Cutting-Edge Electrocatalysts for CO2RR

A world-wide growing concern relates to the rising levels of CO2 in the atmosphere that leads to devastating consequences for our environment. In addition to reducing emissions, one alternative strategy is the conversion of CO2 (via the CO2 Reduction Reaction, or CO2RR) into added-value chemicals, such as CO, HCOOH, C2H5OH, CH4, and more. Although this strategy is currently not economically feasible due to the high stability of the CO2 molecule, significant progress has been made to optimize this electrochemical conversion, especially in terms of finding a performing catalyst. In fact, many noble and non-noble metal-based systems have been investigated but achieving CO2 conversion with high faradaic efficiency (FE), high selectivity towards specific products (e.g., hydrocarbons), and maintaining long-term stability is still challenging. The situation is also aggravated by a concomitant hydrogen production reaction (HER), together with the cost and/or scarcity of some catalysts. This review aims to present, among the most recent studies, some of the best-performing catalysts for CO2RR. By discussing the reasons behind their performances, and relating them to their composition and structural features, some key qualities for an “optimal catalyst” can be defined, which, in turn, will help render the conversion of CO2 a practical, as well as economically feasible process.


Introduction
The day-to-day increase in levels of CO 2 , predominantly produced by anthropogenic processes, poses serious environmental issues, and simply relying on nature (i.e., photosynthesis via plants) is no longer sufficient to stabilize the CO 2 present in the air. The current CO 2 concentration in the Earth's atmosphere is approximately 417 ppm, and it has almost doubled since the 1760s [1]. This has led to devastating effects in terms of climate change, which in turn is causing melting glaciers, rising sea levels, destruction of natural habitats, and flooding, among others [2]. In order to improve the sustainability of our environment, besides limiting anthropogenic CO 2 emissions, one effective solution to reduce CO 2 levels relies on its capture and subsequent electrocatalytic conversion (also known as CO 2 reduction reaction, or CO 2 RR) into value-added chemicals. These products can be very diverse, as schematically shown in Figure 1, and can be used as fuels, as versatile chemical intermediates, preservatives, pesticides, in cosmetics, and more [3].
Current research is focusing more and more on optimizing CO 2 RR to make it practical but also economically feasible. However, there are several challenges to face, the main one related to the energy costs of the reduction step (the CO 2 molecule is thermodynamically stable, ∆ f G 298K = −394 KJ/mol) [4,5], while achieving a conversion with high selectivity and efficiency.
To give an idea of these challenges, Table 1 reports the standard potential required to convert CO 2 to various products.
Another challenge to face during the CO 2 conversion process is the concurrent production of hydrogen, which commonly derives from the competing hydrogen reaction (HER). Clearly, this reaction reduces efficiency and needs to be suppressed to maximize the final yield and selectivity of the desired products. Current research is focusing more and more on optimizing CO2RR to make i practical but also economically feasible. However, there are several challenges to face, th main one related to the energy costs of the reduction step (the CO2 molecule i thermodynamically stable, ΔfG298K = −394 KJ/mol) [4,5], while achieving a conversion with high selectivity and efficiency.
To give an idea of these challenges, Table 1 reports the standard potential required to convert CO2 to various products. Table 1. The standard potentials for the electrochemical reduction of CO2 [6].
These challenges can be mitigated by employing a suitable catalyst that is able t facilitate the breakage of the CO2 bonds, address the conversion toward specific product (selectively), suppress the HER, and remain stable throughout the process.
This review aims to present some recent studies on selected, most promising catalyst for CO2RR. By discussing the reasons behind their performance, we wish to delineate th next generation of electrocatalysts for rendering the conversion of CO2 an economicall  Table 1. The standard potentials for the electrochemical reduction of CO 2 [6]. These challenges can be mitigated by employing a suitable catalyst that is able to facilitate the breakage of the CO 2 bonds, address the conversion toward specific products (selectively), suppress the HER, and remain stable throughout the process.

Reaction
This review aims to present some recent studies on selected, most promising catalysts for CO 2 RR. By discussing the reasons behind their performance, we wish to delineate the next generation of electrocatalysts for rendering the conversion of CO 2 an economically feasible process.

Mechanism for Electrochemical Reduction of CO 2
In general, the electrochemical reduction of CO 2 in an aqueous solution is a multielectron transfer process that enables CO 2 to convert into several gaseous and liquid products, as shown in Figure 1 [7,8]. The final product obtained is dependent on several factors, such as the nature of the electrocatalyst (discussed in later chapters), and the electrolytic reaction conditions, including the electrolyte used [9], the applied potential, and the type of cell used for the setup (i.e., flow cell [10], H-cell [11]). To simply explain the mechanism in three steps: (1) CO 2 adsorbs and interacts with surface atoms of the catalyst; (2) CO 2 is activated, and the reduction proceeds with aid of the catalyst-initiated proton transfer to generate intermediates such as *CO 2 , *COOH, *CO, and others; and (3) the final product is desorbed, and the recovery of the catalyst surface takes place. These intermediates are crucial for CO 2 RR to lead to the final desired product (Figure 2). For instance, the generation of so-called C1 products involves the *CO 2 intermediate interacting with a proton to form *COOH favoring the production of CO. Other intermediates, such as *CO, are invaluable to both the C1 and C2 pathways. To obtain C1 products, *CO can obtain a proton to generate *CHO intermediate, followed by three proton/electron transfers to form CH 3 OH. Moreover, the *CO intermediate can also participate in an additional step known as the C-C coupling step, leading to the production of C2 products [12]. electrolytic reaction conditions, including the electrolyte used [9], the applied potential, and the type of cell used for the setup (i.e., flow cell [10], H-cell [11]). To simply explain the mechanism in three steps: (1) CO2 adsorbs and interacts with surface atoms of the catalyst; (2) CO2 is activated, and the reduction proceeds with aid of the catalyst-initiated proton transfer to generate intermediates such as *CO2, *COOH, *CO, and others; and (3) the final product is desorbed, and the recovery of the catalyst surface takes place. These intermediates are crucial for CO2RR to lead to the final desired product ( Figure 2). For instance, the generation of so-called C1 products involves the *CO2 intermediate interacting with a proton to form *COOH favoring the production of CO. Other intermediates, such as *CO, are invaluable to both the C1 and C2 pathways. To obtain C1 products, *CO can obtain a proton to generate *CHO intermediate, followed by three proton/electron transfers to form CH3OH. Moreover, the *CO intermediate can also participate in an additional step known as the C-C coupling step, leading to the production of C2 products [12].

