Carbon Dioxide Conversion on Supported Metal Nanoparticles: A Brief Review

: The increasing concentration of anthropogenic CO 2 in the air is one of the main causes of global warming. The Paris Agreement at COP 21 aims to reach the global peak of greenhouse gas emissions in the second half of this century, with CO 2 conversion towards valuable added compounds being one of the main strategies, especially in the ﬁeld of heterogeneous catalysis. In the current search for new catalysts, the deposition of metallic nanoparticles (NPs) supported on metal oxides and metal carbide surfaces paves the way to new catalytic solutions. This review provides a comprehensive description and analysis of the relevant literature on the utilization of metal-supported NPs as catalysts for CO 2 conversion to useful chemicals and propose that the next catalysts generation can be led by single-metal-atom deposition, since in general, small metal particles enhance the catalytic activity. Among the range of potential indicators of catalytic activity and selectivity, the relevance of NPs’ size, the strong metal–support interactions, and the formation of vacancies on the support are exhaustively discussed from experimental and computational perspective.


Introduction
Carbon dioxide (CO 2 ) is the second most abundant greenhouse gas in the Earth's atmosphere due to the vast and excessive emissions from human activities related to the burning of fossil fuels [1] with concomitant environmental problems [2]. It is estimated that around 80% of the world's energy demand is supplied by fossil fuels. These anthropogenic emissions are considered to be the major contributors to climate change which may cause extreme events (high or low temperature, dryness, etc.) that are beyond human control. Some technologies have been introduced to reduce the CO 2 emissions [3], although they have been increased since the 1960s to a current level of more than 400 ppm in 2020 [4], as illustrated in Figure 1. Moreover, predictions indicate that CO 2 emissions will continue to increase until at least 2040 [5], with devastating consequences for the environment.
It has become urgent to mitigate the harmful effects of CO 2 emissions, CO 2 capture and storage (CCS), and especially its conversion towards valuable fuels and precursors. Many studies have been conducted with the aim of providing the effective capture and sequestration of CO 2 [6,7]. The deployment of CCS schemes is a multifaceted problem that requires the shared vision and efforts of governments, policy makers, and economists, as well as scientists, engineers, and venture capitalists [8,9]. The Intergovernmental Panel on Climate Change estimates that CO 2 emissions into the atmosphere could be reduced by 80-90% for a modern conventional power plant equipped with carbon capture and storage technology [10]. CO 2 adsorption is considered one of the most promising technologies for CCS [11], where natural zeolites, metal organic frameworks (MOFs) [12], ionic liquids [13,14], and Fe 3 O 4 -graphene are the conventional adsorbents given their capacity to adsorb CO 2 (222-2640 mg CO 2 /g sorbent) at a laboratory scale [15,16]. Nevertheless, it seems clear that efforts should be directed towards the potential use of CO 2 as an economical feedstock [17][18][19], because this means that two targets would be hit with one shot [20][21][22][23]. The number of emitted tons of CO 2 can be reduced while producing useful chemicals [24][25][26]. In addition, the net amount of CO 2 mitigated by conversion with renewable energy is 20-40 times greater than sequestration over a 20-year period [27,28]. To substantially reduce CO 2 emissions via catalytic conversion, only reactions that produce fuels or commodity chemicals can be considered viable and economically feasible solutions. The demand for fine chemicals is simply not great enough to effectively reduce emissions through a CO 2 conversion process [29]. Due to this, Figure 2 summarizes different ways of CO 2 conversion. CO 2 reduction has become an interesting option since the CO produced could be used as feedstock in Fischer-Tropsch synthesis [30]. This is an industrial catalytic process used to produce synthetic hydrocarbons to be used as fuels, which has been known since the 1940s. Furthermore, CO is also a key ingredient of syngas, a mixture of CO, CO 2 , and H 2 that is used in many industrial processes to produce tons of chemicals such as methane (CH 4 ) [31,32] or methanol (MeOH) [33]. In this sense, the CO 2 conversion to MeOH emerges as the most important method of CO 2 recovery, since it is the most direct route for CO 2 utilization. In addition, MeOH is a fuel for batteries and a precursor to many interesting chemicals [34], and it is very remarkable that the use of methanol as a transportation fuel presents economic advantages with respect to hydrogen-based fuel cells [35], which unfortunately have not been generalized yet in the automotive world because of the cost of Pt anodes [36]. Finally, the direct hydrogenation of CO 2 to alkane species (CO 2 -Fischer-Tropsch) is possible in a reactor, since it is thermodynamically easier than the reverse water gas shift reaction (WGSR) because the overall process is exothermic [37].
Many examples of the use of nanoparticles (NPs) supported on metal oxides to collect CO 2 are known. Different methods can be employed to synthesize NPs of different sizes and shapes, including top-down and bottom-up approaches [38]. As an example, on the use of NPs to convert CO 2 , the RWGS (CO 2 + H 2 → CO + H 2 O) reaction occurs on well-dispersed NPs supported on metal oxides to maximize the interface area between the particle and support [39]. Table 1 summarizes selected RWGS catalysts with conversion greater than 20%.   Regarding CO 2 methanation, several groups [72-75] investigated the use of small Ru particles supported on Al 2 O 3 . They concluded that Ru/Al 2 O 3 is an excellent catalyst, obtaining 96% of methane yields without CO, a slightly higher yield than the obtained with Ni/Al 2 O 3 [76]. It was revealed that the higher the number of Ni particles, the higher the CO production. The addition of vanadium improved the catalytic activity of Ni/Al 2 O 3 towards CO 2 methanation, slightly increasing the methane yield, and again, without detecting the CO generation [77]. In contrast, Italiano and coworkers detected the formation of a NiAl 2 O 4 spinel, which lowers the quantity of active Ni species, decreasing the catalytic

ZrO 2
Wambach and collaborators [82] exhaustively studied and compared the activity of several metal NPs supported on zirconia catalysts to investigate their performance in CO 2 hydrogenation. They reported that Cu and Ag promoted MeOH formation, while Ni, Ru, and Rh tended to catalyze CO 2 methanation. Note that these metals also promote the formation of methane using alumina as a support [77,78]. Less selectivity and low activity were found using Pd, Pt, and Au, in which the simultaneous formation of all the byproducts (MeOH, CO, and CH 4 ) was reported. Regarding Cu NPs, their deposition on ZrO 2 exhibits slightly higher activity for the conversion of CO 2 into MeOH, displaying comparable yields to the commercial catalysts using a very low amount of Cu (1-2 wt %) [ [93]. Finally, it was found that a mixture of Pt and Co NPs supported on ZrO 2 favors methane production [94].

