Dual activation and C-C coupling on single atom catalyst for CO₂ photoreduction

Abstract

An excellent single-atomic photocatalyst, Ti@C₄N₃, is theoretically found to effectively convert CO₂ to C₂H₆ by density functional theory (DFT) calculations and non-adiabatic molecular dynamics (NAMD) simulations. The Ti@C₄N₃ photocatalyst has remarkable stability both thermally, chemically, and mechanically. Electronically, it has strong absorption properties (λ = 327.77 and 529.61 nm), suitable band positions, and a long photogenerated electron lifetime (τ_e = 38.21 ps), allowing photogenerated electrons to migrate to the surface. Notably, the high-valence active site effectively activates two CO₂ through dual activation: Under light irradiation, the weakly adsorbed CO₂ undergoes photo-induced activation by the photoelectron of conduction band minimum (CBM); without light, the high Lewis acidity of the Ti site induces CO₂ activation through back-donating π-bond. Contrast simulation results uncovered that dual activation of CO₂ is attributed to the thermal and photonic synergy. Furthermore, two activated CO₂ species under light easily couple to form oxalate with the barrier of 0.19 eV, and further reduced to C₂H₆ with a low activation energy of 1.09 eV.

Introduction

The latest few years has witnessed a significant global push towards achieving carbon neutrality as a response to the escalating impact of greenhouse gases (e.g. CO₂) on the planet^1,2. Compared to other strategies (directly reducing CO₂ emission, CO₂ capture and storage, chemical conversion of CO₂ triggered by heat or electric energy), the conversion of solar-powered CO₂ photoreduction to value-added chemicals or fuels^3,4 (e.g. CH₃OH, CH₄, C₂H₄, and C₂H₆) is an appealing and green solution to ameliorate energy shortages as well as reduce CO₂ emission. Essentially, however, activation of CO₂ is both thermodynamically and kinetically difficult because of the two delocalized ${{\rm{\pi }}}_{3}^{4}$ bonds with the bond length of 1.16 Å. Thus, it is challenging to transform CO₂ selectively into other chemicals at mild reaction conditions.To this end, much effort has been made to investigate the developments of photocatalysts for CO₂ reduction reaction (CO₂RR), e.g. metal nanoparticle⁵, metal oxides⁶ and single site catalysts⁷.

Photocatalysis is a light-driven process that uses a semiconductor catalyst to generate excited electrons and holes when exposed to photons with sufficient energy. These charge carriers trigger various redox reactions, resulting in the formation of end products. Within this context, CO₂ photoreduction involve the C₁ products⁸ (e.g. CO⁷, CH₃OH⁹, CH₄¹⁰), C₂ products^11,12 (C₂H₆¹³, C₂H₅OH¹⁴, CH₃COOH¹⁵, C₂H₄³ and so on) and C₃ products¹⁶ (C₃H₈¹⁷, CH₃COCH₃¹⁸ and so on). Compared to C₁ product, however, rational design of high-performance (e.g. high selectivity) photocatalysts for multi-carbon product is highly challenging. In this vein, single-atom catalysts (SACs)¹⁹, featuring catalytic activity adjustability as well as designed synthesis diversity²⁰, shown good photocatalytic performance for CO₂RR, e.g. Er₁/CN-NT²¹, Fe SAS/Tr-COF²². Generally, an active site on heterogeneous catalysts can activate only one CO₂, leading to C₂ coupling steps requiring the synergistic collaboration of adjacent sites. Thus, C₂ coupling step during CO₂RR process is of great difficult to trigger in catalysts with highly discrete sites (e.g. SACs) due to limited reductive electronic sources as well as high atomically dispersion of active sites^23,24,25,26.

Herein, we theoretically designed an outstanding single-atom photocatalysts Ti@C₄N₃ by a combination of DFT and NAMD simulations. The supporting Ti@C₄N₃ photocatalyst features excellent stability, good absorption properties, suitable band positions and long photocarrier lifetime. Notably, real-time time dependent DFT (rt-TDDFT) simulation reveal that high-valence Ti site renders the dual activating CO₂ on Ti@C₄N₃: high Lewis-acidity of Ti site can thermally induce activation of CO₂ by π back-donating bond without light; and under visible light irradiation the weak-adsorbed CO₂ undergoes photo-induced activation by the photoelectron from CBM. Catalytic studies reveal that the two activated CO₂ is easily coupled to be oxalate, which is further reduced to be C₂H₆.

