In the previous studies, we have discovered the existence of strong metal support interaction between metals and the transition metal carbides and nitrides25–27. In all the studied systems, the catalysts exhibited several common properties, including the electron deficient supported metal centers (charge transfer from metal to support) and the capability of maintaining the fine dispersion of metal centers even after high temperature treatment in the activation atmosphere. Evoked by the SMSI effect in the TMCs/TMNs supported catalysts28, 29, we focused our notice in verifying whether the molybdenum nitride as the alternative host material can spit metals from 3D nanoparticles to sub nanometer or 2D layers, even to single atoms, which is called the reversing sintering effect16. The AIMD simulation was applied to investigate the structural evolution of a Ni nanoparticle on the γ-Mo2N support under thermal perturbation (Fig. 1 and Video S1-S6). The temperature factor was set at 590°C which is in accordance with the commonly used activation temperature of traditional M/γ-Mo2N(C) catalysts25, 30. The initial structure of the Ni/γ-Mo2N model at 0 ps was constructed by 19-atom Ni particle placed on a Mo terminated γ-Mo2N(111) surface. Due to the strong binding of the Ni with the γ-Mo2N substrate, an instant collapse of the 3D structure of Ni particles was observed. Within 3ps, the Ni atoms have spread into a raft-like monolayer on the γ-Mo2N (Fig. 1a). No significant structure changes of the raft like particles occurred as the simulation time prolonged. In comparison, when the surface of molybdenum nitride was covered with oxygen, the NiO species were found to be incapable to spread into under-coordinated species (Fig. 1b), indicating the strong interaction between Ni and molybdenum nitride requires both materials under reduced state. Meanwhile, no wetting phenomenon is observed on the Ni/CeO2 system (Fig. 1c), except for a small shape reconstruction from the cubic into hemispheric particles. It can be seen from Fig. 1d that the relative energy change of the redispersion process of Ni on the γ-Mo2N is about − 18.0 eV. In contrast, the ΔE of the re-shape procedure in the Ni/CeO2 model is only − 3.0 eV. The significant difference of the structure evolution of Ni19 particles on nitride and oxide supports is controlled by the thermodynamics of the Ni-γ-Mo2N and the Ni-CeO2 interfaces.
In order to test the theoretical prediction of AIMD results, a model catalyst composed with the pre-synthesized Ni nanoparticles (4 nm)31 and γ-Mo2N was prepared. The Ni particles were synthesized using the high temperature liquid phase synthesis method with oleylamine and tributylphosphine as the capping reagents31. The loading of Ni on the support is controlled at ~ 2wt%, and further confirmed by ICP-OES. The TEM image of the as-synthesized Ni NPs demonstrated that the average size is about 4.0 nm with a narrow size distribution and sphere-like shape (Fig. 2a), but the Ni NPs were partially oxidized after deposited onto the passivated FCC structured γ-Mo2N support (Fig. 2b). The HAADF-STEM images of the fresh 2%Ni-4 nm/γ-Mo2N catalyst were also collected along with the EDX elemental mapping. It can be seen from the STEM and elemental mapping images that the supported Ni appeared at the surface of γ-Mo2N and maintained their original size and shape (Fig. 2c-d).