SiO 2
Wang and coworkers prepared a long-lived Cu/SiO 2 catalyst synthetized using an ammonia evaporation method, showing excellent performance for CO 2 hydrogenation and large catalyst stability. CO, MeOH, and only a low amount of CH 4 were found at 260 • C, although the MeOH production drastically decreased at high temperatures [95]. Very recently, Shawabkeh and collaborators computationally studied the interaction of CO 2 with a Cu/SiO 2 catalyst, concluding that CO 2 is physiosorbed and bent on the oxide surface, owing to moderate interaction with one of the oxygen surface atoms [96]. A Cu/SiO 2 catalyst prepared via flame spray pyrolysis exhibited comparable catalytic performance to an active Cu/ZrO 2 catalyst for MeOH synthesis from CO 2 , although the conversion was only 5.2%. This experimental investigation showed that Cu NPs avoid CO dissociation and allow its hydrogenation to MeOH [97]. The important role of Cu particles was also suggested by Wang and collaborators [98]. Lam  using surface organometallic chemistry starting from a material consisting of isolated Zn 2+ surface sites dispersed on SiO 2 , and then generating CuZn alloys. This Cu-Zn/SiO 2 material displayed high catalytic activity, methanol selectivity, and higher conversion compared to the benchmark Cu/ZnO/Al 2 O 3 and most other catalysts [99]. Very recently, Fayisa and coworkers found that Pt could be an effective promoter to enhance the catalytic performance of Cu/SiO 2 for the hydrogenation of CO 2 to MeOH [100].
The combination of Cu and Ni particles on SiO 2 was investigated for CO 2 methanation, concluding that higher Ni content increases CO 2 conversion and CH 4 selectivity, and Cupromoted samples favor the CO selectivity [101]. Therefore, it seems clear that Ni NPs favor methanation, whereas Cu NPs enhance hydrogenation into MeOH. The catalytic activity of Ni supported on SiO 2 for CH 4 production was reported by several authors [102][103][104], showing the high conversion of carbon dioxide and 100% selectivity to CH 4 ( Figure 4). The work by Wu and collaborators revealed the importance of Ni coverage and size, since at low Ni loading (0.5 wt%), the system showed comparatively higher catalytic activity for CO 2 hydrogenation with large CO selectivity, whereas increasing the content to 10 wt%-9 nm particles-a selectivity switch was observed, favoring CH 4 production [105]. The promotion of Ni/SiO 2 with other metal particles such as Fe, Co, and Zn was studied by Dias and Perez-Lopez, showing that Fe and Co enhanced CO 2 conversion and selectivity towards CH 4 , while Zn did not favor CO 2 conversion and promoted the production of CO [106]. Huang et al. recently investigated the combination of Ni and Pt particles on SiO 2 , reporting that CH 4 was the main product only when the Ni/Pt molar ratio was higher than 9, owing to the fact that Pt NPs can directly dissociate CO 2 to CO [107]. The addition of Sn to Ni/SiO 2 slightly improved the activity and the catalytic stability at 650 • C, reducing the carbon formation, although a very small amount of Sn had a negative effect on the activity and stability of the Ni/SiO 2 catalyst [108]. Finally, Kattel and coworkers reported that Pt NPs are able to enhance the overall CO 2 conversion due to the CO 2 binding energy on SiO 2 with oxygen vacancies [109].

TiO 2
Titanium oxide is not an inert support. It exhibits great potential as an ideal and powerful photocatalyst for various significant reactions due to their chemical stability, nontoxicity, and high reactivity [110]. TiO 2 -supported metal catalysts have attracted interest due to the characteristics of TiO 2 NPs, showing high activity for various reduction and oxidation processes [111].
As reported using other oxide supports, the deposition of Cu on TiO 2 enhances the catalytic activity of these systems towards CO 2 hydrogenation to CO, MeOH, and CH 4 [112][113][114][115][116][117][118]. López-Caballero and collaborators used DFT simulations to prove that small Cu clusters catalyze the direct dissociation of CO 2 due to C=O bond activation and the subsequent reduction in the energy barrier for bond breaking [119]. The same tendency was observed in all these investigations. The deposition of Cu NPs on TiO 2 photocatalysts showed significantly higher photoactivity for CO 2 reduction, and it was more efficient than bare TiO 2 . Nevertheless, the use of big NPs of Cu does not favor CO 2 conversion. In order to achieve enhanced photoactivity, the concentration of Cu particles on TiO 2 must be low [120]. It is important to mention that Liu et al. reported that the deposition of Cu on TiO 2 nanotubes shows different results for the RWGS reaction with respect to the deposition on TiO 2 NPs. This different behavior was attributed to the formation of more O vacancies on the anatase (001) surfaces of nanotubes [121].
Xie and collaborators evaluated the effect of noble metal NPs on the photocatalytic activity of TiO 2 for CO 2 hydrogenation. The rate of CH 4 formation increased as follows: Ag < Rh < Au < Pd < Pt [122]. Jin et al. showed that group XI of transition metals enhances CO selectivity, while group X promotes the methane production [123]. Li and coworkers studied Pd/TiO 2 catalysts, showing the importance of Pd-TiO 2 interaction and the importance of controlling the size of Pd particles [124]. The performance with respect to CO 2 hydrogenation was better than bare TiO 2 . In addition, Pd particles not only can capture migrating electrons to separate electron−hole pairs but also promote the activation and bending of CO 2 . Regarding Ni NPs, Zhou and coworkers reported the excellent performance of Ni particles on TiO 2 for the methanation of CO 2 , highlighting the important role of Ni (111) exposed faces and CO as a key intermediate [125]. Li et al. found that Ni/TiO 2 NPs enhance the formation of CO 2 δ− species in the CO 2 methanation process, showing 76% conversion. Furthermore, they used DFT simulations to prove that electrons are transferred from Ti to Ni [126]. The study of Vrijburg and collaborators showed that the addition of Mn to Ni/TiO 2 catalysts leads to significantly enhanced CO 2 methanation activity [127].
