Understanding zeolite deactivation by sulfur poisoning during direct olefin upgrading

The presence of sulfur contaminants in bitumen derived crude oils can lead to rapid catalyst deactivation and is a major problem faced by downstream refiners. Whilst expensive hydrotreating steps may remove much of the sulfur content, it is important to understand how catalyst deactivation by sulfur poisoning occurs and how it may be mitigated. Here we report a mechanistic study of sulfur poisoning over a zeolite catalyst promoted with silver and gallium Lewis acids. Olefin upgrading, an essential process in the refinement of heavy oils, is used as a model reaction. Access to the zeolite inner pores is blocked by bulky, weakly adsorbed sulfur species. Pore access and thus catalyst activity is restored by increasing the reaction temperature. We also show that a simple alkaline treatment greatly improves both the sulfur tolerance and performance of the catalyst. These findings may enhance the rational design of heterogenous catalysts for olefin upgrading. The deactivation of zeolite catalyst by sulfur contaminants is a major problem in direct olefin upgrading of crude oil. Here the authors perform mechanistic analyses of zeolite ZSM-5 catalysts to reveal the detrimental effect of sulfur on silver and gallium Lewis acid sites in the catalyst.

D epletion of conventional oil sources has led to a surge in demand for feasible alternatives. Therefore, unconventional crude oils such as bitumen have attracted much attention in recent years. Reserves of bitumen oil sands were estimated to be 24 billion barrels at the end of 2015 in Canada alone (retrieved from: https://www.aer.ca/providing-information/ data-and-reports/statistical-reports/st98). The viscosity and density of heavy crude oils necessitates partial upgrading via high temperature cracking prior to transportation to refineries downstream. However, the partial upgrading of heavy oils leads to the formation of high quantities of olefinic compounds as byproduct. The chemical instability and high concentration of olefins in partially upgraded streams can lead to the formation of highly undesirable polymeric deposits following prolonged periods in storage or during long distance pipeline transportation 1 .
A major problem faced by refiners is the sulfur content of heavy hydrocarbon feeds such as bitumen-derived heavy oils and product streams such as delayed coker light naphtha 2 . Reducing sulfur contaminants is of vital importance as unconverted sulfur poisons may not only lead to rapid deactivation of zeolitic catalysts for downstream refining 3 but also contribute towards the production of harmful SO x emissions. Sulfur content has been strongly correlated to the olefin content in such feeds 4 . Therefore, it is desirable to produce a heterogeneous catalyst that is not only active for the selective upgrading of olefins, but that is also resistant to deactivation via sulfur species.
Perhaps the most well-known class of catalytic materials for oil upgrading are zeolites. Zeolite catalysts, such as HZSM-5, have also displayed promising activity towards the selective upgrading of olefins. Modification with metal dopants such as gallium (Ga) can increase activity and promote increased yields of valuable aromatic products 5 . However, the high temperatures typically employed industrially often result in significant coke formation, although the thermal stability of ZSM-5 catalysts may be enhanced by modifications such as phosphorous addition 6,7 and alkaline treatment 8 .
A common theme throughout the literature is the use of hydrogen gas for refining processes, despite the natural unavailability and cost of H 2 , which are severe drawbacks to hydrotreatments. It is well known that H 2 is mainly produced via catalytic steam reforming of methane, yielding syngas 9 . We have previously shown that by employing a HZSM-5 catalyst loaded with silver (Ag) 10,11 , gallium (Ga) 10,11 , palladium (Pd) [12][13][14] and iridium (Ir) 14,15 , olefin feedstocks may be directly upgraded to paraffins, naphthenes and aromatics under CH 4 environment 10 .
Herein, we report the effect of sulfur contaminants on the catalytic direct upgrading of olefins using a Ag-Ga/ZSM-5 zeolite catalyst. It is found that sulfur poisoning is temporary and occurs only when sulfur species are present in the reaction medium. Catalyst poisoning is shown to be significant in the presence of bulky sulfur molecules such as dibenzothiophene (DBT), which inhibits access to the zeolite micropores by weakly adsorbing to Ag and Ga species located on the external surface, preventing key steps from occurring in the olefin upgrading reaction. Sulfur poisoning is shown to be temperature-dependent and full catalyst activity is recovered upon increasing reaction temperature to 400°C . We also show that catalyst performance and sulfur poisoning are mitigated by introducing mesoporosity into the catalyst through a simple post-synthetic alkaline treatment.

Results
Direct olefin upgrading with a simulated feed. Initial investigations with a simulated feed (composition given in Supplementary Table 1) yields an 83 wt% olefin reduction with high selectivity towards aromatics (Supplementary Fig. 1). Adding 5 wt% diethyl sulfide (DES) and DBT to the reaction mixture (sulfur to metal ratio of 1:0.001 for both Ag and Ga metals) represses both olefin reduction and aromatic formation. Interestingly, catalyst activity is recovered upon removal of sulfur from the reaction medium and Ag-Ga/ZSM-5 is only temporarily and partially deactivated by sulfur. 1-Decene model compound studies. The deactivation mechanism was probed using 1-decene as a model compound. The reader is directed to Supplementary Tables 2-6 for major products of the reactions discussed below and Supplementary Note 1 for the equations used for performance evaluation. Fig. 1a shows that at 330°C, olefin reduction is relatively low (35 wt%) compared to the simulated feed (80 wt%), likely due to the lower olefin content and co-reactants present. Olefin reduction increases as a function of temperature, with values of 35 wt% and 91 wt% achieved at 330 and 400°C, respectively. Fig. 1b displays the relationship between reaction temperature and selectivity towards both paraffins and aromatics. At 330°C, paraffins are the major products, with concentrations of 9 wt% and 22 wt% of aromatics and paraffins observed in the product oil, respectively. Increasing reaction temperature promotes aromatic formation until it plateaus at 400°C (71 wt%). Analogous reactions were then run with DBT and DES contaminants present. Olefin reduction is essentially switched off at 330°C (Fig. 1c), however, sulfur poisoning is diminished with increasing temperature and is negligible at 400°C. The extent of catalyst deactivation at 330°C was investigated via reuse with fresh 1-decene. As with the simulated oil feed discussed earlier, catalyst activity is almost entirely recovered, with olefin reductions of 35 wt% and 30 wt% for fresh and reused catalyst, respectively (Fig. 1e). We conclude that for direct olefin upgrading over Ag-Ga/ZSM-5, sulfur deactivation is only temporary and, furthermore, is only observed when sulfur contaminants are directly present in the reaction medium. This may suggest that DBT and DES contaminants desorb from key active sites for olefin transformation, particularly into aromatics, as reaction temperature increases from 330 to 400°C.
