Water inhibition and role of palladium adatoms on Pd/Al2O3 catalysts during methane oxidation

Supported palladium catalysts are the most active ones toward the complete oxidation of methane. However, the presence of relevant amounts of water hinders the catalytic activity and long-term stability, due to competition between methane and water for the active sites. Hence, understanding the inhibition effect of water on methane oxidation is mandatory to improve these catalysts. We present an in situ ambient pressure X-ray photoelectron spectroscopy study of methane oxidation on Pd/Al 2 O 3 in presence and in absence of water. The inhibition effect of water is demonstrated by combining reactivity tests with electron microscopy and photoelectron spectroscopy measurements. In the presence of water, the redox activity of palladium decreases. Water competes with methane for the catalytically active sites, poisoning the surface with hydroxyl groups and hindering the generation of the coordinatively unsaturated palladium active sites. Multiple palladium oxide species displaying different reactivity are identified. A new cationic palladium species, assigned to adatoms, is detected, which shows higher reactivity with methane than particulate palladium oxide but also high inhibition by water.


Introduction
Natural gas (NG) has gained interest over the years because of its lower CO 2 and particulate matter emissions than traditional fuels and the highest energy density of its major component, methane, among all hydrocarbons [1,2,3].However, the high global warming potential of methane requires the development of efficient catalytic after-treatment systems to control emissions [4,5].Catalytic combustion is the main after-treatment technology used to convert the residual methane [6] and is largely based on palladium supported catalysts [7][8][9][10].Palladium oxide is considered to be the main active phase, with C -H bond activation occurring on a coordinately unsaturated palladium atom [11][12][13][14].Palladium-based catalysts sinter at temperatures above 450 • C [11][12][13][15][16][17][18][19] and are poisoned at lower temperature (T < 450 • C) by water, whose concentration is high in the exhaust of NG vehicles [7,20].The impact of water on the methane oxidation reaction over palladiumbased catalysts has been studied since the 1990s; water and methane compete for adsorption on the active sites [8,21,22].A typical kinetic rate equation (Eq.( 1)) [8,21,22] is: with x ranging from − 0.9 to − 1.1, depending on the experiment; the negative order reflets the inhibition effect of water [8,21,22].At higher temperature, the reaction order for water increases to 0, suggesting that it does no longer compete for adsorption [8,21,22].A structural explanation of the inhibition mechanism of water is still under debate.It has been proposed that water adsorption leads to the formation of hydroxyls on the palladium oxide surface, thus poisoning the adsorption site for methane.The influence of the supporting material and of the composition of the gas mixture is still unclear.As an example, the exchange of hydroxyl groups between the support material and the supported metal has been investigated [23][24][25][26].Increasing the temperature to above 450-500 • C leads to water desorption as a consequence of the recombination of the hydroxyl groups and to the reactivation of palladium active sites [24,26].Ambient pressure X-ray photoelectron spectroscopy (APXPS, also known as near ambient pressure X-ray photoelectron spectroscopy -NAPXPS -) is a reliable tool for studying the effect of water on palladiumbased catalysts during methane oxidation.Due to the surface sensitivity of APXPS, it is possible to perform in situ measurements of the solid-gas interface of a catalyst exposed to reaction mixtures at different temperatures.For example, this technique has been used successfully to characterize the catalytic behavior of different systems for the water gas shift reaction [27][28][29], CO oxidation reaction [30][31][32][33] and methane oxidation reaction [14,34,35].Thus far, model catalysts, i.e. palladium foils [14,21] and single crystals [10], have been employed to determine the behavior of palladium during methane oxidation in the presence and absence of water.Model catalytic surfaces are suitable systems to investigate because (differential) charging is minimal and photoemission signals are sharp.Spectroscopic investigations on these systems provide important information about the nature and structure of catalytically active sites and the reaction mechanism typical of more complex catalysts, constituted by palladium nanoparticles dispersed on a support such as the one used in this work (1 wt% Pd/Al 2 O 3 ).So far, XPS measurements of actual Pd/Al 2 O 3 catalysts have been carried out primarily ex situ, that is, under high vacuum conditions before and / or after catalysis.Interestingly, some studies revealed the presence of palladium species displaying a Pd 3d binding energy higher than that conventionally assigned to Pd 2+ (in PdO), especially when palladium is supported on alumina [36][37][38].However, there is no agreement on the assignment and the high-binding energy species (high-BE species) are associated with either palladium (IV) oxide [36,39] or, more often, Pd n+ (n > 2) [40][41][42][43].Another possibility is that palladium strongly interacts with its support, leading to the formation of aluminum palladate [37,44] or palladium diffusing into the metal oxide lattice [45][46][47], depending on the support.Such a high-BE component in the Pd 3d signal is usually detected for catalysts with a low metal loading (<1% wt) [36,48].Stability tests revealed that it is approximately 10 times more likely to be reduced with respect to the palladium (II) oxide [41].