Kohno and collaborators reported that Rh/TiO 2 produced CO and a very small amount of CH 4 , thus enhancing the RWGS reaction [128]. Zhou and coworkers concluded that metal-support interactions using anatase or rutile phases of TiO 2 can critically control metal-support interactions in Ru/TiO 2 systems and their catalytic performances for CO 2 hydrogenation. They reported that annealing Ru/rutile-TiO 2 in air can enhance CO 2 conversion to methane. However, the use of anatase decreases CO 2 conversion and promotes CO formation due to the strong metal-support interaction [129]. Zhang and collaborators demonstrated that the pretreatment temperature of Ir/TiO 2 catalysts can tune the selectivity of CO 2 hydrogenation, enhancing CH 4 production or even generating only CO. Encapsulated Ir NPs favor CO, while exposed Ir particles promote methane production [130]. The deposition of Pt has been extensively studied by several authors. Liu et al. found that Pt deposition on TiO 2 ultrathin nanosheets exhibited excellent photocatalytic efficiency for the conversion of CO 2 into CH 4 and CO. Strong metal-support interactions due to the formation of O vacancies improved the capability to bind CO 2 [131]. Zhang and coworkers suggested that the methanation process is the preferred reaction for Pt/TiO 2 catalysts, with the ability to activate CO 2 without H 2 assistance. They suggested that the methanation reaction proceeds via the activation of carbon dioxide, the subsequent CO 2 δ− hydrogenation into HCOO − moiety, its dissociation to CO, and finally, the hydrogenation to CHO, with CH 4 being the final product [132]. Kattel and coworkers also studied Pt deposition on TiO 2 , concluding that the effect of Pt NP is better for TiO 2 supports than for SiO 2 . Carbon monoxide is the main product in both Pt-supported oxide supports, although Pt/TiO 2 is slightly more selective to CH 4 than Pt/SiO 2 [109]. Qiu-Ye and coworkers also found that the deposition of Pt improved the efficiency of CO 2 conversion to CH 4 [133]. The recent work of Permporn and collaborators also confirmed the excellent performance of Pt. They loaded Pd, Pt, Cu, and Ni on TiO 2 , concluding that Pt/TiO 2 is the best system for CO 2 Catalysts 2023, 13, 305 9 of 28 activation and conversion, since Pt provides the largest work function when formed in heterojunction with TiO 2 [134]. In conclusion, Pt/TiO 2 does not catalyze MeOH production; otherwise, it favors CO and CH 4 generation.
With respect to bimetallic combinations, Neatu et al. revealed that the deposition of Au and Cu particles on a TiO 2 photocatalyst is an extremely efficient material for the solar-light reduction of CO 2 to CH 4 [135]. Reñones et al. showed that bimetallic Au and Ag NPs supported on TiO 2 are good catalysts for the photocatalytic conversion of CO 2 using water as the reducing agent, reporting that the supported bimetallic particles switch selectivity, enhancing methane production [136]. Singhal and Kumar reported that the deposition of two metal particles on TiO 2 , for instance, AgPd significantly improves methane formation [137].

ZnO
The commercial catalyst for CO 2 hydrogenation towards MeOH is composed of Cu together with ZnO NPs supported on alumina [138]. Therefore, Cu-ZnO interaction has been exhaustively investigated to elucidate the key interactions for this excellent performance. One of the most relevant works was carried out by Kattel and collaborators. Model systems Cu (111), ZnCu(111), ZnO/Cu(111), Cu/ZnO(000ī), and ZnO/Cu/ZnO(000ī) were experimentally synthesized for systematic comparison, including theoretical models of ZnCu and ZnO/Cu(111) [139]. This work reported that ZnO/Cu(111) presents the highest catalytic activity, as well as ZnCu (111), since the Zn is oxidized to ZnO at reaction conditions due to the direct dissociation of carbon dioxide and Zn oxidation. This system showed higher activity than the model of the commercial catalyst, Cu/ZnO(000ī), suggesting that the inverse system (ZnO NPs on Cu surfaces) is more effective than the traditional Cu NPs on ZnO supports. The morphology of a ZnO support was evaluated by Lei et al., concluding that filament-ZnO support is more active than ZnO nanorods after the deposition of Cu NPs [140]. Mahapatra et al. discussed the size of ZnO NPs, concluding that large ZnO NPs are not as chemically active as small ZnO NPs [141]. The work of Le Valant and collaborators revealed that CuZn is not an active site. The key is the presence of a ZnO x shell, in agreement with the work of Kattel [139]. The co-precipitated Cu-ZnO catalyst produces a large amount of CO according to Le Valant et al. [142]. Marcos and collaborators reported that Cu/ZnO/ZrO 2 is a promising system for MeOH production [143]. It is likely that CO 2 is activated by the generated oxygen vacancies of ZnO, and the Cu phase at the interface assists the molecular rearrangement. The use of a ZnO plate shows higher MeOH selectivity than a ZnO rod, although the latter shows slightly high catalytic activity [144]. Phongprueksathat et al. presented a new method to synthetize a Cu/ZnO catalyst in the absence of alumina, showing excellent catalytic activity [145]. Very recently, Guzmán and coworkers reported that Cu NPs and ZnO exhibited a synergistic effect in hydrogenating CO 2 with respect to pure Cu-based catalysts, in which MeOH and CO were the only products, displaying that the selectivity to MeOH changes from 100% at 200 • C to 23% at 300 • C [146].
Going beyond Cu NPs, in 1993, Sakurai and coworkers reported that Au/ZnO showed the highest selectivity to date and yields of CO 2 hydrogenation to MeOH in comparison with other metal oxide supports [147]. Chen et al. studied Au deposition on ZnO, concluding that increasing Au loading is directly related with an increase in the Au particle size, leading to a decrease in catalyst activity but enhanced selectivity to MeOH [148]. Hartadi and coworkers reported that the activity of the Au/ZnO catalyst for CO 2 hydrogenation is significantly higher than that for CO hydrogenation, and they concluded that both reaction pathways can be produced simultaneously [149]. Later, Abdel-Mageed and collaborators found a rapid formation of O-vacancies in the ZnO surface region in the initial stages of the reaction with the subsequent formation of negatively charged Au sites. This fact implies an increase in CO adsorption and MeOH production [150]. To complement these conclusions, Behm and coworkers demonstrated that increasing the ZnO particle size and maintaining the Au NP size and loading at constant values, the activity and selectivity increased [151].