Noble Metal-Based Nanosized Catalysts for CO 2 RR
A suitable catalyst for CO 2 RR must guarantee high selectivity while maximizing efficiency (yield/conversion). In this respect, noble metal-based catalysts such as Au [13], Ag [14], Pd [15], Rh [16], and Ir [17] were proved to have both excellent activity and high selectivity towards the formation of CO and formate (see Table 2). For instance, Au nanoparticles (NPs) with sizes ranging between 4 and 10 nm [13] led to a selective electrocatalytic reduction of CO 2 to CO, with FE 90% at −0.67 V (Figure 3). Furthermore, Au nanoparticles performed better than bulk Au; in fact, as observed by Kauffman et al. [19], bulk Au is barely active, showing a faradaic efficiency of 3% towards CO at a similar potential of −0.675 V vs. RHE. Interestingly, when Au NPs were embedded in a matrix of butyl-3-methylimidazolium hexafluorophosphate (BMIM-PF 6 ), the FE towards CO increased by 7%, whilst simultaneously the H 2 production was inhibited [13]. This suggests that the electrocatalytic performance for CO 2 reduction is size-dependent but can be further enhanced in the presence of a functional support. Zhu et al. [13] attributed the better activity of the 8 nm nanoparticles to a higher number of edge sites (active sites for CO formation) than corner sites (active sites for hydrogen evolution). Lu et al. [14] studied the activity of nanoporous silver, comparing its performance with that of polycrystalline silver at different potentials. Similar to what observed by Zhu et al. [13], the nanoporous silver electrocatalyst was also more active, with a higher production of CO (FE of 92% at −0.6 V) compared to polycrystalline Ag (FE of 1.1% at the same potential). The study at lower overpotentials showed a decreasing efficiency toward CO formation and a greater production of hydrogen. On the other hand, higher overpotentials led to the formation of formate alongside CO, although hydrogen was still the primary product.
(POCOP)Ir(H)(HSiR3) CH4 --For instance, Au nanoparticles (NPs) with sizes ranging between 4 a led to a selective electrocatalytic reduction of CO2 to CO, with FE 90% at − 3). Furthermore, Au nanoparticles performed better than bulk Au; in fact, Kauffman et al. [19], bulk Au is barely active, showing a faradaic efficiency CO at a similar potential of −0.675 V vs. RHE. Interestingly, when Au NPs w in a matrix of butyl-3-methylimidazolium hexafluorophosphate (BMIM towards CO increased by 7%, whilst simultaneously the H2 production was This suggests that the electrocatalytic performance for CO2 reduction is s but can be further enhanced in the presence of a functional support. Z attributed the better activity of the 8 nm nanoparticles to a higher numbe (active sites for CO formation) than corner sites (active sites for hydrogen et al. [14] studied the activity of nanoporous silver, comparing its perform of polycrystalline silver at different potentials. Similar to what observed by the nanoporous silver electrocatalyst was also more active, with a higher CO (FE of 92% at −0.6 V) compared to polycrystalline Ag (FE of 1.1% at the s The study at lower overpotentials showed a decreasing efficiency toward and a greater production of hydrogen. On the other hand, higher overpoten formation of formate alongside CO, although hydrogen was still the prima  (c) Potential-dependent FEs of the C-Au on electrocatalytic reduction of CO 2 to CO. (d) Current densities for CO formation (mass activities) on the C-Au at various potentials. Reprinted (adapted) with permission from [13]. Copyright 2013 American Chemical Society.
Through computational studies, the role of surface morphology to maximize electrocatalytic performances was studied using palladium nanoparticles with sharp geometric features. Here, the predictions from DFT calculations were determined, and it was found that structures with more edges and grain boundaries have enhanced catalytic activity [15]. In this study, the Pd(211) plane exhibited the lowest energy barrier and, as a result, led to high catalytic activity towards formate production while suppressing CO formation. Following these theoretical predictions, Klinkova et al. [15] undertook experiments to develop Pd NPs with different shapes (see SEM images in Figure 4), and different types of stabilized facets: (100) plane-enclosed nanocubes (NC), (110) plane-enclosed rhombic dodecahedra (RDs), NPs with mixed low-index facets, and branched NPs enclosed by high-index facets (BNP). Klinkova et al. [15] showed experimentally that branched (BNP) Pd NPs surrounded by high index facets performed in agreement with theoretical predictions, reaching an impressive selectivity and a FE of 97% towards formate production at −0.2 V. Importantly, no CO production was detected. electrocatalytic performances was studied using palladium nanoparticles with s geometric features. Here, the predictions from DFT calculations were determined, an was found that structures with more edges and grain boundaries have enhanced cata activity [15]. In this study, the Pd(211) plane exhibited the lowest energy barrier and, result, led to high catalytic activity towards formate production while suppressing formation. Following these theoretical predictions, Klinkova et al. [15] under experiments to develop Pd NPs with different shapes (see SEM images in Figure 4), different types of stabilized facets: (100) plane-enclosed nanocubes (NC), (110) pl enclosed rhombic dodecahedra (RDs), NPs with mixed low-index facets, and branc NPs enclosed by high-index facets (BNP). Klinkova et al. [15] showed experimentally branched (BNP) Pd NPs surrounded by high index facets performed in agreement theoretical predictions, reaching an impressive selectivity and a FE of 97% tow formate production at −0.2 V. Importantly, no CO production was detected. These studies confirm that noble metal-based catalysts have high faradaic effici and selectivity, at moderate potentials toward CO and formate, and the main factors influence the catalytic activity are the size of NPs, their potential-dependent selecti shape, and surface morphology. However, their main disadvantages cannot be igno i.e., their costs and scarcity, which make them less suitable for large-scale applicati along with their tendency to be poisoned.

