SSZ‐27: A Small‐Pore Zeolite with Large Heart‐Shaped Cavities Determined by Using Multi‐crystal Electron Diffraction

Abstract The high‐silica zeolite SSZ‐27 was synthesized using one of the isomers of the organic structure‐directing agent that is known to produce the large‐pore zeolite SSZ‐26 (CON). The structure of the as‐synthesized form was solved using multi‐crystal electron diffraction data. Data were collected on eighteen crystals, and to obtain a high‐quality and complete data set for structure refinement, hierarchical cluster analysis was employed to select the data sets most suitable for merging. The framework structure of SSZ‐27 can be described as a combination of two types of cavities, one of which is shaped like a heart. The cavities are connected through shared 8‐ring windows to create straight channels that are linked together in pairs to form a one‐dimensional channel system. Once the framework structure was known, molecular modelling was used to find the best fitting isomer, and this, in turn, was isolated to improve the synthesis conditions for SSZ‐27.


Introduction
Anumber of novel high-silica zeolite structures have been found through exploratory synthesis trials that combine an organic guest molecule,u sually aq uaternary ammonium compound, with av ariety of synthetic conditions.W eh ave recently described how the inorganic conditions chosen in the hydrothermal reaction can bias the crystalline product in terms of the sub-units likely to occur in the zeolite structure. [1] Fore xample,s yntheses can be tuned or modified to produce zeolites with relatively moderate SiO 2 /Al 2 O 3 ratios (SAR) and small pores running multi-dimensionally between cages (where the organic guest is found in the as-synthesized zeolite). This is particularly relevant in the search for materials with high performance in DeNOx reactions in the SCR (selective catalytic reduction) components in diesel emission systems, [2] in which there is high interest in novel small-pore zeolites.High hydroxide levels and the presence of one of anumber of alicyclic ring derivatives produce zeolites yielding an optimum DeNOx catalyst once the counter ions have been exchanged with copper or iron cations.I nt he various zeolite framework types discovered over the last few decades, [3] there is aw ide range of channel and cage architectures to choose from. Adescription of various unique zeolitic features with examples are given in Table S1 (Supporting Information).
Herein we explore how the synthesis of ad iquaternary organic structure-directing agent (SDA) results in different isomeric versions of the SDA, and how asmall difference can lead to the formation of ad ifferent zeolite framework. Figure 1s hows the synthesis path that produces the isomers and their labelling from an earlier work. [4] Initially,wefocused on the production of zeolite SSZ-26 (CON)and showed that it could be made from two well-described isomers (isomers I and III). [4] Thec oncept for developing these isomers was to break the fit of an SDAi nto either 1) ac age accessed by smaller portals,o r2 )a single large pore.Apast study, juxtaposing the size of the SDAa nd the SAR conditions, shows ab reakdown into 5types of zeolite products. [5] At the time,w es ucceeded in creating am aterial with intersecting pores.Infact, this was the first discovery of azeolite (SSZ-26) with both large-and medium-sized pores. [6] During the course of the synthesis explorations for SSZ-26, we occasionally encountered adifferent product. Early microporosity studies on the impurity phase,w hich we termed SSZ-27, indicated that it had only small pores and selective hydrocarbon uptake. At the time,these materials were not of much developmental interest for catalysis or for separations,s of urther characterization of SSZ-27 was not pursued. However,interest in small pore zeolites has recently taken ad ramatic upward turn. We therefore revisited the synthesis of SSZ-27, and show here that the conformation of one of the isolated isomers of the diquaternary SDA(isomer III) is such that it fits particularly well into one of the cavities of SSZ-27. Here we present the reaction conditions and modelling results that led to the synthesis of SSZ-27 and to an understanding of how it forms, and its structure determination from electron diffraction data.

