Molecular behaviour of methanol and dimethyl ether in H-ZSM-5 catalysts as a function of Si/Al ratio: a quasielastic neutron scattering study

Qualitative and quantitative differences are found in methanol and dimethyl ether mobility in H-ZSM-5 catalysts of varying Si/Al ratios (Brønsted acid site concentrations) using quasielastic neutron scattering.


Introduction
Limited gasoline supply in the 1970s led to an increase in oil prices and motivated the search for alternative means of fuel production. 1 The conversion of methanol over H-ZSM-5 catalysts was then discovered as a feasible route for gasoline production in 1977. 2 The first process plant, constructed in New Zealand in 1985 was under continuous operation for at least 6 years. 1 Recently, emphasis on producing value-added chemicals such as olefins have led to a resurgence in the construction of process plant units in China. 3 Through gasification of biomass and coal reserves, or steam reforming of natural gas with subsequent syngas liquefaction, methanol conversion is now a viable route for the production of value-added chemicals such as olefins and fuels. 2,3 For instance, DICP's MTO technology uses SAPO-34 zeotype catalysts for olefin formation from methanol in a fluidised bed reactor. 3 However, a high selectivity to olefins can be obtained over H-ZSM-5 zeolite catalysts when process conditions are tuned towards low pressures 4 and high temperatures. 5 Although SAPO-34 catalysts are the most carbon efficient in producing light olefins 6 , H-ZSM-5 samples provide highest propylene yield, the most diverse product distribution and offers the possibility to re-introduce heavier olefins to boost yield of lighter olefins making efficient use of the cracking chemistry. 7 It is generally accepted that the hydrocarbon pool mechanism regulates product distribution during steady-state methanol-to-hydrocarbon (MTH) conversion in H-ZSM-5. 8,9 This is also known as the dual cycle mechanism as it consists of linked reaction cycles involving olefin and aromatic chemistry.
However, mechanistic understanding of the early stages of the conversion of methanol, comprising of an induction period and a transition regime are not well defined. [10][11][12][13][14] Initially, methanol undergoes an equilibration reaction leading to the formation of dimethyl ether (DME) and water. 15 Readily available oxygenates (methanol and DME), water, and aromatics formed from impurities (acetone, ethanol) in the feed 16,17 compete for sites and their initial coverages are determined from their competitive adsorption. 18 Two of the initial species, that is methanol and DME, are also sources of surface methoxy species 19 which are known to proliferate the direct conversion of oxygenates to olefins [20][21][22] when MTH conversion is tuned towards olefin formation over H-ZSM-5 catalysts. Thus, a detailed understanding of the adsorption/desorption of methanol and DME at the acidic sites would provide insights into their conversion to surface methoxy species and availability for the formation of the first C-C bond.
Optimisation of olefin selectivity is one of the key challenges in MTH chemistry over zeolite and zeotype catalysts. Once the primary olefins are formed, methylation, cracking, hydrogen transfer and cyclisation chemistries regulate olefin selectivity. 23 Of these reactions, methylation is (one of the) most important. Methylation by DME is faster than by methanol due to the abundance of methoxy formed on DME adsorption leading to an additional methylation route. 15,24 In the 3D pore architecture of H-ZSM-5, olefin selectivity is governed by the adsorption/desorption/diffusion of oxygenates and surface reaction of active species at the Brønsted sites. Evidently, these processes differ depending on the location of the sites and their intrinsic acidity (which in turn may differ as a function of Si/Al ratio). The concentration of Brønsted acid sites is one of the main descriptors for olefin selectivity. 25 These factors are considered in the development of microkinetic models 26 which set a conceptual framework where the catalytic cycle can be decoupled, and their interdependencies analysed for selectivity guiding principles. As the hydrocarbon pool ultimately controls the reaction, understanding the processes governing the initial formation provides a mechanistic basis for the modelling of the reaction proceeding afterwards. 27 With respect to the initial adsorption of reactant species over zeolite protons, Blaszkowski and van Santen 28 observed, using DFT calculations, that DME is formed from methanol through an associative pathway. Omojola et al. 19 observed that although methanol adsorbs more easily than DME on H-ZSM-5 catalysts (as shown by higher adsorption constants), the activation energies of desorption of DME are higher than methanol by ca. 10, 4, 6 kJ mol -1 over the highest temperature binding sites of H-ZSM-5 catalysts of Si/Al = 25, 36 and 135 respectively (near vacuum conditions were employed in a temporal analysis of products reactor). In this study, two types of binding sites were observed in H-ZSM-5 catalysts with high Si/Al ratios such as H-ZSM-5(135) and three binding sites were observed in catalysts with lower Si/Al ratios, i.e. H-ZSM-5 (25) and (36) using a detailed elementary plug flow reactor model and a Redhead model. The Redhead method 29 accounts for desorption alone, while the plug flow reactor model accounts additionally for re-adsorption and bed hydrodynamics. These binding sites are reproduced over working catalysts with a fundamental difference in the number of adsorbing molecules per active site in comparison to fresh catalysts. Molecular adsorption on low temperature binding sites and dissociative adsorption (leading to the reversible formation of surface methoxy groups) on medium and high temperature binding sites were observed for methanol and DME adsorption. Jones and Iglesia 30 reached similar conclusions, observing that the mode of adsorption depends on temperature and pressure with the dissociative pathway being dominant at higher temperatures and lower pressures. However, these studies consider adsorption/desorption processes in isolation whereas the influence of diffusion of species could further regulate the availability of oxygenates for primary olefin selectivity, especially in porous materials.
