TiO2 catalysts

Journal of Catalysis 301 (2013) 141–153

Contents lists available at SciVerse ScienceDirect

Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat

Mechanistic study of low temperature CO2 methanation over Rh/TiO2 catalysts Alejandro Karelovic ⇑, Patricio Ruiz Institute of Condensed Matter and Nanosciences – Molecules, Solids and Reactivity (IMCN/MOST), Université catholique de Louvain, Croix du Sud 2/17, 1348 Louvain-La-Neuve, Belgium

a r t i c l e

i n f o

Article history: Received 12 November 2012 Revised 7 February 2013 Accepted 11 February 2013

Keywords: CO2 methanation Operando-DRIFTS Effect of particle size Rh/TiO2 Low temperature Hydrogenation Reaction order

a b s t r a c t CO2 methanation at low temperature and atmospheric pressure was studied over Rh/TiO2 catalysts focusing on the effect of Rh particle size on the activity and reaction mechanism. Catalysts with different Rh contents (0.5–5 wt.%) were prepared in order to obtain different mean cluster sizes. The activity was measured between 85 and 165 °C, with a H2/CO2 ratio equal to 4. The rate of methane production per surface Rh atoms increases as metal particle size increases up to ca. 7 nm. Beyond this size, the rate does not change appreciably. Higher activation energies (up to 28.7 kcal/mol) are obtained for catalysts with small cluster size (ca. 2 nm), whereas for larger particles (>7 nm), the activation energy is lower and does not change with size (ca. 17 kcal/mol). Reaction order with respect to CO2 is near zero for large clusters, whereas it decreases to 0.36 for lower size clusters. From the analysis of adsorbed species using operando-DRIFTS, it is proposed that smaller Rh particles tend to bind CO(ads) intermediate stronger than larger ones. The activation energy for the dissociation of adsorbed CO species does not vary with Rh particle size, which suggests that smaller particles are not intrinsically less active, but they present less active sites than larger ones. The study of the kinetic parameters permits to propose that CO(ads) dissociation is aided by the presence of H species and that a likely surface intermediate is Rh carbonyl hydrides. Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction The recycling of CO2 has become an important topic in recent years. As carbon dioxide is one of the main contributors to greenhouse effect and hence to climate change, there is a growing interest in its use as a feedstock in chemical processes [1–3]. Although CO2 can be converted to valuable chemical products through organic syntheses, the volume of such a production cannot even approach the large volumes of fuels consumed in the world whose production is two orders of magnitude larger [1,2]. Hence, it is important to convert CO2 to fuels or raw materials which must be also easily transportable. Methane is suitable for this because it benefits from the existing infrastructure for transport and storage of natural gas. CO2 methanation (CO2 + 4H2 M CH4 + 2H2O) has been studied extensively, using different types of metals and supports. Noble metals, especially Ru and Rh, are very active and selective at low temperatures [4–8]. Regarding the choice of support, it has been shown that Rh/TiO2 is one of the most active catalysts for the reaction, being up to one order of magnitude more active than Rh/SiO2 and Rh/Al2O3 [9]. The enhanced activity of Rh/TiO2 systems has been attributed to an electronic interaction between the metal and the support [10,11] or to the interaction of Ti3+ ions located ⇑ Corresponding author. E-mail address: [email protected] (A. Karelovic). 0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcat.2013.02.009

at the edge of TiO2 islands with the CO adsorbed on Rh. It has been suggested that these interactions facilitate the breaking of C@O bond, thus increasing the overall activity [12,13]. Different mechanisms have been proposed for CO2 methanation. The first mechanism involves the adsorption of CO2 on the support and its reaction with H(ads) species formed in the metal which leads to formate intermediate (COOH) at the metal–support interface. The formates can give rise to CO(ads) species which are subsequently hydrogenated to methane [14–17]. The second mechanism involves the direct dissociation of CO2 to CO(ads) and O(ads) on the metal surface, with CO(ads) being subsequently hydrogenated to CH4 [4,8,18,19]. The dissociation of CO(ads) has been generally recognized as the rate-determining step of the reaction [7,13,18,20]. The question whether the dissociation of CO(ads) proceeds via an step assisted by hydrogen or not has also been discussed [21–25]. Additionally, few works have addressed the effect of metal particle size in CO2 methanation reaction. Some authors have shown that smaller particles give the higher intrinsic activity [6], whereas others reported increased rates when metallic clusters are large [26–29]. The present work aims to clarify some aspects of the activity of Rh catalysts toward CO2 methanation. The objective is to study the effect of Rh particle size on the activity and mechanism of CO2 methanation at low temperatures using kinetic measurements and operando-DRIFTS experiments. We show that large metal particles (>7 nm) provide the higher intrinsic activities. The study of

142

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

the reactivity of surface reaction intermediates allows us to propose that smaller metal particles bind CO(ads) intermediates stronger than larger ones, leading to more negative reaction orders with respect to CO2 and higher apparent activation energies. Moreover, the energy barrier for CO(ads) dissociation does not seem to depend on particle size. 2. Experimental 2.1. Catalyst preparation and characterization The catalysts were prepared by wet impregnation. RhCl3H2O (Alfa Aesar, CAS 20765-98-4) was used as metallic precursor, and the support was TiO2 (Degussa P-25). Five grams of support was suspended in 250 ml of distilled water. Appropriate amounts of Rh precursor were added to obtain catalysts with varying metal contents (0.5, 0.8, 1, 2, 3, and 5 wt.%). After stirring for 4 h and evaporating the solvent under reduced pressure in a rotavapor at 40 °C, the samples were dried at 110 °C overnight and then calcined in a static-air oven at 450 °C for 4 h (heating ramp 10 °C/ min). Catalysts were afterward ground and sieved to obtain the appropriate grain size to be used in catalytic tests. H2 chemisorption at 35 °C was used to measure the amount of exposed Rh atoms. Experiments were performed with an ASAP 2010C apparatus from Micromeritics. Ca. 0.15 g of catalyst was loaded into a Pyrex tube and subsequently heated in He (Praxair 4.8, 20 ml/min) at 120 °C for 1 h. After evacuation, the sample was reduced at 350 °C during 1 h in pure H2 (Praxair 4.8, 30 ml/ min) followed by purging with He at the same temperature for 1 h and cooling in He to adsorption temperature. Two isotherms were measured in the range 0.07–90 kPa, the first corresponding to total (irreversible + reversible) adsorption and the second to reversible H2 adsorption. The subtraction of the two isotherms gave the total amount of irreversibly adsorbed (chemisorbed) hydrogen. The amount of surface Rh atoms (lmol/g) was calculated from the amount of chemisorbed hydrogen assuming that the chemisorption stoichiometry is H:Rh = 1 [26]. Dispersion is defined as surface Rh atoms divided by total Rh atoms in the catalyst. The mean particle size of Rh was calculated, supposing they are hemispherical in shape, by the following equation:

dp ¼

6M DqrN0

ð1Þ

where M is the molecular weight of Rh (102.91 g/mol), D is the Rh fractional dispersion obtained as explained above, q is the Rh metal density (12.4 g/cm3), r is the area occupied by a surface Rh atom (7.58 A2/atom), and N0 is the Avogadro constant. X-ray photoelectron spectroscopy (XPS) analyses were performed with a SSI-X-probe (SSX-100/206) photoelectron spectrometer equipped with a monochromatic microfocused Al Ka X-ray source (1486.6 eV) from Surface Science Instruments. The sample powders were pressed into small stainless steel troughs mounted on a multi-specimen holder. The samples were outgassed overnight under vacuum (105 Pa) and then introduced into the analysis chamber where the pressure was around 107 Pa. An electron flood gun set at 8 eV and a Ni grid placed at 3 mm above the sample were used to standardize charging effects. Pass energy of the analyzer was 150 eV, and the spot size was approximately 1.4 mm2. The atomic concentration ratios were calculated by normalizing surface area ratios with sensitivity factors based on Scofield cross-sections. In addition, all binding energies were calculated taking as reference the CA(C, H) component of the C 1s adventitious carbon peak fixed at 284.8 eV. Peak decomposition was performed using the CasaXPS program (Casa Software Ltd., UK) with a Gaussian/Lorentzian (85/15) product function and a