Shorter reaction times of 10-30 min were then performed and additionally, the extent of olefin reduction that occurs during the heat-up phase of the reaction was investigated. This was achieved by allowing the reaction to reach the desired temperature at the normal heating rate and once at the desired temperature, the reaction was immediately quenched by submerging the vessel into an ice bath. Complete sulfur deactivation was observed even after 10 min at 330°C, demonstrating that Ag-Ga/ZSM-5 deactivation via sulfur poisoning is rapid (Supplementary Table 7). On increasing the temperature, the reaction proceeds quickly, with >60 wt% olefin reduction seen during the heat-up phase to 400°C both in the presence and absence of sulfur ( Supplementary  Fig. 2).
Our initial results with a simulated oil feed indicate that coreactants may strongly influence not only the olefin upgrading activity of the catalyst but also deactivation due to sulfur contaminants. Olefin upgrading experiments were therefore performed using heptane and benzene as model paraffin and aromatic co-reactants, respectively. Paraffin co-reactant plays no positive role during direct olefin upgrading at 330°C, whereas aromatic co-reactant may improve olefin reduction by promoting aromatic alkylation reaction between olefin substrate and aromatic co-reactant over Ag-Ga/ZSM-5 (Supplementary Fig. 3 and Supplementary Table 8). Furthermore, aromatics may react with organosulfur contaminants to form alkylbenzene products, inhibiting catalyst deactivation 16 . The reader is directed to Supplementary Discussion for a more detailed analysis.   Table 9). Characterisation of fresh and 1-decene spent Ag-Ga/ZSM-5 agrees with our previous studies on this catalyst 11,17 . The reader is directed to Supplementary Discussion for in-depth discussion and results. The effects of DES and DBT contaminants on the catalyst properties were investigated by characterising spent Ag-Ga/ZSM-5 samples following reaction with 1-decene and sulfur contaminants at both 330 and 400°C. XPS analysis of spent samples exposed to sulfur contaminants shows significant differences in both the Ag 3d 5/2 and Ga 2p 3/2 regions. Fig. 2a shows peak shifts of 1.1 and 1.4 eV in the Ag 3d 5/2 region of 330°C and 400°C spent samples, respectively. Furthermore, a small binding energy (B.E.) shift of 0.3 eV is seen between the spent samples, with peaks centred at 367.9 and 368.2 eV for 330°C and 400°C samples, respectively. Whilst both values have been attributed to Ag 2 S 18 , the reported values for Ag metal are similar and so absolute identification of the Ag species present is not possible solely from this data 19 . The Ga 2p 3/2 region (Fig. 2b) shows a significant shift in B.E. from 1118.8 eV in the fresh catalyst to 1117.8 eV for both spent catalysts. The Ga 2p 3/2 peak at 1117.8 eV has been attributed to α-Ga 2 O 3 20 , however, this may also be assigned to Ga 2 S 3 21,22 , suggesting that Ga 2 S 3 may be formed upon reaction with sulfur contaminants.
The surface concentration of gallium for spent samples appears to be decreased, indicating that some gallium species have migrated into the inner pores. Examination of the S 2s region reveals differences between the spent catalysts (Fig. 2c). The S 2s peak is centred at 226.4 eV following reaction at 330°C, indicative of both Ag 2 S and Ga 2 S 3 23, 24 . Upon increasing reaction temperature to 400°C, the S 2s peak shifts to 227.7 eV. An identical B.E. shift is assigned to the decomposition of Ag 2 S to elemental sulfur elsewhere 23 .
X-ray absorption near-edge structure (XANES) Ga k-edge spectra of Ag-Ga/ZSM-5 are shown in Supplementary Fig. 9.   These values are proposed to correspond to tetrahedral (Ga t ) and octahedral (Ga o ) Ga 3+ species, respectively 25,26 . Upon exposure with 1-decene and sulfur contaminants, the E 0 position remains stable at approximately 10,375 eV, suggesting that there is no shift in the position of the Ga t peak compared with the reference. However, the Ga o peak appears to shift to approximately 10,378 eV during the heat-up phase and is shifted even further to 10,377 eV for all sulfur-exposed samples, evidenced by the shift in the white line position (Supplementary Table 10). Furthermore, the white line intensity decreases with increasing temperature. The observed changes to the Ga o peak suggests that whilst much of the Ga 3+ species remain in a tetrahedral coordination local environment, migration to the inner pores of the catalyst may also occur 27 . Spent, sulfur-exposed Ag-Ga/ZSM-5 samples were imaged using HR-TEM with elemental mapping (EM). Fig. 3a-e reveals that following reaction at 330°C, sulfur species are bound to large agglomerates containing both Ag and Ga species, likely formed as a consequence of sintering, on the outer surface of the zeolite support i.e. not within the pores. Ag sintering is seemingly increased at 400°C and surface sulfur species are primarily bound to these Ag clusters ( Fig. 3f-j). It is known that the size of metal clusters may also affect the electronic affinity for sulfur adsorption on Ag particles 28 . Notably, the surface coverage of sulfur appears significantly reduced as the reaction temperature is increased to 400°C, aligning with our experimental observations of recovered catalyst activity at higher reaction temperatures.