In this study, we present in situ APXPS measurements, complemented with electron microscopy and activity determination of a 1 wt% Pd/Al 2 O 3 catalyst in presence and in absence of water.We performed experiments switching on and off oxygen from the methane oxidation reaction feed, thus enabling identification of the most reactive palladium species towards methane oxidation.The strong poisoning by water occurs on all oxidic palladium species and is paralleled by the formation of hydroxyls, which, at high temperature, leads to water desorption.In situ XPS offers the unique opportunity to detect multiple oxidic components of palladium and to follow their evolution as a function of temperature and the reaction environment.

Catalyst synthesis
A 1 wt% Pd/Al 2 O 3 sample (labelled S1), was prepared according to Bugaev et al. [49].The γ-alumina support was made by annealing commercially available Boehmite (Capatal® B Alumina, Sasol) in air at 600 • C for 8 h (heating ramp: 10 • C/min).Palladium was then deposited on the γ-alumina support by wet impregnation.The impregnation solution was prepared by the addition of 0.44 g of ammonia solution (28 wt%; Aldrich) and 127 mg of Pd(NO 3 ) 2 ⋅2H 2 O (99 % purity, SigmaAldrich) to 4 mL of deionized water.3 g of γ-alumina was then impregnated with the above mentioned aqueous solution of Pd 2+ .The sample was first dried at room temperature for 2 h, then overnight at 110 • C, and finally calcined in air at 350 • C for 5 h (heating ramp: 2 • C/min).Finally, the catalyst was reduced in a flow of 5 vol% of hydrogen in argon (flow rate: 50 mL/min) at 300 • C (2 • C/min) for 2 h.In the following, sample S1 investigated with co-dosed water will be named S1-W, while S1 investigated without co-dosed water will be named S1-NW.

Characterization methods
XPS measurements were carried out at the in situ spectroscopy (ISS) beamline (Swiss light source synchrotron, Paul Scherrer Institute, Switzerland), using the solid-gas interface experimental endstation [50].Linearly polarized light was used throughout the experiments.The samples were prepared as pellets using an embedded gold mesh as the support.Around 20 mg of powder were used for each pellet and were pressed for 1 min under 1 ton.The pellets were then fixed to the manipulator and placed in the chamber.The heated sample holder available at the ISS beamline allows precise control of the temperature from room temperature to 600 • C by means of a 976 nm, 25 W laser projected onto the back of the sample holder.The temperature was monitored by a Pt100 sensor in direct contact with the sample holder.Reference measurements were performed by filling the analysis cell with a mixture of N 2 and O 2 (4:1 ratio at a total pressure of 1.1 mbar) at a temperature of 50 • C.This procedure limits charging, avoids differential charging and prevents the accumulation of adventitious carbon.Depending on the reaction conditions, the selected gas mixture was introduced: with a 1:4:8 ratio for CH 4 , O 2 and N 2 without co-dosed water (S1-NW) and a 1:4:8 ratio for CH 4 , O 2 and H 2 O with co-dosed water (S1-W).In both cases, the total pressure in the chamber was maintained at 1.2 mbar.The temperature was then increased stepwise.In the absence of co-dosed water in the reaction environment, at each step, we first measured the gas mixture containing all the gases; then oxygen was removed to determine whether changes took place on the catalyst surface.If so, then oxygen was reintroduced to find out whether the observed modifications were reversible before moving to the next temperature and repeating the measurement procedure.With co-dosed water, oxygen was removed first, followed by water.In both cases, high resolution photoemission spectra were acquired to check for modifications of the line shape.Furthermore, the reversibility was checked by reintroducing water and oxygen before changing the temperature.
