Mixed RuxIr1−xO2 Supported on Rutile TiO2: Catalytic Methane Combustion, a Model Study

With a modified Pechini synthesis, mixed RuxIr1−xO2 is grown on rutile‐TiO2 with full control of the composition x, where the preformed TiO2 particles serve as nucleation sites for the active component. Catalytic and kinetic data of the methane combustion over RuxIr1−xO2@TiO2 and unsupported RuxIr1−xO2 catalysts reveal that the least active catalyst is RuO2@TiO2 (onset temperature: 270 °C), while the most active catalyst is Ru0.25Ir0.75O2 with an onset temperature below 220 °C. Surprisingly, even Ru0.75Ir0.25O2@TiO2 is remarkably active in methane combustion (onset temperature: 230 °C), indicating that little iridium in the mixed RuxIr1−xO2 oxide component already improves the activity of the methane combustion considerably. We conclude that iridium in the mixed RuxIr1−xO2 oxide enables efficient methane activation, while ruthenium promotes the subsequent oxidation steps of the methyl group to produce CO2. Kinetic data provide a reaction order in O2 of zero, while that of methane is close to one, indicating that the methane activation is rate limiting. The apparent activation energy varies among RuxIr1−xO2 from 110 (x=0) to 80 kJ ⋅ mol−1 (x=1). This variation in the apparent activation energy may be explained by the variation in adsorption energy of oxygen. Under the given reaction conditions the catalyst's surface is saturated with adsorbed oxygen and only if oxygen desorbs, methane can be activated and the methyl group can be accommodated at the liberated surface metal sites.


Introduction
Methane, the primary constituent of natural gas, is a promising alternative fuel in mobile and stationary applications due to its relatively high abundance, lower cost compared to other fuels and the lowest C/H ratio among organic molecules so that combustion leads to the lowest amount of CO 2 per energy unit. [1,2] Methane is also a promising energy vector for future circular energy economy based on renewable energies. [3] However, methane is a potent greenhouse gas with a global warming factor that is 28-34 times higher than that of CO 2 . Therefore, the use of natural-and biogas based fuels in transportation requires an efficient oxidation catalyst in particular under lean conditions to prevent methane-slip to the atmosphere. [4] Unfortunately, methane is the most difficult hydrocarbon to catalytically oxidize so that a relatively high temperature is needed for the reaction to proceed with an acceptable rate. [5,6] While several studies about non noble-metal catalysts such as perovskite type materials with high activity have been reported over the last two decades, [7,8,9] palladium based catalysts are generally considered to be the most promising candidates for methane combustion. [1,[10][11][12][13] Especially at low reaction temperatures Pd is more active than rhodium and platinum. [3,10,12] There is general consensus that higher temperatures stabilize the metallic Pd as the (more) active phase, while at lower temperatures PdO governs the stable activity. [14] Pd or rather PdO has shown even higher activities when supported on γ-Al 2 O 3 , demonstrating beneficial interactions between the support and the active component. [15] Since Pt and Rh show also high activity in methane combustion (although less active than Pd), [3] the methane combustion activity may be traced to properties typical for the platinum group members to which ruthenium and iridium belong.
Recently, an oxidized Ir(100) single-crystalline surface was reported to be surprisingly efficient in the low-temperature methane activation. In temperature programmed reaction (TPR) experiments the oxidized surface of Ir(100) was covered with methane (and oxygen) at temperatures below À 173°C and then the products were recorded with mass spectrometry while ramping the sample to 427-527°C. [16,17] The active phase has been assigned to a IrO 2 (110) layer, as previously predicted by Wang et al. on the basis of density functional theory (DFT) calculations. [18] We should emphasize here that these TPR experiments are mere transient and not catalytic experiments; for catalytic experiments one needs to demonstrate steady state conversion under flow reaction conditions. The methane activation process was further examined by in-situ DRIFTS and Raman experiments for IrO 2 nanoparticles in a constant flow of pure CH 4 , emphasizing the important role of the oxidation state of iridium. [19] Metallic iridium (supported on alumina) has briefly explored as a part of catalyst screening of noble metals for methane combustion at 475°C, revealing, however, iridium to be inferior to the remaining platinum group members. [20] While IrO 2 is efficient for methane activation, [16] its total oxidation capability is quite poor as demonstrated by a recent study on the CO oxidation over Ru x Ir 1À x O 2 powder catalysts with the composition x ranging from 0 to 100 mol %. [21] Quite in contrast, RuO 2 is well-documented to be an efficient oxidation catalyst. [22] Therefore, we anticipated that mixed Ru x Ir 1À x O 2 may offer synergy effects for the combustion of methane by promoting the methane activation step by iridium, while the subsequent oxidation steps towards CO 2 are promoted by ruthenium centers.
In order to have full control of the composition of the mixed Ru x Ir 1À x O 2 oxide component we designed a modified Pechini synthesis route where preformed rutile-TiO 2 powder was used to provide nucleation sites for the formation and deposition of Ru x Ir 1À x O 2 , thereby producing a supported Ru x Ir 1À x O 2 @TiO 2 catalyst with well-defined composition x. In this study we present catalytic and kinetic data for the methane combustion over supported Ru x Ir 1À x O 2 @rutile TiO 2 catalysts in comparison with data from unsupported powder Ru x Ir 1À x O 2 . This paper focuses on the intrinsic activity in the combustion of methane, and how this activity can be modified by mixing RuO 2 with IrO 2 .