Non-Noble Metal-Based Electrocatalysts for CO2RR
The costs related to CO2RR can be clearly lowered by using non-noble metal-b electrocatalysts. In this respect, several catalysts have been tested, including pure Sn Ni [21], Fe [22], Zn [23], and Cu [24] (see Table 3). Loading non-noble metal nanopart on suitable carbon supports (e.g., carbon black or graphene) ensures a higher surface and porosity, which eases CO2 transportation and subsequent reduction. Zhang et al. discovered that 5 nm SnO2 NPs loaded onto graphene, rather than carbon black, enhan activity [20], as evidenced by the higher FE achieved toward formate when a graph support was used (93.6%), compared to carbon black (86.2%), both at −1.8 V, possibly to the higher conductivity of graphene. These studies confirm that noble metal-based catalysts have high faradaic efficiency and selectivity, at moderate potentials toward CO and formate, and the main factors that influence the catalytic activity are the size of NPs, their potential-dependent selectivity, shape, and surface morphology. However, their main disadvantages cannot be ignored, i.e., their costs and scarcity, which make them less suitable for large-scale applications, along with their tendency to be poisoned.

Non-Noble Metal-Based Electrocatalysts for CO 2 RR
The costs related to CO 2 RR can be clearly lowered by using non-noble metal-based electrocatalysts. In this respect, several catalysts have been tested, including pure Sn [20], Ni [21], Fe [22], Zn [23], and Cu [24] (see Table 3). Loading non-noble metal nanoparticles on suitable carbon supports (e.g., carbon black or graphene) ensures a higher surface area and porosity, which eases CO 2 transportation and subsequent reduction. Zhang et al. [20] discovered that 5 nm SnO 2 NPs loaded onto graphene, rather than carbon black, enhanced activity [20], as evidenced by the higher FE achieved toward formate when a graphene support was used (93.6%), compared to carbon black (86.2%), both at −1.8 V, possibly due to the higher conductivity of graphene.
This study showed that the type of loading matrix used plays a key role by providing better conductivity, a stronger electronic interaction between the support and the metal nanoparticles and enhancing electronic donation. In this specific case, one can speculate that the stronger electron-donating ability of graphene, compared to carbon black, improves the CO 2 reduction on the Sn surface. Unfortunately, hydrogen production was significant, thus lowering the overall efficiency. It is interesting to note, SnO 2 NPs below 5 nm showed lower selectivity towards formate (to FE~62%), even lower compared to Sn foil (FE of 30%). It is evident that the particle size of the Sn catalysts has an influence on the CO 2 reduction efficiencies ( Figure 5). The maximum efficiencies achieved by the 5 nm nano-SnO 2 catalyst were explained by the affinity between the surface-bound key intermediates and the catalyst, facilitating CO 2 reduction. Remarkably, SnO 2 NPs were found to be stable during electrolysis and had the ability to continuously produce formate for at least 18 h at 1.8 V. By comparison of the controlled potential electrolysis between the SnO 2 NPs/carbon black and the SnO 2 NPs/graphene, the steady state catalytic current density was twice as high for the latter, signifying the importance of support during CO 2 RR to improve performance. The stability of the catalyst after CO 2 RR was further confirmed by TEM images and LSV, indicating no significant changes regarding its morphology and catalytic property. the CO2 reduction efficiencies ( Figure 5). The maximum nano-SnO2 catalyst were explained by the affinity intermediates and the catalyst, facilitating CO2 reduct found to be stable during electrolysis and had the ability for at least 18 h at 1.8 V. By comparison of the control the SnO2 NPs/carbon black and the SnO2 NPs/graphene density was twice as high for the latter, signifying th CO2RR to improve performance. The stability of the confirmed by TEM images and LSV, indicating no s morphology and catalytic property. Nanoporous zinc oxide (ZnO) prepared via the hy decomposition was also tested as a CO2RR alternative Additionally, in this study, the main products were C ZnO was reduced to Zn, a greater faradaic efficiency of 9 V was achieved, also compared to commercial Zn (F potential. These findings were adduced to the proper surface area and high density of unsaturated, coordinat selectivity toward CO formation was explained by physiosorbed CO2 (a linear molecule) to chemisorbed the surface defects and alkali metal promoted surfaces, f and stabilization of bent CO2 δ− intermediates on the Nanoporous zinc oxide (ZnO) prepared via the hydrothermal method and thermal decomposition was also tested as a CO 2 RR alternative electrocatalyst by Jiang et al. [23]. Additionally, in this study, the main products were CO and H 2 . When the nanoporous ZnO was reduced to Zn, a greater faradaic efficiency of 92% toward CO formation at −1.66 V was achieved, also compared to commercial Zn (FE of 55.5%) at the same applied potential. These findings were adduced to the properties of nanoporous Zn, i.e., high surface area and high density of unsaturated, coordinated surface atoms. Hence, the high selectivity toward CO formation was explained by an increase in the activation of physiosorbed CO 2 (a linear molecule) to chemisorbed CO 2 (a bent molecule), caused by the surface defects and alkali metal promoted surfaces, facilitating the increased formation and stabilization of bent CO 2 δ− intermediates on the coordination unsaturated surface atoms [23]. Based on the studies reported here, besides the mentioned advantages, non-noble metal-based catalysts showed a higher FE and selectivity towards CO and HCOO − . However, their performances were limited by their low selectivity and efficiency towards hydrocarbons and alcohol formation (so-called 2-C, 3-C, and 4-C products). Table 3. Some examples of different non-noble metal-based electrocatalysts for CO 2 RR.

