Modulation of Photocatalytic CO₂ Reduction by n–p Codoping Engineering of Single-Atom Catalysts

Abstract

Transition metal (TM) single-atom catalysts (SACs) have been widely applied in photocatalytic CO₂ reduction. In this work, n–p codoping engineering is introduced to account for the modulation of photocatalytic CO₂ reduction on a two-dimensional (2D) bismuth-oxyhalide-based cathode by using first-principles calculation. n–p codoping is established via the Coulomb interactions between the negatively charged TM SACs and the positively charged Cl vacancy (V_Cl) in the dopant–defect pairs. Based on the formation energy of charged defects, neutral dopant–defect pairs for the Fe, Co, and Ni SACs (P_TM⁰) and the −1e charge state of the Cu SAC-based pair (P_Cu⁻¹) are stable. The electrostatic attraction of the n–p codoping strengthens the stability and solubility of TM SACs by neutralizing the oppositely charged V_Cl defect and TM dopant. The n–p codoping stabilizes the electron accumulation around the TM SACs. Accumulated electrons modify the d-orbital alignment and shift the d-band center toward the Fermi level, enhancing the reducing capacity of TM SACs based on the d-band theory. Besides the electrostatic attraction of the n–p codoping, the P_Cu⁻¹ also accumulates additional electrons surrounding Cu SACs and forms a half-occupied d_x²_−y² state, which further upshifts the d-band center and improves photocatalytic CO₂ reduction. The metastability of Cl multivacancies limits the concentration of the n–p pairs with Cl multivacancies (P_TM@nCl (n > 1)). Positively charged centers around the P_TM@nCl (n > 1) hinders the CO₂ reduction by shielding the charge transfer to the CO₂ molecule.

1. Introduction

Artificial photocatalytic CO₂ reduction is an intriguing research area aiming to reduce fossil consumption and mitigate the greenhouse effect [1,2]. Photocatalysts are the crucial materials for readily converting CO₂ into environmentally friendly fuels via photolysis using abundant solar energy [3,4,5]. With maximal atom-utilization efficiency, transition metal (TM) single-atom photocatalysts (SACs) exhibit excellent catalytic performance comparable to precious metals [6,7,8,9,10,11]. Due to its sufficient optical absorption and adequate activation centers, two-dimensional (2D) bismuth oxyhalide BiOX (X = F, Cl, Br, I) is an outstanding host to anchor the TM SACs photocatalyst and accelerate the CO₂ reduction [12,13,14,15,16]. For TM-doped bismuth oxyhalide, the dopant of TM SACs can form impurity levels in the forbidden band, promoting the generation and separation of photogenerated carriers [17,18]. However, the most outstanding advantage is that doped TM SACs regulate the surface state, which enhances the CO₂ adsorption by promoting the charge transfer between the activation centers and absorbed CO₂ molecules [17,18,19]. For example, isolated Cu SACs on BiOBr (Cu@BiOBr) establish a strong built-in electric field which serves as an electron trap to facilitate charge transfer and stabilize charge carriers. As a result, 0.5% Cu@BiOBr has a higher CO₂ absorption uptake (2.7 cm³ g⁻¹) than BiOBr (2.3 cm³ g⁻¹) [20]. In Co-SAC-doped Bi₃O₄Br, the isolation of Co²⁺ by replacing Bi³⁺ enables Co@Bi₃O₄Br layers to be more negatively charged, improving the CO₂ adsorption and stabilizing the *COOH intermediate on the surface [17]. Charge localization induced by Fe SAC doping strengthens CO₂ bonding strengths and improves the CO₂ absorptive capacity in both porous Bi₅O₇I Micro-flower and B_i4O₅I₂ [21,22].

Some issues remain to be addressed in understanding the CO₂ photolysis by the TM SACs on BiOX. The defect configurations around the TM SACs are ambiguous, especially whether halogen atoms are present or not near the TM impurity sites [17,18,19,20,21]. Both experimental and theoretical works confirm that the [Bi₂O₂]²⁺ layer is sandwiched by two X⁻ layers in BiOX [23,24,25,26,27,28,29]. However, the unlocked bismuth (Bi) surface without a halogen covering is usually used to simulate photocatalytic CO₂ reduction around the TM SACs [17,18,19,20]. This divergence highlights the necessity of assessing the effects of halogen vacancy (V_halogen) on the microstructures and reduction performance around the TM SACs. V_halogen usually acts as an n-type defect, whereas TM SACs exhibit p-type characteristics and form negatively charged centers to promote the reduction reaction [23,24].