Shirley non-linear sigmoid-type baseline. The following peaks were used for the quantitative analysis: O 1s, C 1s, Ti 2p, and Rh 3d. 2.2. Operando-DRIFTS In situ diffuse reflectance infrared Fourier transform spectroscopy (in situ DRIFTS) spectra were collected on a Bruker Equinox 55 infrared spectrometer equipped with an air-cooled MIR source with KBr optics and a MCT detector. Spectra were obtained by collecting 200 scans with a resolution of 4 cm1 and are presented in absorbance mode without any manipulation. A background was recorded before starting the experiment by placing an Al mirror in the sample holder. In that way, the signals due to impurities in the cell windows or gases inside the spectrometer were subtracted. Samples were placed without packing or dilution inside a cell with controlled temperature and environment reflectance (Spectra-Tech 0030-103) equipped with ZnSe windows. Different mixtures of gases could be sent to the cell (He, H2, CO2) whose flow rates were controlled by high precision gas rotameters. The gases at the outlet of the cell were analyzed by a quadrupole mass spectrometer (Balzers QMS 200) by following the evolution of the m/z = 2 (H2), 15 (CH4), 18 (H2O), 28 (CO), and 44 (CO2). The catalysts used in these experiments were previously reduced under H2 flow (30 ml/min) in a fixed-bed reactor at 350 °C for 1 h and then grounded finely, before infrared experiments. Once the catalyst was placed in the cell, it was flushed in He (20 ml/min) for 10 min and then reduced again in a flow (20 ml/ min) of a reactive mixture consisting of H2 (5 vol.%) with He (95 vol.%), during 1 h at 300 °C. After that, the cell was cooled to 50 °C in the same gas mixture. Reaction was carried out introducing to the cell 20 ml/min of a mixture of CO2 (10 vol.%), H2 (40 vol.%) diluted in He. The sample was kept at 50 °C for 20 min, which was sufficient to ensure constant spectra. (a) Steady-state experiments: The system was then stepwise heated (5 °C/min) to 100 °C and 150 °C maintaining the sample 20 min at each temperature. (b) Transient experiments: The catalyst was heated under reactive mixture at 5 °C/min to the desired temperature (100, 130, or 150 °C), and after 20 min at that temperature, the flow was changed to H2 (8 ml/min) diluted in He (10 ml/ min) or pure He (18 ml/min). Spectra were recorded every 5 min. The response time of the mass spectrometer following a change in gas concentration was less than 60 s. 2.3. Catalytic activity measurements Catalytic tests were carried out using a quartz reactor (Ushaped) with 0.4 cm internal diameter. A section in the center of the tube is expanded with a diameter of 1 cm, in which the catalyst (200 mg, 200–315 lm particle size) was placed and supported by a quartz frit. A thermocouple was in contact with the central part of the catalyst bed and was used to measure and control the temperature. Heat and mass-transfer effects were ruled out using the criteria recommended by Vannice [30]. Thus, one can safely assume that the reaction took place in fully kinetic regime. The reaction was carried out at atmospheric pressure, by reducing the catalyst in a 30 ml/min flow of pure H2 during 1 h (ramp 10 °C/min). Afterward, the reactor was cooled to 50 °C, and the reaction mixture (20 ml/min) was admitted (CO2 (10 vol.%), H2 (40 vol.%) diluted in He). Measurements were performed at various temperatures between 50 and 200 °C (with 1 h at each temperature to ensure steady state). Exit gases were analyzed with a Varian CP-3800 gas chromatograph (CH4 and CO2 were detected using FID and TCD detectors, respectively). All transfer lines were maintained at 120 °C to avoid water condensation.

143

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

Reaction rates were determined always when CO2 conversion was less than 10% to ensure differential reactor conditions. The specific reaction rate is defined as the number of moles of CH4 produced by mole of Rh by second. The intrinsic reaction rate (turnover frequency, TOF) is defined as the number of molecules of CH4 produced by surface atom of Rh by second. The number of Rh surface atoms was obtained from dispersion data. The dependence of reaction rates on H2 and CO2 partial pressures was determined in a series of tests where the partial pressure of one reactant was kept constant, whereas the partial pressure of the other was varied, at constant temperature (150 °C). The concentration of He was varied accordingly in order to keep the total flow at 20 ml/min. In such tests, the mass of catalyst was varied so as to have similar CO2 conversions. TiO2 (200–315 lm, Degussa P25) was used to dilute the catalyst in order to keep constant the bed volume. The inertness of the TiO2 diluent was verified by using the pure support in the catalytic test. No activity was detected at temperatures below 300 °C. At temperatures higher than 300 °C, only traces of CO were observed. 3. Results 3.1. Catalyst characterization The catalysts present specific surface areas between 41 and 47 m2/g, and the solids are macroporous with a pore volume ranging between 0.26 and 0.30 cm3/g. Pure TiO2 calcined at the same conditions than the catalysts presented a specific surface area of 50 m2/g and 0.22 cm3/g of pore volume. H2 chemisorption data are presented in Table 1. Rh dispersions vary between 0.06 and 0.61, for catalysts with 5 wt.% and 0.5 wt.% Rh loading, respectively. Consequently, the mean size of Rh particles was calculated and found to vary between 2 and 19 nm, showing that a variation of one order of magnitude is obtained for the size of Rh clusters. The amount of Rh surface atoms varies between 18.5 and 32.4 lmol/g, that is, extreme values are separated by a factor less than two.

Table 1 Fractional dispersion and estimated size of Rh particles for the synthesized Rh/TiO2 catalysts determined from H2 chemisorption. Catalyst Rh loading (wt.%)

H/Rh

d Rh (nm)

Exposed Rh (lmol/g)

0.5 0.8 1 2 3 5

0.61 0.26 0.24 0.17 0.06 0.06

2 4 5 7 17 19

29.8 20.3 22.8 32.4 18.5 28.0

XPS results are shown in Table 2. It is worth noting that XPS atomic concentrations and binding energies of C 1s (284.8 eV), Ti 2p (458 eV), and O 1s (531.2–531.3 eV) did not show significant changes in the course of the different treatments applied, and therefore, they are not shown. The decomposition of Rh 3d peak was performed according to a procedure described elsewhere [5]. The measured spectra were fitted by two doublets, one fixed at 307–307.3 eV assigned to Rh in a metallic form (Rh0 species) [31–33] and another that can be related to different oxidized species [32,33]. Oxidation states are presented for catalysts calcined, reduced, and recovered after reaction. For calcined catalysts, the majority (ca. 90%) of Rh surface species are in oxidized Rh3+ state. After H2 treatment and/or reaction, the catalysts are in a more reduced state. Comparing catalysts after reaction, it is observed that the fraction of Rh0 species is about 55% for catalysts with Rh content of 1 wt.% or lower and that this fraction increases up to 74% for Rh(5 wt.%)/TiO2 catalyst. The XPS Rh/Ti atomic ratios increase with the Rh content. There is a decrease in Rh/Ti XPS atomic ratios after reduction treatment, although after reaction, the ratios do not further decrease appreciably. When comparing XPS Rh/Ti atomic ratios with the bulk Rh/Ti atomic ratios, it is observed that XPS ratios are lower than bulk ones. The difference is more pronounced in the case of catalysts with higher Rh content. We did not detect a significant amount of Cl on the samples by XPS after reduction or catalytic tests. The

Table 2 XPS results for Rh/TiO2 catalysts after different treatments (C: calcined, R: reduced, S: Spent). Samples Rh/TiO2

Rh/Ti XPS

Rh/Ti bulk

Rh 3d5/2 oxidized

Rh 3d5/2 reduced

Position (eV)

FWHM

%

Position (eV)

FWHM

%

0.5 wt.% C R S

0.018 0.017 0.016

0.0039

309.0 308.8 308.9

2.40 3.29 3.45

86 60 45

307.3 307.3 307.1

1.44 1.71 1.85

14 40 55

0.8 wt.% C R S

0.031 0.028 0.029

0.0063

309.1 308.5 308.5

2.22 3.18 3.01

83 52 46

307.3 307.0 307.0

1.52 1.68 1.69

17 48 54

1 wt.% C R S

0.046 0.032 0.038

0.0078

309.1 308.7 308.5

2.32 3.19 3.11

85 65 45

307.3 307.1 307.0

1.50 1.70 1.74

15 35 55

2 wt.% C R S

0.079 0.066 0.068

0.0158

309.2 308.7 308.6

2.22 3.22 3.17

83 45 35

307.3 307.1 307.0

1.81 1.70 1.76

17 55 65

3 wt.% C R S

0.129 0.103 0.085

0.0240

309.2 308.7 309.0

2.25 3.21 3.11

86 30 29

307.3 307.0 307.0

1.71 1.73 1.87

14 70 71

5 wt.% C R S

0.183 0.129 0.107

0.0409

309.2 308.5 309.0

2.60 3.09 2.70

91 38 26

307.3 307.1 307.0

1.80 1.66 1.83

9 62 74

144

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

samples analyzed after calcination step contained some traces (not quantifiable) of Cl species.

3.2. Activity results The activity of the different catalysts is shown in Table 3. At the conditions used in this study (low temperatures and atmospheric pressure), the selectivity toward methane was always 100%. The activity expressed as moles of methane produced per atoms of Rh in the catalysts shows an increase with Rh loading, with a maximum for Rh loadings between 2 and 3 wt.%. When the rates are normalized by the surface Rh atoms (turnover frequency), the variations among the catalysts are much more significant. At 150 °C, catalysts with higher Rh loading present turnover frequencies up to 30 times higher than catalysts with the lower Rh content. Similar results are observed at 120 °C. As the Rh mean particle size scales somewhat proportionally with Rh content, the same trends can be observed when activity is plotted as a function of particle size. In Fig. 1a, it is observed that the specific rate increases with increasing particle size for particle sizes up to 7 nm. After that value, the lack of catalysts with particle sizes between 7 and 17 nm does not permit to assess the maximum in activity which could be located in that interval. In the case of turnover frequency (Fig. 1b), it is observed that it increases by a factor of 10 for particle sizes from 2 to 7 nm. Further increase in particle size (up to 19 nm) produces a threefold augmentation in TOF.