The variation in MeOH production can be rationalized by optimizing the concentration of oxygen vacancies. On the other hand, the use of ZnO particles enhances the Ni/CeO 2 catalyst; it was reported that the addition of ZnO into a Ni/CeO 2 catalyst markedly alters the catalyst's properties and hampers the methanation of CO 2 , albeit it favors the production of CO via the RWGS reaction. Liao and coworkers recently investigated Ni and Ru deposition on different ZnO support morphologies, concluding that ZnO nanoplates show strong metal-support interactions since Ru NPs exhibits slightly lower interaction with the support and enhance CH 4 selectivity [152]. The deposition of Ru on ZnO enhances CO 2 methanation with respect to the bare support, where RWGS is the preferred reaction mechanism according to Dreyer et al. [153]. Nevertheless, the Ru/CeO 2 system presents great activity. Regarding Pd NPs, the use of light irradiation on Pd/ZnO catalyst showed a higher MeOH yield [154].
In summary, we have clearly highlighted the important role of ZnO for MeOH production. ZnO-supported NPs work remarkably better than ZnO surfaces. The main evidence of this is the structure and performance of the commercial catalyst for CO 2 hydrogenation to MeOH, where Cu and ZnO NPs are deposited on alumina, which works as an innocent support.

CeO 2
Ceria has been used as a catalyst support for a long time due to its unique structural properties resulting from the stability of Ce 4+ and Ce 3+ species, which promotes the formation of oxygen vacancies [155]. Therefore, ceria does not act as an innocent support; otherwise, these O vacancies are the key in most reaction processes [156].
One of the pioneering works in the use of the Cu/CeO 2 catalyst was carried out by Graciani and coworkers, combining experiments and DFT simulations. The computed energy barriers for the HCO→H 2 CO, H 2 CO→H 3 CO, and H 3 CO→H 3 COH hydrogenation steps were 5.3, 3.5, and 5.0 kcal/mol, respectively, which were very easy to overcome at 450 K. In contrast to the commercial Cu/ZnO/Al 2 O 3 catalyst, the reaction pathway using Cu/CeO 2 was presented in general exothermic steps [157]. In the work of Wang et al., Cu NPs were introduced to facilitate the O vacancy concentration on the CeO 2 support due to the formation of Ce 3+ species, promoting the photocatalytic activity of the Cu/CeO 2 system, ∼26 times larger than the bare CeO 2 support [158]. Lin and coworkers investigated the Cu deposition on ceria nanorods, revealing that the (110) termination of ceria showed better performance because it enhanced the oxygen vacancy formation. This surface termination also promoted more effective CO 2 activation and the formation of bidentate carbonate and formate moieties, characterized by DFT simulations [159]. Figuereido et al. reported that a higher concentration of Cu NPs on the surface under reaction conditions was more reactive [160]. The addition of a second metal NP to Cu/CeO 2 did not show the same behavior; the work of Yan and collaborators evaluated the addition of W to Cu/CeO 2 for CO 2 hydrogenation. W increased and stabilized the concentration of Ce 3+ , enhancing the methanol production and its selectivity [161]. In contrast, the addition of In into Cu/CeO 2 decreased the catalyst activity dispersion of Cu NPs and the formation of oxygen vacancies on CeO 2 due to the obstruction of In [162]. The work of Yang et al. revealed that the bimetallic Cu-Fe NPs supported on mixed CeO 2 -Al 2 O 3 showed the higher conversion of CO 2 than the monometallic NPs, boosting H 2 dissociation and CO 2 activation simultaneously [163].
Wang and collaborators published a series of papers investigating the Au-CeO 2 system. In 2013, they found that CO 2 interacts with pre-reduced Au/CeO 2 , reporting the presence of oxygen, thus suggesting the direct dissociation of the RWGS reaction at temperatures above 200 • C [164]. In another study, it was concluded that the RWGS reaction was dominated by the hydrogenation of carbon dioxide at lower temperatures (120 • C) [165]. The deposition of Au exhibits high activity, 10 times higher towards the RWGS reaction under light irradiation compared to the photothermal reaction, exhibiting a selectivity of around 100% to CO according to Lu et al. [166]. Rezvani et al. studied CO 2 hydrogenation to MeOH on Au/CeO 2 , showing slightly lower activity than a Au/ZnO catalyst under similar conditions. The key of the Au/CeO 2 system is the support reducibility, i.e., the O vacancy generation and the subsequent Au NP dispersion [167]. On the other hand, although this paper reviews the most relevant binary systems based on metal NPs supported in one support, it is important to mention the role of ceria in ternary systems, as in the case of Au/CeO 2 /TiO 2 . The work of Yang and collaborators reported that the addition of 0.1 monolayers of ceria to TiO 2 support stabilized the formation of small Au nanoparticles, promoting the catalyst activity in both CO and MeOH production and improving their selectivity towards MeOH with respect to Cu NPs, as illustrated in Figure 5  The deposition of Ni particles on ceria has been extensively studied to carry out CO 2 methanation. Several experimental studies report that ceria is the best support for Ni nanoparticles since this system exhibits higher catalytic activity than other oxide supports [169][170][171]. Due to the oxygen vacancy generation and the dispersion of Ni particles, Ni/CeO 2 was reported as the best system for CO 2 methanation, with almost 100% selectivity to methane. Rui et al. synthetized Ni/CeO 2 catalysts via the decomposition of a nickel precursor using gas discharge plasma, reporting very high activity due to the rich interfacial Ni sites, showing 99% selectivity to methane [172]. Bian and coworkers carried out a kinetic study, again revealing that the oxygen vacancies were thought to be active sites for the formate route [173]. The work of Jomjaree and collaborators studied the effect of ceria morphology, showing that Ni supported on ceria nanopolyhedrons displayed the highest activity for CO 2 methanation. They expected ceria nanorods to exhibit the highest catalytic activity due to their high surface area, large amounts of oxygen vacancies, and very strong interaction with Ni. Nevertheless, this strong interaction has a negative impact on CO 2 conversion at low temperatures [174]. Vavroutis et al. reported that, surprisingly, larger Ni particles (20 nm) exhibit the highest catalytic activity supported on ceria nanorods compared to Ni particles from 10 to 25 nm, obtaining CH 4 selectivity of 99% and a yield of 92%, the highest ever reported below 300 • C [175]. The study of Lin and collaborators also evaluated the importance of Ni particles on ceria surfaces. The Ni particle of 8 nm exhibited superior methanation selectivity over the 4 and 2 nm NPs. In addition, the methanation activity in terms of TOF was 10 times and 70 times higher than the 4 and 2 nm NPs, Catalysts 2023, 13, 305 12 of 28 respectively. The 8 nm Ni NPs supported on CeO 2 surfaces were revealed to enhance the formate hydrogenation [176]. Winter et al. studied the addition of Fe in a 3:1 molar ratio of Ni to Fe, reporting similar activity to the Ni catalyst with slightly improved CO selectivity. The catalysts that contained the same amount of Ni and Fe improved the selectivity to CO, but the catalytic activity was lower [177]. Sun and collaborators [178] investigated yttrium doping to Ni/CeO 2 . CO 2 methanation showed that the 2 wt% Y-doping of the Ni/CeO 2 catalyst exhibited the best CO 2 conversion and high selectivity towards CH 4 , since this amount of Y enhances the formation of oxygen vacancies in ceria. Doping with La has been recently investigated by Alvarez-Galvan and coworkers, reporting that the optimum value of La concentration is 10%, which corresponds to a maximum oxygen vacancy concentration and the highest CO yield [179]. Moreover, again, they reported the importance of O vacancy formation on ceria support. The system reached an average conversion of 52% and 100% CO selectivity. On the other hand, it has been observed that Ni NPs supported on mixed CeO 2 and Al 2 O 3 are a very stable and active catalyst with high CO 2 conversion for the RWGS reaction [180].
