Figure 1 shows the replotted contour image of the current density of FA, CO, and H₂ production in electrocatalytic CO2R over the modified Cu cathode [35]. Two interesting features are noted in Figure 1. First, FA is the major CO2R species on Cu-based cathodes as its value is about tenfold higher than that of CO. Second, the hot zone (the red area) of the hydrogen current occurs at the left-handed side of the figure that corresponds to the region with the lowest CO current. This clearly indicates that electrocatalytic hydrogen production profoundly competes with CO evolution. Importantly, the hot zone of hydrogen evolution appears in the region with a moderate FA current (0.5~2.5 mA/cm²). This means that FA formation is less affected by hydrogen evolution than that of the CO formation [13]. It must be mentioned that increasing operation potential inevitably leads to high H₂ yield (Figure 2) [37]. Due to thermodynamics restrictions, FA and CO evolution in CO2R is suggested to be carried out at a relatively low potential condition.

Engineering electrochemical properties of cathode materials enhances the selectivity. Designing an efficient catalyst for CO2R is a highly technical challenge due to the strong completion from the hydrogen evolution reaction (HER) [3]. Factors such as catalytic reactivity, product selectivity, Faradaic efficiency, catalytic stability, and reduction mechanisms are crucial to controlling the efficiency of CO2R [38]. A systematical evaluation of the effect of catalyst structure on reaction selectivity is hence highly desirable [4]. Taking the electrocatalytic CO2R molecular system as an example, the energy required to dissociate an M–H bond to form a hydride is the key parameter in CO2R selectivity [2,4,39,40]. Figure 3 shows pathways regulating the transfer of either two protons (for CO evolution) or two electrons (for formic acid formation) and in both cases hydrogen evolution is always the major competitive side reaction.

In this framework, the preferential interaction between the catalytic metal center and CO₂ over protons is responsible for the selectivity for CO evolution. While the moderate hydricity facilitates CO₂ insertion into M–H bonds for FA production, strong hydride donors catalyze H₂ formation [2,4,39,40]. It has been suggested that catalytic activity requires the presence of a weakly coordinating solvent molecule that can readily become dissociated during the catalytic cycle as to provide a vacant coordination site for water binding and assisting C−O bond cleavage [4]. Generally, H₂ and FA formations are favorable reactions in aqueous solutions [39]. In fact, by plotting the hydricity as a function of individual free energy for the one-electron reduction of the parent species, a linear correlation appears indicating high FA selectivity over CO formation. Importantly, this correlation extends over a wide range of metals, ligand architectures, structural geometries, and overall charge of the metal hydride [40]. High overpotential is always found in CO2R [41]. Despite hydrogen evolution always being a competing reaction in CO2R, it is still worthy of scientific investigation on the hydrogenation of CO₂ to FA and dehydrogenation of FA as a practical hydrogen storage pathway [5]. FA formation essentially increases the density of hydrogen gas [6]. This opens a practical alternative, such as the direct formic acid fuel cells (DFAFC) [42]. Major heterogeneous metal catalysts, such as In, Sn, Hg, and Pb exhibit high FA selectivity [5,6,41,42]. Electrochemical CO2R on polycrystalline Sn surfaces exhibits high FA selectivity too. Formation of *OCHO at Sn surfaces is the key intermediate for FA production due to optimal *OCHO binding energy. The results suggest that oxygen-bound intermediates are critical to understanding the mechanism of CO₂ reduction to HCOO– on metal surfaces [43].

