Initially, 0.11 g of Ruthenium (Ruthenium (III) chloride hydrate) was prepared. 3 ml of water was added by dripping to 1 g of multi-walled carbon nanotubes (purchased as Nanocyl-3100). Then, 1.32 ml of ethylene glycol was added to the carbon nanotube and soaked. Finally, the ruthenium solution in ethylene glycol solution was added, and the paste formed was dried under vacuum during three days. The calcination was performed under airflow at 100 ml∙min⁻¹, raising the temperature from 30˚C to 350˚C at 10˚C/min, and held for 2 hours. The reduction was performed with hydrogen flow at 100 ml∙min⁻¹ from 30˚C to 400˚C, at 10˚C/min, and kept for 2 h. The catalyst was denoted as 5%Ru/MWCNT.

We also prepared and functionalized CNT sample that was impregnated with the same metallic precursors, according to the procedure reported previously [32].

The specific surface area of the catalyst was determined using nitrogen physisorption at −196˚C in the NOVA 1200 equipment (Surface Area & Pore Size Analyzer) of Quantachrome Instruments. Prior to the physisorption analysis, the materials were pretreated under vacuum at 200˚C for 17 h. The specific surface area was determined using the BET methodology.

Transmission Electronic Microscopy (TEM) analyses were performed in a JEOL JEM 2100 Microscope, with maximum acceleration voltage 200 kV, and resolution of 0.23 nm at the point and 0.14 nm on the network, with maximum magnification of 1,500,000 times.

X-ray diffraction (XRD) analyses were performed in a Rigaku equipment with an acquisition in the radiation incidence angle range of 2˚ < 2θ < 90˚. Identification phases in the sample were obtained using the PDF (Powder Diffraction File PDF2-2003) datasheets as database through the Search-Match program, comparing to their positions in the experimental data.

The Raman spectra were acquired in the range of 100 - 3500 cm⁻¹ with a Renishaw in Via confocal Raman microscope in backscattering geometry equipped with a CCD detector (−70˚C) and HeNe laser supplying the photon at 532 nm. The laser excitation power was kept below 10 mW on the sample surface to minimize the local heating. The total acquisition time was 300s. All the spectra of the samples after reaction were recorded under identical conditions at room temperature.

The surface of the catalysts was studied using X-ray photoelectron spectroscopy (XPS). X-ray photoelectron spectra were recorded on a Leybold-Heraeus LHS 10 spectrometer using a non-monochromatized Al-K-α source (1486.7 eV).

Thermal analyses of the samples before and after reaction were performed on a DTG-60H Shimadzu with a TG-DTG simultaneous analyses accessory. The measurements were conducted under a flow of nitrogen/oxygen mixture (55 mL/min N₂/8 mL/min O₂) from 25 to 1000˚C at a heating rate of 10˚C/min.

The tests were performed in a Microactivity-Effi (PID company) with a Hastelloy X tubular reactor of 9.1 mm diameter and length of 370 mm at atmospheric pressure, varying the temperature range and weight hourly space velocities (WHSV). The feed flows downwards through the catalytic bed and a condenser before running the test. The temperature was measured with a thermocouple and a PID controller. The total mass was 100 mg (20 mg of catalyst and 80 mg of carbide). Prior, the catalyst was reduced in-situ with pure H₂ flow (99.999%, Air Products) for 120 min, at 10˚C∙min⁻¹ up to 450˚C. After purging with N₂ flow, the gas was switched to the feed mixture (CO₂:O₂:N₂). The exit gases were measured online in a gas chromatograph (GC2010 Plus Shimadzu) with a Carboxen 1010 plot column.

The specific surface area of the catalyst Ru/MWCNT was 285 m²/g, and the pore volume 1.33 cm³/g. The surface area of the functionalized sample 5% Ru/MWCNTf (functionalized) was 371.8 m²/g. The isotherms of the Nanocyl sample and the corresponding pore size distribution are shown in Figure 1. It displays a type IV adsorption curve with hysteresis, which is associated with the development of a pore network. The MWCNTf displayed similar profiles (not shown).