Results

Stability of Ti@C₄N₃. Via three-step strategies (binding energy, band gap and adsorption energy of CO₂), we systematically screen out Ti@C₄N₃ as a potential photocatalyst (Detailed information see Supplementary Information file). Furthermore, we conduct a comprehensive stability investigation of Ti@C₄N₃ in three aspects: thermal stability, chemical stability in aqueous liquids, and mechanical stability. Firstly, Ab-initio molecular dynamic (AIMD) simulations were firstly carried out for a duration of 10 ps at a temperature of 500 K and indicate that both the energy and bonds of the structure fluctuates within a constant range, shining light on excellent thermal stability up to 500 K (see Fig. 1a). To further identify the thermal stability of SACs, an essential factor is the potential for single-atom agglomeration on the substrate. Studying the aggregation process of Ti atoms on the C₄N₃ surface, the climbing image-nudged elastic band (CI-NEB) method was employed to identify diffusion pathway and calculate diffusion barriers (see Supplementary Fig. 9). The results reveal that, when a single Ti atom aggregates with its nearest neighboring Ti atom, the first Ti atom initially diffuses from the current N₃ hollow site to a middle site between them, encountering a barrier of E_a = 2.38 eV. Subsequently, the other Ti atom also migrates to the vicinity of this middle site to agglomerate with the first Ti atom, forming a Ti-Ti distance of d_(Ti-Ti) = 2.603 Å, with the uphill barrier being E_a = 3.93 eV, which represents the rate-determining step (RDS) for aggregation. The entire process of aggregation between the two Ti atoms is endothermic, involving an enthalpy change of 4.79 eV. These results indicate that the aggregation of single Ti atoms on C₄N₃ substrate is exceptionally challenging both thermodynamically (∆H = 4.79 eV) and kinetically (E_a = 3.93 eV). Consequently, Ti@C₄N₃ remains thermodynamically robust under photocatalytic conditions. To gauge chemical stability on aqua liquid environment, a solid−liquid interface model is constructed in Supplementary Fig. 10. The equilibrium structure was used to calculate the kinetic activation energy of Ti atom leaching from N₃C site. The “slow-growth” method²⁷ involves selecting the reaction coordinate in which the Ti-N bonds are elongated (see Fig. 1b). The analysis reveals that the free energy progressively increases with the displacement of the Ti atom from the surface, culminating in a value of 2.47 eV at the reaction endpoint, shedding light on the extreme difficulty of the three Ti−N bonds breaking under aqua environment²⁸. Next, mechanical stability are performed by calculating the independent elastic stiffness tensor components. For this two-dimensional (2D) lattice with tetragonal symmetry, the independent elastic constants are C₁₁ = C₂₂ = 136.08 GPa and C₁₂ = 8.96 GPa, fully meeting the Born−Huang criteria (C₁₁ > 0 and C₁₁ > | C₁₂ | )²⁹. Moreover, Ti@C₄N₃ feature the smaller Young’s modulus of 136.25 GPa than those of many other 2D materials (1000 GPa for graphene³⁰, 330 GPa for MoS₂³¹ and 279 GPa for h-BN³²). Meanwhile, the Poisson’s ratio (0.066) of Ti@C₄N₃ reflects its isotropically higher stiffness than that of Zn metal (0.27). Generally, Ti@C₄N₃ catalyst has excellent mechanical stability.

Fig. 1: Calculation of stability.

a Total energy fluctuation of Ti@C₄N₃ during AIMD simulation at 500 K. b Free energy profile during of Ti atom dissociated from C₄N₃ surface in aqueous solvent. c, d Polar diagrams of Young’s modulus and Poisson’s ratio of Ti@C₄N₃, respectively.

Optical properties and Photocarrier Dynamics of Ti@C₄N₃: Electronically, compared to the main contribution of N atoms for valance band maximum (VBM) and CBM of C₄N₃, the VBM of Ti@C₄N₃ is also concentrated on the 2p states of N atoms, while the CBM is accumulated at 3d state of Ti atom (see Fig. 2a, b). The pristine C₄N₃ features conductor property, where the 2p states of N atoms occupy across the Fermi level, while for Ti@C₄N₃ the 3d state of Ti atom occupy the Fermi level above instead of 2p states in N atoms and effectively reduce band dispersion at the vicinity of Fermi level, rendering to open band gap of Ti@C₄N₃ (see Fig. 2c). It can be seen from the band gap that Ti@C₄N₃ (E_g = 0.97 eV) from HSE-06 calculations has two distinct absorption peaks (327.77 and 529.61 nm) show in Fig. 2d, which are redshifted by 52.37 and 254.21 nm compared to the maximum absorption peak of pristine C₄N₃ (λ = 275.40 nm). Therefore, the Ti atoms loading enables the enhancement of C₄N₃ in the absorption of visible light, fully improving the ability to capture sunlight.