In terms of studying the local, and nanoscale dynamical behaviour of the initial species in H-ZSM-5 catalysts, techniques based on molecular modelling 31 and neutron spectroscopy 32 can offer particularly insightful qualitative and quantitative observations.
Recent molecular modelling studies of the local dynamical behaviour of the initiation species have employed quantum mechanical molecular dynamics (QMMD) simulations to study framework methoxy formation from both methanol and DME in the presence of water and extra methanol 33 , showing that the proton transfer to methanol which initiates the process is accelerated in the presence of excess methanol molecules. However, the reaction with DME is not assisted by the presence of extra methanol. Other studies have probed the above-mentioned methylation reactions, focussing on benzene, where higher methanol loadings encourage methylation from protonated methanol clusters, though this has a higher free energy barrier than methylation from a single methanol molecule in the H-ZSM-5 intersections. 34 Furthermore, studies modelling isobutene reactions with the methanol/DME compared methylation to the susceptibility of hydrogen transfer to produce isobutane. A particularly interesting observation was that the protonation of isobutene by the zeolite acid site, followed by the hydride shift between the methyl group of the oxygenate and the t-butyl cation occurred simultaneously with methanol, but in two separate steps with DME. 35 Examples of neutron spectroscopy contributions include those from O'Malley et al. 36 who used a combination of quasielastic neutron scattering (QENS) and vibrational spectroscopy (INS) to show that the immobility of methanol in H-ZSM-5 catalysts at low loadings was due to its complete room temperature conversion to framework methoxy species, in contrast to zeolite HY of the same Si/Al ratio where the methanol hydroxyl stretches were maintained and diffusion coefficients were later quantifiable. 37 Later studies into the effect of MTH catalysis, and hydrocarbon pool build-up on active species mobility in H-ZSM-5 were carried out by Matam et al, 38 who further showed that methanol is immobile at room temperature when dosed into fresh, and also working H-ZSM-5 catalysts which had previously been tested at 623 K. However, over working catalysts tested at 673 K, isotropic methanol rotation was observed over the QENS instrumental time scale. The observation of methanol mobility in the catalyst tested at the highest temperature was attributed to development of mesoporosity from framework destruction due to dislodgement of lattice Al at 673 K. Earlier QENS studies by Jobic 39 used instruments which probed different timescales, at higher loadings of methanol, to show both 'local' diffusion confined to a sphere corresponding to the H-ZSM-5 channel widths, and the longer range jump diffusion mechanisms through the H-ZSM-5 framework structure. Notably, all the aforementioned studies observe a significant immobile fraction of molecules at all temperatures, indicative of the strong adsorption to Brønsted sites in H-ZSM-5.
While nanoscale studies of methanol diffusion in acidic zeolites as a function of framework structure and catalyst use have been revealing, the molecular mobility as a function of Si/Al ratio (and therefore Brønsted acid site density) has not been studied explicitly, and the mobility of DME, to our knowledge, has not been probed experimentally on the nanoscale. The aforementioned differences in binding energies of less than 10 kJ mol -1 between methanol and DME over H-ZSM-5 (36) and (135) catalysts suggest that neither species is dominant in terms of competition for sites in H-ZSM-5 at these compositions. 27 The mobility of methanol and DME through the catalyst, or indeed their preferred siting in the catalyst framework structure would affect their adsorption to the binding sites and subsequent reaction. This fundamental behaviour of methanol and DME in H-ZSM-5 as a function of Si/Al ratio would therefore be of significant importance in both catalyst design and, in combination with the more macroscale adsorption/desorption studies 19 would be crucial in the formulation of microkinetic models of the early stages of the MTH/O process.
