Single-Atom Catalysts for the Electro-Reduction of CO₂ to Syngas with a Tunable CO/H₂ Ratio: A Review

Abstract

Nowadays, transition towards green chemistry is becoming imperative. In this scenario, an attractive perspective consists in the generation of CO through the electrochemical reduction of CO₂ under ambient conditions. This approach allows storage of the electrical energy from intermittent renewable sources in the form of chemical bonds, and simultaneously reduces greenhouse gas emissions, giving carbon a second chance of life. However, most catalysts adopted for this process, i.e., noble metal-based nanoparticles, still have several issues (high costs, low current densities, high overpotentials), and in the view of generating syngas through co-electrolysis of H₂O and CO₂, do not enable a widely tunable CO/H₂ ratio. Single-atom catalysts with N-doped carbon supports have been recently introduced to face these challenges. The following review aims to answer the demand for an extended and exhaustive analysis of the metal single-atom catalysts thus far explored for the electro-reduction of CO₂ in aqueous electrolyte solution. Moreover, focus will be placed on the objective of generating a syngas with a tunable CO/H₂ ratio. Eventually, the advantages of single-atom catalysts over their noble metal-based nano-sized counterparts will be identified along with future perspectives, also in the view of a rapid and feasible scaling-up.

1. Introduction

Since the last century, the global average atmospheric concentration of CO₂ has been dramatically increasing at an exponential rate, reaching, at present, over 417 ppm, which is about 50% higher than the level at the beginning of the Industrial Revolution. The dramatic increase of CO₂ in the atmosphere is a major consequence of the growth in the consumption of fossil resources. Among the emitted gases, CO₂ has a significant detrimental impact on the environment, accounting for over 80% of all emissions [1,2,3,4]. Since dependence on non-renewable sources is still necessary, it is essential to overcome the problems related to the high amount of CO₂ in the atmosphere. Currently, there are two major approaches to overcome these problems [5]. The first consists in the capture and geological storage of CO₂ in underground reservoirs. However, the feasibility of this approach is significantly hampered by safety, space, and cost issues. The second technology is related to the valorization of CO₂ into low-molecular-weight chemicals [6] or fuels [7] The main methods for the chemical conversion of CO₂ are four: thermochemical, electrochemical, photoelectrochemical, and photocatalytic. Among them, the electrochemical reduction of carbon dioxide offers some advantages over the other methods [8,9]. Firstly, the electro-reduction of CO₂ (defined in literature as CO₂RR) can occur under mild conditions, such as atmospheric pressure and room temperature. Furthermore, this process is the most sustainable, since the electricity needed can be supplied through a renewable resource, such as sunlight, without generating further CO₂, and because the co-reactant is water, which is an abundantly, economically, and easily available natural resource. In addition, the process can be accurately controlled by varying, for instance, the electrode potential, hence higher conversion efficiencies can be more easily obtained. Moreover, the electrochemical system is modular and compact and can be easily scaled up. It is clear, therefore, that being able to efficiently turn CO₂ into value-added molecules is by far one of the most valuable, effective, and feasible methods [10]. A scheme of the circular economy involving CO₂ capture and CO₂ electro-reduction to chemicals is represented in Figure 1. As a consequence of the above-mentioned facts, the electro-reduction of carbon dioxide to value-added molecules has been extensively explored in many studies. Although this route is highly promising, several challenges have still to be faced before its large-scale applicability becomes a certainty. Apart from the high stability of the CO₂ molecule under ambient conditions [6], the major limitation of the process is the low performance of the electrocatalysts so far adopted, in particular their low catalytic activity, poor selectivity, and poor stability over time [11].

As will be explained later in this review, the reduction of carbon dioxide in an aqueous solution to carbon monoxide (CO₂RR to CO) is the most likely reaction to occur, hence the most explored one in the current literature. Furthermore, by properly tuning this reaction and the simultaneous cathodic water-splitting reaction to hydrogen (known as HER) in the same process, CO₂ and H₂O can be converted into valuable and “greener” syngas.

Indeed, around 1.7 × 10⁸ tons of syngas are globally produced every year, primarily through steam methane reforming and coal gasification, with a global market that is expected to grow to 72.4 billion US dollars by 2022 at a compound annual growth rate of 9.1% [12,13]. Depending on its CO/H₂ ratio, a synthesis gas could be used as important feedstock to several industrial processes, such as methanol synthesis (when the CO/H₂ ratio is 0.5) and Fischer-Tropsch synthesis (for instance, when the CO/H₂ ratio is equal to or even less than 0.5) [14,15]. Although cost-effective, it is worth noticing that current industrial processes for syngas generation are fossil-fuel-dependent, hence not “optimized” from an environmental point of view. In addition, the electrochemical co-production of CO and H₂ can lead to syngas with a tunable CO/H₂ ratio, hence ensuring the integration of a sustainable process into production lines of different chemicals. In order to produce syngas, the CO₂RR must be selective towards CO. Nanostructured catalysts consisting of noble metals, such as Au and Ag nanoparticles (NPs), have proved to be highly active and selective towards the CO production from the co-electrolysis of CO₂ and H₂O, leading to syngas with a high CO/H₂ ratio [16,17,18]. However, the high costs and/or the low natural availability of these noble metals significantly hinder their application on an industrial scale. Gold-based catalysts are clearly highly expensive (gold cost is around 59 US dollars per gram). As for silver, although less expensive than Au (its current price is about 0.79 US dollars per gram), it is a finite resource whose availability is due to further decrease in the years to come. In addition, high electrocatalytic activity on these metals corresponds again to high overpotentials, since both Au and Ag show a low COOH* intermediate bond strength along with a low CO adsorption strength in virtue of the linear relationship between the adsorption energies of the CO₂ intermediates and CO on transition metal surfaces [19]. Furthermore, both Au- and Ag-based electrocatalysts are extremely selective for CO generation at low overpotentials (instead, at extremely high overpotentials, HER become more favored than CO₂RR on these metals). This high CO selectivity at low overpotential can become a drawback when aiming at producing syngas with a more tunable CO/H₂ ratio, for instance, with a CO/H₂ ratio lower than 70–60%.

In order to overcome these issues, i.e., to attain improved CO₂ reduction and/or co-electrolysis of CO₂ and H₂O, an effective alternative to these noble-metal-based nanoparticles is the use of metal-based single-atom catalysts (SACs). Recently, it has been demonstrated that by further reducing the size of nanoparticles to the sub-nanometer range, nanocluster catalysts can be obtained [20,21]. By virtue of the strong quantum confinement effects at this size, these catalysts exhibit unique properties, such as several valence states and discrete electronic configurations, which strongly depend on the number of atoms constituting the cluster. These properties, along with the increased number of under-coordinated metal atoms, i.e., the increased amount of active catalytic sites, make nanocluster catalysts highly active towards a wide range of chemical reactions, such as CO₂RR and HER [22,23,24,25]. The ideal case, corresponding to the minimum size and the maximum metal-utilization of the nanocluster is represented by the metal single atom: at this size, individual metal atoms become accessible, and therefore active for catalysis [26]. More generally, in order to guarantee high dispersion and to avoid agglomeration, nanoparticle catalysts are typically dispersed on a support surface. Furthermore, the interaction between nanoparticles and support can lead to a modification of the electronic environment of the metal particles, hence increasing the interactions between catalyst surface and reaction intermediates and changing the catalytic activity of the former [27]. It has been observed that when the size of the nanoparticles has been reduced to the atomic range, these metal-support interactions are maximized, which leads to the formation of novel catalytic active sites, the creation of new reaction pathways, and ultimately, the elevation of catalytic performance [28,29]. Consequently, the catalytic activity, selectivity, and stability of nanostructured catalysts, and above all, of SACs, can be controlled by choosing the right supports and by tuning the coordination environment.

