Recent progress of theoretical studies on electro- and photo-chemical conversion of CO₂ with single-atom catalysts

Abstract

The CO₂ reduction reaction (CO₂RR) into chemical products is a promising and efficient way to combat the global warming issue and greenhouse effect. The viability of the CO₂RR critically rests with finding highly active and selective catalysts that can accomplish the desired chemical transformation. Single-atom catalysts (SACs) are ideal in fulfilling this goal due to the well-defined active sites and support-tunable electronic structure, and exhibit enhanced activity and high selectivity for the CO₂RR. In this review, we present the recent progress of quantum-theoretical studies on electro- and photo-chemical conversion of CO₂ with SACs and frameworks. Various calculated products of CO₂RR with SACs have been discussed, including CO, acids, alcohols, hydrocarbons and other organics. Meanwhile, the critical challenges and the pathway towards improving the efficiency of the CO₂RR have also been discussed.

Recent progress of theoretical studies on electro- and photo- chemical conversion of CO₂ with single-atom catalysts has been presented, from the intermediates to the final products.

1. Introduction

Due to the increasing fossil fuel combustion and energy consumption, carbon dioxide (CO₂) emission has led to serious global environmental problems such as global warming, sea level rising and ocean acidification.¹ The shortage of fossil energy has become a serious problem which needs to be solved urgently. Fortunately, it has been found that CO₂ can be used as a chemical feedstock to synthesize a variety of valuable chemical products or be converted into specific fuel substances such as methanol and methane. It will not only decrease the pressure on energy demand, but also reduce the amount of greenhouse gases.

Many efforts have been devoted to the conversion of CO₂ in recent years.^2–5 The electrocatalytic CO₂ reduction reaction (CO₂RR) has been widely applied for both industry and academia because of its high efficiency and activity.^6,7 The investigations normally focus on the electrochemical overpotential, current efficiency and products selectivity.^8,9 Nowadays, due to the requirement of energy regeneration and environment protection, photocatalysts have been paid more and more attentions because of the significance in the future applications of CO₂RR.^10–15

The concept of single-atom catalysis (SACs), firstly proposed by Zhang, Li, Liu et al.,¹⁶ has been widely used in heterogeneous catalysis,^17–20 which have shown the rosy prospects to achieve high activity, selectivity and ∼100% atom utilization. From then on, it had developed preternaturally fast on enabling heterogeneous reduction products.^21–27 The – radical intermediate formed from CO₂ activation will affect the reaction pathways due to its instability and readily protonation, and further influence the activity and selectivity of electro- and photo-chemical reaction process. Some efforts have been taken to adjust the interaction between the intermediates and catalysts, such as improving the activities of catalyst surface and reaction environment.²⁸ For SACs, the unsaturated coordination can effectively regulate the relevance of CO₂ and correlative intermediates with particular active sites, which is promising for highly efficient CO₂RR catalysts.^29–31 However, the metal monatom has high surface energy that tends to aggregate particles with low surface energy during the reaction. Therefore, the preparation of stable SACs is still a great challenge. It has been proved that theoretical calculations methods such as density functional theory (DFT), molecular dynamics (MD) and Hartree–Fock (HF) are important for the data reliability. For instance, Qiu et al. revealed that Co–N₅ formed by inducing electron delocalization of Co–N₄ sites using nitrogen-rich carbon nitride as supporting material can improve the stability and conductivity by DFT calculations. The catalyst synthesized according to this prediction can remain for 40 h with excellent performance in CO₂ conversion to CO with >99% selectivity.³² Metal–organic framework (MOF) materials, with periodic structure and designability of functional groups, have also been found an ideal carrier to stabilize metal monatom when coordination chelating sites were introduced. Hai et al. summarized theoretical studies of MOFs, in which the performances of electrocatalysts for CO₂RR were involved. The MOFs showed the better catalytic performance than that in the respect of potential barriers, overpotentials, the energy barrier of C–C dimerization, selectivity and activity for CO₂RR.³³ Thus, it is promising to investigate CO₂RR with SACs from the perspective of theoretical calculations, which can guide the catalysts design. The intermediate products of CO₂RR include but not limit *COOH, *CO, *CHO, *OCH₂, *OCH₃, *COH, *CHOH, *CH₂OH, *CH₃OH, *CH₂ (as shown in Fig. 1). The reduction of CO₂ can generate CO, methane, ethylene, acetylene and other organics.

Fig. 1. The intermediate and final products of quantum-theoretical studies on electro- and photo-chemical CO₂RR with SACs.

Above all, we need to provide an updated review to investigate the theoretical research on electro- and photo-chemical conversion of CO₂ with SACs. The framework is also described with reaction products, which will provide guidance for the experimental design of catalysts.

2. Electrochemical conversion of CO₂

In Electrochemical CO₂RR (eCO₂RR), electrocatalyst plays a key role because of its promising application on selectively synthesizing various C₁, C₂, and other polycarbonate compounds. However, the kinetics of eCO₂RR is limited due to the stable linear molecule structure of CO₂. At the same time, the transfer of multiple electrons and protons causes intricate final products, including HCOOH,³⁴ CO,³⁵ HCOH,³⁶ CH₃OH,³⁷ CH₄,^38,39 as well as competitive hydrogen evolution reaction (HER).⁴⁰ In order to improve the activity, selectivity and stability of the catalyst, the theoretical study on eCO₂RR can be commenced from two aspects, the design of the catalyst and the expand of the reaction mechanism.

At present, most of catalyst designs focus on the final products or reaction characteristics with different types of catalyst supports and active sites. It is proposed two strategies to improve: (1) increase the number of active sites; (2) enhance the intrinsic activity of single active site.⁴¹ In addition, based on the theoretical calculations, the catalysts can be designed and predicted in microscopic level. The reaction energy barriers of intermediates for different catalysts are calculated to further investigate the reaction pathway and mechanism of CO₂RR.