Copper-Based Catalysts for CO 2 RR
Among non-precious metals that have been studied, copper-based materials are considered the best electrocatalyst choice for the conversion of CO 2 , also thanks to lower costs (compared to noble metals) and a relative abundance, with over 200 years of supply still left [36]. The abundance of Cu has a direct effect on its price, estimated at 0.20 GBP/oz (as of 29 March 2023) compared, for instance, to palladium or platinum, with prices estimated at around 1170 GBP/oz and 788 GBP/oz, respectively [37].
Carbon-based materials, including porous carbon, graphene, carbon nanotubes, and modified diamond, have shown different performances for CO 2 reduction due to their different crystallinity, surface area, tunable chemical and physical properties, and good conductivity [38]. For instance, oxide carbon nanotubes can selectively convert CO 2 into acetic acid with a faradaic efficiency of 71.3% [39]. However, from an economic perspective, the cost of these materials makes them not economically advantageous for practical applications; e.g., the price of 5 g of carbon nanotubes can range from GBP 59 to 259 [40]. Similarly, graphene and modified diamond are also expensive and thus less suitable for large-scale applications [41,42]. On the other hand, the Cu and Cu-based materials reported in this manuscript have been selected because they are readily available, inexpensive, and abundant. In addition to these advantages, the activity of Cu-based catalysts for electrocatalytic CO 2 reduction has been established [43]. The unique 3d electronic structure of Cu allows a suitable amount of binding energy between the catalyst and the CO 2 molecule for the activation of the CO 2 to generate an activated *CO species. Furthermore, Cu-based catalysts are able not only to reduce CO 2 simply into CO or short-chain hydrocarbons, but they also allow C-C coupling (by *CO dimerization), thus allowing the preparation of more complex C2+ products.
An overview of the timeline of the development of copper-based materials is reported in Figure 6. As for other metals discussed previously, pure copper also lacks selectivity toward specific products, which is why a second metal is usually incorporated to modify it structure, for instance, by metal doping, alloying, changing morphology (e.g., core/shell) adjusting crystallinity [58], and more, with the aim to exploit the resulting synergistic strain, and alloying effects (Figure 7). The incorporation of one or more metals into the Cu structure has been shown to improve stability, increase selectivity and activity, and minimize energy consumption, overall leading to a more efficient CO2RR, as discussed in the following chapters. As for other metals discussed previously, pure copper also lacks selectivity towards specific products, which is why a second metal is usually incorporated to modify its structure, for instance, by metal doping, alloying, changing morphology (e.g., core/shell), adjusting crystallinity [58], and more, with the aim to exploit the resulting synergistic, strain, and alloying effects (Figure 7). The incorporation of one or more metals into the Cu structure has been shown to improve stability, increase selectivity and activity, and minimize energy consumption, overall leading to a more efficient CO 2 RR, as discussed in the following chapters. There are several ways to synthesize Cu electrocatalysts; these include the use of microwave [59] and electron beam irradiation [60], laser ablation [61], thermal decomposition [62], in situ chemical synthetic routes [63], use of microemulsions [64], metal salt reduction [65], and sol-gel based processes [57], just to cite some. However, some of these methods are costly, lengthy, unable to control particles' size, morphology, and/or crystallinity, or just not suitable for large-scale manufacture.

Copper-Based Alloys
Kim et al. [66] investigated the effect of the addition of gold to copper when preparing Cu-Au nanoparticles and found that the presence of gold influences the overall activity toward CO2RR. They showed that the more Au is incorporated into Cu nanoparticles, the more the formation of methane and ethylene declines, while the FE toward CO increases and H2 production is inhibited. By tuning the composition of Au-Cu bimetallic nanoparticles, the degree of stabilization of the intermediates on the nanoparticle surfaces is also affected, and, as a result, different products are favored. Kim et al. [66] also pointed out that the tested Au3Cu showed a high FE of 65% for CO, similar to pure Au NPs of similar sizes [13]. The outcome was elucidated considering the correlation between the nanoparticle's composition and two other factors: (1) the electronic effect and (2) the geometric effect. The change in the electronic structure of the catalyst is influenced by the electronic effect on the binding strength of intermediates. The geometric effect is the local atomic arrangement at the active site. The way the active site is configured can affect the binding strength of intermediates. Therefore, both geometric and electronic effects must work synergistically to improve CO2 reduction [66]. Moreover, the stability of the Au3Cu catalyst was deduced from the total current as a function of time at −0.73 V vs. RHE. Despite stability in the current for 10 h, a steady decline in activity/selectivity towards CO production was observed following the first hour.
Ma et al. [67]    There are several ways to synthesize Cu electrocatalysts; these include the use of microwave [59] and electron beam irradiation [60], laser ablation [61], thermal decomposition [62], in situ chemical synthetic routes [63], use of microemulsions [64], metal salt reduction [65], and sol-gel based processes [57], just to cite some. However, some of these methods are costly, lengthy, unable to control particles' size, morphology, and/or crystallinity, or just not suitable for large-scale manufacture.

