Molybdenum Speciation and its Impact on Catalytic Activity during Methane Dehydroaromatization in Zeolite ZSM‐5 as Revealed by Operando X‐Ray Methods

Abstract Combined high‐resolution fluorescence detection X‐ray absorption near‐edge spectroscopy, X‐ray diffraction, and X‐ray emission spectroscopy have been employed under operando conditions to obtain detailed new insight into the nature of the Mo species on zeolite ZSM‐5 during methane dehydroaromatization. The results show that isolated Mo–oxo species present after calcination are converted by CH4 into metastable MoCxOy species, which are primarily responsible for C2Hx/C3Hx formation. Further carburization leads to MoC3 clusters, whose presence coincides with benzene formation. Both sintering of MoC3 and accumulation of large hydrocarbons on the external surface, evidenced by fluorescence‐lifetime imaging microscopy, are principally responsible for the decrease in catalytic performance. These results show the importance of controlling Mo speciation to achieve the desired product formation, which has important implications for realizing the impact of CH4 as a source for platform chemicals.

Abstract: Combined high-resolution fluorescence detection X-ray absorption near-edge spectroscopy, X-ray diffraction, and X-ray emission spectroscopyh ave been employed under operando conditions to obtain detailed new insight into the nature of the Mo species on zeoliteZ SM-5 during methane dehydroaromatization. The results showthat isolated Mo-oxo species present after calcination are converted by CH 4 into metastable MoC x O y species,w hich are primarily responsible for C 2 H x /C 3 H x formation. Further carburization leads to MoC 3 clusters,w hose presence coincides with benzene formation. Both sintering of MoC 3 and accumulation of large hydrocarbons on the external surface,e videnced by fluorescencelifetime imaging microscopy, are principally responsible for the decrease in catalytic performance.T hese results show the importance of controlling Mo speciation to achieve the desired product formation, which has important implications for realizing the impact of CH 4 as asource for platform chemicals.
The increasing availability of cheap natural gas has attracted growing interest towards direct routes for the conversion of methane into high-value chemicals. [1] Catalytic routes that have been investigated include dehydroaromatization, oxida-tive coupling,and partial oxidation, but are currently not (yet) economically viable. [2] One of these routes,m ethane dehydroaromatization (MDA), is particularly promising for the direct conversion of CH 4 into aromatic compounds and H 2 using metal-exchanged zeolites such as Mo/H-ZSM-5, since it contains acid sites as well as Mo species possessing dehydrogenation and CÀCc oupling functionalities. [1][2][3] It is generally accepted that CH 4 is activated on the exchanged Mo species, forming C 2 H 4 .S ubsequently,C 2 H 4 reacts on the (remaining) Brønsted acid sites and is converted into aromatic compounds,a lso leading to coke formation by the consecutive reaction of aromatic derivatives with light olefins. [3c,4] Although active species are proposed to originate from either (MoO 2 ) 2+ monomers or (Mo 2 O 5 ) 2+ dimers, [3a,c, 5] there is also adebate as to whether the active sites are oxidic,carbidic (MoC x ), or oxycarbidic (MoC x O y )i nn ature. [3a,d, 4, 6] Recently, combined UV/Vis absorption and Raman spectroscopies and DFT calculations have shown the formation of monomeric species upon calcination, demonstrating that the debate over the active sites is still ongoing. [7] In addition, there is no clear understanding of the catalyst deactivation mode,c onsidered to be the main limitation for the commercialization of the process. [1,2] Herein, we present an operando time-resolved combined X-ray diffraction (XRD) and high energy resolution fluorescence detection (K a -detected) X-ray absorption near-edge spectroscopy (HERFD-XANES) study during the MDA reaction on Mo/H-ZSM-5. Thea dvantage of using these techniques in combination is that local structure information around the Mo ions can be considered alongside changes in long-range order,t hat is,t he zeolite framework. Thus any change in catalytic performance can be immediately understood in terms of the structural evolution of the catalyst allowing us to pinpoint active and inactive species,r espectively,t hereby providing at enet for future catalyst development. Of particular importance is the application of scarcely used X-ray emission spectroscopy (XES), which is able to distinguish between ligands surrounding metal ions when they possess asimilar atomic number (Z), that is,Cversus O. [8] In this work we have used the Mo K b valence-to-core (vtc) emission bands (recorded on quenched samples) in conjunction with HERFD-XANES to be able to unambiguously determine the existence of both MoC x and MoC x O y during the course of reaction. Importantly,b ym easuring under operando conditions we have been able to put this species evolution into the context of the evolving catalytic activity. Figure 1a and Figures S2, 3i nt he Supporting Information are the MS data obtained during MDAat677 8 8C.