2.1. eCO₂RR to CO

As one of the most fascinating conversion reaction methods, eCO₂RR to CO has attracted lots of attentions due to the Fischer–Tropsch (FT) type reaction.⁴² Meanwhile, it's a key point to enhanced the selectivity for the eCO₂RR to CO in liquid environment.⁴³ Most of these works aimed at clarifying the reaction paths and selectivity trends, yet the mechanism of catalytic is still debatable which will not be further studied in this paper. eCO₂RR to CO is a 2-electron reduction reaction consisted with two basic steps.⁴⁴ CO₂ reductively adsorbed on the surface of catalyst involved a proton coupled electron transfer process. As a result, *COOH intermediate is formed. The adsorbed intermediate further turned into *CO when another electron proton transfer, finally desorbed from the catalyst surface. Recently, SACs have been widely studied for eCO₂RR to CO because of its superior activity and selectivity compared with traditional catalysts. Table 1 summarizes the key parameters of eCO₂RR to CO with various SACs, including the reaction energy (ΔE) of the main steps, Faraday efficiency, over potential and max current density.

The key parameters on eCO₂RR to CO with different SACs.

SACs	ΔE (eV)	Faraday efficiency (%)	Overpotential (V)	Max current density (mA cm−2)	TOF (h−1)	Calculation method	Ref.
FeN5	0.77	97.0	0.35	—	—	DFT	45
CoPc	−0.1	99	−0.8	—	—	DFT	1
Co–N5	0.67	99.2	−0.73	4.5	480.2	DFT	46
Ni–N3–C	0.66	95.6	−0.65	6.64	1425	DFT	47
Fe–N4	0.94	86.5	−0.75	5	—	DFT	7
Pd–N4	—	55	−0.5	—	—	DFT	48
Cu–N4–NG	1.28	80.6	−1	—	—	DFT	49
Zn SAs/N–C	∼0.6	94.7	0.33	121.5	8190	DFT	50
ZnO@ZIF-NiZn	—	98	−1	34.3	9366	DFT	51
Cu–N4	−0.9	96	−0.7	—	—	DFT	52
N–C–CoPc NR	−0.34	85.3	−0.7	30(0.18 V)	—	DFT	53
Co@Pc/C	0.89	84	−0.9	28	—	DFT	54
SA-Fe/NG	∼0.5	97	−0.5	—	—	DFT	55
Fe–N4	0.29	94	0.58	4.59(−0.58 V)	1630	DFT	56
Fe–N/CNF	—	95	−0.53	4.71	3104	DFT	57
F-CPC	—	88.3	−1	37.5	—	DFT	58
Ni–N4	0.89	98	−0.5	35.9(−1.35 V)	—	DFT	59
Ni/Fe–N–C	0.47	98	−0.7	23.7(−1.0 V)	7682	DFT	60
Ni–NG	0.12	95	0.55	50	2100	DFT	61
Ni–N–C	∼1.7	85	−1	200	—	DFT	62
Ni–SA–NCs	—	99	−0.8	50(−1.0 V)	—	DFT	63
Ni@NC-900	—	96	−1	−17	—	DFT	64
Ni–N@NPC	1.53	98.44	0.67	30.96	2825	DFT	65
NiPc–OMe	∼0.4	99.5	−0.61	−150	—	DFT	66

Besides, the single-atom metal-composite materials have exhibited unique properties in eCO₂RR. The metal centers dispersed atomically and coordination-unsaturated are favorable for providing active sites. Wang et al. synthesized a graphene-based Fe–N₅ SCAs as seen in Fig. 2a. The weak binding energy of CO is conducive to dissociation from the adsorption site of the catalyst. The valence electrons of Fe atom are reduced due to the axial pyrrole nitrogen ligand at Fe–N₅ site, which weaken the back donation from the 3d orbital of Fe atom to the π* orbital of CO to realize the rapid desorption. High CO-selectivity is achieved as a result. Fig. 2b shows the diverse rate determining steps of Fe–N₄ and Fe–N₅ catalysts. The plane charge density by O–C–Fe–N (Fig. 2c) indicated the decrease of charge density of Fe atom. It confirms that the oxidation state of Fe in Fe–N₅ result in a weaker CO adsorption strength.⁴⁵ Deng et al. have doped different transition metals (Mn, Fe, Co, Ni and Cu) in the M–N₄ coordination phthalocyanine (MePcs). According to the results of DFT calculation (Fig. 2d), the rate determining steps for Co, Fe, and Mn reaction paths are the dissociation of *CO on the catalyst surface. But for Ni and Cu, the formation of *COOH exhibits the highest energy barriers among all the steps in CO₂RR. Fig. 2e describes the linear relationship between the adsorption energies of *CO and the reaction energies of CO desorption or *COOH formation. The two linear relations reveal an inverted volcano curve, and CoPc is the closest to the peak of volcano. Thus, the Co center in CoPc is the optimist reaction site by calculation. Fig. 2f shows the intermediate from CO₂ adsorption on the catalyst surface to the final continuous hydrogenation to CO*.¹ Cao et al. reported a method of anchoring atomically dispersed Co on polymer-derived N-doped hollow porous C spheres. The Co–N₅ plays a critical role in the CO₂ activation, *COOH formation and CO dissociation. Its catalytic activity is 15.5 times by CoPc.⁴⁶

Fig. 2. (a) The structures of single-atom FeN₄ and FeN₅ catalysts. (b) The free energy of the optimized intermediate for the electroreduction of CO₂ to CO. (c) The plane charge density distribution of O–C–Fe–N. (d) The free energy of eCO₂RR on several SACs. Illustration shows the adsorption energy of *COOH. (e) CO dissociation and *COOH formation tendency of catalysts doped with transition metals. (f) The adsorption configurations of intermediates in liquid environment. The balls in green, blue, grey, red and white represent Co, N, C, O and H atoms, respectively (reproduced from ref. 1 with permission from John Wiley and Sons, copyright 2018; reproduced from ref. 45 with permission from John Wiley and Sons, copyright 2019).