Modified Pechini synthesis
The synthesis of unsupported Ru x Ir 1À x O 2 was described recently. [21] The supported catalysts (Ru x Ir 1À x O 2 @TiO 2 ) were synthesized with a modification of the Pechini route [23] by adding pure rutile-TiO 2 (particle size < 100 nm) before complexation of the ruthenium-and iridium-cations by citric acid. Rutile TiO 2 as carrier was chosen since we expected a high dispersion of rutile Ru x Ir 1À x O 2 from previous surface science studies of RuO 2 and IrO 2 on rutile TiO 2 (110). [24,25] The added rutile-TiO 2 support particles are trapped in the carbon network after polymerization of citric acid and ethylene glycol. The ruthenium and iridium cations either nucleate directly on the support surface or mixed ruthenium-iridium particles nucleate first and then adhere to the support material, resulting in both cases in highly dispersed supported ruthenium-iridium mixed oxide particles after final calcination.
Throughout the manuscript the supported samples are referred to as Ru_x@TiO 2 , with x being the nominal composition of ruthenium in mol % changing from 0 mol % to 100 mol % in steps of 25 mol % (pure iridium sample is referred to as Ir_100@TiO 2 instead of Ru_ 0@TiO 2 ) while unsupported powder catalysts are abbreviated by Ru_x. The relative amount of active component with respect to the support is chosen to be 5 mol %. This amount represents a reasonable compromise for having enough active component for in-depth characterization and catalytic measurements, while being low enough to ensure that most of the active component is supported on rutile TiO 2 . Further details on the preparation of supported materials can be found in the supporting information.

Characterization techniques
The samples were degassed in vacuum for 12 h at 120°C before conducting Kr-physisorption experiments at À 196°C with the Autosorb 6 of Quantachrome. The Brunauer-Emmett-Teller (BET) method was employed to quantify the specific surface area (referred to as BET), regardless of whether this comes from the active component or from the carrier.
The CO-pulse-experiments (CO up-take experiments) were conducted in a home-built apparatus in order to determine the number of active sites provided by the active component of the rutheniumiridium mixed oxide samples. In general, the titration experiment quantifies the amount of accessible surface noble-metal atoms per gram catalyst according to the number of adsorbed CO molecules. [26][27][28] With the approximation that each of these ruthenium/iridium sites is an active center for the methane conversion the number of active sites is equal to the number of adsorbed CO molecules. Even if this approximation of every metal site being an active center might not entirely be correct, it still provides the relative concentration of the number of actives sites between the different samples for a proper normalization of the catalytic STY (space time yield) data to the accessible surface of the supported active component.
X-ray diffraction (XRD) measurements were carried out in Θ-2Θ geometry (Bragg Brentano) on a Panalytical X'Pert PRO diffractometer with a Cu Kα source (40 kV, 40 mA) with a step size of 0.013°in 2Θ and a scanning speed of about 0.8°min À 1 . LaB 6 standards (NIST) were added to correct the 2Θ shift due for instance to different sample holder positions. The position of the rutile rutheniumiridium mixed oxide (101) reflection can be used to determine the composition of that oxide by Vegard's law. [21] The ratio of iridium to ruthenium concentration in the near-surface region of the Ru_x@TiO 2 samples was quantified by X-ray photoelectron spectroscopy (XPS) (PHI VersaProbe II). Deconvolution of the XP spectra are performed using the CASAXPS software.
Low resolution transmission electron microscopy (TEM) was performed on a Philipps CM30 instrument operated at 300 kV. For detailed structural, morphological and chemical analysis, a Cs probe-corrected scanning transmission electron microscope (STEM) was employed. Further details about the used characterization techniques are collected in the supporting information.

Flow reactor for methane combustion reaction
The catalytic and kinetic experiments are conducted in a homebuilt reactor system (cf. Figure S1). With mass flow controllers (MFC, MKS Instruments 1179 C) the desired reaction feed is mixed, consisting of CH 4 and O 2 balanced by N 2 , and fed into the reactor that consists of a quartz tube, 6 mm inner diameter, placed in a ThermConcept tube furnace. The catalytic experiments are designed to collect microkinetically controlled activity data so that the reactor is employed in a differential way (further description is given in the supporting information). The purities of used gases CH 4 (Linde) and O 2 (NipponGases) are 4.7 and 4.0, respectively. The carrier gas nitrogen is generated by the Hampson-Linde cycle so that it must be dried and purified prior to admission to the mass flow system. A nondispersive infrared (NDIR)-sensor detects the volumetric concentration of CÀ H bonds giving the portion of CH 4 . CO 2 is also detected and used primarily to quantify the space time yield [Eq. (1)] where _ V total is the total volume flow rate. In this work the catalytic performance is quantified by the space time yield STY ([mol (CO 2 ) · h À 1 · kg cat À 1 ]) which is easily calculated by the detectable variables. Under the chosen reaction conditions, we do not detect any trace of CO (NDIR) and H 2 (GCMS) in the product stream. For a direct comparison among the studied samples the STY can also be normalized to the number of actives sites (CO-pulse-experiments) that is proportional to the turnover frequency (TOF). A mass flow meter downstream the reactor measures the apparent total flow rate _ V total,apparent which can be converted to the actual _ V total , besides the CO 2 concentration, required to quantify the space time yield [Eq. (4)]. Further details are provided in the supporting information. With a height of the catalyst bed of about 1.5 mm and a diameter of 6 mm the gas hourly space velocity is about 1.64 · 10 5 or 345.000 ml · g À 1 · h À 1 if normalized to the average mass of catalyst The catalyst bed has been prepared by placing the pure sample (10 mg to 30 mg) on a 1 mm thick layer of quartz sand in order to obtain a catalyst bed as flat as possible that is important for accurate kinetic data. Analogous the catalyst material is covered with a 1 mm thick layer of quartz to prevent the nano-powder from carrying away downstream. The entire reactor is placed vertically in the oven to maintain a stable shape of catalyst bed during operation.

