Heteroatom-Doped Porous Carbon-Based Nanostructures for Electrochemical CO2 Reduction

The continual rise of the CO2 concentration in the Earth’s atmosphere is the foremost reason for environmental concerns such as global warming, ocean acidification, rising sea levels, and the extinction of various species. The electrochemical CO2 reduction (CO2RR) is a promising green and efficient approach for converting CO2 to high-value-added products such as alcohols, acids, and chemicals. Developing efficient and low-cost electrocatalysts is the main barrier to scaling up CO2RR for large-scale applications. Heteroatom-doped porous carbon-based (HA-PCs) catalysts are deemed as green, efficient, low-cost, and durable electrocatalysts for the CO2RR due to their great physiochemical and catalytic merits (i.e., great surface area, electrical conductivity, rich electrical density, active sites, inferior H2 evolution activity, tailorable structures, and chemical–physical–thermal stability). They are also easily synthesized in a high yield from inexpensive and earth-abundant resources that meet sustainability and large-scale requirements. This review emphasizes the rational synthesis of HA-PCs for the CO2RR rooting from the engineering methods of HA-PCs to the effect of mono, binary, and ternary dopants (i.e., N, S, F, or B) on the CO2RR activity and durability. The effect of CO2 on the environment and human health, in addition to the recent advances in CO2RR fundamental pathways and mechanisms, are also discussed. Finally, the evolving challenges and future perspectives on the development of heteroatom-doped porous carbon-based nanocatalysts for the CO2RR are underlined.


Effect of CO2 on the Environment and Human Health
CO2 emissions come from industrial activities (32%), building operations (28%), transportation (23%), building materials and construction (11%), and other sources (6%) and are one of the main reasons for global warming and climate change, which affect humans and the environment significantly [1,55]. CO2 absorbs a lower heat than other greenhouse gases, but it remains in the atmosphere longer and acts as a blanket in the air, trapping heat in the atmosphere and warming up the Earth's temperature and increasing the temperature of the ocean (0.07 °C/decade) [1,55]. Increases in Earth's temperature lead to

Fundamental Parameters for CO 2 RR Performance
The fundamental parameters for estimating the CO 2 RR performance are the Faradaic efficiency (FE), overpotential, current density, durability, and energy efficiency [1,53].

Faradaic Efficiency (FE)
The FE is calculated from the following equation: where n, m, F, and Q are the number of electrons, total mole amounts of the product, Faraday constant (96,485 C/mol), and amounts of cumulative charge during CO 2 reduction. The FE determines the CO 2 RR selectivity for various gas/liquid products (i.e., CH 4 , CO, HCOOH, and CH 3 OH), so a higher FE for specific products is highly required to reduce the cost of isolation products. A catalyst with an inferior HER ability is desired to yield a great FE and high selectivity, which could be enhanced via the integration of heteroatom dopants into the carbon skeleton structure and modifying the hydrophobicity of electrodes. Organic and hybrid electrolytes can also improve the FE due to their great CO 2 adsorption ability and higher solubility of CO 2 during CO 2 RR; however, the effect of electrolytes on CO 2 RR activity and selectivity is not yet resolved and remains ambiguous [34,[49][50][51][52][53].

Overpotential
The CO 2 RR half-reaction under applied potential, which is a nonspontaneous process driven by more negative potentials than standard potentials in actual electrocatalytic conditions, results in various products. The difference between the standard thermodynamic potential of a specific half-reaction and the applied potential in the half-cell reaction is the overpotential. This is often used to activate inert CO 2 molecules to form a bent *CO 2 •− anion radical and allow electron transport during the CO 2 RR process. The lower overpotential is desired to mimic the practical process and reduce the CO 2 RR production cost, which could be achieved by modulating carbon-based catalysts' morphology and composition. The CO 2 RR can occur through a single and multiple proton-coupled electron transfer process, so carbon-based materials with tunable structures and properties are thermodynamically feasible for single electron and multiple electron transfer processes [34,[49][50][51][52][53].

Partial Current Density
The partial current density (j p ) represents the current density needed for the production of specific products, and it is calculated using the following equation: where j is the total current density, so the effective catalyst should deliver a high j p under a low applied potential. The j p depends on the inherited catalytic properties of HA-PCs (i.e., electrical conductivity, interaction with the electrode, and CO 2 adsorption) in addition to the electrolyte and cell design [34,[49][50][51][52][53]. The flow cell system allows prompt CO 2 feeding into the cathode with the assistance of a gas diffusion electrode, resulting in an outstanding j p . Notably, to date, the obtained total current density is lower than 1 A, which is still far beyond commercial requirements [34,[49][50][51][52][53].

Durability
The durability of FE and current density are some of the essential factors hinging on CO 2 RR with maintained selectivity and activity under prolonged potentiostatic polarization. Notably, durability tests are carried out for a few hours, which is insufficient for the large-scale CO 2 RR applications; however, currently, most of the catalysts are exposed to significant attenuation in the current density, FE, and selectivity, owing to blocking and deactivation of the active catalytic sites and structural degradation. In light of the commercialization scale, the stability tests should be conducted for several weeks or months (≥1000 h) along with carrying out various in situ and ex situ characterizations to understand the degradation mechanism and solve this critical issue. Carbon-based catalysts, with their impressive chemical-physical stability, are promising to solve the stability issues in CO 2 RR, but they are not investigated enough. The modification of the electrochemical cell to allow continued CO 2 flow, refreshing the electrolyte solution, preventing gas accumulation over the cathode, and gathering liquid products, can enhance the long-term stability of porous carbon-based catalysts [34,[49][50][51][52][53].

Energy Efficiency (E eff )
The E eff is a crucial factor in estimating the economic efficiency of CO 2 RR as it is the percentage of the energy stored in the chemical. The E eff is calculated based on the equilibrium potential (E eq ) using the following equation: Thus, a great E eff together with outstanding catalytic activity drives the CO 2 RR halfreaction at a low η, negligible ohmic potential drop, and high FE. The high E eff is optimized via adjusting the electrolyte, membrane, and cell structure in addition to carbon-based electrode conductivity. Notably, the relationship between the structure/composition of HA-PC catalysts and E eff is not yet studied [34,[49][50][51][52][53].

Turnover Frequency (TOF)
TOF is the amount for generating specific products over single active sites per unit time during CO 2 RR process, so it is an imperative factor toward the intrinsic activity of HA-PCs. The TOF is calculated from the following equation: TOF = N p /N c where N p and N c are the product's mole number and the mole number of the catalyst's active site, respectively. A high TOF value indicates the presence of multiple active sites.

Template-Based Method
The template-based method is the most effective approach for the rational synthesis of HA-PCs with well-defined morphology, surface area, and ordered porosity driven by the template shape, structure, and properties. There are various hard templates (i.e., MgO, AlO, CaCO3, ZnO, and SiO2), soft templates (i.e., surfactants, polymers, and ionic liquids), and self-templates (i.e., biomass and metal-organic frameworks (MOFs)) or their hybrids [60][61][62][63]. The carbon precursors should be initially infiltrated into the template and annealed at an elevated temperature, followed by chemical etching of the template in acid or alkaline solutions. The source for heteroatoms can be mixed initially with the carbon source to produce heteroatom-doped porous carbon nanostructures.