Synthesis
Our point of departure was the synthesis of the diquaternary SDA, shown in Figure 1. Theconcept was to build on the established organic synthesis chemistry of Ginsberg and others [7,8] to build fused molecules called propellanes.I nt his way we could make derivatives that might promote the construction of channels in the zeolite framework that would extend in more than one direction. We could convert the resulting dione to ad i-tertiary amine in one step using the Leuckart reaction. [9] These diamines can then be quaternized using methyl iodide.N ote that in the formation of the diamine,the isomeric possibilities (there are 3atthis step) are fixed. Individual isomers were recovered by selective recrystallization methods and their structures determined using single-crystal X-ray diffraction data. [4] Fort he zeolite synthesis,w eu sed as et of inorganic component ratios consistent with earlier work on high silica zeolites like ZSM-5 (MFI), where ar educed hydroxide to silica ratio was needed to create structures rich in 5-ring subunits.A ni mportant finding proved to be the need to reduce the alkali content while retaining the amount of hydroxide, and this was achieved by replacing the alkali cations with the quaternary ammonium centres.T his change reduces the intrusion of high-silica zeolite competitors such as Mordenite and Magadiite,w hich require sodium in the synthesis. [10] As we were exploring the synthesis conditions for SSZ-26, we occasionally encountered SSZ-27 as ab y-product. SSZ-27 formed well-defined homogenous crystals ( Figure 2). A comparison of the X-ray powder diffraction patterns for SSZ-26 and SSZ-27 is shown in Figure 2. Experimental details of the SDAa nd zeolite synthesis can be found in the Supporting Information.

Structure Determination using Electron Diffraction Data
Preliminary characterizations of SSZ-27 using XRPD revealed several unindexed peaks,w hich were identified as coming from quartz and SSZ-26 (CON)i mpurity phases (Supporting Information). In view of the fact that obtaining apure sample of SSZ-27 was proving to be difficult (more on this below), we opted to collect electron diffraction (ED) data using the continuous rotation method, [11,12] which offers ac ombination of fast acquisition times and high data completeness,b ecause all of reciprocal space is sampled. We used ar ecently developed crystal tracking routine to ensure that the crystal stayed in the electron beam during rotation [13] implemented in the software Instamatic. [14] This makes it easier to collect data with high rotation ranges,a nd at the same time enables more accurate integration of the reflection intensities for structure refinement. Diffraction data were collected on eighteen isolated crystals between 200 and 1000 nm in size over three sessions (Supporting information  Table S2 and Figure S2). Unit cell determination, indexing, and intensity integration were performed using XDS [15] (Supporting information Tables S4 and S5). Thep atterns of fourteen crystals could be indexed with very similar unit cells,  with mean lattice parameters of a = 24.12 (14) , b = 13.81(6) , c = 25.07(10) , b = 115.19(18)8 8 in space group C2=m,matching the unit cell found for SSZ-27 in our preliminary XRPD assessment (Supporting Information). Thec ell parameters of SSZ-27 were optimized using the synchrotron XRPD data and the Pawley [16] profile-fitting routine in TOPA S5, [17] yielding ag ood profile fit with a = 23.29 , b = 13.37 , c = 24.38 , b = 114.238 8.S everal of the fourteen ED data sets could be used for structure determination using the programs FOCUS [18,19] and SHELXT, [20] revealing the framework structure of SSZ-27 (Figure 3a). Ther emaining four crystals had au nit cell matching SSZ-26 (CON). Adetailed characterization of the SSZ-26 crystals can be found in the Supporting Information.
Because the crystal system of SSZ-27 is monoclinic,none of the ED data sets had ac ompleteness of more than 90 %, which is required for agood structure refinement. Therefore, to improve data completeness and redundancy, several data sets were merged. We performed hierarchical cluster analysis (HCA) on the fourteen data sets using the algorithm described by Giordano and co-workers [21] to find the optimal selection of data sets for merging. We have found this approach to be useful for dealing with large multi-crystal ED data sets. [22] Thedistance metric t,which defines the similarity between the data sets,i sd erived from the correlation coefficients of the common reflection intensities (CC I )i n pairs of data sets,and the "average" linkage method is used. Thec lusters can be visualized in ad endrogram (Figure 4a), which facilitates finding as uitable cut distance,a lthough finding the right value involved some trial and error. Finally, acut distance t = 0.32, equivalent to CC I ¼ 0:95,was used. In our experience with HCA, clusters with t < 0.40 usually result in useful data sets.Herein, two clusters were obtained, where the largest cluster,consisting of 10 data sets (shown in red in Figure 4a), comprised the highest data quality (assessed via CC 1=2 [23] )a nd completeness (Supporting Information Table S6). Thes cripts for cluster analysis are available online from https://github.com/stefsmeets/edtools.