Characterization of mixed ruthenium-iridium supported on rutile-TiO 2
With XPS we investigated the actual near-surface composition of the supported ruthenium-iridium catalysts (Ru_x@TiO 2 ) for varying nominal composition x (cf. Figure 1); in these spectra the background and the C1s component are subtracted for clarity reasons. From the energetic positions of the main components in the Ru3d spectra (cf. Figure 1a) we can clearly conclude that ruthenium is always in the 4 + oxidation state with no sign of metallic Ru (cf. Figure 1c). The Ir4 f spectra are dominated by Ir 4 + (cf. Figure 1b) but exhibit also some minor contribution of metallic Ir (cf. Figure 1c). Similar results were reported for the unsupported mixed ruthenium-iridium oxide samples [21] and were interpreted in terms of agglomerates of oxide particles adhering to a buried mixed metal particle. Therefore, the observed methane combustion activity can solely be ascribed to the mixed oxide phase. In Figure 1c detailed deconvolutions of experimental Ru3d and Ir4 f XP spectra are shown exemplarily for the Ru_25@TiO 2 sample; the deconvolution of the other samples can be found in the supporting information (cf. Figure S2, Table S1).
Within the XP spectra of a single sample, the intensities of Ru3d and Ir4 f are strictly correlated so that from these spectra the mean ruthenium and iridium composition as well as the composition x of the mixed oxide component Ru x Ir 1À x O 2 @TiO 2 Figure 1. XP spectra of Ru_x@TiO 2 samples in the spectral region of Ru3d (a) and Ir4 f (b). The Ru3d the spectra shown are derived by subtracting the background as well as the C1s signal derived by deconvolution using the CASAXPS software. The Ir4 f spectra have been derived by subtracting the background. c) Deconvolution of the Ru3d and Ir4 f spectra exemplified with the Ru_25@TiO 2 sample. Note that both spectra have identical intensity axis, thus emphasizing the relative intensity of Ru3d and Ir4 f. of each sample can be determined quite accurately; recall that the active components form solid solutions as for the unsupported system. [21,29] For determining the mean composition, the integral intensity of Ru3d and Ir4 f (without C1s in case of Ru3d and background) are used, while for determining the composition x of the mixed oxide, only the integral intensity of Ru 4 + 3d and Ir 4 + 4 f are taken. The thus estimated compositions remarkably agree with the nominal composition as given by the molar ratio of the used precursors in the synthesis (see . Table 1), emphasizing the high level of control of the composition of the supported mixed oxides.
However, the intensities among the Ru3d and Ir4 f of different samples are not strictly correlated due to slightly varying experimental conditions and to varying dispersion of the active component supported on the rutile-TiO 2 . Yet, a comparison of the experimental Ru3d and Ir4 f spectra in Figure 1a,b reveals that the intensity variation among the Ru3d (Ir4 f) spectra qualitatively reflects the concentration of the nominal ruthenium (iridium) of all samples, but the pure Ru_100@TiO 2 , whose intensity is close to that of Ru_50@TiO 2 .
The experimental spectra can, however, be normalized to the integral Ti2p intensity and additionally to the actual concentration of iridium and ruthenium as given in Table S2 (cf. Figure S3). From these normalized spectra one can determine the relative near-surface amount of the active component (Ru + Ir)/(Ru + Ir + Ti) that nominally should be 5 mol %. We can clearly see in Table 1 that these values are several times higher than 5 mol %, thus indicating substantial dispersion of the active component. Also obvious is that the dispersion of pure Ru_100@TiO 2 is the lowest among the mixed samples Ru_ x@TiO 2 . Further details are provided in the supporting information (cf . Tables S3, S4).
With powder XRD we examine the structure of the supported catalysts (cf. Figure 2a). The XRD scans are governed by the rutile-TiO 2 support, but there is also a faint but clearly visible (101) reflection of mixed ruthenium-iridium oxide that shifts with increasing ruthenium concentration to higher diffraction angles. In order to be able to quantify this shift, we employed LaB 6 with its sharp reflections to calibrate the 2-theta axis. This allows us to determine the chemical composition of the mixed ruthenium-iridium oxide utilizing a Vegard plot (cf. Figure 2b). [21,30] These composition values together with the full width half maximum (FWHM) of the (101) reflection of the mixed oxide and the mean crystallite sizes derived from the Scherrer equation are collected in Table 2. Actually, for mixed oxides, one needs to apply the William-Hall analysis instead of the Scherrer equation to extract the mean crystallite size. However, the William-Hall analysis requires several mixed oxide reflections in XRD to disentangle the crystallite size from microstrain. Since, there is only one Ru x Ir 1À x O 2 related reflection in our experimental XRD data the Scherrer equation is used as a first estimation for the crystallite size.
