In situ formation of cocatalytic sites boosts single-atom catalysts for nitrogen oxide reduction

Significance

The use of oxide-supported isolated Pt single atoms as catalytic active sites is of interest due to their maximum atom utilization efficiency and unique catalytic properties for the removal of NO_x. However, relationships between the structure of active center, the dynamic response to environments, and catalytic functionality have proved difficult to experimentally establish. This research reveals that the in situ formation of adjacent cocatalytic sites boosts single-atom catalysts for the conversion of NO_x during the catalytic reaction of NO_x reduction with H₂. The presence of cocatalytic sites tunes and optimizes the catalytic performance at the atomic scale that can benefit the environment and human health by removing NO_x efficiently.

Abstract

Nitrogen oxide (NO_x) pollution presents a severe threat to the environment and human health. Catalytic reduction of NO_x with H₂ using single-atom catalysts poses considerable potential in the remediation of air pollution; however, the unfavorable process of H₂ dissociation limits its practical application. Herein, we report that the in situ formation of Pt_Ti cocatalytic sites (which are stabilized by Pt–Ti bonds) over Pt₁/TiO₂ significantly increases NO_x conversion by reducing the energy barrier of H₂ activation. We demonstrate that two H atoms of H₂ molecule are absorbed by adjacent Pt atoms in Pt–O and Pt–Ti, respectively, which can promote the cleave of H–H bonds. Besides, Pt_Ti sites facilitate the adsorption of NO molecules and further lower the activation barrier of the whole de-NO_x reaction. Extending the concept to Pt₁/Nb₂O₅ and Pd₁/TiO₂ systems also sees enhanced catalytic activities, demonstrating that engineering the cocatalytic sites can be a general strategy for the design of high-efficiency catalysts that can benefit environmental sustainability.

Nitrogen oxide (NO_x) from stationary sources and vehicle exhaust is the primary source of acid rain, greenhouse effect, photochemical smog, and fine particulate matter (1–3), which has a substantial impact on the global environment and human health. A promising modern technology for the removal of NO_x is selective catalytic reduction with typical reductants (H₂, NH₃, and HC), with the use of H₂ as the reducing agent, in particular, being able to attract considerable attention due to its many distinct advantages, including sustainability, cleanliness, and superior lower temperature activity (4). It is assumed in previous studies of NO_x reduction by H₂ that the reaction mechanism begins with the activation of H₂, following that the generated active hydrogen species cleave the N–O bond in NO_x (5, 6). Therefore, the efficient activation of H₂ is the critical step in the removal of NO_x (7); however, it is still a challenge.

Recently, Pt-based catalysts, especially Pt single-atom catalysts (Pt₁/Al₂O₃, Pt₁/WO₃, and Pt₁/NC, etc.) with the maximum atom utilization efficiency and unique catalytic properties (8–11), show remarkable performance due to the superior activity of H₂ activation in catalytic hydrogenation reaction and reverse water–gas shift reaction. Despite the increasing consensus that isolated Pt sites are the dominant active sites (12), the nature of single Pt active sites is still under debate, and it still needs to optimize their ability of H₂ activation for real applications. Generally, Pt single atoms possess various chemical coordination environments (Pt–O_x) on oxide supports; in this case, some unstable Pt–O bonds will be interrupted in the high-temperature reducing atmosphere, accompanied by the in situ formation with the new coordination (monatomic metal sites and metal nanoparticle sites, etc.) (11, 13). Moreover, it has been reported that the catalytic performance will be promoted by constructing the cocatalytic sites next to single-atom sites in oxygen reduction reaction (14, 15). The reason is that the adjacent cocatalytic sites can regulate the physical/chemical property of isolated active sites to boost the activation of gas molecule (14–16). For example, adjacent Fe nanoparticles of cocatalytic sites can significantly optimize the electronic structure of isolated Fe–N–C sites, thus assisting the O₂ activation on the Fe–N–C sites and improving the performance of single-atom catalysts (17). These findings suggest that the in situ formation of Pt species with new coordination evolved from unstable Pt–O coordination will facilitate the activation of H₂ molecules in the reduction of NO with H₂; however, it is rarely discussed.

