Elucidating the structure of the W and Mn sites on the Mn-Na 2 WO 4 /SiO 2 catalyst for the oxidative coupling of methane (OCM) at real reaction temperatures

.


Introduction
The increasing reserves of natural gas and the lack of an efficient industrial process for methane upgrading have renewed interest in the oxidative coupling of methane (OCM) [1,2].Since the unprecedented work of Keller and Bhasin in 1982 [3], OCM has been known as a promising route to convert methane directly into C 2 hydrocarbons (ethane and ethylene).However, the industrial deployment of the OCM route remains limited by the low C 2 yield resulting from inherent thermodynamic and kinetic limitations.Indeed, the high stability of methane and the faster formation of more thermodynamically stable CO x compounds compared with that of C 2 hydrocarbons result in low, not economically competitive, C 2 yields [2,4,5].Among hundreds of materials tested for OCM, Mn-Na 2 WO 4 /SiO 2 is considered as the state-ofthe-art catalyst, exhibiting high stability (~500 h on stream) and C 2 yields (14-27%) [6,7].However, for the OCM route to be economically viable, a single-pass C 2 yield of 30% and a C 2 selectivity of 90% are necessary [8,9].
Although a fundamental understanding of the Mn-Na 2 WO 4 /SiO 2 catalyst is crucial to design improved formulations, the literature remains unclear about the nature of the active sites and the mechanism for the selective activation of methane.On the one hand, several studies have ascribed the performance of the Mn-Na 2 WO 4 /SiO 2 catalyst for OCM to W 6+ sites with distorted tetrahedral (T d ) oxygen coordination [10][11][12].However, the proposed mechanisms for the selective activation of methane on these sites are inconsistent.Wu et al. suggested that the activation of methane occurs at lattice oxygen in the T d -W 6+ sites via a W 6+ MW 4+ redox cycle, where gas-phase oxygen completes the cycle [13].Conversely, Jiang et al. proposed a W 6+ MW 5+ redox cycle along with a previous activation of gas-phase oxygen at the Mn 3+ sites with octahedral (O h ) oxygen coordination [14].The O h -Mn 3+ would act as oxygen storage-release sites via a Mn 2+ MMn 3+ redox cycle, supplying activated oxygen species to the T d -W 6+ sites.On the other hand, Lunsford et al. proposed that the active sites are the Mn-O-Na sites instead of T d -W 6+ [15].The activation of gas-phase oxygen would occur at the Mn-O-Na sites via dissociative adsorption, forming Mn-O-NaÁÁÁO Á sites, where methane then activates by cleaving C-H bond.However, W appears as a crucial performer: Elkins et al. recently showed the poor OCM performance of catalysts without W [16].The absence of Na 2 WO 4 , Mn 2 O 3 , or acristobalite phases has been associated with worse OCM performances (i.e., lower C 2 yields) [17][18][19][20].The interplay between W, Mn and Na oxides on silica appears critical for this reaction.
The identification of active species by XRD is limited to the observation of crystalline phases with size larger than ca. 4 nm, being unable to report the potential presence of other structures such as (i) molecularly dispersed oxides, (ii) amorphous phases, (iii) nanocrystallites smaller than ca. 4 nm, or (iv) molten phases.Additionally, some crystalline phases with weak diffraction pattern intensities, such as Mn 2 O 3 , may be overwhelmed by the intense pattern of other crystalline phases, such as Mn 7 SiO 12 .Structureactivity/selectivity relationships based on ex situ XRD evidence may be inadequate because the crystalline phases identified at room temperature may not be the ones present during reaction.Actually, Mn-Na 2 WO 4 /SiO 2 is highly temperature-dependent: crystalline Na 2 WO 4 and a-cristobalite phases undergo transitions during heating, as evidenced by differential scanning calorimetry (DSC) [23], high energy X-ray diffraction computed tomography (XRD-CT) [24,25] and Raman spectroscopy [21,26].The XRD-CT studies fail to reveal the structure of the W sites at temperatures above 680 °C due to molten phases [25] and the X-ray absorption near edge structure (XANES) spectroscopy measurements were conducted in ex situ conditions [27].Thus, determining the structure of the catalyst at temperatures where the OCM reactions can thermodynamically occur is critical to understand the nature of the species involved in the catalytic cycle.
While the Mn-Na 2 WO 4 /SiO 2 system is very dynamic, only a very recent OCM contribution provides complete information about its dynamic states present at relevant OCM temperatures (>700 °C) [21].This work underscores the relevance of characterizing during activation and reaction to: (i) monitor the mean oxidation state and molecular geometry (distortion degree) of the W and Mn sites to reveal the potential existence of temperature-induced structural changes and (ii) evaluate the role of each catalyst component (silica support, W, Mn, and Na sites).