The FWHM decreases considerably with the chemical composition x (Ru x Ir 1À x O 2 @TiO 2 ), that is associated with an increase of the crystallite size of the mixed ruthenium-iridium oxide particles. For Ir_100@TiO 2 the crystallite size is the smallest with 7 nm, while with increasing ruthenium concentration the Table 1. XPS analysis: Mean composition (independent of oxidation state) of the Ru_x@TiO 2 samples, composition of the oxide component (Ru x Ir 1À x O 2 @TiO 2 ), and the overall noble metal (ruthenium + iridium) content given in mol % with respect to the support material TiO 2 . Further details of the quantification are given in the supporting information (cf. This size effect is most prominent with the pure RuO 2 @TiO 2 system (referred to as Ru_100@TiO 2 ) with a mean crystallite size of 22 nm. However, one needs to recall that the XRD technique overestimates the crystallite size in that already small concentrations of large crystallites may dominate the intensities of XRD scans, while diffraction from small crystallites contribute mainly to the background intensity. Therefore, a microscopic technique, such as TEM is needed to countercheck for the size of supported particles (cf. Figure 3). In Figure 3, we can recognize that the dispersion of the supported mixed ruthenium-iridium particles varies with composition. For the case of Ir_100@TiO 2 , clearly many supported particles are discernible with a narrow size distribution. Quite in contrast, pure Ru_100@TiO 2 does not show any sign of supported particles. Instead, most of RuO 2 forms unsupported particles as shown with element mapping (cf. Figure S5). In between these extremes supported particles are clearly visible, but it is also obvious that its concentration decreases with increasing ruthenium content. In Table 2 we summarize the average particle sizes derived from TEM images in Figure 3. The particles size from TEM is systematically smaller than the XRD-derived ones that is explainable since XRD intensities are dominated by the bigger particles.
High resolution STEM together with element mapping supports this view. For pure Ir_100@TiO 2 there is no indication that unsupported IrO 2 particles are formed, while already for the Ru_25@TiO 2 agglomeration of mixed ruthenium-iridium particles is apparent (cf. Figure 4), although most of the particles adhere to the rutile TiO 2 support particles. Element mapping in Figure 4 reveals that all supported particles comprise a mixture of ruthenium and iridium; with 30 � 10 mol % ruthenium that is reconciled with the nominal concentration of 25 mol % ruthenium. A separation into pure RuO 2 and IrO 2 can be ruled out since XRD reveals an average crystallite size of 8 nm (Table 2) for the active component in Ru_25@TiO 2 and such big particles of pure RuO 2 and IrO 2 conflicts with the identical intensity distribution of ruthenium and iridium in the EDS mappings in Figure 4.
The morphology of the pure Ru_100@TiO 2 sample is studied by HRTEM providing evidence that RuO 2 forms a thin layer (1-2 nm thick) of RuO 2 on rutile TiO 2 ( Figure S6), while most of the RuO 2 agglomerates, forming larger particles that do not adhere to the TiO 2 support (cf. Figure S5). These larger RuO 2 agglomerates are responsible for the sharp (101) reflection in XRD (cf.   Table 2). For the comparison of activity data among various supported Ru_x@TiO 2 samples it is required that the space time yield STY is normalized to the active surface area. For mixed ruthenium-iridium powder samples [21] this can be accomplished by measuring the BET surface area. However, for supported catalysts this approach is not reasonable since the catalytically inactive support dominates the BET surface area. Therefore, we measured the relative active surface area by CO uptake experiments, assuming that the number of adsorbed CO molecules is strictly correlated with the number of active sites of the active component and independent of the actual composition of the mixed oxide catalyst. In Table 3 we summarize these experimental results.
As suggested by HRTEM and XRD, the active surface areas of Ir_100@TiO 2 and Ru_25@TiO 2 are the highest, while the ones of Ru_50@TiO 2 and Ru_75@TiO 2 are the lowest. Quite surpris-ingly, the active surface area of Ru_100@TiO 2 is high and apparently conflicts with the large average RuO 2 particle size of 22 nm as derived from XRD. However, HRTEM reveals that the TiO 2 particles are partly covered by a thin RuO 2 layer of 1-2 nm thickness. These layers are likely to be responsible for the observed high active surface area, but do not contribute to the XRD pattern.  Table 3. In the light-off experiments the temperature is linearly ramped with 1°C · min À 1 up to T10 where the conversion X of the methane combustion reaction reaches 10 % with a reaction feed of 2 sccm methane and 8 sccm O 2 balanced by 90 sccm N 2 ; generally, the temperature sweep rate needs to be low enough to maintain steady state conditions during the temperature ramp.