Regarding other transition metals, Xie and collaborators [181] studied the role of Co nanoparticles on different ceria morphologies, finding that ceria nanorods show great catalytic activity, as it was found for Ni particles [174]. This work shows that Co 0 and oxygen vacancies improve the catalytic performance, with 91% selectivity in CH 4 . Nguyen et al. suggested that CO 2 methanation proceeds by means of the CO 2 associative pathway, in which carbonate, bicarbonate, and formate intermediates are detected. However, direct dissociation was also observed on Co NPs [182]. López-Rodríguez and coworkers reported that 2.5 wt. Ru% is the optimal loading for Ru/CeO 2 , since it exhibits good CO 2 adsorption and dissociation capacity with the efficient hydrogenation of intermediates on the ceria surface [183]. López-Rodríguez et al. completed this investigation by studying the role of Ru and ceria for CO 2 methanation including DFT simulations [184]. They revealed that both metal and supports can activate and dissociate H 2 almost without an energy barrier, leading to high H and OH coverages on the catalyst. This fact is suggested to be positive for ceria but negative for Ru. Different Ru sizes were tested by Guo et al., who reported that Ru NPs show the best performance for CO 2 methanation [185]. The combination of Ru with Fe NPs promoted the RWGS reaction and the formation of CO. Again, the importance of oxygen vacancies on ceria was highlighted [186]. Iron NPs were found to be active for carbon dioxide conversion on mixed CeO 2 -Al 2 O 3 supports with high selectivity for RWGS at low temperatures [187]. Moreover, the addition of Ni and Cu particles enhances the catalytic activity and the catalyst robustness. As it was found for all the supports reviewed in this work, Ni doping promotes CO 2 methanation and Cu enhances the formation of CO. The use of Pd dimers supported on CeO 2 displayed a selectivity of 99% towards ethanol [188]. Jiang et al. revealed that the presence of Pd NPs promotes the formation of oxygen vacancies on the ceria support. Moreover, they performed DFT calculations showing that MeOH formation is likely from the formate (HCOO*) pathway via C−O bond cleavage in H 2 COOH*, with the reduction of HCOO* to HCOOH* as the rate-limiting step [189]. The deposition of small Pt particles on ceria provides evidence of strong metal−support interactions in Pt/CeO 2 (111) and Pt/CeO 2 powders, leading to systems that bind CO 2 well at room temperature [190]. Nevertheless, in 2004, Goguet and coworkers displayed that large amounts of CO on Pt/CeO 2 lead to carbon deposition with subsequent deactivation [191].
Finally, it is important to highlight that for single-metal-atom catalysts, the minimum size of metal NPs has been employed for CO 2 conversion using ceria support. Zheng and coworkers reported that doping with Ti NPs enhances the catalytic activity of a single Rh atom supported on CeO 2 for ethanol production [192].

Transition Metal Carbides
Transition metal carbides (TMCs) are formed when carbon atoms, produced, for example, by the decomposition of hydrocarbon molecules, are incorporated into metal Catalysts 2023, 13, 305 13 of 28 interstitial sites. TMCs have unique physical and chemical properties which combine the characteristic features of three different classes of materials: they show the extreme hardness and brittleness of covalent solids, high melting temperatures and simple crystal structures, typical in ionic crystals, and electronic and magnetic properties similar to transition metals [193,194]. TMCs have become a family of materials with an increasing role in heterogeneous catalysis in recent decades because of their chemical properties and low cost. The first and key landmark in this topic was the work of Levy and Boudart, who suggested that tungsten carbides displayed Pt-like behavior in several catalytic reactions such as hydrogenation, dehydrogenation, isomerization, and desulfurization reactions [195]. Since the pioneering work of Levy and Boudart in the 1970s, the number of reactions catalyzed by TMCs has greatly increased [196,197]. Apart from their catalytic activity per se, one of the most promising properties of some TMCs is their capability to modify the electronic structure of small supported metal particles and dramatically increase their catalytic activity through strong metal↔support interactions [198].