In this work, using the Pt₁/TiO₂ catalyzed reduction of NO with H₂ as an example, we propose that adjacent Pt_Ti cocatalytic sites (which are stabilized by Pt–Ti bonds) can significantly boost single-atom catalysts for H₂ activation, as well as reduce the energy barrier for N–O activation to accelerate the catalytic reaction of NO reduction. The formation of Pt_Ti sites in Pt₁/TiO₂ is directly identified by in situ high-angle annular dark-field-scanning transmission electron microscopy (HAADF-STEM) and X-ray absorption fine structure (XAFS) under the real reaction condition. Meanwhile, theoretical studies reveal the surface reaction mechanism of NO reduction boosted by Pt_Ti sites, which promote the efficient activation of NO by active H species. Furthermore, the superior de-NO_x performance (ratios differ by 20%) through the formation of cocatalytic sites is also confirmed in Pt₁/Nb₂O₅ and Pd₁/TiO₂ catalysts, which provides the possibility to promote environmental sustainability and protect human health.

Results

Characterization of Pt₁/TiO₂ Single-Atom Catalyst.

Pt single atom (0.22 wt%) supported on TiO₂ nanosheets (denoted as fresh Pt₁/TiO₂) was prepared by immersion method as the model catalyst for the reduction of NO by H₂ (SI Appendix, Table S1) (18, 19). The fresh Pt₁/TiO₂ was first characterized by X-ray diffraction and high-resolution transmission electron microscopy (SI Appendix, Figs. S1–S3). The results showed that the characteristic peaks mainly indexed to anatase TiO₂ phase, and no discernible peak of platinum was detected. The HAADF-STEM clearly showed the dispersion of individual Pt atoms (marked by the yellow circles) on TiO₂, and energy-dispersive X-ray spectroscopy (EDS) mapping images revealed that the Pt species were uniformly dispersed on the surface of TiO₂ nanosheet (Fig. 1A and SI Appendix, Fig. S4). Notably, there were some adjacent Pt atoms in fresh Pt₁/TiO₂ (marked by the orange circles), which might induce the dynamic migration of Pt atoms (11, 13). As expected, the position of Pt atoms was found to change significantly after reaction (denoted as used Pt₁/TiO₂), but Pt atoms (marked by the orange circles) were confirmed by HAADF-STEM to distribute in the form of single atoms (Fig. 1 E and I–L and SI Appendix, Fig. S5). All Pt atoms in fresh Pt₁/TiO₂ were anchored in bright lattice fringe (Ti atomic columns) of TiO₂ (101), while those in used Pt₁/TiO₂ were trapped in both bright lattice fringes and dark lattice fringes (O atomic columns). Additionally, line-scan intensity analysis (Fig. 1 M and N) also demonstrated the position evolution of Pt atoms, i.e., the Pt atoms originally anchored only on Ti atomic column and evolved into atomic columns of both O and Ti after reaction (taken along the lines 1 and 2 in Fig. 1 A and E, respectively).

Fig. 1.

HAADF-STEM characterization of Pt₁/TiO₂. (A and E) HAADF-STEM images of fresh Pt₁/TiO₂ (A) and used Pt₁/TiO₂ (E). (B–D) In situ HAADF-STEM images and relevant intensity profiles of Pt₁/TiO₂ without reaction in Ar atmosphere at 125 °C (1 bar). (F–H) In situ HAADF-STEM images and relevant intensity profiles of Pt₁/TiO₂ under the reaction at 125 °C (NO:H2 ~ 1:1, 1 bar). (I–L) EDS mapping of used Pt₁/TiO₂. (M and N) The intensity profiles obtained in A and E, respectively.

To further unravel the dynamic structural evolution of Pt single atom under real reaction condition, in situ HAADF-STEM was used to monitor the reaction process, in which small probe currents and minimal acquisition times were used in observation to minimize the electron beam effects (20). For fresh Pt₁/TiO₂, the positions of Pt_a and Pt_b atoms (marked by white circle) remained unchanged in 600 s (125 °C, under Ar atmosphere), which was also confirmed by the results of line-scan intensity analysis that the distance of Pt_b atom and adjacent Ti atomic column was changeless (~3.5 Å, displayed in Fig. 1 B–D, Inset), free from the interference of temperature. Then, the gas mixture of NO and H₂ (volume ratio NO:H₂ ~ 1:1, 1 bar) was introduced, and Pt_A and Pt_B atoms were found to be trapped in bright lattice fringe of TiO₂ (Fig. 1F), corresponding to the initial state of Pt single atoms in fresh catalyst. After 60 s of exposure to the gas mixture, we found that the Pt_B atom moved close to Pt_A atom and was anchored in dark lattice fringe (Fig. 1G), accompanied by the distance of Pt_B atom and the adjacent Ti atom column decreased from 3.6 Å to 2.1 Å (Fig. 1 F and G, Inset). With increasing exposure time, the Pt_B atom was finally anchored at the position of O atom column after 600 s gas exposure as shown in Fig. 1H, during which the position of Pt_A atom remained stable. The observation was consistent with the result of ex situ HAADF-STEM (Fig. 1E), which visually confirmed the phenomenon of dynamic structural evolution.