Fig. 2: Optical properties.

a VBM and b CBM distributions of C₄N₃ and Ti@C₄N₃, respectively. c Distribution of C, N and Ti atoms on the energy band structure of C₄N₃ and Ti@C₄N₃ respectively. The coordinates of the highly symmetric points are as follows: G (0.00 0.00 0.00), M (0.50 0.00 0.00), K (0.33 0.33 0.33). d Optical adsorption spectra of pristine C₄N₃ and Ti@C₄N₃, their absorption coefficients are illustrated by blue and pink lines, respectively.

Furthermore, the temperature dependence of the optical properties of Ti@C₄N₃ was investigated. According to the electron-phonon interaction³³, we utilize one-shot method to sample the cartesian coordinates of atoms at different temperatures^34,35,36, and finally obtain the average values of samples at different coordinate sets at a given temperature ranging from 273.15 K to 373.15 K with the step size of 10 K. The band gap (see Supplementary Fig. 11a) of Ti@C₄N₃ calculated by the GW method³⁷ indicated that the band gap decreases with the increase of temperature, compatible with most of semiconductors. In addition, the temperature does not have a notable impact on the probability of electron transition from VBM to CBM (see Supplementary Fig. 11b), as evidenced by the variation trend in sum result of the squares of the dipole transition matrix elements (P²) at different k points. Furthermore, the effect of electron-phonon interaction on carrier mobility was indirectly assessed by the effective mass of the carriers (see Supplementary Fig. 11c, d). Photogenerated electron effective mass is temperature-insensitive with around 4m_o, whereas photogenerated hole effective mass decreases with temperature. The effective masses of photogenerated electrons and holes tend to approach each other with increasing temperature, resulting in faster carrier recombination during transport and contributing to the low quantum efficiency commonly observed in semiconductor photocatalytic materials. Essentially, non-radiative relaxation and molecular thermal vibrations in semiconductors can lead to photothermal effects, which can reduce the apparent activation energy of photocatalysis and promote carrier mobility or reactant mass transfer. However, excessive temperature can lead to a stronger photothermal effect, causing the effective mass of photogenerated electrons to exceed that of holes, which can further accelerate carrier recombination and reduce photocatalytic activity. Thus, the lower m_e^* than m_h^* demonstrated Ti@C₄N₃ keeps high photocatalytic activity up to 373 K.

The dynamics of photogenerated carriers is an important factor affecting the quantum efficiency of photocatalytic reaction. Ab-initio NAMD simulations show that after light excitation the photogenerated electrons accumulate on the CBM of Ti@C₄N₃, due to the separation of photogenerated carriers. By fitting the long-time evolution curve with P(t) = exp(-t/τ)³⁸ (see Supplementary Fig. 12), the lifetime (τ_e) of the photogenerated electron is 38.21 ps. The picosecond-range lifetime of photogenerated electrons in Ti@C₄N₃ semiconductor suggests that they have ample time to migrate to the catalyst surface and engage in the reduction reaction. These results demonstrated that the generation of Ti@C₄N₃ photogenerated carriers is very fast (femtosecond magnitude), much higher than the carrier trapping and compounding (picosecond magnitude), which is consistent with the characteristic time of photocatalysts reported in other literatures^39,40. Thus, ensuring enough photogenerated carriers to migrate to the surface and undergo surface charge transfer for the photocatalytic reaction. To sum up, Ti@C₄N₃ photocatalyst has excellent photocatalytic activity.