Cu is known to have relatively low CO selectivity because the CO produced is further reduced to several multi-carbon oxygenates (i.e., ethanol, acetate, and n-propanol) [41]. Specifically, sulfur-modified copper catalysts (Cu–S) exhibit positive correlation between particle size and selectivity toward FA evolution [44]. Nanostructured porous dendritic Cu-based catalysts show stable and selective conversion of CO₂ into FA at high current density with low overpotential [45]. The relatively low CO selectivity on Cu surfaces results from consecutive CO electroreduction activity [46]. That is, in CO₂-saturated aqueous solutions, polycrystalline Cu catalysts produce a mixture of compounds. Indeed, H₂ evolution is dominated at low overpotential, CO and FA formation mainly occurs at high overpotential, while hydrocarbons, ethanol, acetate, and n-propanol formation happen at the most extreme overpotentials [47,48]. In a CO₂ free environment, CO is reduced to hydrocarbons and multi-carbon oxygenates over the Cu catalyst [49,50]. Interestingly, oxide-derived Cu (Cu catalysts prepared by reducing Cu₂O) shows much higher H₂ selectivity than polycrystalline Cu [51]. Similarly, aqueous electrochemical CO reduction to C2 products by face-to-face coordinated thiol-terminated metalloporphyrins on copper electrodes exhibits 83% Faradaic efficiency and 1.34 mA/cm² current density at −0.40 V vs. RHE. This is a significant improvement in both selectivity and activity by one order of magnitude over parent copper surfaces or copper functionalized with porphyrins in an edge-on orientation [52]. In a similar system, oxide-derived copper (OD-Cu) electrodes exhibit a high CO reduction performance by producing ethanol and acetate with >50 % Faradaic efficiency at −0.3 V vs. RHE [53].

In a short summary, the selectivity of abiotic CO2R in the aqueous phase is highly sensitive to the surface chemistry of Cu-family catalysts. CO2R occurring at the aprotic surface tends to yield CO as the major product. By contrast, CO2R happening at the protic surface is prone to produce hydrocarbons and multi-carbon oxygenates as the major products. In this case, the selectivity is strongly affected by the involved reaction pathway. In the former case, the reduction is achieved through the charge transfer, while in the latter case the hydride transfer is mainly responsible for the CO2R. The adsorption affinity between reduced intermediates and Cu-family catalysts is another critical factor regulating the CO2R selectivity. High adsorption affinity slows the desorption of the reduced intermediates, which enables their consecutive reduction. This explains that the hydrocarbons and multi-carbon oxygenates as the major products are found in this case. It is thus concluded that increasing CO2R selectivity could be achieved through the modification of surface hydrophobicity and adsorption affinity.

In the iron-based CO2R, introducing the extra elemental Fe plate profoundly decreases the overpotential of the microbial electrosynthesis system (MES) [54]. In this MES system, all produced CO₂ is reduced to formate at the cathode and vast hydrogen is produced during the digestion of waste activated sludge. This is attributed to the high selectivity toward formic acid evolution over CO and methane production to the reduction of H⁺ at the cathode due to the slow methanogensis in Fe-C MES [54]. In the case of photocatalytic CO2R, enhancing sunlight conversion efficiency is always accompanied with improving CO2R selectivity [55]. Similar to the strategy adopted in the dye-sensitized photoelectrochemical cells [56], increasing selectivity in FA formation is usually achieved through coordinating active metals with covalent organic frameworks. For example, the columnar orientation COFs (covalent organic frameworks) provides a high-efficient charge carrier transport through the ordered π-electronic pathway, which improves electron transfer from COF to metal moiety and thus increases the reactivity [55]. The results of density functional theory computations further reveal that COFs decorated with electron-donating substituents favor CO₂ reduction by decreasing the hydricity of the Rh–H bond. This results in a lower hydride transfer barrier toward formic acid production [57] because the selectivity toward CO or HCOOH production is dependent on the coordination environment of the metal ion being capable of cleaving the C–O bond in the metal–CO₂H intermediate [55]. Specifically, an electron-rich coordination environment breaks the C−O bond to form CO, whereas an electron-deficiency coordination environment tends to enhance the C−O bonding thereby enhancing FA formation [58]. That is, if the center metal in covalent-organic frameworks (COFs) is a strong π-donor, such as Co(II), it usually tends to promote CO evolution. By contrast, a weak π-donor, such as Zn(II), favors HCOOH production [55]. A catalyst exhibiting low adsorption energy for HCOO* (i.e., a large energy difference between the two adsorbed CO₂ reduction intermediates, namely HCOO* and COOH* and large H* adsorption energy) would have high FA selectivity [59]. Ajmal et al. (2019) have studied the selectivity of the CO₂ reduction reaction over bimetallic CuZn alloy catalysts and reported that the Faradaic efficiency and partial current density of FA on Cu_0.5Zn_0.5 (equal molar ratio of Cu and Zn) are enhanced by nearly 4 and 5 times, respectively, that of Cu foil [60]. The high selectivity of the CuZn bimetallic alloy catalyst is originated from the synergistic effect of Cu and Zn. In this case, the Zn (a weak π-donor) is likely to create a large energy difference for the adsorption of two CO₂ reduction intermediates, namely, HCOO* and COOH*. An et al. (2019) have studied CO₂ reduction over Sn/SnOx catalysts and reported a maximum FA Faradaic efficiency of 89% at −1.7 V (vs. Ag/AgCl) in a 0.1 M CO₂-saturated KHCO₃ solution [61]. The authors further concluded that Sn(IV) and Sn(II) species are mainly responsible for controlling the overpotential and suppressing H₂ evolution toward improved FA selectivity. Chatterjee et al. (2019) have reported that nanoporous Pd-based alloys (np-PdX, X = Co, Ni, Cu, and Ag) exhibit FA selectivity following the order: np-PdAg > np-PdCu > Pd/C > np-PdNi > np-PdCo [62]. They have concluded that the composition-dependent behavior was governed by CO adsorption strength associated with the presence of transition metal alloying components near the Pd-skin surface and a composition-dependent change in the near surface H-sorption capacity. Interestingly, the free-standing np-PdCo and np-PdNi catalysts are able to sustain a high formate partial current density (>20 mA-cm⁻²) with high CO poisoning tolerance while exhibiting insignificant loss of the active area [62]. This further highlights the importance of durability and resistance of catalysts against CO poisoning during CO2R.