Chemical Nature of Evolution over Pt Single Atoms.

To identify the chemical nature of the evolution over Pt single atoms, we employed XAFS, density functional theory (DFT) calculations, in situ Raman spectroscopy, in situ CO adsorption diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), and quasi in situ X-ray photoelectron spectroscopy (XPS) measurements. According to the results of X-ray absorption near-edge structure (XANES) curves of Pt L₃-edge (Fig. 2A), the white-line intensities of both fresh Pt₁/TiO₂ and used Pt₁/TiO₂ were between Pt foil and PtO₂, indicating that the valences of Pt atoms were 2.6 (fresh Pt₁/TiO₂) and 1.3 (used Pt₁/TiO₂) (Fig. 2 A, Inset), which proved that Pt single atoms were in the lower valence states after reaction. Moreover, for fresh Pt₁/TiO₂, the extended X-ray absorption fine structure (EXAFS) displayed a major peak around 1.5 Å, attributed to Pt–O coordination (Fig. 2B), suggesting that Pt in fresh Pt₁/TiO₂ catalyst was largely present as isolated single atoms on TiO₂. However, besides the dominant peak of Pt–O around 1.5 Å, a new peak appeared at 2.2 Å on used Pt₁/TiO₂ catalyst (Fig. 2B). Notably, the wavelet transform (WT) EXAFS (SI Appendix, Fig. S6) showed that the intensity between nanophase resolved 2 and 3 Å was localized in a different k range from Pt–Pt coordination (5 to 7 Å⁻¹ compared to 8 to 12 Å⁻¹), implying that the EXAFS peak around 2.2 Å was attributed to Pt–Ti coordination (11, 21). Combined with the observation of HAADF-STEM, it also confirmed that more than one Pt species were present in used Pt₁/TiO₂, which further revealed the structural evolution of Pt atoms in fresh Pt₁/TiO₂ during the reaction.

Fig. 2.

Chemical nature of evolution over Pt single atoms. (A) Pt L₃-edge XANES spectra of Pt foil, PtO₂, fresh Pt₁/TiO₂, and used Pt₁/TiO₂. (B) FT-EXAFS spectra of Pt foil, PtO₂, fresh Pt₁/TiO₂, and used Pt₁/TiO₂ at the Pt K-edge. (C) Top views of TiO₂ (101) surface model adopted in this work. And possible Pt–O coordination was labeled. (D and E) The optimized structure models of fresh Pt₁/TiO₂ (D) and used Pt₁/TiO₂ (E). (F) The formation mechanism of Pt–Ti bonds calculated by DFT. (G) EXAFS R space-fitting curve and the original curve over fresh Pt₁/TiO₂. (H) EXAFS R space-fitting curve and the original curve over used Pt₁/TiO₂. (I) In situ Raman spectra of Pt₁/TiO₂ under the reaction at 125 °C (volume ratio NO:H₂ ~ 1:1).

To further validate the existence form of Pt single atoms, we first screened various Pt–O structures stabilized on the (101) crystal plane of TiO₂ over fresh Pt₁/TiO₂. Notably, TiO₂ (101) surface only had five-coordinated Ti_5c and two-coordinated O_2c on the outermost surface (Fig. 2C), thus the possible O-coordinated active sites of Pt_5c–O and Pt_2c–O were finally constructed (22, 23), i.e., 1) Pt single atom substituted one Ti_5c on the surface and bonded with five O atoms and 2) Pt single atom adsorbed on the surface of TiO₂ and bonded with two two-coordinated surface O_2c, respectively. Based on EXAFS analysis, it implied that Pt_2c–O (coordination number = 2) and Pt_5c–O (coordination number = 5) structures were coexisted on fresh Pt₁/TiO₂. Pt species were easily reduced in hydrogen-involved reactions (13), and the symmetry of the highest occupied molecular orbital in Pt_5c sites matched antibonding of H₂ well based on DFT calculations (SI Appendix, Figs. S7–S9). Notably, the electrons from 5d orbitals should be donated to the antibonding orbital of H₂ molecules in the process of H₂ activation, which indicated that symmetric matching of orbitals was necessary in the coupling process (SI Appendix, Fig. S7). Therefore, the symmetric matching of orbitals in the coupling process suggested that Pt_5c sites were the active centers for H₂ dissociation (24), which could accelerate the migration of O atoms and provide a reasonable site for the migration of lower coordinated Pt species (Pt_2c) to form Pt–Ti bonds with subsurface Ti atoms. Therefore, the model with the adjacent coexistence structure of Pt_5c and Pt_2c sites was established to analyze the interaction between Pt_5c and Pt_2c sites (Fig. 2D).

