Exploring the Role of Non-Metal Doping in g-C₃N₄ for CO₂ Reduction: A DFT Investigation

Abstract

The electrochemical reduction of CO₂ (CO₂RR) to valuable chemicals and fuels is a promising strategy for addressing environmental challenges. Graphitic carbon nitride (g-C₃N₄) is a promising electrocatalyst for CO₂ reduction. However, poor electron transfer and low CO₂ affinity often limit its catalytic performance. In this study, we employ density functional theory (DFT) calculations to systematically investigate the effect of various non-metal dopants (B, P, O, and S) on the electronic structure and CO₂ adsorption properties of g-C₃N₄. Our results demonstrated that O-C₃N₄ preferentially catalyzes the formation of HCOOH with a low limiting potential of −0.12 V. Meanwhile, S-C₃N₄ efficiently promotes the generation of CH₂O, CH₃OH, and CH₄ at a limiting potential of −0.58 V, as well as CO at −0.77 V. These findings provide valuable insights toward the rational design of effective non-metal-doped g-C₃N₄ catalysts for efficient CO₂ conversion.

1. Introduction

Carbon dioxide (CO₂) is vital for life on Earth due to its role in the carbon cycle, but excessive levels can harm the environment and human health. The persistent concern over CO₂ emissions has led to intense research into carbon capture and utilization technologies, making it a critical focus area. Among these, carbon dioxide reduction (CO₂RR) is one of the crucial processes for achieving carbon recycling and mitigating environmental issues stemming from excessive CO₂ emissions [1], and it has garnered widespread attention [2,3]. Electrochemical CO₂ reduction involves multiple transfer processes, including multi-electron and multi-proton pathways, that lead to the formation of various products. The products of CO2RR are formed through multiple reaction pathways, including carbon monoxide (CO) and formic acid (HCOOH), formed via two-electron pathways; formaldehyde (HCHO), formed via four-electron pathways; methanol (CH₃OH), formed via six-electron pathways; methane (CH₄), formed via eight-electron pathways; and ethylene (C₂H₄), formed via twelve-electron pathways, as well as some other products [4]. These products are especially significant due to their industrial relevance and their potential to mitigate CO₂ emissions. However, the performance of CO₂RR is limited by several factors, including its poor product selectivity and high overpotential. Thus, searching for CO2RR electrocatalysts with high selectivity and activity is significant.

In the past few decades, continuous efforts have been devoted to improving the performance of CO₂RR. Metal-free materials are becoming highly promising alternatives for various electrochemical reactions due to their excellent activity, high-temperature stability, and low cost [5,6,7,8,9,10]. As an eco-friendly and sustainable semiconductor material, graphitic carbon nitride (g-C₃N₄) possesses excellent electronic characteristics, favorable chemical durability, and a low production cost, factors which are attracting increasing attention [11,12,13,14]. It is considered an organic semiconductor, mainly composed of carbon and nitrogen, which are among the most abundant elements on Earth [15]. g-C₃N₄-based materials have proven to be excellent candidates for various applications, including photocatalysis, electrocatalysis, energy conversion, and environmental remediation [16,17,18,19,20,21,22]. Previous studies have demonstrated that incorporating transition metals (TMs) could significantly enhance the catalytic activity of g-C₃N₄ in CO₂ reduction [23,24,25,26]. However, TM-embedded g-C₃N₄ is generally toxic and expensive, considerably limiting its large-scale application. Therefore, it remains a question of interest whether non-metal-doped g-C₃N₄ is an efficient electrocatalyst for CO₂RR.

Recently, numerous experimental [27,28,29,30,31,32,33,34,35,36] and theoretical [27,31,36,37,38,39,40,41,42] studies have been conducted on the structure and electronic properties of pristine g-C₃N₄ and non-metal-doped g-C₃N₄. In 2017, Fu et al. [43] reported on the efficiency of photocatalytic CO₂ reduction using hierarchical porous O-doped g-C₃N₄ nanotubes (OCN-Tubes). The hierarchical OCN-Tubes showed excellent photocatalytic CO₂ reduction performance, with a methanol evolution rate of 0.88 µmol g⁻¹ h⁻¹, which was five times higher than bulk g-C₃N₄ (0.17 µmol g⁻¹ h⁻¹). In 2020, Wei et al. [40] studied the electronic characteristics of pristine and boron-doped monolayer g-C₃N₄. They found that the energy band gap was reduced from 3.06 eV to 1.33–1.80 eV because of the introduction of boron impurities at three sites (N2, C1, interstitial). Furthermore, the interstitial doping of B atoms significantly promotes the adsorption of pollutants by g-C₃N₄. Ranjbakhsh et al. [27] synthesized and investigated the mechanisms of CO₂ conversion into value-added products using P-doped g-C₃N₄ (PCN). They found that methane could be easily desorbed from the PCN surface. The conversion of CO₂ to CH₄ on the P-doped g-C₃N₄ had a Gibbs activation energy of 3.26 eV. In addition, Wang et al. [44] reported that photoactivity could be improved by S doping to produce CH₃OH. Moreover, Arumugam et al. [45] reported that CH₄ was the main product of CO₂ reduction by B-, P-, O-, and S-doped g-C₃N₄ catalysts, with CH₄ yields of 28.50, 31.80, 38.00, and 55.10 nmol/(mL_H2O·g_cat) for B-, P-, O-, and S-doped g-C₃N₄, respectively, and 22.10 nmol/(mL_H2O·g_cat) for pristine g-C₃N₄.