Acid site analysis of spent sulfur-exposed samples was performed using NH 3 -TPD. Following reaction at 330°C, two new peaks appear in the NH 3 -TPD profile of Ag-Ga/ZSM-5 centred at approximately 370°C and 415°C (Fig. 4b), neither are indicative of the weak or strong acid sites seen on the fresh sample (Fig. 4a). Our TEM observations suggest that sulfur species are still present on this sample, principally bound to both Ag and Ga species on the external surface. It is plausible that NH 3 may therefore preferentially adsorb to these sulfur species yielding the differences in the NH 3 -TPD profile of this sample compared with fresh Ag-Ga/ZSM-5. Only weak and medium acid sites are present on the 400°C spent sample (Fig. 4c), in keeping with our previous findings that strong acid sites remain occupied post reaction 11 . The disappearance of the peaks at 370°C and 415°C upon increasing reaction temperature indicates that much of the sulfur species have desorbed from Ag-Ga/ZSM-5.
To further validate this, TPD experiments were performed on spent catalyst samples using He as carrier gas. Following reaction with sulfur poisons at 330°C, two peaks are visible between approximately 350 and 500°C in the TPD profile (Fig. 4d). Whereas, peaks are only observed in the much lower temperature range of approximately 100-320°C for the 400°C spent sample, implying that the active sites for olefin feed adsorption and transformation on the catalyst surface are temporarily blocked by sulfur poisons at low temperatures and become readily available again when reaction temperature is increased (e.g. >330°C), probably due to decomposition and desorption. Quantification of acid sites reveals an increase for sulfur-exposed, spent samples (Supplementary Table 11) with values of 567, 1585 and 777 μmol NH 3 g −1 for fresh, spent 330°C and spent 400°C samples, respectively. These data further indicate that sulfur species capable of strongly adsorbing NH 3 are still present on all spent samples, but that significant desorption occurs at reaction temperatures >330°C.
Additional TPD experiments were performed following adsorption of 1-decene, DBT and DES at room temperature. We note from the data shown in Fig. 4e that all samples show desorption peaks at temperatures attributed to weaker acid sites. Closer examination of the TPD profiles shows that following 1decene adsorption, two weak, broad desorption peaks are observed at 200-400°C. When DES and DBT are adsorbed on the catalyst, either alone or in a mixture with 1-decene, two distinct desorption peaks are seen at 200-350°C and interestingly the intensity of these peaks is far greater compared to the desorption peaks seen following 1-decene adsorption alone. These data imply that sulfur contaminants, when present in the reaction medium, may strongly compete with 1-decene olefin substrate for adsorption upon weak and medium acid sites of Ag-Ga/ZSM-5 catalyst.
The strength of adsorption of DBT and DES on Ag-Ga/HZSM-5 was investigated using thermogravimetric analysis (TGA). DBT and DES were first adsorbed onto the catalyst surface and the resultant sample was subjected to TGA-DSC under N 2 flow, holding the temperature at 330°C, where greatest catalyst deactivation due to sulfur poisoning occurs, before ramping to 500°C. Supplementary Fig. 10   exothermic heat flow at approximately 160 and 330°C. The first peak centred at approximately 160°C might indicate the decomposition of DES potentially chemically bonded to the catalyst surface, as this is significantly higher than the 98°C boiling point of DES. The second sudden mass loss centred at approximately 330°C corresponds with the boiling point of DBT indicating that DBT is weakly adsorbed on the catalyst surface. These data indicate that DES may desorb from Ag-Ga/ZSM-5 at relatively low temperatures whilst DBT remains adsorbed at temperatures <330°C and may be chiefly responsible for observed Ag-Ga/ZSM-5 deactivation. Additionally, the decomposition of DES at temperatures >160°C and the weakly chemisorbed nature of DBT may rationalise the restoration of the olefin reduction performance of spent Ag-Ga/ZSM-5 once it is charged in a sulfurfree medium. Our TGA data suggests that DBT desorbs from the surface of the spent catalyst at 330°C. Therefore, when the spent catalyst is charged into a sulfur-free medium, desorption may occur from the surface at the reaction temperatures investigated (330-400°C), thus freeing up the active sites for olefin upgrading. The calculated adsorption energies are shown in Table 1. The adsorption energy of DBT on Al-OH and Si-OH are −139.2 kJ mol -1 and −136.6 kJ mol −1 , respectively. When DBT adsorbs on Ag or Ga atom, the adsorption energy was calculated at −216.0 kJ mol -1 and −216.9 kJ mol −1 , respectively. The increased negative adsorption energy of these metal sites suggests that DBT will preferentially adsorb on Ag and Ga species located on the external surface of Ag-Ga/ZSM-5. These data strongly agree with our XPS, TEM and TPD data and as such we tentatively conclude that DBT is mainly bound to Ag and Ga surface species during 1-decene upgrading reaction at 330°C over Ag-Ga/ZSM-5. The calculated energies may also indicate the strength of adsorption of DBT over Ag-Ga/ZSM-5. The energies of physisorption and chemisorption of DBT over various sites of carbon nanotubes have been reported as 128.0-332.2 kJ mol -1 and 544.3-637.2 kJ mol −1 , respectively 29 . Our DFT calculations may therefore complement our TGA results, indicating that DBT only weakly interacts with the catalyst surface. Furthermore, these data may further rationalise the restoration of the olefin upgrading performance of Ag-Ga/ZSM-5 upon reuse at temperatures of 330°C and above in a sulfur-free medium.