To determine the formation mechanism of Pt–Ti bonds, the process of structural evolution was conducted by DFT calculations (Fig. 2F). First, the H₂ molecule was adsorbed on Pt_5c sites (−2.64 kcal/mol) and would dissociate into two adsorbed H* with an activation barrier of 15.41 kcal/mol (TS1). Then, one adsorbed H* would be transferred to the Pt_5c–O₁ to form H*–Pt_5c–O₁–H*, with a Pt_2c–O₁ bond distance of 2.16 Å (the original bond distance of Pt_2c site with two bridge O atoms was 2.03 Å) (stage III). Next, the other H* was transferred to O₁ atom to form H*–O₁–H* with an activation barrier of 31.94 kcal/mol (TS2). In this case, the distance of Pt_2c–O₁ bonds was stretched to 2.34 Å (stage IV). As a result, the Pt_2c atom could move and occupy O₁ position by removing H₂O molecule (H*–O₁–H*), leading to that the Pt_2c atom only bonded with Ti atoms instead of O atoms (donated as Pt_Ti), and the optimization of used Pt₁/TiO₂ structural model is shown in Fig. 2E. The above calculations well explained the cleavage of Pt_2c–O bonds and the formation of Pt–Ti bonds, leading to that the Pt_2c site in fresh Pt₁/TiO₂ was evolved to the Pt_Ti site in used Pt₁/TiO₂. More importantly, the formation of Pt–Ti bonds played an important role in suppressing the aggregation of Pt single atoms; the corresponding calculated structural parameters of transition states are shown in Fig. 2F and SI Appendix, Figs. S10–S12. The ab initio molecular dynamics simulation was further performed to prove that the structural model could serve as single-atom catalysts with high stability (SI Appendix, Figs. S13 and S14). Moreover, the Hirshfeld charges of Pt atoms were also calculated (SI Appendix, Figs. S15 and S16) (25). Because the Pt species were reduced, the charge of Pt_2c and Pt_5c sites, respectively, decreased from 0.204 and 0.611 eV to 0.117 and 0.508 eV, which was consistent with the XANES results and confirmed the formation of low-valence Pt single atoms.

Furthermore, quantitative least-squares EXAFS best-fitting analysis (Fig. 2 G and H and SI Appendix, Figs. S17 and S18 and Tables S2 and S3) was performed to verify the rationality of DFT calculations (26). We should always be aware that the structure information obtained by XAFS was an average result of the detected Pt element due to the coexistence of Pt_5c and Pt_2c (or Pt_Ti) sites in our work (27), that was, Pt atoms only bonded with oxygen atoms with an average coordination number of ~4.7 over fresh Pt₁/TiO₂. Obviously for used Pt₁/TiO₂, the coordination number of Pt–O path decreased from 4.7 to 4.0 after reaction, which was caused by the evolution from Pt_2c–O to Pt–Ti coordination. These results revealed that Pt_5c atoms coordinated strongly with oxygen through occupying the position of Ti atoms, while Pt_2c atoms initially coordinated with the surface oxygen of TiO₂ and evolved to Pt–Ti coordination (Pt_Ti sites) during the reaction.

Then, in situ Raman spectroscopy was used to further reveal the formation of Pt_Ti sites during the reaction on Pt₁/TiO₂. As shown in Fig. 2I and SI Appendix, Fig. S19, we found that the vibrational characteristic peak (141 cm⁻¹) of O–Ti–O species shifted to a higher wavenumber (149 cm⁻¹) after introducing reactant gas (volume ratio NO:H₂ ~ 1:1) (28). Besides, the quantum chemical calculation of vibrational Raman spectra was performed on fresh Pt₁/TiO₂ and used Pt₁/TiO₂; the results showed that the blue shift from 141 to 149 cm⁻¹ (Fig. 2I) could be attributed to the formation of O–Ti–Pt structure, which disrupted the symmetry of the O–Ti–O network (SI Appendix, Figs. S20 and S21) (29). Therefore, both experimental and DFT results provided strong evidence for the formation of Pt_Ti sites.

