Cobalt particle size effects on catalytic performance for ethanol steam reforming – Smaller is better

Journal of Catalysis 318 (2014) 67–74

Contents lists available at ScienceDirect

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

Cobalt particle size effects on catalytic performance for ethanol steam reforming – Smaller is better Andre L.M. da Silva a, Johan P. den Breejen b, Lisiane V. Mattos a,1, Johannes H. Bitter b,2, Krijn P. de Jong b, Fábio B. Noronha a,⇑ a b

Instituto Nacional de Tecnologia – INT, Av. Venezuela 82, 20081-312 Rio de Janeiro, Brazil Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University, Sorbonnelaan 16, 3508 CA Utrecht, The Netherlands

a r t i c l e

i n f o

Article history: Received 11 March 2014 Revised 22 July 2014 Accepted 26 July 2014

Keywords: Steam reforming of ethanol Cobalt on carbon nanofibers (Co/CNF) Cobalt particle size effects Catalyst stability Coke formation

a b s t r a c t The effect of the cobalt particle size in the ethanol steam reforming reaction at 773 K for hydrogen production was investigated using cobalt on carbon nanofiber catalysts. It was found that the turnover frequency increases with decreasing Co particle size, which was attributed to the increasing fraction of edge and corner surface sites with decreasing size. Regarding catalyst stability, a decrease in deactivation rate was observed with decreasing cobalt particle size. This was caused by a significantly lower amount of carbon deposition on the smallest Co particles than on larger ones, as concluded from transmission electron microscopy measurements. The reduced amount of carbon deposition is ascribed to a lower fraction of terrace atoms, proposed to be responsible for initiation of carbon deposition on catalysts with large (>10 nm) Co particles. Therefore, it was concluded for this non-noble metal that the smallest particles perform best in catalysis of ethanol steam reforming. Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction The understanding of particle size and shape effects on catalytic activity and selectivity requires a fundamental insight in the relation between surface structure and catalyst performance [1–5]. This relation does also depend on the type of bonds of reactants involved. For example, for p-bond activation an optimum particle size is generally found, below which the surface-specific activity drops. Examples are provided in the ammonia synthesis [6] and Fischer–Tropsch catalysis [7–9] in which the N2 and CO bonds have to be activated, respectively. In case of the ammonia synthesis, an optimum metal particle size could be attributed to a maximum in the number of specific surface sites facilitating p-bond activation [6]. For r-bond activation, e.g., in hydrogenolysis and CH4 activation reactions, the particle size–performance relations are less unambiguous. For noble metals both a decrease [10,11] and an increase [12–18] in surface-specific activity with decreasing particle size has been observed. For non-noble metals, however,

⇑ Corresponding author. Fax: +55 (21) 2123 1166. E-mail addresses: [email protected] (K.P. de Jong), [email protected] (F.B. Noronha). 1 Present address: Universidade Federal Fluminense (UFF), Departamento de Engenharia Química e de Petróleo, Rua Passo da Pátria 156, 24210-240 Niterói, Brazil. 2 Present address: Wageningen University, Wageningen, The Netherlands. http://dx.doi.org/10.1016/j.jcat.2014.07.020 0021-9517/Ó 2014 Elsevier Inc. All rights reserved.

information on the effect of particle size on catalyst performance in r-bond activation reactions is scarce [19,20]. This is most likely related to the hampered reducibility of small non-noble metal particles. Moreover, the small particles may interact strongly with an oxidic support, which may result in the formation of stable mixed oxides during calcination, reduction or reaction. With the advent of inert nanostructured carbon support materials such as carbon nanofibers [21] and carbon nanotubes [22]; however, a possibility to study the intrinsic effects of the particle size of non-noble metals on activity and selectivity is provided. In the current study carbon nanofibers (CNF) were chosen as support material to study the effect of cobalt particle size on activity and selectivity in r-bond activation in the steam reforming (SR) of ethanol reaction. In this case, the use of an inert support such as CNF or other carbonaceous materials [23] is fundamental since the majority of oxides are active for SR of ethanol. This approach using an inert CNF material is similar to the one applied in the study of the Co particle size effects in the p-bond activation in Fischer–Tropsch catalysis [7,24]. The SR reaction comprises a complex reaction network, including several reaction intermediates [25–29]. Recently, Ferrin et al. [30] showed using density functional theory (DFT) calculations for ethanol hydrogenolysis that the C–C bond cleavage, proposed to be the rate-determining step (RDS), of the CH–CO intermediate was facilitated on step sites. As the number of step sites increases