Kinetic data of methane combustion over mixed ruthenium-iridium supported on rutile-TiO 2
These light-off temperatures T10 as well as T2 (X = 2 %) and T5 (X = 5 %) values are collected in Table 4. Ru_100@TiO 2 with RuO 2 as the active component clearly reveals the lowest activity in methane combustion. However, it is equally evident that not the pure Ir_100@TiO 2 (with IrO 2 as the active component) but instead Ru_25@TiO 2 is the most active catalyst in methane combustion for conversions lower than 10 %. When normalizing  the activity data to the number of active sites (cf. Figure 5b), Ir_ 100@TiO 2 , Ru_50@TiO 2 , and Ru_75@TiO 2 show similar activity that is substantially higher than pure Ru_100@TiO 2 , but still significantly lower than that of Ru_25@TiO 2 .
The activity data shown in Figure 5a,b can also be represented as Arrhenius plots (cf. Figure 5c,d) in order to extract kinetic data such as the apparent activation energies E act and the pre-factors STY 0 which are compiled in Table 4. The apparent activation energies are independent of the normalization procedure, but the pre-factor values STY 0 , STYn 0 depend of course on the normalization [Eqs. (4) and (5)] and are separately listed in Table 4.
The highest activity is paralleled by the highest values of the pre-factors, while the apparent activation energies counteract these values by high apparent activation energies for active catalysts and lower values for the less active catalyst. The prefactor for Ru_100@TiO 2 is three order of magnitude lower than that of Ru_25@TiO 2 , making Ru_100@TiO 2 substantially less active than Ru_25@TiO 2 . A similar pattern in the apparent activation energies and pre-factors was encountered for the CO oxidation over mixed ruthenium-iridium oxide catalysts. [21] To conclude the kinetic study, we performed experiments to determine the reaction order R.O. (cf. Figure 6). Here the ln(STY) is shown as a function of the logarithm of the reaction feed. The reaction temperature is chosen for each sample in a way that the conversion is X = 10 % for a reaction mixture of 8 sccm O 2 and 2 sccm methane. After reaching T10 the reaction order is determined by changing the volumetric concentration of oxygen from 8 sccm to 2 sccm in steps of 2 sccm while keeping the methane concentration constant at 2 sccm. Subsequently, the oxygen concentration is kept constant at 2 sccm and the methane concentration is decreased in steps of 0.5 sccm from 2 sccm to 1 sccm. The oxygen concentration is then increased in steps of 2 sccm back from 2 sccm to 8 sccm while the methane concentration is fixed at 1 sccm. Finally, the methane concentration is raised from 1 sccm to 1.5 sccm to 2 sccm, thereby returning to the starting gas composition. This protocol enables an investigation of the reaction order of oxygen and methane over a wide range of reaction conditions from oxidizing (Vol %(O 2 ) = 8 %, Vol %(CH 4 ) = 2 %), stoichiometric (Vol %(O 2 ) = 4 %, Vol %(CH 4 ) = 2 %) to reducing (Vol %(O 2 ) = 2 %, Vol %(CH 4 ) = 2 %) conditions. The reaction orders are summarized in Table 4. It turns out that the reaction order in O 2 in all cases is zero, while the reaction order in methane is close to unity. From these values we infer that the oxygen supply on the catalyst's surface is not rate limiting, while methane activation seems to dominate the reaction kinetics. Together with the apparent activation energies E act these kinetic data culminates in the formal kinetic law [Eq. (6)]: with the methane concentration [CH 4 ]. This kinetic law is consistent with that found for methane reaction on Ni-based catalysts. [31] For comparison reason we carried out similar kinetics experiments for unsupported mixed ruthenium-iridium oxide powder catalysts with the same composition x as studied here and already characterized in a previous work. [21] These results are summarized in the supporting information (cf. Figure S7 and S8, Table S5), and similar conclusions can be drawn about activity and kinetics as for the supported catalyst. Pure RuO 2 turns out to be the least active catalyst in the methane combustion, while the Ru_25 sample is the most active catalyst. Also, the apparent activation energies and the reaction orders vary similarly as the corresponding supported samples" indicating that the rutile-TiO 2 support does not significantly affect the intrinsic catalytic activity by additional metal-support interactions.
In Figure S9 light off experiments for both the unsupported and supported material are depicted. While for low conversion up to 10 % mainly the intrinsic activity (microkinetics) is reflected, the performance of the catalysts at higher conversion is increasingly more governed by heat and mass transfer.  Table S5. Transport limitations seem to occur quite late allowing Ru_25, Ru_75 and Ir_100 to reach 90 % conversion at temperatures lower than 475°C. In the case of the supported materials the trend of the light-off curves changes. Table 5 summarizes the light off temperatures T50, T70 and T90 for both unsupported and supported materials.
The lowest T90 value is revealed for the unsupported Ru_25 catalyst that underlines its high catalytic activity throughout the  After the catalytic and kinetic experiments, the Ru_x@TiO 2 catalysts were post-characterized by TEM (cf. Figure S10). A comparison with the as-prepared Ru_x@TiO 2 catalysts reveals that no reaction-induced alterations are discernible, i. e., the catalysts are stable under the applied reaction conditions.

Synthesis
We successfully synthesized mixed Ru x Ir 1À x O 2 oxide supported on rutile-TiO 2 with well-defined composition x. In all cases 5 mol % of active component was employed, in order to allow for in-depth characterization of the active component by XRD, XPS, TEM, and CO uptake experiments. Most notably, XPS and also STEM element mapping indicate that the composition x of mixed Ru x Ir 1À x O 2 corresponds to that of the molar ratio of iridium and ruthenium precursors employed in the synthesis, clearly demonstrating our high level of control on the composition x. We should emphasize that a similar level of control of the composition x of mixed Ru x Ir 1À x O 2 cannot be achieved by simple impregnation methods.