Besides, CO was used as the probe molecule in DRIFTS to characterize the Pt-based catalysts, which presented three bands centered at 2,090, 2,062, and 2,055 cm⁻¹ (Fig. 3A). Among them, the band at 2,090 cm⁻¹ was assigned to CO adsorption on Pt single atoms, and that at 2,055 cm⁻¹ was assigned to the linear CO adsorption on Pt nanoparticles (30). Besides, the CO adsorption band was preserved on Pt_Ti sites at 2,062 cm⁻¹ over used Pt₁/TiO₂ (31), further showing the formation of Pt–Ti bonds after reaction. Meanwhile, the evolution of chemical environment of Pt atoms and the surface of TiO₂ during the reaction was measured with quasi in situ XPS. After we introduced reactant gases (volume ratio NO:H₂ ~ 1:1), the binding energy of Pt 4f on Pt₁/TiO₂ decreased rapidly (76.0 and 72.8 eV vs. 75.4 and 71.9 eV) (Fig. 3B), while there was no shift of the characteristic peak of Ti 2p (Fig. 3C). The O 1s XPS spectra were divided into two peaks centered at 531.4 and 529.7 eV, which were assigned to oxygen vacancies (denoted O_ads) and lattice oxygen (denoted O_latt), respectively. Notably, the ratio of lattice oxygen and oxygen vacancy did not change in the hydrogen-involved reaction, as well as the results of in situ electron paramagnetic resonance (Fig. 3D and SI Appendix, Figs. S22–S25 and Table S4). These quasi in situ studies of XPS clearly showed that the singly dispersed Pt species were reduced in the mild reducing atmosphere, which could be related to the structural evolution of Pt single atoms based on DFT and EXAFS results. Additional study using hydrogen temperature-programmed reduction had revealed an outstanding low-temperature catalytic activity of Pt-based catalysts in the reduction of NO with H₂ (SI Appendix, Fig. S26). In summary, all the DFT calculations and characterization data discussed above systematically confirmed the structural evolution of Pt single atoms during the reaction, leading to the formation of Pt_Ti sites near Pt_5c single atom sites.

Fig. 3.

Identification of Pt single atoms and quasi in situ studies of surface chemistry. (A) In situ CO adsorption DRIFTS of TiO₂, fresh Pt₁/TiO₂, used Pt₁/TiO₂, and Pt_NP/TiO₂. (B–D) Quasi in situ XPS spectra of Pt 4f (B), Ti 2p (C), and O 1s (D) of Pt₁/TiO₂ under different conditions.

Correlation of Cocatalytic Site and Catalytic Activity.

The catalytic activities of NO reduction with H₂ over fresh Pt₁/TiO₂, used Pt₁/TiO₂, and Pt_NP/TiO₂ were evaluated on a fix-bed reactor. More importantly, cyclic experiment was used to explore the correlation of cocatalytic site and catalytic activity. As shown in Fig. 4A and SI Appendix, Fig. S27, used Pt₁/TiO₂ exhibited better catalytic activity (ratios differ by 20%) and N₂ selectivity than those of fresh Pt₁/TiO₂ in the temperature range of 75 to 150 °C due to the formation of Pt_Ti sites, and the performance of used Pt₁/TiO₂ tended to stabilize in the third cycle. Then, in situ DRIFTS studies were further performed on fresh Pt₁/TiO₂ and used Pt₁/TiO₂ to reveal the effect of Pt_Ti sites on the reaction (SI Appendix, Fig. S28), and the intensity of NO adsorption on used Pt₁/TiO₂ was clearly observed to be stronger than that on fresh Pt₁/TiO₂. Notably, the active intermediates of adsorbed NO species would be consumed to generate –NH₂ amide species more quickly on used Pt₁/TiO₂, and the latter could further react with NO molecule to form N₂ in the reduction of NO with H₂. The above results confirmed the critical role of cocatalytic sites in promoting the process of NO conversion. Meanwhile, used Pt₁/TiO₂ showed a superior catalytic performance and Pt species were still isolated and uniformly dispersed in the presence of oxygen (SI Appendix, Figs. S29 and S30). Moreover, the apparent activation energy of reaction catalyzed by used Pt₁/TiO₂ (E_a = 50.6 kJ/mol) was also smaller than that of fresh Pt₁/TiO₂ (E_a = 63.6 kJ/mol) (Fig. 4B). Besides, for comparison, Pt nanoparticles (0.23 wt%) supported on TiO₂ nanosheets (denoted as Pt_NP/TiO₂) were also synthesized, which showed inferior catalytic activity and turnover frequency compared with single-atom catalysts (SI Appendix, Figs. S31–S34 and Table S5) (32–34).

Fig. 4.