This work presents a systematic study to elucidate the electronic and molecular structure of the W and Mn sites on the Mn-Na 2 WO 4 /SiO 2 catalyst at relevant OCM reaction temperatures.For this purpose, the mean oxidation state, site symmetry, and distortion degree of the W and Mn sites on the conventional 2 wt% Mn-5 wt% Na 2 WO 4 /SiO 2 catalyst, and on Mn-free 5 wt% Na 2 WO 4 /SiO 2 , and Mn-and Na-free 3.1 wt% WO 3 /SiO 2 catalysts were determined via XRD and Raman and XANES spectroscopies at room temperature (RT) and in situ temperature-programmed oxidation (TPO).This study contributes to elucidating the nature of the active sites of this complex system.

Catalyst characterization
The chemical composition was measured via inductively coupled plasma optical emission spectroscopy (ICP-OES) using a Plas-maQuant PQ 9000 spectrometer from Analytik Jena.The specific surface area was calculated via nitrogen (N 2 ) physisorption experiments using an ASAP 2020 (Micromeritics) instrument at liquid N 2 temperature.The catalysts were degassed at 300 °C (10 °CÁmin À1 ) for 24 h before the adsorption experiments.The specific surface areas were calculated using the multiple-point Brunauer-Emmett-Teller (BET) method to analyze the N 2 adsorption isotherms in a relative pressure range of 0.05-0.3.
X-ray powder diffraction (XRD) measurements were performed on a PANalytical X'Pert Pro diffractometer using CuK a radiation (k = 1.5406Å) generated at 40 kV and 30 mA as the X-ray source.Crystalline phases were identified using the JCPDS (Joint Committee on Powder Diffraction Standards) database.RT-XRD patterns were collected in a Bragg angle (2h) range of 5-90°with a step size of 0.02°and a counting time of 50 s per step.In situ TPO-XRD patterns were isothermally recorded in a high-temperature chamber (Anton Paar XRK900) every 50 °C from 50 °C to 800 °C during heating at 10 °CÁmin À1 under 100 cm 3 Ámin À1 flow (molar O 2 :Ar = 1:4).In situ measurement conditions were: 2h range of 5-90°, step size of 0.05°, and a step counting time of 20 s.
In situ TPO-Raman spectra were taken with a confocal Renishaw inVia Qontor instrument equipped with a cooled CCD detector and three laser excitations (785, 514, and 405 nm).The 405 nm laser was chosen to minimize potential sample fluorescence and register Raman spectra at high temperatures.The spectral resolution was near 1.5 cm À1 , and the wavenumber calibration was checked using the silica standard band at 520.5 cm À1 .The laser was focused on the catalysts with a confocal microscope using an ultralong working distance x20 objective (Olympus LMPlanFL N 20X).Typically, ~0.05 g of each catalyst (180-250 lm) was loaded into a Linkam reaction cell that consists of a fixed-bed microreactor with a quartz window and O-ring seals, which was cooled by flowing water.The laser power on the catalyst was kept below 5 mW to prevent local heating.For the in situ TPO-Raman study, each catalyst was initially dehydrated under 60 cm 3 Ámin À1 flow of O 2 :Ar (molar ratio 1:2) at 200 °C (10 °CÁmin À1 ) for 30 min and then cooled down to 25 °C.The catalyst was then heated to 800 °C (10 °CÁmin À1 ) under the same oxidizing gas flow with simultaneous acquisition of Raman spectra (1 accumulation of 20 s) every 25 °C.The TPO-Raman study mimics the heating process typically used in the literature for testing the steady-state OCM performance of the conventional Mn-Na 2 WO 4 /SiO 2 catalyst [15,17,18].The spectra were analyzed using PEAXACT software.

Steady-state OCM catalytic tests
Steady-state reaction studies were performed in a fixed-bed catalytic reactor.Typically, 0.1 g of each catalyst (180-250 lm) was diluted with 0.4 g of carborundum (SiC, 180-250 lm, Sigma-Aldrich) and supported in a quartz U-shape reactor tube (I.D. = 10 mm) between two pieces of quartz wool.The reactor was placed into an electric furnace, and the reaction temperature was measured and controlled using a thermocouple attached to the outside wall of the reactor in a position corresponding to the middle of the catalyst bed length.Both inlet and outlet gas lines were heated at 120 °C to prevent condensation.Gaseous products were analyzed using an on-line gas chromatograph (Shimadzu GC-2014) equipped with two channels of separation and detection: (i) a polar carboxen 1010 PLOT capillary column (30 m Â 0.32 mm #35789-02A) coupled with thermal conductivity (TCD) and flame ionization (FID) detectors; and (ii) a CP-Molsieve 5A column (25 m Â 0.53 mm #CP7538) coupled with a TCD detector.Before starting the reaction, each catalyst was heated up to reaction temperature under 60 cm 3 Ámin À1 of a mixture O 2 :He:N 2 (molar ratio 2:3:1).Such is the heating process typically used before testing the OCM activity of the Mn-Na 2 WO 4 /SiO 2 catalyst because increasing temperature under oxidizing conditions has been proposed to enhance the C 2 selectivity [15,17,18,31].The reaction started by feeding 73 cm 3 Ámin À1 of a mixture CH 4 :O 2 :He:N 2 (molar ratio 2:1:1.5:0.5) to the reactor.N 2 was used as a GC internal standard.The OCM catalytic activity was measured at 650, 700, 750, and 800 °C after 30 min to allow the reaction to reach the steadystate at each temperature.The water produced during the catalytic test was separated from the gaseous products before entering the GC using a condenser trap cooled by a cryostat.The conversion of methane (X CH4 ) and the selectivity to C 2 hydrocarbons (S C2 ) were calculated from the mole number of the detected compounds, derived from a carbon balance of the system as follows: Replicate experiments involving reloading catalyst into the reactor resulted in standard deviations near 0.9% for both X CH4 and S C2 .The turnover frequency (TOF) was calculated for each catalyst at 650, 700, 750, and 800 °C as the number of reacted CH 4 moles per mol of W in the reactor per second, as described in the Supporting Information.It was assumed that all W atoms of the catalyst, calculated from the W concentration measured by ICP-OES analysis (Table S1), interact with methane.