Hard Templates
With its unique physiochemical properties, the ZnO template can generate porous carbon with 1D (i.e., nanorods and nanotubes) and 3D nanostructures via the incorporation of activators (i.e., KOH). ZnO is easily prepared commercially at a low cost via Zn's electrochemical anodic oxidation method under ambient conditions. Moreover, carbon precursors react with ZnO at a high temperature to produce CO2 gas that acts as an activator to enhance the porosity and surface area. Notably, Zn tends to evaporate via an annealing process at 900 °C, resulting in eliminating the need for acids or alkali solution to remove the ZnO template. Porous core/shell carbon microrods with a high surface area of 660 m 2 /g were prepared using a ZnO microrod template and glucose as a carbon source (Figure 2a-c) [64].  (a-c) SEM images and preparation process of porous core/shell carbon microrods using ZnO microrods template. Reprinted with permission from [64]. Copyright 2015 Wiley. TEM image and BET surface area of porous carbon prepared using (d,g) colloidal SiO 2 , (e,h) ABA-15, and (f,i) MMT templates. Reprinted with permission from [65]. Copyright 2013 American Chemical Society. (j) The fabrication process with its related SEM and TEM images of hierarchical porous carbon nanofibers formed using CaCO 3 template. Reprinted with permission from [66]. Copyright 2016 Elsevier. ZnO microrods on Ni-foil were initially designed by chemical bath deposition, then coated with Ni metals by electro-deposition, and then coated with glucose via autoclave at 180 • C for 3.5 h, followed by annealing at 500 • C under Ar (Figure 2a-c). At 500 • C, glucose is carbonized into carbon nanospheres coated with porous microrods formed from ZnO's evaporation. Three-dimensional flower-like hierarchical porous carbon nanostructures with a surface area of 761.5 m 2 /g, pore volume of 0.49 cm 3 /g, and microporosity of 49% were prepared via a ZnO template in the presence of HO as an activator [67]. With its unique crystal structure, MgO acts as a template for producing porous 2D sheets, 3D clusters, and 2D/3D nanostructures [68]. Mesoporous 3D carbon nanosheets with a BET surface area (883 m 2 /g) and pore size (6-8 nm) were synthesized using a MgO template and coal tar pitch as the carbon source, followed by washing with 10% HCl to remove MgO [69]. Porous carbon nanocages (CNC670) with a surface area of 2053 m 2 /g and pore size of 3-7 nm were obtained using MgO as the template and benzene as a carbon source via annealing at 670 • C and then removal of MgO by HCl [70]. Notably, annealing at a high temperature of 700, 800, and 900 • C led to decreasing the surface area to 1854, 1633, and 312 m 2 /g, respectively. Other Mg-based materials could also be used as a template for the production of HA-PCs, because they can produce in situ MgO templates. Hierarchical porous carbon nanosheets with a surface area of 2300 m 2 /g were formed using starch as a reductant and carbon source while Mg(NO 3 ) 2 was an oxidant and in situ provided a MgO template [71]. Similarly, layered porous carbon nanosheets with a surface area of 1312 m 2 /g were formed via the combustion of pectin with Mg(NO 3 ) 2 [72]. Three-dimensional carbon nanocage networks with a surface area of 1470-1927 m 2 g −1 and pore size of 15-24 nm were formed via annealing of the pitch with Mg 5 (CO 3 ) 4 and KOH treatment [73]. Porous carbon sheets with a surface area of 3145 m 2 /g and abundant micropores were formed via annealing of coal tar pitch with Mg(OH) 2 and in situ KOH activation [74]. Carbon nanocages with a surface area of 3368 m 2 /g and pore volume of 1.7 cm 3 /g were formed using Mg metal and CO 2 as precursors [75].
There are various types of SiO 2 -based templates (FSM-16, MCM-48, KIT-6, and SBA-15) that can drive porous carbon nanostructure formation with a well-defined shape and porosity. Co/N-doped mesoporous 3D carbon nanostructures with a surface area of 568 m 2 /g and pore size of~12 nm were formed using colloidal SiO 2 as a template and vitamin B12 as a template (VB12/Silica colloid) (Figure 2d-i) [65]. Under the same condition, using SBA-15 produced porous nanorods (VB12/SBA-15) with a surface area of 387 m 2 /g and pore diameter of 3.5 nm, while 2D carbon nanosheets with a surface area of 239 m 2 /g and pore size of 4.5 nm were obtained using montmorillonite silica as a template (VB12/MMT) (Figure 2d-i) [65]. N-doped porous carbon nanosheets with a surface area of 1676 m 2 /g and large pore volume of 2.13 cm 3 /g were formed via SiO 2 spheres as a template and polyaniline as the C/N source, followed by KOH activation at 850 • C [76]. Without KOH activation, the same group synthesized 2D mesoporous carbon layered nanosheets with a surface area of 582.7 m 2 /g using mesoporous SiO 2 nanoplates as a template and coal tar pitch as the carbon source, which implies the significant role of activation on the enhancement of the pore volume and surface area [77]. Si-based precursors such as tetraethylorthosilicate (TEOS) could be used to allow in situ formation of SiO 2 via the hydrolysis process and could act as in situ templates for the production of porous carbon [78]. A mesoporous porous carbon microsphere with a surface area of 659-872 m 2 /g and mesopore diameter of 3.2-14 nm was fabricated using TEOS within resorcinol-formaldehyde polymer microspheres and NH 4 OH as a catalyst, followed by annealing [79].

Ca-Based Templates
CaCO 3 can act as a template and activator due to its ability to decompose at 500 • C to release CO 2 (activator); the CaO template forms hierarchical porous carbon with multiple pores (i.e., micro-, meso-, and macropores) after the removal of CaO. Hierarchical porous carbon nanofibers (HPCNFs-3-1) with a surface area of 679 m 2 /g and pore volume of 0.41 cm 3 /g were prepared via the electrospinning of polyacrylonitrile (PAN) (as the C/N source)/N, N'-dimethylformamide (DMF)/tetrahydrofuran (THF) with nano-CaCO 3 (as a template) followed by carbonization at 800 • C for 2 h and then template removal by 1.5 M HCl (Figure 2j) [66]. Under annealing, nano-CaCO 3 decomposes to release CO 2 that produces microspores and mesopores. Then, CaO removal by HCl creates macropores and changes the ratios of PAN/DMF/THF, having an insignificant effect on the surface area and pore volume of the resultant porous carbon fibers. N-doped hierarchical porous carbon with a surface area of 1091 m 2 /g, pore volume of 0.52 cm 3 /g, and N content of 10.59% was synthesized via annealing of gelatin and graphene oxide with CaCO 3 at 900 • C and then activation by KOH. The activation substantially affected the surface area and pore volume because, without activation, they decreased to 433 m 2 g and 4.5 cm 3 /g, respectively [80]. Cornstalk rind (whole plant) was annealed with CaCO 3 and K 2 C 2 O 4 (activator) at 800 • C to hierarchical porous carbon with a surface area of 2054 m 2 /g and pore volume of 1.382 cm 3 /g [81]. The surface area (1419-2054 m 2 /g) and pore volume (0.3704-1.382 cm 3 /g) were regulated via adjustment of the activation ratio by K 2 C 2 O 4 but without activation of the surface area (482 m 2 /g) [81]. The following equations propose the reaction between CaCO 3 and K 2 C 2 O 4 : Similarly, crumpled carbon network-like nanosheets with a surface area of 1822 m 2 /g and pore volume of 4.11 cm 3 /g were formed using annealing anthracene oil with CaCO 3 as a template coupled with KOH activation [82]. Other Ca-based sources such as Ca(NO 3 ) 2 and Ca(OH) 2 could be used as templates and activation for directing the formation of porous carbon nanostructures, but they have rarely been reported.

New Templates
Various new templates such as dry ice (CO 2 ), MXene, and melamine can act as a template for producing HA-PCs. Dry CO 2 produces porous carbon with a lower surface area (100 m 2 /g), so other templates or activation are needed to enhance the surface area and porosity. Mixing phenolic resin, triblock copolymer F127, and Ti 3 C 2 T x MXene followed by annealing and chlorination formed a 2D/2D porous heterostructure with a surface area of 1021 m 2 /g and a pore volume of 58% (Figure 3a) [83]. F127 and phenolic resin molecules are assembled into spherical micelles, which penetrate the interlayers of Ti 3 C 2 T x to form Ti 3 C 2 T x -micelle@resol. Then after annealing, Ti 3 C 2 T x -micelle@resol is converted into porous carbon (Ti 3 C 2 T x -OMC), and the removal of Ti in Ti 3 C 2 T x -OMC produces MXene-derived porous carbon (MDC-OMC) [83]. The chlorination of MDC-OMC at high temperature severely impacted the increment of the surface area and pore volume of MDC-OMC [83].
Hierarchical porous carbon tunable porosity (i.e., micro-, meso-, and macroporous), a surface area of~2500 m 2 /g, and a pore volume of~11 cm 3 /g were formed via annealing of colloidal silica with sucrose at 1000 • C followed by removal of SiO 2 before being activated by CO 2 at 900 • C [85]. Honeycomb-like porous carbon nanosheets with a surface area of 2038 m 2 /g and pore volume of 1.07 cm 3 /g were synthesized from the coal tar pitch using melamine as a soft template coupled with KOH activation [86].  Hierarchical porous carbon tunable porosity (i.e., micro-, meso-, and macroporous), a surface area of ~2500 m 2 /g, and a pore volume of ~11 cm 3 /g were formed via annealing of colloidal silica with sucrose at 1000 °C followed by removal of SiO2 before being activated by CO2 at 900 °C [85]. Honeycomb-like porous carbon nanosheets with a surface area of 2038 m 2 /g and pore volume of 1.07 cm 3 /g were synthesized from the coal tar pitch using melamine as a soft template coupled with KOH activation [86].
N,O,S-enriched hierarchical porous carbon foam with a surface area of 2685 m 2 /g was made using 1,3,5-trimethyl benzene (TMB) and Poloxamer 407 (F127) as a soft template in the presence of graphene oxide, dopamine (DA), and cysteine through the freeze-drying and chemical etching method (Figure 4a) [87]. F127 and TMB are assembled on graphene oxide into micelles coated with polydopamine obtained from the self-polymerization of dopamine into the surface of spherical micelles driven by the shearing force. Regardless of the difficulty of this approach, an additional activation step by KOH is needed.
N,O,S-enriched hierarchical porous carbon foam with a surface area of 2685 m 2 /g was made using 1,3,5-trimethyl benzene (TMB) and Poloxamer 407 (F127) as a soft template in the presence of graphene oxide, dopamine (DA), and cysteine through the freeze-drying and chemical etching method ( Figure 4a) [87]. F127 and TMB are assembled on graphene oxide into micelles coated with polydopamine obtained from the self-polymerization of dopamine into the surface of spherical micelles driven by the shearing force. Regardless of the difficulty of this approach, an additional activation step by KOH is needed.