Meanwhile, adjusting the coordination number can effectively influence the catalytic activity of eCO₂RR. For instance, N-doped catalysts verified low coordination N could accelerate the formation of *COOH intermediate and promote eCO₂RR to CO with the theoretical calculations;⁴⁷ C-based SACs could exhibit low activation energy and favorable CO selectivity;⁷ Pd–N₄ SACs also showed satisfactory reaction paths to the dissociation of CO.⁴⁸ Recently, Qiao et al. successfully synthesized a N-doped graphene supported Cu SACs (Cu–N₄-NG) which exhibited more favorable thermodynamics and kinetics than HER.⁴⁹ They further studied the electrocatalytic activity of 3d transition metal doped M–N₄–C (M = Mn, Fe, Co, Ni, Cu and Zn) catalysts for eCO₂RR.⁶⁷ Coincidentally, Lin et al. successfully prepared single atom Zn supported on NC nanofibers, and found that Zn–N₄ could effectively reduce the free energy barrier of *COOH.⁵⁰ Various Ni doped catalysts supporting have been extensively studied and demonstrated excellent performance in CO₂RR.^59–66 Wang et al. found that Ni–N@NPC with high dispersed Ni–N sites and desirable CO₂ adsorption capacity exhibited high faradaic efficiency for CO (about 98.44%) and durable stability over 30 hours. The eCO₂RR activity was indicated by DFT calculation which revealed that Ni–N sites can decrease the kinetic energy barriers for *CO₂ to *COOH.⁶⁵ Sun et al. reported a thermodynamically stable Ni single atom with a Ni–N₃ structure and further wrapped in porous N-doped C shell. The free energy barrier for *COOH on Ni@N₃ is much lower than that of Ni@N₄.⁶⁸

Moreover, Han et al. successfully prepared N, P-co-doped carbon aerogels, which could be used as electrocatalysts to efficiently reduce CO₂ to CO. With high active area and conductivity, it provided a promising approach for non-metallic catalysts.⁶⁹Fig. 3a shows the configuration of N-/P-doped, and N, P-co-doped carbon. As shown in Fig. 3b, the free energy barrier is significantly reduced after doping with heteroatoms. Among them, the pyridinic N exhibited the lowest free energy barrier. Besides that, the adsorption of H* on both tri-pyridinic N and pyrrolic N is strong from the HER free energy in Fig. 3c. Thus, the pyridinic N is the active site in eCO₂RR to CO. Moreover, the free energy changes of eCO₂RR and HER on carbon Co doped with N–P were also calculated (Fig. 3d and e). N, P-co-doped carbon aerogels have considerable activity and selectivity for the eCO₂RR to CO. Furthermore, they synthesized Sn₄P₃/reduced graphene oxide nanocomposites and Cu–Co bimetallic nanoparticles which exhibited excellent catalytic activity and selectivity of eCO₂RR to CO.^70–72

Fig. 3. (a) A series of N/P-doped, and N, P co-doped carbon configurations. (b) Gibbs free-energy (ΔG) diagrams for eCO₂RR to CO. (c) ΔG for HER. (d) ΔG for eCO₂RR to CO based on N, P-co-doped carbon. (e) ΔG for HER based N, P co-doped carbon (reproduced from ref. 69 with permission from John Wiley and Sons, copyright 2020).

In addition, eCO₂RR is thermodynamically more favorable compared with HER. For instance, N-doped carbon catalyst with Ni sites can exhibit nearly 100% CO selectivity and high activity.⁵¹ Atomically Fe³⁺ site facilitate high-efficiency CO₂ electroreduction catalysis which catalyze eCO₂RR to CO under a low overpotential. Zn supported on N-doped carbon catalyst can inhibit the generation of CO and increase the selectivity of CH₄.²⁹ Except for the use of carbon-based materials as the catalyst carrier, boron nitride is similar to graphene and has excellent physical and chemical properties. Therefore, transition metal-doped BN nanomaterials have been investigated in eCO₂RR with SACs. Transition metals (TM = Sc to Zn, Mo, Ru, Rh, Pd and Ag) are doped on BN monolayers. The DFT calculation present that monoatomic TM–BN have excellent catalytic performance as catalysts for reducing CO₂ to CH₄.⁷³

In recent years, the single atom alloy has developed catalytically active element composed of atoms dispersed in host metal with high catalytic selectivity. This catalyst combines the traditional advantages of alloy catalysts with the new features of SACs. The adsorption energy of *CO corresponds to the theoretical overpotential of methane or methanol, so as to achieve the screening of high-performance catalysts,⁷⁴ further effectively reduce the activation barrier of CO₂.⁷⁵ The weak binding energy of CO greatly reduces the possibility of catalyst poisoning and further protonation, leading to perfect CO selectivity and stability.⁷⁶

2.2. eCO₂RR to alcohols

Another strategy to reduce CO₂ is to generate safe and transportable liquid fuels. Alcohols has become an attractive product due to its core position in the fields of national defense, industry and agriculture.⁷⁷ However, because of the complex reaction intermediates, high energy barrier and complicated mechanism of C–C coupling, the selectivity of alcohols is not optimistic.⁷⁸