Catalytic activity of catalysts and DFT calculations. (A) The cyclic experiment over Pt₁/TiO₂ catalysts. The reaction condition is: [NO] = 1,000 ppm, [H₂] = 1,000 ppm, balanced in Ar. GHSV = 60,000 mL g⁻¹ h⁻¹. (B) Arrhenius plot of NO + H₂ reaction on fresh Pt₁/TiO₂ and used Pt₁/TiO₂. (C) DFT calculations of relative energy on Pt_5c–O over fresh Pt₁/TiO₂ for H₂ activation. (D) DFT calculations of relative energy on Pt_5c–O and Pt_5c–Pt_Ti over used Pt₁/TiO₂ for H₂ activation. (E) DFT-calculated potential energy diagrams and the barrier energy for the reduction of NO with H₂ over used Pt₁/TiO₂.

Insights into the NO Reduction Boosted by Cocatalytic Sites.

The DFT calculations were conducted to further evaluate the role of Pt_Ti cocatalytic sites for boosting catalytic performance. Considering that the adsorption and dissociation of reactant gas molecule was the initial and critical step of the reaction (35), the adsorption and activation of H₂ was first examined over fresh Pt₁/TiO₂ and used Pt₁/TiO₂ (Fig. 4 C and D). As shown in Fig. 4C, the adsorption energy and dissociation barrier of H₂ on the Pt_5c site of fresh Pt₁/TiO₂ were −2.64 kcal/mol and 15.41 kcal/mol, respectively. After the formation of Pt_Ti sites, the adsorption of H₂ (−6.77 kcal/mol) was significantly enhanced on Pt_5c site (Fig. 4D). Notably, there were two pathways (Pt_5c–O path and Pt_5c–Pt_Ti path) for H₂ dissociation over used Pt₁/TiO₂, in which energy barriers were only 14.23 kcal/mol and 3.20 kcal/mol, respectively. The results suggested that the adjacent Pt_Ti sites assisted the H₂ activation on the Pt_5c single-atom site (Fig. 4 C and D and SI Appendix, Fig. S35). Meanwhile, it was also found that the barrier of H₂ dissociation along Pt_5c–O path got decreased over used Pt₁/TiO₂ (14.23 kcal/mol) compared to fresh Pt₁/TiO₂ (15.41 kcal/mol), which was caused by the formation of low-valence Pt single atoms.

Besides, the comprehensive DFT calculations were carried out to reveal the surface reaction mechanism of the reduction of NO (H₂ + NO → H₂O + N₂) and the promotion mechanism of Pt_Ti sites over fresh Pt₁/TiO₂ (SI Appendix, Fig. S37), used Pt₁/TiO₂ (Fig. 4E), and Pt (111) surface (SI Appendix, Fig. S38), which was initiated by the adsorption of H₂ and NO and completed by the desorption of H₂O and N₂. The details on the reaction process of each step and the corresponding reaction path expression and reaction energy barrier are presented and discussed in Fig. 4E and SI Appendix, Figs. S36–S39 and Tables S6–S8. For used Pt₁/TiO₂, the catalytic cycle was initiated by the adsorption of NO on the Pt_Ti site and the adsorption of H₂ on the Pt_5c site. Then, one adsorbed H* could be transferred to the adsorbed NO* to form HNO (NO* + H* → HNO) with an activation barrier of 2.08 kcal/mol (TS1), and the other H* would be transferred to the Pt_Ti site to form HNOH with the energy barrier of 5.55 kcal/mol (TS2). The HHNO was generated from the HNOH (TS3, ~31.65 kcal/mol). In the next step, the second H₂ molecule was introduced and activated, forming HNOH with the energy barrier of 31.18 kcal/mol. Thereafter, the third active H* would transfer to the N atom to form NH–HOH (TS5, ~32.57 kcal/mol), which promoted the removal of H₂O and the formed NH₂ structure. Subsequently, the second NO was introduced and adsorbed on the N site, leading to the formation of N–N (NNOHH → NN–HOH → N–N). Finally, the produced H₂O and N₂ were desorbed from the surface and the catalytic cycle was completed. By comparing the energy barriers of each transition state, it could be found that the energy barrier of TS5 (32.57 kcal/mol) was the highest, indicating that this step was the rate-determining step of the whole reaction. Notably, the DFT results suggested that the energy barrier of N–O activation on used Pt₁/TiO₂ (32.57 kcal/mol) was lower than that on fresh Pt₁/TiO₂ (36.50 kcal/mol) and Pt (111) nanoparticle surface (38.58 kcal/mol), revealing that the formation of Pt_Ti sites could significantly boost single-atom catalysts for the reduction of NO.