Copper-Based Alloys
Kim et al. [66] investigated the effect of the addition of gold to copper when preparing Cu-Au nanoparticles and found that the presence of gold influences the overall activity toward CO 2 RR. They showed that the more Au is incorporated into Cu nanoparticles, the more the formation of methane and ethylene declines, while the FE toward CO increases and H 2 production is inhibited. By tuning the composition of Au-Cu bimetallic nanoparticles, the degree of stabilization of the intermediates on the nanoparticle surfaces is also affected, and, as a result, different products are favored. Kim et al. [66] also pointed out that the tested Au 3 Cu showed a high FE of 65% for CO, similar to pure Au NPs of similar sizes [13]. The outcome was elucidated considering the correlation between the nanoparticle's composition and two other factors: (1) the electronic effect and (2) the geometric effect. The change in the electronic structure of the catalyst is influenced by the electronic effect on the binding strength of intermediates. The geometric effect is the local atomic arrangement at the active site. The way the active site is configured can affect the binding strength of intermediates. Therefore, both geometric and electronic effects must work synergistically to improve CO 2 reduction [66]. Moreover, the stability of the Au 3 Cu catalyst was deduced from the total current as a function of time at −0.73 V vs. RHE. Despite stability in the current for 10 h, a steady decline in activity/selectivity towards CO production was observed following the first hour.
Ma et al. [67] studied geometric effects and the way different mixing patterns can influence selectivity. The mixing patterns for the Cu-Pd catalyst were: ordered, disordered, and phase separated, with compositions ranging from 3:1 to 1:3. Here, it was observed that the more Cu present favored the formation of C2 products, confirming that composition has an influence on selectivity. Interestingly, the product formed depended on the type of mixing pattern used (Figure 8). The ordered CuPd presented the best selectivity for CO formation (FE of 80%), whereas the disordered CuPd showed poor selectivity for CO formation. However, the phase-separated CuPd demonstrated selectivity towards C 2 H 4 and C 2 H 5 OH with FE values of~50% and~13%, respectively. Ma et al. [67] hypothesize that the electronic effects induced by the mixing of Pd with Cu resulted in different FE for different products. This was further investigated by collecting surface valence band photoemission spectra of all mixing patterns of Cu-Pd catalysts (including the corresponding monometallic ones), which showed that the phase-separated Cu-Pd catalyst has weaker binding between the CO intermediate and the catalyst surface, whereas monometallic Cu has a stronger CO binding. Despite these differences, both phases separated Cu-Pd and Cu NPs showed similar catalytic selectivity. This study showed that the geometric/structure effect had a more significant influence on selectivity than the electronic effect.
composition has an influence on selectivity. Interestingly, the product formed d on the type of mixing pattern used (Figure 8). The ordered CuPd presented selectivity for CO formation (FE of 80%), whereas the disordered CuPd show selectivity for CO formation. However, the phase-separated CuPd demo selectivity towards C2H4 and C2H5OH with FE values of ~50% and ~13%, respecti et al. [67] hypothesize that the electronic effects induced by the mixing of Pd resulted in different FE for different products. This was further investigated by c surface valence band photoemission spectra of all mixing patterns of Cu-Pd (including the corresponding monometallic ones), which showed that the separated Cu-Pd catalyst has weaker binding between the CO intermediate catalyst surface, whereas monometallic Cu has a stronger CO binding. Desp differences, both phases separated Cu-Pd and Cu NPs showed similar catalytic se This study showed that the geometric/structure effect had a more significant infl selectivity than the electronic effect. The conclusion was that when the Cu atoms were in close proximity to the P it favored alcohol and hydrocarbon formation, whereas when the Cu atom alternating with Pd atoms, it favored CO and CH4 formation. From this study observe how adjustments over composition, alongside geometric and electroni can enhance selectivity. Yet, alloying noble metals (i.e., Au, Pd) to copper is still r expensive and difficult to manufacture on a large scale.

Copper Alloyed with Non-Noble Metals
Alloying Cu with non-precious metals, such as Ni [68], Fe [69], Sn [70], Zn [55,72], and many others, can further lower the production price. Tan et synthesized CuNi nanoparticles embedded in a three-dimensional (3D) nitroge network and found outstanding performances with regards to CO2RR. It was o that CuNi NPs can selectively convert CO2 into CO (FE 94.5%) at low poten outperform their corresponding mono-metal catalysts. The high selectivity was e by the 3D nitrogen-carbon network improving the CO2 adsorption capacity of th The conclusion was that when the Cu atoms were in close proximity to the Pd atoms, it favored alcohol and hydrocarbon formation, whereas when the Cu atoms were alternating with Pd atoms, it favored CO and CH 4 formation. From this study, we can observe how adjustments over composition, alongside geometric and electronic effects, can enhance selectivity. Yet, alloying noble metals (i.e., Au, Pd) to copper is still relatively expensive and difficult to manufacture on a large scale.

Copper Alloyed with Non-Noble Metals
Alloying Cu with non-precious metals, such as Ni [68], Fe [69], Sn [70], Zn [71], Bi [55,72], and many others, can further lower the production price. Tan et al. [68] synthesized CuNi nanoparticles embedded in a three-dimensional (3D) nitrogen-carbon network and found outstanding performances with regards to CO 2 RR. It was observed that CuNi NPs can selectively convert CO 2 into CO (FE 94.5%) at low potentials and outperform their corresponding mono-metal catalysts. The high selectivity was explained by the 3D nitrogen-carbon network improving the CO 2 adsorption capacity of the system (i.e., more CO 2 being adsorbed on the surface of the catalyst) and enhancing the selectivity towards specific products, as well as improving the stability of the catalyst. The performance of the CuNi catalyst was improved with regards to the catalyst's stability, demonstrating a stability of constant potential electrolysis for over 38 h at −0.6 V vs. RHE. From a structural and compositional point of view, the catalyst remains unchanged after CO 2 RR, as confirmed by XRD and TEM. However, even though catalytic performances were outstanding, the synthesis of CuNi NPs embedded in the nitrogen-carbon network is very time-consuming, requiring a total time of about 50 h.