Shown in
Forease of discussion we delineate the MS response into the following time or "event" domains:formation of combustion products,l ight hydrocarbon evolution, and finally the formation of aromatic compounds.T he first responses in time concern the formation of combustion products ( Figure S3), that is,C Oa nd CO 2 in the form of an initial spike from 0-2 min time on stream (TOS) followed by ad ecrease and plateauing in the detected signal between 2-6 min. This is followed by asecond spike between 7-9 min (during this time period the CO response is significantly greater than that of CO 2 )a fter which the signals for these two components tend towards zero.Inthis second tranche of combustion products, H 2 Oi sa lso detected and exhibits as imilar, albeit weaker, response to that of CO.T he response for CH 4 by contrast is observed to climb in intensity up to 7min ( Figure S2), but is followed by asignificant decrease and by an upturn in signal intensity between 9-20 min. Between 20-73.5 min the signal for CH 4 reaches ap lateau before gradually decreasing.T he final trend in the product profile concerns the aromatic compounds ( Figure 1a), initially detected after 10 min, reaching significant quantities at 30 min followed by am onotonic decrease until the conclusion of the reaction. Figure 1b shows the corresponding Mo K-edge HERFD-XANES spectra of Mo/H-ZSM-5 zeolite,a cquired at the same time as the MS data. Additional HERFD-XANES data collected under ac ontrolled environment, before and after calcination, and after quenching the reaction at 4, 9, and 73.5 min are presented in Figure S5. All HERFD-XANES spectra are essentially dominated by astrong pre-edge peak at 20 008.5 eV attributable to a1 s-4d quadrupole/dipole transition and a1s-5p dipole transition at 20 025.1 eV followed by ar elatively featureless post-edge region;t his is consistent with the presence of Mo species dispersed within the zeolite that do not possess long-range order.Acursory comparison of the sample obtained after calcination with the spectrum of aNa 2 MO 4 reference sample, [9] however,shows ahigh degree of similarity consistent with the presence of monomeric [MoO 4 ] 2À species in agreement with the recent study by Gao et al. [7] We note however, that the presence of small amounts of dimeric Mo-oxo species cannot be readily excluded.
Clear changes in the HERFD-XANES data were detected during the entire reaction, including at otal shift of about 7.7 eV in the rising absorption edge ( Figure S6) and acirca 5% drop in the pre-edge peak, indicating areduction/ carburization of Mo with TOS. However, we observe that these changes,p articularly in the edge position, occur in stages which can be directly linked to the product evolution seen in the MS.T he spectrum recorded after 5.5 min of reaction (after the formation of the first set of combustion products) exhibits ac omparatively small change in edge position (about 1.5 eV), whereas the edge shift is of the order of 6.0 eV after 9.7 min (after the second spike in combustion products). Finally between 9.7-71.8 min av ery small shift of about 0.2 eV is observed ( Figures S6,7). These results suggest that Mo reduction/carburization takes place right up until the end of the experiment. Thes pectrum recorded at 71.8 min, particularly the pre-edge peak and edge position (20 005.7 eV), is strongly reminiscent of the trigonally coordinated Mo II -containing Mo 2 Cs tructure but without longrange order.H owever,t ov erify this and perhaps more importantly to discriminate between the stages of reduction versus carburization, it is necessary to examine the K b XES data.