Fig. 2 shows the turnover frequency for methane formation as a function of the inverse of temperature. It is observed a good linear correlation with a slope that permits to obtain the apparent activation energies. The apparent activation energy was obtained for the range 85–165 °C, and results are shown in Table 3. The higher activation energies are obtained for catalysts with low Rh content (28.7 kcal/mol for Rh(0.5 wt.%)/TiO2), whereas at higher Rh contents, the values tend to decrease and stabilize around 17 kcal/ mol (3–5 wt.% Rh). The effect of particle size can be observed in the figure. Catalysts with smaller Rh clusters present the higher slopes, thus the higher apparent activation energies. Fig. 3 shows the dependence of intrinsic reaction rate on H2 and CO2 partial pressures for catalysts with different Rh loadings (i.e., different particle sizes). Fig. 3 shows the effect of CO2 partial pressure on turnover frequency. It is observed that in the case of Rh(3 wt.%)/TiO2 catalyst, the reaction order with respect to CO2 is nearly zero, whereas for Rh(1 wt.%)/TiO2 and Rh(0.5 wt.%)/TiO2, the orders are 0.15 and 0.36, respectively. Correlating with Rh particle sizes, one can observe that the smaller the Rh clusters, the more negative the CO2 order. When H2 partial pressure is varied (Fig. 3b), it is observed that for the same set of catalysts, an increase in H2 concentration give rise to higher intrinsic activities. The orders with respect to H2 are thus positive. Nevertheless, the order varies among the catalysts. The catalyst with smaller Rh particles (Rh(0.5 wt.%)/TiO2) presents the higher reaction order (0.83). An increase in Rh particle size produces a decrease in H2 reaction order, to 0.58 in the case of Rh(3 wt.%)/TiO2 which has a mean Rh cluster size of 17 nm.

Table 3 Catalytic activity results. Specific reaction rates and turnover frequencies for Rh/TiO2 catalysts in CO2 hydrogenation at 120 and 150 °C. Apparent activation energies were obtained in the range 100–165 °C. Catalyst

Rh(0.5 wt.%)/TiO2 Rh(0.8 wt.%)/TiO2 Rh(1 wt.%)/TiO2 Rh(2 wt.%)/TiO2 Rh(3 wt.%)/TiO2 Rh(5 wt.%)/TiO2

120 °C

150 °C

Activation energy (kcal/mol)

Rate  102 (molCH4 molRh1 s1)

TOF  102 (s1)

Rate  102 (molCH4 molRh1 s1)

TOF  102 (s1)

n.m. 0.013 0.022 0.034 0.033 0.023

n.m. 0.049 0.090 0.199 0.524 0.401

0.044 0.090 0.120 0.149 0.143 0.091

0.072 0.348 0.500 0.878 2.266 1.576

(a)

(b)

1

28.7 22.2 19.4 17.8 17.0 16.3

10

0.1

TOF (*100) s -1

Specific rate (*100) s -1

1

0.1

0.01 0.01

100°C 120°C 135°C 150°C 165°C 0.001

0

5

10

15

Particle size (nm)

20

0.001

100°C 120°C 135°C 150°C 165°C 0

5

10

15

20

Particle size (nm) 1

Fig. 1. Catalytic activity results. Rate of methane formation as a function of Rh particle size and temperature. (a) Specific rate (molCH4 molRh s1 ) and (b) turnover frequency (molCH4 molRhsurface s1). Lines are drawn to guide the eye.

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

10 0.5 wt.% (2 nm) 0.8 wt.% (4 nm) 1 wt.% (5 nm) 2 wt.% (7 nm) 3 wt.% (17 nm) 5 wt.% (19 nm)

TOF (102 s-1)

1

0.1

0.01 2.2

2.4

2.6

2.8

1000/T (K-1) Fig. 2. Catalytic activity results. Arrhenius plot of CO2 methanation over Rh/TiO2 catalysts with different Rh loadings.

(a) 10.00 TOF CH4 (s-1) x100

Rh(3%)/TiO 2 (17 nm) TOF

PCO2 0.01

1.00 Rh(1%)/TiO 2 (5 nm) TOF

0.10

PCO2 -0.15

Rh(0.5%)/TiO 2 (2 nm) TOF

PCO2 -0.36

0.01 1

10

100

P CO2 (kPa)

TOF CH4 (s-1) x100

(b) 10.00 Rh(3 wt.%)/TiO2 (17 nm)

1.00

TOF

PH2 0.58

Rh(1 wt.%)/TiO2 (5 nm)

0.10

TOF

PH20.75

Rh(0.5 wt.%)/TiO2 (2 nm) TOF

0.01 1

PH20.83 10

100

P H2 (kPa) Fig. 3. Catalytic activity results. (a) TOF dependence on CO2 partial pressure at 150 °C at constant H2 partial pressure (40.5 kPa). (b) TOF dependence on H2 partial pressure at 150 °C at constant CO2 partial pressure (10.1 kPa). In both cases, the concentration of He was varied to maintain a total pressure of 101.3 kPa and a total flow rate of 20 ml/min.

3.3. Operando-DRIFTS 3.3.1. Effect of temperature In Fig. 4 are presented the spectra recorded at steady state in CO2 methanation for Rh(0.5 wt.%)/TiO2, Rh(2 wt.%)/TiO2, and Rh(3 wt.%)/TiO2 catalysts. The spectra were taken at 50, 100, and

145

150 °C. All the significant changes observed in the spectra are contained in the region between 1200 and 2200 cm1. The region attributed to hydroxyl bands (3000–3700 cm1) did not show significant variations. CHx vibrations which appear between 2800 and 3000 cm1 were not observed in this study. The interaction of the reactive mixture with the surface of the catalyst generates different surface species. In the case of Rh(0.5 wt.%)/TiO2 (Fig. 4a), the interaction at 50 °C leads to the formation of a peak at 2037 cm1 characteristic of CO(ads) species linearly adsorbed on Rh. At the same temperature, other bands are observed, namely a band at 1410 cm1 attributed to carbonate species [14,34] and other at 1540 cm1, which can be associated with carbonate or formate species [7,34]. The band observed at 1620 cm1 is attributed to water adsorbed on the support [35]. In the case of catalysts with higher Rh content, Fig. 4b shows the spectrum at 50 °C for Rh(2 wt.%)/TiO2. The same features observed for Rh(0.5 wt.%)/ TiO2 are found here, except for the apparition of the CO(ads) peak at somewhat higher wavenumbers (2042 cm1), and the apparition of a small peak centered at 1920 cm1 accompanied by a very broad and less-intense peak appearing at 1800 cm1. The first peak has been attributed to CO species in the form of Rh2(CO)3 [18], whereas the latter is commonly associated with bridge-bonded Rh2(CO) [34,36,37]. In the case of a catalyst with larger Rh particles (Rh(3 wt.%)/TiO2), Fig. 4c shows that the peak attributed to linear CO(ads) appears at higher wavenumbers (2048 cm1), and the peak arising from bridged-bonded CO is more intense. When the temperature is raised, the intensities and positions of the bands change. In the case of Rh(0.5 wt.%)/TiO2 (Fig. 4a), at 100 °C the band related to linear CO(ads) shifts from 2037 to 2050 cm1, and at the same time, its intensity increases appreciably. Further increase in temperature to 150 °C does not lead to significant changes in this and the other bands. However, a new band is formed at 1360 cm1, which has been attributed to formate ions on the surface of TiO2 [14,34]. In the case of Rh(2 wt.%)/TiO2 and Rh(3 wt.%)/TiO2 catalysts (Fig. 4b and c, respectively), a raise in temperature to 100 °C increases the intensity of CO(ads) linear band with a concomitant shift to higher wavenumbers. The broad peak associated with bridged-bonded CO (1800 cm1) also increases its intensity overlapping the peak attributed to Rh2(CO)3 appearing at 1920 cm1. 3.3.2. Transient experiments The reactivity of surface species was studied by performing transient experiments for different catalysts at various temperatures. In Fig. 5 are presented the results obtained at 130 °C when the flow of CO2 is stopped, and the sample is exposed to H2 + He flow. In the case of Rh(0.5 wt.%)/TiO2 (Fig. 5a), when the flow of CO2 is stopped, there is a decrease in the intensity of linear CO(ads) with a concomitant shift to lower wavenumbers, from 2056 cm1 to 2034 cm1 after 40 min in H2/He stream. The peak at 1620 cm1 does not change appreciably during the transient, whereas the peaks associated with formate species (1360 cm1) show a very rapid decrease in the first 5 min of the transient and are not detected afterward. For the catalysts with higher Rh content, the peaks attributed to bridged-bonded CO(ads) species are present at steady state in CO2 methanation conditions. When the flow is switched to H2 + He, the intensity of Rh2(CO) band decreases; which allows the apparition of the band due to Rh2(CO)3 species. The position of linear CO band shifts continuously from 2052 to 2035 cm1 in the case of Rh(2 wt.%)/TiO2 and from 2050 to 2031 cm1 for Rh(3 wt.%)/TiO2. The band due to Rh2(CO)3 species also shows a shift to lower wavenumbers reaching 1885 cm1 after 40 min. Fig. 6 shows the transient results for Rh(1 wt.%)/TiO2 catalyst at different temperatures. The peaks arising from CO(ads) species decrease in intensity with time. The rate of decrease depends on tem-

146

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

(b)

2200

1700

1200

2200

Absorbance (a.u.)