Inspired by previous studies, we systematically investigated the mechanisms of CO₂RR using non-metal (B, P, O, and S)-doped g-C₃N₄ (NM-C₃N₄) using density functional theory (DFT).

We analyzed structural stability, electronic properties, and reaction pathways, focusing on key the intermediates of carboxylate (COOH) and formate (OCHO), leading to C1 products (CO, HCOOH, HCHO, CH₃OH, and CH₄). Limiting potential calculations determined the most favorable pathways and product selectivity. Our findings offer fundamental insights into designing efficient and sustainable non-metal-based electrocatalysts for CO₂ reduction.

2. Results and Discussion

2.1. Geometric Structures of Pristine g-C₃N₄ and NM-C₃N₄

According to the symmetry of the tri-s-triazine structure of pristine g-C₃N₄, we investigated various possible doping sites for NM-C₃N₄. Non-metal dopants, including B, P, O, and S, can occupy one of five substitutional positions (C1, C2, N1, N2, and N3) or interstitial positions, as depicted in Figure S1. To evaluate the stability of the NM-C₃N₄ structures, their formation energies were calculated and are summarized in Table S1. The optimized structures of each doping configuration are presented in Figure 1. Figure 1a,b illustrate the most stable structures for B-₃N₄ and P-C₃N₄, both featuring dopants located at the C1 site. In contrast, the optimal doping sites for O-C₃N₄ and S-C₃N₄ are the N2 site (see Figure 1c,d). The primary factors allowing B and P to occupy the C1 position are their capacities to coordinate effectively with the three neighboring N atoms in the pristine g-C₃N₄ lattice. In contrast, O and S atoms prefer the N2 site due to their placement within the same chemical group and their efficient coordination with two adjacent carbon atoms, yielding greater stability compared to alternative doping sites. Additionally, the calculated formation energies for, B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄ are −1.23, 0.33, −0.49, and 0.29 eV, respectively.

In pristine g-C₃N₄, the bond lengths of N1–C1, C1–N2, N2–C2, and C2–N3 measure 1.48, 1.34, 1.34, and 1.40 Å, respectively, consistent with other research [38,46]. The bond lengths of B–N, B–N1, and B–N2 are 1.43, 1.52, and 1.43 Å, respectively, while the bond lengths of P–N, P–N1, and P–N2 are 1.64, 1.75, and 1.65 Å, respectively. In the case of O-doped pristine g-C₃N₄, the O–C1 and O–C2 bond lengths are 1.42 and 1.36 Å, respectively, which agrees with a previous report [38]. For S-doped pristine g-C₃N₄, the S–C1 and S–C2 bond lengths are 1.81, and 1.77 Å, respectively. Moreover, the bond lengths in the non-metal-doped structure are greater than those in the pristine g-C₃N₄ structure, indicating a slight distortion in the heptazine unit. For the P replacement at the C1 site and S replacement at the N2 site, the larger atomic sizes of P and S compared to the atoms they replace cause them to protrude from the substrate, leading to the slight distortion of the local structure. This protrusion can impact the positive formation energies, in contrast, B-C₃N₄ and O-C₃N₄ are more seamlessly integrated into the substrate due to the smaller atomic sizes of B and O, which are closer to the sizes of the atoms they replace. This allows for their smoother integration into the lattice. The uniform incorporation of B and O results in less structural deformation, enhancing stability and preserving the desired properties of the pristine g-C₃N₄ substrate.

Additionally, the thermodynamic stability of all the catalysts was confirmed by ab initio molecular dynamics (AIMD) simulations at 300 K, as shown in Figure S2. All the catalysts demonstrate high thermodynamic stability. Although P-C₃N₄ and S-C₃N₄ exhibit slight deformation, they still maintain overall stability with positive formation energies. In summary, we assume that all our NM-C₃N₄ studies are also stable.

2.2. Electronic Properties of Pristine g-C₃N₄ and NM-C₃N₄

To elucidate the impact of non-metal doping on pristine g-C₃N₄, we calculated the electron localization function (ELF), charge density difference (CDD), density of states (DOS), and band gap of the NM-C₃N₄. The ELF analysis provides insights into the pristine and doped structures’ bonding nature and stabilization mechanisms. The ELF values range from 0 to 1, where 0 (blue) signifies electron delocalization and 1 (red) indicates strong localization. ELF values above 0.5 typically suggest covalent bonding or core electrons, while values below 0.5 indicate ionic bonds. An ELF value of 0.5 is associated with metallic bonds [47]. Our ELF analysis confirms the presence of covalent bonding between non-metal dopants and C or N atoms in g-C₃N₄, as shown in Figure S3. Notably, the non-metal atoms chemisorb onto the g-C₃N₄ framework, forming stable bonds. This finding aligns well with the stability analysis based on formation energy (E_f), despite the positive E_f observed for P-C₃N₄ and S-C₃N₄.

