Introduction

Since the industrial revolution, the CO2 concentration in the atmosphere has been rapidly increasing because of fossil fuel utilization and human activities, resulting in severe climate change due to the significant greenhouse effect of CO2. To address this issue, global carbon emissions need to be reduced, and China has set its goal of achieving carbon neutrality by 2060. Carbon capture, utilization, and storage (CCUS) is an important emission reduction technology that can effectively utilize highly concentrated CO2 emitted by industrial activities. Among the CCUS processes, CO2 utilization via catalytic conversion is promising as it provides a viable solution to close the carbon cycle [1,2,3,4,5,6].

Generally, the CO2 molecule is thermodynamically stable and kinetically inert; thus, CO2 activation is an important scientific challenge. CO2-conversion-related chemical reactions require remarkable energy input in the chemical, thermal, or electrical form. If the energy comes from fossil fuels, then the process would not achieve the net reduction of CO2. Therefore, CO2 conversion should be driven by nonfossil energy sources, such as nuclear energy and renewable energy, such as solar and wind. Catalysis is an efficient method to activate and convert inert molecules and plays a crucial role in CO2 conversion into valuable chemicals and fuels (e.g., CO, formic acid, methanol, methane, ethanol, acetic acid, propanol, light olefins, aromatics, and gasoline). The substantial amount of hydrogen needed in the production of chemicals and fuels from catalytic CO2 conversion (mainly CO2 reduction) should come from water indirectly in thermal catalysis or directly in electrocatalysis and photo(electro)catalysis [2].

To date, many research institutions in the world have devoted tremendous efforts to the field of catalytic CO2 conversion. As one of the leading research institutions for fundamental and applied catalysis in the world, the Dalian Institute of Chemical Physics (DICP) started its research on heterogeneous catalysis for CO2 conversion in the early 1990s in the field of thermal catalysis [7] and has expanded the research to electrocatalytic and photo(electro)catalytic CO2 conversion in the past decade [8, 9] (Fig. 1). Some important concepts (e.g., nanoconfined catalysis and single-atom catalysis) in heterogeneous catalysis proposed by our institute have also been applied to catalytic CO2 conversion [10, 11]. In this review, we summarize our recent progress in heterogeneous catalysis for CO2 conversion into chemicals and fuels in different conversion routes and highlight some representative studies of thermal catalysis, electrocatalysis, and photo(electro)catalysis.

Fig. 1
figure 1

Heterogeneous catalysis routes for CO2 conversion into chemicals and fuels

Thermocatalytic CO2 Conversion

CO2 conversion via thermal catalysis involves a large group of catalytic reactions, including CO2 hydrogenation, dry reforming of methane with CO2, and nonreductive CO2 conversion into fine chemicals. Of these routes, direct CO2 hydrogenation to valuable chemicals and fuels is most promising. As more than 90% of industrial hydrogen is currently produced from fossil fuels via steam reforming of natural gas and coal gasification, which in turn produce substantial CO2, CO2 hydrogenation should proceed using hydrogen produced in a manner free of CO2 emission. Obtaining hydrogen via water electrolysis powered by renewable energy sources provides a viable solution to this issue, thus making CO2 hydrogenation attractive.

CO2 Hydrogenation to C1 Products

Methanol, formic acid, methane, and CO are the four main C1 products synthesized from CO2 hydrogenation. In light of the reaction thermodynamics (Table 1), CO2 hydrogenation to most C1 products is exothermic, except for CO production via an endothermic reverse water gas shift (RWGS) reaction. Although the ∆G0298 K value of CO2 hydrogenation to formic acid is positive, in most studies adding a base to the reaction medium shifts the reaction equilibrium forward, thereby making the reaction easier and yielding formate as the final product. The generation of methanol, methane, and CO from CO2 hydrogenation is often concurrent under their required reaction conditions, thereby developing efficient catalysts with high selectivity toward the desired C1 products and considerably improving activity and stability.

Table 1 Reaction thermodynamics of CO2 hydrogenation to C1 products

Methanol

Methanol synthesis from CO2 and hydrogen is the first target based on the concept of “liquid sunshine” proposed by Bai and coworkers [12]. Methanol can be used as either liquid fuel or platform molecule for subsequent upgrading to higher-value chemicals via emerging technologies, such as the methanol to olefins process [13]. CO2 hydrogenation to methanol occurs at 200 °C to 300 °C and 3–10 MPa, under which conditions the RWGS and CO2 methanation reactions are also thermodynamically favorable. Cu/ZnO/Al2O3 is an industrial catalyst for traditional methanol synthesis from syngas (CO + H2). This catalyst is also applied to methanol synthesis from CO2 hydrogenation. However, Cu/ZnO/Al2O3 is also an active catalyst for the RWGS reaction, resulting in a low methanol selectivity. Moreover, CO2 hydrogenation produces more water, which causes severe sintering and deactivation of Cu catalysts [14]. The methanol selectivity is usually lower than 50% at a considerable CO2 conversion (i.e., > 10%) over traditional Cu/ZnO/Al2O3 catalysts [15].

The reactivity and stability of Cu-based catalysts can be improved by stabilizing active Cu species via metal–support interaction enhancement. Cu catalysts supported by alternative oxides (e.g., SiO2, ZrO2, CeO2, and Ce1−xZrxO2) have been developed [16,17,18,19]. Sun and coworkers [16] developed a Cu+-enriched Cu/SiO2 catalyst using the flame spray pyrolysis (FSP) method. This catalyst achieved a methanol selectivity of 79% at a CO2 conversion rate of 5.2%. A unique shattuckite-like structure with slightly distorted Cu–O–Si texture in the FSP-prepared catalyst was considered to enrich the Cu+ species. High-pressure in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) showed that the Cu+ species could stabilize and facilitate the hydrogenation of the *CO intermediate to methanol via an RWGS + CO hydrogenation pathway.