The prepared model catalysts were used to track the thermal influences on the morphology of the 4 nm supported Ni particles using combination studies of a various of in-situ characterization methods. Both Ni-4nm/γ-Mo2N and Ni-4nm/CeO2 catalysts were treated under the reductive atmosphere during a temperature-programmed heating experiment from room temperature to 590°C. The spectroscopic studies of in-situ quick X-ray adsorption near edge spectroscopy (QXANES) and the ambient pressure X-ray photoelectron spectroscopy (AP-XPS) were performed at the Ni K edge or Ni 3d region to confirm the chemical state and the electronic structure of the supported Ni species (Fig. 3). When the sample was treated in the flow of N2-H2 mixture, a gradual reduction of the sample started at around 360°C based on the QXANES measurement (Fig. 3a, left panel). The sample was fully reduced to Ni (0) state at around 480°C judging from the edge position and “white-line” profiles. With the further thermal treatment, an unusual intensity increasing emerged at the “white-line” of Ni-4nm/γ-Mo2N-590 catalyst (Fig. 3a, right panel), which was probably due to the electron synergistic effect. The higher intensity of the Ni K edge “white-line” signal indicated that the electron density at the Ni site was weakened due to the charge redistribution from Ni to the molybdenum nitride support25. The electron structure change of the Ni-4nm/γ-Mo2N catalysts was further monitored by the surface sensitive AP-XPS technique (Fig. 3b). In the fresh sample, an obvious oxidation layer can be observed at the γ-Mo2N surface. The Mo (V) and Mo (VI) species were about 31% of the total Mo species observed. Meanwhile, most of the Ni species was NiO and only a small amount of Ni(0) species could be seen. After the activation, the signals of molybdenum nitride increased and the reduction of the supported Ni occurred. With the increasing temperature, the Ni(0) binding energy of Ni-4nm/γ-Mo2N-520 shifted ~ 0.4 eV positively25, which is probably related to the charge transfer from Ni species to the γ-Mo2N supports, in good agreement with the changing of “white line” of QXANES characterization. The charge transfer phenomenon has also been confirmed by the DFT calculations, as shown in Fig. 3c and 3d. The electron transfer from Ni to nitride occurred and the electron density at the Ni-nitride interface increased significantly. The Ni atoms closed to the molybdenum nitride were generally under electron deficient state. The formation of the 2D Ni raft-like clusters has maximized this electron interaction, hence the average electron deficiency per Ni atom in the layered Ni structure will be significantly larger than the Ni particles.
Furthermore, in-situ X-ray adsorption fine structure32 (XAFS) of Ni K edge was carried out to evaluate the size of Ni domains and the surrounding coordination environment changes after the thermal treatments at 400 and 590°C for 40 minutes. Figure 4a presents the Ni K edge XANES spectra of Ni-4nm/γ-Mo2N and Ni-4nm/CeO2 catalysts after 590°C reduction. Compared with the Ni and NiO standards, the Ni-4nm/CeO2-590 catalysts exhibited similar pre-edge and near edge features with the Ni foil, indicating the supported Ni species were almost fully reduced after the high temperature reduction. In comparison, the pre-edge feature of Ni/γ-Mo2N-590 catalyst is slightly weaker than the metallic standard and the XANES oscillation appeared at 8365 eV and higher energy regions cannot be described by neither the Ni (0) nor the NiO (Fig. 4a, the right panel). This phenomenon suggested the Ni formed a special electronic and coordination structure completely different from the metallic and oxide standards. The further EXAFS fitting of the steady state XAFS spectra was performed to reveal the detailed coordination structure of supported Ni species. The much stronger intensity of the Ni-Ni coordination peak of Ni-4nm/CeO2 catalyst in R-space FT-EXAFS spectra than that of the Ni-4nm/Mo2N catalyst were observed (Fig. 4b), indicating the size of Ni species on γ-Mo2N is much smaller than that on the CeO2 substrate. Indeed, based on the EXAFS fitting results (Table S1, Figure S1 and S2), the C.N.(Ni-Ni) of Ni-4nm/CeO2-590 was 10.8. While the C.N.(Ni-Ni) of the Ni-4nm/γ-Mo2N-590 was only 4. To track the temperature effect on the coordination shell of the Ni domains on the molybdenum nitride supports, the EXAFS spectra of Ni-4nm/γ-Mo2N-400 at the different temperatures were also collected and presented in Fig. 4c. It suggested that the supported Ni was fully reduced after 400°C reduction, as the major neighbor atoms of Ni was Ni atoms located at about 2.49 Å, corresponding to a typical metallic Ni-Ni coordination. In addition, a Ni-N coordination peak appeared at 1.90 Å in the R space spectrum was identified as the Ni-N coordination (C.N.Ni−N of 1.1). This feature confirmed that the metallic Ni particles were located on the N-interlayer which was predicted by the AIMD simulation (Fig. 1a). With the elevated temperature, a novel Ni-Mo bonding at 2.62 Å appeared in the Ni/γ-Mo2N-590 catalyst. These phenomena also indicated that the Ni has formed direct interaction with the molybdenum nitride support, confirming the similar structure variation as the AIMD simulation presented.