Discussion

To further expand and verify the universal significance of cocatalytic sites, the catalytic activity was also confirmed to be promoted through the formation of cocatalytic sites over a series of single-atom catalysts (Pt₁/Nb₂O₅, Pd₁/TiO₂, and Rh₁/TiO₂). HAADF-STEM and in situ CO adsorption DRIFTS revealed that the single atoms (Pt, Pd, and Rh) remained stable before and after reaction (SI Appendix, Figs. S40–S42). Notably, we found that cocatalytic sites were only formed in Pt₁/Nb₂O₅ and Pd₁/TiO₂ after reaction (SI Appendix, Fig. S42), accompanied by the formation of low-valence metal active center (Fig. 5 A and B and SI Appendix, Figs. S43 and S44). Conversely, the cocatalytic sites were not identified in used Rh₁/TiO₂ (Fig. 5C and SI Appendix, Figs. S42 and S45). Correspondingly, the formation of cocatalytic sites boosted single-atom catalysts for NO_x reduction due to the superior catalytic performance in cyclic experiment (Fig. 5 D–F).

Fig. 5.

Correlation of metal–metal bonds and catalytic activity. (A) Pt 4f XPS spectra of fresh Pt₁/Nb₂O₅ and used Pt₁/Nb₂O₅. (B) Pd 3d XPS spectra of fresh Pd₁/TiO₂ and used Pd₁/TiO₂. (C) Rh 3d XPS spectra of fresh Rh₁/TiO₂ and used Rh₁/TiO₂. (D–F) The cyclic experiment over Pt₁/Nb₂O₅ (D), Pd₁/TiO₂ (E), and Rh₁/TiO₂ (F) catalysts. The reaction condition of D–F is: [NO] = 1,000 ppm, [H₂] = 1,000 ppm, balanced in Ar. GHSV = 60,000 mL g⁻¹ h⁻¹.

Advances in highly active single-atom catalysts for addressing global atmosphere challenge may be realized by designing cocatalytic sites for efficient NO removal with H₂. Taking Pt₁/TiO₂ as an example, in situ HAADF-STEM was used to monitor the reaction process and unravel the dynamic evolution of Pt single atom under real reaction condition. Combined with the EXAFS results, DFT calculations were performed to confirm the existence form of Pt single atoms and clarify the mechanism of structural evolution. The results revealed that Pt_5c atoms coordinated strongly with oxygen through occupying the position of Ti atoms, while unstable Pt_2c atoms initially coordinated with the surface oxygen of TiO₂ and evolved to Pt–Ti coordination with subsurface Ti atoms during the reaction. Besides, DFT results definitely confirmed that the adjacent Pt_Ti cocatalytic sites could significantly enhance H₂ activation on Pt_5c single-atom sites, resulting that the barrier of H₂ dissociation along Pt_5c–Pt_Ti path got decreased to 3.20 kcal/mol. Compared with fresh Pt₁/TiO₂ (36.50 kcal/mol), the presence of Pt_Ti sites also accelerated the whole de-NO_x process by reducing the energy barrier of the NO activation, with the energy barrier of 32.57 kcal/mol. It was consistent with the results of catalytic activity that used Pt₁/TiO₂ exhibited better catalytic activity than fresh Pt₁/TiO₂ (ratios differ by 20%). Furthermore, the formation of cocatalytic sites was also confirmed on other single-atom catalysts (Pt₁/Nb₂O₅ and Pd₁/TiO₂), and the role of cocatalytic sites for enhancing the catalytic activity was revealed. Our findings provided an understanding to rationally design highly active single-atom catalysts by introducing adjacent cocatalytic sites.

Materials and Methods

Synthesis of Fresh Pt₁/TiO₂ Catalyst.

First, 0.5 g TiO₂ was dispersed in 50 mL deionized water, and 2.5 mL (1 mg mL⁻¹) chloroplatinic acid (H₂PtCl₆, Aladdin) solution was dropped into the above suspension under vigorous stirring at 70 °C for 4 h. The mixture solution was then washed with deionized water by centrifugation to remove excess Pt atoms which were not stabilized by the TiO₂ support. Finally, the catalysts were obtained by drying at 80 °C overnight and calcining at 300 °C for 4 h in flowing air after ramping up the temperature at a rate of 1 °C min⁻¹.

Synthesis of Fresh Pt₁/Nb₂O₅, Fresh Pd₁/TiO₂, and Fresh Rh₁/TiO₂ Catalysts.