68

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

with decreasing particle size, a particle size effect in SR catalysis is therefore expected. However, experimental studies about the intrinsic effects of the particle size on the SR reaction are scarce and contradictory. A study by Ribeiro et al. [31] showed an indicative trend of an increasing surface-specific activity (TOF) with decreasing Co particle size (<8 nm) for the SR of ethanol over Co/ SiO2 under mild conditions where ethanol was mainly dehydrogenated to acetaldehyde. Song and Ozkan [32] studied the performance of Co/ZrO2, Co/CeO2, and Co/CeO2–ZrO2 catalysts on the SR. The TOF calculated at 723 K for the three catalysts increased in the order: Co/ZrO2 (0.054 s1) < Co/CeO2–ZrO2 (0.214 s1) < Co/ CeO2 (0.476 s1). Cobalt crystallite size determined by XRD for the reduced and passivated catalysts revealed a correlation between intrinsic activity and crystallite particle size, which followed the order: Co/ZrO2 (15 nm) < Co/CeO2–ZrO2 (20 nm) < Co/ CeO2 (25 nm). However, the supports used are significantly active for SR. Therefore, the careful elimination of the effect of the support is fundamental to determine the intrinsic cobalt particle size effects on the SR reaction. Even more important than activity is catalyst stability which might be dependent on particle size or number of surface ensemble sites too. The effects of particle size on for example carbon deposition in CH4 reforming or decomposition over Ni-based catalysts have been demonstrated, showing that the lowest amount of coke is formed on the smallest particles [33–35]. The number of studies toward an understanding of the effect of particle size on catalyst stability for SR of ethanol, though, is rather limited. Here, we report on the effects of cobalt particle size and related surface structure on the activity, selectivity, and stability in the SR of ethanol reaction using Co/CNF catalysts. The aim of this research was to obtain a quantitative understanding of the relation between the number of specific sites on the surfaces of Co particles, using a geometrical model, and their activity and stability.

pressure, the total hydrogen uptake was determined. Based on that, the cobalt particle sizes were calculated assuming a hemispherical particle geometry and an H/Cos adsorption stoichiometry of 1 [37]. XPS measurements were carried out on passivated Co/CNF samples in a Vacuum Generators XPS spectrometer using Al Ka radiation. The surface atomic ratios were calculated from photoelectron peak areas after correcting for photoionization cross sections and photoelectron mean free paths. Subsequently, the Co particle size was calculated from the XPS intensity ratios of the Co 2p3/2 and C 1s peaks (ICo/IC) using a model based on the work described by Kuipers et al. [38]. TEM analysis was used to study the cobalt particle sizes of the fresh and spent catalysts. The catalysts were suspended in ethanol, and brought onto a carbon support film on a copper TEM grid. The TEM measurements were conducted with a Tecnai 20 FEG microscope operating at 200 kV. Thermogravimetric analysis of the fresh and used Co3 and Co16 catalysts was carried out in a TA Instruments equipment (SDT Q600). The sample was heated under air from room temperature to 1273 K at a heating rate of 20 K/min. Temperature-programmed surface reaction (TPSR) was performed in a microreactor coupled to a quadrupole mass spectrometer (Omnistar, Balzers). The sample (50 mg) was reduced under flowing H2 (30 cm3/min) up to 623 K (5 K/min) and maintained at that temperature for 1 h. After reduction, the system was purged with helium at the reduction temperature for 30 min and cooled to room temperature. A mixture containing H2O/He (60 cm3/min) was passed through the sample as the temperature was raised at 10 K/min to 773 K, and it was kept for 1 h. The reaction products were monitored with a quadrupole mass spectrometer. 2.3. SR reaction

