the Pt system in MeCN-H 2 O mixtures. They showed that the CO 2 concentration could be up to 270 mM in MeCN with a low water concentration, 8 times higher than in aqueous solution (33 mM), 195 as shown in Fig. 29. We will discuss recent work on EC CO 2 reduction with Cu-based heterogeneous catalysts in organic electrolytes. [196][197][198][199] In these studies, the organic electrolyte contained 0.1 M supporting electrolyte, such as tetrabutylammonium tetrauoroborate (nBu 4 NBF 4 ), tetraethylammonium tetrauoroborate (TEABF 4 ), tetraethylammonium triuoromethanesulfonate (TEATfO) and sodium triuoromethanesulfonate (NaTfO), to enhance the conductivity of Ionic liquids. In the past 15 years, ionic liquid has been emerged as a promising candidate for CO 2 capture. 200 Ionic liquid also possesses a large potential window. It has been reported that ionic molecules could complex with CO 2 c À intermediates to reduce the energy barriers or potentials for Agbased systems 201 and could change the selectivity for Bi-based systems. [202][203][204] The controlled selectivity of CO 2 reduction with Cu nanoparticles-modied boron-doped diamond electrode in 1-ethyl-3-methyalimidazolium tetrauoroborate (EMIM$BF 4 ) ionic liquid was also achieved. 205 Using porous (30-40 mm) dendritic copper nanofoam (10 wt% Cu 2 O) in ionic liquid-water mixture [EMIM](BF 4 )/H 2 O (92/8 v/v) as the electrolyte, V. Artero, M. Fontocave and co-workers 206 obtained almost 90% HCOO À at À1.55 V vs. Fc + /Fc with current density À5.0 mA cm À2 , a much lower potential than that required in their former work without ionic liquid. 196,197 Therefore, organic solvents and ionic liquids are better choices for suppressing H 2 evolution in catalyst systems not suitable for aqueous solutions. In aqueous systems, alkaline conditions could promote C-C coupling during CO 2 reduction. 207,208 For the ultimate goal of CO 2 recycling, neutral aqueous solution is the best choice and various concentrations and cations could be applied to tune the activity.
6. EC/PEC CO 2 reduction and H 2 O oxidation as an overall reaction system for Cu-based electrocatalysts

EC
The EC CO 2 reduction and H 2 O splitting produces carbon-based fuels and oxygen. Realizing the overall reaction with one catalyst in a single device is desirable. In such a system, the catholyte and anolyte may be different for specic half reactions. T. J. Meyer and co-workers 209 combined two half reactions catalyzed by Cu(II)/Cu(0) electrode. As shown in Fig. 30, the electrode for CO 2 reduction into CO/HCOO À was Cu(0) lm electrodeposited on a boron-doped diamond electrode. A boron-doped diamond electrode immersed in Cu(II) was used for H 2 O oxidation into O 2 . This report demonstrates that a simple Cu(II)/Cu(0) electrode is sufficient to catalyze CO 2 reduction and H 2 O splitting in neutral aqueous solution with a H-type cell. However, more aspects should be considered when choosing different catholytes and anolytes, for example the membrane used in the Htype cell. By using SnO 2 -modied CuO nanowire electrodes as both the cathode for CO 2 reduction and the anode for the oxygen evolution reaction, J. S. Luo, M. Grätzel and co-workers 61 constructed a complete CO 2 electrolysis system with a bipolar membrane. Bipolar membranes consisting of a cation exchange layer and an anion exchange layer were also investigated in other systems for CO 2 reduction and H 2 O oxidation with different catholytes and anolytes. Bipolar membranes needed a lower cell voltage than monopolar membranes. 210

PEC
Solar energy is the largest source of renewable energy. Coupling solar irradiation with EC CO 2 reduction is of considerable interest. 211,212 In this review, we do not discuss photocatalytic (PC) CO 2 reduction with Cu-based materials; readers interested in this topic are directed to some recent reports, including of ptype CuI, 213 Cu-decorated TiO 2 , 214 CuO nanoclusters on Nb 3 O 8 nanosheets, 215 Au-Cu nanoalloys supported on TiO 2 , 216 carbondecorated Cu 2 O mesoporous nanorods, 217 etc. We will focus on photoelectrochemical (PEC) CO 2 reduction with Cu-based catalysts, where similar photoelectrolysis cells as for PEC water splitting are made for PEC CO 2 reduction.
