Direct structural identification of carbenium ions and investigation of host–guest interaction in the methanol to olefins reaction obtained by multinuclear NMR correlations† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c7sc03657d

Structural identification of carbenium ion intermediates and quantitative investigation of their interactions with zeolite H-ZSM-5 by multinuclear MAS NMR.


Introduction
The methanol-to-olens (MTO) reaction is an important process for the production of light olens (mainly ethylene and propene) from non-petrochemical resources such as coal and natural gas. 1 The reaction is catalysed by microporous solid acids, in particular a wide range of zeolites (e.g. H-ZSM-5 and H-SAPO-34), and has been successfully commercialised since the 1990s. [1][2][3] Nevertheless, there is a need for a deeper understanding of the catalytic active sites and reaction mechanism in order to identify catalyst deactivation pathways and further optimise the catalytic performance. [4][5][6][7][8][9][10][11] Solid-state NMR is a well-developed technique for structural determination and host-guest investigation studies, and has played an important role in increasing our understanding of heterogeneous catalytic processes. [12][13][14][15] In particular, solid-state NMR has enabled the detection of the intermediates formed during the MTO reaction and their interactions with zeolite, which are both important in unravelling the mechanism of the reaction. The intermediates that have been observed previously by NMR are cyclic carbenium ions such as polymethylcyclopentenyl cations and polymethylbenzenium cations on H-SAPO-34, 16 H-SSZ-13, 16 DNL-6 (ref. 17) and b-zeolite, 18 and polymethylcyclopentenyl cations and ethylated cyclopentenyl cations on H-ZSM-5 zeolite. 19,20 These cyclic carbenium ions are crucial intermediates involved in the hydrocarbon pool mechanism 21 in which the cyclic organic species in the zeolite pores act as co-catalysts for the conversion of methanol to olens. 5 More specically, two reaction routes have been proposed namely the side-chain methylation route in which the olens are produced through the methylation of polymethylbenzenium ions and the subsequent elimination of the side chain groups, and the paring route in which the olens are released via the expansion of polymethylcyclopentenyl cations and the subsequent contraction of polymethylbenzenium ions. 16,[22][23][24] The structural identication of the carbenium ions is crucial for the determination of the dominant route and specic reaction path in different zeolites.
Previous work relied on computational methods for the assignments of the 13 C solid-state NMR spectra of the carbenium ions. 16,19 Although these computational methods are recognised as robust approaches for NMR spectral interpretation, 25,26 there is still a direct lack of experimental data supporting these assignments and therefore this may lead to possible misinterpretation of the carbenium ions produced. Additionally, the identication of the carbenium ions was indirectly obtained by digesting the dienes (the deprotonated counterparts of the cyclic carbenium ions) with concentrated sulfuric acid and analysing the obtained solutions by liquidstate NMR. 16,20 This procedure assumes that the states of the carbenium ions in the solutions are the same as those conned in the solid zeolite pores, and also requires independent synthesis of the dienes, and therefore entails prior knowledge of the possible carbenium ions' structures.
Recently, the interactions between the carbon species and the Al sites of the H-ZSM-5 zeolite were qualitatively investigated using spatially encoded 13 C- 27 Al dipolar coupling NMR experiments (employing a S-RESPDOR, Symmetry-based Resonance-Echo Saturation-Pulse DOuble-Resonance sequence 27 ). The work demonstrated the formation of supramolecular reaction centres composed of conned carbon species and the inorganic framework of zeolite which possesses higher reactivity toward methanol in the H-ZSM-5 zeolite. 7 Here, we unambiguously experimentally identied several cyclic carbenium ions on MTO activated H-ZSM-5 zeolite, including a previously undetected 1,5-dimethyl-3-sec-butyl cyclopentenyl cation, using a refocused INADEQUATE (Incredible Natural Abundance DoublE QUAntum Transfer Experiment) 28 NMR sequence. This experiment relies on scalar J couplings and yields through-bond correlations, providing a straightforward pathway for 13 C spectral assignments. Moreover, the interactions between the conned carbon species and the H-ZSM-5 zeolite framework are quantitatively probed via through-space 13

Results and discussion
Structural identication of the conned carbon species in MTO activated H-ZSM-5 In this work, the MTO activated H-ZSM-5 materials were prepared by passing 13 C enriched CH 3 OH over H-ZSM-5 at 285 C for 20 minutes followed by quenching the reaction mixture with liquid N 2 (see the Experimental section in the ESI †). The 13 C CP (Cross-Polarisation) MAS (Magic Angle Spinning) NMR spectra of the 13 C enriched MTO activated H-ZSM-5 are given in Fig. 1a at 9.4 T (and in Fig. S2 † at 20 T) and show multiple signals ranging from 0 to 260 ppm, highlighting the complexity of the conned carbon species. It is worth pointing out that the unusual downeld 13 C resonances (235-260 ppm) are characteristic signals of protonated carbenium ions and these have previously been assigned to several methylated and ethylated cyclopentenyl cations 19,20 whose overlapping resonances prevent their unequivocal assignments.