Ionic Liquids
With their thermal durability, small vapor pressure, and inbuilt heteroatoms, ionic liquids are commonly used as the precursors and templates to prepare heteroatom-doped porous carbon nanostructures. N/B-doped porous carbon nanosheets with a surface area of 2000 m 2 /g and pore volume of 2.75 mL/g were prepared using Bmp-dca and Emim-tcb ionic liquids as carbon precursors and templates with eutectics as a porogen at 1400 °C [90]. Notably, porogen could be easily recovered for reusability, and the obtained surface area (2000 m 2 /g) of the resultant carbon is superior to zeolite, activated carbons, and graphene. Moreover, the salt templating method is feasible for the large-scale and sustainability requirements.
N/S-co-doped hierarchical porous carbon nanosheets with a surface area of 575 m 2 /g and pore volume of 0.55 m 3 /g were formed using the self-assembly and carbonization of [Phne][HSO4], a protic ionic liquid, as the N/C/S source in addition to acting as a template with and soft template of OP-10 and F-127 (Figure 4b,c) [88]. Changing the ratios of [Phne][HSO4]/F-127/OP-10 leads to changing the surface areas and pore volume; meanwhile, OP-10/F-127 is easily decomposed during annealing at 700 °C to produce porous carbon foam with a good distribution of C, N (3.41%), and S (6.65%) (Figure 4c) [88]. This process is applicable for other ionic liquids with and without soft templates for the formation of HA-PC nanostructures.

Ionic Liquids
With their thermal durability, small vapor pressure, and inbuilt heteroatoms, ionic liquids are commonly used as the precursors and templates to prepare heteroatom-doped porous carbon nanostructures. N/B-doped porous carbon nanosheets with a surface area of 2000 m 2 /g and pore volume of 2.75 mL/g were prepared using Bmp-dca and Emim-tcb ionic liquids as carbon precursors and templates with eutectics as a porogen at 1400 • C [90]. Notably, porogen could be easily recovered for reusability, and the obtained surface area (2000 m 2 /g) of the resultant carbon is superior to zeolite, activated carbons, and graphene. Moreover, the salt templating method is feasible for the large-scale and sustainability requirements.
N/S-co-doped hierarchical porous carbon nanosheets with a surface area of 575 m 2 /g and pore volume of 0.55 m 3 /g were formed using the self-assembly and carbonization of [Phne][HSO 4 ], a protic ionic liquid, as the N/C/S source in addition to acting as a template with and soft template of OP-10 and F-127 (Figure 4b,c) [88]. Changing the ratios of [Phne][HSO 4 ]/F-127/OP-10 leads to changing the surface areas and pore volume; meanwhile, OP-10/F-127 is easily decomposed during annealing at 700 • C to produce porous carbon foam with a good distribution of C, N (3.41%), and S (6.65%) (Figure 4c) [88]. This process is applicable for other ionic liquids with and without soft templates for the formation of HA-PC nanostructures.
Three-dimensional porous carbon nanosheets with a surface area of 1593 m 2 g and pore volume of 0.85 cm 3 /g were formed via coal tar using 1-butyl-3-methylimidazolium tetrafluoroborate (BMIMBF 4 ) ionic liquid as a template at 800 • C and in situ KOH activation [91]. Essentially, coal tar pitch transforms into polynuclear aromatic polymers at initial heating and BMIMBF4 decomposes at 500 • C, creating 3D interconnected pores coherently distributed within obtained carbon. The activation pivots significantly in increments of the surface area and pore volume. The utilization of ionic liquid for synthesizing carbon nanostructures is not emphasized enough, and their mechanisms remain ambiguous and not profoundly investigated. In addition, the high cost, the difficulty of preparation, and the air-sensitive nature of ionic liquids are crucial barriers to their commercialization.
Deep eutectic solvents (DESs) are novel ionic liquids consisting of Lewis or Brønsted acids and alkaline eutectic mixture systems. Unlike traditional ionic liquids, they possess outstanding physiochemical merits such as ionic strength, polarity, supramolecular structure, dielectric constant, inferior vapor pressure, biodegradability, low cost, and durability without environmental sound, which enables their utilization as a carbon source, templates, and solvents in the synthesis of porous doped carbon nanostructures [92].
Notably, utilization of DESs in the fabrication of porous carbon nanostructures is rarely reported and not studied enough compared with other templates. N/O-enriched hierarchically nanoporous carbon with a surface area of 1414 m 2 /g and pore volume of 0.55 cm 3 /g was synthesized through the direct annealing of DESs of urea and ZnCl 2 with phenol-formaldehyde resin 85. This allowed a high content of N (8.09 at.%) and O (14.77 at.%) tunable surface area/pore volume via adjusting the composition of the DESs and their ratio to resin. N/O-doped hollow carbon nanorods with a surface area of 1086 m 2 /g were prepared by direct carbonization of DESs (urea, 2,5-dihydroxy-1,4benzoquinone, and ZnCl 2 ) as sources of N/O/C and as a template [93]. ZnCl 2 can act as a dehydration agent and pore modulator as it can generate ZnO as a self-template for the generation of mesopores and macropores.

MOF Template
MOFs act as sources of carbon and self-templates for various kinds of porous carbon nanostructures, owing to their unique properties such as tunable chemical compositions, porosity, and surface area [51]. During the thermal decomposition process, the metal species inside MOFs can be directly used as the template. N-doped hollow carbon nanofibers with a surface area of 443.5 m 2 /g and pore volume of 1.6 cm 3 /g were synthesized via annealing of zeolite imidazole framework (ZIF-8) nanoparticles into electrospun polyacrylonitrile (PAN) at 800 • C ( Figure 4d) [89]. The electrospinning of PAN with ZIF-8 forms nanofibers that are carbonized into porous carbon fiber ( Figure 4e) at high temperatures, while Zn in ZIF-8 can form ZnO that acts as a template, and increasing the annealing temperature leads to decreasing the surface area and pore volume. The N and O contents were about 9.39% and 4.94%, respectively, as confirmed by the elemental mapping analysis (Figure 4e). The derived BET and Langmuir surface areas were 1192 and 1678 m 2 /g, respectively, with a large pore volume of 1.06 cm 3 g −1 , which is high enough to facilitate ion transportation. Electrospinning can be used for spinning any other polymer with different MOFs to form 1D nanostructures (i.e., rods, fibers, wires) driven by various experimental parameters (i.e., applied potential and polymer concentrations). Porous carbon nanorods with a surface area of 1192 m 2 /g and pore volume of 1.06 cm 3 /g were synthesized by thermal K-MOF at 200, 450, and 800 • C. K-MOF generates K 2 O and carbonates at high temperatures, which act as in situ templates and activators. The same concept is feasible for Zn-MOF, Ni-MOF, and Co-MOF because they can also create metal-oxides that act as templates ad activators. We briefly emphasized MOF-derived porous carbon nanostructures because it was deeply discussed before in another review [51].

Biomass-Derived Carbons
Porous carbon nanostructures with various 1D, 2D, and 3D morphologies can be easily prepared via the direct annealing of biomass wastes from plant wastes, animals, insects, and microbes, which are low-cost, biodegradable, earth-abundant, unique porous structures and inherent heteroatoms, producing carbon nanostructures with outstanding electrical conductivity, massive active sites, and unique physicochemical properties. Moreover, biomass with its hierarchal porous structures (i.e., rods, sheets, nanotubes) acts as a carbon source and template for generating various porous morphologies under ambient conditions. N,S co-doped porous carbon nanosheets with a surface area of 1533 m 2 /g and pore volume of 0.92 cm 3 /g were synthesized via the one-step thermal decomposition at 400 • C and KOH activation at 850 • C of the biomass of willow catkin fibers [94].
B/N co-doped carbon nanosheets with a surface area of 416 m 2 /g and a pore volume of 0.76 cm 3 /g were created via annealing of gelatin biomass with boric acid at 900 • C [95]. The same approach is vulnerable to other biomasses. N-doped hierarchical porous carbon with a surface area of 315 m 2 /g and pore volume of 0.65 cm 3 /g was prepared from Bohai shrimp shell with KOH and CaCO 3 at 700 • C [96]. CaCO 3 acts as a self-template, while activation with KOH creates abundant micropores and mesopores [96]. The removal of CaCO 3 significantly affected surface area and porosity. Soft templates and activators can be used with biomass to create porous carbon nanostructures. Moreover, selecting biomass with abundant heteroatoms is essential to allow in situ generation of heteroatom-enriched porous carbon. Notably, various plant wastes have not yet been investigated for the formation of HA-PCs. Meanwhile, previously reported biomass-derived porous carbon nanostructures have rarely been studied for CO 2 RR compared with other applications such as supercapacitors.