Apart from Cu-based catalysts, researchers also considered non-Cu catalytic materials, such as Pd nanoparticles (NPs) confined in carbon nanotubes, Pd/ZnO/Al2O3, Pt/In2O3, MoS2, ZnO–ZrO2 solid solution, and TiO2-supported Cd cluster [15, 20,21,22,23,24,25,26]. Li and coworkers [22] reported a ZnO–ZrO2 solid solution catalyst, which achieved a methanol selectivity of 86% to 91% and excellent stability for 500 h at a CO2 conversion of more than 10% under the investigated reaction conditions (Fig. 2a, b). The in situ spectroscopy and theoretical calculation results indicated that the synergetic effect between Zn and Zr sites in the ZnO–ZrO2 solid solution catalyst contributed to the outstanding methanol synthesis performance. A MoS2 nanosheet catalyst with rich sulfur vacancies developed by Deng and coworkers [25] was able to achieve CO2 hydrogenation to methanol at a low reaction temperature (180 °C) but with a methanol selectivity of 94.3% at a CO2 conversion rate of 12.5% and a stable performance during a long-term test for over 3 000 h.

Fig. 2
figure 2

Reproduced with permission from Ref. [22]. Copyright 2017, American Association for the Advancement of Science. c Catalytic performance and d stability of CO2 hydrogenation to formate over the Ir/AP-POP catalyst. Reproduced with permission from Ref. [28]. Copyright 2019, Cell Press. e CO2 methanation performance of Ni/a-TiO2 and Ni/a-TiO2-NH3. Reproduced with permission from Ref. [37]. Copyright 2019, American Chemical Society. f CO2 methanation performance of Ir/TiO2 reduced at different temperatures. Reproduced with permission from Ref. [40]. Copyright 2020, Royal Society of Chemistry

a Catalytic performance and b stability of CO2 hydrogenation to methanol over the ZnO–ZrO2 solid solution catalyst.

Formic Acid (Formate)

Formic acid usually presents as formate in practical production from CO2 hydrogenation, as thermodynamics can efficiently shift the reaction equilibrium in the forward direction (CO2 + H2 → HCOOH; HCOOH + OH → HCOO + H2O). CO2 hydrogenation to formate is generally achieved via bicarbonate hydrogenation in alkaline environments, but it suffers from high activation energy because of the considerable thermodynamic stability of the bicarbonate intermediate. Huang and coworkers [27] proposed a revised route for direct hydrogenation of CO2 to formate using a Schiff-base-modified Au catalyst, with a turnover number (TON) of 14 470 at 90 °C and 8 MPa. The in situ DRIFTS and density functional theory (DFT) calculation results indicated an unusual CO2 activation pathway involving a weakly bonded carbamate zwitterion intermediate formed through the Lewis base adduct of CO2. However, this reaction pathway only occurred with hydrogen lacking a Lewis base center in a polar solvent, such as water and methanol, used in this work. Huang and coworkers [28] further applied atomically dispersed catalysts to CO2 hydrogenation to formate. A single Ir atom catalyst was prepared using a porous organic polymer with aminopyridine functionalities. It exhibited excellent performance for formate production, with a TON as high as 6784 at 120 °C and 8 MPa (Fig. 2c). Recycling experiments indicated that the catalytic activity did not decrease even after four uses (Fig. 2d). The catalytic mechanism was similar to that of a homogeneous mononuclear Ir pincer complex catalyst.

Methane and CO

CO2 hydrogenation to methane (methanation) and CO (RWGS) are usually two parallel processes, depending on the catalytic materials and reaction conditions. The produced methane via CO2 methanation can serve as the principal component of synthetic natural gas, whereas the produced CO via RWGS can be integrated into syngas conversion systems [29]. Extensive studies on catalysis have been devoted to controlling selectivity toward methane versus CO at a considerable CO2 conversion for given applications [30].

Among the metals (e.g., Ni, Ru, Rh, Pd, Ir, Co, Fe, and Cu) used for CO2 methanation catalysts, Ni and Ru are the most active ones [30,31,32,33]. As non-noble metal catalysts, Ni-based catalysts hold great promise for practical methanation application. However, catalytically active Ni species are prone to sintering and deactivation because CO2 methanation is a strongly exothermic reaction. An important strategy for dealing with this issue is to modulate the strong metal–support interaction (SMSI) between Ni NPs and oxides [34,35,36,37,38,39]. Pan and coworkers [37] determined that CO2 methanation performance was remarkably suppressed and CO was the only product when Ni NPs were supported by conventional anatase (a-TiO2) because of the SMSI-induced formation of a reduced TiOx overlayer around the Ni NPs at high temperature (500 °C). By contrast, pretreating the a-TiO2 support with NH3 and H2 switched the selectivity from CO to methane with one order of magnitude increase in CO2 conversion (Fig. 2e). The reduction pretreatments generated a large amount of Ti3+ species in the a-TiO2 support, which could weaken the SMSI and suppress the formation of the TiOx overlayer, leading to more Ni surface sites accessible to the reactants. As SMSI was first widely recognized for Pt group metals on reducible oxide supports, the growth of the TiOx overlayer was also observed in TiO2-supported Ru and Ir NP catalysts for CO2 methanation [32, 40, 41]. Huang and coworkers [40] demonstrated the increase of the grown TiOx overlayer around Ir NPs with the increase in the reduction pretreatment temperature, along with the selectivity shift from methane to CO (Fig. 2f). Recently, Fu and coworkers reported the growth of a MoO3−x layer over Ru NPs in CO2 hydrogenation reaction gas and at a low temperature (250 °C), but its selectivity followed the aforementioned trend that the presence of an oxide layer overturned the major product from methane to CO, with the selectivity of both being close to unity [42].