Influence of the zeolite structure. The mechanism of olefin aromatisation over zeolite catalysts is complex and multistep. More specifically, the formation of aromatics from n-C 7 + olefins is reported to proceed via isomerisation, cracking, oligomerisation and finally hydrogen transfer 30 . Crucially, oligomerisation reactions have been shown to occur within the zeolite pores and, in blocking access to these pores as a result of high temperature coking, olefin conversion and aromatic formation is greatly suppressed 31 . Our characterisation data and computational studies suggest that DBT preferentially adsorbs upon Ag and Ga species located on the external surface of Ag-Ga/ZSM-5. Given the bulky size of DBT (kinetic diameter of approximately 9 Å) 32 relative to the micropores of ZSM-5 (pore limiting diameter of approximately 5.6 Å) 33 it may be blocking access to the inner pores when 1-decene upgrading is run at 330°C, preventing olefin oligomerisation and cyclisation reactions and in turn, deactivating Ag-Ga/ZSM-5.
To explore this, we first calculated the selectivity towards diolefin and cycloolefin oligomers as a function of all olefins in the product oil. Fig. 5 shows that as reaction temperature increases, the selectivity towards cycloolefins increases whilst diolefin selectivity is decreased. Crucially, in the presence of DBT and DES, the formation of both cycloolefin and diolefin oligomers is inhibited at 330°C, correlating with the deactivation of Ag-Ga/ ZSM-5. The role of the porous structure of ZSM-5 can be investigated by comparing Ag-Ga/ZSM-5 with a Ag-Ga/SiO 2 -Al 2 O 3 catalyst with the same SiO 2 :Al 2 O 3 ratio. Fig. 6a shows that the activity is substantially worse over the mixed metal oxide catalyst, furthermore, aromatic formation is low and paraffins are the main product. Similarly, the effect of surface acidity was investigated by preparing a Ag-Ga/Si-MFI catalyst. Our data suggest that Brønsted acid sites may be vital for 1-decene upgrading to aromatics as essentially no olefin reduction and zero aromatic formation is seen following reaction at 400°C over Ag-Ga/Si-MFI.
The reaction mechanism was further probed by evaluating olefin distribution in the obtained product oils (Fig. 6b). Ag-Ga/ Si-MFI yields predominantly linear monoolefins, whereas Ag-Ga/ SiO 2 -Al 2 O 3 forms lighter, branched monoolefins, demonstrating that surface Brønsted acid sites are necessary for initial activation of 1-decene via isomerisation and subsequent cracking reactions. Comparing Ag-Ga/SiO 2 -Al 2 O 3 with bare HZSM-5 and Ag-Ga/ ZSM-5, the role of Brønsted acid sites located within the zeolite pores becomes clear as oligomerisation reactions proceed on both zeolite catalysts forming cycloolefins and a small amount of diolefins.
The activation of olefins over solid acid catalysts is widely considered to occur through the formation of a carbenium ion following protonation of the C-C olefin double bond 34 . The exact nature of the carbenium ion is still a subject of debate within the literature and is thought to vary for double-bond isomerisation, skeletal isomerisation and cracking reactions [34][35][36] . Catalytic cracking of C 6+ olefins is reported to involve β-scission of a tricoordinated carbenium ion formed over Brønsted acid sites, yielding an olefin and a free, smaller carbenium ion 37 . In the presence of sulfur contaminants at 330°C, negligible cracking activity occurs and 1-decene undergoes isomerisation, almost exclusively ( Supplementary Fig. 14). In the absence of sulfur, cracking reactions proceed and many lighter, branched olefins are seen in the product oil. Prior to any appreciable cracking activity, C 6+ olefins are reported to undergo multiple isomerisation reactions, meaning that they will rapidly adsorb and desorb from the catalyst surface many times before cracking 37 . Our data indicate that DBT may strongly inhibit cracking reactions and limit 1-decene isomerisation over Ag-Ga/ZSM-5 at 330°C, which may be related to the reported requirement for multiple adsorption and desorption steps in the process.
Ag and Ga species are not thought to be involved in the isomerisation, cracking or oligomerisation reactions, evidenced by the inactivity of Ag-Ga/Si-MFI. Rather, Ag and Ga species are thought to actively catalyse dehydrogenation reactions promoting aromatic formation. These data are in good agreement with previous studies, which have concluded that Brønsted acid sites, located within the pores of the zeolite, are chiefly responsible for olefin oligomerisation and cyclisation whilst metal dopants such as Ga enhance the dehydrogenation of cycloolefin and diolefin intermediates 38,39 . The relative rate of cycloolefin dehydrogenation is reported to be eight times that of linear and branched olefins over a Ga/ZSM-5 catalyst 40 , and, if the formation of cycloolefins is inhibited then conversion to aromatics will likely not proceed. This aligns well with our experimental observations in the presence of sulfur. In summary, our data strongly suggest that adsorbed DBT inhibits the formation of olefin oligomers such as cycloolefins and furthermore that this reaction likely occurs within the zeolite pores.