The n–p codoping concept is established based on the electrostatic attraction between the n- and p-type dopants (or defects) with opposite charge states, which affects the ionization, solubility, and charge transfer of the doped semiconductors [30,31,32,33]. For example, the electrostatic attraction within the n–p pair enhances both thermodynamic and kinetic solubilities, creating tunable intermediate bands to effectively narrow the band gap and enhance the visible-light photoactivity of TiO₂ [32]. The n–p pairs limit the applications of Ga₂O₃ by affecting the dopant ionization [33]. On BiOX, n–p codoping can be established via the electrostatic interactions in the combination of p-type TM dopants and n-type V_halogen defects. The effects of n–p codoping on the stability and reducing capacity of TM SACs must be determined. For TM SACs, the d-band theory is universally recognized as a means of evaluating the reducing capacity [34,35]. Consequently, the dependence of the orbital alignment and d-band center of TM_d on n–p codoping deserves systematic study regarding the CO₂ reduction on TM SACs on BiOX.

In this paper, we investigate the effects of n–p codoping on the stability and reducing capacity of TM SACs for photocatalytic CO₂ reduction reaction (CO₂RR) on BiOCl by using first-principles calculation. Previous experimental works found that Fe, Co, Ni, and Cu are all effective SACs for accelerating CO₂RR on BiOX [17,18,19,20]. Consequently, we constructed a dopant–defect combination by pairing these TM SACs and their surrounding Cl vacancies. Thermodynamically, the balance between n- and p-defects is determined by the defect equilibrium, which can be solved by calculating the formation energies of charged defects (ΔH_f) [36,37,38]. Based on the ΔH_f, we find that the Cl monovacancy (V_Cl) is more stable than the Cl multivacancies. n–p codoping is established by the coulomb attraction between the negatively charged p-type TM SACs and positively charged n-type V_Cl. The electrostatic attraction of n–p codoping enhances the stability of TM SACs. Neutral pairs are formed for the Fe, Co, and Ni SACs (P_TM⁰), while the Cu SAC-based pair is dominated by the q = −1 charge state (P_Cu⁻¹). In the n–p pairs, electrostatic interaction settles the electron accumulation around TM SACs. Accumulated electron occupies localized TM_d orbital and upshifts the d-band center toward the Fermi level, enhancing the reducing capacity of TM SACs based on the d-band theory. As a result, the CO₂ absorption is improved along with the enhancement of charge transfer and the decrease in Gibbs free energies. The P_Cu⁻¹ also locates additional electrons surrounding the Cu SACs and forms a half-occupied d_x²_−y² state, further unshifting the d-band center and improving the reducing capacity of TM SACs. The metastability of Cl multivacancies limits the concentration of the n–p pairs with Cl multivacancies (P_TM@nCl (n > 1)). Positively charged centers around P_TM@nCl (n > 1) shield the charge transfer between the P_TM@nCl (n > 1) and the CO₂ molecule, hindering the CO₂ reduction.

2. Computational Details

Density functional theory (DFT) [39] calculations were performed in the Vienna Ab initio Simulation Package (VASP) [40]. The interaction of ions with electrons was determined by projector augmented wave potentials (PAWs) [41]. The electron exchange–correlation energy was treated based on local-density approximation (LDA) [42] because the lattice constants and band structure of BiOCl calculated by LDA are in good agreement with experimental values [43]. The results were also checked based on the Perdew–Burke–Ernzerhof (PBE) functional [44]. The kinetic energy cutoff energy was 500 eV, and the total energy convergence criterion was set to 10⁻⁵ eV. A 4 × 4 × 1 supercell with over 15 Å of lattice constant was chosen to avoid the interactions of adjacent point defects. The K-mesh was optimized based on the minimum energy principle, and 3 × 3 × 1 Monkhorst–Pack grids were found to be sufficient for sampling the Brillouin zone for the supercell. A vacuum space of over 15 Å was applied to avoid the interactions of neighbor images along the z direction.