We have therefore performed quasielastic neutron scattering experiments studying the dynamical behaviour of methanol and DME in H-ZSM-5 catalysts with Si/Al = 36 and 135. We find significant qualitative differences between the two species, and potentially counterintuitive differences in mobility which may be explained by a difference in preferred siting in the zeolite framework.

Materials
The H-ZSM-5 samples used were commercial zeolite catalysts obtained from BP chemicals (Si/Al ratio = 36 and Si/Al ratio = 135) with the bulk crystallinity verified by powder x-ray diffraction in recent studies. 19 The H-ZSM-5 samples were originally received in their powdered ammonium form and had to be calcined to achieve their protonated form. This was achieved through heating under vacuum at 5 °C min -1 to 450 °C and holding for 4 h. The calcined samples were dehydrated at 200 °C under vacuum for 8 h. After cooling to room temperature, methanol was then loaded using helium as a carrier gas at a rate of 100 mL min -1 to a loading of ca. 22 molecules per unit cell. H-ZSM-5(36) and (135) were loaded with DME by the gas over the sample at a flow rate of 100 mL min -1 , until a loading of ca. 14 molecules per unit cell was reached in both samples. The loadings of all systems were determined gravimetrically as follows: H-ZSM-5(36) and (135) catalysts were loaded with ca. 12 wt% methanol, or DME to match typical conditions studied in kinetic experiments such as those used by Lercher and co-workers. 40,41 Specifically, the loading was ~14 wt% methanol on H-ZSM-5(36), ~13.5 wt% methanol on H-ZSM-5(135), 12 wt% DME on H-ZSM-5(36), 10wt% DME on H-ZSM-5(135). We note from previous literature that MTH related processes such as methylation reactions are zero order with respect to methanol when co-fed with olefins 42, 43 as with DME is co-fed with olefins 44 or an order of less than 0.5 with respect to methanol suggesting full surface coverage. 45 We therefore aimed to

Results and Discussion
Methanol QENS spectra as a function of the momentum transfer vector Q are shown in Figs. S1 and S2 respectively. The QENS spectra at Q = 1.67 were discounted due to the presence of a Bragg peak in both H-ZSM-5 samples at this Q value, which caused issues upon subtraction of the empty zeolite spectra from those of the loaded zeolite. The spectra were fitted to a delta function convoluted with the resolution measurement taken at 10 K, a single Lorentzian function (which was enough to describe the data satisfactorily) and a flat background function. The figures contain the data points, the total fit (black), and the quasielastic component of the spectra (red) given by the Lorentzian function. It is important to note that the presence of any quasielastic components suggest that full room temperature methoxylation cannot have taken place as in reference 33. We note that the loadings of methanol are a factor of >5 higher than those in the referenced study, and as such far outnumber the concentration of available Brønsted acid sites per unit cell. As such, there will be a significant amount of intact methanol in the catalyst pores.
The presence of a large elastic component in all the spectra suggest that either a large fraction of immobile molecules is present, or a localised motion such as rotation is observed or indeed both.
Quantification of the possible localised motions present can be characterised using the elastic incoherent structure factor (EISF), given by equation 1 is the proportion of the total scattered intensity which is elastic.
The experimental EISFs for methanol in H-ZSM-5(36) are shown in Fig. 1 (and S3 without fitted models). The EISF falls lower as the temperature increases, either due to a differing localised motion or an increasing mobile fraction of molecules. Upon inspection, we note that there may be a slight difference in the shape of the EISF at 293 K compared to those at 333 and 373 K, potentially suggesting that a different mode of motion is present at the lowest temperature.  figure S5, the use of any of the pure models is not adequate to fit the data, and as such, the incorporation of an immobile fraction, which considers that a population of the loaded molecules are either static, or moving too slowly to be observed in the timescale probed by the instrument (~1-100 ps), is necessary (detailed in section S1.1).
The models with the optimal immobile fraction are plotted against the experimental EISF in figure S6.