Inner pore accessibility. To confirm the importance of inner pore accessibility in the upgrading of 1-decene, we performed a coking procedure known to selectively coke the inner pores of ZSM-5 17,41 . When this coked Ag-Ga/ZSM-5 catalyst is employed for 1-decene upgrading at 330°C, the catalyst activity is substantially diminished (Fig. 6c). Coking the inner pores of Ag-Ga/ ZSM-5 prevents the formation of cycloolefins and diolefins, in keeping with our hypothesis that the inner pores are pivotal in the formation of these intermediates and therefore, aromatics. The textural properties of coked Ag-Ga/ZSM-5 was investigated by N 2 physisorption ( Table 2). The coking procedure leads to a significant decrease in total surface area, from 491 m 2 g -1 to 266 m 2 g −1 . Micropore surface area was reduced by 46%, accompanied by a 50% reduction in micropore volume,  Fig. 6 Olefin reduction performance and olefin selectivity of various catalysts. a Olefin reduction performance of various catalysts at 400°C using 1-decene as substrate. Black bars indicate olefin reduction, magenta solid line with magenta shaded squares indicates selectivity towards paraffins and magenta dashed line with magenta hollow squares indicates selectivity towards aromatics. b Distribution of olefins observed in the product oil over various catalysts using 1-decene as substrate at 400°C, blue bars indicate linear monoolefins, grey bars indicate branched monoolefins, magenta bars indicate diolefins and black bars indicate cycloolefins. c Olefin reduction performance of Ag-Ga/ZSM-5 and coked Ag-Ga/ZSM-5 catalysts using 1-decene as substrate at 330 and 400°C, black patterned bars indicate olefin reduction at 330°C and black solid bars indicate olefin reduction at 400°C. Magenta solid line with magenta shaded squares indicates selectivity towards paraffins, and magenta dashed line with magenta hollow squares indicates selectivity towards aromatics in product oil. d Olefin reduction performance of Ag-Ga/ZSM-5 and NaOH-treated Ag-Ga/ZSM-5 catalysts using 1-decene as substrate in the presence and absence of DBT and DES at 330°C, black patterned bars indicate olefin reduction in the presence of DES and DBT, and black solid bars indicate olefin reduction in the absence of DES and DBT. Magenta solid line with magenta shaded squares indicates selectivity towards paraffins, and magenta dashed line with magenta hollow squares indicates selectivity towards aromatics in product oil. All error bars represent ±4% calculated as standard deviation confirming that a significant amount of coke deposition occurs within the inner pores of Ag-Ga/ZSM-5.
Surface area analysis of spent, sulfur-exposed Ag-Ga/ZSM-5 shows interesting changes compared with fresh catalyst ( Table 2). Brauner-Emmet-Teller (BET) surface area of the spent 330°C sample decreased by 25% compared with the spent 400°C sample, this is attributed almost exclusively to the microporous surface area. Micropore volume also decreased, with values of 0.07 cm 3 g -1 and 0.1 cm 3 g −1 for 330°C and 400°C spent samples, respectively. These data strongly indicate that access to the inner pores is hindered as a consequence of adsorbed species on the surface of Ag-Ga/ZSM-5 and that as the reaction temperature is increased, these species may desorb, resulting in a significantly increased microporous surface area and micropore volume.
Time of flight secondary ion mass spectroscopy (TOF-SIMS) shows a moderately intense peak at a m/z of 185 ( Supplementary  Fig. 15) in the spectrum of Ag-Ga/ZSM-5 following reaction with 1-decene and sulfur contaminants at 330°C, which may be assigned to DBT (molecular mass = 184.26). The absence of this peak in the spectrum of all other samples may indicate that DBT presence on the surface of Ag-Ga/ZSM-5 is significant following reaction at 330°C. Based upon all the above data, we tentatively conclude that DBT effectively blocks access to the inner pores of Ag-Ga/ZSM-5 at reaction temperature of 330°C, preventing oligomerisation and aromatisation reactions from occurring and deactivating the catalyst.
Increasing the sulfur tolerance of Ag-Ga/ZSM-5. Simple postsynthetic treatments such as reaction with mild alkaline solutions have been shown to be highly effective in the transformation of conventional zeolites into hierarchical forms 42,43 . Treatment of MFI-type zeolites with relatively mild concentrations of NaOH may effectively generate mesoporisity through the dissolution of framework silicon, whilst the framework aluminium and silicon neighbours are largely unaffected, preserving the Brønsted acidity to a good degree 44 . Such materials possess mesopores that are intricately connected with the characteristic zeolite micropores. Mesoporous zeolites have shown increased performance in the aromatisation of olefins due to enhanced micropore accessibility 8 . We therefore performed a simple alkaline treatment on Ag-Ga/ ZSM-5 according to a method reported in the literature 45 . Interestingly, the alkaline treatment not only enhanced the olefin upgrading performance of Ag-Ga/ZSM-5 but also substantially mitigated the effects of sulfur poisoning. Fig. 6d shows that in the absence of sulfur, olefin reduction increases from approximately 35 wt% to approximately 73 wt% and, in the presence of sulfur, olefin reduction increases from 0 to approximately 64 wt%.
The textural properties of NaOH treated Ag-Ga/ZSM-5 are shown in Table 2. The external surface area increased significantly, from 95.5 m 2 g -1 to 161.1 m 2 g −1 following NaOH treatment, whereas the micropore volume decreased from 0.16 cm 3 g -1 to 0.14 cm 3 g −1 indicating that aluminium redistribution and amorphisation occurred during the treatment. A classic type IV isotherm with a characteristic hysteresis loop is present after NaOH treatment of Ag-Ga/ZSM-5, confirming the successful development of mesoporosity (Fig. 7) 43 . X-ray diffraction analysis shows that the MFI crystal structure of Ag-Ga/ZSM-5 is maintained following NaOH treatment (Supplementary Fig. 16).