In a PEC CO 2 reduction system, sunlight, visible light and UV light are three options for the light source. For the purpose of solar energy utilization, sunlight irradiation without bias potential is the ultimate goal. Cu 2 O with a direct band gap of $2.0 eV is a promising material for enhancing CO 2 activity and inhibiting H 2 evolution in PEC CO 2 reduction systems. N. R. Tacconi, K. Rajeshwar and co-workers 218,219 rst reported the utilization of Cu 2 O for CO 2 PEC reduction. They electrodeposited Cu 2 O nanocrystals on CuO nanorod arrays. 95% CH 3 OH was formed with a bias potential of +0.17 using this CuO@Cu 2 O nanorod array as a photocathode (AM 1.5, 70 mW cm À2 ). Modifying the Cu-based photocathode could lead to different enhanced CO 2 reduction products. L. R. Baker and coworkers 220 electrodeposited CuFeO 2 /CuO on FTO and used it as a photocathode. 80% CH 3 COO À was detected with a bias potential of +0.2 V under white-light LED illumination (100 mW cm À2 ). P. D. Yang and co-workers 221 directly assembled Au 3 Cu nanoparticles on the surface of TiO 2 -protected silicon nanowire as a photocathode. Compared with the planar counterpart, lower overpotential or additional bias ($120 mV lower) was needed to drive CO 2 reduction to CO. 80% CO could be achieved at À0.2 V vs. RHE (LED light source with intensity 20 mW cm À2 , wavelength l ¼ 740 nm).
When Cu-based electrocatalysts are used as a cathode, both the cathode and the photoanode could be manipulated to enhance the performance of the overall system. J. S. Lee and coworkers 222 constructed a PEC system comprising a WO 3 photoanode and Cu cathode for CO 2 reduction. 71.6% carbonic products (65% CH 4 ) were obtained at +0.65 V vs. RHE under visible light irradiation (>420 nm). M. Miyauchi, H. Abe and coworkers 223 obtained 79% HCOO À when using Cu-Zn alloy lm as a cathode in 0.1 M KHCO 3 and SrTiO 3 as a photoanode in 0.1 M KCl + 0.01 M NaOH under UV light illumination without applied bias potential. Y. S. Kang and co-workers 224 engineered a (040) facet BiVO 4 photoanode and integrated it with a Cu cathode for CO 2 PEC reduction (Cu cathode|NaCl|BiVO 4 photoanode). Different products were obtained by tuning the bias potential, such as 65.4% HCOO À (+0.75 V), 85.1% HCHO (+0.9 V), 6.89% CH 3 OH and 4.4% C 2 H 5 OH (+1.35 V) in 0.5 M NaCl (AM 1.5).
When using Cu x O as a photocathode or cathode, corrosion will happen. Protective layers such as TiO 2 were applied in studies done well by M. Grätzel's group. J. L. Gong and coworkers 225 introduced a simple strategy by using Cu 2 O as the cathode and TiO 2 nanorods as the photoanode for PEC reduction of CO 2 . 92.6% carbonic products (54% CH 4 , 30% CO, 3% CH 3 OH) were obtained at +0.75 V vs. RHE bias potential (AM 1.5, 100 mW cm À2 ). Through cut-off lter experiments, they conrmed that the photogenerated electrons were not the main reason for Cu 2 O corrosion; instead, photogenerated holes were primarily responsible for the instability of Cu 2 O. As shown in Fig. 31, the photogenerated electrons were consumed at the electrode/electrolyte interface, while the holes moved to the back contact and the counter electrode. Back illumination (the travel distance of the electrons was longer than that of the holes) was favorable for the stability of Cu 2 O.
The photocurrent density is relatively low for the above reported systems, as listed in Table 2. One future aim is to improve the photocurrent density in PEC CO 2 conversion. Constructing a hybrid catalyst consisting of a molecular catalyst and semiconductor material for PEC CO 2 reduction could improve the selectivity and efficiency. 226-228 For a hybrid system, careful design of the semiconductor is also important. As shown in Fig. 32a, M. T. Mayer, M. Grätzel and coworkers 229 designed a heterogeneous catalyst system by  covalently immobilizing molecular Re(bpy) (CO) 3 Cl onto a TiO 2 -protected Cu 2 O photocathode (Cu 2 O/AZO/TiO 2 ) via phosphonate linkers. On the TiO 2 layer there was also a mesoporous lm of 4.5 to 5 mm thickness with 18 nm TiO 2 particles. The catalyst system without mesoporous TiO 2 did not show substantial photocurrents, while mesoporous TiO 2 (enhanced roughness and catalyst loading) exhibited enhanced photocurrents (Fig. 32b). 80-95% CO was achieved under chopped light illumination with photocurrent density of 2.5 mA cm À2 at À1.9 V vs. Fc + /Fc.

PV cells or PV-electrolyzers
The potential bias could be supplied by other forms of energy for PEC electrocatalysis, such as solar PV panels. 36,212,[230][231][232][233] The rst idea to apply PV cells for PEC water splitting into hydrogen and oxygen was proposed in 1995 (ref. 234) and has been widely used, whereas its use for the overall reaction of CO 2  CuO@SnO photocathode/photoanode system (Fig. 33a). At a voltage of 2.38 V (AM 1.5, 100 mW cm À2 ), a solar-to-CO free-energy conversion efficiency peaking at 13.4% with 81% CO selectivity was obtained at À0.55 V vs. RHE (Fig. 33b) and the photocurrent could be up to 12 mA cm À2 .