To solve the defect equilibrium between TM dopants and the compensating V_Cl, the formation energy of charged defects was calculated with the following formula [45,46]:

$$\mathsf{\Delta }{H}_f=E_{tot}(q,\alpha )-E_{tot}(host)+{\displaystyle \sum _i{n}_i}(E_i+{\mu }_i)+q({\epsilon }_{VBM}+E_f)$$

(1)

where E_tot(q,α) and E_tot(host) are the total energy of the defect system and pristine BiOCl, respectively; n is the number of Cr, Bi, and TM dopants; q is the number of electrons transferred from supercell to reservoirs in forming the defect; and μ_i is the chemical formula of constituent i with energy E_i. E_f is the Fermi energy with respect to the valence band maximum (VBM) and ranges from the VBM to the conduction band minimum (CBM). The chemical potentials of Bi (μ_Bi) and TM (μ_TM) were derived from the energies of their corresponding metals. The chemical potentials of O and Cl were limited to avoid the formation of elementary substances and maintain a stable BiOCl compound. μ_Cl and μ_O are defined as:

$$\mathsf{\Delta }{H}_f(\mathrm{B}\mathrm{i}\mathrm{O}\mathrm{C}\mathrm{l})=2{\mu }_{\mathrm{B}\mathrm{i}}+2{\mu }_{\mathrm{C}\mathrm{l}}+2{\mu }_{\mathrm{O}},$$

(2)

$${\mu }_{Cl}\le 0,$$

(3)

where ΔH_BiOCl is the formation energy of BiOCl. Under Cl-poor conditions, the chemical potential of the Cl atom is defined as the energy in the species Cl⁺ (−5.36 eV). The total energies of the charged systems should be corrected for the interaction of the charged defect with the compensating background and its periodic images. We used Makov–Payne (M-P) corrections, formulated as q2 α/2Ɛ L, where L is the linear dimension of the supercell, Ɛ is the static dielectric constant, and α is the Madelung constant.

To evaluate the reaction coordinate of CO₂ reduction, Nørskov’s method [47] was used to calculate the Gibbs free energy difference:

$$\mathsf{\Delta }G=\mathsf{\Delta }E+\mathsf{\Delta }{E}_{ZPE}+T\mathsf{\Delta }S$$

(4)

∆E, ∆E_ZPE, and ∆S denote the absorbed energy difference, the zero-point energy difference, and the entropy difference between the adsorbed state and the corresponding free state for density functional theory calculations, respectively. T is the temperature of the system, 298.15 K.

The reactions of CO₂ adsorption and the *CO₂ plus H formation *COOH process are defined as follows [48]:

$$*+\mathrm{C}{\mathrm{O}}_2\to *\mathrm{C}{\mathrm{O}}_2$$

(5)

$$*\mathrm{C}{\mathrm{O}}_2+H^{+}+{\mathrm{e}}^{-}\to *\mathrm{C}\mathrm{O}\mathrm{O}\mathrm{H}$$

(6)

where * represents adsorption sites, and *CO₂ and *COOH represent adsorption intermediate states. Therefore, the CO₂RR is calculated as:

$$\mathsf{\Delta }{G}_{\mathrm{C}{\mathrm{O}}_2*}=G_{\mathrm{C}{\mathrm{O}}_2*}-G_{*}-G_{\mathrm{C}{\mathrm{O}}_2}$$

(7)

$$\mathsf{\Delta }{G}_{\mathrm{C}\mathrm{O}\mathrm{O}\mathrm{H}*}=G_{\mathrm{C}\mathrm{O}\mathrm{O}\mathrm{H}*}-\frac{1}{2}{G}_{{\mathrm{H}}_2}-G_{\mathrm{C}{\mathrm{O}}_2*}$$

(8)

Charge density redistribution is determined by the charge density difference between the pairs and corresponding isolated systems [49]. The charge density difference is calculated as follows:

$$\mathsf{\Delta }\rho ={\rho }_{AB}-{\rho }_A-{\rho }_B$$

(9)

where ρ_A and ρ_B are the charge densities of isolated systems A and B, respectively, and ρ_AB is the charge density of the pairs.

3. Results and Discussion

3.1. Structures and Stability

In stable BiOCl, the [Bi₂O₂]²⁺ layer is sandwiched by two Cl⁻ layers (Figure 1), and the lattice constant is calculated to be 3.854 Å, agreeing with previous works [13,50,51]. The TM SACs are doped on the BiOX by substituting a bismuth atom based on experiments [16,17,18,19,20]. The TM SAC dopants are denoted as the Bi_TM on the intrinsic BiOX surface as shown in Figure 1b. The dopant–defect pair (P_TM@nCl) comprises the Bi_TM and its surrounding Cl vacancy as shown in Figure 1. The P_TM@3Cl is the combination of a TM SAC and Cl trivacancy. In the Cl trivacancy, the proportion of unlocked Bi atoms is up to 19%, which is large enough to comprise the unlocked Bi region. The structures of the Bi_TM and P_TM@1Cl (P_TM for short) are given in Figure 1b,c, and the P_TM@nCl (n > 1) is shown in Figure S1. To assess the stability of the pairs, the formation energy (ΔH_f) of charged defects is investigated as a function of E_f based on Equation (1). The cathode where the CO₂RR takes place is an electron reservoir and corresponds to the n-type semiconductor. Therefore, we focus on the electron-rich (e-rich) condition where the E_f is close to the CBM. The ΔH_f calculated following Equation (1) also depends on μ_Cl. We focus on the Cl-rich limit (μ_Cl = 0 eV), considering that the electrolyte solution usually accelerates the surface Cl ion exchange in photocatalytic experiments.