Rutile-TiO 2 particles serve here as nucleation centers in the Pechini method. While the dispersion of pure IrO 2 particles turns out to be high, deposition of pure RuO 2 on rutile TiO 2 is less clear-cut. According to HRTEM, RuO 2 forms thin layers on rutile-TiO 2 and in addition unsupported larger RuO 2 particles. This observation is consistent with a previous study, where pure RuO 2 was reported to grow epitaxially on rutile-TiO 2 but not on anatase TiO 2 particles. [32] TEM clearly reveals that the other Ru_x@TiO 2 samples with ruthenium compositions of 25 mol %, 50 mol %, and 75 mol % form supported particles besides agglomerates of unsupported mixed ruthenium-iridium oxide particles with the density of supported particles being significantly lower than that of Ir_100@TiO 2 .
Obviously, pure IrO 2 and pure RuO 2 adhere differently to the supporting rutile-TiO 2 particles. Surface science studies revealed rutile RuO 2 to form a strained pseudomorphic RuO 2 (110) layer adopting the surface lattice constants of the supporting rutile-TiO 2 (110), [24] while rutile IrO 2 (110) grows with much less strain on rutile-TiO 2 (110). [25,33] Since the surface energy of IrO 2 is substantially higher than that of TiO 2 and the interfacial energy IrO 2 /TiO 2 remains high, this may explain why IrO 2 forms particles rather than a wetting film on rutile TiO 2 with a small surface and interface area. Due to the quasi pseudomorphic growth of RuO 2 on rutile-TiO 2 , the interfacial energy RuO 2 /TiO 2 is low so that RuO 2 is now able to wet partly the TiO 2 particles, despite similar surface energies of RuO 2 and IrO 2 . [34] Further growth of RuO 2 proceeds in separate particles with its native lattice constants. Obviously, this is energetically more favorable than increasing the RuO 2 film thickness beyond 1-2 nm.
CO uptake experiments successfully quantify the number of active sites or the relative active surface area for Ru_x@TiO 2 (cf. Table 3). The number of active sites is found to be high for Ir_100@TiO 2 , while with increasing concentration of ruthenium the number of active sites decreases by a factor of 2-3. This decline in active sites can be related to the decreased concentration of supported particles. However, for pure Ru_ 100@TiO 2 the number of active sites (cf. Table 3) is surprisingly high and not reconciled with large RuO 2 particles evidenced by XRD. The high active surface area of Ru_100@TiO 2 is attributed to the observed thin layer growth of RuO 2 on rutile-TiO 2 .

Methane combustion activity
We studied the activity and the kinetics of catalytic methane combustion over supported Ru_x@TiO 2 and unsupported powder Ru_x catalysts with varying compositions x ranging from pure RuO 2 to pure IrO 2 in steps of 25 mol %. To the best of our knowledge, these are the first catalytic methane combustion studies of this mixed oxide system. A comparison between supported and unsupported catalysts reveals practically identical trends in the activity for X = 0 % to X = 10 % and identical kinetics as a function of the composition. Therefore, the rutile TiO 2 support seems to have virtually no impact on the microkinetics of the methane combustion. However, for higher conversions the different dispersions of the active component on the support material seems to differently influence the heat and mass transfer which lead to re-ordering of the light-off curves. We have to bear in mind that this situation may change when the TiO 2 support is replaced for instance by CeO 2 or Al 2 O 3 , where charge transfer and spill-over phenomena are known to be operative. [35] At least for Pd-based catalysts, the standard catalyst for methane combustion, [3] the support has shown to play an important role. [1,36] For Pd embedded in CeO 2 the activity in methane combustion could be increased considerably. [37] Recall that under typical reaction conditions of methane combustion not metallic Pd but rather PdO is the active phase. [38] In our low conversion catalytic tests (cf. Figure 5) the most active catalyst turned out to be Ru_25@TiO 2 (Ru_25) followed by Ir_100@TiO 2 (Ir_100) and then with increasing ruthenium concentration the activity declines steadily with pure Ru_100@TiO 2 (Ru_100) being by far the least active catalyst. Pure IrO 2 is substantially more efficient than pure RuO 2 in methane combustion. This finding is in accordance with surface science experiments which demonstrated that methane can be activated even at low temperature on IrO 2 (110), [20] while under the very same conditions RuO 2 (110) is virtually inactive. [39] Obviously, methane activation is an important step in methane combustion and therefore Ru_100@TiO 2 is the least active catalyst among the homologous series Ru_x@TiO 2 . From these surface science studies it is quite surprising that pure RuO 2 is active at all in methane combustion. But the observed activity of RuO 2 is consistent with a recent study where RuO 2 supported on Gamma Alumina was discussed as alternative catalyst for methane combustion. [40] Pure RuO 2 (110) is, however, known to be an excellent catalyst in the CO oxidation reaction, [22] while on IrO 2 (110) the adsorption of CO was found to be much stronger than on RuO 2 . [41] In fact, RuO 2 has shown to be more active in the CO oxidation than IrO 2 . [21] Therefore, we expected a synergism effect of ruthenium on the activity of methane combustion. Indeed, 25 mol % ruthenium (75 mol % iridium) in the mixed oxide catalysts reveals a significant improvement of the catalytic performance in methane combustion. In order to decide whether this improvement is due to the higher dispersion of Ru x Ir 1À x O 2 on rutile TiO 2 or to an increase in intrinsic activity, we needed to normalize the STY values to the number of active surface sites as provided by CO uptake experiments (Table 3). Since the trend in activity among Ru_ x@TiO 2 is not affected by this normalization procedure, we ascribe the improved performance of the Ru_25@TiO 2 sample in methane combustion to an increase in the intrinsic activity of the mixed oxide. A very similar trend in activity is found for the unsupported Ru_x catalyst (supporting information) when normalizing STY to the BET surface area.