The use of other forms of sustainable energy such as solar energy to drive the overall reaction of CO 2 reduction and water splitting is promising. Cu x O could be used as both a cathode and photocathode aer surface protection, as stated above. It is also promising to use Cu oxide species directly as cathodes and photocathodes, since the Schottky junction between Cu 2 O and Cu could facilitate electron and hole separation, leading to enhanced activity. 236 The morphology and activity evolution of Cu x O is worthy of further investigation because the chemical changes of Cu x O due to photocorrosion and electroreduction processes were indeed pre-activation steps for CO 2 reduction, as discussed in this review for oxide-derived Cu.

Summary and outlook
As part of the response to the energy crisis and environmental issues, the electrochemical reduction of CO 2 has attracted increasing attention from researchers. Until now, Cu-based materials remain the most investigated heterogeneous systems for CO 2 electrolysis due to their distinct advantages for hydrocarbon formation. The high abundance and low cost of Cu will further enable it to become a "star" material in the future.
In this review, the latest progress on Cu-based heterogeneous electrocatalysts for EC CO 2 reduction was discussed. We summarized the benchmark activity for specic products in Section 3 and various types of Cu-based materials reported by different research groups in Section 4. The H 2 evolution reaction is inevitable because its equilibrium potential is lower than that of CO 2 reduction. Apart from engineering the material, therefore, adjusting the electrolyte composition from aqueous to non-aqueous, adding ionic liquid or other additives, and careful pre-treatment of the system are also important. Through the discussions in this review, we hope we could provide useful information to newcomers to the eld through detailed information about the experimental conditions, and to those already experienced in the topic through the comparison data in Table 2 and our focus on the more recent literature about Cu-based heterogeneous electrocatalysts (2013-2019).
In summary, the design of efficient and selective Cu-based electrocatalysts is inspiring but still challenging. Several considerations may also be helpful for engineering efficient systems: (1) PEC CO 2 reduction with Cu-based materials. Here, we could manipulate two aspects, one of which is energy supply, such as solar energy and other forms of renewable and clean energy for lower applied overpotentials. Optimization of this technique may develop commercially feasible CO 2 reduction systems. Another aspect to consider is the Cu (photo)cathode. The photoanode corrosion of copper oxide species is known by many researchers and studies have been done to prolong its stability under irradiation. However, the instability should not be considered a disadvantage when using the copper oxide species as (photo)cathodes for CO 2 reduction. This is because oxide-derived Cu exhibited better performance than its parent counterpart. Moreover, Cu nanostructures possess localized surface plasmonic (LSPR) effects, which have been used for photocatalytic organic synthesis. 237,238 To the best of our knowledge, there is no report yet for CO 2 reduction utilizing Cu LSPR effects. Future efforts should be made towards this.
(2) Complete systems coupling the reduction of CO 2 and H 2 O oxidation, as well as other signicant oxidation reactions. Studying only the half reaction (CO 2 reduction) is not sufficient to achieve a commercially feasible CO 2 reduction system. The performance of anodic reactions also needs to be investigated. It is necessary to explore Cu-based materials as both cathodes and anodes.
(3) Hybrid system of Cu-based materials and metal complexes. Metal molecular catalysts or metal complexes could coordinate CO 2 /intermediates with the metal center and ligands. Covalent-attached metal complexes on Cu-based materials will combine the key features of Cu-based materials and allow molecular-level tunability. This may also address the large overpotential required for the Cu electrode and the instability issue of the molecular catalyst.
(4) 3D Cu porous networks. The pore length and pore size of 3D Cu porous structures will inuence the diffusion of reactants and intermediates. Tailoring the pore length as well as the pore size from the macropore to mesopore, micropore, and nanopore region will tune the diffusion and, in turn, the performance of the catalyst. In addition, through designing abundant active sites in this porous network to prolong the retention time of specic intermediates, we could expect high selectivity for specic products.
(5) Cu nanoclusters. Although the size effects of Cu nanoparticles have been discussed widely, ultrasmall, atomically precisely controlled Cu nanoclusters are rarely reported. Similar to Au and Ag nanoclusters, 239 investigating the reaction mechanism to get a fundamental understanding via the atomic precise control of Cu nanoclusters is critical to design highly efficient and stable Cu nanocluster electrocatalysts.
(6) Design of the ow cell. The fabrication of an efficient ow cell with a designed gas diffusion electrode (GDE) could dramatically promote the activity and stability of CO 2 electrocatalysts. More and more attention has been paid to this area over the past two years, 103,110,185,208,240 and future efforts could be also made towards this aspect.  Electrochemical reduction: CO + HCOOH 70-80% at À0.3 to À0.5 V; a peak of CO 61.8% at À0.4 V, HCOOH 30.7% at À0.6 V, current density 0.5-10 mA cm À2 at À0.3 V to À0.6 V  O-Cu on Si (1 mm) 0.1 M KHCO 3 , 5 sccm CO 2 Cycling Cu foil in KCl : C 2 /C 3 60% (C 2 H 4 34% with partial current density À7.5 mA cm À2 ), C1 6.5% at À1.0 V, current density À10.9 mA cm À2