The synthesis of stable platinum single atoms efficiently supported on the surfaces of iron oxide nanocrystallites has been successfully performed by Qiao et al. in 2011 [25]. Since then, SACs have drawn increasing attention due to their high catalytic performance, with a high reduction in metal loading compared with their nanoparticle counterparts. SACs have been adopted for hydrogen evolution reaction (HER), oxygen reduction reaction (ORR), CO oxidation, and organic synthesis [30,31]. However, it is only recently, with the development of more specific characterization techniques, such as high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), which enables the obtainment of images with atomic resolution, and X-ray absorption fine structure spectroscopy (XAFS), which provides local atomic structural information (oxidation state, coordination number, bond length, atomic species, etc.), that SACs have begun to be considered for CO₂RR catalysis, and above all, for the production of syngas. Therefore, the following review aims to respond to the necessity of an extended and exhaustive analysis of the metal single-atom catalysts thus far adopted for the electro-reduction of carbon dioxide in aqueous solutions to produce CO₂RR. In particular, metal single atoms in the form of M–N–C catalysts (with M: metal single atom; N: nitrogen atom; C: carbonaceous support) are analyzed, as they are the most frequently adopted by virtue of their high efficiency in these reactions. Indeed, N atoms (typically embedded in carbon support by forming several chemical functionalities, such as pyrrolic, graphitic, and pyridinic) contribute to increasing the activity and stability of the metal single atoms by coordinating with them to form peculiar active centers. On the other hand, the contribution to the entire activity and the stability given by carbonaceous supports, such as graphene-, carbon nanotube- and other carbon precursor-(pyrrole, dicyandiamide, glucose etc.) derived supports is undeniable [32]. Indeed, these supports can stably incorporate single metal atoms within their structure through their vacancies and defects, avoiding their agglomeration and possess intrinsic conductivity, which makes them suitable for electrochemical reactions, as they are able to enhance electron transfer [33,34]. In particular, several studies reported the beneficial effect of the presence of carbon supports, which can control the adsorption energy on metal centers through local electron density modification due to electrons donation or withdrawal [35,36,37]. Other types of carbon supports, namely metal-organic framework- (MOF) derived supports, guarantee a high stabilization effect, as well as homogenous distribution of active sites, due to their intrinsic extremely high porosity, i.e., even higher surface area [38].

Moreover, in this review, particular attention is paid to the objective of forming syngas with a highly tunable CO/H₂ ratio. Ultimately, the advantages of SACs over their noble-metal-based NP counterparts are identified, along with future perspectives in view of rapid scale-up. In Figure 2, the structure of the review is reported.

2. Working Principles of the CO₂RR to CO and to Syngas in Aqueous Solution

As shown in

Table 1. Standard reduction potential (E°) at 25 °C, 1 bar and pH = 7 for the CO₂ reaction reductions to several chemicals and the HER.

Chemical Reaction	E°(V vs. RHE)
${\mathrm{H}}_2\mathrm{O}+2{\mathrm{e}}^{-}\to 2{\mathrm{OH}}^{-}+{\mathrm{H}}_2$(HER)	0.00
${\mathrm{CO}}_2+2{\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\to \mathrm{CO}+{\mathrm{H}}_2{\mathrm{O}}^{}$	−0.11
${\mathrm{CO}}_2+2{\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\to {\mathrm{HCOOH}}^{}$	−0.25
${\mathrm{CO}}_2+4{\mathrm{H}}^{+}+4{\mathrm{e}}^{-}\to {\mathrm{H}}_2\mathrm{CO}+{\mathrm{H}}_2{\mathrm{O}}^{}$	−0.07
${\mathrm{CO}}_2+6{\mathrm{H}}^{+}+6{\mathrm{e}}^{-}\to {\mathrm{CH}}_3\mathrm{OH}+{\mathrm{H}}_2{\mathrm{O}}^{}$	0.02
${\mathrm{CO}}_2+8{\mathrm{H}}^{+}+8{\mathrm{e}}^{-}\to {\mathrm{CH}}_4+2{\mathrm{H}}_2{\mathrm{O}}^{}$	0.17
${\mathrm{CO}}_2+12{\mathrm{H}}^{+}+12{\mathrm{e}}^{-}\to {\mathrm{C}}_2{\mathrm{H}}_4+4{\mathrm{H}}_2{\mathrm{O}}^{}$	0.06

, carbon dioxide can be reduced in an aqueous solution under ambient conditions to produce carbon monoxide, formic acid, methane, or the other listed chemicals, by applying a certain theoretical thermodynamic cell potential [39,40]. Among all the chemicals obtained from the CO₂ electro-reduction, carbon monoxide is the most likely and easiest to be produced from the thermodynamic point of view, since only two electrons are involved in the reaction, even though an additional potential for its formation is always required and it mainly depends on the type of catalyst chosen for the reaction [41]. As shown in Equation (1), the rate-limiting step of the reaction is the formation of the CO₂⁻* radical. After its formation on metals such as Au and Ag, a H⁺ proton attacks the oxygen-end of the radical to form a COOH* intermediate (* indicates the adsorption site) which adsorbs on the catalyst surface through its C atom; see Equation (2). The addition of another proton and electron to COOH* leads to CO* intermediate, as can be seen in Equation (3), which can desorb (see Equation (4)) unless its binding strength to the catalyst surface is too high [42,43].

CO₂ + e⁻ → CO₂⁻*

(1)

CO₂⁻* + H⁺ → COOH^*

(2)

COOH^* + H⁺ + e⁻ → CO*

(3)

CO* → CO

(4)

On the other hand, occurring at a theoretical reduction potential close to zero, the hydrogen evolution reaction (HER) is in inevitable competition with the CO₂RR, hence lowering the selectivity to CO and other chemical products. Many past studies have indeed considered the HER as a parasitic reaction to be suppressed. However, an interesting alternative consists of not suppressing the HER, but favoring the electrochemical co-production of CO and H₂ by obtaining syngas by means of an appropriate catalyst.

3. Electrochemical Measurement System

A typical electrochemical measurement system for evaluating the catalyst performance towards HER and CO₂RR to CO can be considered as made up of three main components:

• An electrochemical workstation, which provides electric power to the cell and records the corresponding electrochemical parameters.
• An electrochemical reaction cell. The most widely adopted electrochemical system is the so-called batch-type, H-type cell [44], whose anodic and cathodic compartments are separated by an ionic membrane, and in which the gas-phase chemicals diffuse to the electrode surface through an aqueous electrolyte (as shown in Figure 3). It is worth noticing that, when it comes to liquid electrolyte systems, the selectivity towards either CO or H₂ can also be achieved by changing the type of electrolyte. As can be seen from most studies, the most adopted electrolyte for obtaining a high CO selectivity is KHCO₃, since CO₂ replenishment at the electrode is favored by it [45]. On the other hand, the main limitation of an H-type configuration is that the mass transportation of gas species is rather low [46,47]. In order to enhance mass transportation, continuous-flow reactors have recently been demonstrated as an effective solution. These flow cells can deliver reactants to the electrodes and remove products from them continuously, hence increasing the mass transport, and are typically divided into two main types: membrane reactors and microfluidic reactors (see Figure 4) [48].
• Online/outline gas chromatography (GC), which provides a quantitative analysis of the gaseous products, and nuclear magnetic resonance (NMR), typically used to analyze possible liquid products.

4. Single-Atom M–N–C Catalysts for CO₂RR to CO and Electrochemical Production of Syngas

In the following paragraphs, the SACs in the form of M–N–C catalysts reported in the literature of the last few years were summarized with a particular focus on nickel, cobalt and iron-based SACs, as they are the most typically adopted metals. In particular, most of them were studied for the CO₂RR to CO, since only recently attention has been placed on taking advantage of both HER and CO₂RR to CO for producing syngas. All the catalysts were tested at ambient temperature and pressure, and in most of the studies, a KHCO₃ aqueous solution was adopted as the electrolyte.

Table 2. Evaluation parameters of the most recent metal SACs for the CO and syngas electro-production from CO₂ in aqueous electrolyte solution at ambient pressure and temperature.