Activation Mechanism of CO₂. As a crucial prerequisite for catalytic reaction, the reactant adsorption properties were of great importance for the activation⁴¹. In general, the adsorption strength is associated with activation degree of reactant. The CO₂ molecule, featuring linear configuration with two ${{\rm{\pi }}}_{3}^{4}$ bonds, possesses the non-bonding highest occupied molecular orbital (HOMO) with energy level −10.36 eV and the σ-type and π-type antibonding lowest unoccupied molecular orbital (LUMO) with energy level of 0.36 eV and 0.65 eV respectively (see Supplementary Fig. 15). Activating CO₂ poses a challenge as transferring electrons from the deep-energy level HOMO to the active site of the catalyst incurs a significant energetic penalty. Conversely, the active site of the catalyst donates electrons to the LUMO of CO₂ to facilitate the activation process.

The AIMD simulation results provide insight into the activation process of CO₂ by Ti@C₄N₃. When CO₂ approaches the catalyst, its bond angle decreases from 180° to about 135° within 30 fs, and the two C-O bond lengths gradually elongate from 1.16 Å to 1.30 Å and 1.20 Å, respectively (see Supplementary Fig. 16a, b). The chemisorption characteristic is revealed by the obtained adsorption energy of −1.05 eV, indicating complete activation of the CO₂ molecule both energetically and structurally through thermal-induced activation. Bader charge analysis indicates that the catalyst donates 0.65 |e| electrons to the adsorbed CO₂ by back donating π bond (see Fig. 3a). The energy levels of Ti d orbitals and CO₂ π^* orbitals are matched, leading to partial occupation of the formed d-π^* orbitals and spin-polarization near the Fermi level. This results in a 0.2 eV energy level reduction in spin-up d-π^* states compared to spin-down d-π^* states (see Supplementary Fig. 17a). The well-dispersed π^* states of CO₂ below the Fermi level also validate activation states. Despite the adsorption of CO₂, the Ti site retains a charge of 2.31 |e| due to the high valence of Ti (IV) ions.

Fig. 3: Activation mechanism of CO₂ molecules.

Optimised structure of CO₂ adsorption configuration, differential charge density, bader charge and the orbital interaction of CO₂ and Ti@C₄N₃, a the adsorption of only one CO₂ molecule, b the adsorption of two CO₂ molecules. The yellow and blue areas show the accumulation and depletion of charges with the isosurface of ±0.025 e bohr⁻³. The bond angle of CO₂-2 molecule on Ti@C₄N₃ evolve with time at c light-300K, d light-0K and e no-light-300K, respectively. Snapshots at three representative times (0, 20, 60 and 80 fs) during the CO₂-2 activation process on Ti@C₄N₃ are shown in the insets ①~④; Time evolution of electron populations for the atoms at the reaction center in the period of 0~80 fs at f light-300K, g light-0K and h no-light-300K, respectively; i Piecewise temporal evolution of electron populations at light-300K. The O atom near the Ti atom is labeled O1, and the O atom far from the Ti atom is labeled O₂.

In photocatalytic reactions, the gas hourly space velocity (GHSV) of CO₂ is typically high, which increases the probability of multiple collisions between the substrate and catalyst. Therefore, it is necessary to investigate the multiple adsorptions of CO₂ on the catalyst. The AIMD simulation results of dual CO₂ adsorption on Ti@C₄N₃ indicate that after 10 ps, the first adsorbed CO₂ (CO₂-1) still features structural bending of about 45° and bond-length stretching of about 0.13 Å (see Supplementary Fig. 19a, b). However, the second CO₂ (CO₂-2) molecule maintains a linear configuration with weak physisorption of −0.39 eV adsorption energy. Electronically, the CO₂-2 only involves 0.03 |e| electrons from the substrate and therefore attains a non-activated state (see Fig. 3b). Supplementary Fig. 17b shows that the Ti d orbitals can only partially match the π^* orbitals of the CO₂-2 above the Fermi level, revealing its weak adsorption on the catalyst. In summary, for high GHSV of CO₂ gas, Ti@C₄N₃ activates only one CO₂ molecule, accompanied by weak adsorption of another CO₂ molecule.