XPS measurements were performed at 1 kinetic energy (KE = eV): the Pd 3d spectral region was measured at a photon energy (hν) of 890 eV, and the O 1s spectral region was measured at hv = 1090 eV.We made use of a powder sample, which showed a constant surface charging (even in the mbar gas atmosphere).As a consequence, the kinetic energy values of Pd 3d and O 1s core level peaks shifted negatively by approximately 25 eV.The estimated spectral resolutions at hν = 890 and 1090 eV were 0.45 and 0.50 eV, respectively.Al 2p was always used as a reference peak to align the binding energy (BE) scale (centroid of the Al 2p peak at 74.5 eV) [37,51], additionally, Al 2p was also acquired at hv = 620 eV to have it at the same KE as the other two core level peaks.In photoelectron spectroscopy, the same KE value corresponds to the same probing depth.In this case, the inelastic mean free path (λ) of photoelectrons in our setup geometry, estimated using the QUASES-IMFP-TPP2M software [52], is 1.1 nm in palladium (II) oxide and 1.7 nm in Al 2 O 3 , the main contributor.Considering the measurement geometry of the setup, where photoelectrons are measured at an emission angle of 60 • with respect to the sample surface, we estimate that 86 % of the XPS signal comes from 1*λ and 90 % from 2*λ.All XPS fittings were carried out by removing a Shirley background and using Gaussian peak shapes.The Pd 3d spin-orbit splitting was set at 5.3 eV.Ca 2p, a sample contamination found during the experiments and appearing near the Pd 3d spectral region, has a spin-orbit splitting set at 3.5 eV but did not change as a function of the reaction conditions.Fitting parameters are available in Tables S1, S2, S3 and S4 for Pd 3d fitting for S1-NW, Pd 3d for S1-W, O 1s for S1-NW and O 1s for S1-W, respectively.
The catalytic activity was measured using a quartz glass plug-flow reactor (i.d.= 6 mm) loaded with the sieved catalyst (50 mg; 150-200 μm) diluted with cordierite (150 mg; 100-150 μm) and fixed between two quartz wool plugs.The temperature of the catalyst bed was measured and controlled with a thermocouple inserted in the quartz reactor and positioned in the middle of the bed.The outlet gas composition was monitored by mass spectrometry (MS, InProcess GAM400).The sample was pre-treated in a flow of 0.5 vol% CH 4 and 2 vol% O (balance Ar), while the temperature was increased from 150 to 600 • C A. Boucly et al. (10 • C/min) and maintained at 600 • C for 30 min.The temperature was then decreased to the desired value, and the gas composition was adapted to the experiment as follows.When the experiments were carried out without co-dosed water, all the lines and the reactor were heated at 150 • C overnight before pre-treatment to ensure complete water desorption.The reactive feed consisted of 0.5 vol% CH 4 and 2 vol % O 2 (balance Ar) with or without co-dosing 4 vol% H 2 O.All experiments and pre-treatments were performed using a weight hourly space velocity (WHSV = flow/mass catalyst ) of 90 Lh − 1 g − 1 .The heating/cooling ramp experiments were performed at 5 • C/min.During the reducing pulses, the flow of O 2 was stopped every 10 min for 5 min and replaced with Ar.
Scanning transmission electron microscopy (STEM) micrographs were acquired using a JEOL Grand ARM 300 aberration-corrected microscope operating at 300 kV.The high-angle annular dark-field (HAADF) detector was used for the micrograph acquisition.The average size distribution was analyzed using the ferret diameter of elliptical shapes (N = 200).

Results
Fig. 1 shows STEM micrographs of S1, S1-W, and S1-NW samples.After synthesis, only small and highly dispersed nanoparticles of ca.1.5 nm are observed.The images acquired after in situ XPS experiments show an increase in the average particle size for both S1-W (2.1 nm) and S1-NW (2.5 nm) as well as the presence of larger aggregates (50-60 nm); Fig. 1e-f suggest that some degree of sintering occurred, though to a slightly lower extent in the presence of water.