By reducing the size of metal-based catalysts to single atom level,^79,80 Wang et al. successfully prepared cobalt atom electrocatalysts/nanoparticles and carbon/selenium on N-doped porous carbon (Co_SA–HC, Co_NP–HC, Se@Co_SA–HC, respectively). The catalyst exhibits distinguished coulombic efficiency, high reaction rate and strong stability. Fig. 4a shows the synthetic schematic diagram of Se@Co_SA–HC, which is inserted selenium into carbon particles with hollow structure. The intermediate and free energy changes after structural optimization are observed in Fig. 4b. The transformation of are exothermic process, while are endothermic. In addition, the free energy of reaction of on Co–NC is lower than that of NC, which exhibits thermodynamically more favorable for the reduction of Se. By calculating the phase change barriers of on Co–NC and NC (Fig. 4c and d), Co atoms are the active center of the catalyst.⁸¹ In addition, SACs with uniformly dispersed active metal centers are highly unsaturated coordinated and exhibit quantum confinement effect,⁸² resulting in a high electron density distribution near the Fermi level, which contributes to the rapid electron transfer in the catalytic reaction process.^83,84 MXene anchored SACs by selectively etching hybrid A layers (Al and Cu) for eCO₂RR have been studied by DFT calculation. The monoatomic Cu with unsaturated electronic structure effectively reduces the energy barrier of the rate determining step. The rate determining step of the reaction path is the conversion of HCOOH* to the intermediate of adsorbed CHO*, which was helpful for eCO₂RR to CH₃OH. As shown in Fig. 4e, through the charge density difference of SACs–Cu–MXene, Cu has a positively charged Cu–O₃ structure and the designed reaction path. Based on DFT calculations (Fig. 4f), SACs can significantly reduce the reaction barrier of eCO₂RR.⁸⁵ Co@Cu SACs would also exhibit ideal performance in converting CO₂ to CH₃OH. The narrowed Co d-band stabilizes the adsorption of CO₂ on the surface and further enhances the bonding of COH* to the active site. While stabilizing the intermediate, it reduces the activation energy of COH* for further hydrogenation. It effectively prevents the C–C coupling and enhanced the selectivity of the reaction to CH₃OH.⁹

Fig. 4. (a) Schematic of Co/N doped hollow porous carbon. (b) The energy distribution of Li₂Se compounds/cluster reduction on NC/Co–NC carriers. (c) The charge density of SACs–Cu–MXene. The green/orange area indicates the increase/decrease in electrons. The reaction path for the eCO₂RR to CH₃OH. (d) The free energy change trend of eCO₂RR to CH₃OH (Reproduced from ref. 81 with permission from Springer Nature, copyright 2020; reproduced from ref. 85 with permission from American Chemical Society, copyright 2021).

Various efforts have been made on reducing the activation energy to promote eCO₂RR to CH₃OH. The normal reaction path for calculation is *CO₂ → *CO → *CHO → *CH₂O → *CH₂OH and *CH₃O → CH₃OH.⁸⁶ It could significantly enhance CO₂ protonation ability by reducing the distance between adjacent monomers and catalysis.⁸⁷ By DFT calculations, Ding et al. have discovered that TM (TM = Fe, Co, Ni, Cu, Ir and Pt) could effectively improve the catalytic activity and enhance the redistribution of electrons. Besides that, they could further activate the subsequent reaction intermediates, and finally selectively generate CH₃OH.⁸⁸

The instability safety limits the reserves of precious metal catalysts, which is the obstruction of industrial production, as well large-scale preparation. In addition, the stability is also an insurmountable gap for non-precious metals. Meanwhile, lots of attempts have been researched on metal catalysts, the main advantages of which catalysts are low overpotential, high current density, large specific surface area and less metal consumption. Cobalt phthalocyanine could be immobilized on carbon nanotubes to selectively catalyze eCO₂RR to CH₃OH. The electrocatalytic reaction followed domino process: CO₂ → CO by two electrons, CO → CH₃OH by four electrons.⁸⁹ Meanwhile, TM dimer doped on graphene with adjacent single vacancies could reduce overpotential and enhanced current density in eCO₂RR, which further improve the catalytic performance.² By constructing the network inter-perforation and self-supporting structure, the probability of single atoms at the interface is greatly increased, high current density can be achieved at a lower metal single atom content as a result, which has certain practical application prospects.⁹⁰ Besides that, new materials including 2D, 0D, and Janus material have been found to be promising as efficient electrocatalysts for eCO₂RR. For instance, Chen et al. have developed tricycloquinazoline based 2D c-MOFs successfully, which showed high selectivity toward CH₃OH with Faradaic efficiency up to 53.6%. The eCO₂RR performance is even better than Cu- and MOFs-based electrocatalysts.⁹¹

In recent years, ethanol with higher energy density as an ideal fuel has also been a research hotspot.⁹² Copper clusters supported by catalyst had a mixed charge state of Cu⁰ and Cu^x+. The coverage of *CO on the surface was increased, making it possible to couple C–C on the Cu_t⁰–Cu_b^x+ atomic interface, and finally generated C₂H₅OH.⁹³ The construction of dual active sites by loading single atom on carrier provided a new attempt for the design of the catalyst through the synergistic effect of the single-atom center and substrates.⁹⁴

At present, the design strategies of CO₂ to alcohols monatomic catalysts mainly include by SACs in covalent organic frameworks. However, due to the complicated intermediates and reaction pathways, the mechanism of eCO₂RR is still ambiguity. Meanwhile, varieties of by-products will be generated in the related reaction path. The reasonable design of the carrier, the regulation of the reaction conditions, and even the structure of the carrier are all significant effects on the selectivity of products.

2.3. eCO₂RR to acids

Formic acid is one of the basic organic chemical raw materials. It is widely used in the industry of pesticide, leather, rubber, dyestuff and so on. It also has the potential application for hydrogen storage and fuel cells.⁹⁵ In addition, formic acid is also an indispensable intermediate for the synthesis of higher alcohols. In fact, CO₂ and H₂ can generate formic acid relatively easily except the kinetic obstacles of reaction process that need to be resolved.⁹⁶ Therefore, it is very necessary to find suitable, efficient and stable catalysts.

