Topotactic reactions play a pivotal role in the innovative engineering of various types of crystalline materials including zeolites, ceramics, alloys and battery cathodes1-18. Conventional synthetic routes often encounter challenges in producing these materials, making topotactic reactions a key strategy for their preparations. While obtaining crystal structures of the starting and final phases is crucial, unravelling the atomic-scale structural changes in reaction intermediates is equally important for understanding the underlying mechanisms of these reactions5,10,15,19-24. This subtle understanding is imperative for advancing the tailored design and synthesis of materials through topotactic reactions.
A topotactic transformation involves the displacement or exchange of atoms within a crystal structure, typically maintaining the overall framework of the original structure. It leads to products and intermediates with similar unit cells and subtle structural differences. Furthermore, intermediate structures often exhibit partially occupied and/or disordered atomic sites. While single-crystal X-ray diffraction (SCXRD) has been powerful for determining atomic structures of large crystals (>5×5×5 μm3)10,22-28, powder X-ray diffraction (PXRD) encounters challenges due to peak overlap11,14,29,30, hampering the detection of subtle structural changes during topotactic reactions.
In contrast, electrons interact with matter about 104 times more strongly than X-rays31, making them well-suited for studying crystals that are either too small for SCXRD or too complex for PXRD. Although high-resolution electron microscopy has provided valuable insights into topotactic transformations at atomic-scale5,8,16,20,32-35, its application is often semi-quantitative and limited to simple structures observable through 2D projections along specific directions. Over the past two decades, three-dimensional electron diffraction (3D ED), sharing conceptual similarities with SCXRD, has emerged as a powerful technique for structural studies, effectively overcoming limitations associated with crystal size and complexity36. Notably, 3D ED has become the most important technique for structure determination of novel zeolites.36-38.
Zeolites are a class of microporous materials with well-defined pore structures and channel systems in molecular dimensions. They are widely used as catalysts, adsorbents, and ion-exchangers. Most of zeolites are synthesized as submicrometre- or nanometre-sized crystals and have complex framework structures. The framework structures of zeolites are composed of TO4 tetrahedra (T= Si, Al, B, Ge, Ti, etc.). These tetrahedral atoms can be selectively removed from or incorporated into zeolite framework structures through topotactic reactions. This distinctive structural characteristic makes topotactic reactions a unique strategy for preparation of novel zeolites that are unfeasible through conventional synthetic routes11-14. One of them is the ADOR (assembly-disassembly-organization-reassembly) strategy proposed by Čejka and Morris et al. that produced a series of new zeolites (IPC-n series of zeolites)11. By employing the heating or pressure induced topotactic transformations, Corma et al. and Yu et al. have prepared novel zeolites ITQ-50 and ZEO-34,13, respectively.
Despite extensive efforts to understand zeolite formation mechanisms, knowledge has primarily been limited to the nanometre level39-43. The direct visualization of atomic-scale structural transformations during the topotactic reactions remains elusive, impeding a comprehensive understanding of transformation mechanisms of zeolite topotactic reactions.
In this study, we present the synthesis of the two extra-large pore silicate zeolites, ECNU-45 and ECNU-46, and their structure determinations using 3D ED. Both ECNU-45 and ECNU-46 contain 24-ring channels defined by 24 TO4 tetrahedra (T = Si). They represent the first examples of pure silica zeolites with pores ≥ 24-ring. This is particularly noteworthy, as only three reported zeolites ITQ-37 (30-ring, -ITV), ITQ-43 (28-ring, -IRT), and SYSU-3 (24-ring, -SYT) are known to have pores ≥ 24-rings, and all of them are Ge-rich silicogermanates, thermally and hydrothermally unstable44-46. ECNU-46 was derived from ECNU-45 by topotactic transformation. While ECNU-45 features a 3D 24 × 10 × 10-ring channel system, ECNU-46 exhibits a 1D 24-ring channel connected to 10-ring pockets. Importantly, we demonstrate for the first time that time-resolved 3D ED can capture topotactic structural transformations at the atomic scale, from the starting phase, through various reaction intermediates at distinct time points, to the final product. These detailed structure evolutions throughout the entire topotactic reaction provide new insights into the atomic-scale mechanisms governing the zeolite transformations.
ECNU-45 was hydrothermally synthesized using 1,1,6,6-tetramethyl-1,6-diazacyclododecane-1,6-diium hydroxide as the organic structure-directing agent (OSDA; Fig. 1 and Supplementary Fig. 1; see Supplementary materials), and used as the starting phase. The final product ECNU-46 was prepared via topotactic reactions of ECNU-45 in an acidic solution of HCl/EtOH/H2O (1 M) at 190 ℃ for 24 h (Fig. 1). While the framework of ECNU-46 was stable after calcination at 550 °C in air, that of ECNU-45 collapsed after calcination where all OSDAs were removed (Supplementary Fig. 2).