CO 2 RR Pathways
CO 2 is a valuable molecule, and CO 2 RR can generate wide ranges of high value-added hydrocarbons and gases (i.e., CO and CH 4 , CH 3 OH, C 2 H 5 OH, and HCOOH) driven by the applied potentials (Table 2) and electron transfer number (Scheme 1). This is based on the catalyst compositions and experimental parameters (i.e., applied potential, cell type, and electrolyte) via different CO 2 RR pathways. The CO 2 RR pathways on carbon-based catalysts generally include multiple interactive steps of CO 2 adsorption and charge transfer, including electron and proton transfer in addition to product dissociation, which comprises migration of the products from the catalyst surface. The CO 2 molecule is very stable thermodynamically with a great dissociation energy of~750 kJ/mol for C=O (which needs significantly greater energy), which is greater than that of other carbon-based chemical bonds, such as C-H (~430 kJ/mol) and C-C (~336 kJ/mol). Notably, the applied potential in CO 2 RR is often much more negative than the equilibrium potential, resulting in high overpotential. This is because the CO 2 adsorption comprises the rearrangement of a linear CO 2 molecule to a bent radical anion via one-electron transfer to a CO 2 molecule (i.e., . During this process, the C=O bond is strongly unstable, and the electrons are shared between the CO 2 and the catalyst. The obtained CO 2 •− radical is highly reactive and easily reacts with H 2 O in the electrolyte to produce HCO 2 • that is perturbed and tends to transform to HCO*, which is consequently released from the active catalyst sites. These multiple electron transfers and protonation processes occur on carbon-based catalysts to produce various products. The E 0 for most CO 2 RR half-reactions is close to 0 V vs. RHE (Table 2), so the undesired side products because of the HER (2H + + 2e − → H 2 , E 0 , 0.42 V) result in the reduced electrocatalytic CO 2 RR activity of carbon-based catalysts. The CO 2 RR pathways can lead to the formation of various products driven by electron transfer and protonation under different applied potentials (Scheme 1). Experimental studies using in situ analysis and computational calculations (i.e., density functional theory (DFT)) were used to confirm the CO 2 RR mechanisms and pathways. Mainly after the CO 2 activation, protonation can occur for O to form *COOH or C to produce *OCHO. HA-PC catalysts *COOH pathways are preferred over *OCHO, so they usually produce CO in addition to HCOOH via the two-electron reduction process with rare CH 3 CH 2 OH. This is due to the absence of stable adsorbed states for the *HCOOH and *CO on the heteroatom-doped carbon surface for facilitating protonation of the *COOH intermediate. Despite the great progress in CO 2 RR, its mechanism and the boundary between each reaction pathway remain ambiguous owing to the absence of effective methods to detect the intermediates.  to confirm the CO2RR mechanisms and pathways. Mainly after the CO2 activation, protonation can occur for O to form *COOH or C to produce *OCHO. HA-PC catalysts *COOH pathways are preferred over *OCHO, so they usually produce CO in addition to HCOOH via the two-electron reduction process with rare CH3CH2OH. This is due to the absence of stable adsorbed states for the *HCOOH and *CO on the heteroatom-doped carbon surface for facilitating protonation of the *COOH intermediate. Despite the great progress in CO2RR, its mechanism and the boundary between each reaction pathway remain ambiguous owing to the absence of effective methods to detect the intermediates.

Half Reactions of CO2R
The proposed CO2RR pathways on HA-PCs.

Advantages of Heteroatoms
There are various metal-based catalysts for CO2RR; however, their high cost, earth scarcity, and ability for the HER are critical barriers to large-scale applications and sustainability requirements. To this end, noble metals (i.e., Pt, Pd, Ru, and Au) possess great activity for the HER, while Au, Ag, and Sn, for example, produce mainly C1 (e.g., CO and HCOOH) rather than C2 products via a two-electron transfer pathway. Cu/Cu-O forms low-carbon hydrocarbons and oxygenates (i.e., CO with relatively high overpotentials).

Advantages of Heteroatoms
There are various metal-based catalysts for CO 2 RR; however, their high cost, earth scarcity, and ability for the HER are critical barriers to large-scale applications and sustainability requirements. To this end, noble metals (i.e., Pt, Pd, Ru, and Au) possess great activity for the HER, while Au, Ag, and Sn, for example, produce mainly C1 (e.g., CO and HCOOH) rather than C2 products via a two-electron transfer pathway. Cu/Cu-O forms low-carbon hydrocarbons and oxygenates (i.e., CO with relatively high overpotentials). HA-PCs are promising electrocatalysts for the CO 2 RR with remarkable catalytic activity, long durability, and high selectivity. Distinct from metal-based catalysts, porous carbon materials possess numerous advantages such as tailorable structures/properties, rich specific surface chemistry, excellent surface area to volume ratio, thermal-chemical-physical durability, and low toxicity.
Moreover, C can be easily synthesized in high yield from inexpensive and natural abundant resources that meet the sustainability requirements. During CO 2 RR, HA-PCs tune the adsorption of CO 2 , induce electrolyte-electrode interaction, provide massive catalytic active sites, maximize atomic utilization, and tolerate the adsorption of intermediates and products. The HER activity of porous carbon catalysts is inferior, so the undesired effect of HER during CO 2 RR is neglected. The CO 2 RR is carried out in aqueous electrolyte solutions, so the electrode's wettability or hydrophilicity is crucial to controlling the transportation of hydrated CO 2 to the active sites. Heteroatom dopants enhance the hydrophilicity of the C-electrode, which can promote the accessibility of reactants to the active sites; make active sites more accessible, thus maximizing atomic utilization during the CO 2 RR. Integration of heteroatoms (i.e., P, N, O, B, and S) into carbon skeleton structures is essential to detect the limitation of electrical conductivity of carbon-based catalysts to empower their CO 2 RR activity. This is due to electron-donating or electron-withdrawing properties of heteroatoms that modulate the electronic characteristics.

Nitrogen (N) Configuration and Effects
The N atom is an ideal heteroatom for modulating the properties of carbon due to its size (155 pm) being close to that of the carbon atom (170 pm). In addition, N has a larger electronegativity (3.04) than carbon (2.55), so N-doped C (N-C) can attract electrons, create multiple active sites, and augment the electronic/ionic conductivity of carbon. The N atom in the C skeleton structure is mainly pyridinic (398.5 eV), pyrrolic/pyridonic (400.1 eV), quaternary, or graphitic (401.1 eV), and pyridine-N-oxide (403.2 eV), that can be easily identified by X-ray photoelectron spectroscopy (XPS) [99]. These N species (Scheme 2a) are highly active sites for CO 2 RR and other catalytic applications. N-doping introduces Lewis basicity on C's surface that endows the CO 2 adsorption significantly.

Boron (B) Configuration and Effects
The smaller size of the B atom (85 pm) and its lower electronegativity (2.04) than C lead to the retention of the planar structure of C. The XPS showed two in-plane binding structures in B-doped C, including graphitic B at 200.5 eV, B atoms substituting for C atoms in the hexagonal ring, and "boron silane" B at 198.5 eV, referring to the place of B in the co-conjugated system (Scheme 2b) [99]. B-doping persuades charge polarization in the C framework, stabilizes the negatively polarized O atoms in CO 2 , and consequently enhances the chemisorption of CO 2 on C during the CO 2 RR process due to the relatively large electropositivity between B and C atoms.

Sulfur (S) Configuration and Effects
Stone-Wales defects ease the formation of S-doping into carbon. The relatively larger size of the S atom (180 pm) and higher electronegativity (2.58) than C lead to improving C's electrical conductivity and provide a higher spin density, edge strain, and charge delocalization. There are four types of S dopant in C, including S1, S2/S3, S4/S5, and S6 for adsorption of S on the C surface, the substitution of C by S at the edges, formation of the S/S oxide at the edges, and S-containing ring connecting sheets, respectively, as confirmed by DFT on a graphene model (Scheme 2c) [100]. The formation energies for these species are different, but S1 is the most durable structure.

Phosphorus (P) Configuration and Effects
The P atom has a lower electronegativity (2.19) and greater electron-donating ability than the C atom and generates positive charges on the P dopant and negative charges on positively charged C, introducing new functional groups (P + -C − ) which change the structure of C, resulting in accelerated charge transfer and the reduced binding energy of intermediates. Also, the P atom, with its size (195 pm) being larger than C, leads to various defects in C skeleton structures, which act as active sites during CO 2 RR. Usually, P-doping occurs through the substitution of C by P. The XPS can detect P dopant at 130.2 eV, P-C at (132.5 eV), and P-O at (134.0 eV).