As a very ideal electrocatalyst, indium (In) can efficiently reduce CO₂ to formic acid, but it cannot be further promoted because of the high potential and low FE. In SACs based on N-doped carbon matrix derived from the metal–organic framework have found to be excellent in this regard. The catalyst has well-dispersed In^δ+–N₄ atomic interface sites, which can efficiently catalyze eCO₂RR to formic acid. In^δ+–N₄ atomic interface has lower free energy for the formation of intermediate HCOO*, which is beneficial to improve the catalytic activity for eCO₂RR. It provides a feasible strategy for precise adjustment of indium catalysts at the atomic level for efficient eCO₂RR.⁹⁷ Meanwhile, TM (TM = Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu and Zn) SACs can stably be supported on porphyrin framework. This new catalyst designed strategy on eCO₂RR to acids provides a different enlightenment for experimental design.²¹ Hai et al. found that, for some MOFs, the selectivity of HCOOH depended on the adsorption energy difference between *HCOO and *COOH by DFT calculations.⁹⁸ Thus, the biggest obstacle is that HER would severely inhibit eCO₂RR.⁹⁹

Graphite carbonitride (g-C₃N₄) contains unique chemical composition, π-conjugated electronic structure and strong nucleophilic ability. Thus, it can be used as multifunctional catalyst in traditional organic catalytic reaction, such as green carrier and energy storage material. Based on first-principles calculations, systematic studies have been carried out on the geometric configuration, electronic structure and eCO₂RR catalytic performance of graphdiyne-supported alkali metals which had great potential as SACs for eCO₂RR. The conversion mechanism by constructing model of S-doped Cu has been calculated, which could effectively enhance the adsorption strength of *CO to further selectively generate HCOOH,¹⁰⁰ as well as Ru SACs on layered double hydroxide.¹⁰¹ low-coordination active center and oxygen vacancies could selectively generate HCOOH while inhibiting HER.¹⁰²

The heterogeneity of the catalyst support surface limits the further development of SACs, including synthesis and catalytic performance. In recent years, SACs have begun to construct active centers at the molecular level to improve the stability of the active centers and realize the “quasi-homogeneous” application. Zhang et al. successfully prepared a porous organic polymer containing pyridine-amide groups with Ir SACs on the porous organic polymer. It exhibits excellent catalytic activity in the CO₂RR to formic acid with excellent cycle stability. The SACs have the same catalytic reaction mechanism as the homogeneous catalysis process.¹⁰³ The solid micelle can increase the number of active sites of the carrier species, meanwhile, the reaction conditions are more tolerant, and the product can be easily separated.¹⁰⁴

2.4. eCO₂RR to other organics

In addition to C₁ and C₂, the products of eCO₂RR also included olefins,¹⁰⁵ alcohols, ketones,¹⁰⁶ aldehydes,¹⁰⁷ and even C₃ and C₄ products.¹⁰⁸ Compared with the C₁ product, the multi carbon products have more considerable energy density. Moreover, multi carbon products play an important role in the chemical industry.¹⁰⁹ Ethylene is a basic raw material for the synthesis of plastics, rubber, fiber and ethanol, which occupies an important position in the petrochemical industry. However, eCO₂RR to C₂₊ also faced with the obstacles of C–C coupling complex process, as well as the problem of adsorbate coverage on the surface.

CO₂ can be adsorbed on the surface with strong adsorption strength. Metal doped catalysts could show high selectivity and activity for the formation of CH₃OH.¹¹⁰Fig. 5a and b show the possible reaction pathway about C₁ and C₂.¹¹¹ Recently. The reaction mechanism of C₂H₄ was studied by constructing asymmetric.¹¹² Cu/C can significantly promote coupling and maintain a moderate binding energy between the intermediate and the carrier.¹¹³ For instance, Cu–C₃N₄ has been studied on the adsorption behavior of intermediates in the eCO₂RR process. The catalyst can effectively reduce the onset potential with excellent catalytic activity. The generation rate of C₂ is significantly better than that of Cu–NC.¹¹⁴ The excellent selectivity for C₂H₄ on the Cu-based MOFs for eCO₂RR has been reported by Yang et al. with isolated Cu–S motifs in MOF based precatalysts.¹¹⁵ At present, the C₃ and C₄ products have also begun to arouse widespread concern on multi-carbon products. Dismukes et al. synthesized nickel phosphide compounds, which were used to reduce CO₂ in aqueous solutions and studied the catalytic performance of eCO₂RR. The product selectivity has a significant change by increasing the phosphorus content. They proposed a reaction pathway for the efficient synthesis of multi-carbon products from CO₂ through formic acid and formaldehyde intermediates without CO intermediates.¹⁰⁸

Fig. 5. The eCO₂RR reaction pathways to (a) C₁ products and (b) C₂ products on Cu₂@CN catalyst (reproduced from ref. 111 with permission from John Wiley and Sons, copyright 2020).

Although lots of works have been done on eCO₂RR in recent years, most of them only stayed in the laboratory stage. How to find the catalyst with high efficiency and selectivity becomes a challenge.¹¹⁶ Besides, most of the current catalyst designs have limitations due to that CO₂RR has complicated reaction paths and produces intermediates.¹¹⁷ In addition, there is still lack of clear reaction mechanism. Many obstacles need to be overcome for SACs with excellent performance to be promoted and applied to industrial production.¹¹⁸ Above all, based on what has been achieved so far, we believe that there will be a qualitative breakthrough in the practicality of eCO₂RR in the future.

3. Photochemical conversion of CO₂

Investigating photochemical CO₂RR (pCO₂RR) with SACs is a pressing environmental issue due to the increasing concentration of CO₂ in the atmosphere. Since Inous et al. first reported pCO₂RR to organic compounds in gas–solid reaction systems, a great deal of researches have been devoted to the development of photocatalysts for pCO₂RR.¹¹ Photocatalysts mainly include the following types: (1) metal oxides, mainly composed of d⁰ structured (Ti⁴⁺, Zr⁴⁺, Nb⁵⁺, Ta⁵⁺, W⁶⁺, and Mo⁶⁺) and d¹⁰ (In³⁺, Ga³⁺, Ge⁴⁺, Sn⁴⁺, and Sb⁵⁺) metal cations,^12,119 such as TiO₂, ZrO₂. XTaO₃ (X = Li, Na, K), InVO₄, ZnGa₂O₄, etc.;^120–125 (2) metal sulfides such as XS (X = Cd, Zn);^126–129 (3) metal nitrides such as GaN,¹³⁰ TaON;¹³¹ (4) layered metal hydroxides (referred to as LDH), such as FeNi-LDH/g-C₃N;¹³² (5) metal organic framework materials (MOFs), such as MOF-525-Co;¹³³ (6) non-metallic semiconductors, such as g-C₃N₄,¹³⁴ and graphene.¹³⁵

Moreover, metal oxides,^16,136–138 MOFs,^139,140 C-based materials,^141,142 and LDH^143,144 are the common carriers for SACs. It has been reported that TiO₂,^145–147 CeO₂,^148,149 MOFs,^150–152 graphene,^153,154 graphitic carbon nitride,^155–157 LDHs^158,159 have shown high activity, stability and selectivity in pCO₂RR. The strong interaction between the single metal atom and the carriers is beneficial to improve the catalytic performance. Meanwhile, the application of pCO₂RR with SACs is an effective way to improve the rate of photocatalysis and achieve carbon emission reduction.