The pore size distribution curves suggest various types of pore sizes. The small pores around 4 nm are aggregated pores of CNTs. Regarding the pore size distribution, mostly can be considered bimodal (micro and mesopores) types, which agreed with Ma et al. [30].

The images of the transmission electron microscopy (TEM) in Figure 2 show that the impregnation of ruthenium nanoparticles over carbon nanotubes was well succeeded, using this preparation method, as proposed in this work.

Noteworthy is that the depositions of Ru particles (Figures A and B) are well dispersed on the external wall of the CNT.

In fact, transmission electron microscopic analyses generate two-dimensional images, allowing determining the location of the particles in some selected regions.

We calculated the average diameter of the nanotubes and particles from the images using the ImageJ software and obtained the particle size distribution from a representative group of 12 images.

The average diameter for the Ru located at the external wall of the carbon nanotubes was 2.3 ± 0.6 nm, very close to the calculated by the X-ray diffraction, XRD (1.7 nm).

Figure 3 displays the average external and internal diameter of the carbon nanotubes, 13.4 ± 2.7 and 6.8 ± 1.6 nm, respectively. The sample space presents 32 tubes.

Figure 4 shows the X-ray diffractograms for the MWCNT and of the catalyst. According to the literature [33], the MWCNT present characteristic peaks at 2θ ≈ 26.2˚, 43.0˚, 55.0˚, and 77.0˚, which represent, respectively, the (002), (100), (004) and (110) planes. The first peak indicates the presence of the effect of different overlapping graphene sheets, in a concentric mode [34], characteristic of multi-layered carbon nanotubes.

The profile obtained for MWCNT (Figure 4) showed the peaks of the planes (002) and (100) at 2θ equal to 26.05˚ and 43.03˚, respectively. In addition, such peaks are broad, characteristic of amorphous materials, as expected for carbon nanotubes.

The diffractogram of the catalyst with ruthenium showed the presence of two peaks with low intensity at 34.8˚ and 54.8˚, which are assigned to RuO₂ (Ref. code 40-1290 and overlap identified by “*” in Figure 5) present at the external surface of the MWCNT, in agreement with the images of electron transmission microscopy analyses. For comparison, the XRD of the functionalized MWCNTf diffractogram is presented in supplementary information S1.

The crystallite size of ruthenium was calculated using Scherrer’s equation [35]. For this, the peak of 34.8˚ was used, which resulted in a crystallite size of 1.7 nm in agreement with TEM result of 2.3 ± 0.6 nm. Rojas et al. [36] found metallic Ru particles of 2.5 nm on the Ru/MWCNT sample.

The thermogravimetric analyses (TGA) and their derivate (DTG) profiles of the Ru/MWCNT catalysts are shown in Figure 6, presenting the weight loss of the pure carbon nanotube (MWCNT) and of the catalyst (Ru/MWCNT).

The TGA analysis (red curve) shows that the support (MWCNT) and the catalyst (5% Ru/MWCNT) present slightly different profiles. For the support, the mass loss was approximately 84% at 515˚C, and for the catalyst, was 92% at approximately 600˚C, which indicates decomposition of the carbonyl and carboxylic groups present in carbon nanotubes, in accordance with Saleh et al. [37].

Raman spectroscopic analyses provide information about the chemical composition and crystalline structure of the sample [38]. Figure 7 shows the spectra of the carbon nanotube and the catalyst. The spectrum of the MWCNT showed the peaks at bands D (≈1340 cm⁻¹), G (≈1572 cm⁻¹), 2D (≈2663 cm⁻¹), G + D (≈2919 cm⁻¹), in accordance with Zdrojek et al. [39] and noise bands between 100 and

200 cm⁻¹, as observed by Ivanova et al. [40]. However, the spectrum of the catalyst sample showed extra peaks, which could be correlated to impurities (≈2330 cm⁻¹, 800 cm⁻¹ and 270 cm⁻¹).