Structure determination of ECNU-45 and ECNU-46
Because of the small crystal size of ECNU-45 and ECNU-46, 3D ED was applied to determine their crystal structures. Both structures have similar unit cell and the same space group P-62c (Supplementary Fig. 3). The unit cell parameters were further refined by Pawley fit from the PXRD data to be a = 19.9616 and c = 13.8771 Å for ECNU-45 and a = 19.7848 and c = 14.1435 Å for ECNU-46 (Supplementary Fig. 4). While the low angle PXRD profiles are similar for ECNU-45 and ECNU-46, the high angle (2θ > 15°) peak intensities are considerably different (Fig. 1). This indicates that their framework structures are different despite the same space group and similar unit cells. The unit cell parameters of ECNU-45 and ECNU-46 are also similar to those of zeolite EMM-23 (a = 19.7480 and c = 13.8585 Å, framework type: -EWT)47, which has a different space group P31c.
The structures of ECNU-45 and ECNU-46 were solved and refined using 3D ED data, which were found to be highly related but distinct in topology (Fig. 2, Supplementary Fig. 5, and Supplementary Table 1). While ECNU-45 has a 3D intersecting channel system defined by 24´10´10-rings (Figs. 2a and j), ECNU-46 has a 1D 24-ring channel system connected to 10-ring pockets (Figs. 2b and k). The frameworks of both ECNU-45 and ECNU-46 are constructed by columnar building units along the c-axis, as indicated in Figs. 2d and e. Each 24-ring channel is built from six such columns, and each column is associated to three 24-ring channels (Supplementary Figs. 6a and b). The column (#1) in ECNU-45 contains 10-ring pores (Fig. 2d and Supplementary Fig. 6c), which connect the 24-ring channels to form a 3D interconnected 24 × 10 × 10-ring channel system. Both ECNU-45 and ECNU-46 have eight symmetry-independent TO4 tetrahedral sites; six of them (T1 to T6) have similar connectivity and positions (differ by < 0.40 Å for T1-T5 and 0.68 Å for T6). The remaining two sites (denoted T7 and T8 in ECNU-45 and T10 and T12 in ECNU-46) are different (Figs. 2a and b; Supplementary Figs. 6a and b; Supplementary cif files). Although their positions are very different (differ by 1.42 Å), T7 in ECNU-45 and T10 in ECNU-46 share the same connectivity to three T sites (T3, T5, T6) (Figs. 2a and b; Supplementary cif files). However, the fourth connectivity is different; T7 is connected to T8, while T10 is connected to T12. T8 is partially occupied with a very low occupancy (0.27) and connects to two T7, while T12 is fully occupied and connects to three T7 and one T12 (columns #1 and #2 in Fig. 2; Supplementary Figs. 6c and d). The new T10-T12 and T12-T12 connections block the 10-ring channels and result in a 1D 24-ring channel system in ECNU-46.
The framework of ECNU-45 is also similar to that of EMM-2347, both have 3D intersecting channels, defined by 24 ´ 10 ´ 10-rings and 21 ´ 10 ´ 10-rings (Figs. 2a and c), respectively. All T-sites (T1 to T8) in ECNU-45 also exist in EMM-23 (column#3, Figs. 2d and f). The difference is that EMM-23 has a higher occupancy (0.67) occupancy at the T8 site than ECNU-45 (0.27). In addition, EMM-23 has an additional T site (T9, occupancy 0.33) (Fig. 2f)47. The refined framework composition per unit cell is [Si60O109(OH)22] for ECNU-45 and [Si62O118(OH)12] for ECNU-46, compared to [Si64O116(OH)24] for EMM-23 reported early47. The free diameters of the 24-rings in ECNU-45 and ECNU-46 are very similar, 12.6 × 8.2 Å and 12.3 × 8.6 Å, respectively (Figs. 2g and h). In comparison, the 21-ring pore opening in EMM-23 is much smaller (10.1 × 2.7 Å), which has a trilobe-shape resembling three fused 10-ring channels (Fig. 2i). The free diameters of 10-ring channels in ECNU-45 and EMM-23 are similar, 5.8 × 4.6 Å and 5.7 × 5.1 Å, respectively.
Structure evolution by time-resolved 3D ED and insight into the reaction mechanism
The topotactic transformation from a 3D 24 × 10 × 10-ring channel system in ECNU-45 to a 1D 24-ring channel system in ECNU-46 involves displacement, addition and removal of framework atoms, bond formation and bond breakage in the framework. It is therefore important to follow the structure evolution at atomic scale throughout the entire topotactic transformations to understand the reaction mechanism. To achieve this, we investigated the reaction intermediates prepared in an acidic medium (HCl/EtOH/H2O, 1 M) at 190 ºC at different treatment time points, namely 0, 1, 2, 4, 6, 8, 10, and 24 h by combining SEM, PXRD, solid-state 29Si MAS NMR and thermogravimetric analysis (TGA). More importantly, we applied time-resolved 3D ED to follow the detailed structural changes at atomic scale.