A 2D 13 C-13 C refocused INADEQUATE spectrum is displayed in Fig. 1a and gives correlations mapping out the carbon skeleton of each of the carbenium ions. In this J-based experiment, two directly bonded 13 C nuclei share a common frequency in the double quantum (vertical) dimension at the sum of their 13 C individual frequencies in the single quantum (horizontal) dimension. 28 The peaks observed in the INADEQUATE spectrum notably allowed us to explicitly identify the three methylated carbenium ions, namely the dimethylcyclopentenyl cation I, trimethylcyclopentenyl cation II and pentamethylbenzenium cation III (see Fig. 1b) which have been previously proposed but were identied based on a combination of 1D 13 C CP NMR spectra, GC-MS (Gas Chromatography-Mass Spectrometry) and DFT (Density Functional Theory) calculations. 19 More explicitly, the dimethylcyclopentenyl cation I can be identied through correlations C1(I) (249 ppm) -C2(I) (147 ppm), C1(I) (249 ppm) -C3(I) (48 ppm) and C1(I) (249 ppm) -C4(I) (25 ppm) as identi-ed in purple in Fig. 1a (and in Fig. 1c for all horizontal traces). A similar approach is used to directly establish the carbon connectivities in cations II (Fig. S3 †) and III (Fig. S4 †).
A previously unprecedented observed 1,5-dimethyl-3-secbutyl cyclopentenyl cation IV, is also identied on H-ZSM-5 as revealed by the C3(IV) (255 ppm) -C8(IV) (35 ppm), C8(IV) (35 ppm) -C9(IV) (9 ppm), C8(IV) (35 ppm) -C10(IV) (23 ppm), C10(IV) (23 ppm) -C11(IV) (13 ppm) correlations and correlations amongst C1(IV) to C8(IV) observed in the 2D 13 C-13 C refocused INADEQUATE spectra (red lines in Fig. 1a and S5 †). A cyclopentenyl cation with a tert-butyl group has previously been reported to be involved as a key intermediate in the aromaticbased paring route proposed by theoretical modelling for the formation of isobutene in the MTO reaction, however, it was not experimentally observed. 24 In contrast, some work proposed that butenes are formed through an alkene-based cycle involving the methylation/cracking of alkenes. 31,32 Hence, the experimental identication of cation IV provides direct support for the aromatic-based paring route for butene formation in H-ZSM-5. The elimination of the sec-butyl group of cation IV is likely to produce but-1-ene and but-2-ene which are also the products of the MTO reaction. 33 Several correlations relating to 13 C signals in the 235 ppm to 260 ppm region (black lines in Fig. S6 †) are also obtained and can be assigned to some additional carbenium ions, most likely cyclopentenyl cations. 19,20 The correlation involving the weak signal at 225 ppm may arise from the polymethylcyclohexenyl cations. 34 However, the lack of correlations relating these signals with the aliphatic region of the 13 C NMR spectrum limits the complete assignments of these signals and is probably due to the low concentration of these carbenium ions as evidenced by their weak signal intensities in the 1D 13 C CP MAS spectrum. The low concentration may also account for the apparent absence of ethylated cyclopentenyl cations 20 in our MTO activated H-ZSM-5.