As is well known, photocatalytic reaction proceeds under continuous light irradiation⁴². Whether it is possible for weakly adsorbed CO₂-2 molecule to be activated under light? Thus, ab initio rt-TDDFT molecular dynamics simulation under light illumination⁴³ (see Supplementary Fig. 22) was used to study the kinetically activation process of dual CO₂ adsorption on Ti@C₄N₃ catalyst. Fig. 3c–i depicts the dependence of geometrical factors (bond angles) and electronic populations over irradiation time. To investigate the underlying mechanism for dual activation under light, we divide the electron density into each reaction center atom by Hirshfeld charge analysis. The whole reaction process involves three distinct stages (see Fig. 3f): (1) Electron excitation (0~20 fs, white span): Under the activation of the laser pulse, C₄N₃ substrate lose about 0.03 e⁻ charge (at t = 17 fs). As shown in Fig. 3i, the Ti site initially acquires around 0.01 e⁻ charge from C₄N₃ substrate at t = 15 fs. Simultaneously, the weakly-adsorbed CO₂-2 molecule gains 0.01 e⁻ charge, while the Ti atom experiences a decrease of approximately 0.02 e⁻ from 17 fs to 20 fs. This suggests a potential charge transfer between Ti atom and CO₂-2. During this process, the bond angle of CO₂-2 becomes bent to 165° (see Fig. 3c). (2) Charge transfer (20~60 fs, pink span): In this stage, electronic populations of related species undergo significant changes compared to the first stage. Firstly, a Ti atom transfers approximately 0.05 e⁻ charge to a CO₂-2 molecule, causing the CO₂-2 bond angle to bend from 165° to 157° within 20~40 fs. This indicates that the singular Ti site in Ti@C₄N₃ functions as a carrier bridge, providing reductive photoelectrons to activate CO₂-2. Subsequently, C₄N₃ substrate also contributes about 0.05 e⁻ charge to CO₂-2 within 40~60 fs, resulting in a maximum gain of approximately 0.10 e⁻ charge at t = 60 fs. This reveals a charge transfer process from Ti@C₄N₃ support to CO₂-2. Concurrently, the CO₂-2 bond angle further bends from 165° to 152° at 20~60 fs, and the C-O1 bond gradually lengthens to around 1.35 Å with oscillations. This behavior indicates that the CO₂-2 molecule undergoes vibrational excitation following the charge transfer in this stage. (3) CO₂ activation deeply (60~80 fs, bule span): During the initial 10 fs (t = 60~70 fs) of this stage, the geometrical structure of CO₂-2, including bond angles and bond lengths, remains unchanged, and the electronic population of CO₂-2 shows minimal changes, displaying only slight oscillations. During this time, Ti@C₄N₃ acquires approximately 0.15 e⁻ charge at 67 fs. Meanwhile, thermal-activated CO₂-1 experiences a loss of approximately 0.22 e⁻ charge. As a result, the bond angle of CO₂-1 bends to 148° (see Supplementary Fig. 24), and the C-O bonds elongate to 1.50 and 1.40 Å, respectively (see Supplementary Fig. 23a). These changes indicate that CO₂-1 further amplifies the activation process by providing electronic feedback to Ti@C₄N₃. Furthermore, at t = 70~80 fs the CO₂-2 acquired another 0.15 e⁻ charge, resulting in the bond angle bending from 152° to 145° and the bond length of C-O elongating to 1.37 Å. Throughout this process, both Ti@C₄N₃ and CO₂-1 experience a reduction in their charges. Hence, the substantial activation of CO₂-2 can be attributed to the electron transfer from both CO₂-1 and Ti@C₄N₃ to CO₂-2, leading to strengthened interaction between the two CO₂ molecules and enabling CO₂ coupling. Furthermore, the attached simulation video (see Supplementary Video) visually illustrates the dynamic processes of CO₂ activation. This provides evidence that the Ti@C₄N₃ photocatalyst can potentially activate two CO₂ molecules simultaneously when exposed to light.