As discussed previously, CO production via RWGS is also likely even with the most active methanation catalysts (e.g., Ni and Ru). Huang and coworkers [43] theoretically investigated the origin of such selectivity changes. They demonstrated that the C–O bond scission of the main intermediates (e.g., HCOO–, –COOH, and M–CO) was a key factor that determined methane/CO selectivity, which was correlated with the difference between the dissociation barrier and the desorption energy of metal carbonyls. With this knowledge, one can improve CO selectivity by decreasing the coordination number of supported metals or using metals at the lower right corner of the transition metal series. This theory likely explains the preferred formation of CO versus methane over TiO2-supported metals with low loadings [43, 44] and Au-based catalysts [45].

CO2 Hydrogenation to C2+ Products

The synthesis of C2+ products with two or more carbon atoms from direct CO2 hydrogenation via C–C coupling and carbon chain growth is attractive, as these products have a high energy density and can be used as basic chemicals for polymer synthesis or liquid fuels for transportation. Similar to CO hydrogenation, CO2 hydrogenation also suffers from low selectivity toward specific C2+ products [46,47,48,49]. Here, we discuss selective CO2 hydrogenation to several subgroups of C2+ products, including light olefins, aromatics, gasoline, and higher alcohols (e.g., ethanol). Because the pathway involved in the growth of the carbon chain for any given C2+ product contains many successive reaction steps, tandem catalysis, which utilizes more than one catalyst to promote two or more mechanistically distinct reaction steps [50], has been used as a promising strategy for achieving high selectivity.

Light Olefins

Light olefins (C2=–C4=), including ethylene, propylene, and butylene, are essential chemicals for the production of polymer materials, which are conventionally produced from steam cracking of naphtha and dehydrogenation of light alkanes. Light olefins can also be produced from CO2 hydrogenation via a modified Fischer–Tropsch synthesis (FTS) pathway. However, the selectivity of C2–C4 hydrocarbons (including olefins and paraffins) is limited by the Anderson–Schulz–Flory (ASF) distribution (typically < 58%) [29]. For instance, modifying FTS catalysts (e.g., Fe) with structural and electronic promoters (e.g., Na and Mn) can only achieve a selectivity of < 50% for light olefins in hydrocarbons at a CO2 conversion of approximately 40% [51, 52]. Thus, obtaining high selectivity toward light olefins while suppressing the production of alkane and CO (via RWGS) from CO2 hydrogenation is challenging.

A concept of catalyst design for a bifunctional oxide–zeolite system proposed by Bao and coworkers [53] has been demonstrated to effectively surpass the ASF distribution limit. This concept is also applicable to CO2 hydrogenation. Pan and coworkers [54] reported a bifunctional catalyst composed of ZnO–Y2O3 oxide and SAPO-34 zeolite, which achieved selective CO2 hydrogenation to light olefins with a selectivity of 83.9% in hydrocarbons at a CO2 conversion rate of 27.6% at 390 °C. Li and coworkers [55] developed a ZnZrO/SAPO tandem catalyst composed of ZnO–ZrO2 solid solution and Zn-modified SAPO-34 zeolite. With this catalyst, light olefins were produced from CO2 hydrogenation with a selectivity as high as over 80% in hydrocarbons and a suppressed CO selectivity of 47% at a CO2 conversion rate of 12.6% at 380 °C (Fig. 3a, b). The tandem catalysis mechanism involved the formation of CHxO* species via CO2 hydrogenation over the ZnO–ZrO2 solid solution and the transformation of migrated CHxO* species to light olefins over the Zn-modified SAPO-34 zeolite.

Fig. 3
figure 3

Reproduced with permission from Ref. [55]. Copyright 2017, American Chemical Society. c Catalytic performance and d stability of CO2 hydrogenation to aromatics over the ZnAlOx/H-ZSM-5 tandem catalyst. Reproduced with permission from Ref. [56]. Copyright 2018, Springer Nature. e Catalytic performance and f stability of CO2 hydrogenation to gasoline over the Na-Fe3O4/HZSM-5 multifunctional catalyst. Reproduced with permission from Ref. [62]. Copyright 2017, Springer Nature. g, h Performance of CO2 hydrogenation to ethanol over the bifunctional Ir1-In2O3 single-atom catalyst. Reproduced with permission from Ref. [64]. Copyright 2020, American Chemical Society

a, b Performance of CO2 hydrogenation to light olefins over the ZnZrO/SAPO tandem catalyst.