The superior performance of Ru_25@TiO 2 in methane combustion points towards an intimate interplay of methane activation (by IrO 2 ) and the subsequent formation of an oxygenated reaction intermediate from the methyl fragment, such as CO, CH 2 O, or CHO 2 or the final oxidation step to CO 2 by RuO 2 . Of course, RuO 2 may equally promote the oxidation of the abstracted H from methane to form water. This may be evident when again considering corresponding surface science experiments: Hydrogen adsorption and subsequent annealing of the sample leads to water formation between 127°C to 227°C for RuO 2 (110), [42] but results in a broad water desorption feature ranging from 127°C to 477°C for IrO 2 (110). [43] Also quite surprising is the observation that Ru_75@TiO 2 and Ru_75, although being not the most active catalyst, are substantially more active than Ru_100@TiO 2 and Ru_100 (Figure 5d) by decreasing the reaction temperature T10 by 80°C. This means that already a relatively small concentration of iridium improves the low temperature methane activity of RuO 2 considerably.
The observed kinetics (for conversions lower than 10 %) of the methane combustion over unsupported Ru_x (Table S3) and supported Ru_x@TiO 2 (cf. Table 4) are virtually identical, corroborating the view that the rutile-TiO 2 support does hardly affect the catalytic behavior of the active component. From the kinetic data of Ru_x@TiO 2 in Table 4, the catalysts can be grouped in two categories, one with a high concentration of ruthenium (50 mol %, 75 mol %) and the other with low ruthenium concentration (0 mol %, 25 mol %, 50 mol %). The apparent activation energies and the pre-factors of the ruthenium-rich catalysts (80-90 kJ · mol À 1 ) are significantly lower than those of the iridium-rich catalysts (105 � 2 kJ · mol À 1 ). The most active catalyst Ru_25@TiO 2 reveals the highest apparent activation energy and the highest pre-factor, resulting in a catalyst that at higher reaction temperature is even more active than the others. Obviously, the higher apparent activation energy is overcompensated by the high value of the pre-factor. This compensation effect is known as the Cremer-Constable relation. Similar apparent activation energies as for the present iridium-rich catalysts were also reported for supported Pd and Pt based catalysts. [3] For methane combustion over both Ru_x@TiO 2 and Ru_x, the reaction order in methane is close to one, while that in oxygen is zero. Similar reaction orders in methane and oxygen are reported for Pd-based catalyst. [3] These reaction orders for the total methane oxidation reaction over Ru_x@TiO 2 and Ru_x agree remarkably well with those reported for other noble metal catalysts. [44] For the unsupported material Ru_25 and Ru_75 remain the most active catalysts in the entire temperature and conversion range. Due most likely to transport limitations which vary among the supported materials with their varying dispersions and particle sizes, the catalytic activity of Ru_25@TiO 2 decreases with increasing conversion and Ir100@TiO 2 and Ru_75@TiO 2 become the most promising catalysts. At high conversions, the catalytic performance of Ru_75 and Ru_75@TiO 2 approaches even that of pure Ir100 and Ir100@TiO 2 catalysts.
In Table 6 the catalytic performance (T50 and T90) of Ir100@TiO 2 and Ru_75@TiO 2 in the catalytic methane combustion is compared to supported Pd catalysts , defining the benchmark catalyst for low temperature methane combustion. Unfortunately, a direct comparison of T50 and T90 values is hampered by differences in the mass of catalyst, the loading of the active component, the gas composition fed to the reactor, and the contact time with the catalyst material; all these factors affect the light-off curves.
From the T50 and T90 values summarized in Table 6 we conclude that the activities of both Ir_100@TiO 2 and Ru_75@TiO 2 are comparable to that of Pd@Al 2 O 3 . A direct comparison of the Ir_100@TiO 2 and Ru_75@TiO 2 with Pd@TiO 2 reveals that T90 is 44 to 66°C lower. However, it needs to be noted that in our study the methane fed is two times higher and the amount of the active component is about 2 to 3 times lower than in the study of Pd@TiO 2 . Besides activity, a critical issue in catalytic methane combustion is catalyst poisoning by water, sulfur and other constituents in the exhaust gas [51] that needs to be studied for mixed ruthenium/iridium oxide catalysts in the future. We do not expect that IrO 2 catalysts will replace Pd-based catalysts for the methane combustion due to the limited amount of mined Ir (6 t/a). On the other hand, RuO 2 is not an efficient oxidation catalyst for methane combustion, and under oxidizing conditions, RuO 2 is vulnerable to overoxidation to form volatile RuO 3 and RuO 4 above 500-600°C. [52] However, mixing 25 mol % or less of IrO 2 into RuO 2 improves the activity steeply and thermally stabilizes RuO 2 . Since Ru is significantly more abundant than Ir (30 t/a) and about eight times less expensive than Pd, RuO 2 with little IrO 2 could therefore be a promising option for catalytic methane combustion.

