Morphology of ZSM-5-MT catalyst
ZSM-5-MT catalyst was synthesized via a one-pot hydrothermal process (as shown in Methods). The three-dimensional (3D) morphology of ZSM-5-MT catalyst is revealed by the 3D electron tomographic reconstruction in Fig 1a and Supplementary Video 1. In such ZSM-5-MT, a tenon-like protrusion vertically grows on the (010) surface of a underlying ZSM-5 crystal as a mortise subunit (as shown in the schematic model Fig. 1b). The morphology of ZSM-5-MT can also be confirmed by the annular dark field (ADF) STEM images (Fig. 1c-e) and the SEM images (Supplementary Fig.1) of various 2D projections. In the X-ray diffraction (XRD) results in Supplementary Fig.2, we only detect the pure MFI zeolite phase in both ZSM-5-MT and traditional short-b-axis ZSM-5 (ZSM-5-Sb)33-35. Thus, the mortise-tenon morphology should be caused only by the intergrowth of ZSM-5 lattices. Consistent pore structures of ZSM-5-MT and ZSM-5-Sb are confirmed by the gas physisorption experiments. As shown in Supplementary Fig.3 and table 1, there is no evidence that additional mesopores are introduced in ZSM-5-MT, indicating that the tenon and mortise subunits are closely connected without any gaps.
Then, we used the iDPC-STEM to identify local structures and study such intergrowth behavior in detail. In the iDPC-STEM image obtained from the purple frame in Fig. 1d, we can observe three different areas according to different projected structures as shown in Fig. 1f. In this projection, the pure tenon area shows the typical (010) surface with the ordered 10-membered rings of straight channels, while the pure mortise area shows the (100) surface with the sinusoidal channels. It indicates that the crystallographic axes a and b are rotated by 90° around c in space to form the tenon and mortise respectively. Meanwhile, in the intergrowth area in Fig. 1f, we find the overlapping contrast of (010) and (100) lattice planes where the straight channels are blurred by the perpendicular lattices above. It means that the mortise and tenon subunits grow into each other in a perpendicular intergrowth structure just like the structure of a traditional mortise-tenon junction.
We also investigated the connection of channels in tenon and mortise using the images with different defocuses. Fig. 1g and h are obtained from the same area marked by the red frame in Fig. 1e but with different defocuses. In Fig 1g, the (100) surface of mortise subunit is in focus and the sinusoidal channels are clearly imaged in accordance with the structural model. While, in Fig. 1h, the (010) surface of tenon subunit with straight channels is in focus after changing the defocus. Combining with the imaged structures in Fig 1g and h, it can be deduced that the straight channels in mortise and the sinusoidal channels in tenon are perfectly connected without any dislocations and regardless of local lattice mismatching. We can also confirm such channel connectivity from another projection where the (010) surface of mortise subunit and the (100) surface of tenon subunit can be imaged with different defocuses. As marked by the red hexagons in Fig 1i and j, the projected positions of the channels in mortise and tenon are coherently located. Moreover, we also revealed other local structures in the ZSM-5-MT, including surface terminations, surface steps and step-edge sites (Supplementary Fig. 4). These results provide an overall understanding on the morphology of ZSM-5-MT and reveal such intergrowth behavior with confirmed channel connectivity preliminarily.
Resolving the atomic interface structures of ZSM-5-MT catalyst
Atomic structures of intergrowth interfaces can be investigated by the imaging from the [001] direction of ZSM-5-MT. In this projection (Fig. 2a), it is possible to observe the interfaces to reveal how two areas with different lattice orientations (tenon and mortise) grow into each other. Fig. 2b is the iDPC-STEM image obtained from the blue area in Fig. 2a, which clearly shows the lattice characteristics in this projection. Fig. 2c gives two structural models with a 90° rotation in the [001] projection, where the straight and sinusoidal channels are marked out. The green arrows in Fig. 2c show a characteristic pattern (projected Si-O island), which helps us to identify the lattice orientations. Such characteristic patterns can be observed from the iDPC-STEM and simulation (Fig. 2d). Then, at the intergrowth area, these rotating lattices with two orientations will overlap. If the characteristic arrows in upper and lower lattices are the same, the characteristic patterns are retained in the overlapped image. If not, a square pattern appears in the overlapped image instead. These results are shown in the image, simulation and model in Fig. 2e-g. Based on our analysis on image characteristics, we can identify two areas with different lattice orientations (tenon and mortise) according to the green arrows in Fig. 2b. Meanwhile, we can also use the red frames to outline an intergrowth area that contains the interfaces between the tenon and mortise subunits, where the image is just the overlap of the lattice images in two subunits growing into each other.