Based on previous experiments, surface halogen atoms significantly influence the charge transfer between activation sites and absorbed molecules [52,53]. The stability of Cl vacancies is evaluated prior to the dopant–defect pairs. Figure S2a illustrates the formation energy of the Cl monovacancy V_Cl, Cl di-vacancy (V_2Cl), and tri-vacancy V_Cl (V_3Cl) with varied charge states as a function of E_f. The E_f ranges from the VBM and CBM. One can see that the V_Cl is energy favorable in these three kinds of Cl vacancies if E_f is above the mid-gap (i.e., for the e-rich condition). In the V_3Cl, 3/16 of Bi atoms are uncovered by the surface Cl atoms so that the V_3Cl is large enough with regard to the unlocked Bi region. Consequently, the V_Cl is more stable than the unlocked Bi region in the BiOCl cathode. The V_Cl is capable of forming a donor defect with q = +1 charge state (V_Cl⁺¹, with the superscript referring to the charge state) when the E_f is in most of the gap and becomes a neutral defect only when the E_f is closer to the CBM. The charge transition level Ɛ(+1/−1) is only 0.07 eV above the CBM, indicating that it is a donor defect. The Fermi level of the V_Cl⁰ is located in the conduction band in the spin–polarized density of states, which also indicates the donor defect characteristic.

Considering the stability of the V_Cl, we first investigate the P_TM, which pairs the Bi_TM and surrounding single V_Cl, as shown in Figure 1c. The formation energy ΔH_f is calculated to investigate the stability and ionization of the P_TM and Bi_TM. If E_f is close to the CBM, as shown in Figure 2, the Bi_TM⁻¹ dominates the Bi_TM for all the Fe, Co, Ni, and Cu SACs, indicating the p-type impurities characteristic of these TM SACs on pristine BiOCl in the e-rich condition. When the positively charged V_Cl⁺¹ and negatively charged Bi_TM⁻¹ move close to each other, as shown in Figure 1c, n–p codoping is established due to the coulomb attraction and forms the P_TM pairs. Due to the electrostatic attraction of the n–p codoping, the ΔH_f of the P_TM is lower than the relative Bi_TM as shown in Figure 2. Consequently, the V_Cl facilitates the stability of the TM SACs on the BiOCl. When E_f is close to the CBM, neutral pairs (P_TM⁰) are energetically favorable for the Fe, Co, and Ni SACs, while the negatively charged P_Cu⁻¹ with charge state q = −1 is most stable for the Cu SACs. The P_Cu⁻¹ possesses the lowest E_f (as low as −1.56 eV) in all the defects, suggesting that the P_Cu⁻¹ can exist in great quantities on the BiOX cathode. The q = −1 charge state indicates that the P_Cu⁻¹ further traps electrons from the surrounding lattice besides the inner electrostatic interaction in the pairs. The Cl-poor condition, where ΔH_f is elevated by 5.36 eV depending on the μ_Cl, as shown in Figure S3, hinders the TM SACs from anchoring on the BiOCl.

The electrostatic interaction is the key factor for establishing n–p codoping. We investigate the electrostatic interaction in the pairs by calculating the charge density redistribution. The charge density redistribution is determined as the charge density difference of the BiOCl system after and before forming P_TM pairs. In Figure 3, the yellow and blue isosurfaces refer to the electron accumulation and dissipation regions, respectively. Before combining into P_TM pairs, isolated TM SACs are all negatively charged, and V_Cl is positively charged based on the ΔH_f in Figure 2 and Figure S3a. The negative charge density localizes around the TM SACs in the Bi_TM, while the positive charge localizes around the vacancy site in the V_Cl as shown in Figure 3e,f. In all the P_TM pairs, negative and positive charges also separated at the adjacent TM SACs and V_Cl site as shown from Figure 3a–d. An obvious coulomb interaction is elicited between the closer positive and negative charges, forming more stable n–p codoping.