As illustrated in Figure S4, the charge density difference (CDD) analysis indicates that B and P atoms lose electrons. In contrast, O and S atoms gain electrons, producing distinct oxidation and reduction effects. B and P exhibit electron depletion, transferring 1.58e and 2.03e to the g-C₃N₄ surface (see

Table 1. Bond distances of B- or P-N1/N2 and O- or S-C1/C2 (D in Å), Bader charge analysis of Cu and C atoms (Q, in e), and band gaps with HSE06 (eV). Positive and negative charge values represent gained and lost electrons.

Surfaces	Q(Non-Metal)	Q(N1/N2)	Q(C1/C2)	Band Gaps
g-C3N4	-	−0.75/−0.89	1.22/1.09	3.06
B-C3N4	1.58	−0.87/−1.02	-	1.50
P-C3N4	2.03	−0.97/−1.22	-	2.06
O-C3N4	−0.76	-	1.18/1.24	1.07
S-C3N4	−0.05	-	0.86/1.05	1.14

). In contrast, O and S act as electron acceptors, with O gaining 0.76e and S gaining 0.05e. These charge redistribution trends reflect differences in electronegativity, classifying B and P as electron donors and O and S as electron acceptors.

Furthermore, the total density of states (TDOS) and partial density of states (PDOS) of the pristine g-C₃N₄ and NM-C₃N₄ were calculated by the HSE06 function. As shown in Figure 2, the calculated band gap of pristine g-C₃N₄ is 3.06 eV, consistent with a previous theoretical study [40] and close to the experimental value of 2.70 eV [48]. For pristine g-C₃N₄, the PDOS analysis reveals that the valence band (VB) predominantly comprises N p-orbitals. In contrast, both C and N p-orbitals contribute to the conduction band (CB). The symmetric TDOS around the Fermi level indicates that the material is non-magnetic. A significant reduction in the band gap is observed upon doping with non-metal atoms (B, P, O, and S). The band gaps of B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄ are reduced to 1.50 eV, 2.06 eV, 1.07 eV, and 1.14 eV, respectively. In the cases of B and P doping, the Fermi level shifts toward the valence band due to their donor-like behavior, which introduces shallow acceptor states. Conversely, O and S doping result in a Fermi level shift toward the conduction band, attributed to their acceptor-like nature. This change introduces mid-gap states and narrows the band gap, which is favorable for enhancing electronic conductivity and charge transfer. Additionally, the PDOS profiles show strong hybridization between the non-metal dopants and adjacent C or N atoms, as evidenced by overlapping peaks near the Fermi level. This hybridization indicates strong electronic coupling and improved charge delocalization in the doped systems. Notably, the appearance of pronounced peaks near the Fermi level in O-C₃N₄ and S-C₃N₄ suggests the presence of localized states that may act as active sites for CO₂ adsorption and activation. The narrowed band gaps and the emergence of mid-gap states benefit the CO₂ reduction reaction (CO₂RR), as they can facilitate efficient charge carrier transport and enhance electrochemical activity. Moreover, the alignment of the Fermi level near the CB in O-C₃N₄ and S-C₃N₄ can promote electron transfer to adsorbed CO₂ molecules, thereby facilitating the key reduction steps. Therefore, these doped g-C₃N₄ systems, particularly the O- and S-doped variants, are expected to exhibit improved electrocatalytic activity for CO₂RR, correlating well with their predicted ability to generate valuable products such as HCOOH, CH₂O, CH₃OH, and CH₄ at low overpotentials.

2.3. CO₂ Adsorption on Pristine g-C₃N₄ and NM-C₃N₄

The initial step of electrocatalytic CO₂ reduction involves the adsorption of CO₂ onto catalyst surfaces. The potential adsorption sites for CO₂ on the catalysts are illustrated in Figure S5. Firstly, as reported in Table S2, CO₂ is adsorbed on surfaces with adsorption energies ranging from −0.15 to −0.38 eV. The most stable CO₂ adsorption onto pristine g-C₃N₄, B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄ is displayed in Figure 3. CO₂ molecules are adsorbed on all surfaces with bond lengths of 2.97–3.26 Å. To further examine the adsorption mechanism, charge density difference analyses were conducted. Electron transfer from the catalyst surfaces to the CO₂ molecules was observed, suggesting that NM doping enhances charge redistribution. This electron transfer could facilitate CO₂ activation and promote subsequent electrochemical reduction, potentially improving CO₂RR performance.

2.4. Electrocatalytic CO₂ Reduction to C1 Products

The Gibbs free energy changes ($\Delta G$) for forming various C1 products were calculated to evaluate the efficiency of different reaction pathways in CO₂RR. Figure 4 illustrates the reaction mechanisms leading to key C1 products, including CO, HCOOH, CH₂O, CH₃OH, and CH₄, highlighting how CO₂RR conditions can be optimized to yield specific products. We compared the $\Delta G$ of possible intermediates at each reduction step to identify the most favorable reaction pathways. This analysis revealed the energetically most favorable path to the target products. Furthermore, we compared each pathway’s maximum free energy change ($\Delta G_{max}$) corresponding to its potential-limiting step (PLS). Pathways with lower $\Delta G_{max}$ values were considered more selective [49,50], a crucial criterion for optimizing CO₂RR processes and enhancing catalytic activity.