The best fit to experimental data at 293 K was the model of isotropic rotation with a fraction of ~43%   The rotational diffusion coefficient is lower than that obtained at 325 K in previous work studying methanol in H-ZSM-5 38 potentially reflecting both the lower temperature in this study and also the presence of mesopores in the above reference generated by framework damage due to the MTH process taking place.
The experimental EISF was then fit at 333 and 373 K. Fig. S8 (section S1.1) shows the experimental EISF of methanol in H-ZSM-5(36) at 333 K. The same models described above are applied to the data.
It was clear that none of the models in isolation can fit the data alone. The 3-site model falling above the experimental points at all Q values and also exhibiting a different shape to the experimental data.
While the pure isotropic rotation model can fit the data at the lower Q values, it deviates at higher Q values. The pure model of confined diffusion falls well below the experimental data at all Q-values.
When an immobile fraction is incorporated into the models ( With regard to calculating a diffusion coefficient, the most reliable way for confined systems is to take the HWHM values at the low Q plateau point, which is (4.33D s )/(r conf 2 ). 50 Using r conf = 2.75 Å we obtain the self-diffusion coefficient (D S ) values listed in Table 1, in the range of ~8.6 -9 x 10 -10 m 2 s -1 . We note that these values for a confined/localised diffusion coefficient are lower (by a factor of ~3) than those obtained by Jobic in ref. 39 We consider this may be due to the higher loading used in these experiments, where intermolecular interactions upon confinement can lower the diffusivity of sorbed species. 54,55 With translational diffusion only being exhibited at two temperatures of the measured range, an activation energy of confined diffusion cannot be obtained by an Arrhenius plot from these data.
We now consider the behaviour of methanol in the more siliceous H-ZSM-5 sample (Si/Al ratio = 135), with the QENS spectra showin in Fig. S10. As with H-ZSM-5(36) there is a significant elastic component to all the spectra, suggesting that a localised motion/significant immobile fraction is present.   able to fit the data within the error bars at almost all Q values suggesting at this temperature that DME is able to diffuse translationally in H-ZSM-5 (36). This contrasts to methanol behaviour in this zeolite sample, which appears to have the more localised motion of isotropic rotation.
However, it is clear that a large fraction of the molecules (similar to methanol) are still static, or not exhibiting motion in the 1-100 ps timescale probed by the instrument. It is probable that this fraction is sterically hindered by the constrictive H-ZSM-5 channels, where the population may be located long term, or slowly travelling through the channel system to reach another intersection where they remain locally and observably mobile for longer periods. There is also a high liklihood that a portion of these 'immobile' molecules will be interacting strongly with the Brønsted acid sites, as shown to be very favourable in previous QM/MM embeded cluster calculations. 56 However, we must note that the radius of the confining sphere calculated to fit the data best was actually 3.2 Å rather than 2.75 Å, suggesting the sphere in which DME was diffusing was ~6.4 Å in diameter, larger than the H-ZSM-5 channels. We therefore suggest at this point that the DME is more likely diffusing in the spherically shaped channel intersections which are quoted as being up to to 8-9 Å in diameter 57-60 which will be discussed in more detail later.
The EISFs of DME in H-ZSM-5(36) at all temperatures are plotted in Fig. 6 where the model of diffusion   Table 1. We note that the diffusion coeffcients (in the range of ~9.2 -9.9 x 10 -10 m 2 s -1 ) are slightly higher than those obtained by methanol in the same system, but this is a reflection of the larger sphere used in the calculation. The increase in diffusivity for dimethyl ether is somewhat counterintuitive given the larger size of the DME molecule. But if one were to consider the possibility that its diffusion over this timescale is located in the larger volume intersection rather than the more The QENS spectra for DME in H-ZSM-5(135) are shown in Fig. S14. The EISF plots for DME in H-ZSM-5(135) are shown in Fig. 8. As with DME in H-ZSM-5(36) the best fitting model at each temperature is that of diffusion confined to a sphere, with a mobile fraction of 0.64, 0.67 and 0.68 respectively. We note that these mobile fractions are slightly higher than those obtained for H-ZSM-5(36) reflecting the fewer adsorption sites present. However, when considering the proportion of DME molecules mobile in the 1-100 ps timescale between H-ZSM-5(36) and H-ZSM-5(135) is significantly less than that observed for methanol between the two samples. This suggest that steric hindrance of the larger DME molecule (particularly if there is a fraction located in the H-ZSM-5 channels rather than the  intersections) is the dominant factor in preventing its diffusion, rather than strong interactions with Brønsted sites as suggested for methanol. The most significant observation however, is that the radius of the sphere necessary to fit the EISF to the Volino-Dianoux model is 4 Å (a diameter of 8 Å), as opposed to the 3.2 Å in H-ZSM-5(36), which will be discussed in more detail below. The broadenings as a function of Q for DME in H-ZSM-5(135) are shown in Fig. 9. The broadenings also fit to the Chudley-Elliot jump diffusion model as for H-ZSM-5(36), with very similar residence times. However, we note that the jump distance is larger (3 Å compared to ~2.5 Å ) which may reflect the larger sphere available for diffusion. The plateuing of the broadenings corresponds to a spherical radius range of 3.6 -4.3 Å, again corrobrating the fitting of the EISF to a sphere of diameter 8 Å.