3.2. Modulation of CO₂ Reduction

CO₂ absorption (*CO₂) is usually the rate-determining step in the CO₂RR [15,18]. We investigate CO₂ absorption on defective BiOCl systems to evaluate the effects of n–p codoping on the photocatalytic CO₂RR. From Figure 4a, one can see that the CO₂ is physically absorbed on the pristine BiOCl surface, and hardly any charges transfer between the BiOCl and the CO₂. Only the surface charge density distribution of the BiOCl is disturbed by absorbed CO₂. The charge transfer between the V_Cl⁺¹ site and absorbed CO₂ is enhanced, but still only physical absorption occurs with over 3.5 Å of distance as shown in Figure 4b. The Gibbs free energy difference (ΔG) is also calculated based on Equations (5) and (6) to evaluate the CO₂ absorption. Negative ΔG refers to the chemisorption of CO₂ on the photocatalyst. In Figure S4, one can see that the ΔG is 0.16 eV on the V_Cl⁺¹, lower than that on pristine BiOCl (0.32 eV). Consequently, the V_Cl⁺¹ mildly improves the CO₂ absorption. For the CO₂ absorption on the dopant–defect pairs, dramatic charge transfers are found in the neutral P_Fe⁰, P_Co⁰, and the negatively charged P_Cu⁻¹ as shown in Figure 4c, Figure 4d, and Figure 4e, respectively. The lengths of C–TM bonds between the absorbed CO₂ and these pairs are 1.90 Å, 1.97 Å, and 1.97 Å as shown in

Table 1. Calculated TM–C bond length (Å) for absorbed *CO₂ (d_(*CO2)) and *COOH (d_(*COOH)), bond angle of absorbed *CO₂ molecule (θ(°)), and d-band center (eV). “/” indicates that the distance between physically absorbed CO₂ and the activation site is beyond 3.50 (Å).

Defect	d(*CO2) (Å)	d(*COOH) (Å)	d-Band Center (eV)	Θ (°)
BiFe0	/	1.91	−2.54	179.75
BiFe−1	1.90	1.83	−1.26	145.77
PFe0	1.90	1.83	−1.27	147.15
BiCo0	/	1.85	−1.81	179.53
BiCo−1	/	1.80	−2.45	148.32
PCo0	1.97	1.81	−0.83	147.70
BiNi0	/	1.86	−3.85	179.66
BiNi−1	/	1.81	−2.09	179.15
PNi0	3.31	1.84	−1.46	177.90
BiCu0	/	1.94	−2.81	179.77
BiCu−1	/	1.93	−2.31	179.18
PCu0	/	1.93	−2.31	179.13
PCu−1	1.97	1.88	−1.45	153.71
PCu@2Cl0	/	1.88	−3.12	149.73
PCu@3Cl0	/	1.85	−1.99	179.40

, while the ΔG sharply declines to −0.52 eV, −0.83 eV, and −0.54 eV for the P_Fe⁰, P_Co⁰, and P_Cu⁻¹, respectively, as shown in Figure 5. As a result, the n–p codoping facilitates the absorption of the CO₂ on the P_Fe⁰, P_Co⁰, and P_Cu⁻¹. The charge transfers in the P_Ni⁰ are slightly weaker, and the distance between the absorbed CO₂ and Ni SACs (3.31 Å) is larger than that in other P_TMs, but the ΔG still declines to −0.31 eV. Consequently, the CO₂ absorption is also accelerated by the n–p codoping in the P_Ni⁰.