Catalyst	Transition Metal Loading [wt% or at%]	Onset Potential (vs. RHE) [V]	Cathodic Potential (vs. RHE)[V]	Corresponding Maximum CO Faradaic Efficiency [%]	Corresponding Total Current Density and Maximum Total Current Density [mA/cm2]	Durability	Range of CO/H2 Ratio (When Evaluated)	Reference
N–C	-	∼−0.50	−0.52	∼91	0.47	-	-	[49]
Ni–N–C	Ni: 0.24–0.46 wt%	∼−0.40	−0.67	93	3.90	No significant current loss within 12 h of test at −0.67 V	-	[49]
Ni–N–C	Ni: 2.83 wt%	∼−0.45	−0.75	∼96	∼10 (maximum ∼25 at −0.95 V)	No significant current loss within 9.5 h of test at −0.75 V	-	[50]
Ni–N–Gr	Ni: 2.20 wt%	−0.50	−0.70	∼90	∼0.50 (maximum ∼4 at −2.20 V)	No significant current loss within 5 h of test at −0.65 V	-	[51]
NiN–GS	-	−0.35	−0.82	93.20	∼35 (maximum ∼65 at −0.95 V) [values in mA/mg]	No significant current loss within ∼20 h of test at −0.70 V	-	[52]
Ni–NG	Ni: 0.44 at%	−0.31	−0.62	95	11 (maximum ∼28 at −0.90 V)	No significant current loss within∼20 h of test at −0.64 V (overpotential)	-	[53]
SA–Ni/N–CS	Ni: ∼1.08 wt%	∼−0.40	−0.80	95.10	∼9 (maximum ∼32 at −1.20 V)	No significant current loss within∼25 h of test at −0.80 V	-	[54]
Ni–N4–C	Ni: 1.41 wt%	−0.40	−0.81	99	28.60 (maximum 36.20 at −0.91 V)	About 25% current loss within 30 h of test at −0.81 V	-	[55]
Ni SACs/N–C	Ni: 1.53 wt%	−0.57	−0.90	72	∼10 (maximum 20 at −1.20 V)	No significant current loss within 60 h of test at −1 V	-	[56]
Ni–N–MEGO	Ni: 6.90 wt%	−0.29	−0.70	92.10	∼17.50 (maximum ∼25 at −0.76 V)	29.1% current loss within 21 h of test at −0.55 V	-	[57]
NiSA–N–CNT	Ni: 20.7 wt%	−0.27	−0.70	91	(maximum ∼36 at −0.65 V)	12% current loss within 12 h of test	-	[58]
NiCoSA–N–CNT	Ni + Co: 13.20 wt%	−0.27	−0.70	∼30	-	-	-	[58]
NiPtSA–N–CNT	-	−0.27	−0.70	∼5	-	-	-	[58]
A–Ni–NSG	Ni: 2.80 wt%	∼−0.15	−0.50	97	∼1.40 (maximum ∼110 at −1.10 V)	No significant current loss within 100 h of test at −0.61 V	-	[59]
Ni–CTF	Ni: 0.14–0.17 at%	−0.48	−0.90	∼95	∼1 (maximum 4.50 at −1.20 V)	About 50% current loss within 3 h of test at −0.65 V	-	[60]
NiSAC/NiNP CNRs	Ni: ∼20 wt%	−0.40	−0.60–−0.80	95.2/10	∼13–∼27.5 in the range −0.60 V–−0.80 V	No significant current loss within 8 h of test	0.11 − 19 by varying the Ni–N sites/ Ni NPs ratio from 0.20 to 1	[61]
Ni/NCNTs-50	-	−0.32	−0.6 − −1	∼100	∼5 − ∼35 (maximum ∼50 at −1.20 V)	No significant current loss within 20 h of test	-	[62]
NiSA/PCFM (GDE cell, electrolyte flow cell)	Ni: 0.85 wt%	∼−0.20	−1	∼88	∼180 (maximum ∼360 at −1.35 V)	No significant current loss within 120 h of test	-	[63]
NI–N–C (GDE, electrolyte flow cell)	Ni: 10.20 wt%	-	−0.70	∼90	(maximum ∼700 at −2.20 V)	No significant CO FE loss within 20 h of test at a JCO of 200 mA/cm2	-	[64]
Co–N2	Co: 0.25 wt%	∼−0.20	−0.68	95	∼22.5 (maximum ∼45 at −0.90 V)	No significant current loss within 60 h at −0.63 V	-	[65]
Co–HNC	Co: 3.40 wt%	∼−0.25	Above −0.80	∼35	Above 10 (maximum ∼25 at −1.40 V)	No significant current loss within 24 h at −0.80 V and −1.40 V	It remains stable at 0.5 in the potential range −0.7 − −1 V.	[66]
Co–N5/HNPCSs	Co: 3.54 wt%	−0.30	−0.79	99.40	∼5 (maximum ∼18 at −0.90 V)	No significant current loss within 10 h of test	-	[67]
Co–N–C	Co: 2.52 wt%	∼−0.40	−0.67	∼46	∼6 (maximum∼ 18 at −0.88 V)	-	-	[50]
Co@CoNC-900	Co: ∼2 wt%	∼−0.40	−0.80	∼47	(maximum ∼8 at −0.80 V)	No significant current loss within 12 h of test at −0.05 V	4–1 by varying potential range −0.3 V–−0.8 V at 900 °C annealing temperature	[68]
CoNi–NC	Co: 1.2 wt%; Ni:1.2 wt%	-	−0.60	∼55	∼12 (maximum ∼74 at −1 V)	No significant current loss within 20 min of test at −0.60 V	0.23–∼3.30 by varying potential in the range −1 V–−0.5 V and Co/Ni precursors ratio in the range 0.2–5	[69]
CoSA–N–CNT	Co: 7.6 wt%	−0.27	−0.70	∼20	-	-	-	[58]
CoFeSA–N–CNT	Co + Fe: 11.50 wt%	−0.27	−0.30	∼30	-	-	-	[58]
Fe–N–G/bC	Fe: 7.67 wt%	∼−0.40	−0.66	∼95.80	∼7 (maximum ∼33 at −1.20 V)	No significant current loss within 12 h of test at −0.66 V	-	[70]
Fe/NG-750	Fe: 1.25 wt%	∼−0.30	−0.57	∼−80	∼1 (maximum ∼7.30 at −0.80 V)	No significant current loss within 10 h of test at −0.60 V	-	[71]
Fe–N/CNF-2	Fe: 0.33 wt%	∼−0.27	−0.53	∼95	∼7 (maximum ∼18 at −0.80 V)	No significant current loss within 24 h of test at −0.53 V	-	[72]
Fe–N–C	Fe: 0.1 at%	−0.29	−0.58	93	∼4.50 (maximum ∼15 at −0.90 V)	No significant current loss within 20 h of tests at −0.58 V	-	[73]
C–AFC©ZIF-8	Fe: 1.47 wt%	−0.33	−0.43	93	∼5 (maximum ∼23 at −0.85 V)	The half-wave potential decreases by 20 mV after 10,000 cycles	-	[74]
Fe–N–C-0.5	Fe: 2.91 at%	∼−0.30	−0.68	∼98	∼7 (maximum ∼20 at −0.90 V)	No significant current loss within 55 min of test at −0.58 V, −0.63 V and −0.68 V	-	[75]
Fe–SA/NCS-700	Fe: 0.89 at%	−0.22	−0.45	87	∼3 (maximum ∼35 at −1.35 V)	No significant current loss within 10 h of test at −0.45 V	-	[76]
Fe–N–C	Fe: 3 wt%	∼−0.40	−0.60	74	∼1.5 (maximum ∼11 at −1.20 V)	No significant current loss within 12 h of test at −0.60 V	0.25–∼3 by varying potential in the range −0.6 V–−1.0 V at 950 °C carbonization temperature	[77]
Fe–N–C	Fe: 2.14 wt%	∼−0.30	−0.49	∼87	∼0.2 (maximum ∼18 at −0.88 V)	-	-	[50]
Fe0.5d	Fe: 1.50 wt%	−0.30	−0.60	91	4.5 (maximum ∼19 at −0.90 V)	No significant current loss within 6 h of test at −0.60 V	0.2–4 by varying the Fe–N4 active centers percentage from 0 to 100%	[78]
AD-Sn/N-C1000	Sn: 1 wt%	∼−0.40	−0.60	91	∼2.50 (maximum ∼17.60 at −1 V)	No significant current loss within 25 min of test at −0.60 V	-	[79]
ZnNx/C	Zn: 0.10 wt %	−0.13	−0.43	95	∼3 (maximum ∼37.50 at −1.10 V)	No significant current loss within 75 h of test at −0.43 V	-	[80]
Pd–NC	Pd: 2.95 wt%	-	−0.50	55	-	No significant current loss within 3.5 h of test at −0.60 V	-	[81]
Ag2–G	-	−0.25	−0.70	93.40	11.87 (maximum 44.30 at −1 V)	No significant current loss within 36 h of test at −0.70 V	--	[82]
Cr–N–C	Cr: 2.22 wt%	∼−0.50	−0.85	∼72	∼5 (maximum∼ 13 at −1 V)	-	-	[50]
Mn–N–C	Mn: 2.74 wt%	∼−0.55	−0.59	∼70	∼0.20 (maximum ∼7.50 at −0.95 V)	-	-	[50]
Mn–C3N4/CNT	Mn: 0.17 wt%	∼−0.30	−0.55	98.8	∼15 (maximum ∼38 at –0.90 V)	No significant current loss within 20 h of test at −0.55 V	-	[83]
Ni/Fe–N–C	Ni: 0.97 wt%;Fe: 0.34 wt%	∼−0.45	−0.70	98	9.50 (maximum 23.70 at −1 V)	No significant current loss within 33 h of test at −0.70 V	-	[84]
Cu–N4–NG	Cu: 4.7 wt%	∼−0.40	−1.00	80.6	∼5.40 (maximum ∼9 at −1.30 V)	No significant current loss in the first hour of test at −1.00 V	-	[85]
Cu–N–C (H-type cell)	Cu: 0.61 wt%	∼−0.40	−0.67	98	∼4 (maximum ∼22 at −1.10 V)	No significant current loss within 20 h of test at −0.67 V	-	[86]
Cu–N–C (electrolyte flow cell)	Cu: 0.61 wt%	∼−0.10	−0.67	99	∼75 (maximum ∼130 at −1.13 V)	-	-	[86]
Cu–N4–C/1100	Cu: 1.98 wt%	∼−0.50	−0.90	98	∼7 (maximum ∼15 at −1.20 V)	No significant current loss within 40 h of test at −0.80 V	-	[87]

summarizes the main electrochemical parameters and results from these SACs.