2.4.1. First Protonation to HCOOH and CO

In the first protonation step, CO₂ can be protonated at either the C or O site, forming formate (*OCHO) and carboxylate (*COOH) intermediates, respectively. Each intermediate can subsequently be protonated at either the C or O site, leading to the formation of other intermediates, as illustrated in Figure 4. Theoretically, the *OCHO or *COOH species can undergo further protonation to produce the final C1 product. We calculated the first protonation energies for both the *OCHO and *COOH intermediates to identify the most favorable pathway. Through a two-electron reduction process, CO₂ can thereby be converted into either HCOOH or CO.

Figure 5 and Figure 6 illustrate the reaction pathways of CO₂ reduction to HCOOH and the optimized structures of intermediates involved in HCOOH and CO production. For HCOOH, two potential pathways are identified: the COOH pathway (CO₂ → *COOH → *HCOOH → HCOOH) and the OCHO pathway (CO₂ → *OCHO → *HCOOH → HCOOH). The results indicate that the first protonation to form the *COOH intermediate requires a significantly lower $\Delta G$ compared to the *OCHO intermediate formation on both pristine g-C₃N₄ and S-doped C₃N₄, with values of 0.84 eV and 0.18 eV, respectively. In contrast, B-C₃N₄, P-C₃N₄ and O-C₃N₄ exhibit downhill trends in $\Delta G$ for *COOH formation. In the second protonation, the reaction from *COOH to *HCOOH on pristine g-C₃N₄ is also downhill, whereas B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄ require $\Delta G$ values of 1.93, 0.73, 0.12, and 0.02 eV, respectively. These findings suggest that HCOOH production is more likely to proceed via the COOH pathway rather than the OCHO pathway, with the $\Delta G_{max}$ values for pristine g-C₃N₄, B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄ being 0.84, 1.30, 1.93, 0.12, and 0.17 eV, respectively.

In the formation of CO through the CO₂RR pathway, the sequence can be outlined as CO₂ → *COOH → CO. The $\Delta G$ for the first protonation step (CO₂ → *COOH) mirrors that in the HCOOH generation reaction. For the second protonation, the *COOH → *CO step requires $\Delta G$ values of 0.06, 2.38, 1.33, 0.97, and 0.77 eV for pristine g-C₃N₄, B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄, respectively. Our results identify the PLS for CO production as the CO formation step, with $\Delta G_{max}$ values of 2.38, 1.33, 0.97, and 0.77 eV on B-C₃N₄, P-C₃N₄, O-C₃N₄, and S-C₃N₄, respectively, whereas for pristine g-C₃N₄, the first protonation step is the PLS. In evaluating the catalytic activity of each surface, we focus on the PLS for each pathway. Overall, the non-metal doping of pristine g-C₃N₄ decreases the $\Delta G$ of the first protonation step, enhancing CO₂RR efficiency. Consequently, the two-electron CO₂RR pathway favors HCOOH production via the COOH pathway over CO, with relatively low $\Delta G_{max}$ values observed across all surfaces.

2.4.2. The Other C1 Products for CO₂RR

To elucidate the thermodynamic catalytic activity, we calculated the different CO₂RR pathways for various C1 products on both pristine g-C₃N₄ and NM-C₃N₄. Detailed free energy diagrams for CO₂RR on these materials are depicted in Figure 7. These diagrams provide insight into the energy changes associated with each step of the CO₂RR process on these materials. The corresponding optimized structures for the favorable reaction pathways are presented in Figure 8.

Figure 7a shows the Gibbs free energy profile for CO₂RR on the pristine g-C₃N₄ surface. The initial step (CO₂ → *COOH) is critical in determining the PLS for producing various C1 products, including HCOOH, CO, CH₂O, CH₃OH, and CH₄, with a $\Delta G_{max}$ of 0.84 eV. The pathway for CH₂O production is CO₂ → *COOH → *HCOOH → *CHO → *CH₂O, while CH₃OH is formed through *CH₂O → *CH₂OH → *CH₃OH, and CH₄ proceeds via *CH₃OH → *CH₃ → *CH₄. HCOOH is more favorable than other products due to its easier absorption from the catalyst surface.

In contrast, Figure 7b displays the Gibbs energy profile for B-C₃N₄, where the second protonation step, *COOH → *HCOOH, serves as the PLS for producing HCOOH, CH₂O, CH₃OH, and CH₄, with a $\Delta G_{max}$ of 1.93 eV. The pathways for CH₂O, CH₃OH, and CH₄ follow the sequences CO₂ → *COOH → *HCOOH → *CHO → *CH₂O, *CHO → *CHOH → *CH₂OH → *CH₃OH, and *CH₂OH → *CH₂ → *CH₃ → *CH₄, respectively. Although CH₄ is thermodynamically favored due to its smaller free energy change, the high $\Delta G_{max}$ value limits the catalytic activity of B-C₃N₄, indicating a poor selectivity and limited effectiveness for CO₂RR.