Fig. 2 shows the conversion curves during methane oxidation obtained on Pd/Al 2 O 3 under lean conditions in the absence and presence of co-dosed water.Water shifted the temperature required to reach 50 % methane conversion by approximately 50 • C to higher values confirming its inhibition effect on the catalytic activity [52].Under both reaction conditions, methane conversion exhibited a positive hysteresis, typical of exothermal reactions.Cut-off experiments consisting in the periodic removal of O 2 for 5 min every 10 min, were carried out to observe the impact of the removal and re-introduction of oxygen on the catalyst.In all the experiments, the reintroduction of oxygen always results in the recovery of activity, suggesting that the catalyst reduction is completely reversible.In the absence of co-dosed water (Fig. S1), no conversion takes place up to 250 • C. From 300 to 450 • C, methane conversion increases in the lean phases of the pulses from 5 % to ca. 80 %.In the presence of co-dosed water (Fig. S2), the same conversion levels are reached, but at temperatures that are about 50 • C higher, in line with the methane conversion data (Fig. 2).Cut-off experiments also show that the  A. Boucly et al. sample is active towards the steam reforming reaction above 250 • C, as conversion still occurs and hydrogen is detected when oxygen is removed.To confirm that the catalyst reduction is completely reversible and to study the chemical composition of the palladium catalysts during reaction (both with and without co-dosed water) and in the switch-off experiments, APXPS measurements have been carried out.
Figs. 3 and 4 show the Pd 3d photoemission spectra and their deconvolution recorded in situ under catalytic conditions (0.5 vol% methane and 2 vol% oxygen) without co-dosed water (sample S1-NW) and in the presence of 4 vol% water (sample S1-W).The detailed fitting parameters are shown in Tables S1 and S2.The gas composition was identical to that used in the catalytic measurements shown in Fig. 2.However, the amount of catalyst, the configuration of the reaction cell, and the total pressure were different so it is not possible to quantitatively correlate the conversion (Fig. 2) with the spectroscopic data.
Fig. 3a shows the Pd 3d core level spectra of S1-NW.The reference Pd 3d spectrum (50 shows two features that can be deconvolved with four doublets (spin-orbit splitting of 5.3 eV and branching ratio of 1.5; Table S5) [14,36,51,53].The first peak at a BE of 335.5 eV (Pd 3d 5/2 , green) corresponds to metallic palladium (MPd) [36,51,53].The second peak centered at 336.1 eV (Pd 3d 5/2 , orange) is attributed to surface-palladium oxide (SPdO) [36,51,53], while the peak at 337.0 eV (Pd 3d 5/2 , red) is ascribed to bulk-like palladium (II) oxide (PdO) [36,51,53].The fourth peak at the highest binding energy of 338.5 eV (Pd 3d 5/2 , blue) has not yet been unambiguously assigned (see SI Table S6).However, it can be related either to cationic palladium with an oxidation state higher than 2+, [40][41][42] or cationic palladium adatoms in strong interaction with the alumina support [37,44].Based on the low abundance and dominant presence at low palladium loading, we tentatively assign this peak component to the latter situation and we label it as "palladium adatoms".This assignment is further supported by the formation of Pd-Al and Pd-O bonds in surface science studies [54][55][56][57][58], compatible with palladium adatoms interacting with the alumina support.The Pd 3d peak of this species has also a broader full width at half maximum (FWHM) than the one assigned to SPdO species.This indicates that it might be the result of a distribution of species in a similar but not identical chemical state, reflecting the interaction of palladium atoms with oxygen vacancies on alumina (vide infra).The data acquired at 450 • C originate from a second sample (used to check the experiment reproducibility), which did not show any calcium contamination.While there can be a small uncertainty in the proportion of each species (error bars are reported in the histograms of panel b), there is no uncertainty about the four peaks needed to obtain a good correlation with the experimental data, as well as the observed behavior.The redox processes are completely reversible, as the catalyst reverts back to an equally oxidized state when oxygen is reintroduced into the feed (Fig. S3).As the temperature increases, these effects become more pronounced.At 250 • C, the fraction of palladium adatoms changes significantly.In the presence of oxygen, it is the most intense signal (46 %), while when oxygen is removed it decreases to 21 %.Palladium adatoms are the only reactive species at this temperature, because the fraction of MPd shows the opposite trend and increases from 11 to 24 % upon oxygen removal, while SPdO increases by 15 % and PdO does not change.At 300 • C, the fraction of palladium adatoms decreases from 42 to 20 % upon oxygen removal.However, at this temperature, the signal of PdO also shows a significant decrease (from 27 to 11 %).The increase of MPd fraction from 0 to 33 % indicates that a substantial reduction of cationic palladium takes place within the probing depth of XPS.At 450 • C, the palladium signal is dominated by palladium adatoms (almost 60 %).When oxygen is removed, adatoms almost disappear, and the main component is MPd (ca.60 %).As observed at lower temperatures, the signal assigned to the SPdO increases, indicating that hydroxyls (the binding energy of palladium hydroxyls is superimposed to that of SPdO [14]) form as a consequence of hydrogen extraction from methane.