XRD analysis.The qualitative phase identification of the synthesized catalysts was performed via XRD analysis at room temperature (RT) and temperature-programmed oxidation (TPO).Fig. 1 shows the RT-XRD patterns of the WO 3 /SiO 2 , Na 2 WO 4 /SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts.
The WO 3 /SiO 2 catalyst maintains the amorphous silica phase of the support, exhibiting only diffraction peaks related to the crystalline monoclinic (space group P2 1 /n) c-WO 3 phase (Fig. 1a).
The thermal stability of the crystalline phases identified at room temperature was monitored by TPO-XRD analysis.Fig. 2 shows the in situ TPO-XRD patterns of the WO 3 /SiO 2 catalyst and the diffraction patterns of some WO 3 polymorphs that can form upon heating.The qualitative phase identification is challenging due to the low crystallite size of the WO 3 NPs (loading slightly above ''monolayer" coverage) and the faster step counting time compared with the RT-XRD measurements.However, the crystalline WO 3 phase can be monitored by following two groups of the main diffraction peaks in the 22°< 2h < 25°and 32°< 2h < 34°windows.Upon heating, the diffractograms hardly change until 400 °C, when the three diffraction peaks at 32°< 2h < 34°exhibit similar intensity, suggesting the presence of the orthorhombic (Pnma) b-WO 3 phase.A distinct diffraction pattern characterized by two peaks at 2h = 22.88°and 23.87°, and 2h = 33.23°and33.84°is observed at 650 °C, suggesting the presence of the tetragonal (P4/ncc) a-WO 3 phase, which remains stable until 800 °C.These temperature-induced (monoclinic) c M (orthorhombic) b M (tetragonal) a-WO 3 phase transitions are reversible [39,40].On the other hand, the diffraction peaks of the cubic Na 2 WO 4 phase disappear at 600 °C, while the diffraction peaks of the lower symmetry orthorhombic Na 2 WO 4 phase arise and remain present until 650 °C.Further heating leads to the complete disappearance of Na 2 WO 4 reflections, suggesting its melting.DSC and Raman spectroscopy results reported in the literature show that these transitions are reversible [21,23], which emphasizes the importance of in situ studies, as fresh or spent catalysts at room temperature would not provide relevant information about the catalyst structure during OCM.
At room temperature, the dehydrated WO 3 /SiO 2 catalyst exhibits Raman bands typically ascribed to the crystalline WO 3 , in agreement with RT-XRD (Fig. 4a, left) [42].The crystalline WO 3 phase has three major Raman bands at 805, 713, and 265 cm À1 , corresponding to the symmetrical (m s ) and asymmetrical (m as ) stretching modes, and the bending (d s ) mode of the bridging W-O-W bond, respectively [43].The surface density of W atoms on the WO 3 /SiO 2 catalyst (~0.7 W-atomsÁnm À2 ) is above de dispersion limit surface coverage.Thus, the presence of the dispersed WO x phase is unlikely, as confirmed by Raman spectroscopy (see Fig. S4).
XRD and Raman uncover W sites with different environments; the W sites are octahedrally coordinated in the WO 3 /SiO 2 (O h -W 6+ sites in the WO 3 phase) and tetrahedrally coordinated in the Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 (T d -W 6+ sites in the Na 2 WO 4 phase) [43,49].The position of the most intense Raman band denotes the highest bond order (shortest W-O bond) of tungsten species [49].Thus, the crystalline WO 3 phase (O h -W 6+ sites) exhibits the most intense Raman band, related to the symmetrical (m s ) stretching mode, at lower wavenumber (805 cm À1 ) than the crystalline Na 2 WO 4 phase (T d -W 6+ sites) (ca.926 cm À1 ) due to the lower bond order of the former [43,49].The different symmetry of the W sites in the synthesized catalysts is due to the presence of Na + cations [12]  Na-promoted tungsten oxide catalysts [26]: the W sites exhibited T d symmetry in the dispersed Na-WO 4 and the crystalline Na 2 WO 4 phases.