Figure 7c shows the Gibbs energy profile for P-C₃N₄, where HCOOH and CH₂O are the primary products. The PLS for their production is CO₂ → *COOH, with a $\Delta G_{max}$ of 0.73 eV, indicating the lowest $\Delta G_{max}$. The preferred pathway for HCOOH production is CO₂ → *COOH → *HCOOH, with a desorption energy of 0.35 eV. For CH₂O, the pathway proceeds as CO₂ → *COOH → *HCOOH → *CHO → *CH₂O, with a desorption energy of 0.86 eV. Due to the difficulty in desorbing CH₂O compared to the subsequent protonation step, P-C₃N₄ slightly disfavors CH₂O production. The minor products CH₃OH and CH₄ have higher $\Delta G_{max}$ values of 1.38 and 1.47 eV, respectively, following the pathways CO₂ → *COOH → *HCOOH → *CHO → *CHOH → *CH₂OH → *CH₃OH and CO₂ → *COOH → *HCOOH → *CHO → *CHOH → *CH₂OH → *CH₂ → *CH₃ → *CH₄. The desorption energies for CH₃OH and CH₄ are 0.02 eV and −0.09 eV, respectively, suggesting the good selectivity of P-C₃N₄ for HCOOH production.

The Gibbs energy profile for O-C₃N₄ is shown in Figure 7d, where HCOOH emerges as the primary product, following the pathway CO₂ → *COOH → *HCOOH. The PLS for HCOOH production is the *COOH → *HCOOH step, with a $\Delta G_{max}$ of 0.12 eV and a desorption energy of 0.31 eV. For minor products such as CH₂O, CH₃OH, and CH₄, the corresponding pathways are CO₂ → *COOH → *HCOOH → *CHO → *CH₂O for CH₂O, CO₂ → *COOH → *HCOOH → *CHO → *CH₂O → *CH₂OH → *CH₃OH for CH₃OH, and CO₂ → *COOH → *HCOOH → *CHO → *CH₂O → *CH₂OH → *CH₃OH → *OH + CH₄ → *H₂O for CH₄. Both pathways have a $\Delta G_{max}$ of 0.73 eV, with the PLS occurring at the *CHO → *CH₂O step. The desorption energies are −0.23 eV for CH₂O, −0.25 eV for CH₃OH, and −0.01 eV for CH₄, indicating favorable conditions for desorption. These findings suggest that O-C₃N₄ exhibits good selectivity and excellent catalytic activity for HCOOH production.

Finally, Figure 7e displays the Gibbs energy profile for S-C₃N₄, where HCOOH is the primary product along the CO₂ → *COOH → *HCOOH pathway. The PLS for HCOOH formation is *COOH → *HCOOH, with a $\Delta G_{max}$ of 0.17 eV and a desorption energy of 0.12 eV. Either *HCOOH desorption or further protonation to generate *CHO occurs with the same energy. Therefore, the reaction proceeds from *CHO through successive proton–electron (H⁺ + e⁻) transfer steps, leading to the formation of CH₂O, CH₃OH, and CH₄. It can be observed that, among the various intermediates in the CO₂RR process, there is one pathway leading to the four-electron product CH₂O, two pathways to form CH₃OH, and four pathways to produce CH₄. The most favorable pathways are as follows: The formation of CH₂O proceeds via CO₂ → *COOH → *HCOOH → *CHO → *CH₂O. For CH₃OH and CH₄, the initial reduction steps follow the same sequence as CH₂O formation, continuing along the pathways CO₂ → *COOH → *HCOOH → *CHO → *CH₂O → *CH₂OH → *CH₃OH and CO₂ → *COOH → *HCOOH → *CHO → *CH₂O → *CH₂OH → *CH₃OH → *CH₃ → *CH₄, respectively. The protonation step from *CHO to *CH₂O is identified as the PLS for the formation of all three products, with a $\Delta G_{max}$ of 0.58 eV. The desorption energies are −0.26 eV for CH₂O, −0.29 eV for CH₃OH, and −0.23 eV for CH₄, indicating that these products can be readily released from the surface. These results suggest that S-C₃N₄ exhibits a low $\Delta G_{max}$ for HCOOH production. Interestingly, S-C₃N₄ can also facilitate the formation of CH₂O, CH₃OH, and CH₄, with a $\Delta G_{max}$ of 0.58 eV compared to just 0.17 eV for HCOOH.

While our DFT calculations focus on a fixed, low concentration of dopants (one non-metal atom per 2 × 2 supercell), the analysis captures the intrinsic catalytic behavior of each doped system. Notably, the dopant identity strongly influences the competition among various CO₂RR pathways. For example, O-doping favors HCOOH formation via the *COOH pathway due to a low potential-limiting step ($\Delta G$ = 0.12 eV), whereas S-doping facilitates deeper reduction pathways, enabling the formation of CH₂O, CH₃OH, and CH₄ with a $\Delta G_{max}$ of 0.58 eV. These results reveal that dopant type governs activity, selectivity, and product distribution. We acknowledge that dopant concentration may further affect these trends by modulating surface electronic properties and active site density. However, a systematic investigation of varying dopant levels is beyond the scope of this work and will be explored in future studies. As supporting evidence, experimental work by Zhu et al. [51] demonstrated that CO₂RR activity in S-C₃N₄ varied significantly with dopant concentration, with peak performance observed at a thiourea-to-dicyandiamide molar ratio of 0.4. This underscores the importance of future studies combining concentration effects with pathway analysis.