Metal Nanoparticles on TiC(001)
The first attempt to deposit metal NPs on TMCs involved the deposition of Au on TiC, carried out by Roldan-Cuenya and coworkers [199,200]. Theoretical calculations evidenced strong Au↔TiC interactions, in which TiC activated the deposited Au particles [198]. This fact promoted the deposition of different metal particles such as Cu or Ni. It was experimentally and computationally tested that small NPs show better results towards CO 2 hydrogenation than big NPs [201,202], and usually, these systems have been modeled with square and planar clusters containing four metal atoms. Focusing on CO 2 conversion, it is important to mention that CO 2 does not show strong interaction with metals, although the deposition of small metal particles on TiC exhibits the excellent conversion of CO 2 to CO and MeOH. Vidal and coworkers demonstrated that TiC produces a slightly lower amount of MeOH and CO as products of CO 2 hydrogenation with respect the Cu/ZnO(0001) system, a model of the commercial catalysts, but larger than the Cu(111) surface. However, Cu/ZnO(0001) cannot reach the catalytic performance of Au/TiC and Cu/TiC systems, with the MeOH and CO production being 5-12 times higher than the model of the commercial catalyst, as illustrated in Figure 6a [203]. CO 2 activation plays a very important role in the catalyst's performance. DFT simulations showed a binding energy of −0.62 eV on a bare TiC(001) surface, where the molecule is bent and both C=O bonds are equally elongated (1.29 Å). The interaction became stronger when the CO 2 molecule was adsorbed on top of the metal-supported particle, Au 4 /TiC(001) and Cu 4 /TiC(001), being −0.68 and −1.12 eV, respectively. According to these studies, the deposition of small particles of Au and Cu metal particles on TiC(001) increased the catalytic activity of CO 2 hydrogenation compared to a bare Cu(111) surface and Cu/ZnO(0001). Thus, TiC enhanced the reactivity of the supported metals by 1-2 orders of magnitude. The major product was CO, which was produced via the RWGS reaction. In addition, a substantial amount of methanol was also produced, especially using Cu NPs. Nevertheless, the deposition of Ni NPs changed the selectivity of the reaction. The main product for CO 2 hydrogenation using the Ni/TiC system was CH 4 , which was not observed using Cu-or Au-supported particles [204]. The amount of produced CO was even higher with respect to Cu and Au supported on TiC. A recent computational study revealed that CO 2 interaction is stronger in Ni/TiC(001) systems in comparison to bare Ni (111) and TiC(001), with lower dissociation energy barriers. According to experimental studies, small and 2D Ni particles deposited on TiC(001) exhibited larger binding energies and lower dissociation barriers for CO 2 compared to 3D particles [205,206].  [203,204,207,208]. Copyright ACS, Elsevier B. V., and Royal Chemistry Society.

Metal Nanoparticles on Molybdenum Carbides
Despite the great performance of titanium carbide material, it is a cumbersome support due to the difficulty of obtaining catalytically active supported NPs on working conditions. In this sense, molybdenum carbides, in particular cubic (δ) MoC and orthorhombic (β) Mo 2 C, have emerged as excellent alternatives because they are more active and do not require special conditions for their synthesis. Experimental and theoretical investigations were performed to explore the suitability of these Mo carbide phases. The performance of β-Mo 2 C for CO 2 conversion was impressive, since it is able to activate and dissociate CO 2 and even CO without the hydrogen assistance [209,210]. Owing to the fact that the β-Mo 2 C(001) single-crystal surface contains two possible terminations, C or Mo, DFT calculations were essential to elucidate the contribution of each different termination to CO 2 hydrogenation, revealing that a Mo-terminated surface is very active and responsible for large amounts of catalytic activity, while a C-terminated surface is able to activate carbon dioxide [207]. Figure 7 represents the CO 2 adsorption modes on C-and Mo-terminated surfaces, clearly showing that C-terminated surfaces enlarge one of the C=O bonds. Theoretical calculations showed that the energy barrier for the first C=O scission was only 0.21 eV on the Mo-terminated surface and 0.50 eV for the C-terminated surface, in which the second one was more energy-demanding: 0.86 eV for Mo-terminated and higher than 1.5 eV for Cterminated surfaces [207]. The major products for CO 2 hydrogenation were CH 4 and CO, while the amount of methanol was very low according to the experiments (see Figure 6b). The deposition of Cu NPs on β-Mo 2 C drastically modified the catalysts' selectivity. The CO 2 conversion increased under the presence of H 2 , increasing the production of MeOH as well with respect to bare surfaces ( Figure 6c) and decreasing the amount of produced CH 4 . Posada-Pérez and coworkers reported that the larger the Cu NP, the lower the CH 4 production [207]. Nevertheless, big Cu particles and Cu monolayers clearly decreased the catalytic activity of β-Mo 2 C. Theoretical models of Cu-supported clusters were reported by Illas and coworkers [211], showing that Cu n clusters in contact with β-Mo 2 C(001) adopted a planar configuration, independently of the surface termination. The Cu 4 NPs were anchored to a Mo-terminated surface to model the Cu/β-Mo 2 C system. In detail, Posada-Pérez et al. determined that Cu NPs on Cu/β-Mo 2 C catalysts have a dual function [207]. On the one hand, Cu NPs occupy active surface sites for direct CO 2 and CO dissociation, blocking the fast conversion to CH 4 . On the other hand, DFT simulations conclude that Cu NPs open a new pathway to generate MeOH. Cu-supported particles cannot directly dissociate CO 2, but they favor its hydrogenation to COOH and HCOO intermediates, generating a new route to produce MeOH at the Cu-carbide interface. The experimental work of Heracleous and coworkers proved that the addition of 20 wt% Cu slightly decreases the catalysts' reactivity, although it enhances the MeOH selectivity with respect to methane [212]. Furthermore, Zhang and coworkers computationally proved that Cu-supported particles facilitate CO hydrogenation to an HCO moiety compared to bare β-Mo 2 C [213], in agreement with the studies of Posada-Pérez [207,208]. The study of Jing and coworkers [214] also demonstrated that the addition of Cu exhibits better catalytic activity for the hydrogenation of CO 2 , and Zhang et al. demonstrated that Cs and Cu supported on β-Mo 2 C have a positive impact on the selectivity and activity [215]. As it was found with metal NPs supported on TiC, Cu/β-Mo 2 C produced larger amounts of MeOH than the model of the commercial catalysts. However, the commercialization of Cu/β-Mo 2 C is not viable due to its large deactivation. The large reactivity of the carbide support, allowing the CO 2 and CO direct dissociations, promotes O deposition, which decreases the catalytic activity of the system [216,217]. On the other hand, the use of supported Ni particles on Mo 2 C was explored for CO 2 methanation. In this particular case, the Ni/Mo 2 C system was supported on Al 2 O 3 . Again, the large influence of metal↔carbide interaction on the catalytic activity was demonstrated, since the conversions of CO 2 for the reduced Ni-Mo/Al 2 O 3 catalyst and Ni-Mo 2 C/Al 2 O 3 catalysts were 5.3% and 13.8%, respectively, with a corresponding selectivity in CH 4 of 10.0% and 98.1% [218]. Therefore, the Ni↔Mo 2 C interaction is more efficient for CO 2 methanation than Ni↔Mo interaction.