The band G (graphite) corresponds to the vibrations of the sp² bonds between the carbons in the same plane and the band D to the disorder in the carbon structure, that is, the sp³ bonds between the carbons. The ratio between the areas of these peaks (I_D/I_G) allows determining the quality of carbon nanotubes [41], representing the density of structural defects. The higher the ratio, the more defects (sp³ connections) are present in relation to the expected (sp² connections). The ratio of the pure support was 1.98.

This ratio increased after the addition of ruthenium (2.80). In fact, it indicates that there is an exchange of double bond for single bond in the carbons, with the formation of new covalent bonds. Figure 7 shows the peaks in the region at lower wavenumbers, indicating the presence of the metal oxide, probably ruthenium oxide, RuO₂ (set of 2 small peaks, between 510 and 1300 cm⁻¹) as indicated in the figure and in accordance with the literature [42]. In fact, RuO₂ is probably the main oxide present at the external surface of the MWCNT. Literature reported different oxidation states of the Ru, but in particular the Ru⁺⁴ and Ru⁺⁸, besides other oxidation states that were reported in literature. For comparison, the Raman spectrum of the functionalized MWCNTf is also presented in supplementary information S1.

The XPS analyses were performed for identifying the oxidation state of ruthenium oxide deposited on the outer surface of the CNT, and also identifying the functional groups of the MWCNT. All binding energies were corrected, referencing the C1s (284.6 eV) peak of the contamination carbon as an internal standard. The XPS experiments were obtained for the functionalized MWCNT with the same content of 5%Ru. The XPS results are important to explain the presence of the RuO₂ at the surface, not well seen in the XRD and Raman spectra. In fact, Figure 8(a) shows the screening spectrum of carbon, ruthenium and have oxygen.

Figure 8(b) shows the deconvolution of C1s peak. Results showed graphitic carbon at 284.3 eV [41]. Peaks at 284.5 and 290.5 eV are assigned to the carbon atoms (C-C) attached to groups containing oxygen (C=O) in their structure [43]. The electronic transition of orbitals π → π* was associated with the peak at 290.5 eV. These results indicate the external surface and also that the graphite layers retained their original structure [28] [31].

The spectrum the Ru 3d is characterized by a pair corresponding to the 5/2 and 3/2 spin-orbit components located at 281.4 and 285.6 eV, respectively (not

shown). Figure 8(c) displays the O 1s core level spectrum of the RuO₂ on distinct maxima located at 530.1 eV, associated with oxygen adsorbed on the surface and 532.0, 532.8 and 534.0 eV, assigned carboxylic and hydroxyl groups. These peaks were related to the metallic oxides on the external surface of the CNTs.

The catalyst (5% Ru/MWCNT) was tested in the CO oxidation reaction, for different conditions (3 groups), as presented in Table 1, to show the influence of temperature and space velocity. The conversion and selectivity results are presented in this table and illustrated in Figure 9. Each experiment was carried out in triplicate with the same sample and conditions, and constant mass.

Tests for condition C3 were obtained at WHSV 13,285 (mL/min.g) and mass of 20 mg, consequently, low contact time. The CO and O₂ conversions with temperatures between 50 and 270˚C are displayed in Figure 9. In fact, the conversions of CO and O₂ increased, from 50˚C, reaching the maximum conversions of 75% and 100%, respectively, at 150˚C, but the selectivity of CO₂ was around 85%.

The results for condition C1, for lower space velocity (WHSV) 2000 (mL/min∙g)

and CO:O₂ = 1:0.5, and temperatures between 300˚C and 500˚C, showed conversions of carbon monoxide 100%, and of O₂ around 95%. However, the CO₂ yield was higher than 100%.

Conditions C2 were tested, diluting the reagents at the feed entrance (CO:O₂:N₂ = 1:0.47:5.91), by increasing the volume of the inert gas N₂ and the space velocity of 2657 mL/min.g, and increasing the temperature from 200˚C to 400˚C. The carbon monoxide and O₂ reached maximum conversions of 100%, suggesting total oxidation of CO; however, the CO₂ selectivity was 88%.