K b XES experiments were performed at specific reaction times,the results of which are given in Figure 1c.Asaresult of the longer acquisition times required, data were collected on quenched samples (see the Supporting Information). The K b XES spectrum of MoO 3 was very similar to those of the zeolite-based samples before and after calcination, with one intense feature at about 19 961 eV attributed to the K b2 transition (4p!1s), and two weak bands at higher energies (19 980 and 19 996 eV), corresponding to K b'' and K b4 vtc transitions.A lthough K b4 might arise from either 4d to 1s or Mo pd ensity of states to 1s transitions, [8] K b'' has been suggested to originate from ligand 2s to metal 1s transitions. [8b] Conversely,t he spectrum of MoO 2 exhibited Following as imilar trend, the spectrum recorded after 4min of reaction contained aK b'' peak of reduced intensity (about 62 %weaker), which was also shifted in energy (circa 0.7 eV; Figure 1c). On the basis of previous work which showed that the intensity of K b'' decreases with increasing average Mo À Ob ond length whereas the energy position decreases with decreasing oxidation state,wepropose that at this stage ap artial reduction in the Mo oxidation state and ap artial replacement of Of or Cl igands occurs. [8b] Based on the observations in the K b XES spectra (particularly when considered against the spectra for MoO 3 ,M oO 2 ,and Mo 2 C), we propose that these results provide compelling evidence for the presence of partially carburized MoC x O y species.F rom the corresponding HERFD-XANES data, we estimate an average oxidation state of + 5( Figure S10) which would be consistent with the formation of MoC x O y complexes with a[MoCO](O fr ) 2 structure (Scheme 1; O fr indicates framework oxygens). Interestingly,the MS data showed the formation of C 2 H x /C 3 H x at this reaction time,demonstrating that MoC x O y species are also able to activate CH 4 ,y ielding entirely light hydrocarbons.
After 25 min, the K b'' band was not detected, while K b4 was further shifted to lower energies (about 3.5 eV; Figure 1c), resulting in as pectrum very similar to that of Mo 2 C ( Figure S13) and indicating the formation of trigonally coordinated MoC x (Scheme 1). [10] Thea bsence of aK b'' peak suggests an absence of Mo-O interactions;o nt his basis,w e conclude that these MoC x species are not attached to the zeolite framework. As indicated before,further changes in the HERFD-XANES were detected between the appearance of aromatic compounds and the end of the reaction, suggesting that Mo centers were not completely reduced or else that reduced Mo species sintered with TOS. Nevertheless,t he estimated oxidation state only appears to decrease from + 2.1 to + 2b etween 9a nd 73.5 min ( Figure S10), suggesting that the changes detected are primarily due to Mo agglomeration rather than further reduction. This is in line with the conclusion that MoC x species are not attached to the framework, as well as with previous studies showing the formation of carbide nanoparticles on the external surface,and smaller clusters on the zeolite channels. [11] Based on the HERFD-XANES/XESr esults,acomplete pathway for the evolution of the Mo species is given in Scheme 1. Isolated Mo-oxo centers present after calcination are converted into MoC x O y species during the initial contact with CH 4 ;t hese are still attached to the zeolite framework and present varying stoichiometry depending on the extent of the carburization. As carburization of Mo proceeds,MoC x O y species partially detach from the framework, being present as [MoC 3 ](O fr )c omplexes.L onger reaction times eventually lead to the transition to aM oc arbide phase,i nvolving the formation of MoC 3 sites not connected to the framework. As such, these MoC 3 species are not stable and easily agglomerate with TOS. Ultimately,s intering of MoC 3 leads to the migration of the clusters to the outer zeolite surface and the formation of larger Mo 2 C-like nanoparticles. [3a,11a] According to our observations,i ti sn ow possible to correlate the type of Mo species and the catalytic behavior. As shown in Figure 1a,a lthough the maximum benzene formation is only reached after the complete carburization of Mo,M oC x O y species formed at short reaction times are also able to activate CH 4 ,a se videnced by the formation of C 2 H x and C 3 H x .Itappears,however, that MoC x sites are necessary for the formation of aromatic compounds,inagreement with the proposal that MoC x species assist in C 2 H 4 conversion on the Brønsted sites. [12] In relation to this,the decrease observed in benzene formation after circa 30-35 min may be related to the agglomeration of Mo;s intering leading to ad ecrease in the active surface area, affecting dehydrogenation where which MoC 3 species are thought to play ar ole.I mportantly, this could be also due to adecrease in the amount of Brønsted sites (or else ar educed accessibility to these sites).