Aromatics

Aromatics (CnH2n − 6) are compounds that contain at least one benzene ring. Benzene, toluene, and xylenes of aromatic compounds are important bulk chemicals. Generally, aromatics are industrially derived from petroleum. Liu and coworkers [56] proposed a route for the synthesis of aromatics directly from CO2 hydrogenation. They reported a ZnAlOx/H-ZSM-5 catalyst with an aromatic selectivity of 73.9% and a low methane selectivity of 0.4% among carbon products (excluding CO) at a CO2 conversion rate of 9.1% (Fig. 3c). This catalyst also showed good stability during a test for 100 h (Fig. 3d). Mechanistic studies indicated that methanol and dimethyl ether were formed via hydrogenation of formate species over ZnAlOx sites, migrated to H-ZSM-5 sites, and subsequently converted into olefins and aromatics. When replacing H-ZSM-5 with Si-H-ZSM-5 in the tandem catalyst, 58.1% p-xylene in xylenes could be achieved. Li and coworkers [57] developed a ZnZrO/ZSM-5 tandem catalyst, which achieved an aromatic selectivity of 73% at a CO2 conversion rate of 14% and a suppressed CO selectivity of 44%. Sun and coworkers [58] combined a Na-Fe3O4 CO2-FTS catalyst with ZSM-5 zeolites with rich Brønsted acid sites as a tandem catalyst for CO2 hydrogenation to aromatics. This catalyst achieved a selectivity of 75% for light aromatics, and p-xylene accounted for as high as 72% of xylenes. A tandem catalysis process comprising CO2 methanation and methane aromatization with two connected reactors [59] was also demonstrated for CO2 hydrogenation to aromatics, with a benzene production rate of 0.68 µmol/(g min) at a high CO2 conversion rate of 92%.

Gasoline

Long-chain hydrocarbons in the gasoline range (C5–C11) can also be produced via direct CO2 hydrogenation. Using this route, Sun and coworkers [60,61,62,63] conducted numerous studies using multifunctional composite catalysts composed of CO2-FTS catalysts and zeolites. They developed a Na-Fe3O4/HZSM-5 catalyst that could achieve direct CO2 conversion into gasoline with a selectivity of up to 78% in hydrocarbons and only a small amount of methane (4%) at a CO2 conversion rate of 22% under industrially relevant conditions (Fig. 3e) [60]. Three types of active sites (Fe3O4, Fe5C2, and acid sites) at appropriate proximity to the multifunctional catalyst cooperatively catalyzed the tandem and synergetic catalytic conversion of CO2 into long-chain gasoline fractions. The multifunctional catalyst also showed remarkable stability for 1000 h and held great promise as a potential industrial catalyst for CO2 conversion into liquid fuels (Fig. 3f). Recently, Sun and coworkers further combined the Na-Fe3O4 CO2-FTS catalyst with two zeolites (SAPO-11 and ZSM-5) in single-bed, dual-bed, and triple-bed arrangements for the production of gasoline with a high fraction of multibranched isoparaffins [62]. The triple-bed configuration composed of Na-Fe3O4, SAPO-11, and ZSM-5 achieved a gasoline selectivity of 71.7% in hydrocarbons at a CO2 conversion rate of 31.2%, with an isoparaffin selectivity as high as 38.2%, corresponding to a research octane number (RON) value of 91.7.

Ethanol

CO2 conversion into ethanol has been realized using modified methanol synthesis or FTS catalysts. Over alkali-metal-modified methanol catalysts, methanol production is still high, leading to unsatisfactory ethanol selectivity. Over modified Fe-based or Co-based FTS catalysts, alcohol production follows the ASF distribution with a typical ethanol selectivity of approximately 35%. Thus, achieving high ethanol selectivity remains a challenge. Huang and coworkers [64] designed a bifunctional Ir1–In2O3 single-atom catalyst, which showed an excellent ethanol selectivity of > 99% and an initial turnover frequency (TOF) of 481 h−1 (Fig. 3g, h). Characterizations and DFT calculations indicated that isolated Ir sites in combination with adjacent oxygen vacancies forming Lewis acid–base pairs could activate CO2 to produce *CO intermediates over Ir sites. By contrast, In2O3 was an active catalyst for methanol synthesis and produced CH3O* intermediates. Ethanol could be formed over this bifunctional single-atom catalyst by coupling *CO and CH3O* adjacent sites. Recently, Huang and coworkers [65] developed a K0.2Rh0.2/β-Mo2C single-atom catalyst, which exhibited an ethanol selectivity of 72.1% at 150 °C.

Other CO2 Conversion Routes

Apart from hydrogenation, CO2 can also be converted and utilized through other important routes. The dry reforming of methane with CO2 provides an attractive route to simultaneously convert the two greenhouse gases into syngas [66]. Current studies mainly focus on the stability issue at a high reaction temperature, for instance, developing robust Ru/MgAl2O4 and Ni/h-BN catalysts [67, 68]. As a weak oxidant, CO2 is also able to assist the dehydrogenation of light alkanes, including ethane, propane, and isobutane, during which CO2 is reduced to CO [69,70,71]. Thermochemical splitting of CO2 into CO and O2 utilizing concentrated solar energy is also a promising route for CO2 reduction [72]. The catalytic function of metal species (e.g., IrOx and Pd) that can activate CO2 is still necessary in the CO2 splitting reaction even at high temperatures [73, 74]. Some nonreductive CO2 conversion routes can also produce valuable fine chemicals, such as organic carbonates [75, 76], cyclic carbonates [77,78,79,80], carboxylic acids [81, 82], and amides, via the N-formylation reaction [83].