We calculated the yield, using equation: Y(j) = X(CO) × S(j), and results are shown in Figure 9(d), with increasing conversion. As shown yield increased linearly with the conversion.

The thermogravimetric analysis (TGA) was performed after reaction in air atmosphere (Figure 10). Noteworthy is that the profile shows gain of mass from 250˚C up to 700˚C, and the DTG profile shows a maximum peak at 320˚C, decreasing then up to 750˚C, and afterwards increases without mass alteration. It evidences an oxidation step of the metallic particles starting at 450˚C up to 700˚C. According to these results, the mass gain is suggested to be attributed to the carbon formation, and principally to the oxidation of Ru species [11]. These last results may explain the phase transformation without decomposition of the CNT.

The thermogravimetric analysis (TGA and DTA) results indicate between 200˚C and 400˚C the reduction besides the oxidation of RuO₂ to RuO₄ due to the gain of mass, as shown in Figure 9, that is:

RuO 2 + O 2 → RuO 4 (5)

In fact, it supports the original hypothesis of spectroscopic analyses that the metallic Ru nanoparticle are transformed into core-shell structure (Ru@RuO₂) under reaction conditions, according to Joo et al. [28].

Moreover, Figure 10 shows gain of mass between 450˚C and 720˚C and DTA profile evidence transformation of phases attributed mainly the oxidation of RuO₂ with a gain of mass to RuO₄, in good agreement with TGA and DTA results of the original catalyst in Figure 6(b) and in accordance with the XPS results. The O1s core level spectrum of the RuO₂ displays oxygen adsorbed on the surface, which is associated with oxygen from Ruthenium oxides. There are no references reported in the literature concerning this behavior.

According to the literature [45] authors suggested that the mechanism of CO oxidation on the RuO₂ surface is explained by the adsorption and reaction of atomic oxygen species on the coordinative unsaturated Ru atoms and bridge position surface oxygen atoms as verified by Mars-van Krevelan.

Therefore, we propose a reaction mechanism that follows the scheme presented below. In the first stage the RuO₂ present as nanoparticles can be reduced in the presence of CO molecule, with the formation of Ru⁰ metallic. Besides this RuO₂ can also be reduced with the formation of a carbonate specie ( Ru 2 CO 3 − ) and in sequence to the formation of Ruthenium carbonate (C₃O₉Ru₂) that is inert, but can be decomposed. Then the metallic Ru⁰ is oxidized regenerating the RuO₂, which completes the oxi-reduction stage. However, the MWCNT present OH⁻ hydroxyls as evidenced by XPS, which can combine with the carbonate and formation of CO 3 − + OH → HCO 3 − + O − , releasing O⁻ at the surface of the CNT and react with CO to form CO₂, which was observed in all experiments, indicating higher yields than realeased from the direct oxidation of CO with oxygen.

The following scheme represents the mechanism of the CO oxidation on RuO supported on MWCNT:

RuO 2 + CO → Ru 0 + CO 3 − (3)

RuO 2 + CO → Ru 2 CO 3 − → C 3 O 9 Ru 2 (4)

RuO 2 + O 2 → RuO 4 (5)

Ru 0 + O 2 → RuO 2 (6)

CO 3 − + OH → HCO 3 − + O − (7)

CO + O − → CO 2 (8)

Therefore, this agrees with Park et al. [45], suggesting the mechanism of CO oxidation on the RuO₂ surface through the adsorption and reaction of atomic oxygen species (O⁻), as identified by XPS, which can be attributed to the coordinative unsaturated Ru atoms.

However, for higher temperatures 300˚C - 400˚C one observed higher release of CO₂, which can be explained by the Bourdourd reaction and carbon oxidation ( 2CO → C + CO 2 ; C + O 2 → CO 2 ), that were observed by TGA analysis too.

In fact, this support MWCNT does not supply oxygen like the reducible mixed oxides support. But, it can be attributed to the presence of Lewis sites (Equation (7)) with the adsorption of molecular oxygen and release of atomic oxygen for the reaction with CO (Equation (8)), but depending also on the temperature range can influence the CO₂ yield.