Temperature-induced structural modifications of the W sites can be followed by measuring the position of the m s vibration band because changes in the oxidation state and/or distortion degree impose variations on the bond order of tungsten species [49].Concerning the Na 2 WO 4 phase transitions: the m s vibration of the Na 2 WO 4 /SiO 2 catalyst red shifts with increasing temperature but undergoes two sharp blue shifts near 650 and 750 °C, suggesting successive temperature-induced cubic ?orthorhombic ?molten Na 2 WO 4 transitions.The cubic ?orthorhombic Na 2 WO 4 transition is also corroborated by the blue shift of the d s vibration from 309 to 319 cm À1 and the split of the m as vibration band at 810 cm À1 into two smaller ones at 847 and 820 cm À1 (Fig. 4b, left) (for a more detailed view see Fig. S5) [50,51].The bands at 923, 847, 820, and 319 cm À1 related to the orthorhombic Na 2 WO 4 phase loose intensity with a further temperature increase and the m s band blue shifts, suggesting the melting of the Na 2 WO 4 phase near 750 °C (Fig. 4b, right).
The temperature-induced c ? b ?a-WO 3 , a ?b-cristobalite, and cubic ?orthorhombic ?molten Na 2 WO 4 transitions observed by the TPO-Raman study take place at higher temperatures than those observed by TPO-XRD analysis.This may be due to the different heating ramps imposed by the equipments.Note that, for instance, the complete formation of the orthorhombic Na 2 WO 4 phase requires some 4 min or 40 °C in a continuous 10 °CÁmin À1 heating ramp under oxidizing conditions to occur, according to DSC measurements [21].
Monitoring the structure of the W sites on the conventional Mn-Na 2 WO 4 /SiO 2 catalyst during the heating via Raman spectroscopy is challenging due to the broadness of the bands (Fig. 4c, left).The UV-vis spectra of the three catalysts (Fig. S6) show that the addition of Mn increased the absorption (broad band near 400 nm).The sequential cubic ?orthorhombic ?molten Na 2 WO 4 transition in the Na 2 WO 4 /SiO 2 catalyst is also apparent in the dehydrated Mn-Na 2 WO 4 /SiO 2 catalyst.The m s vibration shows an overall red shift with increasing temperature up to near 700 °C, and then, two sudden blue shifts above 700 and 780 °C, associated with the sequential cubic ?orthorhombic ?molten Na 2 WO 4 transitions (Fig. 4c, right).These transitions take place at higher temperatures than on the Na 2 WO 4 /SiO 2 catalyst (above 650 and 750 °C, respectively), suggesting the relevance of Mn on the thermal evolution of W sites.The a ?b-cristobalite phase transition and the presence of the crystalline Mn 2 O 3 phase on the dehydrated Mn-Na 2 WO 4 /SiO 2 catalyst throughout the heating had to be elucidated via chemometric component analysis using indirect hard modeling method with PEAXACT software (Fig. S7).
The a ?b-cristobalite phase transition occurs near 250 °C, as in the Na 2 WO 4 /SiO 2 catalyst, while the crystalline Mn 2 O 3 phase remains stable up to 800 °C, in agreement with the TPO-XRD findings elsewhere.
Although the cubic ?orthorhombic ?molten Na 2 WO 4 phase transitions have already been reported in the literature, understanding the structure of the W sites in the molten Na 2 WO 4 phase is crucial because this phase is present at temperatures where the OCM reactions are thermodynamically relevant (>700 °C).
In situ TPO-XANES Spectroscopy.The temperature-induced electronic and structural changes of the W and Mn sites were also studied via in situ TPO-XANES spectroscopy to complete the characterization and elucidate if the shift of the m s vibration band observed in the TPO-Raman study is due to a change in the oxidation state, the distortion degree of the W sites, or both.depicts the in situ XANES spectra at the W-L 3 edge of the bulk WO 3 , Na 2 WO 4 , and WO 2 reference materials recorded at 25 °C and the WO 3 /SiO 2 , Na 2 WO 4 /SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts recorded at 25, 700, 750, and 800 °C.Insets in the upper right-hand corner show a more detailed view of the edge region.
The electron transition from the 2p 3/2 state to a vacant 5d state appears as an intense dipole-allowed peak in the absorbance curve, known as the ''white line" [52].To calculate the mean oxidation state of the W atoms, the W-L 3 edge position (i.e., the absorption threshold) for the bulk WO 3 , Na 2 WO 4 , and WO 2 reference materials and the WO 3 /SiO 2 , Na 2 WO 4 /SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts was determined by using the intersection with the energy axis of the second derivative with respect to the energy of the normalized XANES curve.Certainly, the energy shift on the W-L 3 edge position in tungsten species is due to differences in the mean oxidation state of the W atoms [53,54].
The W-L 3 edge position measured for the bulk WO 3 , Na 2 WO 4 , and WO 2 reference materials was 10205.3,10205.6, and 10203.3eV, respectively.Thus, the W-L 3 edge position depends not only on the mean oxidation state, WO 3 (W 6+ ) vs. WO 2 (W 4+ ), but also on the local structure of the W atom, WO 3 (O h -W 6+ ) vs. Na 2 WO 4 (T d -W 6+ ), in agreement with the literature [52].Therefore, to estimate the mean oxidation state of the W sites at each temperature, the W-L 3 edge position measured for the WO 3 /SiO 2 catalyst was interpolated to a straight line between the edge values recorded for the bulk WO 3 (O h -W 6+ ) and WO 2 (O h -W 4+ ) reference materials, as shown in Fig. S8a, and that of the Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 catalysts was calculated using the bulk Na 2 WO 4 (T d -W 6+ ) reference material instead of the WO 3 (O h -W 6+ ), Fig. S8b.Table 1 summarizes the edge position and mean oxidation state of the W sites on each catalyst at 25, 700, 750, and 800 °C.