In detail, along with the durability performance and the range of CO/H₂ ratio of the syngas produced (when evaluated), the onset potential, the total current density, and the corresponding maximum CO Faradaic Efficiency are reported as main parameters for measuring the catalyst performance.

The onset potential is defined as the applied potential at which electrochemical reaction begins to occur, and reaction products start to be formed on the surface of the electrode. From this parameter is derived the definition of overpotential of an electrochemical reaction, which is defined as the difference between the onset potential and the thermodynamic reduction potential of the reaction. It is clear, therefore, that an onset potential as close to zero as possible with respect to the reversible hydrogen electrode (RHE) would be desirable [49]. An onset potential of around −0.10 V vs. RHE is the lowest reported so far for this type of catalyst [88]. The current density is defined as the electric current per unit of surface or geometric area of the electrode. The higher the current density measured at a certain potential, the higher the reaction rate of the electrochemical reaction. Obviously, a high-performance electrocatalyst must be able to generate a high current density (most precisely, a high partial current density related to the formation of the desired product, in this case, either CO or H₂) at a potential as close as possible to the standard reduction potential, which corresponds to an overpotential equal to zero.

As can be observed from Table 2, maximum current densities reported in the majority of literature are below 100 mA/cm², i.e., low values, being at least an order of magnitude lower than industrial current densities. This is due to the adoption in the majority of the reported studies of a batch-type H-cell configuration, in which CO₂ solubility limitations represent a critical issue. This has been solved only in a few cases, as will be explained later in the review.

Eventually, the Faradaic efficiency (FE) is the parameter adopted to quantify the yield of a chemical produced through the electrochemical route with respect to all the produced species, and therefore the catalyst selectivity. FE is the ratio of the number of charges required to form a certain amount of desired product to the total charge over a specific time interval. The Faradaic efficiency for a specific product can be calculated as indicated in Equation (5):

$$FE=\frac{F·e·mol}{j·A·t}·100\%$$

(5)

where F is the Faraday constant (96,485.33 s⋅A⋅mol⁻¹), e is the number of electrons involved in the reaction, mol is the number of moles of desired product, j is the current density (A⋅m⁻²), A is the area of the electrode (m²), and t is the reaction time (s). Ideally, the sum of the Faradaic efficiencies of all the generated species should be equal to 100%, to guarantee a Faradaic balance.

As can be observed from Table 2, the majority of the SACs reported in the literature exhibit high FE towards CO, with maximum FE_CO values above 70%, which can become a drawback when aiming at producing syngas with high H₂/CO ratios. As will be better explored later in this review, there are few studies in which a better control of this ratio has been obtained, therefore reaching higher Faradaic efficiencies towards hydrogen and lower Faradaic efficiencies towards CO in accordance with the desired CO/H₂ ratio.

4.1. Nickel-based Single-Atom Catalysts

Because of the very strong adsorption energy of carbon monoxide on it, the Ni (111) bulk surface can be easily poisoned by adsorbed CO. Therefore, instead of showing high CO₂RR activity towards CO production, Ni nanoparticles consisting of Ni metals/metal oxides/hydroxides are highly active towards HER in aqueous solutions [89,90]. On the other hand, Ni SACs can show some of the highest activities towards CO₂RR to CO [91,92] As far as Ni SACs is concerned, Su et al. [51], as an alternative to organometallic electrocatalysts, synthesized Ni single atoms dispersed on N-doped graphene (named as Ni–N–Gr) by means of rapid heat treatment (900 °C, 1 min) of Ni-pentaethylenehexamine and graphene oxide (GO) under an inert atmosphere. This synthesis approach led to robust Ni–N bonds, where Ni atoms coordinated with N atoms in carbon-based materials are expected to serve as catalyst centers for CO₂ reduction, while proton adsorption behavior on Ni was found a key cofactor inducing the high activity of the catalyst. The X-ray photoelectron spectroscopy (XPS) and extended X-ray absorption fine structure (EXAFS) results confirmed the existence of isolated Ni atoms coordinated with N, with a loading of about 2.2 wt% embedded in the carbon matrix. This Ni SAC turned out to be highly selective towards CO, with a CO FE of over 90% at −0.70 V vs. RHE. Yang et al. [59] synthesized Ni single atoms dispersed on N-doped graphene with and without being co-doped with S (defined as A–Ni–NG and A–Ni–NSG respectively), by pyrolysis of L-cysteine, melamine, and nickel acetate in argon. The as-obtained catalyst reached a CO FE of 97% at around −0.5 V, a specific current of 350 A per g_catalyst at a 0.61 V overpotential, a low onset potential, and good stability, with only 2% of current loss after 100 h of the test. The dispersion of Ni atoms on the graphene was confirmed by HAADF-STEM, which showed Ni atoms as bright spots with a size of about 0.2 nm within the carbon matrix. In both A–Ni–NG and A–Ni–NSG, X-ray absorption spectroscopy and density functional theory (DFT) calculations enabled the discovery that the geometry of the catalyst around Ni(I) centers was highly distorted because of a non-centrosymmetric ligand strength, which increased the adsorption strength of CO₂ and CO₂ reaction intermediates on A–Ni–NSG. Results demonstrated the existence of delocalization of the unpaired electron in the Ni 3d orbital and of a spontaneous charge transfer from Ni(i) to the 2p orbital of the carbon in the carbon dioxide molecule to form intermediate adsorbed species.

One of the major issues in the synthesis of SACs is the high surface energy of the metal sites causing their uncontrolled aggregation, which is favored either by the growth of synthesis temperatures generally over 700 °C or by the increase in the metal loading, consequently leading to a reduction in the catalytic performance. On the other hand, in nanomaterials many defects can be found, such as vacancies, step edges, caves, doped defects, and lattice defects, which can alter the electronic structure and improve the catalytic performance of the catalysts [93,94,95]. These defects can act as anchor sites for the stabilization of metal atoms, therefore combining single metal atoms, and support defects can guarantee both high stability and high catalytic activities due to synergistic effects [96]. For instance, Qiao et al. [97] prepared an Au/FeO_x-type catalyst and discovered that the isolated Au atoms anchored at Fe vacancy sites exhibit very high binding energies, being highly positively charged. This leads to the formation of strong covalent bonds between the Au atoms and the lattice oxygen atoms of FeO_x, hence stabilizing the atoms on the support surface. Therefore, the strong metal–support interactions can be seen as a consequence of the presence of defects created on the surfaces of the oxides.

In regard to the M–N–C-based SACs (with M as generic transition metal), which are most frequently adopted for the CO₂RR catalysis, Jiang et al. [52] synthesized Ni single atoms anchored to a graphene shell for the CO₂RR to CO, leading to a maximum CO FE of 93.2% at −0.82 V vs. RHE with a CO partial current density of 20 mA/mg. The aforementioned catalyst (NiN-GS) was obtained through electrospinning of polymer nanofibers with Ni and N precursors homogeneously distributed. Transmission electron microscopy (TEM) images confirmed that HER-active Ni NPs are formed within the carbon shell by preventing the core from participating in the CO₂ reduction. Three-dimensional atom probe tomography (APT) confirmed the existence of atomically dispersed Ni sites anchored on the graphene vacancies. In particular, among all the Ni single atomic sites, only 0.2% were directly anchored to N, while 98% of them were embedded within the carbon defects in the graphene vacancies. As a consequence of DFT calculation results, the authors proposed that the most active sites towards CO production are the Ni atoms embedded within the carbon defects in the graphene vacancies, whereas N atoms promoted the entrapment of Ni atoms by generating defects in the graphene structure.

Furthermore, Jiang et al. synthesized Ni single atoms coordinated on graphene nanosheets through their vacancies (Ni-NG), obtaining a catalyst with high electrochemical performance [53]. The authors further explained that layered GO nanosheets were selected as support material for Ni SACs because of the high density of defects on GO, the high surface area, and the high negatively charged area, as well as the presence of a two-dimensional structure allowing clear characterization of the atomic sites. They firstly incorporated Ni ions into GO nanosheets, and then annealed them at 750 °C using ammonia as a reducing reagent and source of N dopants. The result is a catalyst that exhibited a CO FE of over 95% at −0.62 V vs. RHE, mainly due to its Ni–N atomic sites.

Yuan et al. [54] prepared Ni single atoms anchored to nitrogen-doped hollow carbon spheres (SA–Ni/N–CS). SiO₂/polydopamine spheres were chosen because they can efficiently adsorb Ni²⁺ ions coming from Ni(NO₃)₂·6H₂O precursors on their cavities. After pyrolysis, they were transformed into N–C hollow spheres. The as-prepared SA–Ni/N–CS catalyst exhibited an excellent CO faradaic efficiency of 95.1% at −0.8 V vs. RHE and 24 h stability without any current loss.