The spectroscopic evidence has given a detailed description on the electronic properties and coordination environment changes of the reverse-sintering of supported 4nm Ni particles over the molybdenum nitride surface. Electron microscopy is expected to directly monitor the structure evolution of the 4nm Ni particles downsizing into the raft-like clusters. However, in conventional in-situ TEM/STEM experiments, thin and light supports, for instance, activated carbon, should be used or one has to find a proper orientation where electron beam is parallel to interface between nanoparticles and supports33, 34. In this work, the relatively lower Z-contrast of Ni particles than the Mo makes it difficult for the conventional environmental scanning transmission electron microscope (STEM) imaging techniques to distinguish the Ni from the γ-Mo2N substrate25. Moreover, it is hard to derive the surface information from the recorded 2D projection when samples were thick (above 100 nm) or in irregular shape. The new electron microscopy technique of environmental probe-corrected scanning transmission electron microscope equipped with secondary electron detector was applied to achieve the simultaneous acquisition of SE image and STEM image (STEM-ADF, STEM-BF) with an atomic spatial resolution (below 1Å)35. Utilizing the surface sensitive low energy secondary electron caused by interaction between the primary beam and object, SE images showed a powerful ability to analyze surface morphology on bulk materials, regardless of the thickness and Z contrast of the metal on the support36. With the assistance of the SE-STEM method, we managed to observe the structural evolution of Ni particles on γ-Mo2N directly in-situ during the high temperature treatment under the designated atmosphere.
As shown in Fig. 5a, the γ-Mo2N support is highly porous, and the supported nanoparticles were well dispersed on the surface of the γ-Mo2N at 400 ºC in the simulated activation atmosphere. The corresponding BF and ADF-STEM did not show useful information since the support is too thick (Figure S3). On the SE images, the supported nanoparticles showed an irregular polygon shape (about 5–7 nm) with clear edge (Fig. 5). No significant morphology change was observed with the time evolution at 400 ºC (Fig. 5a and b). The detailed analysis of the fast Fourier transformation (FFT) of SE and BF images of the Ni-4nm/γ-Mo2N at 400 ºC were shown in Fig. 6a and b. Due to the penetration depth differences of secondary electron and traditional electron probes, the FFT patterns of SE and STEM reflect the structural information of the near surface species and the bulk of the sample respectively. As shown in the Fig. 6a, the region 1 shows a typical pattern of FCC structured γ-Mo2N with a d-spacing around 2.0 Å, which is the bare support. While region 2 is confirmed unexpectedly as Ni4N particles (d-spacing 2.6Å, Figure S4) rather than the metallic Ni. This phenomenon suggested that the formation of Ni-N bonding changes the bulk structure of loaded Ni species. While the FFT of the BF image at region I is identical to the SE one, confirming the bulk phase of region 1 is the substrate. In contrast the features of both γ-Mo2N and Ni4N could be seen at region II in the FFT of STEM images, demonstrating the bulk of the catalyst below region II is composed by both the Ni4N particles and the nitride support. These results also demonstrate that the loaded Ni particles are 3D particles with considerable thickness at 400°C. With the elevated activation temperature, the supported nanoparticles on the surface begins to fade, or even disappear. The SE images of 520°C (Fig. 5d, g and h) proves that the pores in the γ-Mo2N substrate have disappeared, probably been covered by the loaded species, indicating the Ni has formed raft-like clusters and spread over the support (Video S7). The FFT patterns of SE and STEM (Fig. 6c and d) further confirmed the thermal induced redispersion phenomenon. The SE FFT pattern at region III showed a typical Ni4N features with d-spacing around 2.5 Å, suggesting the surface of region 3 is still covered by the Ni species. However, the STEM FFT pattern at the same region only showed features belong to γ-Mo2N, which demonstrates that the Ni particle is possibly too thin to generate diffraction patterns. Compared with the electron diffraction patterns collected at 400°C, we could confirm that the Ni particles have been spread into 2D layers on the nitride substrate under the thermal treatment. The normalized intensity profiles of the 520°C sample’s surface were much weaker than those of the 400 and 500°C samples (Figure S5), which is another evidence that the Ni particles have reverse sintering at high temperature. The EDS mapping on STEM of ex-situ samples of Ni-4nm/γ-Mo2N-590 catalysts were also carried out to further confirm the spontaneous dispersion of Ni NPs on the γ-Mo2N (Figure S6)
As the SE-STEM characterization has confirmed the existence of Ni4N during the thermal treatment, it is important to further confirm whether the Ni4N is able to further reverse sintering to undercoordinated structures. The AIMD simulation has been done using the same setting to the Ni19/γ-Mo2N over a novel Ni4N(Ni20N5 cluster was truncated from bulk Ni4Nm/γ-Mo2N model surface (Fig. 6e, Video S8-9). The spherical Ni4N turns directly into layered structure with maximized Ni-N-Mo bonding with in 5 ps time evolution. The relative stabilization energy is calculated as 26.8 eV, even larger than the energy change of Ni19/γ-Mo2N system (Fig. 6f), ensuring that the loaded Ni particles will be converted into layered structure under thermal perturbation.