2.5. Competition Between CO₂RR and HER on Pristine g-C₃N₄ and NM-C₃N₄

To evaluate catalytic activity and selectivity for CO₂RR, it is essential to consider the competing hydrogen evolution reaction (HER). We investigated the free energy barriers of the HER based on the Volmer–Heyrovsky mechanism. Figure 9 and Figure S6 show the free energy diagram and optimized structure of H* adsorption on each surface. Our results showed that the HER occurs readily on pristine g-C₃N₄ with a $\Delta G$ of 0.04 eV, indicating a high tendency for HER and poor CO₂RR selectivity. However, non-metal doping significantly alters this behavior, suppressing HER and enhancing the preference for the CO₂RR pathway. Specifically, B-C₃N₄ and P-C₃N₄ exhibit greater thermodynamic favorabilities toward CO₂RR over HERs. O-C₃N₄ and S-C₃N₄ also show a clear preference for CO₂RR, particularly in producing HCOOH. The HER remains slightly more favorable regarding free energy under standard conditions for other C1 products (CO, CH₂O, CH₃OH, and CH₄). However, in practical applications, HER can often be suppressed by tuning the pH or electrolyte composition [52].

These findings confirm that non-metal doping effectively modulates the electronic structure and surface adsorption behavior of g-C₃N₄, thereby suppressing HER and improving CO₂RR efficiency. Among the doped catalysts, S-C₃N₄ emerges as the most promising material for producing CH₂O, CH₃OH, and CH₄, exhibiting a remarkably low limiting potential ($U_L$) of −0.58 V. This performance surpasses that of other non-metal-doped systems and rivals or exceeds certain metal-based catalysts. Additionally, O-C₃N₄ stands out for HCOOH production, achieving the lowest U_L of −0.12 V, significantly outperforming previously reported systems such as Ni-C₃N₄ [26], Ag@g-C₃N₄ [53], and B-C₃N [54]. This exceptional activity is attributed to its stabilizing of key reaction intermediates, such as *OCHO and *COOH, which lowers the energy barriers and promotes the selective formation of desired C1 products.

The comparative analysis of the limiting potentials of CO, HCOOH, CH₂O, CH₃OH, and CH₄ on pristine and doped g-C₃N₄ systems, shown in Figure 10, highlights the critical role of dopants in product selectivity. The overall trend in product preference is as follows: HCOOH > CH₂O, CH₃OH, CH₄ > CO. This confirms HCOOH as the most favored product, while CO is the least preferred.

The doped variants show significantly improved catalytic performance compared to pristine g-C₃N₄, which exhibits a relatively high limiting potential (−0.84 V) for the initial CO₂ activation step (CO₂ → *COOH). In particular, the P-, O-, and S-doped systems exhibit remarkably lower limiting potentials of −0.73 V, −0.12 V, and −0.17 V, respectively, for HCOOH production, indicating their enhanced activity and reduced energy barriers for CO₂ reduction. These systems also promote the formation of valuable products such as CH₂O, CH₃OH, and CH₄. While B-C₃N₄ also shows performance improvements and leads to diverse product distributions, it still requires higher overpotentials than its P-, O-, and S-doped counterparts. Among all the systems investigated, non-metal dopants, especially O- and S-, exhibit remarkably low limiting potentials and support the formation of various valuable reduction products. As summarized in

Table 2. Comparison of PLS, limiting potential ($U_L$), and product of CO₂RR with previously reported catalysts.

Catalysts	PLS	${\mathit{U}}_{\mathit{L}}$ (V)	Product
*Pristine g-C3N4	CO2 → *COOH	−0.84	HCOOH, CH2O, CH3OH, and CH4
*B-C3N4	*COOH → *HCOOH	−1.93	HCOOH, CH2O, CH3OH, and CH4
*P-C3N4	*COOH → *HCOOH	−0.73	HCOOH and CH2O
*O-C3N4	*COOH → *HCOOH	−0.12	HCOOH
*S-C3N4	CO2 → *COOH	−0.17	HCOOH
	*CHO → *CH2O	−0.58	CH2O, CH3OH, and CH4
HSi-C3N4 [55]	*OCH2OH → *CH2O	−0.65	CH4
Ag@g-C3N4 [53]	*OCHO → *HCOOH	−0.37	HCOOH
Cu@g-C3N4 [53]	*COOH → HCOOH	−0.37	HCOOH
Co-C3N4 [26]	*CH2O → *CH2OH	−0.65	CH3OH
Fe-C3N4 [26]	*HCOOH → *CHO	−0.91	CH3OH and CH4
Ni-C3N4 [26]	*HCOOH → * + HCOOH	−0.74	HCOOH
BN-C3N [54]	CO2 → *HCOO	−0.28	HCOOH
KBCN [56]	*CO2 → *COOH	−0.86	CH3OH

* This work’s Catalysts.