Another important parameter that influences catalytic activity and selectivity is the coordination number of the metallic atoms with N. In particular, Ni single-atom-based catalysts are reported to be highly active towards CO₂RR when each Ni atom is coordinated with four N atoms. Li et al. [55] synthesized a catalyst with exclusive Ni–N₄ active moieties through a topo-chemical transformation method. According to this strategy, a Ni-doped graphitic carbon nitride (g-C₃N₄) was synthesized starting from a mixture in the water of dicyandiamide, NH₄Cl and NiCl. Afterward, it was dispersed in a glucose solution, heated at 180 °C for 10 h, and eventually, annealed at 1000 °C for 1 h to coat Ni-doped g-C₃N₄ on a carbon layer, hence preventing Ni atoms from aggregation. Because of the high presence of Ni-N₄ active sites, the Ni–N₄–C catalyst exhibits a maximum CO FE of 99% at −0.81 V vs. RHE with a total current density of 28.6 mA/cm². The high contribution of Ni–N₄ sites for the CO₂RR to CO was confirmed since the poisoning of the Ni atoms through thiocyanate significantly reduced catalytic activity (see Figure 5).

Su et al. [60] proposed the synthesis of Ni single atoms on a covalent triazine framework (CTF), a conjugated polymer with 1,3,5-triazine linker units. The catalyst, named as Ni-CTF, achieved good electrocatalytic performance and a high CO selectivity (CO FE of about 95% at −0.90 V vs. RHE). The authors demonstrated that the lower coordination number of Ni atoms on CTF can increase the binding strength of reaction intermediates during CO₂RR, since such single atoms are more reactive. Indeed, the coordination number of the metal species in the catalyst can be used as an additional control parameter to improve the performance towards CO₂RR.

Another strategy to synthesize SACs is to adopt a highly porous metal-organic framework (MOF) as template, such as ZIF-8 [Zn(MeIm)2], a zeolitic imidazolate framework obtained from Zn²⁺ nodes and 2-methylimidazole. Zhao et al. [56] prepared Ni atoms dispersed on a N-doped carbon matrix (named as NiSACs/N–C) for CO₂RR by a ZIF-assisted method based on an ionic exchange between Ni ions and Zn nodes within the ultrafine pores of ZIF-8 (as summarized in Figure 6). During the pyrolysis step, the low-boiling-point Zn nodes evaporated, leaving the N-rich defects to be occupied by the neighboring Ni²⁺ ions, hence preventing them from agglomeration. As a result, Ni single atoms were formed, each of them coordinated with three N atoms homogeneously distributed on carbon matrix. The presence of Ni–N₃ active centers guaranteed good CO₂RR catalytic activity with a selectivity towards CO of over 71.9% at a 0.89 V overpotential, as well as high stability.

Hou et al. [62] successfully synthesized Ni single atoms stabilized on N-doped carbon nanotubes (Ni/NCTs) through a self-sacrifice template method carried out by pyrolyzing ZnO/ZIF-NiZn core/shell nanorods and by subsequently removing Zn species via acid etching. Compared with the Ni single atoms supported on porous bulk carbons prepared with a similar approach, the Ni/NCTs catalyst showed much better performance towards CO₂RR to CO, with a more positive onset potential of −0.323 V, a much larger CO current density of 34.3 mA/cm² at −1.0 V vs. RHE and a CO FE of almost 100% in the potential range −0.6 V–−1 V. According to the authors, this is mainly due to the hierarchically porous structure and high surface area of the nanotubes, as well as to the Ni–N active centers.

Another challenge to overcome in the development of SACs is the necessity of increasing the atomic loading, which is typically less than 2 wt% because of the insufficient number of anchoring sites on the support, as well as because the higher the metal loading is, the more likely the agglomeration of atoms during synthesis would be. Recently, Cheng et al. [57] synthesized Ni SACs on an extremely porous and surface-defect-rich microwave exfoliated graphene oxide (Ni–N–MEGO), reaching a Ni SACs loading of 6.9 wt%. The support was endowed with a highly porous surface area (2649 m² g⁻¹), which provided a high number of anchor sites for single Ni atoms; the large amount of nanopores, whose average size was less than 6 nm, contributed to reducing aggregation and stabilizing the SA Ni–N centers during high-temperature annealing. The catalyst showed a high CO FE of 92.1% at −0.70 V vs. RHE as well as a low onset potential of 0.29 V. In a later study, the authors managed to synthesize a Ni SACs with the highest atomic loading of 20 wt% [58]. The catalyst was obtained by means of a multistep pyrolysis process, starting from a mixture of dicyanamide and Ni acetylacetonate, and the 20 wt% loading was obtained when the annealing temperature reached 800 °C. The result was Ni single atoms dispersed on the walls of bamboo-like carbon nanotubes (NiSA-N-CNT), with an estimated coordination number for Ni–N of 3.8 ± 0.7, suggesting that Ni–N₄ active sites were formed during synthesis. Single-atom-N-CNT-based catalysts prepared by combining other metals, such as Co, Fe and Pt, show the superiority of the catalytic activity of the first Ni-based catalyst favored by the presence of more active Ni–N₄ centers.

As previously mentioned, the typical electrochemical system adopted in all the aforementioned studies is the H-type cell. One main limitation of such a configuration, in which CO₂ molecules have to diffuse through the liquid electrolyte to reach the electrode surface, is that the equilibrium concentration of CO₂ in an aqueous solution at 1 bar and 25 °C is about 30 mM, which corresponds, considering the occurrence of the CO evolution reaction, to a maximum achievable current density of around 60 mA/cm², a figure that is one order of magnitude below typical industrial current densities [98]. A solution to achieving higher current density values consists in continuous-flow reactors, where CO₂ is directly supplied from the gas phase through gas diffusion electrodes (GDEs), which integrate a catalyst layer with a porous substrate, allowing both an increase in the mass transport phenomena and the enhancement of the electrochemically active surface area available. Yang et al. [63] synthesized Ni single atoms on a porous carbon fiber membrane (NiSA/PCFM) through the electrospinning method. By virtue of its high porosity, excellent mechanical strength and good flexibility, NiSA/PCFM could combine gas-diffusion and catalyst layers into a single architecture and work as a gas diffusion electrode. Loaded on the cathode compartment of a flow electrolyte cell, the catalyst generated a maximum total current density of about 360 mA/cm^2, comparable with industrial applications, with a CO selectivity of around 88% at −1.0 V vs. RHE, and long-term stability without any significant current losses for 120 hours (see Figure 7). Möller et al. [64] tested their Ni–N–C electrocatalyst, synthesized by pyrolysis of a mixture of polyaniline and carbon- and Ni-precursors, in an industrial-like flow electrolytic cell and by spry-coating the catalyst ink on a gas diffusion electrode in order to overcome the limitation of the CO₂ solubility in water under ambient temperature and pressure. The Ni–N–C catalyst exhibited a maximum CO FE of almost 90% between CO current densities of 100 and 200 mA/cm², and maintained the 200 mA/cm² for 20 h of tests.

After a long period of literature results reporting studies about catalysts for CO evolution from CO₂RR, only recently has attention been given to the development of an efficient Ni SAC for CO₂ electrochemical conversion into syngas with a tunable CO/H₂ ratio. Zhu et al. [61] synthesized N-doped carbon nanorods containing both Ni single atoms and Ni nanoparticles (named as NiNP CNRs), which favored CO₂RR to CO and HER respectively. With such a catalyst, syngas with a wide range of CO/H₂ ratios could be obtained by simply tuning the ratio between Ni NPs and Ni single atoms, which could be achieved by controlling the acid etching time after pyrolysis of the nickel precursor. Indeed, by changing the etching time from 2 hours to 48 hours, the ratio between Ni–N sites and Ni NPs changed from 1:4.9 to 1:0, and subsequently, a syngas with a CO/H₂ ratio ranging from 1:9 to 19:1 and with a total current density higher than 15 mA/cm² and stability for over 8 h could be obtained.