The preparation of fresh Pt₁/Nb₂O₅ catalyst is same as that of fresh Pt₁/TiO₂ catalyst except that Nb₂O₅ (Aladdin, ≥99% purity) replaced TiO₂ as support. And the preparation of fresh Pd₁/TiO₂ and fresh Rh₁/TiO₂ catalysts is same as that of fresh Pt₁/TiO₂ catalyst except that palladium chloride (PdCl₂, Aladdin) and rhodium chloride (RhCl₃, Aladdin) solution replaced H₂PtCl₆ as precursor.

In Situ HAADF-STEM.

The HAADF-STEM images and elemental mapping results were obtained by a Titan 80-300 scanning transmission electron microscopy (STEM) operated at 100 kV and 1 × 10⁻⁸ ~ 1 × 10⁻⁹ Pa with double-aberration correctors and four EDS detectors. For in situ HAADF-STEM, the sample was encapsulated into a chip from the DENS solutions company before the experiments, whose specified range of error for temperature was ≤5%. The holder was then inserted into Titan 80-300 STEM and the air tightness of the system was checked. Argon and reaction gas (H₂:NO ~ 1:1) were introduced to a system of direct-current plasma-enhanced chemical vapor deposition at 125 °C. The chamber pressure was 1 bar.

X-Ray Absorption Spectra Collection and Data Processing.

The X-ray absorption find structure spectra (XAFS) experiment was conducted on the beamline 1W1B station in Beijing Synchrotron Radiation Facility. The EXAFS data were analyzed by ATHENA module implemented in the IFEFFIT software packages (36), and relevant quantitative structural parameters were obtained by ARTEMIS module of IFEFFIT software packages (37).

Method and Model for DFT Calculations.

First-principles calculations were executed by DFT from cambridge sequential total energy package (CASTEP) package with the plane-wave pseudopotential method (38). The interaction of ion core and valence electron was proved by Vanderbilt-type ultrasoft pseudopotential (39). The exchange–correlation interactions were treated by the generalized-gradient approximation (GGA) with spin-polarized Perdew–Burke–Ernzerhof scheme (40). The energy cutoff for the plane-wave basis was set to be 400 eV. The convergence thresholds between optimization cycles for energy change and maximum force were set as 10⁻⁵ eV/atom and 0.03 eV/Å, respectively. According to the surface energy values, the (101) surface was the most abundant one with high thermodynamic stability, followed by (100) and (001) surfaces (41). The corresponding p(2 × 2) TiO₂ (101) surface was constructed using the same method of our previous work (22, 24, 42). The slab contained three Ti₂O₄ layers, i.e., 48 O atoms and 24 Ti atoms. A 15-Å vacuum layer between the images in the direction of the surface normal was utilized. The models of Pt-loaded catalyst were constructed based on the experimental results (especially the results of XAFS) together with the results of our previous theoretical studies (22). A previous study had determined that GGA was reliable for transition state calculations in anatase TiO₂ (101) surface when compared with GGA+U calculations (22, 43, 44); hence, only GGA was adopted in this part (SI Appendix, Fig. S46 and Table S9). The Hirshfeld charge population analysis was used to determine the atomic charge. The adsorb energies of adsorbed species were defined as:

$${\mathrm{\Delta E}}_{\mathrm{ads}} = {\mathrm{E}}_{\mathrm{mol}/\mathrm{sub}} - {\mathrm{E}}_{\mathrm{sub}} - {\mathrm{E}}_{\mathrm{mol}},$$

where E_ads was the total energy of the species adsorbing on the substrate. E_mol/sub was the total energy of the complex consisting of substrate and adsorbed species. E_sub and E_mol were the energy of isolated substrate and adsorbed species, respectively.

Theoretical vibrational Raman spectra were performed within density-functional perturbation theory as implemented in Quantum-ESPRESSO code, which was an integrated suite of computer codes for electronic-structure calculations and materials modeling, and the acronym ESPRESSO stands for opEn Source Package for Research in Electronic Structure, Simulation, and Optimization. (45). The local density approximation to the exchange–correlation function with norm-conserving pseudopotential was employed for the calculation of phonon frequencies (46); 2, 12, and 18 electrons were considered as valence electrons in O, Ti, and Pt elements, respectively. To achieve the converged results, a plane-wave kinetic energy cutoff of 70 Ry (952 eV) was used for the wave functions, and the convergence threshold was set to 10⁻¹⁰ eV and 10⁻¹² eV in the electron and phonon self-consistent calculation, respectively. A Monkhorst–Pack k-point mesh of 1 × 1 × 1 was used to sample the Brillouin Zones for slab systems in Raman spectra relative calculations.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The shared data (including information, code analyses, sequences, etc.) are not included for the data involved in our manuscript, and all the necessary study data are included in the article and/or SI Appendix.