Catalysis Science & Technology Accepted Manuscript
The confined diffusion coefficients calculated from the plateauing of the broadenings at low Q are listed in table 1. We note that at all temperatures the D s is slightly higher than that of the H-ZSM-5 (36) in the range of 9.75 -10.8 x 10 -10 m 2 s -1 . The difference however is very small particularly when the error values associated with each value are considered, and is more a reflection of the larger sphere radius used in the Volino-Dianoux model to calculate D s . As with the methanol samples, the errors of the D s at all temperatures overlap, and as such any activation energies of confined diffusion (Table 1) must be treated with caution. We have tentatively suggested that the DME motions in the H-ZSM-5 (36) and (135) This fitting supports the use of radii larger than that of the H-ZSM-5 channels to fit the confined diffusion models for the EISF, and its attributing to DME location/diffusion in the intersections. It also supports that the use of a larger sphere to fit to the EISF of DME in H-ZSM-5(135) sample is due to the relative absence of Brønsted acid sites in this catalyst, allowing for a larger intersection void space for DME to diffuse in. The observations for DME are significant as they suggest a qualitative difference in DME behaviour compared to methanol, where mobile DME is primarily in the intersections (likely due to the increased molecular dimensions) compared to methanol which appears to remain in the H-ZSM-5 channels, despite both species showing confined diffusion at higher temperatures of similar rates.

Implications of QENS derived dynamical values for kinetic modelling and MTH/MTO catalysis
We now consider the potential implications of the QENS observations from this study on the development of future microkinetic models of the MTH/MTO process. We define microkinetic models in the full sense that molecular events such as adsorption, diffusion, desorption and reaction steps are decoupled without any assumptions on rate-limiting steps. As mentioned in the introduction, For particles in a fixed bed reactor (mesoscale-macroscale level), the adsorption and desorption of species is affected by sorbate concentration across catalyst particles, bed concentration gradients, carrier gas flow rate, re-adsorption and fluid hydrodynamics. 66,67 The ideal or non-ideal plug flow reactor model 19 which is used to describe the behaviour of industrial catalysts in fixed bed reactors, is based on material balance for a reaction component, for a differential element of volume, which is later integrated across the whole reactor. 68  level, these bulk self-diffusivities are placed in the respective partial differential equations to simulate the temporal behaviour of adsorbing and reacting species in a porous particle/pellet. 69,70 On the timescales of this study (1-100 ps), self-diffusion coefficients confined to a spherical region local to the active site (D s (local) ) can be obtained from the QENS broadenings according to equation 2 adapted from the Volino-Dianoux model 50 below.

(2)
Where Γ is the width of the QENS fitted Lorentzian and r conf is the radius of the confining sphere.
We note that the aforementioned rotational diffusion coeffcients (D R ) may also provide another descriptor for the interaction strength with the pore wall and steric constriction in the channels, where a higher D R suggests weaker interaction and less steric constriction. Incorporating local dynamics such as D R and D S (local) into models which also account for the meso-macroscale has, hitherto, not been carried out. These experimentally obtained local descriptors may also be probed using molecular dynamics simulations and crucially may even be used to help tune the forcefields on which these simulations are based.