(c)

Absorbance (a.u.)

Absorbance (a.u.)

(a)

1700

2200

1200

1700

1200

Wavenumber (cm-1)

Wavenumber (cm-1)

Wavenumber (cm-1)

Fig. 4. DRIFT spectra as function of temperature at steady-state conditions. Rh(0.5 wt.%)/TiO2 (a), Rh(2 wt.%)/TiO2 (b), Rh(3 wt.%)/TiO2 (c). From bottom to top: 50, 100, and 150 °C.

(a)

(b)

Absorbance (a.u.)

Absorbance (a.u.)

Absorbance (a.u.) 2200

(c)

1700

1200

2200

Wavenumber (cm-1)

1700

1200

2200

Wavenumber (cm-1)

1700

1200

Wavenumber (cm-1)

Fig. 5. DRIFT spectra taken after changing from CO2 (2 ml/min) + H2 (8 ml/min) + He (10 ml/min) to H2 (8 ml/min) + He (10 ml/min) flow. From bottom to top: 0, 5, 10, 15, 20, 25, 30, 35 and 40 min after gas switching. Rh(0.5 wt.%)/TiO2 (a), Rh(2 wt.%)/TiO2 (b), Rh(3 wt.%)/TiO2 (c).

(a)

Absorbance (a.u.)

Absorbance (a.u.)

Absorbance (a.u.) 2200

(c)

(b)

1700

Wavenumber (cm-1)

1200

2200

1700

Wavenumber (cm-1)

1200

2200

1700

1200

Wavenumber (cm-1)

Fig. 6. DRIFT spectra taken after changing from CO2 (2 ml/min) + H2 (8 ml/min) + He (10 ml/min) to H2 (8 ml/min) + He (10 ml/min) flow for Rh(1 wt.%)/TiO2 catalyst. From bottom to top: 0, 5, 10, 15, 20, 25, 30, 35, and 40 min after gas switching. 100 °C (a), 130 °C (b) and 150 °C (c).

147

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

perature. Namely, when the experiment is performed at 100 °C (Fig. 6a), the intensity of both linear Rh-CO and bridged Rh2(CO) shows a very slow diminution with time. At higher temperatures, the peaks decrease more rapidly, and at a higher rate at the beginning of the transient. After 20 min, CO(ads) species are clearly less reactive toward H2. The peak due to linear Rh-CO shifts toward lower wavenumbers as a result of the decrease in its intensity. The peak arising from Rh2(CO)3 species appears after 15 min for the experiment at 130 °C, showing a shift to lower wavenumbers. At 150 °C, the peak disappears quickly, as the other CO(ads) species. The peak attributed to formate species (1360 cm1) also shows a decrease in its intensity when exposed to H2 stream. Fig. 7 shows the intensity of the peak associated with linear CO(ads) as a function of time when the gas flow is changed from H2 + CO2 to H2. The intensity (in absorbance units) has been corrected for the baseline and normalized by the intensity of the peak before gas switching. Absorbance can be successfully used as a way to estimate adsorbate coverage provided that absorbance is less than 60%, which is fulfilled in our case [38]. The rate of decrease in CO(ads) concentration depends on the temperature and Rh content of the catalysts. Rh(0.5 wt.%)/TiO2 catalyst presents the lower rate of CO(ads) disappearance in the transient experiment, and the rate increases when Rh content is increased. The trends observed for CO(ads) reactivity are approximately the same obtained for the steady-state activity in CO2 methanation (Fig. 1b). The residual intensity of the CO peak can be clearly observed in Fig. 7c. Approximately 20% of adsorbed CO species of Rh(0.5 wt.%)/TiO2 do not react with H2 even after 1 h on stream. In the case of catalysts with larger particles, a lower residual intensity of the CO(ads) peak is observed, reaching only 5% in the case of Rh(3 wt.%)/TiO2. The initial rate of CO(ads) dissociation can be calculated from the slopes of the curves in Fig. 7. The variation of the rate of CO(ads) dissociation with temperature allows the determination of the activation energy of this process. The results are plotted in Fig. 8 as a function of Rh particle size and compared to the apparent activation energy of the CO2 hydrogenation reaction obtained in fixed-bed reactor. It is observed that the activation energy of CO(ads) dissociation does not vary significantly with particle size, being about 15 kcal/mol for all the catalysts studied. However, the apparent activation energy of the overall CO2 methanation reaction presents a different behavior; namely, it is constant for catalysts with particle sizes larger than 7 nm (ca. 17 kcal/mol)

(b)

1

CH4 formation C=O dissociation

Activation energy (kcal/mol)

25

20

15

10

5

0 0

5

(c)

1

1

0.8

0.8

0.7

0.7

0.7

0.3

Rh(0.5%) Rh(1%) Rh(2%) Rh(3%)

0.2 0.1 0

0

10

20

30

40

time (min)

50

60

Relative intensity

0.8

0.4

0.6 0.5 0.4

0.6 0.5 0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

0

10

20

and increases sharply for smaller particles, reaching 28.7 kcal/ mol for Rh(0.5 wt.%)/TiO2 catalyst (2 nm particle size). In Fig. 9 are shown the MS responses of methane concentration during the transient experiment in which a H2 + CO2 flow is changed to H2. The results at 150 °C show the concentration at steady state before the transient. It is observed that the order of activity presented in this figure is qualitatively similar to that shown in Fig. 1. Moreover, the apparent activation energies calculated from the experiments in DRIFTS cell (not shown) match the values presented in Table 3 for fixed-bed reactor. These results confirm the absence of diffusion limitations in the DRIFTS cell. When the conditions are changed from steady-state H2 + CO2 flow to H2, the methane signal at the exit of the cell presents a dynamic behavior that depends on the catalyst Rh cluster size. In the case of Rh(3 wt.%)/TiO2 catalyst, one can observe a slight augmentation

0.9

0.5

15

Fig. 8. Apparent activation energy for CH4 formation at steady-state conditions compared and the activation energy obtained for CO(ads) dissociation by operandoDRIFTS, as a function of Rh particle size.

0.9

0.6

10

Rh particle size (nm)

0.9

Relative intensity

Relative intensity

(a)

30

20

30

40

time (min)

50

60

0

0

10

20

30

40

50

60

time (min)

Fig. 7. Evolution of the relative intensity of Rh-CO linear peak as a function of time after the gas flow is changed from CO2 (2 ml/min) + H2 (8 ml/min) + He (10 ml/min) to H2 (8 ml/min) + He (10 ml/min). The data was obtained at 100 °C (a), 130 °C (b), and 150 °C (c). Intensity was normalized by the intensity of the peak right before the gas switching.

148

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

(a)

H2 + CO2

(b)

H2

MS signal m/z = 15 (a.u.)

MS signal m/z = 15 (a.u.)

Rh(3%)

H2 + CO2

Rh(2%)

Rh(1%)

H2

Rh(2%) Rh(3%) Rh(0.5%)

Rh(0.5%)

0

Rh(1%)

10

20

30

40

0

10

20

30

40

time (min)

time (min)

Fig. 9. MS signal of methane (m/z = 15) at the exit of the DRIFTS cell when the CO2 (2 ml/min) + H2 (8 ml/min) + He (10 ml/min) flow is changed to H2 (8 ml/min) + He (10 ml/ min). The gas switching was done at t = 10 min. Data at 150 °C (a) and 130 °C (b).