The acidity of NaOH-modified Ag-Ga/ZSM-5 was analysed by pyridine DRIFT and NH 3 -TPD. Although exact Lewis and Brønsted acid site quantification is not possible solely from our DRIFT spectra, as dictated by the Beer-Lambert Law, the signals obtained are shown as a function of absorbance to indicate any obvious changes in acid site distribution. Supplementary Fig. 17 shows that the area of the peak located at approximately 1540 cm −1 , assigned to pyridinium ion (Brønsted acid sites) 46,47 relative to the area of the peak located at approximately 1450 cm −1 assigned to covalently bonded pyridine (Lewis acid sites) 46, 47 is diminished, indicating a relative increase in the number of Lewis acid sites. Quantification of NH 3 -TPD curves (Supplementary Fig. 18 and Supplementary Table 12) reveals a slight decrease in the strong acid sites and a significant increase in the number of weak acid sites, consistent with the findings of others, suggesting an increase in Lewis acidity occurs following treatment of Ag-Ga/ ZSM-5 with NaOH 45 .  External surface area (m 2 g −1 ) Micropore surface area (m 2 g −1 ) Micropore volume (cm 3 g −1 ) HZSM Such Lewis acid sites have been proposed to generate primarily at the mesoporous surface of ZSM-5 catalysts, rather than inside the micropores 45 . Our support control experiments have shown that the acid sites located inside the micropores of Ag-Ga/ZSM-5 are key for efficient transformation of olefin substrate into aromatics. Furthermore, our catalyst characterisations and computational analysis strongly suggest that DBT preferentially adsorbs upon Ag and Ga species located on the external surface of Ag-Ga/ZSM-5. Therefore, we reasonably assume that increased access to the micropores of Ag-Ga/ZSM-5 through the generation of a hierarchical mesoporous structure affects the olefin reduction performance of the catalyst and its resistance to sulfur poisoning via adsorbed DBT to a far greater extent than the generation of additional EfAl Lewis acid sites located primarily at the mesoporous surface. Consequently, the enhanced catalytic activity of NaOH-modified Ag-Ga/ZSM-5 in the presence of sulfur contaminants is reasonably assumed to be primarily due to the introduction of mesoporisity and a hierarchical structure.

Discussion
The introduction of Ag and Ga Lewis acid sites into microporous zeolite catalysts, such as ZSM-5 is known to increase dehydroaromatisation activity. Extraframework Ag + Lewis acid sites are proposed to stabilise π-complexed intermediates, enhancing aromatic formation 48,49 . Ga-promoted ZSM-5 catalysts are known to increase dehydrogenation activity towards olefin oligomers, significantly increasing aromatisation activity compared with their parent zeolite counterparts 31 . Whist evidence exists for methane incorporation over Ag and Ga modified zeolites, methane is not thought to play an active role in this study.
Ga-methoxy species, formed after methane activation, are reported to provide a route for incorporating methane into aromatic products 50 . Given that the formation of aromatics over Ag-Ga/ZSM-5 is blocked by sulfur poisoning at 330°C, such a reaction pathway is proposed to not be relevant to the scope of this study. Nuclear magnetic resonance studies have shown that at reaction temperatures < 450°C methane plays no active role in the transformation of olefins over Ag/HZSM-5 catalyst 49 . Furthermore, we have previously observed that under the reaction conditions investigated in this study, methane participation is negligible during the aromatisation of 1-decene, rather, methane participation was only seen when employing small olefin substrates over a Ag-Ga/ZSM-5 catalyst 11 . N 2 control experiments were performed to confirm the role of methane during 1-decene upgrading over Ag-Ga/ZSM-5. Supplementary Table 13 shows no appreciable difference in olefin upgrading performance under N 2 and CH 4 environments. On this basis, methane is not thought to play any noticeable role in this study.
Diolefin and cycloolefin formation is proposed to be the ratelimiting step in olefin aromatisation over Ga/ZSM-5 catalysts 40 . Once activated upon the external surface protonic acid sites of ZSM-5 catalysts, olefin oligomerisation and cyclisation steps occur within the inner pores of the zeolite, confirmed by our support material control experiments. We have shown through our catalyst characterisations and computational analysis that DBT may effectively bind to surface metal species sites such as Ag and Ga and, in doing so, can block the access to the inner pores of Ag-Ga/ZSM-5 due to its bulky size. Consequently, at a reaction temperature of 330°C in a sulfur-containing reaction medium, the crucial oligomerisation of branched and linear monoolefins is prevented, resulting in little to no olefin reduction activity, this  Fig. 8 Proposed sulfur deactivation mechanism. Proposed mechanism for the deactivation of Ag-Ga/ZSM-5 via sulfur poisoning with DBT contaminant during direct upgrading of 1-decene at 330°C deactivation pathway is summarised in Fig. 8. Upon increasing reaction temperature much of the adsorbed DBT is thought to desorb from Ag and Ga species on the external surface, enhancing access to the inner pores and restoring catalyst activity. Additionally, TGA analysis of sulfur-adsorbed Ag-Ga/ZSM-5 strongly indicates that DES may be effectively chemisorbed to the surface of the catalyst but will readily desorb and decompose at 160°C and above, well below the reaction temperatures investigated in this study. Whereas DBT may be primarily weakly chemisorbed or physisorbed upon the surface of Ag-Ga/ZSM-5 and its desorption occurs at 330°C and above. The temporary nature of sulfur deactivation and subsequent restoration of olefin upgrading activity is reasonably proposed to be due to the desorption of weakly adsorbed DBT from the surface of Ag-Ga/ZSM-5 once charged in a sulfur-free medium at reaction temperatures of 330°C and above. Finally, we have also shown that in performing a simple NaOH treatment, mesoporosity may be introduced into Ag-Ga/ZSM-5, resulting in a greatly enhanced olefin reduction and aromatic formation during 1-decene upgrading at 330°C and that deactivation by bulky sulfur poisons such as DBT may be almost entirely mitigated.

Methods
Catalyst preparation. Ammonium type ZSM-5 (NH 4 -ZSM-5) obtained from Zeolyst, USA (CBV 8014) was converted to hydrogen type ZSM-5 (HZSM-5) by calcination in static air at 600°C for 3 h. A silver-and gallium metal-modified HZSM-5 reported in our previous studies 11,17 was prepared by wetness impregnation technique. In a typical synthesis, a certain amount of AgNO 3 (Sigma-Aldrich, 99.9+%) and Ga(NO 3 ) 3 ·xH 2 O (Sigma-Aldrich, 99.9%+) were dissolved in 10 g deionised (DI) water to form an aqueous solution of metal precursors and HZSM-5 support was then impregnated with the metal precursor to give the desired metal weight loading. The obtained wet powder was first dried in an oven at 92°C overnight, followed by calcination at 600°C for 3 h in static air to give the final catalyst.