Previous work by Jobic 39 has shown how one may use different QENS instruments to differentiate between diffusivity confined to an area local to the active site (or at least a portion of the channel approximated by a sphere) using an instrument probing the ~1-100 ps timescale, and how one may track jump diffusion throughout the framework using a higher resolution instrument which probes motion over ~1-10 ns (though at the expense of the wider energy window which allows for faster local motions to be tracked). Indeed, motions over the scale of hundreds of nanoseconds would be necessary for larger species of the hydrocarbon pool, requiring the neutron spin-echo technique to probe motions over this timescale. 71 It is therefore not possible for a single QENS instrument to bridge the gap between direct molecule-Brønsted acid interactions studied with adsorption energies and vibrational spectroscopy, and the microscale probed by adsorption/desorption and mass transport studies. However, a range of instruments may provide descriptors for dynamical behavour at or around the active site (rotation or confined local diffusion), and jump diffusion through cage windows and along the channels of the catalyst framework. Incorporating these descriptors into microkinetic models allows for the more nuanced kinetic modelling of the system across scales previously mentioned, enabling us to more accurately model changes induced by varying the Si/Al ratio of the zeolite catalyst, as demonstrated by this study. There is indeed great potential for these techniques to enable the more detailed modelling of such complex catalytic systems as a function of catalyst composition, which are necessary for understanding factors important to activity and selectivity in MTH/MTO processes over zeolite catalysts.
We now consider the direct incorporation of relevant diffusion coefficients which have been In terms of DME, Perez et al. 24 constructed a kinetic model for the reaction of DME to olefins over a H-ZSM-5 zeolite catalyst in a fixed bed reactor under a pure DME feed, and through co-feeding with helium, methanol, and water. In their reaction scheme, the transformation of DME to olefins and methanol to olefins was decoupled. The kinetic constant of DME conversion to olefins (at 623 K) was found to be 20 times greater than that of methanol conversion. As with Ortega et al. 45  We may use the Wagner-Weisz-Wheeler Modulus (M w ) 73 to determine the influence of intracrystalline diffusion on reaction kinetics at the micropore level.
Where r A is the observed reaction rate (mol m cat -3 s -1 ) , L is the characteristic length (related to the catalyst particle diameter, m), D E is the effective diffusivity directly substituted with the extrapolated values from our QENS studies (m 2 s -1 ) and C a is the concentration of reactant (mol m -3 ).
We have incorporated them into calculations which are based on a pure feed at ca. 1 atm (as indicated in the reference 24) in table 2 to calculate M w . We also obtain a diffusion coefficient 9.36 ×10 -10 m 2 s -1 for methanol in H-ZSM-5(135) using activation the activation energy of diffusion of 0.58 kJ mol -1 (this was not possible for methanol in H-ZSM-5(36) as only two self-diffusion coefficient could be obtained, ruling out an Arrhenius plot). We once again note the need for caution in using these activation energies due to the error bars associated with the each diffusion coefficient, however have confidence in a value on the order of 10 -9 -10 -10 m 2 s -1 . We assume also that the kinetic parameters are constant with respect to crystal size and Si/Al ratio since they are intrinsic as reported by Perez et al.   While a detailed discussion and analysis of the potential activity and product distribution as a function DME and methanol mobility/siting within the catalyst is outside of the scope of this study, there may well be some immediate insights gleaned from these experiments. We note that methanol has been previously shown to adsorb and desorb relatively easily from these catalyst samples 19 but according to our QENS measurements moves relatively slowly (or rotates close in proximity to binding sites) in the channel system of H-ZSM-5. DME adsorption on the other hand, has been shown to be slower than methanol, and its desorption requires higher temperatures potentially due to higher activation energies of desorption over all types of binding sites 19 , but also potentially due to a higher steric hindrance in the H-ZSM-5 channel system as shown in our QENS experiments (concluded through its only observable mobility occurring in the channel intersections).
One may tentatively consider that despite DME and methanol having similar steric access to all regions of the catalyst framework, DME is more able to remain in all regions of the H-ZSM-5 framework (channels and intersections), leading to an increased interaction with more of the different Brønsted acid sites which vary in strength as a function of location. The formation of the first C-C bond which drives the formation of primary olefins from the equilibrium mixture of methanol, DME and water, may therefore take place in a situation where DME would be bound to a wider range of vacant sites, with a higher abundance and coverage given its differing adsorption and desorption behaviour.