The special chemical properties of the under-coordinated Ni species derived from the strong interaction between supported Ni and the molybdenum nitride substrates have been utilized in the catalytic hydrogenation of carbon dioxide. It is a common sense that the Ni based catalysts were highly active in the methanation reactions of CO and CO2 with exclusive high selectivity to the product methane37–39. Ever since the discovery of the CO2 methanation by Sabatier and coworkers in 190240, Ni has been considered to be the typical inexpensive metal for this hydrogenation conversion process. Mechanism studies have confirmed that the domains of metallic Ni particles should be above 2 nm in order to achieve the optimal methane formation rate and efficiency38, 41. Indeed, when using the activated Ni-4nm/CeO2 as the catalyst, the selectivity of methane was over 70% at 400 °C (CO2 conv. ~8%), and ~ 60% at 500 °C (CO2 conv. ~12%). While in comparison, the production of methane has been significantly suppressed on the activated Ni-4nm/γ-Mo2N catalyst. In the range from 250 to 500°C, the selectivity of CH4 has never exceeded 3%, even at the low temperature region at which the reverse water-gas shift reaction is thermodynamically unfavorable. When the working temperature went above 400°C, the CO2 was chemoselectively converted into CO (S(CO) > 99%) (Fig. 7a). What’s more, the extraordinary selectivity toward CO remained above 95%, when decreasing the mass space velocity to 7500 ml/g/h (Fig. 7b) or tuning the CO2/H2 ratio (1:3 to 1:6) in the gas feed (Fig. 7c). Under a typical condition, the activated Ni-4nm/γ-Mo2N catalyst exhibited stable activity in the hydrogenation of CO2 for over 50 hrs. Neither the CO2 conversion decay nor the decreasing on the selectivity of CO has been observed (Fig. 7d). The completely selectivity reversion of the Ni/γ-Mo2N catalysts suggested that the reverse sintering phenomenon induced by the strong interaction between the Ni and Mo2N was able to significantly change the intrinsic catalytic properties of Ni based catalysts and prolong the stability of Ni catalysts (Fig. 6e). It has also been demonstrated the reverse sintering effect in the Ni/γ-Mo2N catalyst is able to enhance the water splitting activity by increasing the exposure of the active metal sites. Comparing the fresh catalyst with the Ni-4nm/γ-Mo2N-590 catalyst, the hydrogen evolution rate increased by about 10 times (Figure S7). Therefore, the reverse sintering effect can be used to maximize the dispersion and utilization efficiency of metals using simple treatments.
In conclusion, we have demonstrated using theoretical calculation and ab initio molecular dynamic simulation methods that, the pre-synthesized 4 nm Ni particles are able to reverse sintering into under-coordinated Ni species after the high temperature activation procedure driven by the strong interaction between the γ-Mo2N and the Ni. The existence of both reduced Ni particles and the bare γ-Mo2N is important for the formation of the highly dispersed Ni species. In-situ structural characterizations have confirmed the dispersion of Ni occurred after the reduction of bulk phase Ni and the removal of surface passivated O-layer of γ-Mo2N. The reverse sintering effect for Ni-4nm/γ-Mo2N catalyst has a positive effect on the chemoselective hydrogenation of CO2 to CO. Compared with the Ni-4nm/CeO2 reference catalyst (CO Sel.%~29%), the activated Ni-4nm/γ-Mo2N catalyst exhibited over 96% CO selectivity at an even higher CO2 conversion. This reverse transformation in the catalytic performance compared with traditional Ni based catalysts can accounts with the dispersion and wetting phenomena of Ni nanoparticle on γ-Mo2N, which serves as an excellent example of the potential application of reverse sintering effect in the high temperature favorable reactions.