4.2. Cobalt-Based Single-Atom Catalysts

Co single atoms dispersed on N-doped carbon-based materials have been recently explored as good catalysts for the CO₂RR to CO. In contrast to Ni SACs, the typical Co–N₄ moiety does not represent the most active center for CO production. Wang et al. [65] synthesized atomically dispersed Co catalysts on N-doped porous carbon through pyrolysis of a bimetallic Co/Zn ZIF. By varying the pyrolysis temperature, the authors managed to obtain SACs with different coordination numbers for Co: Co–N₄, Co–N₃, and Co–N₂, were synthesized at 800, 900, and 1000 °C, respectively, along with Co NPs for a better comparison. On Co–N₂, a CO current density of 18.1 mA/cm² was reached at −0.63 V vs. RHE, which is a value 23.3, 7.3, and 2.0 times higher than the ones obtained on Co–N₄, Co–N₃, and Co NPs respectively. Co–N₂ exhibited a maximum CO FE of 95% at −0.68 V vs. RHE, while Co–N₃ and Co NPs reached a maximum CO FE of 63% at −0.53 V vs. RHE and of around 7% at all the analyzed potentials, respectively. Furthermore, the Co–N₄ catalyst was found to be rather inactive towards CO₂RR. DFT calculations revealed that, on the Co–N₂-based catalyst, the CO₂⁻* intermediate is more easily formed, which means a higher CO₂RR catalytic activity. On the other hand, on Co NPs the formation of the H* intermediate is much more favored, hence hampering the CO₂RR. As for the Co–N₄ catalyst, the adsorption energies of CO₂⁻* and H* are high, which means that both HER and CO₂RR require high overpotentials to occur on it. The H* adsorption energy on Co–N₄ is anyhow lower than the one related to CO₂⁻*, which makes the catalyst superior towards HER.

Most recently, Pan et al. [67] synthesized Co single atoms anchored on polymer-derived hollow N-doped porous carbon spheres with Co–N₅ active centers (named as Co–N₅/HNPCSs, see Figure 8), which showed a CO FE of 99.2% and 99.4% at −0.73 V vs. RHE and −0.79 V vs. RHE, respectively. Similarly to the previous study, a comparison with SACs synthesized at different pyrolysis temperatures (Co–N₄ and Co–N₃, obtained at 400 and 600 °C respectively) showed the higher performance of the Co–N₅: indeed, the CO FE decreased as the Co–N_x coordination number dropped to 3 and 4. DFT calculations confirmed the importance of the crucial role of the Co-single-atom active sites, which must possess a high COOH* intermediate adsorption strength as well as a moderate CO adsorption energy.

Most recent studies are focused on the adoption of Co-based SACs for syngas production. These recent studies focus on the idea that the bifunctionality of the catalyst is the key to enabling a tunable co-generation of CO and H₂. Daiyan et al. [68] synthesized a Co-single-atom-decorated, N-doped graphitic carbon shell encapsulating a Co NPs core (Co@CoNC-900) capable of producing a stable H₂/CO ratio ranging from 0.25 to 1 in the −0.3 V–−0.8 V vs. RHE potential range. As for this catalyst, the authors recognized the following active sites: the graphitic shell encapsulating the metallic Co core for the CO₂RR and the Co–N₄ moieties on the top of the shell for HER. They also tested the effect of the annealing temperature on the H₂/CO ratio from ∼1.5 (when the annealing temperature is 800 °C) to 1 (when the annealing temperature is 900 °C) to ∼0.5 (when the annealing temperature reaches 1000 °C, within a wide potential range). On the other hand, Song et al. [66] proposed that the CO₂RR active sites of the catalyst are the Co–C₂N₂ moieties, while other nitrogen groups, such as graphitic and pyridinic N, can favor HER. In detail, the authors synthesized a Co single-atom catalyst supported on a nitrogen-doped 3D hollow carbon structure (named as Co-HNC) for the syngas production through a 900 °C pyrolysis of a ZnO/ZIF core/shell precursor. The chosen precursor and the synthetic process enable the obtainment of a catalyst with a tridimensional hollowed structure and a sponge-like thin shell with a hierarchical porous system, which favors the mass transport phenomena and makes the active sites highly exposed. The result is a catalyst able to guarantee the formation of syngas with a stable CO/H₂ ratio of 1/2 in the potential range −0.7–−1 V vs. RHE, and a CO FE of around 35% for applied potentials above −0.8 V.

4.3. Iron-based Single-Atom Catalysts

Among single-atom catalysts for the CO₂RR to CO, Fe-based single-atom catalysts have been also recently explored as single atoms dispersed on N-doped carbon supports. Fe SACs can exhibit high selectivity towards CO, with a maximum achievable CO FE of about 98% at −0.68 vs. RHE [75], comparable to the values achieved with Co-based and Ni-based SACs. However, unlike them, Fe–N–C electrocatalysts exhibit the highest CO selectivity at low overpotentials, from approximately −0.43 V vs. RHE to –0.68 V vs. RHE. Nonetheless, the current densities recorded on these Fe-based catalysts are rather low, with a maximum total current density (CO and H₂) of around 35 mA/cm² at −1.35 V vs. RHE. Zhang et al. [70] adopted electrochemically exfoliated graphene-based foils, melamine, and 1-butyl3-methylimidazolium tetrachloroferrate ([BMIM]FeCl₄) to obtain Fe atoms along with Fe₃C nanocrystals embedded into a graphene nanosheet/bamboo-carbon nanotube matrix, by electrochemical charging and ball-milling followed by annealing. The result was an effective catalyst (named Fe–N–G/bC) towards CO₂RR, which achieved a CO FE of about 95.8% at −0.66 V vs. RHE and a total current density of around 33 mA/cm² at −1.20 V. The authors explained the good activity as being a result of the Fe–N_x active sites and the Fe₃C phase, which could lead to synergistic effects in combination with the atomic active sites. Moreover, the hierarchical porous carbon matrix facilitated reactant accessibility and fast pathways for charge and mass transport. Zhang et al. [71] immobilized Fe atoms on N-doped graphene support (Fe/NG) by annealing a mixture of graphene oxide and FeCl₃ in an Ar/NH₃ atmosphere at 700–800 °C, which did not lead to the formation of Fe nanocrystals. The existence of Fe single atoms combined with nitrogen atoms in the graphene layers was confirmed by means of both HAADF-STEM and XAFS, and the high selectivity of the catalyst towards CO (80% CO FE at −0.57 V vs. RHE) could be attributed to a synergistic effect between these two atomic species on graphene. Eventually, DFT helped understand that the synergetic effect of the Fe–N₄ moieties and nitrogen-doping on the graphene surface (the Fe-N₄ + 2N_g(2) active sites) was the main reason for the high electrochemical performance (see Figure 9). A similar result was found also by Zhu et al. [76], whose catalyst was constituted by single atomic iron evenly dispersed on N-doped carbon nanosheets obtained by pyrolysis of hemin-doping polyaniline (named as Fe–SA/NCS-X, with X referring to the pyrolysis temperature). Even in this case, DFT results indicated that the presence of graphitic N synergistically improved the CO₂RR activity of the Fe–N₄ active center by reducing the free energy barrier for the formation of the COOH* intermediate.

Cheng et al. [72] synthesized an Fe–N–C nanofiber encapsulating iron nitride nanoparticles through its Fe and N co-doped carbon layers (named as Fe–N/CNF) through electrospinning, carbonization, acid-leaching, and nitridation under NH₃. The as-obtained catalyst showed high catalytic activity towards CO₂RR, especially that obtained after a 20-min-nitridation time (Fe–N/CNF-2) which exhibited over 95% of CO FE at −0.53 V vs. RHE. DFT calculations revealed that the desorption of the adsorbed CO* from the Fe–N-based shell is easier when the iron nitride core is included, but this is not an advantage: the binding energy of adsorbed CO on the Fe atom is too strong to enable CO desorption. On the other hand, in the catalyst with the Fe₂N core (Fe₂N@Fe–N₄–C), an overpotential of 0.19 V is required for CO₂ adsorption, but the adsorbed CO can leave more easily the catalyst surface. Following the same reasoning, HER is not favored on the Fe₂N@Fe–N₄–C catalyst.

In general, understanding the chemistry of the active sites and the origin of the high CO₂RR activity of these Fe–N_x–C catalysts remains a source of scientific interest due to their complexity deriving from high-temperature treatments. For instance, Pan et al. [73] synthesized a Fe–N–C catalyst with a low onset potential of −0.29 V vs. RHE, and explained its high performance through the presence of edge-hosted Fe–N₂₊₂−C₈ moieties bridging two armchair-like graphitic layers as active sites for CO₂RR to CO instead of the traditionally proposed Fe–N₄–C₁₀ sites anchored on a compact carbon plane. The easier occurrence of the CO₂RR is due to the presence of both Fe centers and C atoms with dangling bonds next to N, which adsorb CO*and OH* intermediates respectively, hence leading to the cleavage of the C–O bonding.