Another hypothesis to explain the increase of SPdO, which is supported by electron microscopy, is that part of the palladium adatoms sinter into SPdO.The signal of PdO decreases but does not disappear, suggesting that exposed active sites react with methane, while those more in the bulk of NPs retain their oxidation state.The diffusion of oxygen from PdO to the surface, where it participates in the methane oxidation, can also explain the increase of the SPdO [15,23].
The sintering effect observed on S1-NW, followed by the evolution of the Pd 3d:Al 2p ratio (Table S7), becomes significant at high temperatures as the ratio decreases.Under reaction conditions, only a half of the initial palladium signal is detected with respect to the Al 2p signal generated by the alumina support.Interestingly, the Pd 3d/Al 2p ratio decreases further when oxygen is removed and it increases once oxygen is reintroduced, suggesting re-dispersion of palladium particles.
The reference Pd 3d spectrum (50 • C) of the S1-W sample (Fig. 4a) displays the same peak components as in S1-NW (see Table S5).However, when in presence of co-dosed water, the fraction of palladium adatoms decreases.Considering that there are no strong modifications at 150 • C for the S1-NW when the flow of oxygen is stopped (no reduction) and that the inhibition effect of water shifts redox reactions to higher temperatures, we selected 200 • C as the first characterization temperature on S1-W.Fig. 4 shows that there are negligible changes in the spectra line shapes in the three investigated reaction mixtures.At 250 • C, palladium is fully oxidized, showing a strong contribution of PdO (45 %), followed by palladium adatoms (39 %) and SPdO (16 %).Reduction to MPd is negligible (4 %) when oxygen is removed from the feed but becomes noticeable upon the removal of water, when only methane and nitrogen are dosed.MPd increases to about 23 % of the signal, similar to the value measured for S1-NW at 200 • C in the absence of oxygen (20 %).Increasing the temperature to 250 • C leads to an increase in the fraction of PdO when all gases (methane, oxygen and water) are present (8 % higher signal than that measured at 200 • C), a similar behavior as that of the S1-NW sample.Upon removal of oxygen and water, the fraction of PdO decreases by approximately 15 %.The Pd 3d:Al 2p ratio (Table S7) decreases indicating that some degree of sintering takes place already at this temperature.To achieve a significant reduction of palladium when only oxygen is removed from the reaction mixture, it is necessary to increase the temperature to 450 • C (24 % of the Pd 3d signal is assigned to MPd).This is also observed in the O 1s spectra (Fig. S5); the component attributed to hydroxyls and adsorbed water almost disappears.Based on the literature, hydroxyls and adsorbed water should give two distinct peaks (separated by approx.1.0 eV) [59].However, because both signals are due to the presence of water and they tend to become undistinguishable at higher temperature, we decided to fit their signals with a single peak.The goal was to estimate the effect of co-dosed water on the O 1s spectra and as a function of temperature.Palladium adatoms are more inert than in the case of S1-NW, where relevant variations are observed upon modification of the reaction environment due to the inhibition effect.Furthermore, the increasing trend with temperature (from 50 • C to 250 • C) observed in the fraction of PdO reverts, and its value returns to the initial one.Both observations indicate that water favors the dispersion of nanoparticles, especially when compared to S1-NW.The Pd 3d:Al 2p ratio (Table S7) shows a smaller decrease than that of S1-NW.Despite the temperaturedependent decrease, there are negligible variations in the Pd 3d:Al 2p ratio upon oxygen removal.As an example, at 450 • C in the absence of oxygen, the ratio for S1-W is 2 times larger than for S1-NW.Such a difference is also confirmed by microscopy (Fig. 1), showing a smaller average nanoparticle size for S1-W.As opposed to the experiment performed without co-dosed water, in presence of water the fraction of SPdO is relatively stable.Similar to the experiment carried out in absence of co-dosed water, the consumed fraction of palladium adatoms is higher than that of PdO (Fig. S4, top).However, this becomes evident only at 450 • C due to the water inhibition effect.