Furthermore, we resolved such interface structures in detail. In the [001] projection, we can observe not only the characteristic patterns but also the contrast of O atoms between them. Based on the structural model in Fig. 2h, there is an inherent missing of O atoms at the interface between the 90°-rotation lattices. In order to study such atom missing, we used the profile analysis of iDPC-STEM images to give the positions and intensities of O peaks as shown in Fig. 2i. The normalized intensities of O peaks can describe the different numbers of atoms in these O atom columns semi-quantitatively. We have made statistics on the intensities of these O atom columns in single-crystal (tenon or mortise) and intergrowth areas respectively according to the profile data in Supplementary Fig. 6-9. As shown in Fig. 2j, the statistical intensities are all higher than 0.6 for the single-crystal area, while the results for the intergrowth area are varying in a wide range (0.2-1). It indicates the obvious missing of O atoms at the interface in this intergrowth area. This can be interpreted that the increasing distances between Si or Al atoms due to the lattice mismatch at the interface make them impossible to be bonded by O atoms. These results reveal the interface structures of tenon-mortise intergrowth in atomic precision, and the non-tetra-coordinated Al atoms formed by the missing of O atoms will generate more Lewis acidity to improve the performances of ZSM-5-MT catalyst.
Inherent Lewis acidity in ZSM-5-MT catalyst
In order to establish the structure-property relation of ZSM-5-MT catalyst, we studied the inherent Lewis acidity in ZSM-5-MT compared with traditional ZSM-5-Sb without intergrowth. First, the acid property was investigated by the temperature-programmed NH3 desorption (NH3-TPD) in Supplementary Fig. 10. The NH3-TPD results show that the peaks of both weak and strong acid sites in ZSM-5-MT shift to higher temperatures than those in ZSM-5-Sb, which demonstrates an unusual change in the acid properties of ZSM-5-MT. Then, the acid sites in ZSM-5 zeolites are detected by Fourier transform infrared (FTIR) spectroscopy of adsorbed pyridines and carbon monoxides (CO). These two different probe molecules were used together for a deeper knowledge on the nature of acid sites. The FTIR spectra using pyridines as probe molecules are shown in Fig 3a. The bands at 1430-1470 cm-1 and 1544 cm-1 can be ascribed to coordinately bond pyridines (pyridines interacting with LASs) and pyridinium ions (pyridines interacting with BASs) respectively15,36,37. As we summarized in Supplementary Table 2, the concentration of BAS in ZSM-5-MT is nearly equal to that in ZSM-5-Sb, while the concentration of LAS in ZSM-5-MT is one-order of magnitude higher than that in ZSM-5-Sb. Meanwhile, the signal of LAS in ZSM-5-MT is slightly blue shifted (~17 cm-1) with regard to the signal in ZSM-5-Sb, which implies the formation of additional acid sites with Lewis acidity in ZSM-5-MT.
To finely distinguish between multiple LASs, we further detected these acid sites using CO as probe molecules for the FTIR spectroscopy (Fig. 3b). The peak at 2129 cm-1 indicates the physisorbed CO and the peak at 2168 cm-1 results from the CO interacting with the BASs in zeolites38. Interestingly, the ZSM-5-MT shows two bands at ~2338 cm-1 and 2358 cm-1, which are attributed to the CO interactions with penta-coordinated and tri-coordinated LASs respectively after the dehydration of zeolites during sample pretreatment (Fig.3c), while the ZSM-5-Sb does not. Such Al-CO interactions can be promoted by the de-coordination of water from the octahedral Al sites. Combining with the iDPC-STEM imaging results, it can be concluded that these penta-coordinated and tri-coordinated LASs are closely related to the inherent missing of O atoms at the tenon-mortise interfaces, which fully agrees with our prediction on Lewis acidity.