Recent work 18 has shown that even during the competitive adsorption of oxygenates and aromatics over H-ZSM-5 catalysts, DME still has a higher surface coverage at typical MTH/O conditions. This higher coverage may suggest a more DME-mediated pathway towards the formation of primary olefins in the absence of diffusion constraints. The higher diffusion coefficients obtained in this study for DME compared to methanol (derived from the differing location of DME allowing more space to be mobile in) suggests that even if a DME-mediated pathway to olefin formation were to occur, its mobility may not limit the establishment of significant surface coverages throughout the pore architecture of H-ZSM-5 catalysts. Previous work 24,74,75 as well as the industrial DICP MTO process 3 in the archived literature support the use of DME, or a mixture of methanol/DME as feedstock for primary olefin formation over zeolite and zeotype catalysts. Our qualitative and quantitative insights from these QENS measurements may provide a more fundamental understanding of the consequences of using this varied feedstock. Finally, we consider the importance of probing temperatures closer to that of the catalytic reaction (> 573 K for the MTH process). Our local self-diffusivities extrapolated to 623 K are listed in table 2, though we have explained the need for caution with these values, and note that they neglect the inevitable buildup of the hydrocarbon pool which will affect mobility in the real system. The potential for methoxylation reactions to occur even at temperatures close to 373 K would cause issues with the measurement due to the increasing concentration of water in the system, which would be mobile over the same timescales as methanol (though DME measurements may be less problematic). One possible option would be to use the siliceous analogue of H-ZSM-5, silicalite-1, to gauge the effect of the MFI framework on species diffusivity without reaction complications at these higher temperatures. Future, more realistic studies must embrace the complexity of the hydrocarbon pool build-up and its effects on starting species behaviour at catalysis relevant temperatures, building on previous work which attempted this at ambient temperatures. 38 We consider that future experimental developments such as reliable flow-through neutron sample environments may allow for the in situ dosing of species to the working catalyst, and that this dosing may be combined with the strategic deuteration of the catalyst and the build-up of a deuterated hydrocarbon pool, such that its scattering cross-section is negligable compared to an incoming hydrogenous reactant stream. At a time where the flux to neutron instruments is ever increasing, allowing for a fast measurement of diffusion coefficients before the reactant stream is consumed, selective measurements of relevant species' mobility under increasingly realistic conditions may well be an acheivable goal for neutron scattering experiments in the coming years.

Summary and Conclusion
Quasielastic Measurements of DME motion showed that in H-ZSM-5 (36), diffusion confined to a sphere is observed at all temperatures. This contrasts with methanol in this sample, which showed rotational motion at 293 K, suggesting that DME is freer to diffuse at 293 K despite the increase in molecular dimensions.
However, DME showed diffusion confined in a sphere of diameter 6.2 Å, with immobile fractions of 41-34% with increasing temperature (far smaller increases in mobile populations with temperature compared to methanol). The confined radius is larger than that of the H-ZSM-5 channels, suggesting that the DME dynamics are confined to the H-ZSM-5 intersections. D s values were obtained in the range of 9.2 -9.9 x 10 -10 m 2 s -1 , slightly larger than those for methanol in the same sample, however this is more indicative of the larger sphere radius derived for the Volino-Dianoux model used to calculate the D s . In the H-ZSM-5(135) sample, confined spherical diffusion was also observed, with slightly lower immobile fractions (36-32% with increasing temperature). The D s values calculated were slightly higher (9.7 -10.8 x 10 -10 m 2 s -1 ) that those obtained in the H-ZSM-5(36) sample. The sphere to which diffusion was confined in H-ZSM-5(135) was even larger than that of the H-ZSM-5(36) sample, with a diameter of 8 Å (the increase in D s is again, due primarily to the larger sphere in the Volino-Dianoux calculating model). This suggests that if DME diffusion is confined to the intersection in H-ZSM-5(135) as with H-ZSM-5 (36), that the intersection has a larger free volume for diffusion, probably due to the lack of Brønsted acid sites present. This was supported by an investigation of the atomic configuration of the intersection from an experimental silicalite sample, and geometry optimised H-ZSM-5 structure.
The study illustrates the complex nature of sorbate behaviour in the methanol-to-hydrocarbons process even under fresh catalyst conditions. The catalyst composition and concentration of Brønsted sites can have qualitative differences in behaviour of methanol at ambient temperatures, and significant differences in the size of 'immobile' fractions at higher temperatures between samples. In contrast, despite the larger molecular dimensions of DME, the diffusion of molecules which are mobile on the instrumental timescale does not appear to be significantly hindered with increasing Brønsted acid site concentration, potentially due to their location in the larger channel intersections of the framework, which may differ in free volume depending on catalyst composition.