Discussion
The changes of the Pd 3d line shape observed in the absence of codosed water at 150 • C (S1-NW) upon removal of oxygen can be explained by a temperature-mediated removal/desorption of surface carbon contaminants.As a consequence, signals originating from the deeper layers of the nanoparticles are less attenuated and increase with respect to the surface ones.As the intensity of both the MPd and the PdO species increases, it is assumed that these species are more distributed towards the bulk of palladium particles.On the contrary, the stability of the SPdO signals and the decrease of the signals of the palladium adatoms indicate that both species are either at the surface of the nanoparticles (coordinatively unsaturated palladium forming surface oxide) or correspond to dispersed entities (palladium adatoms, whose dimension is smaller than the escape depthapprox.0.9 nm -of XPS at the kinetic energy of Pd 3d).The slight decrease in the fractions of the PdO and palladium adatoms at 200 • C, mirrored by an increase in MPd, suggests that the reaction starts between 150 and 200 • C. The decrease in the PdO signal can be assigned to the transfer of oxygen from the bulk to SPdO, explaining the stability of the fraction of these latter species [14].At 250 • C, the reported increase in the SPdO fraction in the absence of oxygen could be due to the formation of surface hydroxyls, whose binding energy is superimposed onto that of the SPdO.Surface hydroxyls form during the reaction between SPdO sites and adsorbed methane and are stable up to 450 • C on a palladium foil [14].The temperature-mediated sintering of palladium adatoms, could also explain the increase in the SPdO fraction.Such an hypothesis is supported by the analysis of the Pd 3d:Al 2p ratio (Table S7), which decreases above 250 • C, suggesting that palladium starts to sinter.At 300 • C in the reaction mixture, the signal of MPd completely disappears, while that of the cationic species increases, indicating quantitative oxidation of palladium nanoparticles within the probing depth of XPS.
The bulk and, in particular, the SPdO fractions increase with temperature when the catalyst is exposed to the reaction mixture.This indicates that the nanoparticles undergo structural modifications.Larger nanoparticles, exposing a greater number of SPdO species as well as a slightly higher amount of PdO underneath, form.This is confirmed by electron microscopy, showing an increase in the average diameter from 1.5 nm to 2.5 nm for the S1-NW sample compared to the as-prepared state (Fig. 1).This experiment, carried out without co-dosed water shows that palladium adatoms are more susceptible to reduction than PdO, as they are quantitatively reduced at 250 • C suggesting that this species is more active towards methane.Conversely, higher temperature (300 • C) is required to obtain a significant reduction of PdO when oxygen is removed.With regard to the SPdO component, which also corresponds to the signal of surface hydroxyls, no significant changes are detected; the temperature is too low to favor hydroxyls recombination.
The initial fraction of palladium adatoms in S1-W is lower than in S1-NW.This suggests that at low temperature (T < 300 • C), water hinders the dispersion of palladium NPs to adatoms in the presence of methane and oxygen.The absence of changes upon removal of oxygen and water at 200 • C suggests that the adsorption of methane is hindered by the water poisoning of cationic sites.This proves the competition between water and methane for catalytically active sites in the low temperature range, which influences the catalytic activity and reaction menchanism, as also recently pointed out by Velin at al. [60] Surface hydroxyls and palladium oxide-/surface oxide-related components should also be visible in the O 1s spectrum.However, as shown in Fig. S5, the O 1s spectral region is dominated by the signal of the lattice oxygen of the alumina support at 531.3 eV.Because the powder catalyst used in this work contains only 1 wt% palladium, the estimated contribution of palladium oxide to the O 1s peak would be extremely low (0.7 % of the total O 1s peak area, based on the Pd 3d area corrected by the total cross section and photon flux).Therefore we did not add a palladium oxiderelated component in the fitting of O 1s.While the peak deconvolution seems to indicate a component at higher binding energy (533.4eV), assigned to both the hydroxyls and the adsorbed water, its intensity is extremely low compared to the signal of lattice oxygen.Due to the low peak intensity, hydroxyls and adsorbed water cannot be separated.However, the intensity of such a component decreases as the temperature increases, suggesting that a fraction of hydroxyls starts to recombine and/or adsorbed water is being removed above 250 • C (Fig. S5).At 250 • C, the competition between methane and water is still in favor of the latter, because it is necessary to stop the water supply to induce a reduction of palladium.Compared to the experiment performed in absence of co-dosed water, the formation, recombination, and desorption of the hydroxyl groups shift to higher temperatures with water.Furthermore, the stability of the fraction of palladium adatoms indicates that water tends to also hinder the formation/stabilization of the most active cationic sites.At 450 • C and upon oxygen removal, the fraction of palladium adatoms finally starts to decrease, together with that of PdO, indicating that methane activation is no longer poisoned by water.