Copper-Based Core/Shell Systems
Changing structure and morphology can also lead to improved performance. Thus, alongside copper alloys, core/shell systems were studied. Compared to alloys, core-shell structures enable additional control over the core size and the thickness of the shell, which can influence the electrochemical reduction of CO 2 . Li et al. [70] synthesized Cu core/SnO 2 shell NPs using the seed-mediated method, where the synthesis of Cu NPs is followed by the decomposition of Sn(acac) 2 [70]. One advantage of this method is that it allows precise control over size and leads to core/shell structures with adjustable shell thickness [73]. The core copper NPs were kept at a controlled size of 7 nm, and the thickness of the SnO 2 shell was changed by adjusting the amount of Sn(acac) 2 added, with ±1 nm final precision. This study showed that both activity and selectivity were thickness dependent ( Figure 9). In fact, when the thickness of the SnO 2 coating was above 1 nm, the favored product was formate (FE of 85% at −0.9 V). Whereas, when the SnO 2 coating was below 1 nm, the main product was CO (FE 93% at −0.7 V). Unfortunately, the drawback of this method is the instability of the seeds [74]. Yet, Cu/SnO 2 (0.8 nm shell) demonstrated good stability for 10 h at −0.6 V, and the core/shell structure remains intact even after CO 2 RR, as confirmed by CV data (before and after CO 2 test showing similar surface redox potential) and TEM elemental mapping. (i.e., more CO2 being adsorbed on the surface of the catalyst) and enhancing the selectivity towards specific products, as well as improving the stability of the catalyst. The performance of the CuNi catalyst was improved with regards to the catalyst's stability, demonstrating a stability of constant potential electrolysis for over 38 h at −0.6 V vs. RHE. From a structural and compositional point of view, the catalyst remains unchanged after CO2RR, as confirmed by XRD and TEM. However, even though catalytic performances were outstanding, the synthesis of CuNi NPs embedded in the nitrogen-carbon network is very time-consuming, requiring a total time of about 50 h.