Electrocatalytic CO2 Conversion

The conversion of CO2 into chemicals and fuels via electrocatalysis can directly couple CO2 utilization and the storage of renewable energy. Although the co-conversion of CO2 with other molecules (e.g., N2) has also been demonstrated [84], currently, the main electrocatalytic conversion route is still electrocatalytic CO2 reduction or CO2 electrolysis [85]. Here, it should be noted that there is no strict difference between CO2 electroreduction and CO2 electrolysis. CO2 electroreduction is usually preferred by the catalysis community, whereas CO2 electrolysis is often used by the electrochemistry community. Rather than using H2 produced by water electrolysis in thermocatalytic CO2 hydrogenation reactions, the electrocatalytic CO2 reduction reaction utilizes active hydrogen species directly derived from water activation to achieve the reduction of CO2. Compared with thermal catalysis, it can drive electrocatalytic CO2 reduction under mild conditions, e.g., room/low temperatures (< 100 °C) and ambient pressures. By contrast, high-temperature CO2 electrolysis is used in some application scenarios, such as O2 production for space missions on Mars in the 1960s and direct electrolysis of high-temperature flue gas containing highly concentrated CO2.

High-Temperature CO2 Electrolysis to CO and Syngas

Figure 4a shows the energy demand for CO2 electrolysis to CO (CO2 → CO + 1/2O2) at different temperatures [86]. As the enthalpy change (∆H) is constant regardless of temperature, the electrical energy demand (∆G) decreases, whereas the heat demand (TS) increases with the increase in the reaction temperature. Therefore, a high-temperature CO2 electrolysis process heated with industrial waste heat and Joule heat inside an electrolysis cell can effectively lower the electrical energy consumption. Thermodynamic analysis indicates that the formation of CO and/or syngas becomes more favorable at high temperatures compared with other CO2 reduction products. High-temperature CO2 electrolysis is usually conducted in solid oxide electrolysis cells (SOECs) using oxygen-ion-conducting or proton-conducting solid oxides as electrolytes at high operating temperatures (up to 1000 °C). Different electrode reactions occur at cathodes and anodes when oxygen ions or protons are used as ion carriers. SOECs using proton-conducting electrolytes can work at low-to-intermediate temperatures (400 to 700 °C), at which reduced products, such as hydrocarbons, can also be generated from the point of view of thermodynamics. Because of the poor stability of proton-conducting electrolytes under CO2 electrolysis conditions, to date, most studies have been conducted using oxygen-ion-conducting SOECs [87, 88]. When CO2 and water are co-fed to SOECs, the final product can be syngas with different H2/CO ratios, even hydrocarbons (e.g., methane and ethylene) with the aid of other metal catalysts for syngas conversion [88, 89].

Fig. 4
figure 4

Reproduced with permission from Ref. [86]. Copyright 2019, Wiley–VCH. b Density and size of RuFe alloy NPs on SFRuM perovskite surfaces. c CO2 electrolysis performance and d stability of SFRuM after redox treatments. Reproduced with permission from Ref. [104]. Copyright 2021, Wiley–VCH. e CO2 electrolysis current density and f electrochemical impedance spectroscopy at 1.4 V. Reproduced with permission from Ref. [105]. Copyright 2021, American Chemical Society

a Temperature dependence of energy demand for CO2 electrolysis to CO (CO2 → CO + 1/2O2).

The state-of-the-art cathode material for CO2 electrolysis in SOECs is yttria-stabilized zirconia-supported Ni (Ni/YSZ). However, the Ni/YSZ cathode suffers from carbon deposition, grain coarsening, and Ni oxidation under the reaction conditions (i.e., CO2 atmosphere and high temperature), leading to poor stability [90, 91]. Recently, mixed ionic and electronic conductors, such as perovskite oxides (ABO3), have been used as alternative cathode materials because of their good redox stability and high carbon deposition resistance; however, they show inferior CO2 electrolysis performance compared with the Ni/YSZ cathode. The performance of perovskite oxides has been improved by element doping [92,93,94,95,96], anchoring single-metal sites on the surface [97], constructing catalytically active oxide–oxide interfaces via infiltration [98, 99] and metal–oxide interfaces via metal decoration [100], and in situ exsolution of metal NPs [101,102,103,104]. The in situ exsolution treatment can produce many metal–alloy NPs (e.g., Rh, Pd, Pt, Fe, Ni, Co, FeNi, CoFe, and RuFe) on perovskite surfaces under reducing environments; however, the number of NPs is low because of the sluggish diffusion of B-site dopant cations in the bulk perovskite. Through repeated redox treatments, Wang and coworkers [104] promoted the exsolution of RuFe alloy NPs on Sr2Fe1.4Ru0.1Mo0.5O6−δ (SFRuM) perovskite. The dynamic exsolution and redissolution processes of the RuFe alloy NPs were imaged by in situ secondary electron scanning transmission electron microscopy at 600 to 850 °C in a 10-Pa H2 or O2 atmosphere. The density of exsolved RuFe alloy NPs was as high as approximately 21,000 µm−2 after four redox cycles, whereas the NP size did not change (Fig. 4b). The in situ grown RuFe@SFRuM interface after four redox cycles increased the current density of CO2 electrolysis (approximately 0.65 A/cm2) by 74.6% and contributed to high stability for 1000 h at 1.2 V and 800 °C (Fig. 4c, d). Zhu and coworkers [105] reported the in situ dispersion of an Au layer used as a current collector to Au NPs (down to 2 nm) at 800 °C by applying high cell voltages. During this transformation, Au/YSZ was reconstructed into nano-Au/Zr–suboxide interfaces, which resulted in a decrease in the polarization resistance by a factor of 75, along with a 38-fold improvement in CO2 electrolysis current density (Fig. 4e, f).