When comparing the fraction of MPd that forms both in the presence and absence of co-dosed water, (lower part of Fig. S4) the negative impact of water is clear.The extent of palladium reduction is always greater in the case of S1-NW within the same temperature range.Data in Fig. S4 (lower plot) also demonstrate that the shift towards higher temperature to observe approximately the same palladium reduction extent between S1-NW and S1-W is by 200 • C.This value is greater than that estimated from the light-off curves (50 • C, Fig. 1), which we attribute to significant differences in the reactor geometry, gas flow, and absolute gas pressures that can lead to differences in the local concentrations of the gases and the surface coverage of the active species.

Palladium adatoms (high-BE palladium species)
As reported above, the Pd 3d photoemission spectra unequivocally display four components.Beside three species commonly observed with A. Boucly et al. palladium reference samples (foils and single crystals), e.g.MPd, SPdO, and PdO, we need to consider an additional high binding energy component, which we assigned to palladium adatoms stabilized by the interaction with the alumina support [37,44] Such a high binding energy palladium species has been reported several times and has been attributed to palladium in close interaction with the support [36][37][38].The reactivity of this species towards methane oxidation has been the focus only of a few studies.Based on activity measurements of catalysts containing high-BE palladium species (attributed to palladium (IV) oxide), Gao et al. [61] reported that palladium (IV) oxide is less active than PdO in the oxidation of methane.In contrast, Khader et al. [46] and Liu et al. [62] attributed the high catalytic activity of their Pd/CeO 2 catalysts to the presence of such high-BE palladium, stabilized by the interaction with the ceria support.These differences may indicate that the support plays a fundamental role in the reactivity of high-BE palladium species, with ceria being more efficient than alumina.In our case, in situ XPS data acquired without co-dosing water (Fig. 3b) show that palladium adatoms are reactive toward methane oxidation.This is demonstrated by their high reducibility compared to that of PdO: at 250 • C, the percentage of palladium adatoms decreases from 46 to 21 % upon oxygen removal, while that of PdO is relatively stable.To obtain a significant reduction of PdO, it is necessary to reach higher temperatures (≥300 • C).Similar results have been obtained when both palladium adatoms and PdO species were exposed to carbon monoxide [41], further supporting the assignment of these species to adatoms, which are stabilized by the strong interaction with alumina.Therefore, the synthesis strategy we have followed in this work favors the formation of palladium adatoms on alumina, which show high reactivity toward methane oxidation at low temperature in absence of water in the reaction mixture.We can postulate that the impregnation of a pre-calcined alumina support in a corresponding low weight loading of the palladium precursor, followed by drying and calcination in air (oxidizing environment), facilitates the dispersion of the metal and formation of adatoms.This is in good agreement with the reported positive correlation between oxidizing pretreatments and the dispersion of metal nanoparticles [63].When water is co-dosed to the reaction mixture, the behavior changes.The fraction of palladium adatoms detected at low temperature is ca.20 % lower than in absence of water.This suggests that water hinders the formation and stabilization of palladium adatoms and heavily poisons them.Such an effect becomes more evident when considering the behavior during the switching experiment performed at 250 • C: the fraction of palladium adatoms shows negligible changes, also upon water removal, whereas the fraction of PdO decreases.A significant decrease in the signal of palladium adatoms occurs only at 450 • C, where methane activation becomes favored.At 450 • C, the fraction of palladium adatoms is larger than that of all the other cationic species, proving that high temperature is required to mitigate water poisoning and to reactivate these sites but also their stability with respect to the other palladium species.The observed sintering behavior of the catalysts is less significant in presence of co-dosed water, especially when oxygen is removed, which indicates that water favors the dispersion of palladium nanoparticles at high temperature.
In summary, palladium adatoms can activate methane at low temperature in the absence of water and their activity is higher than that of PdO.Water hinders formation of these sites at low temperature and competes with methane for adsorption producing the shift of methane conversion curves to high temperature.