The increase in resolution offered in the vertical dimension of the 2D spectrum enables a more accurate determination of the 13 C chemical shi values of the different carbon sites from these carbenium ions (Table S4 †). Those signals are usually poorly resolved in the 1D CP MAS NMR spectrum (Fig. 1a and Fig. S2 † for data at 9.4 T) even at a high magnetic eld (see Fig. S2 †).
In the 2D 13 C-13 C refocused INADEQUATE spectrum, signals ranging from 120 to 140 ppm show strong correlations with each other and with signals in the 13-22 ppm region (maroon lines in Fig. S6 †). These correlations are attributed to correlations amongst carbons of the benzene rings and between benzene ring carbons and alkyl group carbons of neutral aromatic species, respectively. Aromatics with various types and numbers of substituted alkyl groups have very close chemical shis, 35 which makes individual assignments of signals from these aromatics challenging due to a lack of resolution.
Signals at 60 and 51 ppm have strong intensities in the 1D CP MAS spectrum and show no correlations with any other signals in the 2D spectrum (Fig. 1a). This observation is consistent with their assignments to dimethyl ether and residual adsorbed methanol, respectively. 20 Correlations among peaks between 10 and 45 ppm (orange lines in Fig. S6 †) can be assigned to alkyl groups of aromatics or carbenium ions. 20

Quantitative investigation of the interactions between the conned carbon species and the H-ZSM-5 framework
The interactions between the conned carbon species and zeolite are initially investigated by recording 13 C{ 27 Al} S-RESPDOR data, in which the 13 C-27 Al dipolar couplings are reintroduced by the SR4 2 1 recoupling sequences 36 on the 13 C spins, and provide access to the intramolecular distance. With 27 Al irradiation and at a recoupling time of 15 ms, the S-RESPDOR dephasing DS/S 0 obtained at 9.4 T can be clearly observed, indicating spatial proximities between the 13 C and 27 Al spins. More specically, signals from 0 to 40 ppm, corresponding to the alkyl groups of both carbenium ions and aromatics, all have similar signal reduction DS/S 0 due to being coupled to the 27 Al spins (Fig. 2a) and are integrated against the recoupling time in Fig. 2b. Due to the small concentration of 27 Al atoms in ZSM-5 (SiO 2 /Al 2 O 3 ¼ 50), the 13 C spins are unlikely to be coupled with multiple 27 Al spins and a single spin pair model is used to t the S-RESPDOR data 27 (see ESI † for further  Table S4 † for details) and weak intensities of these 13 C signals to make them more easily visible. Signals corresponding to carbenium ions (black) and to other neutral carbon species (blue) are highlighted to distinguish them. The assignments of the different carbenium species are given in different colours. Asterisks (*) denote spinning sidebands.  Fig. 3a and S8 †) show multiple broad signals at À102, À106, À112 and À117 ppm which are characteristic of the (SiO) 3 SiOH (Q 3 ), Si(OSi) 3 (OAl), Si(OSi) 4 (Q 4 ) and the crystallographically inequivalent Si(OSi) 4 (Q 40 ) sites, respectively, of which the Si(OSi) 3 (OAl) sites contribute to the Brønsted acid sites. 37 Note that no 29 Si signal for the T n sites of the type R-Si(OSi) n (OH) 3Àn (typically observed around À60 ppm (ref. 38)) could be detected on H-ZSM-5, indicating that the conned carbon species are not directly covalently bonded to the 29 Si nuclei.