Oxygen (O) Configuration and Effects
The oxygen (O) atom has a lower atom size (152 pm) than C but a greater electronegativity (3.44 pm) that enhances positive charges on C to form (C + -O − ) after doping, which can attract electrons, create multiple active sites, and enhance the electronic/ionic conductivity of C. Also, introducing O into the carbon framework creates abundant structural defects and modulates the inherent electronic structure, allowing tuning adsorption/desorption of reactants and intermediates during CO 2 RR. Moreover, O-doping creates various functional groups (i.e., C-O, C=O, and C-OH) which can promptly activate oxidants allowing in situ generations of reactive oxygen species, activating H 2 O 2 to in situ form oxygenated species that facilitate the CO 2 RR kinetics. The O-doping occurs via partial substitution of C with O to form C-O at (532 eV) and C=O at 533 eV species, as confirmed by the XPS, which are highly active sites for CO 2 RR. The FTIR can also easily detect the O-doping.

Single Heteroatom-Doped Porous Carbon Materials
Because the stable sp 2 or sp 3 hybridized carbon atoms cannot activate CO 2 molecules, pristine carbon materials possess poor electrocatalytic activity for CO 2 RR. Therefore, incorporating heteroatoms with different electronegativity into carbon provides a great opportunity to modulate the electronic properties and thus improve their catalytic performance [50,102]. Porous carbons have unique advantages such as high specific surface area and tunable pore structure, which is beneficial for mass transfer and abundant active sites, thereby increasing the catalytic activity [103][104][105]. According to the reported literature, the frequently used single dopant can be classified into nitrogen, boron, sulfur, and fluorine [33]. Table 3 shows a detailed comparison of the electrocatalytic performance of binary HA-PCs toward CO 2 RR.

Nitrogen-Doped Porous Carbon Materials
Nitrogen is one of the most widely studied dopants for carbon materials, based on the fact that the nitrogen atom has a similar atomic size to the carbon atom and higher electronegativity of 3.04 compared to the carbon atom's 2.55. N-doping can induce more electrons into the carbon matrix and thus improve the electrical conductivity. Moreover, more active sites can be created after the N-doping [106][107][108]. As displayed in Figure 5A, the N dopants can exist in four types: pyridinic N, pyrrolic N, graphitic N, and oxidized N [109]. Liu et al. developed a series of N-doped carbon catalysts with tunable types and contents of nitrogen dopants to uncover the correlation between N species and catalytic performance toward CO 2 RR [110]. Electrochemical tests coupled with X-ray photoelectron spectroscopy identified that the CO 2 RR activity is proportional to the content of pyridinic N, whereas no noticeable relevance was observed on other N species. Moreover, the free energy calculated by density functional theory (DFT) calculations in Figure 5B revealed that the COOH could interact with pyridinic N in optimal bonding strength, benefiting the reduction of CO 2 to form *COOH, and further to *CO. Apart from the pyridinic N content, porous carbon materials' pore structure also affects the electrochemical performance [106,109,[111][112][113][114][115]. For instance, N-doped nanoporous carbon sheets were synthesized by Yao et al. via a hydrothermal reaction and calcining Typha in NH 3 [116]. It was found that the calcination temperature has a significant effect on the pore structure and N atom type. The optimal sample possesses the highest surface areas and pore volume, exposing abundant accessible pyridinic N, thereby delivering a much higher selectivity of 90% for CO at a lower overpotential of −0.31 V. However, there remains controversy about the critical active N sites for CO 2 RR [117]. For example, Liu and co-workers verified that the stable graphitic nitrogen atoms restricted in the micropores for coal-based metal-free electrocatalysts could effectively convert CO 2 into CO [118]. Huang's group demonstrated that pyridinic and graphitic N are the active sites for CO 2 RR by calcinating oxygen-rich Zn-MOF-74 precursors at different temperatures [119].
Up to now, the reported products for N-doped porous carbon materials toward CO 2 RR mainly focus on C1 compounds such as CO, CH 4 , and HCOOH [121,122]. It remains a challenge to stabilize the active C1 intermediates for their coupling to generate multicarbon products. It is worth mentioning that Song et al. developed a nitrogen-doped ordered cylindrical mesoporous carbon (denoted as c-NC) as a high-efficient catalyst for the electroreduction of CO 2 to ethanol with nearly 100% selectivity [120]. To reveal the unique structural effect of c-NC, an inverse mesoporous N-doped carbon with a similar pore structure and N content, namely i-NC, was synthesized as a control. It can be seen from Figure 5C that both c-NC and i-NC electrodes exhibit an obvious reduction peak in CO 2 saturated 0.1 M KHCO 3 solution, whereas no reduction peak was observed in Ar, indicating that the CO 2 RR occurs on c-NC and i-NC. Figure 5D shows the Faradaic efficiency (FE) of CO 2 RR on c-NC and i-NC catalysts at a potential between −0.40 and −1.00 V. The main product for c-NC and i-NC catalysts is ethanol from CO 2 RR and H 2 from the hydrogen evolution reaction (HER), accompanied by a small amount of byproduct CO. This demonstrates that the mesoporous structure of both c-NC and i-NC contributes to the generation of ethanol. For c-NC, ethanol is the dominant product at the potential between −0.40 and −0.90 V, and the ethanol FE attained the maximum of 77% at −0.56 V. Meanwhile, the competitive reaction over c-NC for CO 2 electroreduction to CO can be neglected. By contrast, the CO 2 electroreduction to ethanol over i-NC is only dominated at the potential of −0.40 and −0.50 V, and the maximum ethanol FE reaches 44% at −0.50 V. After that, the HER becomes dominant, accompanied by a substantial amount of CO. Electrochemical impedance spectroscopy (EIS) was measured at −0.56 V vs. RHE to determine the charge transfer resistance (R ct ) on c-NC and i-NC catalysts ( Figure 5E). The smaller R ct value for c-NC (3.8 Ω) than that for i-NC (8.5 Ω)) implies the easier transportation of electrons to the cylindrical surface of c-NC. DFT calculations were performed to check the possible CO 2 RR reaction pathways. Both pyridinic and pyrrolic N sites can promote the adsorption/activation of CO 2 molecules and generate CO* intermediates ( Figure 5F). The calculated reaction energy for CO* formation on pyridinic and pyrrolic N sites is −1.68 and 0.12 eV, respectively, implying the preferential formation of CO* on pyridinic N sites. The CO* intermediates can be stabilized by the cylindrical surface of c-NC with high electron density, thereby restricting the CO generation. The dimerization of CO* intermediates can proceed to form OC-CO* intermediates with reaction energies of −1.31 and −0.34 eV on pyridinic and pyrrolic N sites, respectively, signifying the favorable C-C bond formation on pyridinic N sites. the stable graphitic nitrogen atoms restricted in the micropores for coal-based metal-free electrocatalysts could effectively convert CO2 into CO [118]. Huang's group demonstrated that pyridinic and graphitic N are the active sites for CO2RR by calcinating oxygen-rich Zn-MOF-74 precursors at different temperatures [119]. Up to now, the reported products for N-doped porous carbon materials toward CO2RR mainly focus on C1 compounds such as CO, CH4, and HCOOH [121,122]. It remains a challenge to stabilize the active C1 intermediates for their coupling to generate multi-carbon products. It is worth mentioning that Song et al. developed a nitrogen-doped ordered cylindrical mesoporous carbon (denoted as c-NC) as a high-efficient catalyst for the electroreduction of CO2 to ethanol with nearly 100% selectivity [120]. To reveal the unique structural effect of c-NC, an inverse mesoporous N-doped carbon with a similar pore structure and N content, namely i-NC, was synthesized as a control. It can be seen from Figure 5C that both c-NC and i-NC electrodes exhibit an obvious reduction peak in CO2 saturated 0.1 M KHCO3 solution, whereas no reduction peak was observed in Ar, indicating that the CO2RR occurs on c-NC and i-NC. Figure 5D shows the Faradaic efficiency (FE) of CO2RR on c-NC and i-NC catalysts at a potential between −0.40 and −1.00 V. The main product for c-NC and i-NC catalysts is ethanol from CO2RR and H2 from the hydrogen evolution reaction (HER), accompanied by a small amount of byproduct CO. This demonstrates that the mesoporous structure of both c-NC and i-NC contributes to the generation of ethanol. For c-NC, ethanol is the dominant product at the potential between −0.40 and −0.90 V, and the ethanol FE attained the maximum of 77% at −0.56 V. Moreover, the electron-rich cylindrical surface can accelerate the subsequent single and multiple proton-electron transfers to form the OC-COH* intermediate and ethanol. In short, the cylindrical surface with a high electron density and highly active N sites for c-NC synergistically benefits the highly efficient and highly selective production of ethanol. In addition, Yuan and co-workers reported the use of nitrogen-doped porous biochar from plant moss to catalyze CO 2 RR into CH 4 , CH 3 OH, and CH 3 CH 2 OH at a high current density and low overpotential [123]. Except for the above-mentioned C2 product, Li et al. first revealed that C3 hydrocarbons could be formed during CO 2 RR when nitrogen sites are situated close to each other in the micropore space [124]. Though great progress has been made for N-doped porous carbon materials, N sites' high spin density also benefits the competitive HER, resulting in moderate FE and low partial current density.