H 2 + CO 2

He or H 2

H2/He

MS signal m/z = 15 (a.u.)

in CH4 concentration followed by a continuous decrease in the following 15 min. For catalysts with smaller Rh particles, a clear peak in CH4 is observed at the beginning of the transient. The magnitude of the peak is higher when Rh particles are smaller. Similar results are obtained when the experiment is performed at 130 °C (Fig. 9b). Fig. 10 shows the response of DRIFTS spectra to a change in gas flow from the steady-state conditions (H2 + CO2) to pure He at 150 °C for Rh(3 wt.%)/TiO2 catalyst. Comparing the spectrum taken right before gas switching (t = 0 min) to that after 5 min in He, an increase in intensity in all the peaks is observed. It can be clearly observed that the bands present at the beginning of the transient do not change appreciably in intensity during the He transient. The MS trace of methane during the precedent experiment is presented in Fig. 11 and compared with data obtained in H2(40%)/He

He

0

10

20

30

40

time (min) Fig. 11. MS signal of methane (m/z = 15) at the exit of the DRIFTS cell. Comparison between changes from CO2 (2 ml/min) + H2 (8 ml/min) + He (10 ml/min) to He (18 ml/min) or H2 (8 ml/min) + He (10 ml/min) at 150 °C. The gas switching was done at t = 10 min. Data for Rh(3 wt.%)/TiO2 catalyst.

transient. It is observed that after changing to He flow, the concentration of methane decreases rapidly, whereas in the case of the transient in H2/He, the decrease is less pronounced indicating that methane continues to be formed after the gas switching. It has to be noted that the baselines for the two experiments are different due to the different carrier gases used (H2/He vs. He). 4. Discussion 4.1. Activity of Rh/TiO2 catalysts

Fig. 10. DRIFT spectra taken after changing from CO2 (2 ml/min) + H2 (8 ml/ min) + He (10 ml/min) to He (18 ml/min). From bottom to top: 0, 5, 10, 15, 20, 25, 30, 35, and 40 min after gas switching. Data for Rh(3 wt.%)/TiO2 catalyst.

The intrinsic activity of Rh/TiO2 catalysts increases with particle size when particles are smaller than 7 nm (Fig. 1a). Up to our knowledge, the dependence of intrinsic activity on metal particle size in the case of Rh catalysts in CO2 methanation has not been studied previously. Nevertheless, some studies have been pre-

149

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

sented regarding other supported transition metal catalysts at higher temperatures and with other gas compositions (e.g., with CO). For Ru/TiO2 and Ru/Al2O3 catalysts, the turnover frequencies for CO2 consumption were reported to increase for particles from 1 to 4 nm, whereas further increase in size did not produce changes in activity [27]. We recently showed that in the case of Rh/c-Al2O3 catalysts, the intrinsic activity increased slightly with Rh particle size [5], which differs with the results presented here for Rh/ TiO2. Furthermore, it was reported that the intrinsic activity of Ru/TiO2 catalysts was tenfold higher for catalysts with smaller Ru nanoparticles (2.5 nm) compared to catalysts with particles of 9.5 nm [6]. From these results, it is evident that the support, the type of metal, and the operating conditions affect the dependence of intrinsic CO2 methanation rates with respect to the size of metal particles. The fact that intrinsic activity increases with particle size implies that a maximum in specific activity has to be observed. According to our results (Fig. 1a), this maximum could be located between 7 and 17 nm, but certainly, more experimental data are required to fully assess this issue. Interestingly, a similar behavior as the reported here has been found in the case of CO hydrogenation to methane or heavier hydrocarbons, a reaction which has been much more studied than CO2 hydrogenation. In the case of Ru/Al2O3 [39] and Co/Carbon [40] catalysts, the intrinsic activity in CO hydrogenation has been reported to increase steeply when particles are small (<6 nm), whereas that further increase in particle size do not significantly affect the activity. The fact that a very similar behavior is observed for CO2 methanation strongly suggests that the two reactions proceed by similar mechanisms. The magnitude of turnover frequencies obtained in this study agrees well with the results published by Solymosi et al. [9] for a Rh/TiO2 catalyst (properties presented in Table 4). Extrapolating to 150 °C using their kinetic equation, a turnover frequency of 0.013 s1 is obtained, which compares well with data presented in Table 3 for the high-loaded catalysts (Rh(3 wt.%)/TiO2 and Rh(5 wt.%)/TiO2) whose intrinsic rates are 0.023 and 0.016 s1, respectively. The catalyst used by Solymosi et al. had a dispersion of 0.223 (Table 4). The catalysts presented in this manuscript that are comparable in dispersion (Table 1) are Rh(1 wt.%)/TiO2 (dispersion 0.24) and Rh(2 wt.%)/TiO2 (dispersion 0.17). The turnover frequencies of these catalysts (Table 3) are 0.005 and 0.009 s1, respectively, which is lower by less than a factor of 3 comparing with Solymosi’s catalyst.

The apparent activation energies obtained in this study lie between 16.3 and 28.7 kcal/mol (Table 3). Catalysts with smaller particles present the higher activation barriers. The data obtained are in good agreement with those reported in the literature (Table 4). For Rh/TiO2 catalysts, an activation barrier of 19.4 kcal/mol was reported previously [9]. Other Rh-based catalysts also show similar apparent activation energies (Rh/SiO2 (16.6 kcal/mol) [18], Rh/cAl2O3 (17 kcal/mol) [41]). The variation of apparent activation energy with particle size was also observed in the case of Rh/c-Al2O3 catalysts [5]. If we compare with the results obtained here, it is observed that apparent activation energy depends on Rh particle size (or dispersion), but it is independent of the support properties. This suggests that the reaction steps defining the activation energy are occurring on the metal surface rather than on the support. The dependence of intrinsic reaction rates on H2 and CO2 partial pressures is shown in Fig. 3. Comparing with data for similar catalysts, our data agree in the sense that generally reaction orders with respect to H2 are positive and higher than 0.5. Nevertheless, regarding at Fig. 3b, one can observe that catalysts with smaller particles present higher reaction orders in H2, even 0.83 for Rh(0.5 wt.%)/TiO2 catalyst. If we look at the reaction orders with respect to CO2 (Fig. 3a), the orders are lower than for H2, in agreement with results from the literature (Table 3). For catalysts with larger particles, the order is zero (Rh(3 wt.%)/TiO2 catalyst, 17 nm particle diameter) and when particle size is decreased, the order is negative (0.36 for Rh(0.5 wt.%)/TiO2, 2 nm). This clearly indicates that CO2 behaves as an inhibitor of the reaction and that this effect is enhanced when Rh cluster sizes are smaller. Probably, the CO(ads) species arising from the dissociation of CO2 can be ascribed as the responsible for this inhibition. The fact that the order with respect to H2 is more positive for catalysts with smaller particles could indicate that these particles have a lower coverage of hydrogen compared to larger ones. 4.2. The mechanism of CO2 dissociation In Fig. 4 are shown the surface species formed at steady state when a mixture of CO2 + H2 interacts with the surface of different Rh/TiO2 catalysts with varying metal particle size. It is observed that at temperatures as lower as 50 °C, CO(ads) species are formed, even when at this temperatures, no methane is observed at the effluent of the DRIFTS cell. This suggests that CO2 is easily dissociated on these catalysts. The capacity of CO2 to undergo dissociation

Table 4 Comparison of apparent activation energies and power lawa orders for different Rh-based formulations in CO2 methanation. Rh loading (wt.%)

Rh dispersion

Rh mean particle size (nm)b

Ea (kcal/mol)

Rh/TiO2

1

0.223

5

19.4

Rh foil + TiO2 Rh foil

– –

0 0

– –

17.0 17.0

Rh/ZrO2

2.3

0.51

2

14.9

[41]

Rh/Al2O3

1 1.5 2 3 5 5 2.3 0.5

0.3 0.24 0.18 0.071 0.072 0.302 0.6 0.79

4 5 6 15 15 4 2 5

22.7 19.1 15.8 14.6 14.5 16.2 17.0 20.3

[5]

5 2.3 3.4

0.228 0.27 0.45

4 2 5

17.3 15.9 16.6

Rh/SiO2

a b

x

y

TOFCH4 ¼ AeEa =RTPH2 PCO2 . Calculated from dispersion data according to Eq. 1.

x

y

Reference [9]

0.5 0.5

0.3 0.2

[13] [13]

0.61

0.26

[9] [41] [42]

0.64

0.27

0.53

0.46

[9] [41] [18]

150

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

over Rh catalysts, particularly at low (room) temperatures, has been recently demonstrated over Rh/c-Al2O3 catalysts [8]. Moreover, other studies have shown that CO2 dissociates on Rh foils or Rh single crystals [43,44], as well as over supported Rh catalysts [18,45]. An enhancement in CO2 dissociation has been reported when H2 is present [46,47]. The way in which CO2 transforms to CO(ads) has been a subject of discussion in the literature. Some authors propose a mechanism through formate species [14,15] occurring at the interface between the metal and the support, in which reverse water–gas shift is involved:

CO2 þ HðmetalÞ ! COOHðinterfaceÞ ! COðmetalÞ þ OHðsupportÞ On the other hand, the direct dissociation (CO2 ? CO(ads) + O(ads)) has been reported by several groups over a variety of noble metal-based catalysts [18,19,48]. In our case, the direct dissociation of CO2 seems to be evidenced. We observed the formation of CO(ads) at fairly low temperatures where reverse water–gas shift does not occur. Moreover, CO was never detected in gas phase even at higher temperatures. Additionally, the formation of CO(ads) was not accompanied by the concomitant apparition of formate species. Formates were detected only at 150 °C. Based on these results, we can propose that formates are mainly spectators and have a minor influence on the reaction path. This agrees with the findings of other authors [19] over noble metal-based catalysts. The formation of formate species, thus, would proceed by the reaction of CO(ads) with OH groups of the support [7,49], following the equation:

COðmetalÞ þ OHðsupportÞ $ COOHðinterfaceÞ This proposal fits well with the fact that over supports containing low concentration of OH groups, such as SiO2, formates are not detected [34]. As it is shown in Fig. 4, when temperature is raised, the carbonyl peaks (2042, 1920, and 1800 cm1) increase in intensity and shift their position to higher wavenumbers. The increment in intensity is explained by the enhanced dissociation of CO2 when temperature is raised, leading to a higher concentration of CO(ads) species. The shift toward higher wavenumbers is explained by an increment in CO(ads) coverage which produces an increase in CO(ads) dipole–dipole coupling [50]. 4.3. The reactivity of the CO adsorbed species and the role of the support It can be observed from Fig. 4 that, at 150 °C, the position of the peak associated with CO(ads) in linear form is near 2050 cm1 for all the catalysts presented. However, when the temperature is lower (50 °C), there is a clear difference in the position of this peak. When the catalyst is more dispersed (lower particle size), the peak tends to shift to lower wavenumbers denoting a lower coverage of CO(ads) species and hence a less favorable CO2 dissociation on these catalysts. The intensity of CO(ads) peak at steady state at 150 °C is higher for catalysts with higher Rh loading (i.e., larger Rh particles) (Fig. 4). Even if it is difficult to compare the absolute amounts of CO adsorbed on the catalysts, it can be accepted that catalysts with larger Rh particles present a higher amount of CO adsorbed on them. On the other hand, the amount of surface Rh, obtained from H2 chemisorption (Table 1), is practically the same for Rh(0.5 wt.%)/TiO2 and Rh(2 wt.%)/TiO2 catalysts, 29.8 lmol/g and 32.4 lmol/g, respectively. In the case of Rh(3 wt.%)/TiO2, the amount of surface Rh is even lower (18.5 lmol/g). This observation strongly suggests that catalysts with larger particles, such as Rh(3 wt.%)/TiO2, present a larger proportion of surface sites capa-

ble of adsorbing CO and hence potentially participating in the reaction. The fact that the position of the linear CO(ads) peak does not change appreciably among the catalysts would indicate that the dipole–dipole interaction of CO adsorbed species should be similar and could be associated with CO adsorption in the form of paths on the metallic surface. It is important to underline that an increase in CO(ads) band intensity when metal particle size increased was also observed for Ru/zeolite catalysts, which was related to the higher rate of CO2 dissociation on larger nanoparticles [51]. The peak attributed to CO(ads) species adsorbed in linear form appears at ca. 2050 cm1 for all the samples analyzed. The absence of CO species adsorbed on Rh+ sites, such as gem-dicarbonyl species has been widely reported in the case of hydrogenation of CO2 over noble metals supported on different metal oxide supports [14,16,18,46,47,52]. The fact that the band of CO(ads) in linear form appears at lower wavenumbers (ca. 2050 cm1) compared to 2060–2080 cm1 normally reported for Rh-CO has been attributed to the formation of Rh carbonyl species [46,47,52,53]. Based on these data, the linear CO peak (ca. 2050 cm1) observed in our DRIFT experiments could be associated with such Rh carbonyl hydride species. Rh carbonyl hydride is composed of a CO molecule adsorbed to a Rh atom and one or two H atoms bound to the same Rh atom [46,47]. The apparition of the carbonyl band at 1920 cm1 was attributed to Rh2(CO)3. However, CO(ads) peaks in this range have been also attributed to CO adsorbed onto the metal and interacting with Lewis acids sites of the support such as Ti3+ [12,13], that is, adsorbed on the metal support interface when the support is partially reducible. In our case, this band is observed only in catalysts with large particles, and it is absent on Rh(0.5 wt.%)/TiO2 catalyst (Fig. 4). This would indicate that the band at 1920 cm1 forms on special Rh sites rather than on interfacial sites, because interfacial sites are more numerous on catalysts with smaller Rh particles. Moreover, the band is observed to shift toward lower wavenumbers when its intensity decreases (Fig. 5), which could indicate a decrease in dipole–dipole coupling when coverage decreases. Also, the band has been observed on Rh catalysts that are not partially reducible and thus do not present those interfacial sites, such as Rh/Al2O3 [5] and Rh/SiO2 [18]. When the catalysts working at steady state are exposed to H2 + He flow, the concentration of carbon containing species observed by DRIFTS starts to decrease due the reaction with hydrogen. In Fig. 5, at 130 °C, it is observed that the reactivity of CO adsorbed in bridged form, in the case of catalysts with large Rh particles, such as Rh(2 wt.%)/TiO2 and Rh(3 wt.%)/TiO2, seems to be higher than linear CO(ads) species, because they disappear early in the transient. When the transient experiment is performed at different temperatures, as shown in Fig. 6 for Rh(1 wt.%)/TiO2, it is observed that the peaks due to CO(ads) species decrease in intensity and that rate increases with temperature. As bridged-bonded CO species disappear (1800 cm1), the peak arising from Rh2(CO)3 species is resolved. Afterward, this peak also disappears, however before than the linear CO(ads) peak. According to the temperature in which the transient was performed, the sequence is more or less rapid, but there are no differences in the type of CO(ads) species depending on temperature. The shifts in the position of Rh-CO and Rh2(CO)3 could be due to a decrease in CO dipole–dipole coupling arising from the decrease in CO(ads) coverage during the transient experiment. When the transient experiments were performed using a flow of pure He instead of the reactive mixture (Fig. 10), one can observe that CO(ads) species, in linear and bridged form, do not show significant changes with time. This indicates that they do not desorb nor react with hydrogen to form methane. That is verified in

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

Fig. 11, where no CH4 is seen to evolve after the flow is changed to He, which contrasts with the fact that methane is produced when H2 is present in gas phase. These results indicate that TiO2 does not have the capacity to store some reactive species which would later react with CO(ads) adsorbed to form methane. It is important to underline that in the case of Rh/c-Al2O3 catalysts [5], CO(ads) species are reactive even when no gaseous H2 is present. This suggest that in that case, there are H(ads) species adsorbed on the support capable of reacting with CO(ads) to form CH4. This property was successfully used to increase the activity of a Rh/c-Al2O3 catalyst by mixing it with Ni/C catalyst which is capable of activating hydrogen easily [54]. These observations may help to shed some light in the effect of the support on the reaction rate for this reaction. According to these results, even if TiO2 seems not to be capable of storing H(ads) species that could react with adsorbed CO(ads) in the absence of gas-phase H2, its catalytic activity is up to one order of magnitude higher than that of Rh/c-Al2O3 [5,9]. That would imply that this property do not affect the catalytic activity and hence suggest that the hydrogenation of CO species is not a rate-determining step. 4.4. Mechanism and kinetic aspects From Fig. 7, it can be observed that the rate of CO(ads) dissociation depends on the type of catalyst and temperature. The same order of reactivity shown in Table 2 is observed here. That is, catalysts with smaller particles such as Rh(0.5 wt.%)/TiO2 and Rh(1 wt.%)/TiO2 show a lower rate of CO(ads) dissociation compared to Rh(2 wt.%)/TiO2 or Rh(3 wt.%)/TiO2. That strongly supports the fact that CO(ads) species are intermediaries of the reaction, as has been proposed by several authors [14–16,18,19]. The activation energy for the dissociation of CO(ads) does not vary between the different Rh/TiO2 catalysts and thus does not depend on Rh particle size. The values found in this study (ca. 15 kcal/ mol) compare well with the results of Marwood et al. [55] who reported a barrier of 18 kcal/mol in the case of a Ru/TiO2 catalyst. A value of 17.3 kcal/mol was presented by Mori et al. [56] who studied a Ru/Al2O3 catalyst. On the other hand, a higher activation barrier for CO(ads) dissociation was observed on Ni (23 kcal/mol) [57]. The apparent activation energies observed for the whole CO2 hydrogenation process are higher than the activation energy for CO(ads) removal. Fig. 8 shows apparent activation energies as higher as 28.7 kcal/mol for catalysts with small Rh nanoparticles. When particle size increases, the activation energy decreases, up to ca. 17 kcal/mol, nearly the same as that measured for CO(ads) removal. If we assume that the dissociation of CO(ads) is the rate-determining step, as has been generally accepted

(a)