For comparison, we also evaluated the CO₂ absorption on the isolated Bi_TM⁰ and Bi_TM⁻¹. On the neutral Bi_TM⁰, the CO₂ is only absorbed physically on the TM SAC sites with fewer charge transfers as seen in Figure S5. The distances between the CO₂ and Bi_TM⁰ sites are all beyond 4 Å and are obviously larger than the distances between the CO₂ and the dopant–defect pairs as shown in Table 1. The ΔGs of CO₂ absorption on the Bi_TM⁰ are all positive as shown in Figure 5. The CO₂ is therefore still physically absorbed at the neutral Bi_TM⁰ site. For the CO₂ absorption on the negative Bi_TM⁻¹, we find that the CO₂ is only chemically adsorbed at the Bi_Fe⁻¹ but is still physically absorbed at the Bi_Co⁻¹, Bi_Ni⁻¹, and Bi_Cu⁻¹. In detail, an intense charge transfer is formed between the CO₂ and Bi_Fe⁻¹ site as shown in Figure S6a, while the length of the C-Fe bond is 1.90 Å as shown in Table 1. The ΔG of CO₂ absorption is −0.18 eV in the Bi_Fe⁻¹. On the other hand, relatively few charge transfers are formed between the CO₂ and Bi_Co⁻¹, Bi_Ni⁻¹, and Bi_Cu⁻¹ sites as shown in Figure S6b, 6c, and 6d, respectively. The distances between CO₂ and these activation sites are over 4.0 Å as shown in Table 1. The relative ΔGs of CO₂ absorption are 0.15 eV, 0.23 eV, and 0.12 eV in the Bi_Co⁻¹, Bi_Ni⁻¹, and Bi_Cu⁻¹ systems. Based on previous experiments, the carboxylate pathway is the most common route for the CO₂RR on Fe-, CO-, and Cu-doped bismuth oxyhalide [16,17,18,19,20]. *CO₂ hydrogenation (*COOH) follows close behind CO₂ absorption (*CO₂) in the carboxylate pathway and is reported to be also an important step influencing CO₂ reduction [15,18]. We therefore also assess the ΔG of both *CO₂ and *COOH in all defective BiOCl systems. In Figure 5, one can see that n–p codoping facilitates both the CO₂ absorption and *CO₂ hydrogenation with negative ΔG in the BiOCl systems by Fe, Co, and Cu SAC doping. Such an enhancement of CO₂ photoreductions corresponds with previous experiments [18,19,20]. For the P_Ni⁰-doped BiOCl, the CO₂ absorption is exothermic, but the *CO₂ hydrogenation is endothermic, which is related to the fact that the CO₂ activation prefers the carbide pathway, in which the *COOH is difficult to generate [17].

3.3. Orbital Alignments and d-Band Center of the TM_3d

The TM_3d orbital is usually crucial to the reducing capacity of TM SACs [54,55]. To account for the improvement in the CO₂ absorption by the n–p codoping, we investigate the d-band center of the TM_3d orbital states on the BiOCl systems. Based on the d-band theory, the d-band center close to the Fermi level usually facilitates the reducing capacity of TM SACs [30,31]. For comparison, the d-band centers of the Bi_TM⁰ and Bi_TM⁻¹ are also calculated as shown in Table 1. One can see that the d-band centers of all Bi_TM⁰ dopants are located far away from the Fermi level. For instance, the d-band centers of the Bi_Fe⁰, Bi_Ni⁰, and Bi_Cu⁰ are as low as −2.54 eV, −3.85 eV, and −2.81 eV, respectively. Low d-band centers correspond to the low reducing capacity of the Bi_TM⁰ and result in physical absorption of the CO₂. For the Bi_TM⁻¹, the d-band center of Bi_Fe⁻¹ is significantly shifted from −2.54 eV to −1.26 eV compared with Bi_Fe⁰, resulting in the enhancement from physical absorption to chemical absorption. The d-band center in the Bi_Co⁻¹ decreases from −1.81 eV to −2.45 eV, so the CO₂ remains physically absorbed on the Bi_Co⁻¹ site. Although the d-band centers of the Bi_Ni⁻¹ and Bi_Cu⁻¹ are shifted toward the Fermi level compared with the Bi_Ni⁰ and Bi_Cu⁰, they are still lower than the Fermi level by −2.09 eV and −2.31 eV. Consequently, the CO₂ is also physically absorbed on these systems. As shown in Table 1, the d-band centers are upshifted to −1.27, −0.83, −1.46, and −1.45 eV for the P_Fe⁰, P_Co⁰, P_Ni⁰, and P_Cu⁻¹, respectively, and are comparable to those (−1.56 eV) of the Pt SACs anchored at the edge of graphene, which are excellent photocatalysts for H₂ reduction [56]. Consequently, the modulation of CO₂ absorption on the dopant–defect pairs derives from the enhancement of the reducing capacity of TM SACs with the upshift in the d-band center by the n–p codoping.