Spatial interactions between the conned carbon species and the H-ZSM-5 zeolite framework are further probed by 29 Si detected 29 Si{ 13 C} REDOR experiments (Fig. 3a) which reintroduce the 29 Si-13 C dipolar couplings under MAS. 30 At an evolution time of 6 ms, the intensity of the dephased 29 Si NMR signal S 0 is signicantly reduced vs. the spin echo signal S 0 , indicating spatial proximities between 29 Si and 13 C spins. Fig. S8 † shows that the different 29 Si signals have a similar degree of intensity reduction and these signals cannot be well resolved even at a high eld of 20 T (Fig. S9 †). Therefore, integration of the whole 29 Si NMR signals from À90 to À125 ppm was used to determine the REDOR fraction DS/S 0 as a function of the evolution time (Fig. 3b). The number of retained carbon species and their unknown geometries with respect to the zeolite impose the use of a geometrically-independent REDOR curve model in which only data for a short dipolar evolution time (DS/S 0 < 0.3) are needed. 39 Fitting the 29 Si{ 13 C} REDOR data (see Fig. 3b and ESI †) yields P D i 2 of 6650 AE 1600 Hz 2 which gives an estimated 29 Si-13 C dipolar coupling constant D of 82 AE 10 Hz (assuming a simpli-ed single spin pair model) and a 29 Si to 13 C internuclear  distance of 4.2 AE 0.2Å. This distance is comparable to the one obtained above for 13 C to 27 Al from the 13 C{ 27 Al} S-RESPDOR experiments, showing strong interactions between the conned hydrocarbon species and zeolite framework and providing quantitative information for the proposed supramolecular reaction centres in H- 40 The interactions between the neutral aromatics, carbenium ions and H-ZSM-5 have previously been investigated computationally. [41][42][43][44][45] It was found that it is the connement of pores via long-range van der Waals interactions between the neutral aromatics and zeolite framework that contributes considerably more to the aromatics' adsorption in H-ZSM-5 than the shortrange interactions between the acid OH group of the zeolite and the electrons of the aromatic ring. These previous works proposed that the aromatics prefer to adsorb in the intersection region between the straight and sinusoidal channels in which polycyclic aromatics grow and block the channels, leading to the catalysts' deactivation. [41][42][43] The 29 Si{ 13 C} REDOR spectra (Fig. S8 †) show that different 29 Si sites, including the Si(OSi) 3 (OAl) sites corresponding to the Brønsted acid sites, have apparent similar interactions with 13 C nuclei, which indicates that the connement effects dominate the adsorption of the main hydrocarbon species (neutral aromatics), and that the shortrange interactions between the main hydrocarbon species and the Brønsted acid sites may not be strong enough to make a signicant difference between these acid sites (and others). These observations suggest that the deactivation of zeolite may not result from the direct poisoning of acid sites, but from blockage of the channels due to the accumulation of aromatics, which is consistent with previous calculations. 41 The adsorption model in previous studies showed that the acid O-H bond axis faces the aromatic ring in a nearly perpendicular orientation with the distance between the acidic H and the aromatic ring falling in the 2.2-2.9Å range. 41 Considering an Al-H distance of about 2.4Å (ref. 43) and the size of aromatics, both 27 Al-13 C and 29 Si-13 C distances around 4-5Å can be expected, matching the values measured above.
Carbenium ions were previously proposed to form ion-pair complexes with the Brønsted acid sites. In the DFT optimised geometry of this complex, the 13 C nucleus directly involved in the ionic bonding interaction is around 3.1Å away from the O of the Brønsted acid sites. Considering the Al-O and Si-O bond distances (1.7 and 1.6Å respectively) 44,45 and the local geometry of the complex, distances between this 13 C nucleus and 27 Al/ 29 Si can be estimated to be around 4Å. This 13 C nucleus is on the carbenium ring in the optimised geometry and should be the closest one to the Al sites. Hence, we can expect a longer distance between the dangling alkyl groups of the carbenium ions and the Al sites, satisfying the experimental value of 4.7 AE 0.3Å distance as measured by the 13 C{ 27 Al} S-RESPDOR experiments.

Conclusions
In conclusion, we show here that a 13 C-13 C refocused INADE-QUATE experiment on a MTO activated H-ZSM-5 leads to the unambiguous assignment of the 13 C NMR spectrum and the direct spectroscopic determination of the molecular structures of the retained carbon species inside the zeolite framework. The spatial proximities between these carbon species and the zeolite framework were probed by 13 C{ 27 Al} S-RESPDOR and 29 Si{ 13 C} REDOR experiments for which quantitative analysis reveals carbon-aluminium and carbon-silicon host-guest distances in the range of 4.2-4.7Å, supporting pore connement interactions (Fig. S10 †).

Conflicts of interest
There are no conicts to declare.