Copper-Based Core/Shell Systems
Changing structure and morphology can also lead to improved performance. Thus, alongside copper alloys, core/shell systems were studied. Compared to alloys, core-shell structures enable additional control over the core size and the thickness of the shell, which can influence the electrochemical reduction of CO2. Li et al. [70] synthesized Cu core/SnO2 shell NPs using the seed-mediated method, where the synthesis of Cu NPs is followed by the decomposition of Sn(acac)2 [70]. One advantage of this method is that it allows precise control over size and leads to core/shell structures with adjustable shell thickness [73]. The core copper NPs were kept at a controlled size of 7 nm, and the thickness of the SnO2 shell was changed by adjusting the amount of Sn(acac)2 added, with ±1 nm final precision. This study showed that both activity and selectivity were thickness dependent ( Figure 9). In fact, when the thickness of the SnO2 coating was above 1 nm, the favored product was formate (FE of 85% at −0.9 V). Whereas, when the SnO2 coating was below 1 nm, the main product was CO (FE 93% at −0.7 V). Unfortunately, the drawback of this method is the instability of the seeds [74]. Yet, Cu/SnO2 (0.8 nm shell) demonstrated good stability for 10 h at −0.6 V, and the core/shell structure remains intact even after CO2RR, as confirmed by CV data (before and after CO2 test showing similar surface redox potential) and TEM elemental mapping. Similarly, Zhang et al. [75] explored Cu@Ag core/shell nanoparticles and also found that tuning the thickness of the shell was a crucial contributor to the selectivity of CO2RR. Similarly, Zhang et al. [75] explored Cu@Ag core/shell nanoparticles and also found that tuning the thickness of the shell was a crucial contributor to the selectivity of CO 2 RR. Optimized performances were observed for Cu@Ag 2 NPs with FE of 67.6% and 32.2% toward C2 products and ethylene at −1.1 V vs. RHE, respectively. Increasing the thickness of the Ag shell to Cu@Ag 3 or Cu@Ag 4 resulted in an increase in CO formation (with similar selectivity as pure Ag NPs [14,76]), while C2 products decreased. Again, selectivity was shown to be thickness-dependent but also influenced by the choice of metal. In fact, while the previous study on Cu@Sn showed that CO and formate were mostly preferred, the Cu@Ag system favours the production of C 2 H 4 and other C2 compounds (including acetate and ethanol). One explanation for the observed difference in selectivity can be ascribed to the core/shell structure of the Cu@Ag, which can function as "a tandem catalyst", i.e., CO 2 is first attached to the Ag shell, activated, then reduced into CO, followed by its conversion into C2+ products on the Cu core. The stability of the Cu@Ag electrocatalyst was tested by performing chronoamperometry measurements (i.e., constant applied potential as a function of time under CO 2 reaction conditions). The measurements showed excellent durability with a steady current density and, more importantly, consistent production of C 2 H 4 (FE maintained about~30%) for 14 h at an applied potential −1.1 V vs. RHE for the Cu@Ag 2 catalyst. To prove the stability of the catalyst here, TEM and XRD were performed after CO 2 RR, showing no changes in the composition or morphology of the catalyst. Although desirable products were formed when alloying Ag with Cu, Ag is an expensive, precious metal, making it not practical for large-scale applications.
A different approach explored the influence of the elemental spatial distribution in bimetallic CuO x -ZnO nanowires after an in situ electrochemical reduction by Wan et al. [77]. Interestingly, they discovered that the phase-separated structural distribution possesses better activity, with higher faradaic efficiency towards CO (>90%), and greater stability over time compared to core/shell structures (FE > 80%) ( Figure 10). Unfortunately, the FE declines over time as competitive H 2 production increases [77]. Also interestingly, regarding stability, the phase-separated catalyst demonstrated longer durability with a constant rate of FE CO and a current density stabilized at 16 mA −2 for 15 h. In comparison, the CuZn coreshell structure was relatively unstable for long durations, and after about 2.5 h, a decline in FE CO and a rise in H 2 production was observed (see Figure 10e,f). In fact, the CuZn phaseseparated was more stable than CuZn core/shell, this was suggested by the observation that element redistribution occurred (after 20 min of the CO 2 RR) possibly due to the strain on the Zn atoms, with consequent precipitation of Zn, resulting in morphological changes, such as the appearance of dendritic structures (shown by SEM images) for the core/shell sample. The drastic change in morphology for the core/shell after CO 2 testing was reasoned to account for the observed inactivation in CO 2 performances. On the other hand, CuZn phase-separated reported no morphological transformation or elemental redistribution to be seen after CO 2 RR. Therefore, we can clearly conclude here that the type of structure has an influence on the final stability and catalytic performance [77].
Other studies (e.g., Ren et al. [50]) on Cu-Zn systems for CO 2 RR have found that introducing Zn dopant as a co-catalyst can assist copper catalysts to selectively convert CO 2 into ethanol. In fact, previous studies have shown that carbon monoxide (CO) can be reduced into ethanol on Cu NPs [78]. It was then showed that the presence of Zn dopants can increase the CO-producing sites in situ on oxide-derived copper catalysts during CO 2 RR for selective ethanol formation. It seems that the more Zn is added, the maximum faradaic efficiency towards ethanol increases (FE 29.1%), whereas an excess of Zn leads to a decrease in ethanol formation. Further studies were performed on other co-catalysts such as Ag and Ni on copper-based catalysts (where Ag is selective for CO formation and Ni is inactive towards CO 2 RR), but they showed a much lower selectivity compared to Zn towards ethanol(Zn > Ag > Ni). A suggestion for Ag having a low selectivity towards ethanol compared to Zn was deduced from their different CO binding strengths [79]. Whilst EtOH is a desirable product obtained from CO 2 , the catalytic stability of the Cu 4 Zn catalyst for EtOH production was a minimum of 5 h. In addition, significant morphological changes (confirmed by SEM) were observed following the first hour of CO 2 reduction, which were attributed to the relief of structural strain during the reduction process [50]. SAED patterns from TEM investigation confirmed that CuZn was no longer an alloy, but rather a phase segregation of Cu and Zn was seen after the reduction process.
process [50]. SAED patterns from TEM investigation confirmed that CuZn was no longer an alloy, but rather a phase segregation of Cu and Zn was seen after the reduction process. Other core/shell catalysts prepared and tested for CO2RR are reported in Table 4 with their synthetic methods. Generally, it is observed that they tend to favor C1 products, i.e., CO and formate. This selectivity can be however changed by tuning the thickness of the shell, which seems to influence the final product more than the choice of the metal. Advantage of seed growth: good control of NP size [81]. Advantages of the galvanic displacement method: fine particle and morphology control [82].  Other core/shell catalysts prepared and tested for CO 2 RR are reported in Table 4 with their synthetic methods. Generally, it is observed that they tend to favor C1 products, i.e., CO and formate. This selectivity can be however changed by tuning the thickness of the shell, which seems to influence the final product more than the choice of the metal.  Despite these promising results, the yield towards the preparation of more complex products remains a challenge, i.e., the so-called C2 (C 2 H 4 , C 2 H 5 OH), C3 (C 3 H 6 , C 3 H 7 OH), and C4 (C 4 H 7 , C 4 H 8 O) products.

The Influence of Supports
The electrochemical reduction of CO 2 necessitates that the Cu-based catalyst be placed on a conductive material to maximize electron transport. Common supports are carbon black [13], carbon nanotubes [87], and graphene [88]. To investigate the role of support, Li et al. [88] loaded Cu NPs onto three different matrices, namely Kejen black EC-300 carbon, graphene oxide, and pyridinic-N-rich graphene (p-NG). When the copper NPs deposited onto Kejen black EC-300 carbon were tested, the conversion of CO 2 to C 2 H 4 was below an FE of 10%. However, when the Cu NPs were placed onto the p-NG support, the selectivity and efficiency towards C 2 H 4 increased. The high selectivity towards C 2 H 4 was explained as a result of the synergistic effect between the p-NG support and Cu NPs. It was suggested that since the p-NG structure is a strong Lewis base, it can allow the protons to concentrate around the Cu and facilitates CO 2 to interact with Cu, resulting in CO 2 reduction and C-C coupling for the formation of ethylene. However, when p-NG alone (as a catalyst) was tested for CO 2 RR, it was found that p-NG favors the formation of formate and H 2 , thus highlighting the importance of the interaction between the Cu NPs and p-NG support.
Baturina et al. [48] also explored carbon-supported Cu nanoparticles for CO 2 RR, in particular: Vulcan Carbon (VC), Ketjen Black (KB), and Single-Wall Carbon Nanotubes (SWNTs), where Cu was either merely supported or electrodeposited. The carbonsupported Cu nanocatalysts achieved a selectivity toward higher ratios of C 2 H 4 :CH 4 compared to the electrodeposited (and smoother) Cu films ( Figure 11). The high selectivity towards C 2 H 4 was ascribed to the presence of so-called 'rough surfaces', which present more corners, edges, and defects, along with smaller particle sizes, than the smooth surface of the electrodeposited Cu film. However, the reasoning behind the different stability patterns observed for different catalysts was unexplained. Also according to this study, the production of C 2 H 4 and CH 4 does not significantly change with time, but H 2 production increases due to changes on the surface. The choice of support clearly has an influence on catalytic performances yet using supports like single-wall carbon nanotubes (SWNTs) can limit large-scale applications due to their costs (ranging from 125-300 USD/gram) [89].