The results of computational hydrogen electrode model simulation reveal a striking similarity in CO2R electrocatalytic activity for the Cu₃ vs. Cu₅ and Cu₄ vs. Cu₆ size-selected clusters [63]. The rate-limiting potential of Cu₄ and Cu₆ clusters in CO2R is the proton-electron (H⁺ + e⁻) transfer to CO* (species adsorbed on clusters) to form CHO*, which is also the rate-limiting step on Cu surfaces. On the other hand, with respect to Cu₃ and Cu₅ clusters, removing OH* from the cluster surface (OH* →S+OH)

Several valuable chemicals such as CO, methane, methanol, low olefins, and long-chain carbohydrates could be produced from the gaseous phase CO2R reaction. To achieve selective CO2R, catalysts that are effective in activating both H₂ and CO₂ and stabilizing surface intermediates are needed. To this end, most catalysts are comprised of metallic sites, which are active in splitting adsorbed H₂ (from H₂ to H*) and exhibit a high affinity toward CO₂ adsorption. Additional modification with alkali or noble metals can further change the surface acidity or aid in the formation of an extra alloy phase in the catalyst [73]. The metal/support interfacial sites are highly active in CO₂ hydrogenation due to electron perturbation of the metal and partial reduction of metal oxide via the H-spillover mechanism [74,75].

Wet impregnation and co-precipitation are the two most frequently used methods to synthesize heterogeneous catalysts for CO₂ hydrogenation [76,77]. Nanosized catalysts can be synthesized by flame-spray pyrolysis [78]. The catalysts made by pyrolyzing metal organic precursors can achieve a complex nanostructure with high surface metal loadings [79,80] and the carbon sites can promote the adsorption and activation of CO₂ [81]. Fixed bed reactor configuration is applied in the evaluation of catalytic performance, in which specific amounts of catalysts are distributed over inert particles, such as SiC or silica, to the minimize hot spot phenomenon. After pre-reduction of the catalysts in H₂ flow, the stream that contains certain ratios of H₂/CO₂/inert gas with a specific gas hourly space velocity (GHSV) is injected into the fixed bed reactor. The CO₂ conversion (X_CO2) and selectivity (S) toward CO and low hydrocarbons are measured from on-line gas chromatography, while the long chain hydrocarbons are collected in a cold trap for further quantification.

is regarded as a weak acid in this configuration. It is thus suggested that in addition to the characteristics of the CO2R catalyst, its dispersion over the support is another issue of concern for rationally engineering a CO2R catalyst with high selectivity and efficiency.