Other Metal-Free Heteroatom (S, F, or B)-Doped Porous Carbon Materials
Similar to N atoms, the doping of F and S atoms with electro-withdrawing or B atoms with electron-donating behaviors can also tailor the electronic structure of adjacent carbon atoms, thus improving the electrocatalytic performance toward CO 2 RR [125].
Nitrogen-doped carbon nanotubes (NCNTs) with an average diameter of 30 nm and N content of 5.0% were formed via the liquid chemical vapor deposition (CVD) method using acetonitrile and dicyandiamide with ferrocene at 850 • C under Ar/H 2 . NCNTs promoted the CO 2 RR with a maximum FE of CO of nearly 80% at −0.78 V and an overpotential of −0.26 V, which is comparable to Au and Ag nanoparticles but with a lower overpotential in 0.1 M KHCO 3 [126]. This was significantly higher than that of N-free CNTs revealing inferior CO 2 RR activity (3.5% FE CO ). NCNTs remain stable at −0.8 V for 10 h without any obvious degradation, and the FE varies slightly around 80% [126]. For instance, a fluorine-doped cage-like porous carbon (F-CPC) electrocatalyst was synthesized by a polymer-derived method using SiO 2 spheres as templates ( Figure 6A) [127]. The transmission electron microscope (TEM) image in Figure 6B clearly shows the hollow cagelike structure of F-CPC with a thin carbon shell, in which the bright dots marked by yellow circles represent the rich mesopores on the surface. Elemental mapping images reveal that the carbon and fluorine elements homogeneously distribute throughout the F-CPC ( Figure 6C,D). There is a single peak at 689.8 eV in the X-ray photoelectron spectroscopy (XPS) spectra of F 1s, suggesting the presence of a semi-ionic C-F bond in the F-CPC ( Figure 6E). Compared with the covalent C-F bond, the semi-ionic C-F bond is anticipated to promote the CO 2 RR. The F-CPC catalyst exhibits a maximum CO FE of 88.3% at −1.0 V vs. RHE among all samples ( Figure 6F), originating from the novel structure and morphology. Moreover, the superior stability of F-CPC was confirmed by chronoamperometric curves at −0.9 V vs. RHE ( Figure 6G). Of note, F-CPC can maintain 97% of the initial current density and keep CO FE stable for 12 h. It turns out that the nanocage structure of F-CPC can generate an enhanced electrostatic field and increase the K + ion concentration, thereby lowering the thermodynamic energy barrier for CO2RR. In another report [128], a fluorine interlayer doped carbon (FC)  [127]. Copyright 2020, American Chemical Society.  Of note, F-CPC can maintain 97% of the initial current density and keep CO FE stable for 12 h. It turns out that the nanocage structure of F-CPC can generate an enhanced electrostatic field and increase the K + ion concentration, thereby lowering the thermodynamic energy barrier for CO 2 RR. In another report [128], a fluorine interlayer doped carbon (FC) catalyst was obtained by the facile pyrolysis of the precursor's mixture, and the FC was able reach the maximum FE for CO of 89.6% at −0.62 V. DFT calculations revealed that fluorine interlayer doping can activate neighbor carbon atom defects and contribute to the interaction of COOH* with activated carbon, which is considered as the rate-determining step for CO 2 -to-CO conversion. Likewise, S-doping can also result in higher spin density and charge delocalization owing to the dissimilar electronegativity of sulfur (2.58) and carbon (2.55), which are believed to enhance the electrocatalytic activity for CO 2 RR. S-doped porous carbon nanosheets (CPSs) were prepared by the annealing of poly (4-styrene sulfonic acid-co-maleic acid) sodium salt at 800 • C under N 2 , which showed a higher CO 2 RR current density (7.2 mA/cm −2 ) than that of N-doped CPSs (CP-SNs) (7.2 mA/cm −2 ) [130]. The maximum FE for CO (FE CO ) and CH 4 on the CPSs was about (2.0 % and 0.1%) at −0.99 V vs. RHE relative to the CPSNs (11.3% and 0.18%) due to S and/or N dopants, which stabilize the CO 2 − and COOH* intermediates that promote CO2RR to CO and CH 4 . The durability studies showed that the CPSs maintained only 65% of their FE CO (1.3%) after 2 h, while the CPSNs kept around 72.7% of their FECO (8%). With the relatively large difference between the electronegativity of B and C atoms, B-doping can induce charge polarization and make the carbon framework suitable for adsorbing CO 2 molecules. For instance, B-doped graphene (BG) was synthesized by catalyst by heating the graphene oxide and boric acid (1/5 wt ratio) at 900 • C under Ar, which showed higher CO2RR activity with a current density of~6 mA/cm 2 at 1.6 V than undoped graphene (~1.6 mA/cm 2 ), and Bi (~2.4 mA/cm 2 ) in 0.1 M KHCO 3 can reach an FE for formic acid of 66% at −1.4 V vs. SCE during CO 2 RR [129]. BG mainly allowed the CO 2 RR to formate with an FE of 66% at −1.4 V, which was substantially higher than Bi (FE of 20%). BG remains stable for 4 h without any significant loss in the CO 2 RR activity. B-doped diamond (BDD) thin films were grown on Si(111) wafers by the microwave plasma-assisted chemical vapor deposition (MPVCD) method at 5 kW using B(OCH3) (B-source) and acetone (C-source) with B/C (1/1 atomic ratio) [131]. BDD showed a current density of 0.3 mA/cm 2 at 2.2 V vs. Ag/AgCl in an electrolyte of methanol solution and tetrabutylammonium perchlorate (MeOH-TBAP). Moreover, BDD allowed the CO 2 RR to produce formaldehyde, formic acid, and H 2 with a maximum FE of 74% at −1.7 V, 15% at −1.5 V, and 1.1% at <−1.7 V, respectively [131]. Interestingly, BDD revealed a higher CO 2 RR activity with a greater FE for formaldehyde, formic acid, and H 2 in MeOH-TBAP electrolyte relative to water (0.1 M NaCl) and seawater, respectively. Notably, mono heteroatom-doped carbon-based catalysts for CO 2 RR are rarely reported and are not studied enough (Table 3). Also, other reports did not systemically study the activity and durability as a function of dopant amount or in different electrolytes.

Binary Heteroatom-Doped Porous Carbon Materials
In consideration of the unsatisfying catalytic performance for single heteroatom-doped porous carbon materials, dual heteroatom co-doping may regulate the chemical properties in a wide range and bring great promise toward CO 2 RR by virtue of the synergistic electronic interactions between different dopants. Table 4 shows a detailed comparison of the electrocatalytic performance of binary HA-PCs toward CO 2 RR.