Conclusions
In this work, we investigated in situ the competition between water and methane for active sites in the complete oxidation of methane.The presence of water shifts the reduction of cationic palladium species toward higher temperature when oxygen is removed from the reaction feed.Activity measurements are supported by spectroscopy; there are differences in the behavior (average nanoparticle size) depending on the presence of water in the reaction environment.STEM measurements show that nanoparticles are initially well dispersed and are on average small in size.Sintering is promoted by the temperature, however, water favors the dispersion of nanoparticles at high temperature (above 300 • C).XPS indicates the presence of four palladium species: metallic palladium (MPd), surface palladium oxide (SPdO), bulk palladium oxide (PdO), and a high binding energy palladium one, which is assigned to palladium adatoms stabilized by the interaction with alumina.The latter species have higher reactivity than PdO species towards methane activation in the absence of water in the reaction feed.Their formation at low temperature is hindered by water, which causes the positive shift in temperature, as evident in the methane conversion curve.In situ XPS is a valid method to detect palladium active sites and to follow their evolution in the reaction environment.Considering the presence and activity of palladium adatoms, as well as preserving their high activity towards the oxidation of methane, is of paramount importance to keep the catalyst active.

Fig. 1 .
Fig. 1.STEM images of the catalysts after exposure to different conditions: as prepared S1 (a), mean particle size distribution of the sample as prepared and after the in situ APXPS experiments performed with and without co-dosed water (b), STEM images of S1-NW (c-d), STEM images of S1-W (e-f).

Fig. 3 .
Fig. 3. a) Deconvolution of Pd 3d photoemission spectra of S1-NW acquired at hν = 890 eV with increasing temperature in 0.5 vol% methane and 2 vol% oxygen in the absence of co-dosed water.Colors represent metallic palladium (MPd) in green, surface oxide palladium (SPdO) in orange, bulk oxide palladium (PdO) in red, palladium adatoms in blue, and the Ca 2p contamination signal in grey.b) Percentage of Pd 3d signal of MPd, SPdO, PdO and palladium adatoms.

Fig. 3b displays
Fig. 3b displays the fraction of each species plotted as a function of the reaction environment and temperature.Palladium is almost completely oxidized under initial measurement conditions, showing only a small fraction of MPd (10.8 %).While both SPdO (17.2 %) and PdO (20.6 %) display similar fractions, palladium adatoms are the dominant species (51.4 %).When switching to the reaction mixture at 150 • C, the signals of PdO and MPd increase at the expense of the palladium adatoms, whose contribution decreases from 51 to 43 % (Fig. 3.b).Removing oxygen does not lead to important modifications.At 200 • C, the signal of MPd increases from 10 to 20 % and at the same time, the fractions of both PdO and palladium adatoms decrease slightly.The redox processes are completely reversible, as the catalyst reverts back to an equally oxidized state when oxygen is reintroduced into the feed (Fig.S3).As the temperature increases, these effects become more pronounced.At 250 • C, the fraction of palladium adatoms changes significantly.In the presence of oxygen, it is the most intense signal (46 %), while when oxygen is removed it decreases to 21 %.Palladium adatoms are the only reactive species at this temperature, because the fraction of MPd shows the opposite trend and increases from 11 to 24 % upon oxygen removal, while SPdO increases by 15 % and PdO does not change.At 300 • C, the fraction of palladium adatoms decreases from 42 to 20 % upon oxygen removal.However, at this temperature, the signal of PdO also shows a significant decrease (from 27 to 11 %).The increase of MPd fraction from 0 to 33 % indicates that a substantial reduction of cationic palladium takes place within the probing depth of XPS.At 450 • C, the palladium signal is dominated by palladium adatoms (almost 60 %).When oxygen is removed, adatoms almost disappear, and the main component is MPd (ca.60 %).As observed at lower temperatures, the signal assigned to the SPdO increases, indicating that hydroxyls (the binding energy of palladium hydroxyls is superimposed to that of SPdO[14]) form as a consequence of hydrogen extraction from methane.

Fig. 4 .
Fig. 4. a) Pd 3d photoemission spectra and their deconvolution acquired at hν = 890 eV with increasing temperature for the S1-W sample under 0.5 vol% methane and 2 vol% oxygen with 4 % co-dosed water gas mixture conditions.b) Percentage of Pd 3d signal of MPd, SPdO, PdO and palladium adatoms.