Low-Temperature Electrocatalytic CO2 Reduction

Low-temperature electrocatalytic CO2 reduction occurs at room temperature to low temperature (< 100 °C) using liquid water and/or vapor as a reactant. Table 2 lists the standard potentials of CO2 electroreduction for various products. The CO2 electroreduction reaction is thermodynamically not that difficult, but it is hindered by the hydrogen evolution reaction (HER). Moreover, CO2 electroreduction involves multiple proton and electron transfer steps, leading to highly complex reaction pathways toward a variety of products, including CO, formic acid (formate), methane, methanol, ethylene, ethanol, acetate acid (acetate), and n-propanol. Therefore, selectivity control is a core aspect of CO2 electroreduction studies, particularly for reduced and C2+ products. In terms of practical applications, high selectivity toward given products should be obtained at an industrial current density, with considerable stability.

Table 2 Standard potentials of CO2 electroreduction to various products at pH = 0

Electrocatalytic CO2 Reduction to CO and Formic Acid (Formate)

CO and formate are the most widely reported C1 products from CO2 electroreduction. Although Hori’s seminal works [106, 107] in the 1980s have classified bulk metals into several groups in terms of major products, such as CO, formate, hydrocarbons, or H2, remarkable differences can be observed in nanostructured metal catalysts [108,109,110,111,112,113,114,115,116,117]. For instance, Pd is an active metal for HER, and bulk Pd mainly produces H2 in CO2 electroreduction; however, Pd NPs show impressive performance for CO2 electroreduction to CO or formate, both with a Faradaic efficiency of > 90%, depending on the applied electrode potentials (Fig. 5a). This selectivity switch can be rationalized by potential-induced and intermediate-induced active phase transition under the reaction conditions [8, 118, 119]. The CO production over Pd NPs can be further improved by alloying Pd with second metals [120, 121]. Notably, the electrochemical promotion of the catalysis effect was also observed in CO2 hydrogenation to formate over Pd NPs in the liquid phase under ambient conditions [122]. Inspired by the concept of nanoconfinement catalysis proposed by Bao [10], Wang and coworkers [123] developed an Au(Ag)-CeOx catalyst with abundant metal–oxide interfaces. In situ scanning tunneling microscopy and X-ray photoemission spectroscopy results indicated that these interfaces enhanced CO2 adsorption and hydrogenation to the initial *COO and *COOH intermediates, thus leading to high Faradaic efficiency and partial current density for CO production (Fig. 5b). Another important example is using single Ni atom catalysts [124,125,126,127,128,129,130,131,132] instead of Ni bulk or NPs that are good HER catalysts. Wang and coworkers [124] developed nickel–nitrogen (Ni–N) sites embedded within porous carbon with a Ni loading of up to 5.44 wt% through pyrolysis of Zn/Ni bimetallic zeolitic imidazolate framework-8. This catalyst achieved a high CO current density of 71.5 mA/cm2, corresponding to a TOF of 10,087 h−1 with a CO Faradaic efficiency of more than 92% in an H-cell (Fig. 5c, d). The atomically dispersed Ni–N sites are coordinatively unsaturated and can suppress the HER [125]. Heteroatom-doped carbon materials have also been used for CO2 electroreduction to CO [133,134,135]. Apart from the catalysts themselves, the reaction conditions, such as electrolyte composition and temperature, also play important roles in determining selectivity [136,137,138].

Fig. 5
figure 5

Reproduced with permission from Ref. [118]. Copyright 2017, Springer Nature. b Au–CeOx interface enhances CO2 electroreduction. Reproduced with permission from Ref. [123]. Copyright 2017, American Chemical Society. c, d CO2 electroreduction performance of Ni–N–C catalyst. Reproduced with permission from Ref. [124]. Copyright 2018, Royal Society of Chemistry. e CO2 electroreduction performance of CoPc@Zn-N–C catalyst. Reproduced with permission from Ref. [149]. Copyright 2020, Wiley–VCH. f CO2 electroreduction performance of Cu-CuI composite catalyst. Reproduced with permission from Ref. [161]. Copyright 2021, Wiley–VCH

a Selectivity switch of CO2 electroreduction over Pd/C catalysts.

In addition to the unique Pd catalysts [8, 118, 119], formate can also be produced using a series of catalysts, such as Bi, Sn, and Pb compounds [139,140,141,142,143], Cu alloy with Sn and In [144,145,146], and single-atom alloy catalysts [147]. Wang and coworkers [144] reported an in situ reconstructed hierarchical Sn-Cu/SnOx core/shell (Sn2.7Cu) catalyst, which achieved a current density of 406.7 mA/cm2 at − 0.70 V versus reversible hydrogen electrode (RHE) and high stability at 243.1 mA/cm2 for 40 h at − 0.55 V versus RHE in a flow cell. Xiao and coworkers [147] developed a single-atom Pb-alloyed Cu catalyst (Pb1Cu) that can exclusively convert CO2 into formate with approximately 96% Faradaic efficiency at more than 1 A/cm2. Formate production occurred over the modulated Cu sites rather than the isolated Pb sites in the Pb1Cu catalyst, following an HCOO* pathway. The production of a pure formic acid solution at 100 mA/cm2 for 180 h was further demonstrated with the Pb1Cu catalyst in a 3-cm2 electrode device using a proton-conducting solid electrolyte.

Electrocatalytic CO2 Reduction to Methane

Methane, as a fully reduced C1 product, is usually produced using Cu catalysts at negative potentials [148]. However, some non-Cu catalytic materials can also reduce CO2 to methane. Wang and coworkers [149] demonstrated a cobalt phthalocyanine (CoPc) and zinc–nitrogen–carbon (Zn–N–C) tandem catalyst that can efficiently convert CO2 into methane. The methane production rate of CoPc/Zn–N–C was over 100 times higher than that of CoPc or Zn–N–C alone (Fig. 5e). This catalyst decoupled complicated methane pathways over single active sites (for Cu catalysts) into a two-step tandem reaction. CO2 was reduced to CO over CoPc; then, CO diffused onto Zn–N–C for further hydrogenation to methane.