In order to improve the stability and selectivity of Mo carbides towards the hydrogenation of CO 2 , the reaction was investigated by decreasing the metal:carbon ratio on the carbide. The surface of MoC experimentally synthetized was best described as polycrystalline [219]. Owing to the complex phase diagram of MoC [220], a well-defined (001) termination of the δ-MoC (001) surface is difficult to synthetize. The ideal δ-MoC(001) surface was modeled computationally [209], exhibiting promising behavior: it can activate the CO 2 molecule as it was observed using β-Mo 2 C but with a moderate adsorption energy and a large dissociation energy barrier, which avoids the formation of atomic oxygen and self-poisoning [208,211,221]. As it was observed for TiC and Cu/TiC systems, δ-MoC and Cu/δ-MoC catalysts only produce CO and MeOH, while CH 4 is not detected (Figure 6d). This is because the direct dissociation towards CO is energetically demanding, avoiding the formation of C* + O*, which is directly related with CH 4 production and Mo-carbide deactivation [207,220]. Furthermore, δ-MoC shows slightly more major production of CO and MeOH than TiC(001). The deposition of small Cu NPs increases the CO and MeOH production in both β-Mo 2 C and δ-MoC(001) surfaces (Figure 6e) and increases the selectivity towards both products on β-Mo 2 C, reducing the methane generation. Therefore, the metal:carbon ratio on the carbide plays a relevant role in the function of Cu NPs: on β-Mo 2 C, Cu NPs block surface sites for O generation and CH 4 production and promote CO 2 hydrogenation [222]. In contrast, the deposition of Cu on δ-MoC(001) enhances direct CO 2 dissociation to CO on top of Cu clusters [208]. Figure 6f compares all the carbides and models of commercial catalysts for CO 2 hydrogenation to CO, highlighting the catalytic activity of Cu/δ-MoC.
On the other hand, other transition metals were deposited on molybdenum carbides, although these systems demonstrated high catalytic activity for the water gas shift reaction instead of the reverse water gas shift. For instance, Nagai and coworkers showed that Co supported on Mo 2 C is more active than Mo 2 C, partially avoiding the catalysts' deactivation [223]. Similar results were reported by Rodriguez and coworkers, using Pt supported on δ-MoC(001) [224], showing high activity, stability, and selectivity at low temperature, and better performance than the typical industrial Cu/ZnO catalyst.
In summary, molybdenum carbides in particular and TMCs in general exhibit excellent catalytic activity towards the hydrogenation of CO 2 . Nevertheless, the deposition of metal NPs becomes essential to increase the selectivity towards CO and MeOH, especially when the metal/carbon ratio on the carbide is 2. This is related with the large O-carbide interaction, which enhances CO 2 and CO dissociation. Thus, anchoring metals with lower oxygen binding energy can avoid these dissociation processes, suppressing methane generation [225]. This was observed for other reactions, such as MeOH decomposition studied by Kelly et al. using Ni-, Rh-, and Au-modified WC surfaces [226]. Ni-, Rh-, and Pt-modified WC surfaces preferably catalyze the C-H scission instead of C-O, disfavoring methane production.

Perspective: Single-Metal-Atom Catalysts
One of the general remarks that can be highlighted from this review is the fact that, in general, small metal NPs enhance the catalytic activity of the metal-support system, while the catalytic activity decreases, increasing the NP size. As it has been reported in this work, the support morphology and NP size can drastically modify the reaction selectivity. Focusing on the industrial applications of these heterogeneous catalysts, its facile separation from the products is a very strong point that helps to reduce the operating costs, although it is complicated to control the selectivity. In this sense, homogeneous catalysts avoid this problem since, in general, they are very selective, although the use of precious metals and catalysts' recovery practically discard their commercialization. Single-atom catalysts (SACs) hold great promise to bridge homogeneous and heterogeneous catalysts since, on the one hand, SACs maximize the metal utilization due to the low coordination of the single metal atom [227], and they have tunability of the catalytic site on the support [228][229][230][231][232][233]. On the other hand, they can reach the selectivity obtained with homogeneous catalysts. Furthermore, SACs contribute to lowering the cost, because the amount of metal decreases. Since the seminal work of Zhang and coworkers, who reported in 2011 that the Pt 1 /FeO x SAC was three times more active than its nano-Pt counterpart for CO oxidation [234], SACs have become a new frontier in heterogeneous catalysis.
For example, Au single atoms likely occupy surface ceria vacancy sites, thus achieving higher metal loading [235]. Ceria can stabilize transition metals and favor metal dispersion due to the high number of vacancies, becoming an excellent support for many reactions such as the WGS reaction, CO oxidation, CO 2 reduction, or CH 4 oxidation. The deposition of single metal atoms on a TiO 2 support was exhaustively investigated by Chen and coworkers, reporting that the direct dissociation of CO 2 to CO using a Cu single atom presents an energy barrier lower than 0.2 eV, i.e., the dissociation is almost spontaneous [236]. Using a Pt single atom, the energy barrier is higher (0.43 eV). Very recently, two-dimensional transition carbides (MXenes) have emerged as very active catalysts for several reactions, since the catalytic activity increases with respect to 3D materials [237]. The deposition of single metal atoms shows excellent catalytic activity; for instance, the Pt 1 /Ti 3−x C 2 T y catalyst, in the presence of a diverse range of amines and silanes, could readily fixate CO 2 and yield value-added amides with very high conversion and selectivity [238]. Bare MXenes are very reactive, with high metal:carbon ratios, which implies low stability. The deposition of single metal atoms may enhance the support stabilization and improve the catalytic activity, especially focusing on carbon dioxide conversion. These studies, together with the excellent results reviewed in this work about highly active small metal NPs supported on non-innocent metal oxide and metal carbide supports, encourage the investigation of the deposition of single-metal-atom catalysts for CO 2 conversion.