All synthesized catalysts exhibit a lower W-L 3 edge position than the bulk WO 3 and Na 2 WO 4 reference materials at room temperature, suggesting that the mean oxidation state of the W sites is lower than 6+, in line with the known chemistry of WO 3 [55]: its polymorphic nature and tendency to form Magnéli phases, which stabilize cations with oxidation states below W 6+ .Thus, the WO 3 / SiO 2 , Na 2 WO 4 /SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts have W 6+ sites with some electron polarons (i.e., W 5+ at a W 6+ site).The formation of oxygen-deficient W oxides by single self-trapped electron polarons in small amounts is expected in the crystalline WO 3 and Na 2 -WO 4 phases [56][57][58][59].On the other hand, the increasing temperature under oxidizing conditions did not shift the W-L 3 edge position out of the error in the energy resolution for any catalyst.Thus, the electronic state of the W sites remains as W 5+ -W 6+ sites at relevant OCM temperatures.To summarize, the in situ TPO-XANES spectra indicate that the temperature-induced m s vibration band shift observed in the TPO-Raman study is related to a variation in the distortion degree of the W sites rather than a variation in the oxidation state.
The intensity and shape of the white line also provide information about the oxidation state and symmetry of the W sites.On the one hand, the intensity is associated with the density of unoccupied states, and therefore with the oxidation state.Thus, the white line intensity of the WO 3 (W 6+ ) is higher than that of the WO 2 (W 4+ ) [52].However, the higher intensity evidenced for the synthesized catalysts compared with the bulk WO 3 and Na 2 WO 4 reference materials (even at room temperature) cannot be associated with a higher oxidation state of the W sites, since the edge position remained unchanged.This effect is actually related to the interaction of the WO 3 and Na 2 WO 4 NPs with the support, in agreement with the literature [60].Garcia-Lopez et al. recently observed that the white line intensity of the silica-supported Keggin [PW 12 -O 40 ] 3À and Wells À Dawson [P 2 W 18 O 62 ] 6À heteropolyanions were higher than that of the unsupported heteropolyanions [60].On the other hand, the white line shape depends on the symmetry and distortion of the W atoms [61,62].Although both WO 3 and Na 2 WO 4 reference materials have the same oxidation state (W 6+ , d 0 ), they present different white line shapes.The crystalline WO 3 phase (O h -W) exhibits a broad peak with an indistinct top (Fig. 5a), whereas the Na 2 WO 4 (T d -W) exhibits a sharper, more asymmetrical peak (Fig. 5b-c).These differences are due to the splitting of the W 5d state by the ligand field [61,62].Thus, the broad white line of the O h -W sites consists of two peaks related to the t 2g (d xy , d yz , and d zx ) and e g (d x 2 -y 2 and d z 2) orbitals, whereas the narrow white line of the T d -W sites is due to a smaller splitting of the W 5d state and consists of two peaks related to the e and t g orbitals [60][61][62][63].
To analyze the temperature-induced changes in the distortion degree of the W sites, the white line shape of the synthesized catalysts at each temperature was interpreted by considering the final state of the 5d orbitals following a methodology proposed by Yamazoe et al. [61].The white line was deconvoluted by representing each electron transition to a vacant 5d split orbital with a Lorentz function and the vacuum level with an arctangent function.Two Lorentz peaks, at lower and higher energy, related to the t 2g and e g orbitals for the O h -W 6+ sites or the e and t g orbitals for the T d -W 6+ sites, were adjusted.The two peaks in the second derivative of the bulk WO 3 and Na 2 WO 4 reference materials were considered as the input center values for the Lorentz peaks in catalysts with O h -W 6+ and T d -W 6+ sites, respectively (Fig. S9).The peak ratio of t 2g /e g = 3/2 and e/t g = 2/3 were considered as constraints.Irrespective of the temperature, because the oxidation state of the samples did not change, the arctangent functions adjusted for the bulk WO 3 and Na 2 WO 4 reference materials were used to deconvolute the white line of the WO 3 /SiO 2 catalyst and of the Na 2 WO 4 and Mn-Na 2 WO 4 /SiO 2 catalysts, respectively.The energy gap between both peak centers reflects the 5d state splitting.This analysis was similar to that carried out in previous works reported in the literature [60][61][62][63][64]. Fig. S10 shows a representative example of the fit performed for the bulk WO 3 reference material.The results of all the spectra deconvolutions are reported in Table S3, and Fig. 6 contrasts the measured energy gap values with the phases identified by TPO-XRD at each temperature range.