151

[7,13,18,20], the increase in activation energy when metal particle size becomes smaller could arise from an increase in the energy barrier for dissociating CO(ads) or from a change in adsorptive properties (coverage, adsorption energies) depending on the size of metal clusters. The first option could be discarded if the results presented in Fig. 8 are taking into account, that is, the dissociation barrier of CO(ads) would not change with particle size. Although there is general agreement in the fact that CO(ads) dissociation is the rate-determining step of the reaction, the way in which CO(ads) dissociates has been proposed to proceed by two main paths: (i) direct CO(ads) dissociation and (ii) H-assisted CO(ads) dissociation. In the first mechanism, CO(ads) is dissociated to C(ads) and O(ads) species which are afterward hydrogenated to form methane. This has been proposed to occur over group VIII metal-based catalysts [9,20]. Moreover, it has been recently shown by theoretical calculations that CO(ads) can dissociate more easily than HCO(ads) or COH(ads) species over a corrugated Ru surface [21]. On the other hand, several groups have proposed that the dissociation of CAO bond is assisted by hydrogen. Andersson et al. [22] performed density functional theory calculations and found that the dissociation barriers are lower when COH is the intermediate instead of CO. On the basis of kinetic experiments coupled with theoretical calculations, it was proposed that H-assisted CO dissociation is the predominant path in the case of CO hydrogenation over Co and Fe catalysts [23]. Mori et al. [24,25] proposed the existence of a HnCO intermediate based on an inverse isotopic effect for H2–D2 found for CO dissociation. Nevertheless, the nature of this hydrogenated CO intermediate species is not fully understood. Some authors have proposed the existence of formyl species. These species have been observed by IR over Ru/Al2O3 [19] and Rh/ TiO2 [45], although in the latter case their relevance in the mechanism has not been fully demonstrated. In our experiments, we could not detect any formyl species, even when reaction rates 1 reported here (between 2  107 and 3  105 molCH4 g1 Rh s ) were sometimes lower than those reported in [19] 1 (3:6  106 molCH4 g1 Ru s ), and hence, the coverage of formyl species would be higher in those cases. Another proposal regarding the dissociation of CO consists in the formation of Rh carbonyl hydride, which induces a weakening in the CAO bond facilitating its cleavage [46,58]. According to the infrared results, it could be suggested that the dissociation of CO can be assisted by H through the formation of Rh carbonyl hydride species. In order to discriminate between the two proposed CO dissociation steps (hydrogen assisted or not), we present in Fig. 12 two reaction paths in both of which it is considered that CO2 dissociates to CO(ads) and O(ads) species but that differ in the rate-determining step, namely whether CO(ads) dissociates alone or assisted by

(b)

Fig. 12. Mechanisms proposed for CO2 hydrogenation to methane. (a) H-assisted CO dissociation. (b) Unassisted CO dissociation. Ki and ki denote the adsorption constant or the rate constant of the step i.  Stands for an active site. X represent an adsorbed species.

152

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

H species. In Fig. 12a, H2CO species as have been said in the precedent paragraph could stand for Rh carbonyl hydride species. Langmuir–Hinshelwood rate equation models taking into account the above rate-determining steps were obtained. It was assumed that H(ads) and CO(ads) species were the most abundant reaction intermediates. The coverage of H(ads) and CO(ads) species has been shown to be of the same magnitude in the case of CO2 hydrogenation over Rh catalysts under similar reaction conditions [13,18]. Furthermore, for the derivation of the reaction rate, it was considered that all the steps prior to the rate-determining one are in equilibrium. Regarding at the dependence on H2 and CO2 partial pressures of the models shown in Fig. 12, it is observed that in the case of unassisted CO dissociation, the model predicts that the order with respect to H2 and CO2 can vary between 1 and 0.5. According to the results presented Fig. 3, the reaction order with respect to H2 experimentally obtained is well above 0.5, reaching 0.83 for the catalyst with larger Rh clusters. Moreover, if we compare with the literature data (Table 3), the reaction order with respect to H2 is generally higher than 0.5. These results strongly suggest that direct CO dissociation is not likely to be the dominant path of reaction. In the case of H-assisted CO dissociation, the predicted reaction orders can vary between 1 and 1 in the case of H2 and between 1 and 0.5 for CO2. Our experimental results agree well with these predictions. In the case of H2, the experimentally obtained reaction orders lie in the upper part of the interval (between 0.58 and 0.83). Regarding at equation in Fig. 12a, it is observed that high reaction orders in H2 could arise from a low coverage of H spe0:5 cies (negligible K0:5 2 PH2 term). Consequently, the coverage of CO(ads) species is higher, and thus, the reaction order in CO2 is lower. This agrees well with the experimental results; namely, catalysts with smaller Rh clusters present higher H2 orders (0.83) and lower CO2 orders (0.36) (Fig. 3), whereas in the case of catalysts with larger Rh clusters (Rh(3 wt.%)/TiO2), the pressure dependence in H2 is lower (0.58) and in CO2 higher (0.01). The fact that in the case of catalysts with smaller Rh particles, adsorbed CO inhibits the reaction rate, can also be confirmed regarding at Fig. 9, which shows the evolution of methane concentration when the reactive feed (H2 + CO2) was changed to H2, at the exit of the DRIFTS cell. A peak in CH4 production is observed after the CO2 gas feed is switched off. That takes place with the concomitant disappearance of CO(ads) species, which have been shown to be the direct precursor of CH4. The peak in methane production is larger for catalysts with smaller Rh particles. This indicates that when the surface is depleted of CO(ads) species, the reactivity of small particles increases. This is proposed to arise from the liberation of active sites which permits the activation of hydrogen and the subsequent hydrogenation and dissociation of adsorbed CO. The increase in apparent activation energy with decreasing particle size can also be explained by the stronger adsorption of CO(ads) species on the metal surface. In fact, the equation in Fig. 12a predicts an increase in apparent activation energy when adsorption energy of CO(ads) increases (higher value of K1 constant). The stronger adsorption of CO on small metallic particles has been shown in the case of CO hydrogenation over Co and Ru-based catalysts. It was found by SSITKA experiments [39] that the residence time of adsorbed CO increased as the size of Rh particles decreased for particles smaller than 10 nm. Results were explained by a stronger Ru-CO bond in the case of smaller Ru nanoparticles. According to Fig. 7, the asymptotic value of the relative intensity of CO(ads) band, in transient DRIFTS experiments, is higher for catalysts with smaller nanoparticles such as Rh(0.5 wt.%)/TiO2. This implies that on these catalysts, the proportion of CO(ads) species irreversibly adsorbed is higher, compared to catalysts with larger Rh cluster size. This agrees with previous works [59] on the hydrogenation of CO over Co/C catalysts under conditions in which the

main product was methane. It was reported that smaller Co nanoparticles contained a higher amount of irreversibly adsorbed CO. A higher proportion of sites capable of adsorbing CO intermediates and a lower amount of irreversible adsorbed CO were found for catalysts with large cluster sizes compared to those containing smaller nanoparticles. It has been proposed in the literature that coordinatively unsaturated sites such as steps and kinks are the preferred sites for CO(ads) dissociation [22,60] and play a key role in other reactions such as N2 dissociation in ammonia synthesis or NO dissociation [61]. Since these sites are more numerous over smaller particles, the question of why these particles are less active in CO2 hydrogenation becomes relevant. According to our results, catalysts with smaller Rh nanoparticles have less active sites and also bind CO species strongly. These two facts could lead to a decrease in available H(ads) species and hence to a lower overall activity. In the case of catalysts with larger particles, there are more active sites, and CO is adsorbed weakly leading to more free space to accommodate H(ads) species, which can thus react with CO(ads) and help to dissociate it. The nature of active sites at the moment cannot be elucidated according to our data. We can suggest that these sites correspond only to a fraction of the metallic surface, and they are present in larger numbers over larger Rh particles. According to XPS results, the fraction of oxidized sites after reaction is higher for catalysts with smaller particles (Table 2). That could explain the lower number of active sites over these catalysts. Another possibility is that catalysts with larger Rh clusters present morphologies (e.g., lateral planes) that permit to activate hydrogen more efficiently than small particles, thus increasing the rate of reaction. Nevertheless, the nature of the surface sites must be similar, because the activation barrier of adsorbed CO over the different Rh/ TiO2 catalysts is similar (Fig. 8). This is in line with recent results in the case of CO hydrogenation on cobalt catalysts [62]. It was found that the differences in activity between catalysts with varying particle size were due to differences in the number of active sites which was lower for catalysts with smaller particles. 5. Conclusions The methanation of CO2 was studied at temperatures between 85 and 165 °C and at atmospheric pressure. The rate of methane formation is highly sensitive to the Rh particle size of Rh/TiO2 catalysts. The rate normalized by surface Rh atoms increases as metal particle size increases up to ca. 7 nm. Beyond this size, the activity does not change appreciably. Kinetic parameters such as the apparent activation energy and reaction orders with respect to H2 and CO2 depend strongly on the Rh cluster size. The combination of kinetic data with operando-DRIFTS experiments allows to propose that smaller Rh particles have less active sites and bind CO(ads) species strongly than catalysts with larger ones, leading to lower reaction orders in CO2 and higher apparent activation energies. The inverse is observed for catalysts with larger Rh particles. Moreover, activation barriers for CO(ads) dissociation were found to be similar for all the catalysts studied. On the basis of a kinetic comparison between two proposed reaction paths (hydrogen-assisted and unassisted CO(ads) dissociation), it is concluded that the dissociation of CO(ads) could proceed via a H-assisted path, probably by the formation of Rh carbonyl hydride species. These results point out the similarities between the reaction path and the metal particle size dependence between CO2 and CO hydrogenations. Acknowledgments The authors gratefully acknowledge the ‘‘Direction Générale des Technologies, de la Recherche et de l’Energie (DGTRE)’’ of the

