Structure and active centers of the catalysts
Single atomic Pt supported on defective WO3-x and non-defective WO3 nanosheets (samples denoted as Pt SA-WO3-x and Pt SA-WO3) were prepared as the model catalysts (Supplementary Figs. 1-5)14,15. The high-angle annular dark-field-scanning transmission electron microscope (HAADF-STEM) and energy-dispersive X-ray spectroscopy (EDXS) mapping profiles confirmed the individual Pt atoms (marked by the orange circles) dispersed on the surface of WO3-x nanosheet (Fig. 1a and Supplementary Fig. 6). The intensity profile taken along the orange line in the Supplementary Fig. 7, the intensity of the brighter spot (marked with the orange circle) was higher than that of the surrounding W atomic column, corresponding to the existence of Pt single atoms. The characterization of XPS, EPR, H2-TPR and O2-TPD clearly indicated that the rich surface oxygen vacancies on the WO3-x surface compared with WO3 (Supplementary Figs. 8-11), despite they showed the same phase, similar morphology and close specific surface areas (Supplementary Figs. 2 and 12, Supplementary Table 1). This was consistent with the analysis of HAADF-STEM and corresponding Fast Fourier Transform (FFT) (Fig. 1b and 1c), the smooth and regular atomic arrangement in the crystal lattice of WO3 became rougher and distorted after the introduction of defects16. Moreover, compared to the quantity of Pt loading on WO3-x and WO3 (Supplementary Table 2), it was inferred that existence of oxygen vacancy in the WO3-x greatly improved the anchoring of isolated Pt atoms17-19. Besides, XAS was also carried to study the coordination environment of isolated Pt. However, we could not get any valid information of coordination environment of Pt species because of the overlap of the absorption energy range of Pt and W elements (Supplementary Fig. 13).
Notably, the Pt single atoms on the WO3-x were unstable and easily transformed into Pt clusters with diameters of <1 nm after reaction, which can be confirmed by HAADF-STEM (Fig. 1d, marked by the green circles). For comparison, the size of Pt nanoparticles supported by WO3-x remained unchanged after reaction after the same time (Supplementary Fig. 14)20. Further, the structure evolution of Pt species on the WO3-x was further investigated by CO adsorption DRIFTS. The adsorption of CO bands centered at 2120 cm-1 was ascribed to linearly bonded CO on Ptδ+ single-atom sites with a top geometry and the peak appearing at 2084 cm-1 was assigned as adsorbed CO on Pt nanoparticles (Supplementary Fig. 15)21,22. It was noteworthy that the redshift of the IR peak from 2120 cm-1 to 2106 cm-1 after the reaction (Fig. 1e), indicating Pt clusters were formed during the reaction. Moreover, the in situ ambient pressure X-ray photoelectron spectroscopy (AP-XPS) confirmed the transformation of Pt SA to Pt clusters during the reaction (Fig. 1f). In the ultra-high-vacuum (Supplementary Fig. 16a), the binding energy of 75.43 eV and 72.03 eV were assigned to the Ptδ+ (0 < δ < 2) in the Pt SA-WO3-x23. While at reaction conditions (Supplementary Fig. 16b), additional Pt peaks at 74.08 eV and 70.58 eV could be found and assigned to the Pt0 species, which was similar with that of supported Pt NP and indicated the formation of Pt clusters (Supplementary Figs. 17 and 18)24. This was consistent with the results of HAADF-STEM and in situ CO-DRIFTS. Moreover, by a comprehensive consideration of experimental and theoretical results, the size of Pt4 cluster is basically consistent with that of the Pt clusters observed under HAADF-STEM, and that the Pt4 cluster is the most proper model for the clusters that can stably exist on the surface of WO3-x. Therefore, the Pt4 cluster was selected as a model for the real active centers of the catalysts (Fig. 1g).