The d-band center is determined by the orbital alignment and occupation of the TM-3d. To account for the effects of n–p codoping on the d-band centers, we further calculated the spin-resolved states of density (PDOSs) of the BiOCl systems before and after n–p codoping to illustrate the orbital alignment and occupation. Based on the crystal field theory, the degeneracy of the TM-3d orbital states is broken in a process dependent on the molecular symmetry. As shown in Figure 1b,c, the TM ion is located at the center of a square plane formed by four nearest neighboring oxygen atoms. In this local symmetry, the TM_3d orbitals split into four groups d_xz + d_yz, d_xy, d_z², and d_x²_−y². In the neutral Bi_TM⁰, the oxidation states of the TM dopant are all +3. As shown in Figure 6a, the d electron configuration obeys Hund’s rule in Bi_Fe⁰. Five spin-up d orbitals are fully occupied, while five spin-down d orbitals are entirely empty. In the Bi_Co⁰, Bi_Ni⁰, and Bi_Cu⁰, the d electron configurations break Hund’s rule as shown in Figure 6a. In detail, four spin-up d orbitals and only the spin-down d_xz + d_yz orbital are occupied in the Co_3d, when the electron configuration of Co is changed from 3d⁷4s² to 3d⁶4s⁰ after losing three valence electrons in the Bi_Co⁰. In the Bi_Ni⁰, four spin-up d orbitals and three spin-down d orbitals (d_xz + d_yz orbitals and d_xy orbital) are occupied to form the 3d⁷4s⁰ configuration. In the Bi_Cu⁰, the Cu_3d orbitals are spin-degenerated, and spin-degenerated d_xz + d_yz, d_xy and d_z² orbitals are occupied to change the 3d¹⁰4s¹ to a 3d⁸4s⁰ configuration.

In the dopant–defect pairs P_TM⁰, negatively charged TM dopants and positively charged V_Cl⁺¹ are stabilized by the electrostatic interaction of the n–p codoping. The electron accumulation changes the oxidation state of the TM dopant from +3 to +2. Compared with the electron configuration in the Bi_Tm⁰, electron accumulation is located around the TM while charge depletion forms around the V_Cl⁺¹ and durable electrostatic attraction is established based on the Coulomb’s law as shown in Figure 4. Accumulated electrons occupy the localized empty TM_3d orbital as shown in Figure 6b. In detail, spin-down d_z² is further occupied in the P_Fe⁰ to form the 3d⁶4s⁰ configuration. In the P_Co⁰, the electron from the V_Cl⁺¹ further occupies the spin-down d_xy to form the 3d⁷4s⁰ configuration. Spin-down Ni_d_z² is further occupied, leading to spin-degenerated Ni_3d states in the P_Ni⁰, while the spin degeneration of the Cu_3d states is broken along with the spin-up d_x²_−y² being occupied. Above all, newly occupied TM_3d orbitals are localized and remain at the Fermi level in the P_TM⁰, gathering both the spin-up and spin-down TM_3d orbitals together and bringing them to the Fermi level. As a result, the d-band centers are shifted to the Fermi level compared with those in the Bi_TM⁰. For the P_Cu⁻¹ as shown in Figure 6b, the added electrons also accumulate around the Cu SACs and fill into Cu_d_x²_−y², inducing spin-degenerated Cu_3d orbital states and localized half-occupied d_x²_−y². A stronger localized characteristic of Cu_3d is formed around the Fermi level, further shifting the d-band center to the Fermi level.

The electron accumulation around the TM SACs plays an important role in changing the d-orbital alignment and shifting the d-band center in the P_TM. In contrast to the Bi_Fe⁻¹, the reducing capacity in the Bi_TM⁻¹ is weaker than that in the P_TM. To account for the reduced capacity in the Bi_TM⁻¹, we investigate its d orbital alignment and electron configuration in detail. The electron accumulation is also formed around the Bi_Fe⁻¹ as shown in Figure S7a. The spin–polarized PDOS of TM_3d in the Bi_TM⁻¹ is shown in Figure S8a. The Bi_Fe⁻¹ keeps in line with the P_Fe⁰ on the d-orbital alignment, the 3d⁶4s⁰ configuration, and even Hund’s rule, resulting in a similar d-band center and ΔG for the CO₂ absorption. Although electron accumulation occurs around the Bi_Co⁻¹ as shown in Figure S7b, the 3d⁷4s⁰ configuration stays the same as that in the P_Co⁰, but the spin–polarized d-orbital alignments exhibit a dramatic change compared with those in the P_Co⁰. Five spin-up orbitals are occupied in the Bi_Co⁻¹ as shown in Figure S8a, while only four spin-up orbitals are occupied in the P_Co⁰ (the same as in the Bi_Co⁰) as shown in Figure 6. Such variation changes the high spin state (3μ_B) to a low spin state (1μ_B) as well as induces greater broadening of Co_3d orbital states and then a more negative d-band center in the Bi_Co⁻¹. The electron localization is weakest in the Bi_Ni⁻¹ as shown in Figure S7c. As a result, although the electron configuration in the Ni_3d is same as the cases in the P_Co⁰, the energy level and broadening of Ni_3d, which also influence the d-orbital alignment, are extended to a low energy level in the Bi_Ni⁻¹, resulting in a much lower d-band center than that in the P_Co⁰. The Bi_Cu⁻¹ possesses the same orbital alignment, orbital occupation, and d-band center as the P_Cu⁰ as shown in Figure S8d. Therefore, it possess a similar ΔG for CO₂ absorption. The stronger reducing capacity of the P_Cu⁻¹ derives from the accumulation of additional electrons around the Cu SACs. On the other hand, compared with the Bi_TM⁰, Bi_Fe⁻¹, Bi_Ni⁻¹, and Bi_Cu⁻¹ upshift the d-band center because additional electrons accumulate around the TM SACs and occupy more 3d orbitals. In contrast to the Bi_Co⁰, the Bi_Co⁻¹ moves the d-band center downward with negative energy following the transition of spin states. Consequently, the electron accumulation plays a key role in the d-orbital alignment and the d-band center. The transition of spin states also influences the d-band center.