SEM shows the crystal sizes and morphology remained the same during the topotactic reaction (Supplementary Fig. 7). PXRD indicates considerable structural changes of the reaction intermediates, as shown by the changes of the peak positions and intensities (Fig. 3a; Supplementary Fig. 8a). Pawley fitting of the PXRD patterns shows some minor contraction (0-4 h) followed by expansion (4-8 h) of the unit cell volume (Supplementary Fig. 9). In order to gain a deeper insight into the topotactic reactions and transformations, we further collected 3D ED data on the reaction product obtained at each time point and determined the 3D atomic structure of the intermediate (Fig. 3b-i; Supplementary Table 1). The high 3D ED data quality allowed us to locate partially occupied atomic sites from the difference Fourier maps and determine their occupancies. The refinements converged to reasonable R1-values from 0.1226 to 0.1732 for , and Si-O bond distances (1.55 to 1.66 Å, on average 1.59 Å) and O-Si-O angles (102.7 to 117.0°, on average 109.5°) (Supplementary Table 1).
We compared the atomic structures of the starting material (0 h), reaction intermediates (1-10 h) and the final product (24 h), which provided several new insights of topotactic reactions (Figs. 3b-I; Supplementary videos 1 and 2). We found the reactions involved six topologically independent sites (T7-T12, Fig. 3 and Supplementary Fig. 10). All these T sites are two- or three-connected (Q2 and Q3) and their occupancies changed throughout the reactions. The major changes include shift of Q3 species from T7 to T10, addition of new framework atoms at T9 and T11, shift of Q2 species from T11 to T12, and finally removal of Q2 species at T8 and T9. Consequently, a number of new bonds were formed (in bold, T2-O-T9, T6-O-T9, T8-O-T9, T10-O-T11, T10-O-T12). The Si species at T9 and T11 had very low occupancies (< 23%) and were only observed in the intermediate structures (1-8 h). Except for T10-O-T12, all other new bonds broke and the Si species at T8 and T9 completely dissolved in the later stages of the reactions (6-10 h). The resulting framework ECNU-46 has only two Q3 species (at T3 and T6) compared to that of ECNU-45, which has three Q3 species (at T3, T6 and T7).
Notably, the OSDAs located in the pores of ECNU-45 were removed during the topotactic reactions. TGA revealed a stepwise removal of the OSDAs (Supplementary Fig. 9a), 43% during 0 - 1 h, 11% during 1 - 4 h, followed by the complete removal (34%) of the OSDAs during 4 - 6 h. Although it was challenging to obtain atomic positions of the OSDAs from the 3D ED data, the locations of the molecules could be identified from the difference Fourier maps of the reaction intermediates (Supplementary Fig. 12). We found that the OSDAs were initially located at the intersections of the 10-ring and 24-ring channels (0 h). They remained at the same locations despite 43 % of the OSDAs being removed (1 h). Most of the remaining OSDAs migrated to the center of the extra-large channels (2 - 4 h). No strong extra peaks were observed in the difference Fourier map of the reaction intermediate at 6 h, indicating the absence of the OSDAs. The results obtained from 3D ED data agree with the TGA results and also with the modelling of the OSDA positions (Supplementary Fig. 13).
The structural information of the intermediates obtained from the 3D ED data enables new insights into reaction mechanisms of the topotactic transformation from ECNU-45 to ECNU-46 under the acidic treatment (Fig. 4). We anticipate the interplay between the stepwise removal of the OSDAs, the incorporation of new framework species and subsequent removal of those species played a pivotal role and ensured the framework stability throughout the topotactic reactions. Acid treatment expeditiously eliminates approximately half of the OSDAs from the channels, which facilitates the shift of some Q3 species from T7 to T10 and subsequently incorporation of additional framework species at T9 and T11 to bind and stabilize the T7 and T10 sites, respectively. Meanwhile, the remaining OSDAs predominantly migrate towards the center of the extra-large channels, in proximity to the Q2 species at T8 and T9, thus maintaining good structural stability for the later topotactic reaction. The migration of OSDAs would also facilitate the shift of more Q3 species from T7 to T10, allowing the shift of Q2 species at T11 to T12 to form Q4 species (Supplementary Fig. 10a). This increases the stability of the framework, allowing for further removal of the OSDAs. After the complete removal of the OSDAs (6 h), most Q3 species at T7 shifted to T10 and reacted with Si(OH)4 species at T12 until both T10 and T12 were fully occupied. The unstable Q2 species at T8 and T9 were gradually dissolved. These sequential events led to the transformation of ECNU-45 into ECNU-46 through topotactic reactions. This is in contrast to direct calcination where the framework of ECNU-45 collapsed upon the removal of the OSDAs (Supplementary Fig. 2).
To our knowledge, it is the first time that time-resolved 3D ED has been applied to the study of a highly complex topotactic transformation at the atomic scale. It enabled the visualization of detailed structural changes including atom displacement, addition, and removal, as well as bond forming and bond breaking. It also discovered the interplay between the framework atoms and OSDAs, which provides new insights into the formation mechanism of a zeolite and represents significant advancements in zeolite chemistry. We believe time-resolved 3D ED not only opens a new avenue for studying topotactic transformations in diverse crystalline materials at the atomic scale but also holds immersing potential for accelerating the development of novel functional materials.