Another important issue is not only to identify the active sites but also to guarantee a high exposure of them on the surface support. In the study by Ye et al. [74], through functionalization of the surface of zeolitic imidazolate framework-8 with ammonium ferric citrate (AFC), isolated Fe–N sites were formed on the surface of the catalyst (named as C-AFC©ZIF-8) obtained after a pyrolysis treatment. In comparison with the C-AFC@ZIF-8 catalyst obtained by means of bulk functionalization of ZIF-8 with AFC, C-AFC©ZIF-8 exhibited much higher catalytic activity by virtue of its highly exposed Fe-N active sites on the surface support. In particular, the maximum CO FE of C-AFC©ZIF-8 is 93.0% at −0.43 V vs. RHE (versus the CO FE of 84% at the same potential for C-AFC@ZIF-8) and the total current density reaches its maximum value of around 23 mA/cm² at −0.85 V vs. RHE (versus the value of around 12 mA/cm² at the same potential for C-AFC@ZIF-8) and at the maximum iron content of 1.47%, suggesting the importance of the Fe–N active sites for the catalysis of CO₂RR. Wu et al. [75] synthesized an Fe–N–C catalyst with highly exposed Fe–N_x active sites through pyrolysis of a 3D sea-urchin-like FeOOH-polyaniline precursor, which also guaranteed large specific surface and electrochemically active surface areas. The prepared catalyst achieved a high CO FE of 95% with CO partial current density of 1.9 mA/cm² at an overpotential of 530 mV.

Recently, similarly to Ni and Co SACs, Fe-based single-atom catalysts have been also explored as effective catalysts for the syngas electro-production from CO₂ in an aqueous solution. Zhao et al. [77] prepared a Fe–N–C electrocatalyst from iron chloride and urea-formaldehyde resin through pyrolysis at different temperatures and Fe contents. At optimal carbonization temperature and Fe metallic content of 950% and 3%, respectively, and by varying the applied potential in the range of −0.6 V to −1.0 V vs.RHE, the H₂/CO ratio of the obtained syngas varies from 4 to 1/3, which is a relatively wide range when compared with one of noble metal-based catalyst and other M–N–C catalysts. Huan et al. [78] synthesized Fe–N–C electrocatalysts (named Fe_4.0d, Fe_1.0w, Fe_1.0d, Fe_0.5d, and Fe_0.5d-950 according to the different synthesis conditions) with different proportions of Fe single-atom centers and Fe nanoparticles. The Fe–N₄ active centers catalyzed the CO production, whereas Fe NPs mainly favored HER (as outlined in Figure 10). Therefore, by varying the Fe–N₄ active centers’ percentages from 0 to 100%, syngas with a CO/H₂ ratio in the range between around 0.2 and 4 could be obtained.

4.4. Other Metal-based Single-Atom Catalysts

According to recent studies, Zn single atoms, Sn single atoms, and Cu single atoms coordinated with N showed high activity and selectivity towards CO₂RR to CO. Zhao et al. [79] prepared Sn single atoms dispersed on pyridinic N-doped carbon nanofibers (AD-Sn/N-C1000) through electrospinning and pyrolysis at 1000 °C. Unlike Sn NPs on N-doped nanofibers, which exhibited a 62% FE towards formate, the AD-Sn/N-C1000 catalyst showed a CO FE of 91% at an overpotential of 490 mV. On the other hand, Yang et al. [80] produced a cost-effective Zn–N–C type catalyst with atomically dispersed Zn coordinated with N atoms forming Zn–N₄ active sites (ZnN_x/C). The catalyst was tested in a CO₂-saturated, 0.5 M KHCO₃ solution, showing high catalytic performance for CO₂RR to CO, such as a low onset potential of 0.13 V, a CO FE of 95% at 0.43 V vs. RHE, and high stability without significant current losses for 75 h of the test. DFT calculations confirmed that the high activity of ZnN_x/C can be attributed to a synergistic effect between Zn and N atoms in the Zn–N₄ active centers, which is responsible for reducing the energy barrier for the formation of the COOH* intermediate.

Copper is typically adopted in its bulk form as an effective catalyst for CO₂ electro-reduction to high-molecular products such as alcohols and hydrocarbons. However, recently it has been proved that Cu as single atoms can selectively catalyze the CO formation. For instance, Xu et al. [85] proposed the synthesis, through a two-step pyrolysis method, of single Cu atoms embedded in 2D N-doped graphene support (named as Cu–N₄–NG). In contrast with the bulk Cu catalyst, which exhibited relatively low selectivity towards CO, electrochemical tests proved that Cu–N₄–NG exhibited a much higher CO Faradaic efficiency of 80.6% at −1.0 V vs. RHE. This is mainly due to the presence of Cu–N₄ sites, which highly facilitated the CO₂ activation step, whereas the graphene substrate promoted water dissociation, which provides protons involved in the CO₂RR process. Cheng et al. [87] confirmed the effectiveness of the Cu–N₄ site for the catalysis of the CO₂RR to CO, since it can guarantee an appropriate binding energy strength in COOH*and CO* intermediates, and therefore boost CO generation. In particular, the authors successfully synthesized Cu single atoms dispersed on N–C support through a MOF-assisted method (named Cu-N₄-C/1100). The as-synthesized catalyst achieved a maximum CO Faradaic Efficiency of 98% at –0.9 V vs. RHE and high stability for at least 40 h of the test. Chen et al. [86] prepared Cu single atoms on a nitrogenated carbon-based catalyst (named Cu–N–C). The Cu–N₃ sites of the catalyst were proved to strongly promote the CO* desorption. Indeed, in a gas-tight H-type cell, Cu–N–C reached a high CO faradaic efficiency of 98% at −0.67 V vs. RHE, and high durability (Faradaic Efficiency remains above 90% over 20 h of tests). The same catalyst was also tested in an electrolyte flow cell configuration, exhibiting an even higher CO Faradaic Efficiency of 99.9% at −0.67 V vs. RHE due to the boosted rate of CO₂ diffusion. DFT calculations revealed the reason for the improved performance of the catalyst: the Cu–N₃ site was located on an extended carbon plane with six nitrogen vacancies which stabilize the active site, while three unoccupied N sites were spontaneously saturated by protons during the CO₂RR. Therefore, the hydrogen bonds formed between the O atom of *COOH and the H atom of adjacent protons contribute to dramatically reducing the energy barrier of *COOH formation. After the first proton-coupled electron transfer process, the hydrogen bond disappears and the adsorbed *CO species are easily released, hence guaranteeing the CO production.

Other than Ni, Fe, and Co-based SACs, Pan et al. [50] synthesized Mn–N–C and Cr–N–C catalysts through pyrolysis of a solid obtained by drying a water mixture of urea, citric acid, and metal precursor. The Mn-based and Co-based single-atom catalysts exhibited comparable maximum CO FE (72% and 70%, respectively), while the overpotential of Mn–N–C required for reaching the maximum CO selectivity was 260 mV lower than that of Cr–N–C. In any case, it can be observed that both Mn and Cr showed lower catalytic performance towards CO formation compared to single-atom catalysts, such as Ni and Fe.

Manganese was also the key element in the catalyst prepared by Feng et al. [83]. Instead of an Mn–N₄-based structure embedded in graphene support, the authors proposed the synthesis of a more efficient catalyst consisting of Mn–N₃ sites embedded in graphitic carbon nitride (g-C₃N₄) on carbon nanotubes (named as Mn–C₃N₄/CNT). DFT calculations proved that the unusual Mn–N₃ site is responsible for a higher performance of the catalyst by favoring the formation of the COOH* intermediate. Indeed, Mn–C₃N₄/CNT exhibited a 98.8% CO Faradaic efficiency with a CO partial current density of 14.0 mA/cm² at a low overpotential of 0.44 V vs. RHE in aqueous electrolyte, outperforming all the Mn–N₄-based SACs previously reported in the literature.

Starting from the idea of reducing the amount of Pd in Pd-based nanostructured catalysts for the CO₂RR, He et al. [81] prepared an N-doped carbon-supported Pd SAC (named as Pd-NC) with a low Pd loading of 2.95 wt%. Through several characterization techniques as well as by means of DFT calculations, the authors revealed the unique coordination nature of Pd atoms with N atoms in Pd–N₄ active centers, which helps stabilize and activate the adsorbed CO₂, hence guaranteeing CO generation at low overpotentials. In comparison with commercial palladium on carbon (Pd/C), which shows high selectivity towards HER, the Pd-NC exhibited a moderate selectivity also towards CO₂RR to CO, with a CO FE of 55% at −0.50 V vs. RHE.

A more recent and novel approach is to develop catalysts whose active sites consist of two or three adjacent atoms or few-atom sub-nanoclusters, which can show a comparable or even better catalytic activity compared with single-atom catalysts. For instance, as for CO production from CO₂ electro-reduction reaction, Li et al. [82] reported the synthesis of a dual-atom Ag₂/graphene catalyst (Ag₂–G) for CO₂ electrochemical reduction to CO. The Ag₂–G active site consists of two adjacent Ag atoms, each of them coordinated with three nitrogen atoms, and the AgN₃–AgN₃ sites are linked to the graphene matrix through strong Ag–C bonds. While CO₂ interacts with the Ag single atom of Ag–N₄ sites in the more conventional Ag₁–G catalyst (i.e., single Ag catalyst) only with its carbon atom, both the carbon and the oxygen atoms of carbon dioxide interact at once with the two Ag atoms of AgN₃–AgN₃ sites in Ag₂–G, which makes the reaction intermediates on the latter catalyst more stable. Consequently, Ag₂–G showed high CO₂RR to CO performance, including a low onset potential of −0.25 V vs. RHE, a high CO FE up to 93.4% with a current density of 11.87 mA/cm² at −0.7 V vs. RHE, and stability for 36 h of tests, hence proving its superiority to the single-atom Ag₁/graphene and the traditional Ag NPs.