To examine thermal effects in the photocatalytic reaction, we conducted additional simulations in two conditions: (1) with light and at 0 K (light-0K); (2) without light at 300 K (no-light-300K), alongside the previously mentioned conditions under light and 300 K (light-300K). Compared to that under light-300K, the light-0K simulation features different results (see Fig. 3g). In 0~20 fs, C₄N₃ substrate has almost no electron transfer to Ti atom, Ti transfers about 0.02 e⁻ charge to CO₂-2 in 20~40 fs, CO₂-2 acquires the electron around 0.04 e⁻ charge from C₄N₃ substrate at 40~60 fs. At this point CO₂-2 gradually begins to activate and bends to approximately 160°. At 66~70 fs, C₄N₃ substrate begins to transfer some electrons to Ti atom (about 0.15 e⁻ charge). Subsequently, Ti provides charge to CO₂-2 around 0.13 e⁻ charge during 70~80 fs (see Supplementary Fig. 25). The bond length of CO₂-2 during this whole process fluctuated around 1.25 and 1.30 Å (see Supplementary Fig. 23b), respectively, and the maximum curvature was 158° at t = 45 fs, after which it fluctuated in the range of 160°. Thus, the slower electron-transfer rate at light-0K makes the activation of CO₂-2 worse than at light-300K. Under the no-light-300K, the electron densities of CO₂ and Ti@C₄N₃ do not change significantly, fluctuating roughly around 0 e⁻ charge because the transitions between vibrational levels are insufficient to induce electron transfers. Structurally, there were minimal changes in both the bond angle and bond length of CO₂-2. Consequently, activating two CO₂ molecules under the no-light-300K condition proved to be highly challenging. The combined results from light-300K, light-0K, and no-light-300K simulations reveal that concurrent activation of two CO₂ molecules occurs. Under light-300K, CO₂-2 undergoes significant charge transfer and structural changes compared to the no-light-300K condition, whereas it maintains an approximately linear structure under no light. Comparing light-300K and light-0K, both conditions can activate two CO₂ molecules simultaneously, but charge transfer is slow pronounced under light-0K, and the catalyst-CO₂-2 interaction is weaker. These findings suggest that a synergistic effect between photoexcitation and thermal effects plays a crucial role in the simultaneous activation of two CO₂ molecules. Light conditions are key to this dual activation, while thermal effects facilitate charge transfer via electron-phonon coupling, further enhancing CO₂-2 activation^15,44,45.

Mechanism of photocatalytic CO₂RR. To evaluate the catalytic mechanism of the CO₂RR on Ti@C₄N₃ catalysts, transition state calculations were performed utilizing DFT + U (U_eff = 3.79 eV) calculations. Both proton transfer and proton-coupled electron transfer (PCET) is utilized to describe the reduction mechanism of CO₂^24,46,47. Several possible reaction pathways for the CO₂ reduction, including single CO₂ reduction and dual CO₂ reduction, were identified on the Ti@C₄N₃ (see Fig. 4). In the following ^*, E_a, E_ads and E_des denote the adsorption site, barrier energy, adsorption energy and desorption energy respectively. Firstly, the reduction mechanism of single CO₂ on Ti@C₄N₃ photocatalyst was studied. CO₂ molecule is activated on the catalyst to form a CO₂^* anionic intermediate. Via the first PCET, the adsorbed CO₂^* binds with H proton with two different pathways of HCOO^* and COOH^*. Subsequently, the more stable COOH^* intermediate experiences the C-O bond rupture with the uphill trend of 0.23 eV, resulting in the formation of CO^* and OH^* species co-binding with Ti site. Then through an energy barrier of 0.40 eV, the OH^* generates H₂O^* by the PCET process and desorbed from the catalyst, while under the larger energetic penalty of 1.51 eV, the CO^* is desorbed from the catalyst, which is the RDS for CO generation (see Fig. 5). Thus, it is accessible for subsequent hydrogenation on CO^*. The CO^* is hydrogenated to produce two possible species: COH^* and CHO^* with the associated barrier of E_a = 1.73 eV (COH^*) and E_a = 0.11 eV (CHO^*) (see Supplementary Fig. 26), implying the optimal product of CHO^* rather than COH^*. During the fourth PCET process, the CHO^* will involve these two processes: CHO^* + H⁺ + e⁻ → HCHO^* and CHO^* + H⁺ + e^- → CHOH^* (see Supplementary Fig. 27). The hydrogenation of CHO^* to HCHO^* (E_a = 0.30 eV) is more favorable than that of CHOH^* (E_a = 1.21 eV) from the perspective of kinetics. The larger adsorption energy (E_ads = −2.26 eV) of HCHO^* on Ti@C₄N₃ leads to the desorption difficulty, while via the barrier of 0.70 eV the hydrogenation of HCHO^* intermediate generates the CH₂OH^*, which further form CH₃OH^* product via the 6^th PCET step with the barrier of 0.65 eV. Under the uphill barrier of 0.40 eV, the adsorbed CH₃OH^* enables hydrogenated (7^th PCET) and dehydrated to CH₃^*, which is easily hydrogenated to CH₄ under the 8^th PCET. The CH₄ is more readily desorbed (E_des = 0.56 eV) from the Ti@C₄N₃ catalyst surface than other C₁ products (such as CO, HCHO, CH₃OH), and HCHO^* + H⁺ + e⁻ → CH₂OH^* is a RDS for CH₄ generation. In general, single adsorbed CO₂ molecule undergoes a series of hydrogenation steps to form C₂ intermediate species through four common coupling pathways: CO^*-CO^*, CO^*-COH^* or CHO^*, CO^*-CH_n^* (n = 1, 2, 3), and CH_n^*-CH_n^*11,12,16. Among the four pathways, a common feature is the formation of CO^* intermediates before C-C coupling. However, for SACs, the nearly identical charge distributions between neighbouring C₁ intermediates inevitably result in a strong dipole−dipole repulsion and thereby hinders C − C coupling. Apart from that as the discrete active sites and strong adsorption for crucial intermediates (−1.51 eV for CO^*, −2.74 eV for COH^*, −2.98 eV for CHO^*, −4.38 eV for CH^*, −3.70 eV for CH₂^* and −3.08 eV for CH₃^*), desorption and diffusion of these intermediates are too difficult to achieve C-C coupling through Langmuir–Hinshelwood (L-H) or Eley–Rideal (E-R) mechanisms. In summary, the generation of C₂ products via the four coupling pathways described earlier is particularly challenging when using Ti@C₄N₃ catalysts. Thus, the reduction of individual CO₂ molecule is more inclined to generate C₁ species via the eight PCET steps, where the barriers of RDS are 1.51 eV for CO, 2.26 eV for HCHO, 1.59 eV for CH₃OH and 0.70 eV for CH₄, respectively, Thereinto, CH₄ features the highest kinetic selectivity.