However, the high recombination rate of electrons and holes, the narrow wavelength-range of light to achieve the exited state, and the small specific surface area of conventional semiconductor photocatalysts limit the catalytic rate.^160,161 The photocatalytic performance of catalysts can be improved by co-catalysts and the noble metals, for example that Pt and Pd, are considered to be efficient co-catalysts.^162,163 The high cost and low abundance of precious metals limited the industrial application. Therefore, large specific surface area, stability, high selectivity, ability to inhibit recombination are the critical issues to design practical photocatalysts.^164–166

Single-atom photocatalysts are photoactive materials in which metals are present as SACs uniformly dispersed on the support. The catalytic performance of SACs depends largely on the interaction between the support and metal.¹⁶⁷ For single atom photocatalysts, the introduction of SACs can effectively inhibit the electron–hole recombination, as well as increase the activity and selectivity.^168,169 The favorable SACs dispersion effect and prevention of metal SACs aggregation play important roles in the final catalytic activity of monoatomic photocatalysts. CO₂RR with SACs is a promising strategy for the conversion of CO₂ to hydrocarbon fuels such as CO, CH₄, and CH₃OH.^13,168,170

The pCO₂RR requires the following steps (Fig. 6): (1) the photocatalyst is irradiated by light. It leads to the excitation of electron from valence band to conduction band, leaving a hole in the valence band, which is the photo-electron/hole pair;¹⁷¹ (2) the photogenerated electrons and holes diffuse and separate. The electrons/holes transfer to the reduction/oxidation sites, respectively; (3) CO₂ is adsorbed onto the surface; (4) a redox reaction occurs on the photocatalyst surface, where the electrons reduce CO₂ to CO,^172,173 CH₄,^169,174,175 HCOOH,^176–178 CH₃OH^179–181 or other hydrocarbons,^170,182,183 and the holes oxidize water to molecular O₂ by cavities;^152,154,184 (5) products desorbed from the catalyst surface. The pCO₂RR and potentials are also exhibited. Fig. 7 shows the band edges commonly used in photocatalysts.¹⁸⁵ In this section, we summarize the theoretical studies of SACs in pCO₂RR, including CO₂ to CO, methane, and CH₃OH.

Fig. 6. The mechanism and steps of pCO₂RR. Including the electrons and reduction potential required for the pCO₂RR to different products with respect to the standard hydrogen potential NHE at pH 7.

Fig. 7. CB and VB potentials of common semiconductors with CO₂ and water redox couples at PH 7 (reproduced from ref. 185 with permission from John Wiley and Sons, copyright 2018).

3.1. pCO₂RR to CO

Based on the thermodynamic principle of proton-assisted multi-electron transfer to reduce CO₂, the specific pathway for CO₂ reduction to CO is: CO₂ + 2H⁺ + 2e⁻ → CO + H₂O,^168,184,186 for pCO₂RR relative to the Normal Hydrogen Electrode (NHE) in aqueous solution.¹⁸⁷ The potential level of the conductive band of the photocatalyst determines the production of pCO₂RR.¹⁸⁸ In order to overcome the slow thermodynamic process and promote the reaction rate, the potential (vs. NHE) of the conduction band needs to be more negative than the reduction potential of CO₂. The energy band structure of semiconductor photocatalysts is extremely important for the catalytic efficiency. How to abbreviate the residence time of CO on the catalyst surface and accelerate the desorption from the catalyst surface into the gas phase are the key points to improve the catalyst selectivity.^189–191 The energy band structure of conventional semiconductor catalysts is far from ideal. In recent years, many researches have been conducted to improve the photogenerated electron–hole separation efficiency of photocatalysts by modifying the semiconductors to increase the light absorption rate and enhance the CO₂ adsorption capacity.

The combination of monatomic transition metals and polymeric carbon nitride has been an effective strategy for photocatalytic reactions. Li et al. reported the preparation of copper carbon nitride nanocatalysts (Cu–CCN) based on the doping of monatomic Cu into CN by molten salt and reflux methods to promote the photoreduction of CO₂ to CO.¹⁹² Density functional theory calculations further clarified the mechanism of the pCO₂RR process (Fig. 8a), where the CO₂ adsorbed on the surface of the Cu–CCN catalyst was hydrogenated to *HCOO, *CO, successively. And CO desorbed from the surface, finally. Wang et al. analyzed the mechanism of CN-loaded Cu single-atom pCO₂RR using DFT and found that the active center in the form of C–Cu–N₂ not only promoted charge transfer but also reduced the energy potential energy for pCO₂RR (Fig. 8b).¹⁶⁵ Meanwhile, Ji et al. reported an atomic confinement and coordination design strategy applied to design various metal monatomic catalysts.¹⁹³ Er monatomic catalysts (Er/CN-NT) loaded on carbon nitride nanotubes with tunable metal dispersion density were synthesized for pCO₂RR. According to the computational hydrogen electrode (CHE) model, it is confirmed that the favorable production is CO, instead of CH₄ on rare-earth erbium atom catalysts.