Thet en data sets belonging to the largest cluster were merged using XSCALE [15] and structure refinement was initiated using the lattice parameters obtained from the XRPD data, reasoning that these would be more accurate than those obtained from ED data. Indeed, this immediately resulted in more reasonable average SiÀOb ond lengths. Restraints on the SiÀOdistances and O-Si-O angles were not necessary.T he atomic displacement parameters (ADPs) for all framework atoms were refined anisotropically.Inthe final stages of the refinement, the SWAT instruction was introduced to model the diffuse species in the channels (such as, the SDA, residual water, Na + ). This instruction suppresses the contribution from the strong, low-angle reflections where the contribution of extra-framework species is strongest. This significantly reduced the R1v alue,a nd resulted in more meaningful shapes for the ADPs (Figure 3b,c). Ther efinement of the framework structure of SSZ-27 converged with R1 = 0.178, wR2 = 0.486, and S = 1.47 (Figure 4b). Crystallographic details and the geometry of the zeolite framework are summarized in the Supporting information (Tables S6-S8).
Recently,w es howed that physically meaningful anisotropic ADPs can be obtained from ED data, provided these are of sufficiently high quality. [13] ADPs are known to act as a" fudge factor" for poor quality data. Fort his reason, they are often refined isotropically or their physical meaning is ignored. In the case of SSZ-27, the anisotropic refinement of the ADPs was stable without the addition of restraints and their values are physically sensible (Figure 3b,c). They are slightly elongated along the z-direction, which we attribute to lower data redundancyalong this direction. TheADPs of the Oatoms are slightly larger than those for Si, and elongated in the direction perpendicular to the plane formed by the Si-O-Si bonds.T he fact that reliable ADPs can be obtained for submicron-sized crystals from ED data may open up new  correlation between the SSZ-27 data sets (crystals S1-S14) corresponding to the data shown in Table S4 (Supporting Information). The cut distance set at 0.32 is represented by the horizontall ine. Twoclusters (green, red) can be identified.b ) ffiffiffiffiffiffi ffi I obs p vs. ffiffiffiffiffiffiffi I calc p plot for the refinementofSSZ-27. The refinement details can be found in Table S7 (Supporting Information).
possibilities for studying the atomic vibrations or static disorder in zeolites that do not grow large enough for X-ray single-crystal analysis.F inally,w ea ttempted to find the position of the SDAf rom the ED data. To do so,t he SWAT instruction was removed, and ad ifference map generated. Although the difference map clearly revealed two large clouds of residual electrostatic potential (Supporting Information Figure S3), refinement of the SDAinthe heart-shaped cavity was not stable,sowehad to conclude that the data do not support refinement of the SDAatthis stage (see also the Supporting information).

Framework Structure
Thes tructure analysis confirmed that SSZ-27 is indeed as mall pore zeolite.A sm entioned earlier, the data also revealed the presence of as mall, but noticeable amount of SSZ-26. Thef ramework structure of SSZ-27 is characterized by two types of cavities (Figure 5a,b). Thel arger,h eartshaped cavity consists of 50 T-atoms ([8 2 6 10 5 10 4 6 ]), and has two 8-ring windows with free dimensions of 3.6 5.5 (Supporting Information Figure S4), which are shared with as maller cavity of 42 T-atoms ([8 4 6 8 4 6 ]). Each 42T cavity,delimited by a1 4-ring perpendicular to the z-axis,c onnects four heartshaped cavities together. Figure 6shows the orientation of the heart-shaped cavities with respect to the 42T cavities and the layer they form in the xz-plane.E ach layer is shifted by ( 1 = 2 , 1 = 2 ,0 )w ith respect to the adjacent one.S SZ-27 has ao nedimensional channel system. Ap air of linked straight 8-ring channels runs along the y-axis,alternating between single 42T and two heart-shaped cavities.T he heart-shaped cavities can be viewed as large side-pockets.T his means that despite this being as mall-pore zeolite,t he internal volume is relatively large (Figure 3).
To our knowledge,t he EEI [25,26] framework type is the only one besides that of SSZ-27 with small pores,aonedimensional channel system, and large side pockets.I no ur previous experiments,S SZ-45 (EEI)s howed as urprisingly low N 2 microporosity (0.056 cm 3 g À1 ), [26] which appeared to be limited by the fact that the connection between the cavities involves two 8-rings (i.e.f orms as hort channel) and is relatively rigid. In contrast, the micropore volume of 0.14 cm 3 g À1 found for SSZ-27 is much more in line with what is expected for azeolite (Supporting Information Figure S1). We suspect that the pore configuration consisting of only asingle 8-ring window connecting the large cavities allows for much more efficient diffusion.