Methanol is a valuable CO₂ reduction product as it can be served as biofuel, building blocks in organic synthesis, and fuels for a methanol-based fuel cell. Table 3 shows CO₂ conversion to methanol at a pressure of 1–4 MPa over different metallic meal oxide catalysts. Catalyst Cu/Zn/Al₂O₃ plays an important role in the commercial methanol production from syngas. DFT studies reveal that the stabilization of the transition surface species is the key to achieving high selectivity in the syngas conversion. The formate pathway predominates the total process on the surface of Cu/ZnAl₂O₄ and Cu/Zn/Al₂O₃, in which the rate-limiting step is the formation of H₂COO* and H₂COOH* [94,95]. The CO production via RWGS is the major byproduct in methanol synthesis. Based on the H/D isotope substitution technique, it is known that methanol synthesis and RWGS occurrence take place at difference surface sites on Cu/ZnO/Al₂O₃, Cu/MgO, Cu/SiO₂, and Pd/SiO₂ [96]. Karelovic and Ruiz (2015) have suggested that ZnO loads with larger Cu particles tend to suppress the activity of RWGS as the specific methanol formation rate per surface Cu is independent of Cu particle size, while that of CO is enhanced by smaller Cu particles [97]. Ro et al. (2016) have studied the synergistic effect by dispersing Cu on ZrO₂ and reported that the rate constant of CO₂ conversion on Cu-ZrO₂ interfacial is eight times greater than that on plain Cu [98]. Phongamwong et al. (2017) have added colloidal silica on Cu/Zn/ZrO₂ as geometric spacers to enhance the stability and performance of the Cu-based catalysts in CO₂ hydrogenation [99]. The authors have reported that 1% of SiO₂ loading increases the methanol synthesis activity by 26% and retains 12% more activity after emerging from the steam for 96 h at 280 °C.

Bimetallic Pd, such as PdZn and PdCu, exhibits a high methanol yield similar to Cu-based catalysts even at low temperatures [73]. The Pd dispersed in ZnO shows exceptional stability attributed to the particle size being maintained at 5 nm even after pre-reduction at 400 °C. While the colloidal dispersion technique is effective in stabilizing the interfacial sites of PdZn, its methanol formation yield is 40-fold greater than the same catalysts synthesized by the traditional wet impregnation method [101]. PdCu and PdCu₃ bimetallic catalysts exhibit a methanol formation rate of 0.31 μmol gcat⁻¹ s⁻¹, which is 3.4-fold and 6.2-fold greater than monometallic Pd and Cu, respectively [100]. Results of DFT simulation further indicates that the (111) plane of PdCu is highly active in methanol evolution, particularly at the low-coordinated Pd on the stepped surface [102]. Furthermore, methanol evolution through the formate pathway again is the major catalytic reaction, which can be further promoted by adsorbing small amounts of water to lower the energy barrier through the H-shuttled mechanism [102]. A novel metal-free In₂O₃ catalyst is also known to exhibit high selectivity in reducing CO₂ to methanol [76]. The oxygen vacancies on In₂O₃ are important to CO₂ hydrogenation. Specifically, the In₂O₃ with optimal oxygen vacancy exhibits a methanol yield of 3.7 mol kg-cat⁻¹ h⁻¹, CO₂ conversion of 7.1%, and methanol selectivity of 40%, respectively, at 330 °C and 4 MPa [75]. Martin et al. (2016) have reported that the selectivity toward methanol over In₂O₃ approaches 100% in the temperature range of 200–300 °C [104]. The methanol yield is further promoted by increasing the density of surface oxygen vacancy through Ar sputtering, syngas pretreatment, and ZrO₂ support [104]. Likewise, incorporating a small fraction of Mn into the spinel Co₃O₄ structure greatly enhances the methanol selectivity in CO₂ hydrogenation, likely attributed to the increase in surface basicity [103]. In brief, the significance of the thermal stability of the support in the CO2R to methanol is relatively less profound than that in the CO2R to CO and methane. This is because the former is usually carried out in a relatively lower temperature than that of the latter. In this case, the surface basicity becomes much more significant in affecting the reactivity of CO2R catalysts.