3.4. Discussion

Finally, the effect of Cl multivacancies on the n–p codoping is discussed. Taking the most sable P_Cu⁻¹ as an example, the P_Cu@2Cl and P_Cu@3Cl are pairs with double and ternary Cl vacancies, as shown in Figure S1. Both the P_Cu@2Cl and P_Cu@3Cl exhibit lower formation energy than the P_Cu⁻¹. However, a high ΔH_f hinders the stability and solubility of V_2Cl and V_3Cl, as shown in Figure S2. The solubility of both the P_Cu@2Cl and P_Cu@3Cl is lower than that of the P_Cu⁻¹. In the e-rich condition, both the P_Cu@2Cl and P_Cu@3Cl exhibit the charge state of q = 0, as shown in Figure S2b (denoted as P_Cu@2Cl⁰ and P_Cu@3Cl⁰). The ΔG profiles in Figure 7a exhibit obvious high free energy (0.61 eV and 0.43 eV) for CO₂ absorption at both the P_Cu@2Cl⁰ and P_Cu@3Cl⁰ sites, which is higher than that at the P_Cu⁻¹ and even higher than that at the V_Cl⁺¹. Relatively high d-band centers lead to a weak reducing capacity as shown in Table 1. Obviously positive charge accumulation around the P_Cu@2Cl⁰ and P_Cu@3Cl⁰ is also observed as shown in Figure 7c,d, shielding the electron transfer to absorbed CO₂ and decreasing the reducing capacity. Consequently, the P_Cu⁻¹ enhances the CO₂ absorption, and doping the TM on the unlocked Bi region is not an efficient strategy to facilitate the CO₂RR.

4. Conclusions

In summary, n–p codoping engineering is introduced to account for the modulation of photocatalytic CO₂ reduction on a BiOCl-based cathode by using first-principles calculation. n–p codoping is established via the Coulomb interactions between positively charged TM SACs and the negatively charged Cl vacancy (V_Cl) in the dopant–defect pairs. Based on the formation energy of charged defects, we find neutral dopant–defect pairs for the Fe, Co, and Ni SACs (P_TM⁰) and the q = −1 charge state of the Cu SAC-based pair (P_Cu⁻¹). The electrostatic attraction of the n–p codoping in pairs strengthens the stability and solubility of TM SACs and stabilizes the electron accumulation around the TM SACs. Accumulated electrons occupy the localized TM_d orbital and change the d-orbital alignment, gathering the d-orbital states and shifting the d-band center toward the Fermi level and enhancing the reducing capacity of TM SACs based on the d-band theory. As a result, the CO₂ absorption is improved with the enhancement of charge transfer and the decrease in Gibbs free energies. Besides the electrostatic attraction of the n–p codoping, additional electrons in the P_Cu⁻¹ also accumulate, surrounding Cu SACs and forming a half-occupied d_x²-_y² state, which further upshifts the d-band center and improves photocatalytic CO₂ reduction. For the pairs consisting of TM SACs and Cl multivacancies P_TM@nCl (n > 1), their concentration is limited due to the metastability of Cl multivacancies. The positively charged center around the TM SACs blocks the charge transfer, indicating that doping the unlocked region with TM atoms does not facilitate CO₂ reduction.