Visualizing and analyzing the lattice expansion
To unravel the interaction between NO molecules and the surface of catalysts at the real reaction condition, we carried out in situ high-resolution transmission electron microscopy (in situ TEM, experimental details in supporting information). The NO or H2 or a mixture of NO + H2 (volume ratio NO:H2 = 1:4, 1 bar) was introduced to the microscope column at 150°C (423 K). Then, the real-time structural change on (020) surface edge of the Pt SA-WO3-x and Pt SA-WO3 were visualized by time-resolved TEM images as shown in Fig. 2. For Pt SA-WO3-x (Fig. 2a), the lattice fringes with spacing of 3.8 Å was assigned to a clean WO3-x (020) surface under Ar. After the input of NO, we directly observed the lattice fringes expansion of WO3-x. After 380s input of NO, the lattice fringes spacing expanded from 3.8 Å to 4.1 Å. Then followed by purging with Ar, the lattice fringes shrank to 3.8 Å. While exposed to the NO again, the lattice fringe expanded to 4.1 Å again, showing a reversible process, just like the breath. The corresponding dynamic microscopic evolution of catalyst was also recorded in the Supplementary movie 1, which clearly showed that the NO adsorption significantly leads to lattice expansion. Moreover, when the Pt SA-WO3-x was exposed to the H2 + NO mixture, it was found that the lattice fringe remain expansion throughout the reaction (Fig. 2a and Supplementary movie 1). While for the Pt SA-WO3, no lattice expansion or shrinkage was observed (Fig. 2b and Supplementary Movie 2). Furthermore, from the detailed change of outmost, sub and third lattice spacing of Pt SA-WO3-x and Pt SA-WO3 during the reaction in the Fig. 2c and 2d, we could discover lattice expansion of WO3-x induced by NO occurred mainly at the outmost lattice layer. In addition, to rule out the electron beam effects on the lattice expansion during NO reduction, we performed continuous electron-beam irradiation experiments on Pt-WO3-x without reactant NO and H2. The results clearly indicate that the outmost edge of WO3-x surface is truly stable, and the 10-minute irradiation of the electron beam at 200 keV did not induce obvious lattice fringes spacing changes at the outmost edge of surface (see more details in the Supplementary Fig. 19 and Supplementary Movie 3). Therefore, we confirm that the lattice expansion was caused by the reactant molecules, rather than the electron beam irradiation.
To further reveal the chemical nature of lattice expansion during the reaction, in situ Raman spectroscopy was then performed (Fig. 3). As shown in Fig. 3a, according to the rules of Raman spectra, the stretching of the bond length of W-O could cause the decrease of the vibration frequency of W-O25,26. Then the ex-situ Raman was used to characterize the chemical structure of the catalyst, there was no significant difference between pristine WO3-x/WO3 and Pt-loaded WO3-x/WO3 (Supplementary Fig. 20), the bands at ~815 cm-1 (~810 cm-1 for Pt SA-WO3) and ~703 cm-1 (~715 cm-1) were assigned to W-O-W symmetric and asymmetric stretching frequencies, respectively. The bands at ~266 cm-1 (~269 cm-1) and ~328 cm-1 (~324 cm-1) were related to the W-O-W bending modes25,27. When carrying out the in situ Raman analysis, for the Pt SA-WO3-x, all vibrational characteristic peaks (815 cm-1, 703 cm-1, 328 cm-1 and 266 cm-1) of W-O-W bonds were red shifted (809 cm-1, 695 cm-1, 325 cm-1 and 261 cm-1) after flowing NO at 150°C (423 K) for 30 min (Fig. 3b). This was due to the bond length of W-O was stretched after the adsorption of NO, resulting in the decrease of the vibration frequency of W-O. While after switching the gas atmosphere to NO + H2 (volume ratio NO:H2 = 1:4), there was no further shift of the strong peak at ~809 cm-1 corresponds to the W-O-W (Fig. 3c). And all the peaks could restore the original position after purging with Ar. As a comparison, no obvious peak shift was observed on the Pt SA-WO3 (Fig. 3d and 3e). These results were consistent with those of in situ TEM observations and further elucidated that the NO induced lattice expansion could be ascribed to the stretching of W-O-W bonds. Also, the existence of oxygen vacancies was beneficial for the adsorption of NO and the stretching of W-O-W bonds.