The energy gap for the bulk WO 3 and Na 2 WO 4 reference materials at room temperature is 4.4 and 1.6 eV, respectively (Table S3), in agreement with the values reported in the literature (4.5 and 1.7 eV, respectively) [61,64].Thus, the O h -W sites exhibit a larger splitting of the W 5d state than the T d -W sites, as discussed before.Regarding the general decrease of this parameter observed with temperature, some works attribute the decrease in the energy gap to an increase in the distortion degree of the O h -and T d -W sites [60,61].The distortion of regular O h symmetrical units solves the degeneracy of 5d orbitals and results in smaller splitting of the 5d orbitals [62].In the WO 3 /SiO 2 catalyst, the temperature-induced phase transition of the crystalline WO 3 phase observed in our TPO-Raman study suggests changes in the distortion degree of the O h -W sites.The transition from the highly distorted monoclinic c-WO 3 phase at room temperature to the ideal undistorted cubic WO 3 phase at very high temperatures (~1500 °C) in the sequence c (monoclinic) ?b (orthorhombic) ? a (tetragonal) is associated with changes in the W-O bond length, octahedral tilt, and/or displacement of the W atom out of the center of the octahedron [65,66].The literature recognizes that the tetragonal a-WO 3 phase is less distorted than the monoclinic c-WO 3 system because it presents lower tetrahedron tilting and W displacement from the octahedra center [66].However, Wang et al. recently observed that the distortion imposed on the O h -W sites by doping WO 3 with Rb atoms resulted in a longer W-O bond in the z-axis [65].Therefore, the energy gap decrease observed throughout the d ?c ? b ?a-WO 3 phase transition might be explained by the elongation of the W-O bond in the z-axis.The energy gap of the T d -W 6+ sites narrows after the Na 2 WO 4 melting (>698 °C) for Na 2 WO 4 /SiO 2 and widens in Mn-Na 2 WO 4 / SiO 2 .Thus, although both catalysts have equal T d -symmetry, the T d -W sites on Mn-Na 2 WO 4 /SiO 2 exhibit a wider energy gap (2.2 vs. 1.6 eV at 800 °C), and therefore a lower distortion.After comparing the chemical, textural, and structural properties of these catalysts, it is evident that the exclusive presence of the O h -Mn 3+ sites in the molten Na 2 WO 4 phase may be related to the lower distortion degree of the T d -W sites observed for Mn-Na 2 WO 4 /SiO 2 catalyst.Recent studies have shown that Na + and WO 4 2-ions are unstable, mobile, and interact with other materials after Na 2 WO 4 melting [24].Thus, the interaction of the O h -Mn 3+ sites with the WO 4 2-ions (T d -W 6+ sites) will most likely be present on the catalyst surface wetted with the molten Na 2 WO 4 phase.Finally, to confirm the presence of the O h -Mn 3+ sites and fully understand the molecular and electronic structure of the Mn-Na 2 WO 4 /SiO 2 catalyst at relevant OCM temperatures, we also performed in situ TPO-XANES spectroscopy at the Mn-K edge.Fig. 7 shows the in situ XANES spectra at the Mn-K edge of the bulk MnO, Mn 3 O 4 , Mn 2 O 3 , and MnO 2 reference materials recorded at 25 °C (Fig. 7a) and the Mn-Na 2 WO 4 /SiO 2 catalyst at increasing temperatures up to 800 °C (Fig. 7b).
We can easily observe that the Mn-K edge XANES spectra of the Mn-Na 2 WO 4 /SiO 2 catalyst resemble the spectra of bulk Mn 2 O 3 reference material at 25 °C.Although the in situ TPO-Raman spectra of the Mn-Na 2 WO 4 /SiO 2 catalyst exhibits a band related to the mixed-valent hausmanite Mn 2+ Mn 3+  2 O 4 phase, the Mn 3+ sites dominate the XANES spectrum.Upon heating, the Mn-K edge XANES spectra of the Mn-Na 2 WO 4 /SiO 2 catalyst keep almost unchanged, suggesting that Mn atoms are mostly present as O h -Mn 3+ sites at relevant OCM temperatures [24].Therefore, O h -Mn 3+ and T d -W 6+ electronic states are the most likely cation configuration of the Mn and W sites on the conventional Mn-Na 2 WO 4 /SiO 2 catalyst at real OCM temperatures, just before methane admission into the reactor.