Fig. 8. (a) Schematic of the tentative photocatalytic mechanisms and reaction pathway of Cu–CCN for pCO₂RR. (b) Schematic illustration of Cu/CN hybrid for pCO₂RR. (c) Photocatalytic selective CO₂RR on Ni-TpBpy. (d) Proposed reaction mechanism for the photoconversion of CO₂ into CO on Ni-TpBpy (reproduced from ref. 192 with permission from American Chemical Society, copyright 2020; reproduced from ref. 165 with permission from American Chemical Society, copyright 2020; reproduced from ref. 194 with permission from American Chemical Society, copyright 2019).

In addition to the above-mentioned design photocatalyst strategies, we have continuously developed single-atom photocatalyst design strategies that favor adjustable metal atom coordination and demonstrated their catalytic effects on the pCO₂RR to CO. Zhong et al. synthesized Ni-rich single-atom 2,2′-bipyridine covalent organic frameworks of photocatalyst (Ni-TpBpy) (Fig. 8c and d).¹⁹⁴ Ni-TpBpy was effective in improving the selectivity of CO generation in aqueous solution. Ni atom was stabilized by the dipyridine coordination unit based on the special structure. The hydrogen bonding between ketone node and Ni–CO₂ reaction intermediate promoted the pCO₂RR preferential to H₂ formation. Meanwhile, the adsorption of CO₂ by Ni-TpBpy was greatly improved due to the Lewis acid–base interaction between the loaded Ni ions and absorbed CO₂ molecules. DFT calculations used to investigate the mechanism for pCO₂RR confirmed that Ni-TpBpy could reduce the CO₂ molecules and activate Ni site to Ni–COOH, which would combine with H⁺ to Ni–CO intermediate, followed by CO desorption from Ni–CO. Oxygen vacancies, as common defects in oxides or hydrogens carbide surfaces, has a great impact on pCO₂RR.^195,196 For instance, CO production rate of Co–Bi₃O₄Br was approximately 4 and 32 times than atomic layer Bi₃O₄Br and bulk Bi₃O₄Br.¹⁶⁴ Ni atoms could occupy the oxygen vacancies of ZrO₂, efficiently achieved pCO₂RR to CO, which is 6 and 40 times by that of defective ZrO₂ and complete ZrO₂, respectively.⁶³ Chen et al. have studied the adjustable catalytic performance and the modulation mechanism of Ni²⁺ doped NH₂-MIL-125-Ti (NH₂-MIL-125-Ni_x/Ti) with different Ni²⁺/Ti⁴⁺ molar ratio (x = 0.5–1.5%) via an in situ doping method. They found that Ni around O atoms of Ti oxo clusters of NH₂-MIL-125 can change the electronic structure of the clusters thus lead to the improvements of electron transfer and charge separation. The highest CO₂ conversion rate with 98.6% CO selectivity was exhibited with the molar ratio 1%.¹⁹⁷

So far, researchers have focus on using atomic confinement strategy, coordination design strategy, and defect engineering strategy to increase the active sites with CN, metal–organic backbone, and photocatalysts of monatomic metal to achieve improving photocatalytic activity. Thus, the pCO₂RR still have the obstacle in accelerating the rate of CO desorption, which is also the key to improve photocatalytic selectivity. So, it is also challenging to improve the CO₂ photoreduction activity and product selectivity.

3.2. pCO₂RR to CH₄

CH₄ is the main component of natural gas. Due to the limited petroleum resources, it will be a high-quality gas candidate fuel. pCO₂RR to CH₄ requires 8-electron reduction: CO₂ + 8H⁺ + 8e⁻ → CH₄ + 2H₂O.¹⁹⁸ The CO₂ molecule is very inert, which limits the catalytic conversion. At the same time, the products of CH₄ are frequently accompanied by CO and H₂, so that the efficiency and selectivity will be influenced. Thus, how to improve the quantum efficiency and increase the selectivity of pCO₂RR become the key issue for photocatalysis.

Zhang et al. realized the modular optimization of the metal–organic framework by injecting monoatomic Co SACs on porphyrin,¹³³ so as to synthesize the catalyst with atom dispersion (Fig. 9a). The catalyst significantly improves the electron–hole separation efficiency of the porphyrin unit and the CO₂ adsorption capacity, which realize the directional migration of photo-excitons from the porphyrin to Co center. It provides long-lived electrons for CO₂ molecules adsorbed on the Co center. The addition of unsaturated Co SACs achieves the enhancement of pCO₂RR, thus, greatly improves the selectivity of CH₄.

Fig. 9. (a) Schematic representation for construction of Cu-based ZIF catalyst and representation of pCO₂RR; (b) illustration of LSPR-mediated charge excitation mechanisms; (c) pCO₂RR mechanisms of Ag-HMO; (d)Proposed mechanisms for PEC pCO₂RR on the CeO₂ NPs/CuO NPs composite catalyst (reproduced from ref. 133 with permission from MDPI, copyright 2018; reproduced from ref. 199 with permission from Springer Nature, copyright 2016; reproduced from ref. 200 with permission from Elsevier, copyright 2019; reproduced from ref. 201 with permission from Springer Nature, copyright 2020).

Due to the heat generated by surface plasmon resonance of noble metals, the excited electrons can be easily transferred into the conduction band of the semiconductor,²⁰² which could facilitate the separation of electrons and holes to increase pCO₂RR to CH₄. Liu et al. reported Au–Cu alloy catalysts which was adjusted the structure of co-catalysts and improved the electron capture capacity.²⁰³ Meanwhile, the A₃Bi₂I₉ (A⁺ = Rb⁺, Cs⁺, CH₃NH³⁺) were synthesized by Bhosale et al.,²⁰⁴ which has shown excellent photocatalytic activity for pCO₂RR at the gas–solid interface. Besides, precious metal catalysts such as Ag, Au, and Pt loaded on metal-catalysts as SACs can effectively increase photoactivity. Ag nanoparticles were widely used to enhance photocatalytic reaction rates due to localized surface plasmon resonance (Fig. 9b). The “hot electrons” excited from surface plasmon resonance will be injected into the conduction band of the metal–semiconductor, thus enhancing photocatalytic activity.