NaOH-treated Ag-Ga/ZSM-5 was prepared following the method described in the literature 45 . In a typical synthesis, HZSM-5 was treated in an aqueous 0.2 M NaOH (30 cm 3 g −1 zeolite ) solution at 50°C for 30 min under constant stirring. The obtained powder was ion exchanged three times by stirring in an aqueous 0.1 M NH 4 NO 3 solution (100 cm 3 g −1 zeolite ) for 12 h and converted to its protonic form via calcination in static air at 550°C for 10 h (5°C min −1 ramp rate). The desired amount of Ag and Ga metals was then loaded onto the obtained NaOH-treated catalyst as per the wetness impregnation method described earlier in this section followed by calcination at 600°C for 3 h in static air.
Coked Ag-Ga/ZSM-5 was prepared using the method described in the literature 17,41 . In a typical synthesis, NH 4 -ZSM-5 was ion exchanged three times by stirring an aqueous 0.1 M Ca(NO 3 ) 2 ·4H 2 O (Alfa Aesar, 99%) (100 cm 3 g −1 zeolite ) at 80°C and then calcined in static air at 550°C for 2 h. The obtained powder was placed into a lab-fabricated flow reactor and treated at 600°C under N 2 flow for 2 h followed by an ethylene/N 2 mixture for an additional hour. The obtained sample was ion exchanged three times by stirring in an aqueous 0.2 M NH 4 NO 3 solution (100 cm 3 g −1 zeolite ) at room temperature and converted into its protonic form via calcination under N 2 at 600°C for 3 h. Finally, the desired amount of Ag and Ga metals was loaded onto the obtained coked catalyst as per the wetness impregnation method described earlier in this section followed by calcination at 600°C for 3 h under N 2 environment. SiO 2 -Al 2 O 3 mixed metal oxide support was prepared using tetraethyl orthosilicate (TEOS, 98%, Sigma-Aldrich) as the Si source and Al(NO 3 ) 3 ·9H 2 O (99.9%, Alfa Aesar) as the Al source through co-precipitation using NH 3 ·H 2 O as the alkaline agent. The molar ratio of SiO 2 /Al 2 O 3 was the same as the parent ZSM-5 support of Ag-Ga/ZSM-5. Ag-Ga/SiO 2 -Al 2 O 3 was then prepared as per the wetness impregnation method described earlier in this section followed by calcination at 600°C for 3 h in static air.
SiO 2 -MFI structure was prepared using hydrothermal synthesis method. A mixture of 22.66 g TEOS (98%, Sigma-Aldrich), 26.97 g Tetrapropylammonium hydroxide (TPAOH) (0.1 M, Sigma-Aldrich) and 22.65 g DI H 2 O was stirred for 2 h at 300 rpm and then transferred in an autoclave and held at 180°C for 3 days. The solid was recovered by centrifuge, dried and calcined at 550°C for 3 h in static air. Ag-Ga/Si-MFI was then prepared as per the wetness impregnation method described earlier in this section followed by calcination at 600°C for 3 h in static air.
Olefin upgrading reactions. Olefin upgrading tests were performed in a 300 mL batch reactor manufactured by Parr Instruments capable of tolerating high temperatures (up to 500°C) and pressures (up to 14 MPa). In a typical reaction, 8.0 g of 1-decene (Alfa Aesar, 96%) and 0.08 g of Ag-Ga/HZSM-5 were placed into the reactor and purged with N 2 three times. The reactor was then charged with 1.0 MPa CH 4 and heated to the desired reaction temperature (330-400°C) and held at reaction temperature for 1 h unless it is otherwise mentioned. Following reaction, the reactor was first ice cooled and then air cooled to room temperature. Once at room temperature, the liquid products were separated from solid catalyst via filtration. The resultant solid catalyst was then dried in a desiccator and either stored or recycled for reuse experiments. The liquid product oil composition was determined via analysis with a pre-calibrated gas chromatography mass spectrometer (GC-MS: PerkinElmer GC Claus 680 and MS Clarus SQ 8T) equipped with a paraffins-olefins-naphthenes-aromatics (PONA) column (Agilent HP-PONA). The oven temperature of the GC was programmed as follows: hold at 35°C for 15 min, ramp at 1.5°C min −1 to 70°C, ramp at 3°C min −1 to 150°C and hold for 30 min, finally, ramp to 250°C at 3°C min −1 and hold for 2 min. Liquid products were divided into four major groups: paraffins, olefins, aromatics and naphthenes.
Catalyst characterisation. XPS was performed using an AXIS-Ultra-DLD spectrometer (Kratos Analytical) using a mono-chromated Al Kα X-ray source (1486.6 eV). The XPS spectra of the selected elements were measured with a constant analyser pass energy of 80.0 eV. Calibration of all B.E.s was performed using the C 1s peak (B.E. = 284.6 eV).
HR-TEM was performed on FEI Talos F200X, a high-resolution scanning transmission electron microscope operated between 80 and 200 kV and equipped with high-angle annular dark field detector for Z contrast imaging and SuperX EDS detector for compositional analysis.
XANES measurements were performed at BL44A of the Taiwan Photon Source. This beamline is capable of acquiring quick XANES spectra at a speed of 500 ms per spectrum. A home-built reaction cell is installed at the beamline for in situ XANES measurement. All spectra were acquired in transmission mode. The reactant, composed of 90 wt% 1-decene, 5 wt% DES and 5 wt% DBT, was adsorbed on the catalyst pellet, which was put in the cell. The cell was purged by N 2 flow to remove the air in the cell. The cell was then sealed by closing the inlet and outlet valves. The temperature of the cell was controlled using an external heating system. The temperature was set from room temperature to 330°C at the rate of 60°C min −1 , and held at 330°C for 5 min. It was ramped to 400°C at a ramping rate of 20°C min −1 and kept at 400°C for 20 min. XANES spectra were collected every 500 ms throughout the course of reaction.