The use of highly dispersed SACs is a promising way to increase the catalytic activity of a support. Nevertheless, as it was reported with metal NPs, SACs are very sensitive to the structural and electronic properties of the support. It is commonly assumed that SACs can activate the adsorbed molecules working analogously to homogeneous catalysts, but one should bear in mind that the structure of an active support site and possible defects may play a major role [239]. NPs' deposition on or close to surface vacancies enhance the catalytic activity of the system, together with the capability of the support to anchor reaction intermediates, such as hydroxyl groups, which have been demonstrated to play a relevant role in CO 2 activation. In our opinion, this fact may be more decisive in the performance of SACs. Therefore, to further comprehend and control the catalytic activity and selectivity of SACs, it is mandatory to explore the local coordination environment of the metal on the support, avoiding possible SAC absorption on the material, with the subsequent suppression of the catalytic activity of SACs [240]. The use of SACs may bring new opportunities in CO 2 conversion in particular and green chemistry reactions in general, owing to the precise building of catalytically active sites in dimensions of the sub-nanometer scale.

Conclusions
In the present review, we discussed selected examples of metallic NPs supported on metal oxides and metal carbide surfaces for the conversion of CO 2 , covering the most relevant produced fuels and chemicals. A combined analysis of the state of the art of density functional simulations and sophisticated experiments was used to find suitable metal NP and support combinations for the efficient conversion of CO 2 . The discussion is focused on the evaluation of suitable catalysts to identify opportunities by exploiting these harmful greenhouse gases as economical feedstock, following the main principles of green chemistry.
In this work, we have reported the catalytic activity and selectivity of several metallic NPs supported on different innocent and a priori non-innocent supports, highlighting important descriptors that govern the activity, selectivity and stability of these catalysts, such as metal-support interactions, NP sizes, and support morphology. Some general conclusions can be extracted independently of the metal oxide and metal carbide structure and morphology. In general, Cu stands out among all metal NPs in its enhancement of the RWGS reaction and the subsequent hydrogenation of CO towards MeOH. The use of alumina as a unique support is preferred with respect the ternary catalysts. It was proved that alumina can host hydroxyl moieties, which play a relevant role in the facilitation of CO 2 activation and formate production. Furthermore, it is important to highlight the role of ZnO in MeOH production, where experiments and simulations proved that it is more active as a supported NP than as support. In addition, CeO 2 is also presented as an excellent support for Cu NPs to enhance the conversion of carbon dioxide. The facile transformation of Ce 4+ to Ce 3+ species promotes the formation of oxygen vacancies, an essential fact to become a suitable support for catalytic purposes. Regarding the use of transition metal carbides as supports, the formation of CH 4 using Cu supported in metal carbide particles was computationally and experimentally observed, which in general was not observed using metal oxides. Nevertheless, it was proved that the large catalytic activity of the support enhances the double bond cleavage of CO 2 molecules, favoring the formation of methane, i.e., Cu NPs do not participate in the methanation process. This is because the metal:carbide ratio plays a relevant role in the metal Cu activity. For Mo 2 C, with a metal:carbon ratio of 2, Cu NPs block active surface sites for direct CO 2 and CO dissociations, which lead to the formation of oxycarbide surfaces with the subsequent catalysts' deactivation. Cu NPs open a new catalytic route for CO 2 and CO hydrogenations since they bind weakly-moderately to Cu NPs. For carbides with metal:carbon ratios of 1, such as TiC and MoC, Cu NPs can enhance CO 2 hydrogenation and also promote its direct dissociation to CO. For these systems, the methane production was not observed.
Regarding other metal particles, the deposition of Au NPs follows the same trend as Cu, although its performance is lower than Cu NPs. This is not bad for the prospects of catalyst commercialization, since Cu is cheaper than Au. Some works revealed that Au is more effective for the WGS reaction. The use of Ni NPs clearly promotes CO 2 methanation, independently of metal oxide and metal carbide support. Other metal particles such as Pt, Pd, and Ru have been reported in this review to show excellent catalytic activity for CO 2 methanation, although the take-home message should be simplified, highlighting Cu NPs for CO and MeOH production and the use of Ni NPs for CO 2 methanation. No general conclusions can be extracted for the combination of bimetallic particles, since they do not always improve the catalytic performance of the system, either for MeOH production or CO 2 methanation. Each combination must be evaluated individually.
Another general conclusion that can be gathered from these selected works is the strong dependence of the catalytic activity of the system and the capability of metal oxide supports to generate oxygen vacancies. These defects have a double role: they favor NP dispersion and enhance CO 2 activation. On the other hand, it is well-known and assumed by the research community that strong metal-support interactions govern the catalytic activity of the system, since the stronger the interaction is, the higher the catalytic performance is. Nevertheless, in this review, we reported a few works that suggest that the support morphology with the highest interaction with the metal NPs does not always show the best activity (for instance, CeO 2 nanorods and Ni and Ru supported on Al 2 O 3 ).
The last important descriptor to analyze is the metal NP sizes, concluding that, in general, small metal particles enhance the catalytic activity of CO 2 conversion. Therefore, it seems clear that the efforts should be routed towards research regarding a single-metalatom catalyst, due to its large activity, selectivity, and the reduction in the amount of metal, which clearly benefits its possible commercialization. However, as it has been presented in this review, the number of variables that affect the performance of metal-supported particles (or single metal atoms) is large, such as the formation of oxygen vacancies, the morphology of the support, and the metal-support interactions, among others. Therefore, to investigate the suitability of single-atom catalysts, it is necessary to further explore the interaction of these single atoms with surface defects and different active sites, using computational and experimental techniques, with the goal to correlate the catalytic performance with the coordination site and the coordination environment around the support.
To summarize, the use of CO 2 as feedstock is a clear benefit that helps to reduce the impacts of global warming compared to those of current technologies only based on CO 2 capture. One of the keys to accelerate this reduction is to find low-cost, stable, selective, and highly active catalysts for the potential conversion of CO 2 to fuels and chemicals. Despite the fact that the descriptors that govern this process cannot be overgeneralized, some take-home messages can be derived from our revision. Cu and Ni NPs supported on metal oxides and metal carbides are promising catalysts for the future commercialization of more efficient catalysts, although the full puzzle that combines low kinetic barriers and the thermodynamic stability of catalysts at reaction conditions is not complete. In addition, the inclusion of single metal atoms as a new puzzle piece will promote the investigation of metal-support interactions. We hope that the most relevant descriptors analyzed in this work can help researchers to make progress in CO 2 utilization. [CrossRef] 33. Liu