Molecular Modelling and the Location of the SDA
Them olecular modelling results show that the SDAi s located exclusively in the heart-shaped cavity,w hich fits around the SDA( isomer III) like ag love (Figure 7). At am aximum occupancyo f0 .5 on ap osition of mirror symmetry,there is one SDA( with two possible orientations) per heart-shaped cavity,orfour (8 N + )per unit cell. Based on elemental analyses (Supporting Information Table S12), the Si:Al ratio of the framework was estimated to be 14.5, which translates to 7.5 Al per unit cell, so the charge from the SDA balances that of the framework nicely.T his would indicate that there are no additional organic cations in the 42T cavity, where the second cloud of residual electrostatic potential was found, and that is consistent with the fact that the cavity is too small to accommodate an SDAcation. On the other hand, the elemental analysis also indicates the presence of as mall amount of residual Na + (approximately 1.0 per cell), which  could be located in the 42T cavity with an occupancyo f0 .5. This may indicate that the SDAi sn ot fully occupied, which we have observed before. [27][28][29] Ther est of the 42T cavity is probably filled with residual water.
It has been observed experimentally, [30] and predicted computationally [31] that the SDAcan direct the location of Al in aluminosilicates.W eh ave previously observed this for borosilicates,i nw hich Batoms are preferentially located in close proximity to the N + centre in the SDA. [27,31] In SSZ-27, the SDAappears to be located exclusively in the heart-shaped cavity,which suggests that the Al is likely to be at the surface of this cavity too.W ehypothesize that this means that all the active sites are accessible in the side pockets created by the heart-shaped cavities,whereas the main straight channels are relatively free of active sites,a nd this might reduce the potential for blockage.

NMR Spectroscopy
Figure 8shows the 13 CMAS NMR spectra for the as-made products in runs to make SSZ-26 (with isomer I) and SSZ-27 (with isomer III). Also shown are the spectra for the two separate isomers (I and III) in the crystalline solid state and in solution. There are clear differences in the 13 CM AS NMR spectra of the two isomers,but in both cases the lines become broader and less distinctive when they are entrapped in the SSZ-26 and SSZ-27 framework structures.N ote that the 13 C resonances in the crystalline solid are more complicated than those in solution phase, [4] especially for peaks in the 35-45 ppm region for isomer Ia nd the 45-65 ppm region for isomer III. Thechange may be attributed to the generation of less symmetric crystalline stacking of molecules during solidification. Therefore,t he 13 CNMR spectra of the two zeolites do not provide ac lear view of the structural difference between the isomers.H owever, 13 Cc ross-polarization MAS (CPMAS) spectra in Figure 8c differentiate resonances of methylene carbons in the 5-membered rings (30-45 ppm) as well as the methyl groups.
To unambiguously show which isomer was occluded inside the two frameworks we dissolved the crystalline samples of SSZ-26 and SSZ-27 with HF and extracted the organic material for 1 HNMR analysis.B esides as light shift because of the acidic condition after dissolution with HF,the 1 HNMR spectra (Figure 9) clearly demonstrate that the isomers recovered from SSZ-26 and SSZ-27 can be identified as I and III, respectively.T his proves that both samples contain only as ingle isomer,a nd that no interconversion occurred during zeolite synthesis and crystallization.
On the other hand, the ED structure analysis clearly showed that SSZ-26 is am inor component in the SSZ-27 sample studied. Additional modelling studies (Supporting information Table S13 and Figure S9) show that while isomer Is ignificantly favours the formation of SSZ-26   (À7.6 kJ mol À1 Si)over SSZ-27 (À3.66 kJ mol À1 Si), isomer III only favors SSZ-27 slightly (À8.95 kJ mol À1 Si)o ver SSZ-26 (À8.18 kJ mol À1 Si). Then it is perhaps not surprising that some SSZ-26 formed in the presence of isomer III, albeit more slowly than SSZ-27. Of course,t he chemical environment of isomer III in SSZ-26 will be slightly different from that in SSZ-27 and this will contribute to the breadth of the 13 CNMR signals,m aking ad efinitive interpretation of the NMR spectra difficult. Additional NMR studies using nuclear Overhauser effect spectroscopy (NOESY) to better probe the details of the SDAwithin the zeolite are planned.