Nitrogen, Sulfur Co-Doped Porous Carbon Materials
Although there are few reports about the effect of sulfur dopants into porous carbon materials on CO 2 RR, the co-doping of nitrogen and sulfur has been extensively investigated over the past few years [40,43,130,[132][133][134][135][136]. Yang et al. reported N, S co-doped hierarchically porous carbon nanofiber (NSHCF) membranes as high-efficiency catalysts for electrochemical conversion of CO 2 to CO with 94% Faradaic efficiency at −0.7 V vs. RHE [133]. In their synthesis, NSHCF900 was obtained by electrospinning the mixture of ZIF-8 nanoparticles, trithiocyanuric acid (TA), and polyacrylonitrile (PAN), followed by the carbonization at 900 • C in Ar ( Figure 7A). In contrast, the NHCF900 without sulfur doping can only achieve a CO FE of 63% toward CO 2 RR, highlighting the key role of S species in promoting electrochemical activity. Moreover, DFT calculations showed that the Gibbs free energy of *COOH on pyridinic N adjacent to the carbon-bonded S atom is effectively decreased compared to that on pure pyridinic N atoms ( Figure 7B). This is likely due to the greater spin density and charge delocalization arising from S atom doping. Following this method, Li and coworkers synthesized N, S co-doped hierarchically porous carbon (NSHPC) by pyrolysis of glucosamine hydrochloride and thiocyanuric acid precursor using SiO 2 as a hard template, and they were able to obtain a maximum CO FE of 87.8% at −0.6 V vs. RHE [134]. The N, S co-doped high-surface-area carbon materials (SZ-HCN) were developed by one-step pyrolysis of N-containing polymer and S powder [136]. A partial current density of 5.2 mA/cm 2 at the overpotential of 0.490 V and a maximum CO FE of 93% at −0.6 V for CO 2 RR was achieved on SZ-HCN, which is superior to those on the single N-doped carbon counterpart. beneficial for forming highly active pyridinic N while suppressing graphitic N. The CO FE for each catalyst at various potentials is shown in Figure 7D. NS-C-900 can catalyze CO2RR at a smaller overpotential but with a larger CO FE relative to N-C-900, showcasing the improved reactivity and selectivity after sulfur doping.
Furthermore, it was found that the catalytic activity of NS-C layers strongly depends on the annealing temperature, among which the NS-C annealed at 900 °C can yield the highest CO FE of 92%. This is mainly attributed to doped N and S species' optimal content and structure, which can expose more active sites and accelerate mass transport. DFT calculations were further employed to study the effect of sulfur doping on the inherent activity of various N dopants. They found that the distance between the S atom and pyridinic N is proportionally related to the enhancing effect of sulfur atoms. It was speculated that introducing S atoms can increase the spin density of N-C, which is more conducive to promoting electron transfer and COOH* adsorption, resulting in superior catalytic ability at a lower overpotential. Similarly, Li et al. revealed the relationship between the dispersion of S and N groups in porous carbon with the CO FE, verifying the enhancing effect of sulfur doping on highly active pyridinic N sites for CO2RR [132].

Nitrogen, Phosphorus Co-Doped Porous Carbon Materials
In view of the larger difference between heteroatom P and N in electronegativity, dual-doping into porous carbon materials provides more options to boost the electrocatalytic performance toward CO2RR [137][138][139][140]. Chen et al. fabricated N, P co-doped carbon materials (NPCM-1000) using aniline monomer and phytic acid as nitrogen, carbon, and phosphorus sources via one-pot pyrolysis at 1000 °C ( Figure 8A). As a reference, single Ndoped carbon materials (NCM-1000) were synthesized by replacing phytic acid with HCl. The NPCM-1000 exhibits a similar structure to NCM-1000 but with a higher surface area, pyridine N content, and defects, as revealed by TEM, N2 adsorption-desorption measure- In another study [40], Pan et al. concluded that sulfur addition could significantly boost the electrochemical activity and selectivity for CO 2 RR of N-doped carbon catalysts. A layer-structured carbon nitride-templated pyrolysis strategy synthesized the N, S dualdoped carbon (NS-C) layers. Thiourea and citric acid were used as N, S sources and C sources, respectively. For comparison, the NS-C samples (denoted as NS-C-800, NS-C-900, and NS-C-1000) were prepared at different annealing temperatures of 800, 900, and 1000°C to tune the doped N, S content, respectively. In addition, S-free N-C (denoted as N-C-900) layers were also obtained under identical synthesis conditions to those of NS-C-900, except for replacing thiourea with urea. It is notable from Figure 7C that the S-doping is beneficial for forming highly active pyridinic N while suppressing graphitic N. The CO FE for each catalyst at various potentials is shown in Figure 7D. NS-C-900 can catalyze CO 2 RR at a smaller overpotential but with a larger CO FE relative to N-C-900, showcasing the improved reactivity and selectivity after sulfur doping.
Furthermore, it was found that the catalytic activity of NS-C layers strongly depends on the annealing temperature, among which the NS-C annealed at 900 • C can yield the highest CO FE of 92%. This is mainly attributed to doped N and S species' optimal content and structure, which can expose more active sites and accelerate mass transport. DFT calculations were further employed to study the effect of sulfur doping on the inherent activity of various N dopants. They found that the distance between the S atom and pyridinic N is proportionally related to the enhancing effect of sulfur atoms. It was speculated that introducing S atoms can increase the spin density of N-C, which is more conducive to promoting electron transfer and COOH* adsorption, resulting in superior catalytic ability at a lower overpotential. Similarly, Li et al. revealed the relationship between the dispersion of S and N groups in porous carbon with the CO FE, verifying the enhancing effect of sulfur doping on highly active pyridinic N sites for CO 2 RR [132].

Nitrogen, Phosphorus Co-Doped Porous Carbon Materials
In view of the larger difference between heteroatom P and N in electronegativity, dualdoping into porous carbon materials provides more options to boost the electrocatalytic performance toward CO 2 RR [137][138][139][140]. Chen et al. fabricated N, P co-doped carbon materials (NPCM-1000) using aniline monomer and phytic acid as nitrogen, carbon, and phosphorus sources via one-pot pyrolysis at 1000 • C ( Figure 8A). As a reference, single N-doped carbon materials (NCM-1000) were synthesized by replacing phytic acid with HCl. The NPCM-1000 exhibits a similar structure to NCM-1000 but with a higher surface area, pyridine N content, and defects, as revealed by TEM, N 2 adsorption-desorption measurement, X-ray photoelectron spectroscopy, and Raman spectra, respectively. The electrochemical tests were conducted using a standard three-electrode system using a CO 2saturated 0.5 M NaHCO 3 solution. As expected, the NPCM-1000 exhibits a higher onset potential (−0.38 V) than NCM-1000 (−0.59) V ( Figure 8B). The CO FEs at various potentials of NPCM-1000 and NCM-1000 are displayed in Figure 8C, in which the maximum FE for NPCM-1000 and NCM-1000 is 92% and 14% at −0.55 V, respectively. Moreover, NPCM-1000 shows much higher partial current densities of CO (j co ) than NCM-1000 ( Figure 8D). The Tafel curve is generally used to describe the reaction kinetics of CO 2 RR. The Tafel slope of 122 mV/dec for NPCM-1000 implies the formation of COOH* is the rate-determining step, while for NCM-1000 it is the CO 2 molecular adsorption and desorption ( Figure 8E). Based on DFT calculations, the N, P co-doping can synergistically promote the formation of COOH*. Furthermore, the reaction barrier of CO 2 activation can be more effectively reduced for NPCM-1000 than that for NCM-1000, as calculated from Figure 8F,G. Recently, Liang and co-workers demonstrated a porous N, P dual-doped carbon nanosheet catalyst for CO 2 RR which can attain a high CO FE of 88% and good stability for 27 h at a low overpotential [140]. In addition, they found that the introduction of P can adjust the electronic structure of pyridinic N to hinder the adsorption of *H and contribute to the higher selectivity of CO 2 -to-CO.
formation of COOH*. Furthermore, the reaction barrier of CO2 activation can be more effectively reduced for NPCM-1000 than that for NCM-1000, as calculated from Figure 8F,G. Recently, Liang and co-workers demonstrated a porous N, P dual-doped carbon nanosheet catalyst for CO2RR which can attain a high CO FE of 88% and good stability for 27 h at a low overpotential [140]. In addition, they found that the introduction of P can adjust the electronic structure of pyridinic N to hinder the adsorption of *H and contribute to the higher selectivity of CO2-to-CO.

Nitrogen, Boron Co-Doped Porous Carbon Materials
Since N atoms have a larger electronegativity and B atoms have a smaller electronegativity than C atoms, N/B co-dopants may modulate the electronic structure and create an unexpected effect on CO2RR [33]. As reported previously [141], B, N co-doped nanodiamond (BND) was an efficient and stable electrocatalyst for CO2RR to ethanol. BND revealed a high FE of 93.2% at −1.0 V vs. RHE due to the synergistic effect of B and N codopants. Then, various HA-PCs co-doped with B and N were reported for the CO2RR