DRIFT was performed using a Thermo Scientific Nicolet iS50 equipped with environmental chamber and liquid-nitrogen cooled mercury-cadmium-telluride detector. The gas inlet of a multifunctional reactor system fabricated in our research laboratory was employed for the gas introduction to DRIFT via fluorinated ethylene propylene (FEP) tubing (Thermo Scientific, tubing 890 FEP, 8050-0125). The connection between the FEP tubing and the environmental chamber was accomplished through a short tubing with an inside diameter of 1/4 inch. The gas inlet was connected to a bubbler for liquid feedstock introduction via two three-way valves. Inlet gas could pass through the liquid sample before entering the chamber. First, the control group DRIFT spectra were acquired at 512 scans per spectrum with a resolution of 4 cm −1 . The catalyst was heated to 500°C and held for 30 min under 30 sccm N 2 to remove the impurities, which might be adsorbed on the catalyst surface during storage. When the catalyst was cooled down to 25°C the background spectrum was collected. Following background collections, a 30 sccm N 2 flow was introduced through the aforementioned bubbler filled with pyridine and kept for 20 min in order to introduce pyridine to the environmental chamber as the vapour phase. After that, the gas flow was switched back to 30 sccm pure N 2 . The catalyst was held at 25°C for 25 min before collecting sample DRIFT spectrum under transmission mode.
Surface acidity measurements were performed by NH 3 -TPD using ∼200 mg samples in an Altamira AMI-390 system. NH 3 was selected due to its simplicity, small molecular size and ability to titrate both strong and weak acid sites on the catalyst. Adsorption of NH 3 was performed using a flow of 30 sccm of 0.5% NH 3 / N 2 for 90 min. After flowing He at 30 sccm for 1 h to remove any physically adsorbed NH 3 , TPD was carried out by ramping to 600°C at 10°C min −1 increments and holding for 30 min. A thermal conductivity detector (TCD) determined the amount of desorbed NH 3 . Upon completion of each experiment, seven pulses of 5 mL of 10%NH 3 /He were used to calibrate the TCD response. For 1-decene and/or DES + DBT mixture TPD experiments, conditions were kept identical to those described above, however, He was used as carrier gas throughout and the desired compounds were adsorbed on the catalyst rather than NH 3 .
Identical TPD experiments were also performed using spent catalyst samples following reaction with 1-decene, DES and DBT, however no adsorption step was used.
TGA profiles were used to detect the desorption of chemisorbed species from Ag-Ga/ZSM-5 catalyst following pre-adsorption with DES and DBT. Preadsorption was achieved by placing 1 g of Ag-Ga/ZSM-5 catalyst into a solution of DBT dissolved in DBT (1:1 ratio by mass) and leaving in a sealed container overnight, the resultant catalyst sample was then dried in air at 80°C for 2 h to remove any water. TGA measurements were undertaken with a simultaneous thermal analyser (PerkinElmer STA 6000). The sample was held at 50°C for 1 min and then heated to 100°C at a rate of 10°C min −1 under 30 sccm N 2 flow, held for 30 min, heated to 330°C, held for 10 min and finally ramped to 500°C.
Surface area analysis was performed via N 2 physisorption analysis. The adsorption and desorption of N 2 on catalyst samples was measured using a Quadrosorb SI (Quantachrome Instruments). Samples were outgassed under vacuum at 350°C overnight and brought to 77 K by immersion in a liquid N 2 bath. Total surface area was calculated using a multi-point BET analysis. Pore surface area and pore volumes were calculated using Barrett-Joyner-Halenda analysis. The t-plot method was used with the DeBoer model for the calculation of the statistical thickness to distinguish the contribution of micropores (<2 nm) to the total surface area.
TOF-SIMS was performed using a TRIFT V nano TOF (Physical Instruments) using a 30 keV Au+pulsed primary ion source in bunched mode. The TOF-SIMS analysis on all samples were collected over an area of 400 µm × 400 µm with a total ion dose <1012 ions −1 cm −2 . Mass spectra were collected in the mass-to-charge range of 0-1850 m/z, for positive polarity and negative polarity. Charge compensation was accomplished using 10 eV electrons.
Computational calculations. DFT calculations were conducted by using DMol3 module in Materials Studio. B3LYP was selected as the exchange-correlation functional. The basis set DNP was used in the calculation. By using this basis set, the effect of occupied atomic orbital and valence atomic orbital have been considered and, a polarisation p-function on all hydrogen atoms was applied. Therefore, this basis set could provide reasonable accuracy with acceptable computational cost, which has been heavily used in DFT calculation for ZSM-5 [51][52][53] . During geometry optimisation, the tolerances of energy, gradient and displacement convergence were 2 × 10 −5 Ha, 4 × 10 −3 Ha Å −1 and 5 × 10 −3 Å. The initial ZSM-5 structure was obtained from the database of International Zeolite Association (IZA-SC) 33,54 . To conduct DFT calculation for non-periodic ZSM-5 cluster, the structure was hydrogen-terminated, which suggests all terminal Si and O atoms were saturated by hydrogen atom. During the geometry optimisation, the bond lengths of terminal O-H and Si-H were fixed at 0.95 Å and 1.49 Å, respectively. The initial structure has 96 T-sites. To satisfy a SiO 2 /Al 2 O 3 ratio of 23:1, 8 tetrahedral sites were replaced by Al atom. Based on the position of silicon atom, these T-sites can be classified into 12 groups, which were labelled from T1 to T12. Al atom is preferred located at T2 and T12 sites. To obey the aluminium avoidance principle, four T2 and T12 sites were occupied by Al 55 . The introduced Ag and Ga atom are located close to the Al to form [Al-O-M] (M = Ag, Ga) since metal atom has a preference to react with the Brønsted acid sites 56,57 .

Data availability
The data sets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.