To further investigate the catalyst deactivation mode,the operando XRD data was next examined (Figure 2a). Thefirst observation of note is the lack of change in the XRD patterns with increasing TOS, suggesting that no new phases form and that the zeolite ZSM-5 structure maintains its stability.A closer inspection of the data ( Figure S14) also reveals no shift, broadening,o rr eduction in Bragg reflection intensity,w hich strongly suggests that neither dealumination (unit-cell contraction) or else accumulation of carbon within the micropores (lattice expansion) occurs. [13] However,t he catalyst recovered after 73.5 min is black in color,s uggesting that if carbon deposition inside the pores does not occur,t hen it must certainly do so outside on the external zeolite surface.
To study the possible build-up of carbon, fluorescencelifetime imaging microscopy (FLIM) measurements were performed on the samples recovered at different reaction times,s ince fluorescence microscopy alone is unlikely to differentiate between species with similar emission characteristics.Asseen in Figures 2b and Figures S17-19, only scattering species and patches of species with as hort fluorescence lifetime (about 150 ps) were detected on the calcined sample,

Angewandte Chemie
Communications whereas amixture of both short-and long-lived species (with lifetimes of ns-ms) were detected on the reacted samples, including some particularly long-lived species that did not decay within the 10 ns FLIM window.A st he reaction time increased, the concentration of the very long-lived species was also found to increase,supporting the idea that fluorescence is emitted from complex carbon species on the external surface, in line with previous reports. [14] This may be favored by the absence of steric limitations and promoted by the external acid sites and perhaps also suggestive of Al gradients in the zeolite crystals. [15] Thep resence of carbon on the external surface will certainly limit accessibility to the Brønsted sites inside the channels,i nfluencing the selectivity towards benzene.F urthermore,f rom the image at 73.5 min, it can be seen that in addition to being outside of the zeolite pores,the deposited carbon shows am icroscale distribution and is located at the periphery of the zeolite particles.T herefore, both sintering of Mo and build-up of carbon contribute to the gradual deactivation of the catalyst. Note,h owever, that although steric hindrance would prevent carbon build-up within the pores,wecannot rule out that this occurs at longer reaction times.
In summary we have shown an ovel combination of techniques,c oupling for the first time XRD and HERFD-XANES/XES experiments to investigate,u nder real operando conditions,the direct conversion of CH 4 on bifunctional zeolite catalysts.T his approach has provided important new insight regarding the need to control, or else maintain, Mo active species to achieve the desired product formation.
Although highly transient, the appeal of stabilizing MoC x O y species is that they are highly selective to light hydrocarbons. Although there are issues with stabilizing MoC x O y in the presence of the reaction mixture,itispossible that this could be achieved for example in the presence of co-fed H 2 Oand or O 2 . [16] MoC 3 on the other hand are the species to target when we wish to produce aromatic compounds but again either cofeeding of oxidants to mitigate carbon deposition or else enhancing the interaction with the zeolite is needed for improved stability.Perhaps,however,the most salient observation is that this multi-technique operando approach has been able to correlate the evolution of active species with distinct reaction products,a llowing us to identify clearly the way forward in helping translate the potential of these important catalysts into areality.