Nitrogen, Boron Co-Doped Porous Carbon Materials
Since N atoms have a larger electronegativity and B atoms have a smaller electronegativity than C atoms, N/B co-dopants may modulate the electronic structure and create an unexpected effect on CO 2 RR [33]. As reported previously [141], B, N co-doped nanodiamond (BND) was an efficient and stable electrocatalyst for CO 2 RR to ethanol. BND revealed a high FE of 93.2% at −1.0 V vs. RHE due to the synergistic effect of B and N co-dopants. Then, various HA-PCs co-doped with B and N were reported for the CO 2 RR [142,143]. Zhao's group has reported the integrated design of a N, B co-doped three-dimensional hierarchical porous carbon network with a high doping level by a salt-sugar method [144].
The as-obtained catalyst displays a high CO 2 -to-CO FE of 83% at a low overpotential of 290 mV and good stability over 20 h. Based on the physical and electrochemical characterization, the superior activity and selectivity are first attributed to the unique porous structure, including macropores, mesopores, and micropores, which can offer larger surface areas and more active sites for CO 2 adsorption. Moreover, the N-doping can accelerate the conversion of CO 2 into the CO 2 * intermediate, while the B atoms can facilitate the capture of CO 2 by bonding to the O atoms of the CO 2 * intermediate, and then the conversion of COOH* into CO*. B, N co-doped mesoporous carbon (BNMC) was synthesized through carbonization of the mixed precursors using glucose as the carbon source, urea and dicyandiamide as the nitrogen source, and boric acid as the boron source along with silica as a template ( Figure 9A) [142]. The as-synthesized BNMC annealed at 1000°C (BNMC-1000) possesses a porous surface, as seen from the scanning electron microscopy (SEM) image in Figure 9B. The TEM image of BNMC-1000 reveals the rich mesopores with an average diameter of 25 nm ( Figure 9C). For comparison, control experiments of BNC-1000 without mesopores, NMC-1000 without B-doping, and BMC-1000 without Ndoping were conducted to explore the effect of the porous structure and heteroatom doping. As shown in Figure 9D,E, the BNC-1000 reveals a very low current density and CO FE compared with NMC-1000 and BNMC-1000, indicating the critical role of the mesoporous structure in enhancing CO 2 RR. Furthermore, the current density and CO FE of BNMC-1000 is larger than those of NMC-1000, demonstrating that the N, B dual-doping contributes to improving CO 2 RR relative to single N-doping. They also confirmed that BMC-1000 could reduce CO 2 to formic acid with a different electrochemical selectivity from NMC-1000 and BNMC-1000. The Tafel slope of BNMC-1000 was calculated to be 128 mV/dec, which is smaller than that of NMC-1000 (141 mV/dec), indicating the favorable reaction kinetics on BNMC-1000 through N, B dual-doping ( Figure 9F). DFT calculations reveal that the coupling effect between N and B atoms can tune the projected density of states of adjacent carbon active sites, thereby generating an optimal adsorbed energy of *COOH and *CO on the carbon surface.

Ternary Heteroatom-Doped Porous Carbon Materials
Introducing multiple heteroatoms with different electronegativities and sizes to porous carbon materials can modify the local electronic properties, thereby boosting the electrochemical CO 2 RR. Table 4 shows a detailed comparison of the electrocatalytic performance of ternary HA-PCs toward CO 2 RR. Yang et al. presented a facile synthesis of N, P, S ternary heteroatom-doped carbon (NSP-HPC) via the H 2 SO 4 -H 3 PO 4 binary-acids activation method [145]. The dual heteroatom-doped NS-HPC, NP-HPC, and single-doped N-HPC were also prepared as a control. In a conventional H-cell, NSP-HPC exhibits the lowest onset potential of −0.38 V among the four samples, which corresponds to a smaller overpotential of 270 mV ( Figure 9G). Moreover, NSP-HPC demonstrates the highest CO FE among the control samples at various potentials ( Figure 9H). The maximum CO FE for NSP-HPC is 92%, achieved at −0.7 V, while that for NP-HPC, NS-HPC, and N-HPC is 83%, 90%, and 56%, respectively. More importantly, the current density and CO FE of NSP-HPC can keep constant for 50 h at −0.7 V, reflecting its robust stability. In situ Raman spectroscopy reveals that *COOH is the key intermediate in CO 2 -to-CO conversion. Based on the DFT calculations, the free energy diagrams of CO 2 RR for each catalyst are given in Figure 9I. During the CO 2 -to-CO conversion, the free energy barrier of *COOH is 1.61 eV for N-doped carbon. Once P or S atoms are introduced, the free energy barrier decreases to 1.42 and 0.54 eV, respectively. The N, S, and P doping reveals the lowest *COOH formation energy of 0.05 eV, indicating that the synergistic coupling effect among multiple heteroatoms could benefit the enhanced CO 2 RR. Furthermore, the well-defined hierarchically porous structure of NSP-HPC also plays a key role in improving the CO 2 RR activity.

Conclusions and Outlook
In summary, this review discussed the controlled synthesis and emphasized the rational synthesis of heteroatom (i.e., N, S, O, F, or B)-doped porous carbon nanostructures (HA-PCs) for the CO 2 RR. This includes the CO 2 RR fundamental pathways and engineering methods of HA-PC nanostructures. The effects of dopants (individually or mixed) on the CO 2 RR activity and durability were reported.
The Faradaic efficiency (FE), overpotential (η), partial current density (j), durability, energy efficiency (E eff ), and turnover frequency (TOF) are the main factors that determine the CO 2 RR activity of HA-PCs.
There are a few methods for the rational synthesizing of HA-PC nanostructures: template-based, activation, element-doping, and direct annealing of biomass-based resources, which vary in their productivity for tailoring the porosity (i.e., pore-volume, pore order, and pore size), surface area, and structure of HA-PCs. Hard template methods (Zn-based, Mg-based, Ca-based, and Si-based) are extensively studied for HA-PCs. This is due to their ability to produce uniform porous morphologies with well-controlled porosity and surface area, but their multiple reaction steps and use of hazardous chemicals for etching templates remain a grand challenge. However, the Zn-based template is the most preferred among hard templates because Zn is inexpensive, earth-abundant, and can be easily prepared in high yield and evaporate during annealing without additional etching steps. New templates such as melamine, dry ice, and MXene have also allowed the synthesis of HA-PCs, but they have rarely been reported. Soft templates including ionic/nonionic copolymers (i.e., F127and cetrimonium chloride) and ionic liquids (i.e., Gemini-type) are also used for the preparation of HA-PCs, but they cannot produce uniform porous structures with a high surface area and they are usually accompanied with other methods.
Deriving HA-PCs from biomass is among the most promising approaches due to the low-cost and natural abundance of biomass waste; also, the unique structures and composition of biomass wastes drive the production of HA-PCs with well-defined porosity and composition under ambient conditions that meet the sustainability requirements. Using melamine is highly promising as it allows integration of high N content and constructer carbon-nitride over HA-PCs during carbonization at a temperature above 500 • C. MOFderived HA-PCs are also studied significantly, but their high cost and multiple/complicated reaction steps remain a significant challenge. Moreover, the use of activators is needed to enhance the surface area and porosity.
Various HA-PC nanostructures have been prepared using multiple methods for CO 2 RR, which varied in their performance (Tables 3 and 4). Notably, KHCO 3 and NaHCO 3 are the main electrolytes used for HA-PCs, and NaClO 4 electrolytes were rarely reported; furthermore, CO is the main CO 2 RR product, and other products such as CH 4 , HCOOH, C 2 H 5 OH, and CH 3 OH were rarely found, which implies the selectivity of HA-PCs for producing CO.
The biomass-derived HA-PCs are the most promising for practical applications. Chlorine-promoted N/S co-doped PCs (LC-3) formed from annealing of lignin, urea, melamine, NaCl, and ZnCl2 at 100 • C and impregnating in HCl revealed an FE of 95.9% [134] and N/B-doped porous carbon (BNMC-1000) showed an FE of 95% [142], and N-doped porous carbon from calcination of ZIF-8 (NC1100) showed an FE of 95.4% [107]. N-doped PCs are the most reported and most active compared to other HAs dopants, and they are the most active for CO 2 RR. The highest active HA-PCs are N-doped porous carbon (NPC-1000) formed using annealing of Zn-MOF-74 and with melamine at 1000 • C showing an FE of 98.4% [119] and N-doped porous carbon (N/C-Cl-1100) obtained from halogen-assisted annealing of ZIF-8 with KCl at 1100 • C revealing an FE of 99.5% [117]. This is due to their porosity and abundance of N-species (i.e., pyridinic-N and graphitic-N). B/N-doped nanodiamond (BND) produced only ethanol with an FE of 93.2% [141]. Despite the noticed progress in HA-PC catalysts for CO 2 RR, they are still far from being useful for large-scale applications, so various perspectives and challenges should be addressed:

•
The current preparation approaches of heteroatom-doped porous carbon-based nanocatalysts involve multiple reaction steps, energy consumption, and hazardous reagents, making them impractical. Thus, they should be prepared using green materials under ambient conditions to meet sustainability requirements. • Using biomass wastes is a promising approach to synthesizing HA-PCs with tunable porosity and surface area under ambient conditions; however, they are rarely reported for CO 2 RR.

•
The CO 2 RR performance of HA-PCs is mainly measured in CO 3 -based electrolytes, so other organic, ionic liquid, and hybrid electrolytes should be studied to produce liquid products other than CO. Moreover, the effect of electrolytes and cell design on the CO 2 RR of HA-PCs has not yet been reported.

•
Computational studies could be conducted with experimental studies to allow the synthesis of novel HA-PCs and to examine their CO 2 RR activity, mechanism, and pathways.
Author Contributions: All authors contributed equally to this work. Collecting the data and writing-original draft preparation, Q.L. and K.E.; collecting data, writing, and editing, K.E. and W.L.; supervision and project administration, W.L. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.