Moreover, to further demonstrate the lattice expansion at molecular level, the DFT calculations were conducted. As shown in Supplementary Figs. 21-23, three aspects were considered for building the models: 1) the adsorption of NO on the (010) surface of non-defective WO3 and defective WO3-x; 2) the role of oxygen vacancies on defective WO3-x; 3) the adsorption of NO on Pt4 cluster supported on defective WO3-x28. Results indicated that Pt cluster had the strongest adsorption capacity for NO, and the adsorption energy of one NO on non-defective WO3 was lower than those of adsorption on vacancy sites of defective WO3-x and on Pt4 centers (-0.51 eV vs. -2.71 eV vs. -3.56 eV). However, the strength of lattice expansion of W-O-W did not follow this order. The length of surface W-O-W bond (equal to lattice spacing) expanded from 3.79 Å to 3.91 Å after adsorption of one NO on the oxygen vacancy site, while the strength for adsorption on Pt4/WO3-x was very weak (from 3.83 Å to 3.85 Å) and there was no change in lattice spacing of non-defective WO3 (from 3.73 Å to 3.73 Å). Besides. we also calculated and found that the lattice expansion was positively correlated with the number of oxygen vacancies (Supplementary Table 3). The above theoretical calculations revealed that the lattice expansion was mainly caused by the adsorption of NO on the oxygen vacancy sites, which induced the stretching of W-O-W bonds.
Correlation of activity and lattice expansion
To elucidate the relationship of catalytic performance and lattice expansion, we also chose reduction of NO with H2 as the model reaction (Fig. 4)29. Before designing of proper experiments, several tests about catalytic performance have been performed concerning the effect of supports and the active centers (Fig. 4a and 4b). Firstly, both defective WO3-x and non-defective WO3 supports exhibited very poor activity in the whole temperature range, although the defective WO3-x had a strong adsorption capacity for NO (see the results of NO-TPD in Supplementary Fig. 24). While introducing Pt species on the catalysts could dramatically improve the catalytic activity. Moreover, both of Pt SA-WO3-x and Pt NP-WO3-x exhibited much better deNOx catalytic activity than Pt SA-WO3 and Pt NP-WO3 in the low temperature of 50-250°C. Specifically, comparison of the Temperature programmed surface reaction (TPSR) results of NO + H2 reaction on the Pt SA-WO3-x and Pt NP-WO3-x catalysts at 150°C (Supplementary Fig. 25), both of them had excellent N2 selectivity (Supplementary Fig. 26), but Pt SA-WO3-x exhibited better catalytic performance. These results proved that Pt was the main active center for NO adsorption and activation and the oxygen vacancy mainly exhibited as the adsorption site of NO30.
Accordingly, the lattice expansion was mainly related to oxygen vacancies of support and had no relation to the loading of Pt atoms or particles on support. Besides, the accurate loading amounts of single atom were quite difficult to control on WO3-x and WO3. Based on the above considerations, an experiment was designed to further confirm the relationship of catalytic performance and lattice expansion. We used WO3-x and WO3 to load the same amount of Pt nanoparticles (0.3 wt%), and there was no obviously difference in the dispersion state of Pt particles observed by TEM (Supplementary Fig. 27), which ensured that there was almost no difference in the Pt active centers between the two catalysts. The result indicated that Pt NP-WO3-x show better catalytic activity than Pt NP-WO3 (ratios of differ by 20%, Fig. 4c). In this case, by excluding the effect of active centers and support, we can further correlated the activity difference with three reasons: 1) lattice expansion of Pt NP-WO3-x induced by the adsorption of NO, 2) the pre-adsorbed NO on the oxygen vacancy could easily transferred to Pt active sites for activation, 3) the activated H* on Pt species spills over the adsorbed NO on the oxygen vacancy for further reaction. Then DFT calculations were carried out to evaluate these viewpoints, we created an oxygen vacancy in the ortho position of Pt4 cluster, and then adsorbed a NO molecule on the oxygen vacancy. Then the results show that the adsorbed NO could not be transferred to the Pt4 site to react with the activated H* due to a large energy barrier (Fig. 4d), and vice versa. The activation of H2 and NO tend to occur only on Pt species. Therefore, we could infer that the enhanced catalytic activity of Pt NP-WO3-x was mainly correlated with the lattice expansion effect.