To verify Al sites in terms of their coordination, 27Al magical-angle-spinning (MAS) nuclear magnetic resonance (NMR) was performed under ambient conditions without pre-treatment (dehydration). The 27Al MAS NMR spectra (Fig. 4d) of both ZSM-5-MT and ZSM-5-Sb show sharp peaks at 58 ppm that are attributed to tetra-coordinated Al in bulk framework. It validates that most of Al atoms are incorporated into framework. Meanwhile, the ZSM-5-Sb shows a broad peak of chemical shift at 0 ppm belongs to distorted octahedral Al species. As for the ZSM-5-MT, this peak is higher than that for the ZSM-5-Sb and shifts to 4 ppm corresponding to the boehmite-like structures (given in Supplementary Fig. 11)39. It can be deduced that such boehmite-like structures only exist when the high-density LASs are very close in space to connect, for example, when the LASs are concentrated at the interface of zeolite intergrowth. Based on the FTIR and 27Al MAS NMR results, the inherent LASs caused by the missing of O atoms in ZSM-5-MT are confirmed experimentally, which is also consistent with the imaging results above.
Catalytic performances of ZSM-5-MT catalyst
To address the effect of these AlFR Lewis acid sites, we tested the catalytic performances of ZSM-5-MT and ZSM-5-Sb catalysts in methanol conversion at high conversion rate (≥ 99%). The gas product selectivities are given in Fig. 4a-b and Supplementary Fig.12, which indicates that two catalysts with different architectures exhibit quite different catalytic performances. The selectivities of propylene (25-29%) and butene (12-13%) over ZSM-5-MT are higher than those over ZSM-5-Sb (18-19% and ~ 11%, respectively). Meanwhile, the ZSM-5-MT shows lower selectivities of C1-C5 alkanes (32-38%) and aromatics (~ 8%) than the ZSM-5-Sb (42-44% and ~ 12%, respectively). It is known in the hydrocarbon pool (HP) mechanism that the aromatics-based cycle produces almost equal amounts of ethylene and propylene, while the olefins-based cycle will produce much more propylene than ethylene40-42. Therefore, the relative contribution of two cycles can be represented by the ratio of ([propylene]-[ethylene])/[ethylene] ((P-E)/E). As shown in Fig. 4a and b, the descriptor (P-E)/E ratio of ZSM-5-MT is much higher than that of ZSM-5-Sb, implying that the olefins-based cycle is more significantly enhanced in ZSM-5-MT. And, another criterion for determining the relative contribution of two cycles is the hydrogen transfer index (HTI)10. As shown in Supplementary Fig.13, the C4-HTI and C5-HTI of ZSM-5-MT are lower than those of ZSM-5-Sb, indicating the less vigorous hydrogen transfer reactions in ZSM-5-MT. Thus, both of these two criteria demonstrate that the enhanced olefins-based cycle is induced in ZSM-5-MT. To explicitly illustrate the hydrogen transfer (HT) capacity, we further evaluated the catalytic performances of ZSM-5-MT and ZSM-5-Sb catalysts in propylene conversion (Fig. 4c). The ZSM-5-MT catalyst shows lower selectivities of both propane (~ 16%) and butane (~ 6%), which are the predominant HT products in propylene conversion, than the ZSM-5-Sb catalyst (~ 22% and ~ 7% respectively). This comparison reveals that the introduction of AlFR-LAS-enriched interface in a mortise-tenon ZSM-5 catalyst significantly inhibit the olefin-induced hydrogen transfer reaction, thus resulting in a higher propylene selectivity in methanol conversion. Overall, we validated that the modulation of AlFR LASs via interface engineering can be used to design efficient methanol-to-olefins catalysts.