Molecular insight
From kinetic data only, we cannot decipher the actual reaction mechanism of methane combustion over Ru x Ir 1À x O 2 . Instead, we would need to perform operando spectroscopy experiments to identify the reaction intermediates and to conduct first principles kinetics simulations to simulate the experimental kinetics. However, from experimental kinetic data we can draw some mechanistic conclusions about the catalytic reaction. For instance, from a reaction order in methane of one and a reaction order in oxygen of zero we can safely conclude that methane activation constitutes the rate limiting step in the combustion process. A similar conclusion was previously drawn for supported Ir-clusters. [53] A reaction order in oxygen of zero means that always enough surface oxygen is available at the active sites, thus favoring a Mars-van-Krevelen mechanism over the Langmuir-Hinshelwood mechanism. For a firm conclusion about the question of Langmuir-Hinshelwood versus Mars-van Krevelen mechanism one needs to conduct isotope labelling experiments of 16 O and 18 O [31,54] which are, however, beyond the scope of the present paper. First principle calculations for methane combustion over Ru x Ir 1À x O 2 are not available. However, for a similar catalyst system PdO(101), a recent first principle microkinetic study [38] has shown that the reaction mechanism constitute a combination of Langmuir Hinshelwood and Mars van Krevelen mechanism depending on the reaction temperature.
In the methane activation step two different surface sites are involved. Vacant undercoordinated metal sites (likely Ir sites) are required to adsorb the methyl group, while undercoordinated surface oxygen sites can accommodate the abstracted hydrogen species. Therefore, these surface species participate in the catalytically active sites, i. e., the number of active sites is proportional to the number of surface metal sites.
There are DFT studies available for the methane activation on IrO 2 (110). [16,18] From these calculations the apparent activation energy for methane activation turns out to be negative since the adsorption energy of intact methane is with 60 kJ · mol À 1 higher than the actual activation barrier for dissociation (about 30 kJ · mol À 1 ). This result is not reconciled with the observed apparent activation energies of 80-110 kJ · mol À 1 depending on the composition of the mixed RuÀ Ir oxide. However, this variation in apparent energy fits remark-ably well to the adsorption energy of oxygen on IrO 2 and RuO 2 . From thermal desorption spectroscopy [41,55] of IrO 2 (110) and RuO 2 (110) the OÀ Ir bond is shown to be slightly stronger than the RuÀ O bond. [56] Under lean methane oxidation reaction conditions the catalyst's surface is likely saturated and therefore blocked by adsorbed oxygen so that methane activation is suppressed. The apparent activation energy may therefore correspond to the activation of oxygen desorption, thereby liberating metallic adsorption sites for methane activation.

Conclusions
We present here stable activity data of methane combustion on mixed ruthenium-iridium based catalysts Ru_x@TiO 2 supported on rutile-TiO 2 in comparison with those on unsupported Ru_x catalysts. Both types of mixed catalysts are prepared by a (modified) Pechini synthesis in order to ensure full control of the composition x in the mixed Ru x Ir 1À x O 2 . From the direct comparison of supported and unsupported catalysts, it turns out that the rutile-TiO 2 support does not affect the catalytic performance in methane combustion.
Pure Ir_100@TiO 2 is much more active in the combustion of methane than pure Ru_100@TiO 2 , with the former exhibiting an onset temperature of about 220°C that is similar to onset temperatures encountered for Pd and Pt based catalysts. [3,44] The most active catalyst among the series of Ru_x@TiO 2 and Ru_x turned out to be Ru_25@TiO 2 (Ru_25) with an apparent activation energy of 105 kJ · mol À 1 and a required reaction temperature of 313°C to achieve 10 % conversion at GHSV of 164.000 h À 1 (corresponding to 345.000 ml · g À 1 · h À 1 if normalized to mass of catalyst). The catalytic performance of IrO 2 @TiO 2 and Ru 0.75 Ir 0.25 O 2 @TiO 2 in methane combustion in terms of T50 and T90 (the reaction temperature to achieve 50 % and 90 % conversion) is comparable to that reported for typical Pd catalysts (cf. Table 6). Ru_75@TiO 2 and Ru_75, although being not the most active catalyst, are substantially more active than Ru_100@TiO 2 and Ru_100 by decreasing the T10 temperature by 80°C; the onset temperature of Ru_75@TiO 2 is about 230°C. This means that already a small concentration of iridium improves the activity of RuO 2 in methane combustion considerably.
From a reaction order of unity in methane we conclude that the methane activation step is rate determining in the methane combustion over mixed ruthenium-iridium based catalysts. There is evidence that the apparent activation energy is related to the activation energy of molecular oxygen desorption from the catalyst's surface. In this way active metal sites are liberated to accommodate the methyl group in methane activation step. Methane activation by IrO 2 , although mandatory, is however only part of the story. The further oxidation of methyl to CO 2 is apparently another important step which is promoted by the addition of RuO 2 .