A. Karelovic, P. Ruiz / Journal of Catalysis 301 (2013) 141–153

‘‘Région Wallonne’’ (Belgium) and the ‘‘Fonds National de la Recherche Scientifique (FNRS)’’ of Belgium, for their financial support. The involvement of IMCN-MOST in the «INANOMAT» IUAP network sustained by the «Service public fédéral de programmation politique scientifique» (Belgium) is acknowledged. A. Karelovic acknowledges Becas Chile program of CONICYT (Chile) for the PhD grant. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]

[23] [24] [25] [26]

G. Centi, S. Perathoner, Catal. Today 148 (2009) 191. G. Centi, S. Perathoner, Greenhouse Gases: Sci. Technol. 1 (2011) 21. M. Aresta, A. Dibenedetto, Dalton Trans. (2007) 2975. A. Beuls, C. Swalus, M. Jacquemin, G. Heyen, A. Karelovic, P. Ruiz, Appl. Catal. B 113–114 (2012) 2. A. Karelovic, P. Ruiz, Appl. Catal. B 113–114 (2012) 237. T. Abe, M. Tanizawa, K. Watanabe, A. Taguchi, Energy Environ. Sci. 2 (2009) 315. M.R. Prairie, A. Renken, J.G. Highfield, K. Ravindranathan Thampi, M. Grätzel, J. Catal. 129 (1991) 130. M. Jacquemin, A. Beuls, P. Ruiz, Catal. Today 157 (2010) 462. F. Solymosi, A. Erdohelyi, T. Bansagi, J. Catal. 68 (1981) 371. Z.L. Zhang, A. Kladi, X.E. Verykios, J. Catal. 148 (1994) 737. F. Solymosi, I. Tombácz, J. Koszta, J. Catal. 95 (1985) 578. A.T. Bell, J. Mol. Catal. A 100 (1995) 1. K.J. Williams, A.B. Boffa, M. Salmeron, A.T. Bell, G.A. Somorjai, Catal. Lett. 9 (1991) 415. M. Marwood, R. Doepper, A. Renken, Appl. Catal. A 151 (1997) 223. D.I. Kondarides, P. Panagiotopoulou, X.E. Verykios, J. Phys. Chem. C 115 (2011) 1220. P. Panagiotopoulou, D.I. Kondarides, X.E. Verykios, Catal. Today 181 (2012) 138. C. Schild, A. Wokaun, R.A. Koeppel, A. Baiker, J. Phys. Chem. 95 (1991) 6341. I.A. Fisher, A.T. Bell, J. Catal. 162 (1996) 54. S. Eckle, H.-G. Anfang, R.J.r. Behm, J. Phys. Chem. C 115 (2010) 1361. G.D. Weatherbee, C.H. Bartholomew, J. Catal. 77 (1982) 460. S. Shetty, A.P.J. Jansen, R.A. van Santen, J. Am. Chem. Soc. 131 (2009) 12874. I. Chorkendorff, M.P. Andersson, E. Abild-Pedersen, I.N. Remediakis, T. Bligaard, G. Jones, J. Engbwk, O. Lytken, S. Horch, J.H. Nielsen, J. Sehested, J.R. RostrupNielsen, J.K. Norskov, J. Catal. 255 (2008) 6. M. Ojeda, R. Nabar, A.U. Nilekar, A. Ishikawa, M. Mavrikakis, E. Iglesia, J. Catal. 272 (2010) 287. Y. Mori, T. Mori, T. Hattori, Y. Murakami, J. Phys. Chem. 94 (1990) 4575. Y. Mori, T. Mori, A. Miyamoto, N. Takahashi, T. Hattori, Y. Murakami, J. Phys. Chem. 93 (1989) 2039. P. Panagiotopoulou, D.I. Kondarides, X.E. Verykios, Ind. Eng. Chem. Res. 50 (2010) 523.

153

[27] P. Panagiotopoulou, D.I. Kondarides, X.E. Verykios, Appl. Catal. B 88 (2009) 470. [28] M. Kusmierz, Catal. Today 137 (2008) 429. [29] W. Rarog-Pilecka, Z. Kowalczyk, K. Stolecki, E. Miskiewicz, E. Wilczkowska, Z. Karpiniski, Appl. Catal. A 342 (2008) 35. [30] M.A. Vannice, Kinetics of Catalytic Reactions, Springer, New York, 2005. [31] M. Kawai, M. Uda, M. Ichikawa, J. Phys. Chem. 89 (1985) 1654. [32] Z. WengSieh, R. Gronsky, A.T. Bell, J. Catal. 170 (1997) 62. [33] J.S. Brinen, A. Melera, J. Phys. Chem. 76 (1972) 2525. [34] F. Solymosi, A. Erdohelyi, T. Bansagi, J. Chem. Soc. Faraday Trans. 1 (77) (1981) 2645. [35] H. Wijnja, C.P. Schulthess, Spectrochim. Acta A 55 (1999) 861. [36] S.S.C. Chuang, R.W. Stevens, R. Khatri, Top. Catal. 32 (2005) 225. [37] W.M.H. Sachtler, M. Ichikawa, J. Phys. Chem. 90 (1986) 4752. [38] J. Sirita, S. Phanichphant, F.C. Meunier, Anal. Chem. 79 (2007) 3912. [39] J.M.G. Carballo, J. Yang, A. Holmen, S. García-Rodríguez, S. Rojas, M. Ojeda, J.L.G. Fierro, J. Catal. 284 (2011) 102. [40] G.L. Bezemer, J.H. Bitter, H.P.C.E. Kuipers, H. Oosterbeek, J.E. Holewijn, X. Xu, F. Kapteijn, A.J. van Dillen, K.P. de Jong, J Am. Chem. Soc. 128 (2006) 3956. [41] T. Iizuka, Y. Tanaka, K. Tanabe, J. Catal. 76 (1982) 1. [42] P. Panagiotopoulou, D.I. Kondarides, X.E. Verykios, Appl. Catal. A 344 (2008) 45. [43] M.F.H. van Tol, A. Gielbert, B.E. Nieuwenhuys, Appl. Surf. Sci. 67 (1993) 166. [44] B.A. Sexton, G.A. Somorjai, J. Catal. 46 (1977) 167. [45] E. Novak, K. Fodor, T. Szailer, A. Oszko, A. Erdohelyi, Top. Catal. 20 (2002) 107. [46] M.A. Henderson, S.D. Worley, Surf. Sci. 149 (1985) L1–L6. [47] F. Solymosi, M. Pasztor, J. Catal. 104 (1987) 312. [48] L.F. Liotta, G.A. Martin, G. Deganello, J. Catal. 164 (1996) 322. [49] P. Panagiotopoulou, D.I. Kondarides, J. Catal. 260 (2008) 141. [50] J.W. Niemantsverdriet, Spectroscopy in Catalysis, Wiley-VCH, Weinheim, 2007. [51] S. Eckle, M. Augustin, H.-G. Anfang, R.J. Behm, Catal. Today 181 (2012) 40. } helyi, Top. Catal. 55 (2012) [52] M. Tóth, J. Kiss, A. Oszkó, G. Pótári, B. László, A. Erdo 747. [53] S.D. Worley, G.A. Mattson, R. Caudill, J. Phys. Chem. 87 (1983) 1671. [54] C. Swalus, M. Jacquemin, C. Poleunis, P. Bertrand, P. Ruiz, Appl. Catal. B 125 (2012) 41. [55] M. Marwood, R. Doepper, M. Prairie, A. Renken, Chem. Eng. Sci. 49 (1994) 4801. [56] T. Mori, A. Miyamoto, H. Niizuma, N. Takahashi, T. Hattori, Y. Murakami, J. Phys. Chem. 90 (1986) 109. [57] J. Sehested, S. Dahl, J. Jacobsen, J.R. Rostrup-Nielsen, J. Phys. Chem. B 109 (2004) 2432. [58] F. Solymosi, A. Erdohelyi, M. Kocsis, J. Catal. 65 (1980) 428. [59] J.P. den Breejen, P.B. Radstake, G.L. Bezemer, J.H. Bitter, V. Frøseth, A. Holmen, K.P.d. Jong, J. Am. Chem. Soc. 131 (2009) 7197. [60] R.A. van Santen, M.M. Ghouri, S. Shetty, E.M.H. Hensen, Catal. Sci. Technol. 1 (2011) 891. [61] R.A. Van Santen, Acc. Chem. Res. 42 (2009) 57. [62] J. Yang, E.Z. Tveten, D. Chen, A. Holmen, Langmuir 26 (2010) 16558.