, our O- and S-doped g-C₃N₄ surfaces exhibit superior activity, with notably low limiting potentials of −0.12 V and −0.58 V, respectively. These values not only improve upon pristine g-C₃N₄ (−0.84 V) but also outperform several reported systems. These comparisons indicate that our designed O- and S-doped g-C₃N₄ catalysts offer compelling combinations of low energy barriers, favorable selectivity, and scalable non-metal synthesis, reinforcing their promise for CO₂RR applications. This highlights the potential of non-metal-doped g-C₃N₄, which is promising due to its high activity at low overpotentials and its environmentally benign and sustainable nature. Unlike metal-based catalysts, which may face cost, scarcity, and toxicity challenges, non-metal-doped systems offer a green alternative without compromising performance.

To further evaluate catalytic performance and mechanistic trends, we constructed volcano plots (Figure 11) correlating the limiting potential ($U_L$) for HCOOH and CH₄ production with the free energy difference $\Delta G$(*COOH) − $\Delta G$(CO₂). These plots reflect the Sabatier principle, which posits that optimal catalysts exhibit an intermediate binding strength for key intermediates, neither too weak, leading to poor activation, nor too strong, resulting in high energy barriers for product formation or desorption. In this analysis, O-C₃N₄ and S-C₃N₄ surfaces are positioned at the top and middle of the curve, demonstrating a favorable balance between activity and stability. Catalysts positioned in this region are more likely to facilitate efficient CO₂RR. Notably, O-C₃N₄ favors HCOOH formation, while S-doping promotes CH₄ generation, consistent with their respective mechanistic pathways.

To understand the electronic basis for this trend, we further analyzed the *COOH and *OCHO intermediates using ICOHP and PDOS methods, as shown in Figure 12. Interestingly, *OCHO exhibits more negative ICOHP values than *COOH (e.g., −4.73 eV vs. −4.21 eV on O-C₃N₄), indicating stronger orbital overlap. However, this strong binding may hinder further protonation and raise the reaction barrier. In contrast, *COOH exhibits moderately strong bonding, sufficient to activate CO₂ but not so strong as to prevent desorption or subsequent transformation. These insights align well with the observed thermodynamic profiles, which show lower $\Delta G$ and limiting potential for *COOH-mediated pathways. The PDOS results further confirm that *COOH forms well-hybridized states with dopant orbitals, facilitating charge transfer and enhancing reactivity. The volcano plot and bonding analysis provide a unified explanation for the superior CO₂RR activity observed on O-C₃N₄ and S-C₃N₄. Their position near the volcano apex and favorable intermediate bonding characteristics highlight their potential as highly efficient, metal-free catalysts for selective CO₂ conversion.

Although this study evaluates non-metal-doped g-C₃N₄ at a single dopant concentration, the observed trends in limiting potential and product selectivity reflect the intrinsic effects of dopant identity. Our findings underscore the distinct catalytic roles of each dopant, with O promoting selective HCOOH production and S facilitating deeper reduction to CH₄. These outcomes provide predictive guidance for tuning product distribution via rational doping strategies. Future work will extend this framework to explore dopant concentration effects, which have been shown experimentally to significantly influence photocatalytic CO₂ reduction performance in g-C₃N₄ systems.

3. Methodology

3.1. Computational Details

Firstly, all the spin-polarized density functional theory (DFT) calculations were performed using the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation exchange–correlation function [57] and the projector augmented wave (PAW) [58] method via the Vienna ab initio Simulation Package code. (VASP) [59]. Grimme’s dispersion correction (DFT-D3) [60] described the van der Waals interaction. The cutoff energy of the plane–wave basis was set to 450 eV. The forces and electronic energy were converged to within 0.02 eV/Å and 1 × 10⁻⁶ eV, respectively. Accordingly, the k-point was set to a 3 × 3 × 1 Monkhorst–Pack mesh for geometry optimizations. After geometry optimization, the HSE06 hybrid function [61,62] was used to refine the electronic properties. Ab initio molecular dynamics (AIMD) simulations were performed to assess the thermal stability of the surfaces.

3.2. Structural Model and Stability Assessment

The pristine g-C₃N₄ structure was modeled as a 2 × 2 supercell with 32 N atoms and 24 C atoms, cleaved along the (001) plane as shown in Figure 13. A vacuum layer of 15 Å was introduced to prevent interactions between periodic images. The lattice constant of the surface after geometry optimization was a = b = 14.24 Å, which was consistent with previous computational and experimental studies [63,64,65].