Steady-State OCM catalytic tests
Because the most drastic structural variations under oxidizing conditions of the T d -W 6+ sites were observed above 650 °C, the steady-state OCM catalytic tests were performed at 650, 700, 750, and 800 °C.Fig. 8 depicts the steady-state OCM performance in terms of methane conversion (X CH4 ) and selectivity to C 2 hydrocarbons (S C2 ) for WO 3 /SiO 2 , Na 2 WO 4 /SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts and the blank test (an empty reactor with no catalyst under the same conditions), together with the phase transitions of W sites evidenced in our TPO-XRD analysis.The blank test (homogeneous reaction) is selective, albeit not very active; the presence of catalytic materials affords higher conversions, being selectivity dependent on the specific catalyst composition.Thus, at 650 °C, all catalysts exhibit higher S C2 and similar X CH4 than the blank test (X CH4 = 0.6% and S C2 = 25.8%),excluding the Mn-Na 2 WO 4 /SiO 2 catalyst that exhibits a X CH4 = 1.8%, which is above the experimental error at this temperature (±0.4%).At 700 °C, the S C2 in the blank test increases to 54.7%, but the X CH4 remains almost negligible (0.7%), Fig. 8a; the WO 3 /SiO 2 and Na 2 WO 4 /SiO 2 catalysts exhibit an increase in X CH4 (1.8 and 1.4%, respectively) and S C2 (45.1 and 48.1%, respectively), Fig. 8b-8c; and the Mn-Na 2 WO 4 /SiO 2 catalyst exhibits a slight increase in S C2 (37.7%), while the X CH4 remains unchanged (1.9%), Fig. 8d.The S C2 of the WO 3 /SiO 2 and Na 2 WO 4 /SiO 2 catalysts was lower than in the homogeneous reaction due to the higher X CH4 , leading to sequential reactions towards undesired, more thermodynamically stable CO x compounds [2].The X CH4 of Mn-Na 2 WO 4 /SiO 2 and Na 2 WO 4 /SiO 2 catalysts is similar (within the experimental error at this temperature, ± 0.3%), but the S C2 of the Na 2 WO 4 /SiO 2 catalyst is significantly higher.
As the temperature reaches 750 °C, the blank test shows an increase in both X CH4 (3.1%) and S C2 (56.1%),Fig. 8a; the WO 3 / SiO 2 catalyst shows an increase in X CH4 (4.9%) and a decrease in S C2 (42.2%),Fig. 8b; and the Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 catalysts exhibit an increase in both X CH4 (5.3 and 9.5%, respectively) and S C2 (54.7 and 51.6%, respectively), Fig. 8c-8d.The use of the WO 3 /SiO 2 catalyst, compared to the blank test, results in a slightly higher X CH4 (from 3.1 to 4.9%) but a drastic decrease in S C2 (from 56.1% to 42.2%).Thus, the WO 3 /SiO 2 sample is a poorly selective OCM catalyst, in agreement with the literature [12].Actually, the octahedral coordination of the W 6+ sites in the crystalline WO 3 phase has been related to deep oxidation reactions.This feature is even more evident at 800 °C (at higher X CH4 ), Fig. 8b.Finally, all the catalysts and the blank test afforded higher X CH4 and lower S C2 at 800 °C.The Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 catalysts exhibited a similar S C2 (49.3 and 48.2%, respectively), within experimental error at this temperature (±1.2%), but the X CH4 of the Mn-Na 2 WO 4 /SiO 2 was significantly higher (35.8% vs. 22.7%), suggesting that its less distorted T d -W 6+ sites are more active towards methane activation than the highly distorted T d -W 6+ sites in the Na 2 WO 4 /SiO 2 catalyst.The reactivity of the W sites on the catalysts for methane activation can be also analyzed by comparing the TOF values at each temperature, Table 2.
The number of reacted moles of methane per mol of W per second was always higher on the Mn-Na 2 WO 4 /SiO 2 catalyst, reflecting the synergistic effect among the supported Mn, Na, W oxides, in agreement with the literature [6].Multiple works have attempted to elucidate the catalytic contribution of each supported oxide by formulating mono-, bi-, and tri-component oxide catalysts [19,20] and evaluating their OCM performance using different approaches, for instance chemical looping experiments: the Mn-Na 2 WO 4 /SiO 2 catalyst performs OCM reaction by involving the lattice oxygen from reducible W 6+ and Mn 3+ cations in the absence of gas-phase oxygen; however, the catalyst structure drastically changes under methane flow, forming the crystalline MnWO 4 phase [21,67].This phase, with O h -Mn 2+ and O h -W 6+ sites (i.e., wolframite structure), has been associated with the catalyst reduction and deactivation [21].Higher temperatures and very reducing conditions (high CH 4 /O 2 molar ratios in steady-state operation) accelerate the MnWO 4 phase formation [67].
We have characterized the catalyst discriminating the oxidizing and temperature effects on the structure from the effect of the methane flow.However, the T d -W 6+ and O h -Mn 3+ sites observed just before methane admission into the reactor in the present study are expected to be also present in the steady-state OCM  formation) is slower than catalyst re-oxidation [68], ii) the CH 4 /O 2 ratio used was low (i.e., 2), iii) the Mn 2 O 3 phase has been observed during operando OCM studies even at CH 4 /O 2 ratios as high as 10 [67], and iv) the Mn-Na 2 WO 4 /SiO 2 catalyst has been reported to be stable for extended time on stream (450-1000 h) under CH 4 + O 2 flow [6].Comparing the performance of Mn-Na 2 WO 4 / SiO 2 and Na 2 WO 4 /SiO 2 , the higher methane conversion exhibited by the former appears due to the exclusive presence of the O h -Mn 3+ sites, which makes the T d -W 6+ sites less distorted.However, we may not currently discard that Mn and Na play a role in the activity of this catalyst.