5. Summary

In summary, metals such as Ni, Fe, Co, Zn, etc. in the form of single-atom catalysts, specially bound to N through either covalent bonds or coordination bonds and embedded in carbonaceous conductive support, were tested and found to be valuable and effective approaches to reducing CO₂ and producing syngas:

• First of all, they are more economical than traditional Au and Ag-based nanostructured catalysts, consisting of lower quantities of more readily available and low-cost metals dispersed on low-cost and highly conductive supports, such as carbon-based materials. A cost/power analysis performed on a CO₂ plant fed with 13,750 ton/day of CO₂ and producing syngas revealed that, for instance, when adopting a Ni–N–C catalyst rather than an Ag-based catalyst in a flow-cell, less power is required for the reactions to occur [99].
• At the same time, they are highly active towards CO₂RR due to their unique characteristics, which distinguish them from their nanoparticle-based counterparts: low-coordination state and homogeneity of catalytically active sites with maximum metal utilization efficiency, i.e., all the catalytic active sites of the metal are exposed to the reactants.
• Finally, these catalysts exhibit a higher selectivity towards CO₂RR to CO than towards HER when compared with their nano-sized counterparts, as a consequence of the nature of their active sites. Indeed, the metal single atoms are often organized in the form of M–N_x active sites (where M = metal), on which chemisorption of atomic hydrogen is not favored, because: (i) the adsorption step of H⁺ ions require higher energy (typically, a 0.3 eV penalty is recorded for the formation of H_ads) and (ii) HER proceeds through a Volmer–Heyrovsky mechanism rather than through a faster Volmer–Tafel one [49,100]. Furthermore, nitrogen atoms not only contribute to increased activity and selectivity of the transition metal single atoms, but nitrogen incorporation into a carbon matrix can also play an effective role in stabilizing the highly energetic single atoms through coordination bonds, avoiding their aggregation [101].
• Finally, these catalysts exhibit a higher selectivity towards CO₂RR to CO than towards HER when compared with their nano-sized counterparts, as a consequence of the nature of their active sites. Indeed, the metal single atoms are often organized in the form of M–N_x active sites (where M = metal), on which chemisorption of atomic hydrogen is not favored, because: (i) the adsorption step of H⁺ ions require higher energy (typically, a 0.3 eV penalty is recorded for the formation of H_ads) and (ii) HER proceeds through a Volmer–Heyrovsky mechanism rather than through a faster Volmer–Tafel one [99,100]. Furthermore, nitrogen atoms not only contribute to increased activity and selectivity of the transition metal single atoms, but nitrogen incorporation into a carbon matrix can also play an effective role in stabilizing the highly energetic single atoms through coordination bonds, avoiding their aggregation [101].

Moreover, with the aim of producing syngas, the CO/H₂ ratio can be tuned in a wide range, not only by varying the type and morphology of metals and supports, but also by changing the size of the active phase, for instance by reaching a definite proportion of HER-active nanoparticles and CO₂RR-active single atoms, which corresponds to a precise syngas ratio [78].

6. Future perspectives and challenges

Despite the recent advancements, it is clear, however, that there are still some challenges to overcome for further development and scale-up of these catalysts. New routes can be and need to be explored for the development of more efficient catalysts with lower onset potentials and higher current densities at ambient conditions.

Firstly, not only should new metals be explored in monometallic SACs, but also and above all, more focus should be placed on the development of multimetallic SACs, which are expected to show improved catalytic activity by virtue of novel synergistic mechanisms generated by the interaction between different metals, as happens for the nano-sized counterparts [102,103]. Currently, only a few examples can be found in this direction. In particular, Ren et al. [84] synthesized an efficient bi-atomic nickel and iron–nitrogen carbon-based catalyst (Ni/Fe–N–C) for CO production through an ion-exchange strategy, based on the pyrolysis of a Zn/Ni/Fe zeolitic imidazolate framework. Experimental results and DFT calculations revealed that the Ni/Fe–N dual-site works according to a different mechanism. It lowers the energy barrier for both the formation of the COOH* intermediate and the desorption of CO* when a previous CO* is already present on the active site, leading to superior CO₂RR activity. On the other hand, and with the primary aim of producing syngas, He et al. [69] explored the catalytic activity of both Ni and Co single atoms dispersed on N-doped carbons (TM-NC with TM=Co and/or Ni), finding that the Ni-NC catalyst was extremely active towards CO₂RR to CO evolution (>56 mA/cm² at −1.0 V vs. RHE), whereas the Co–NC catalyst exhibited a favorable HER activity (>58 mA/cm² at −1.0 V vs. RHE). On the basis of that, the authors synthesized single-atom catalysts, including Co and Ni, with different Co/Ni ratios (CoN-NC). The CoNi-NC catalysts exhibited a total current density as high as 74 mA/cm² at −1.0 V_RHE with tunable CO/H₂ ratios (ranging from 0.23 to about 3.30) and with the sum of CO and H₂ faradaic efficiencies almost equal to 100 %.

In addition, the combination of different metals with a non-metal, also different from N, is another route that can lead to lowered overpotential for CO formation, as a consequence of a bi-functional effect obtained by combining atoms of different metals, i.e., atoms with different binding energies towards the COOH* and H* intermediates. Indeed, an efficient catalyst for CO₂RR to CO must be able to stabilize COOH* without stabilizing CO* to a similar degree (Figure 11), as it naturally occurs with enzymes that possess two different functional sites: one bonds the C-end of CO* and COOH*, while the other can stabilize the O group in COOH*, favoring the C–O cleavage [24].

Furthermore, the majority of the thus-far reported studies focus on testing the electrochemical performance of the catalysts in a batch-type H-cell configuration. In this case, scaled-up conditions are not reproducible, due to inefficient mass transport phenomena, poor mixing, additional losses due to membrane resistance, and very low current densities. In order to verify the effectiveness of the catalysts on a scaled-up electrochemical environment, flow-reactors, such as membrane reactors and microfluidic reactors, should be adopted for tests, along with incorporating the catalyst within GDE in order to solve the CO₂ solubility limitations, thus far adopted in a few cases [63,64].

Moreover, further attention should be paid to the choice of the support and to the study of its influence on the catalysis of the atomic centers dispersed on it. Currently, most of the developed SACs for CO₂RR are supported on carbon-based materials, such as graphene-based materials and carbon nanotubes. However, many other supports could be analyzed and compared, such as oxides, nitrides, and transition metal dichalcogenides. Moreover, defects, either natural or artificially created, could work as anchoring sites to stabilize single atoms, enabling higher metal loadings and avoiding agglomeration, which is another critical issue for SACs. In particular, among defects, cation and anion vacancies are the most frequently adopted to guarantee the anchorage of metal single atoms to metal compound-based supports. More specifically, metal oxides can use their cation vacancies as anchoring sites. For instance, Ni(OH)₂ has been recently used to anchor Pt single atoms through Ni²⁺ vacancies [104], TiO₂ single-crystal nanosheets were doped with Rh atoms by means of Ti⁴⁺ centers [105], and Al₂O₃ bonded Pt clusters through coordinatively unsaturated pentacoordinate Al³⁺ centers [23]. Deng et al. [106] prepared, through a one-pot chemical method which adopted (NH₄)₆Mo₇O₂₄, H₂PtCl₆, and CS₂ as precursors, isolated Pt atoms anchored on few-layer MoS₂ nanosheets as an efficient catalyst for the HER. HAADF-STEM results highlighted that the single Pt atoms substituted some of the Mo atoms. As previously mentioned, anions vacancies represent the second type of anchoring site. For instance, oxygen vacancies can be formed in metal oxides, as in TiO₂ nanosheets for the stabilization of Au single atoms [107]. Another type of anion vacancy is the sulfur vacancy, which can be obtained by the chemical exfoliation of transition metal chalcogenides, such as MoS₂. Liu et al. [108] synthesized Co single atoms dispersed on MoS₂ monolayers by firstly chemically exfoliating bulk MoS₂ and then mixing these sheets with thiourea-based Co species via sulfur vacancies. The authors observed that the strong covalent bonds between Co atoms and sulfur vacancies of MoS₂ are the main reason for the high activity, stability, and selectivity of the catalyst towards hydrodeoxygenation reaction. Although many studies have been performed, and many other interesting perspectives suggested, these materials have never been tested for CO₂RR to CO or for syngas electro-production. This study may prompt the future development of better-performing SACs towards these reactions.