Crystalline Versus Amorphous Cu Nanoparticles
So far, research has mainly focused on crystalline Cu-based catalysts, however it was observed that amorphous Cu-based catalysts might perform better toward CO 2 conversion compared to crystalline Cu nanoparticles, due to their lower-coordinated atoms, increased number of defects, and reactive sites (Table 5).

Crystalline Versus Amorphous Cu Nanoparticles
So far, research has mainly focused on crystalline Cu-based catalysts, however it was observed that amorphous Cu-based catalysts might perform better toward CO2 conversion compared to crystalline Cu nanoparticles, due to their lower-coordinated atoms, increased number of defects, and reactive sites (Table 5). Duan et al. [58] synthesized amorphous Cu nanoparticles and tested them for electrochemical reduction of CO 2. They found that the FE towards HCOOH, C 2 H 5 OH, and CO were 37%, 22%, and 5.8%, respectively, at −1.4 V, while crystalline Cu NPs achieved a lower FE towards HCOOH and C 2 H 5 OH at much lower potentials (compared to amorphous Cu nanoparticles). More notable, the amorphous Cu NPs exhibited a higher stability for up to 12 h at a constant applied potential of −1.4 V vs. RHE. In addition, the stability of amorphous Cu NPs was seen to be more attractive as negligible changes in composition, morphology, and catalytic property were identified by XRD, TEM, and LSV characterizations, respectively. In contrast, crystalline Cu NPs after 12 h of electrolysis showed poor stability t after 12 h due to an increase in crystallinity, agglomerated particles, as observed by XRD, TEM, and LSV, respectively. Duan et al. [58] suggested the superior performances showed by the amorphous Cu NPs relate to a higher electrochemical active area, arising from the abundant active sites, originating from the distinct electronic structure and intrinsic chemical heterogeneity on the surface of the amorphous Cu. The cyclic voltammetry (CV) data showed that the higher the electrochemical double layer capacitance (EDLC), the higher the electrochemical active area, which led to more active sites and allowed a greater adsorption of CO 2 compared to crystalline Cu NPs. Table 6 reports an overview of some Cu-based catalysts tested in CO 2 RR.

Conclusions
In the present review, selected electrocatalysts for the conversion of CO 2 (referred to as CO 2 RR) into added-value chemicals (including carbon monoxide, formic acid, methane, and ethanol) were discussed. Special emphasis was given to electrocatalysts with improved stability and high selectivity toward specific products over systems with high faradic efficiency (FE) only.
Amongst the various electrocatalysts discussed in this review, more focus was given to Cu and Cu-based catalysts, for these systems (more than others), can be tailored to produce a wider range of multi-carbon-based products, whilst simultaneously ensuring large-scale manufacture.
While, studies on noble metals (e.g., Au, Ag) confirmed good activity, high faradaic efficiency, and selectivity toward CO 2 RR, yet the main products were usually CO and formate. Even though it was shown that the applied potential can affect selectivity, the production of more complex hydrocarbons is still a challenge when using noble metals. In addition, these systems are in general costly, scarce, and have a tendency to undergo poisoning, making them unsuitable for large-scale applications.
From this perspective, non-noble metal-based catalysts are a preferable choice, thanks to their lower costs and readiness, but they can still show low selectivity and efficiency, especially towards hydrocarbons and alcohol formation (the so-called 2-C, 3-C, and 4-C products). Nonetheless, performances can be enhanced by using nanoparticles, rather than bulk systems. Nanoparticles' size was shown to be one of the main factors influencing the system's catalytic activity, alongside the particles' shape, surface arrangements, degree of crystallinity, and general surface order/disorder of the final structure. Even surface roughness was shown to play a role, leading to different catalytic performances; i.e. CO 2 conversion was shown to be favoured by the presence of defects, corners, and edges, commonly found in rough surface-based catalysts, which resulted in better catalytic performances compared to smooth surfaces.
The importance of moderate potentials (0.6-0.7 V) was also shown for non-noble metals, e.g., on Sn catalysts. Adjustments over composition and atomic arrangement, alongside geometric and electronic effects, can enhance selectivity, stability, and efficiency. Studies on CuZn catalysts, for instance, showed how some factors, such as metal element distribution and the use of co-catalysts can influence selectivity. It was also shown that tailoring morphologies may have a significant effect. Studies on core-shell structures (e.g., on CuSnO 2 ) showed that selectivity can be changed by tuning the thickness of the particles' shell, which in some cases influenced the final product more than the choice of the metal itself.
The important role of functional support was also discussed. Supports can serve to improve conductivity and general electron transport in the final system, can improve the diffusion path, as well as enhance stability. For instance, it was observed that using a more structured matrix, such as graphene, led to better overall performance than using mere amorphous carbon.
In our discussion, we have also pointed out how multi-metallic systems (e.g., bimetallic or alloys) showed better electrocatalytic activity, outperforming the pure metals, thanks to synergistic effects, possibly resulting from geometric and electronic factors, and an increased number of active sites. These so-called "tandem catalysis" also showed better selectivity, along with higher activity.
Clearly, rather than a screening approach, a dedicated study is needed to develop outperforming catalysts. The studies undertaken so far have shown that the best choice relies on multi-metallic systems with a suitable choice of metals, composition, size, and atomic arrangement to maximize efficiency and selectivity towards multi-carbon-based products. Special attention should also be paid to the synthetic method and/or preparation technique, for it is surely a contributing factor to the final performance, whilst also considering the eco-friendliness of the process, total time, and final cost.