Electrocatalytic CO2 Reduction to C2+ Products

Compared with C1 products, C2+ products, such as ethylene, ethanol, and n-propanol, have higher energy density and broader availability. The formation of these products involves C–C coupling, a key topic in fundamental catalysis research. More importantly, selectivity control after the C–C coupling step would determine the product distribution among C2+ products [150,151,152,153,154,155,156]. The highly selective production of a specific single C2+ product relies on the proper mechanistic understanding of full reaction pathways. The oxidation state of Cu has been considered to play a critical role in the formation of C2+ products, with the Cu+ species being the most catalytically active because of the presence of Cu0/Cu+ interfaces [157,158,159,160]. Wang and coworkers [161] designed a Cu–CuI composite catalyst with abundant Cu0/Cu+ interfaces by physically mixing commercial Cu NPs and CuI powders. This catalyst achieved a remarkable partial current density of 591 mA/cm2 for C2+ products at − 1.0 V versus RHE in a flow cell (Fig. 5f). The high-rate C2+ generation was attributed to the presence of residual Cu+ and adsorbed iodine species, which could facilitate CO adsorption and subsequent C–C coupling. Although the total C2+ Faradaic efficiency could reach 71% at such a high current density, further efforts should be made to improve the selectivity toward a single C2+ product, such as ethylene and ethanol.

Tandem Catalysis Mechanism in Catalytic CO2 Conversion Reactions

The selective generation of highly reduced products, such as hydrocarbons and oxygenate, via catalytic CO2 conversion is not only desirable but also challenging because of multiple electron–proton transfer steps during these reactions. Tandem catalysis has been demonstrated as an emerging and promising strategy in CO2 conversion into highly reduced products [162]. In principle, a tandem catalyst with two or more catalytically active sites is expected to break the linear scaling relationship of the adsorption of key intermediates, resulting in remarkably improved selectivity toward specific products [53]. As mentioned previously, Wang and coworkers [149] fabricated a CoPc/ZnN4 tandem catalyst (Fig. 6a), in which CO2 was reduced to CO over CoPc and the generated CO diffused onto the ZnN4 site for subsequent hydrogenation to methane. This tandem effect is also applied to promote the formation of C2 products (e.g., ethylene) when combining CO-producing sites with Cu sites into a single catalyst (Fig. 6b), for instance, Cu/Ag [163] and Cu/Ni–N–C [164]. In thermocatalytic CO2 conversion, tandem catalysis plays a vital role in generating highly reduced and multicarbon products [54,55,56,57, 60, 165,166,167]. Generally, the oxide component of these tandem catalysts accounts for the formation of C1 intermediates (e.g., *CO and *CHxO), whereas the zeolite component catalyzes C–C coupling to form light olefins, aromatics, gasoline, and higher alcohols (Fig. 6c).

Fig. 6
figure 6

Copyright 2020, Wiley–VCH. b Reproduced with permission from Ref. [163]. Copyright 2019, American Chemical Society. c Reproduced with permission from Ref. [57]. Copyright 2019, Cell press

Proposed reaction mechanisms for CO2 electroreduction to a methane over CoPc/ZnN4 tandem catalyst and b ethylene over Cu/Ag tandem catalyst as well as c for CO2 Hydrogenation to aromatics over ZnZrO/ZSM-5 tandem catalyst. a Reproduced with permission from Ref. [149].

Development of CO2 Electrolyzers at Industrial Current Densities

Beyond catalyst design, the rational design of efficient CO2 electrolyzers is also of great importance for scaling up CO2 electroreduction toward practical applications. The electrolyzers should be operated at industrial current densities (> 200 mA/cm2), which could be achieved using the gas diffusion electrode configuration that breaks the limitation of CO2 solubility. Flow and membrane electrode assembly (MEA) cells are the two main CO2 electrolyzers [168, 169], of which MEA electrolyzers hold great promise because of their low iR drop, compact structure, improved stability, and high full cell energy efficiency (from electrical energy to chemical energy). With such electrolyzers, it is likely to only feed CO2 and pure water as reactants without using any liquid electrolyte solution [170]. Wang and coworkers [171] developed a 4-cm2 alkaline MEA CO2 electrolyzer working at industrial current densities (Fig. 7a), which achieved a total Faradaic efficiency of 80% for CO2 electroreduction over commercial Cu NPs at 350 mA/cm2 with an energy efficiency of approximately 30% under optimized reaction conditions (Fig. 7b, c). The scale-up demonstration with a geometric electrode size of 100 cm2 at the laboratory level has been achieved by Wang and coworkers. With the continuous technological improvement of the electrode structure, ion exchange membrane, and electrolyzer, current density (CO2 reduction rate) and energy efficiency are expected to significantly increase [172]. To increase single-pass CO2 conversion, large-area electrolyzers and stacks should be assembled, and this has been investigated by some groups [173].

Fig. 7
figure 7

Reproduced with permission from Ref. [171]. Copyright 2020, Springer Nature

a Schematic of MEA CO2 electrolyzer. b, c CO2 electroreduction performance of Cu catalyst measured in MEA electrolyzer.