Hydrothermal and impregnation methods have been used to develop Ag SACs supported on hollandite manganese dioxides (Ag-HMO) for pCO₂RR to CH₄.²⁰⁰ During the calcination process, the Ag atoms enter the HMO lattice tunnel through the anti-Oswald maturation mechanism to promote the formation of single-atom Ag chains, while effectively inhibiting Ag agglomeration to improve the catalytic efficiency. Recently, the heterostructured photocatalysts have been considered as an effective strategy to promote the spatial separation of photogenerated electrons and holes. The catalyst with NiMOF/CN heterojunction structure exhibited a photocatalytic activity 18 times larger than pure carbon–nitrogen.¹⁵² The Pd–Au hetero junction model co-catalyst have also been prepared,¹⁷⁵ in which electron-deficient Pd massively inhibited the competition for H₂ generation and enhanced the electron–hole separation, while the electron-rich Au promote the 8-electron reduction of CO₂ to CH₄ at the Pd–Au interface.

Even though the synthetic strategies are continuously innovative, CO and H₂ are still formed along with the pCO₂RR to CH₄, which reduces the selectivity of CH₄. The key for improving the pCO₂RR activity and CH₄ selectivity is preventing or reducing this phenomenon. Otherwise, this aspect of research would play a great potential in the field of theoretical studies of synthetic photocatalysts for pCO₂RR, making it a viable technology for a carbon-neutral future.

3.3. pCO₂RR to CH₃OH

The two main reaction systems for pCO₂RR by water reducing agent are liquid- and gas-phase systems.^205,206 CH₃OH as the reduced liquid product generally exists in liquid phase system. In addition, photocatalytic reduction of H₂O produces relatively strong hydrogen side reactions, which can reduce the selectivity of the products. Therefore, the effects of different catalytic systems on the reaction products should be taken into account. Kavil et al. synthesized Cu/C–TiO₂ nanocatalysts for pCO₂RR to CH₃OH in seawater using sol–gel method by co-doping Cu and C into TiO₂.²⁰⁷ The pathway followed with thermodynamic principle is:

nTiO₂ + 6hν → 6e_cb⁻ + 6h_νb⁺ R-1

6H⁺ + 6e_cb⁻ → 6H˙ R-2

CO₂ + 6H˙ → CH₃OH + H₂O R-3

6OH⁻ + 6h_νb⁺ → 3/2 O₂ + 3H₂O R-4

The overall reaction is the following:

2CO₂ + 4H₂O → 2CH₃OH + 3O₂ R-5

It is shown that Cu doping into TiO₂ can effectively trap carriers and inhibit electron–hole recombination, thus effectively improve the efficiency and selectivity for CH₃OH generation. The effects of factors on the pCO₂RR by Cu/C–TiO₂ were investigated. Pan et al. used CeO₂/CuO p–n junction photocatalyst with Cu cocatalysts for pCO₂RR to CH₃OH (Fig. 9d).²⁰¹ Under light irradiation, the photoexcited electrons in CuO transfer to cerium oxide because the CB of cerium oxide is more positive than that of copper oxide, while the holes (h⁺) generated in the VB of CeO₂ under light irradiation will migrate to the VB of CuO. By electron–hole exchange between CuO and CeO₂, electron–hole pair complexation is suppressed and the effectiveness of electron reduction of CO₂ on cerium dioxide nanoparticles is increased. The combination of p-type copper oxide and n-type cerium oxide make the carrier in the cerium oxide nanoparticles/CuO nanoparticles catalyst concentration is 108 times higher than that of the copper oxide nanoparticle catalyst, which greatly improves pCO₂RR performance. Meanwhile simple catalyst modification (i.e. cerium dioxide electrodeposition) can significantly increase the carriers' concentration and electron availability, thus chemically reducing CO₂ plasma to valuable commercial chemicals and providing a theoretical model for pCO₂RR.

Cu-based catalysts have been widely applied for pCO₂RR to CH₃OH by different design strategies to increase the active sites and improve the catalytic efficiency. Meanwhile, MOFs are considered as potential light-driven catalysts, which have imidazolium or aromatic carboxylate-based linkers.^208,209 Goyal et al. developed a catalyst (CuZrlm) with Cu combined with zirconium oxide loaded on imidazoline skeletons,¹³ which can effectively improve the photocatalytic performance of imidazolines as potential catalysts for pCO₂RR to CH₃OH. The authors stated that the presence of Cu²⁺ oxidation state and uniform dispersion of the active metal are the main reasons for the improved yield, because Zr introduction can distort the symmetry of the Cu²⁺ oxidation state thereby stimulating charge transfer.

Although pCO₂RR research has made significant breakthroughs in artificial action and photosynthesis, the industrialization is still limited by high costs, carrier limitations, and imperfect analytical characterization techniques. To further improve the performance of pCO₂RR, we should work on finding new carrier materials and designing different photocatalyst structures, as well as proposing new pathways for pCO₂RR to design highly selective and stable photocatalysts.

4. Conclusions and outlook

In this review, we focus on the theoretical studies of electro- and photo-chemical CO₂RR aspects with SACs. The conversion of CO₂ to CO, CH₄, alcohols, acids and other organics have been discussed in detail. The construction of SACs on different unique support, along with the detail catalytic mechanism and improvement of product selectivity were involved.

In spite of gratifying results on the CO₂RR, the challenges of how to improve the activity and selectivity of CO₂RR products, as well as to find green mild conversion environment for large-scale industrial implementation are still the issues need to be solved in the future. The influences of electrode morphology, surface structure and reaction conditions on CO₂RR are still required to be studied to improve the stability of catalysts. For the theoretical calculations, a multi-scale computational model covering all aspects, rather than a collection of theoretical models applicable to specific cases is also needed to achieve.

Author contributions

Q. Yu supervised the project and organized the collaboration. L. Jiang, Q Yang co-wrote the manuscript with input from all authors. Z. Xia and X. Yu summarized the electro-CO₂RR, M. Zhao and Q. Shi summarized the photo-CO₂RR.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.