CO₂ conversion to lower olefins (C₂–C₄), building blocks, and other long chain hydrocarbons in the gasoline range (C₅–C₁₁) or diesel range (C₁₂–C₂₁) has been explored extensively. Again, catalysts that are reactive in both CO₂ hydrogenation and the Fisher–Tropsch reaction will be capable of achieving the CO₂ conversion objectives [105]. Table 4 lists the performance of CO₂ hydrogenation to low and long chain hydrocarbons on Fe-based catalysts. Upon CO₂ hydrogenation to different hydrocarbons, the Fe-based catalysts undergo consecutive phase transitions together with the creation of multivalent charges and this surface reconstruction process further diversifies distinct active sites on the catalyst [84]. The creation of multivalent charges results in the spinel iron (Fe₃O₄) being composed of Fe₃O₄, iron carbides, and α-iron, which enhance the activity of RWGS, carbon chain growth, and olefins’ secondary reactions [106]. It is therefore suggested that an appropriate fraction of Fe₃O₄ and Fe₅C₂ is necessary for the production of high olefin and paraffin [107,108]. Pretreatment of Co-Fe bimetallic under various reducing gases (i.e, H₂, syngas, and CO) is another effective ex-situ modification strategy to precisely control the phase transition [77]. Under the selected reducing environment, CO activation leads to the formation of CoFe alloy and carburized phases (Co₂C and FeC_x) that shift the selectivity toward low hydrocarbons and oxygenate [77]. While most Fe-based catalysts produce lower hydrocarbons, delafossite (CuFeO₂) exhibits high selectivity toward long chain hydrocarbons (C₅⁺) and 85 wt% of the produced long chain hydrocarbons are in the gasoline and diesel range [109]. Doping alkali metals on Fe-based catalysts is also beneficial to chain growth propagation due to stronger CO_x adsorption [106]. Compared to Na-free Fe₃O₄, incorporating 1.18 wt% of Na greatly enhances the CO₂ conversion and selectivity toward light olefin from 29.3 to 40.5% and 0.1 to 40.3%, respectively [110]. The presence of potassium promoter on Fe-Co/Al₂O₃ diminishes the density of hydrogen on metal surfaces, which in turn suppresses the hydrogenation of olefins [111].

Bifunctional catalysts have emerged actively in synthesizing hydrocarbons from CO₂ with great flexibility. Combined In₂O₃ and zeolite (HZSM-5) effectively synthesizes liquid fuels (C₅₊) from CO₂ by suppressing undesired C₁ products [112]. In the process, CO₂ is first hydrogenated to methanol at the reduced site on In₂O₃. Methanol that diffuses to the acidic site of zeolite transforms to hydrocarbons by the hydrocarbon-pool mechanism, which results in 78.6% of C₅₊ hydrocarbons [112]. Replacing the zeolite HZSM-5 by SAPO-34 increases the selectivity toward lower olefins, probably as a result of changes in topology [74]. Notably, the bifunctional catalyst must be packed in granular instead of powder form as the surface acidity of the zeolite support is another important factor controlling the selectivity of CO₂ hydrogenation toward hydrocarbons [112]. HZSM-5/Na-Fe₃O₄ composite also effectively converts CO₂ to hydrocarbons in the gasoline range [110]. The above catalyst composites provide three distinct active sites: Fe₃O₄ sites for RWGS, Fe₅C₂ sites for FTS, and zeolites for oligomerization [105]. Na-Fe₃O₄/HZSM shows little deactivation; the CO₂ conversion and C₅₊ selectivity remain at 27% and 54%, respectively, over 1000 h of operation [105]. Based on the above discussions, it is noted that the surface chemistry of the support is of equal importance to catalysts in CO2R. Interestingly, surface acidity instead of surface basicity is much more important in this case as the development of long chain chemicals strongly relies on the hydride transfer.

In short, there are advances in the development of heterogeneous catalysts for CO₂ hydrogenation in recent years due to a better understanding of the elementary reactions via DFT simulation, catalytic performance of metal-support, and metal-dopant interfacial sites. With appropriate modifications, CO₂ can be hydrogenated to CO, CH₄, CH₃OH, and other low and long chain hydrocarbons. Further challenges are reducing cost and increasing durability of catalysts, reducing cost of renewable hydrogen sources, and carbon dioxide capture.