Although the anion exchange membrane-based (alkaline) MEA electrolyzer holds great promise, it encounters a basic problem in low-temperature CO2 electrolysis, i.e., carbonate formation [174, 175]. CO2 reacts with surface OH ions produced during electrolysis to form carbonate (CO32−/HCO3). Carbonate would precipitate in salt form to block catalyst surfaces, which reduces electrode hydrophobicity, leading to flooding. Moreover, carbonate can transfer to the anode through the anion exchange membrane, thereby leading to significant carbon loss and a low CO2 utilization efficiency [176, 177]. Possible strategies for dealing with the carbonate issue are to use a cation exchange membrane, bipolar membrane, and solid electrolyte buffer layer [168, 169, 178,179,180]. This leads to an emerging direction, i.e., acid CO2 electrolysis that has been demonstrated to suppress carbonate formation [181,182,183,184]. Moreover, the use of a two-step tandem conversion process, i.e., CO2 electrolysis to CO in SOECs and CO electrolysis to C2+ (e.g., ethylene and ethanol) in MEA electrolyzers, is feasible [185]. This tandem process also highlights the full use of individual advantages of high-temperature electrolysis (for the desired CO2 activation) and low-temperature electrolysis (for the desired deep reduction of carbon intermediates). Another promising two-step CO2–to–C2+ conversion process is to connect an acidic MEA electrolyzer and an alkaline MEA electrolyzer in sequence, which combines the advantage of acidic CO2 electrolysis (free of carbonate formation) and alkaline CO electrolysis (improved C2+ production). 

Photo(electro)catalytic CO2 Conversion

The conversion of CO2 and water into chemicals and fuels via solar-driven photocatalysis and photoelectrocatalysis (PEC) with the aid of an efficient photocathode, which mimics the natural photosynthetic process, is generally known as artificial photosynthesis. The products of photocatalytic and PEC CO2 reduction mainly include CO, methane, and ethylene [9, 186,187,188,189]. To date, semiconductor materials have been widely used to fabricate photocathodes for photocatalytic and PEC CO2 reduction; however, the solar-to-fuel energy conversion efficiency is still unsatisfactory for practical applications. To improve the performance of photocatalytic/PEC CO2 reduction, the construction of heterostructure photocathodes comprising multiple components is an important approach. Li and coworkers [190] developed an Au/p-GaN plasmonic heterostructure photocatalyst for selective and unassisted gas phase photocatalytic CO2 reduction to CO under visible-light illumination. A metal/insulator/semiconductor configuration with a nanometer-thick aluminum oxide layer between Au/Cu NPs and p-GaN could remarkably improve the CO production rate (Fig. 8a). Zong and coworkers [191] reported a sandwich-like organic–inorganic hybrid perovskite-based photocathode for PEC CO2 reduction to CO. The photocathode decorated with carbon encapsulation and a cobalt phthalocyanine molecular catalyst exhibited a high photocurrent density of − 15.5 mA/cm2 at − 0.11 V versus RHE under air mass (AM) 1.5 G illumination with a stable CO Faradaic efficiency of more than 80% for 25 h (Fig. 8b). Unbiased PEC CO2 reduction in tandem with an amorphous Si photoanode achieved a total solar-to-fuel energy conversion efficiency of 3.85%.

Fig. 8
figure 8

Reproduced with permission from Ref. [190]. Copyright 2021, American Chemical Society. b Faradaic efficiencies of CO and H2 in unassisted PEC tandem system. Reproduced with permission from Ref. [191]. Copyright 2020, Wiley–VCH

a Photocatalytic CO2 reduction to CO production of Au/p-GaN, Cu/Au/pGaN, and Cu/Au/Al2O3/p-GaN samples under visible-light illumination.

Concluding Remarks

The recent progress in the conversion of CO2 into valuable chemicals and fuels (e.g., CO, formic acid, methanol, methane, ethanol, acetic acid, propanol, light olefins, aromatics, and gasoline) via thermal catalysis, electrocatalysis, and photo(electro)catalysis has been reviewed, with an emphasize on highlighting some representative studies. The structure–performance correlations of typical catalysts used in these reactions have been revealed by combined advanced in situ spectroscopy and microscopy characterizations and DFT calculations. The control in activity, selectivity, and stability in future catalysis research at a fundamental level should aim at addressing issues facing practical applications. To reduce the cost of downstream separation and purification processes, catalytic selectivity toward a single CO2 reduction product/fraction should be further improved at an industrially relevant CO2 conversion rate with considerable stability in the future. Such selectivity control can be achieved by designing new catalytic materials and developing new CO2 conversion routes with coupling multiple external fields (e.g., light, thermal, electrical, and magnetic). From the view of fundamental research, catalysts should be characterized in their working state during CO2 conversion reactions using in situ/operando methods with high temporal and spatial resolutions and sufficient surface sensitivity to gain a proper understanding of their structure–performance correlations.

The scale-up of some catalytic CO2 conversion routes has been demonstrated at the industrial level in China and around the world, e.g., the liquid sunshine demonstration project in Lanzhou, China, which synthesizes methanol from CO2 hydrogenation coupled with photovoltaic-driven water electrolysis; the pilot project of direct CO2 hydrogenation to gasoline in Jining, China; the high-temperature SOEC CO2 electrolysis demonstration by the company of Haldor Topsøe; and low-temperature CO2 electrolysis demonstration by the company Siemens. This exciting industrial progress encourages the research community to further join forces to scale-up other important catalytic CO2 conversion routes, such as CO2 hydrogenation to light olefins and aromatics and tandem CO2 electrolysis (CO2–CO–ethylene) in connected SOEC and MEA electrolyzers as well as connected acidic and alkaline MEA electrolyzers.