Elucidation the effect of lattice expansion
To further evaluate the role of surface lattice expansion for enhancing the catalytic activity, comprehensive DFT calculations were then conducted (Fig. 5). Firstly, we proposed a catalytic cycle for reduction of NO with H2 on Pt4/WO3-x model catalyst (Fig. 5a). The reaction path expression and reaction energy barrier for each step were shown in Table S4. The cycle was initiated by adsorption of NO and H2 on the Pt4 site. The results showed H2 could dissociate into two adsorbed H* directly and generally considered to be barrier less31,32. Then one adsorbed H* could be transferred to the adsorbed NO* to form NOH* (NO* + H* = NOH*) with an activation barrier of 1.19 eV (TS1). Thereafter, the other proton from dissociation of H2 would react with the NOH* (NOH* + H* = N* + H2O) to generate H2O via TS2 (1.03 eV). Subsequently, the other NO could adsorb on the N site and produce N-NO, and then activated and dissociated the second H2 (step Ⅵ). Similarly, the intermediate N-NOH was generated from the HN-NO, which was produced from the combination of H* and N-NO with an energy barrier of 0.94 eV (TS3). However, results showed that the other H* from dissociation of second H2 could not be transferred to the adsorbed N-NOH to form the second H2O due to a large energy barrier. In our experiments, the H2 in the actual reaction process was in excess amount. Thus, we introduced the third H2 molecule, and the N-NOH could react with the active H* to generate another H2O (N-NOH* + H* = N-N* + H2O) via TS4 (1.50 eV), which was the rate-determining step. Finally, the adsorbed N-N could form N2, and the produced H2O and N2 desorbed from the surface and complete the catalytic cycle. However, note that the structure of Pt4 changed from tetrahedral to square after the reaction (Supplementary Fig. 28). Previous DFT studies had shown that higher reactivity of NO decomposition was obtained on Pt cluster, which had a square structure composed of four Pt atoms28. Therefore, we further performed the catalytic cycle for the Pt4 cluster with a square structure, and found it could also complete the reduction of NO with H2 cycle, and follow the same reaction path and keep the stable square structure (Supplementary Figs. 29 and 30, Supplementary Table 5). These results also reveal that the whole reaction mechanism and path was reasonable via both thermodynamic and kinetic pathways.
Furthermore, considering that the lattice expansion was a metastable process, we have established a special model of Pt4/WO3-x (Supplementary Fig. 31) to investigate its effect toward catalytic activity, which artificially stretched the surface W-O-W lattice spacing from 3.8 to 4.0 Å. Then the initial step (adsorption of NO on Pt) and rate-determining step (generation of the second H2O) of the reaction were evaluated. The results showed that when the lattice expansion happened, the adsorption energy of NO and the barrier energy of the rate-determining step decreased from -3.56 eV and 1.52 eV to -2.44 eV and 1.37 eV, respectively (Fig. 5b and 5c, Supplementary Table 6). Therefore, the moderate adsorption energy of NO and the decreased barrier energy would enhance the catalytic activity33. This result further proved the lattice expansion of W-O-W occurring on the interface of the NO-Pt4/WO3-x was beneficial to the activation of NO and promotes the reaction.