Fig. 4: Reaction pathways.

Schematic illustration of the possible reaction pathways for CO₂RR on Ti@C₄N₃.

Fig. 5: CO₂RR energy barrier and reaction intermediates.

Reaction mechanisms of C₁ and C₂ products. b stands for catalyst. Color codes: C, brown; N, silver; Ti, light blue; O, red; H, white.

Just as aforementioned, given for high GHSV of CO₂ gas, the Ti@C₄N₃ can activate two CO₂ molecules through both thermal and photonic excitation. Thus, it is whether two activated CO₂^* directly undergoes C-C coupling reaction under light irradiation. However, it is very difficult to use the ab-initio NAMD simulation to identify the whole real-time coupling process under light. Thus, we hypothesize that photon is only beneficial for the activation rather than the diffusion of intermediate. Based on CI-NEB strategy (see Supplementary Fig. 29), we found that direct coupling of two CO₂^* proceeds with the tiny barrier of 0.19 eV, which is thermodynamically easier than the coupling via previous four pathways (see Fig. 6). Such result is different from the consensus that CO^* is necessary for the formation of multicarbons (C₂₊) products^11,48,49. Supplementary Fig. 30 indicates that the kinetically CO₂^* to CO₂^* coupling (E_a = 0.19 eV) capacity is higher than that of CO₂^* to H^* proton (E_a = 0.23 eV). Furthermore, crystal orbital hamilton population (COHP) results in Supplementary Fig. 31 showed that negative ICOHP of C-C bond (−3.99) and Ti-O bonds (−2.82 and −2.79) sheds light on the stability of OOCCOO^* on the surface of Ti@C₄N₃ catalyst and the existence of strong C-C bond^50,51. Thus, the adsorbed OOCCOO^* intermediate is able to participate in the subsequent reduction to generate multiple C₂ products (e.g. (COOH)₂, CH ≡ COH, C₂H₂, C₂H₄ and C₂H₆). Via the first PCET step (OOCCOO^* + H⁺ + e⁻ → OOCCOOH^*), the barrier of 0.64 eV is required for the formation of OOCCOOH^*, which binds with one proton to form cis HOOCCOOH^* through the next elementary step (see Fig. 5) with the barrier of E_a = 0.86 eV.

Fig. 6: Schematic diagram of two CO₂ molecules coupled.

Comparison of two CO₂ coupling processes in light and no-light.