Conclusions
The structure of the Mn-Na 2 WO 4 /SiO 2 catalyst is highly temperature-dependent and, thus, the association of any OCM activity with crystalline phases observed at room temperature is inadequate.In situ TPO-XRD analysis shows that the crystalline phases identified on the Mn-Na 2 WO 4 /SiO 2 , Mn-free Na 2 WO 4 /SiO 2 , and Mn-and Na-free WO 3 /SiO 2 catalysts at room temperature do not exist at OCM temperatures (>700 °C).The c ? b ?a-WO 3 , a ?b-cristobalite, and cubic ?orthorhombic ?molten Na 2 WO 4 phase transitions occur upon heating in oxidizing conditions.In situ TPO-Raman study shows that the O h -W 6+ sites in the WO 3 / SiO 2 catalyst and the T d -W 6+ sites in the Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 catalysts undergo significant structural changes during these phase transitions, as evidenced by the shift of the symmetric stretching (m s ) vibration band.The m s vibration band shift suggests variations in the bond order of the W sites by changes in their oxidation state or distortion degree or both.In situ TPO-XANES spectra indicate that the m s vibration band shift is due to changes in the distortion degree instead of oxidation state, which remains unchanged (W 6+ ) for all catalysts at heating.Additionally, in situ TPO-XANES spectra confirm the presence of O h -Mn 3+ sites in Mn-Na 2 WO 4 /SiO 2 catalyst even at relevant OCM temperatures, which reduces the T d -W 6+ sites distortion in the molten Na 2 WO 4 phase compared with the Na 2 WO 4 /SiO 2 catalyst.Finally, steadystate OCM tests confirm that the O h -W 6+ sites are not active and the presence of the O h -Mn 3+ sites, and thus less distorted T d -W 6+ sites, makes the Mn-Na 2 WO 4 /SiO 2 catalyst more reactive towards methane activation.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 3 .
Fig. 3.In situ TPO-XRD patterns of the Na 2 WO 4 /SiO 2 and Mn-Na 2 WO 4 /SiO 2 catalysts recorded between 50 °C and 800 °C (10 °CÁmin À1 ) under O 2 :Ar = 1:4 (100 cm 3 Ámin À1 ).Diffraction patterns of the crystalline phases identified at high temperatures are also included: orthorhombic (Pnam) Na 2 WO 4 [ICDD 00-020-1163] (purple box) and cubic (Fd-3 m) b-cristobalite [ICSD 034924] (orange box).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) . The resulting catalysts exhibited Raman bands related to T d -W 6+ sites in the crystalline Na 2 WO 4 phase, in line with XRD results and a recent study by Kiani et al. on the molecular and electronic structure of the W sites on model SiO 2 -supported

Fig. 4 ,
Fig. 4, right, shows the position of the m s vibration band of the crystalline WO 3 phase on the WO 3 /SiO 2 catalyst (a) and of the crystalline Na 2 WO 4 phase on the Na 2 WO 4 /SiO 2 (b) and Mn-Na 2 WO 4 / SiO 2 (c) catalysts as a function of temperature in the TPO-Raman study.The m s vibration band in the WO 3 /SiO 2 catalyst red shifts with increasing temperature but undergoes two sharp blue shifts near 450 and 700 °C, suggesting the temperature-induced c (monoclinic) ?b (orthorhombic) ?a-WO 3 (tetragonal) phase transitions, respectively (Fig. 4a, right).The overall red shift indicates a decrease in the bond order of the O h -W 6+ sites with increasing temperature.This is due to an elongation of the W-O bond.The sequential c.?.b?.a-WO 3 phase transitions are thus corroborated by following the d s vibration band, which monotonically red shifts with temperature up to ~700.°C.The presence of the crystalline a-WO 3 phase at 800 °C is confirmed by the broadening of the m as vibration band (Fig. 4a, left) [42].The TPO-Raman study of the dehydrated Na 2 WO 4 /SiO 2 catalyst monitors the a ?b-cristobalite and cubic ?orthorhombic ?molten Na 2 WO 4 transitions observed via TPO-XRD analysis elsewhere.Regarding the a ?b-cristobalite phase transition, the band at 1075 cm À1 broadens, weakens, and slightly redshifts to 1073 cm À1 , the band at 785 cm À1 slightly red shifts to 782 cm À1 , and the bands at 415 and 228 cm À1 vanish above 250 °C.Thus, the crystalline b-cristobalite phase gives rise to Raman bands at 1073, 782, and 292 cm À1 , which remain constant until 800 °C (Fig. 4b, left) [44].Note that the d s vibration band of the cubic Na 2 -WO 4 phase (~309 cm À1 ) overlaps the b-cristobalite band at 292 cm À1 above 250 °C [26]. Fig.5

Fig. 6 .Fig. 7 .
Fig. 6.Energy gap of the 5d split orbitals at the W-L 3 edge (XANES) as a function of temperature and in relation to the observed crystalline phases for the WO 3 /SiO 2 , Na 2 WO 4 / SiO 2 , and Mn-Na 2 WO 4 /SiO 2 catalysts.

Table 1 W
-L 3 edge position according to the zero-crossing of the second derivative and mean W oxidation state obtained by interpolation.Errors in mean oxidation state derived from incertitudes on edge energy determinations.