To assess the stability of the non-metal atoms (B, P, O, and S) substituted into the pristine g-C₃N₄, the defect formation energies ${(E}_f)$ for replacing a nitrogen atom or carbon atom (1) and for the interstitial dopants (2) were calculated using the following equations:

$$E_f=E_{non-metals-C_3{N}_4}-E_{C_3{N}_4}-{\mu }_{non-metals}+{\mu }_{N or C}$$

(1)

$$E_f=E_{non-metals-C_3{N}_4}-E_{C_3{N}_4}-{\mu }_{non-metals}$$

(2)

where $E_{non-metals-C_3{N}_4}$ is the total energy of the non-metal substitution into the pristine g-C₃N₄ and $E_{C_3{N}_4}$ is the total energy of the pristine g-C₃N₄. ${\mu }_{non-metals}$, ${\mathrm{\mu }}_N$, and ${\mathrm{\mu }}_C$ are the chemical potentials of the non-metals, N, and C atom, respectively. ${\mathrm{\mu }}_N$, ${\mathrm{\mu }}_C$, ${\mathrm{\mu }}_B$, ${\mathrm{\mu }}_P$, ${\mathrm{\mu }}_O$, and ${\mathrm{\mu }}_S$ are determined as follows: ${\mathrm{\mu }}_N$ is defined as half of the energy of an N₂ molecule in the gas phase, ${\mathrm{\mu }}_C$ as the energy of a single carbon atom in pristine graphene, ${\mathrm{\mu }}_B$ as the energy of a single boron atom in the metallic α-B phase [66], ${\mathrm{\mu }}_P$ as the energy of a single phosphorus atom in P₄ [27], ${\mathrm{\mu }}_O$ as half of the energy of an O₂ molecule in the gas phase, and ${\mathrm{\mu }}_S$ as the energy of a single sulfur atom in S₈ [67]. A negative $E_f$ predicted that the formation of all the non-metal doped systems would be more stable and spontaneous. Moreover, the adsorption energies ${(E}_{ads})$ of the gas molecules on the surfaces were calculated by the following equation:

$$E_{ads}=E_{complex}-E_{monolayer}-E_{gas}$$

(3)

where $E_{complex}$, $E_{monolayer}$, and $E_{gas}$ are the total energy of the molecules adsorbed on the monolayer, the bare monolayer, and isolated molecules, respectively. Herein, a more negative $E_{ads}$ value indicates more stable gas adsorption.

Reaction pathways were analyzed using Gibbs free energy (ΔG) corrections based on the computational hydrogen electrode (CHE) model [68], which is described as follows:

$$G=E+ZPE-TS+\int C_{p}dT+E_{sol}$$

(4)

where $E$, $ZPE$, $TS$, $\int C_{p}dT$, and $E_{sol}$ represent the total electronic energy, zero-point energy, entropy correction, enthalpic temperature correction, and solvent effect correction, respectively. $T$ is the temperature, which was set to 298.15 K in this work. These corrections were obtained from vibrational frequency calculations. The $E_{sol}$ value depends on the hydroxy group (OH) binding configuration. In this study, we used an $E_{sol}$ of −0.20 eV when the intermediate had its OH group directly bound to the surface, and of −0.10 eV when the intermediate had an OH group indirectly bound to the surface [69].

The free energy change $(\Delta G)$ between the intermediate states was calculated by the following equation:

$$\Delta G=\Delta E+\Delta ZPE-T\Delta S+\int C_{p}dT+{\Delta E}_{sol}+{\Delta G}_{pH}+{\Delta G}_U$$

(5)

where $\Delta E$, $\Delta ZPE$, $T\Delta S$, and ${\Delta E}_{sol}$ represent the electronic energy difference between the two intermediate states, the change in zero-point energies, the entropic corrections, and the change in solvent effect correction, respectively. ${\Delta G}_{pH}$ and ${\Delta G}_U$ are the change in the pH of the solution and the difference in electrode potential of the electrochemical reaction, respectively. Herein, ${\Delta G}_{pH}=2.303k_{b}TpH$, where $k_b$ and $pH$ are the Boltzmann constant and $pH$ of the reaction condition, respectively. In this work, we assumed that pH = 0; thus, ${\Delta G}_{pH}=0$. Moreover, ${\Delta G}_U=-neU$, where $n$, $e$ and $U$ are the number of transferred electrons, the electron charge, and the applied external potential, respectively. If there was no applied potential in the reaction, then ${\Delta G}_U=0$ [70,71].

Additionally, the limiting potential ${(U}_L)$ represented the minimum applied potential that would make every elementary step exergonic, which could be obtained by the following equation:

$$U_L=-\Delta G_{max}/e$$

(6)

where $\Delta G_{max}$ is the $\Delta G$ of a potential-limiting step (PLS) in the CO₂RR pathway.

4. Conclusions

We employed density functional theory (DFT) to investigate the CO₂ reduction reaction (CO₂RR) mechanisms on pristine and non-metal-doped g-C₃N₄ systems, incorporating B, P, O, and S dopants. Our results reveal that the dopant type and its substitutional site strongly influence catalytic activity and product selectivity. The *COOH intermediate plays a crucial role in C1 product formation, with O-doped g-C₃N₄ (O-C₃N₄) showing outstanding performance for HCOOH production at a low limiting potential of −0.12 V. In contrast, S-doped g-C₃N₄ (S-C₃N₄) exhibits high selectivity for CH₂O, CH₃OH, and CH₄ formation at −0.58 V, and CO at −0.77 V, underlining its versatility for multi-product CO₂RR. Beyond their catalytic efficiency, NM-doped g-C₃N₄ materials offer notable advantages in cost-effectiveness and environmental sustainability compared to traditional metal-based catalysts. Among the materials studied, O-C₃N₄ and S-C₃N₄ emerge as particularly promising, combining high activity and selectivity with the benefits of sustainable, metal-free catalysis for carbon dioxide conversion.