Then, via the intramolecular proton transfer, the cis HOOCCOOH^* can undergo a structural transformation resulting in the production of trans HOOCCOOH^* with an associated energetic penalty of 0.94 eV. The obtained trans HOOCCOOH^* prefers to continue hydrogenation to produce HOHOCCOOH^* (E_a = 1.01 eV) rather than the desorption with energy of 2.56 eV. Next, HOHOCCOOH^* underwent proton transfer to produce HOHOCCOHO^* with the barrier of 0.77 eV. The HOHOCCOHO^* generates species HOHOCCOHOH^* is demands an energy barrier of 1.09 eV, which is further reduced to HOCCOHOH^* by C-O bond fracture dehydration (E_a = 0.27 eV). Due to the structural instability of HOCCOHOH^*, the proton transfer from the hydroxyl group to the C atom with the formation of carbonyl group (E_a = 0.69 eV), resulting in OHCCOHOH^* intermediate. Subsequently, OHCCOHOH^* undergoes PCET dehydration with 0.56 eV barrier to yield OHCCOH^*, which must overcome a 0.60 eV barrier to undergo hydrogenation and form an O-H bond, resulting in the generation of HOHCCOH^*. Rapid hydrogenation of HOHCCOH^* leads to C-O bond breakage to emerge CHCOH^* and H₂O (E_a = 0.09 eV). On Ti@C₄N₃ catalyst, the stronger adsorption of CHCOH^* (E_ads = −1.71 eV) benefits further hydrogenation and dehydration to obtain CHC^* through the barrier of 0.53 eV. The successive hydrogenation of CHC^* → CHCH^* → CH₂CH^* → CH₂CH₂^* → CH₃CH₂^* → CH₃CH₃^*occurs under the barriers of 0.24, 0.06, 0.07, 0.19 and 0.03 eV, respectively. Despite the formation of C₂H₂ and C₂H₄ products, the subsequent hydrogenation process leads to the final production of C₂H₆ due to their adsorption energy (E_ads = 2.02 eV for C₂H₂, 1.75 eV for C₂H₄) compared to the hydrogenation barrier. Generally, during the coupled hydrogenation of two CO₂ on Ti@C₄N₃, various products are formed with different energy barriers of 2.56 eV for (COOH)₂, 1.71 eV for CH ≡ COH, 2.02 eV for C₂H₂, 1.75 eV for C₂H₄ and 1.09 eV for C₂H₆, respectively. Therefore, C₂H₆ is determined to be the optimal product via fourteen-electron reduction, and the generation of the HOHOCCOHOH^* intermediate is the RDS for the entire reaction in both kinetics (E_a = 1.09 eV) and thermodynamics (ΔG_max = 1.57 eV).

Discussion

In summary, by utilizing DFT calculation and ab-initio NAMD simulation, we build the computational frameworks to screen an outstanding single-atom photocatalysts Ti-supported 2D C₄N₃ material, Ti@C₄N₃. Structurally, the Ti@C₄N₃ catalyst shown the excellent stabilities both thermally, chemically and mechanically. Electronically, such catalyst has great potential as a photocatalyst for CO₂ reduction: two main absorption peaks (327.77 and 529.61 nm), suitable band positions (0.002 eV for VBM and −0. 968 eV for CBM) and long photocarrier lifetime (38.21 ps for electron), ensuring that enough photogenerated electrons migrate to the surface and participate in photocatalytic CO₂RR. Essentially, such properties are intimately tied with the doping Ti atom, embodying the fundamental transformations from conductor to semiconductor. Interestingly, under the high GHSV of CO₂, such high-valence of doping Ti ion renders the dual activation of CO₂: Without light, high Lewis-acidity of Ti site thermally induces activation of CO₂ by back-donating π-bond; and under visible light irradiation another weak-adsorbed CO₂ attains the photoelectron from Ti site through CBM and thereby undergoes photo-induced activation. Catalytic mechanism studies systematically reveal that with the barrier of 0.19 eV the two activated CO₂ is easily coupled to be oxalate, which is further reduced to be C₂H₆ (E_a = 1.09 eV). Our finding reveals the possibility of dual activations (thermally induced activation and photo-induced activation) during the photocatalytic CO₂ reductions.

Methods

Ground-State Calculations. DFT calculations (geometrical optimization, electronic properties and catalytic mechanism) were carried out by VASP package⁵² (6.3.1) using Perdew−Burke−Ernzerhof⁵³ (PBE) functional and projector-augmented wave⁵⁴ (PAW) pseudopotential. Detailed computational information was presented in Supplementary Data file.

Excited-state Dynamic calculations. The photocarrier dynamics was simulated by using the ab initio NAMD program (Hefei-NAMD)⁵⁵. Dual activation of CO₂ under light irradiation was identified by rt-TDDFT MD simulation utilizing the TDAP program⁵⁶. Detailed computational information was presented in Supplementary Data file.