Additional Synthesis Studies
In an attempt to confirm the indications of the energy calculations that isomer Ifavors the formation of SSZ-26 and isomer III SSZ-27 experimentally,w ep erformed as et of synthesis experiments where the two isomers (I and III) were used individually with and without seeding ( Table 1). It can be seen that in the case of isomer III, which shows the favourable space-filling (Figure 7), SSZ-27 crystallizes.T he formation is even faster if there is seeding with SSZ-27. In contrast, the other two experiments yield no SSZ-27 and eventually form SSZ-26 after amuch longer run period (4-5 times longer). In the process of determining the crystallization requisites for reproducing the synthesis of SSZ-27, we found that higher temperatures (170 8 8C) favoured SSZ-27 over SSZ-26. The latter is typically made at 160 8 8C. While the zeolite synthesis literature does describe some examples of crystallization of relatively open structures at higher temperatures,the general trend in crystallization is to produce open, metastable structures at short reaction times and lower temperatures and then zeolites with progressively higher framework density (sometimes through structure transformation [33] )a tl onger reaction times and higher temperatures.T he void volumes in zeolite structures have an inverse relationship to framework density,a nd the framework structure of SSZ-26, which crystallizes at the lower temperature,i si nf act more open than that of SSZ-27.
Returning to the theme advanced in the introduction, alicyclic derivatized ring structures provide good guest molecules for the formation of cage-based zeolites under the right synthetic conditions.The zeolite SSZ-13 (CHA)was first discovered with ad erivative of the polycyclich ydrocarbon, adamantane. [34] Similarly,avery large number of piperidine ring derivatives will make SSZ-35 (STF). [35] The underlying aim in the synthesis was to produce guest molecules that have axes long enough to extend along the channels of az eolite.T his is true for the anti,anti isomeric derivative (isomer I) and, in fact, SSZ-26 is the only zeolite made with this SDA. However,o nce both trimethyl ammonium groups have syn,syn configurations (isomer III), it is as though the molecule is beginning to fold in on itself to produce more of acage structure.T his was well-illustrated in our previous work to produce at riquaternary imidazolium derivative off acentral benzene ring, where the imidazolium charged groups folded in on themselves to stabilize the sizable cage of the small pore zeolite LTA. [36] Conclusion Synthesis highlights in this work are that subtle changes in the rigid organo-cation (SDA) stereochemistry can lead to entirely different zeolite (host lattice) structures as the framework develops around the guest molecule.I nt urn, as we continue to increase our database of known zeolite structures,and understand the spatial orientation of the guest SDAm olecules in the lattice,o ur ability to create new pairings improves.T hat is,w ec an predict what type of host environment can accommodate the guest and in which orientation. Molecular modelling of the type seen herein can be used effectively to check on the feasibility of the orientation of the SDAi naprospective lattice.I na ddition, our understanding of how the inorganic conditions of azeolite synthesis bias the potential for specific sub-units in the developing structure provide another parameter in model building for creating space for ag uest. Ther esurgence of discoveries related to ABC-6 zeolites,s mall-pore products with larger cages that host the organo-cations,a re ag ood example.T he SSZ-27 zeolite,r ich in 5-and 6-ring subunits (which typically requires ah igher Si/Al synthesis regime) is found to have ag ood spatial consistencyw ith one particular SDAi somer,a nd then yields ah ighly unusual framework structure.
At the same time,d evelopments in electron diffraction methodology,b oth hardware and software,h ave reached apoint where high-quality data can be collected routinely on al arge number of crystals.A st he XRPD data were insufficient for structure refinement, ED data were collected on eighteen crystals,f our of which belonged to an SSZ-26 impurity phase,c orroborating our conclusions from the molecular modelling studies that SSZ-26 forms equally well using all three isomers.T he other fourteen crystals belonged to the SSZ-27 phase.K ey to this study is the application of hierarchical cluster analysis that leads to an optimal selection of data sets for structure refinement. This,i nt urn, helps to obtain more reliable and precise atomic coordinates of the framework, but also to get ah andle on the atomic displacements of the atoms.T he fact that physically meaningful atomic displacement parameters can be obtained from submicron-sized crystals opens up new possibilities for studying atomic motion (e.g.i nternal vibrations) and disorder (static or dynamic). Thef ramework structure that emerged for SSZ-27 is that of as mall-pore zeolite with ao ne-dimensional channel system connecting heart-shaped cavities with smaller cavities through 8-rings.T hrough molecular modelling,w ew ere able to establish that the SDAi sl ocated exclusively in the heart-shaped cavity.I ti sa ssumed that the other cavity contains water.