2. Experimental 2.1. Catalyst preparation The carbon nanofiber support material (SA = 200 m2 g1, PV = 0.65 mL g1) was obtained from synthesis gas using a 5 wt% Ni/SiO2 growth catalyst. The obtained material was purified in subsequent reflux treatments of 1 M KOH and concentrated HNO3. In the latter step, oxygen-containing groups required to achieve high metal dispersions were introduced [36]. The cobalt catalysts were obtained using incipient wetness impregnation of cobalt acetate tetrahydrate or cobalt nitrate hexahydrate solutions in either water or ethanol, aiming for various Co loadings (0.9–22 wt%) [7]. After impregnation, the samples were dried in static air at 393 K. Subsequently, a reduction was conducted at 623 K for 2 h, in a flow of 30 vol% H2 in N2, followed by a passivation at room temperature by exposure of the catalyst to air. The cobalt particle sizes were determined using H2-chemisorption, X-ray photoelectron spectroscopy (XPS), and transmission electron microscopy (TEM) analysis. Co/CNF catalysts with cobalt particle sizes ranging from 2.6 to 16 nm were obtained. 2.2. Catalyst characterization Cobalt particle sizes were determined by H2-chemisorption using a Micromeritics ASAP 2010C. Before each measurement, the samples were dried in vacuum at 393 K for 2 h followed by a reduction (5 K min1) in H2 flow at 623 K for 2 h. Subsequently, the samples were evacuated at that temperature for 30 min. The H2 adsorption isotherms were measured at 423 K. From extrapolation of the linear part of the isotherm to zero

The SR of ethanol reaction was performed in a fixed-bed reactor at atmospheric pressure. Prior to reaction, the Co/CNF catalysts were reduced at 623 K for 1 h and then purged under N2 at the same temperature for 30 min. All reactions were carried out at 773 K and H2O/ethanol molar ratios of 3.0 and 10.0 were used. The reactant mixtures were obtained by flowing two N2 streams (30 mL min1) through two saturators, one containing water and the other one containing ethanol, which were maintained at the temperature required to obtain the desired H2O/ethanol molar ratios. The partial pressure of ethanol was maintained constant for all experiments. The variation of partial pressure of water was compensated by the decrease of partial pressure of N2. Two different series of experiments were carried out to determine the intrinsic activity and to study the catalyst deactivation. In the first series, different amounts of catalyst (2.5–5.0 mg) and N2 flow rates (120–180 mL min1) were used in order to obtain low ethanol conversions (iso-conversion). 1 Then, reaction rate (molEtOH g1 ) and Turnover Frequency Co s 1 1 (TOF; molEtOH molsurf,Co s ) were calculated taking into account the initial ethanol conversion that was taken after 0.2 h TOS. For the calculation of TOF, the cobalt particle size determined by H2-chemisorption or XPS was used. In the second series, 20 mg of catalyst and a total flow rate of 60 mL min1 were used for all catalysts in order to observe the catalyst deactivation within 24 h time-on-stream (TOS). The catalysts were diluted with inert SiC. The reaction products were analyzed by a gas chromatograph (Micro GC Agilent 3000 A) containing three channels for three thermal conductivity detectors (TCD) and three columns: a molecular sieve, a Plot Q, and an OV-1 column. The ethanol conversion and product distribution were determined from:

69

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

ðnethanol Þfed  ðnethanol Þexit  100 ðnethanol Þfed ðni Þproduced Yi ¼  100 ðntotal Þproduced

X ethanol ¼

ð1Þ ð2Þ

where (ni)produced = moles of i produced (i = hydrogen, CO, CO2, methane, acetaldehyde, or ethene) and (ntotal)produced = moles of H2 + moles of CO + moles of CO2 + moles of methane + moles of acetaldehyde + moles of ethene (the moles of water produced are not included). The selectivity to products containing carbon was defined as the concentration of each product at reactor effluent over the concentration of all products containing carbon, taking into account the stoichiometric reaction coefficients between carbon-containing products and ethanol (e.g., for CO: 1/2; CO2: 1/2; CH4: 1/2; C2H4O: 1; C2H4: 1) [39]. The stability of the catalyst was defined as the ratio between the reaction rate at t0 = 0.2 h and tf = 3 or 24 h TOS (Eq. (3)), assuming first-order kinetics.

Stability ¼ Lnð1  X ðtf Þ Þ=Lnð1  X ðt0 Þ Þ

ð3Þ

3. Results and discussion 3.1. Catalyst characterization The Co/CNF catalysts with their properties and preparation routes are summarized in Table 1. This table shows that the cobalt particle size increases with increasing Co loading. Moreover, from the H2-chemisorption and XPS on the one hand, and TEM analyses on the other hand, consistent particle sizes were found. It is important to notice that previous XANES experiments revealed complete reduction of cobalt oxide even for the highly dispersed catalysts [7].

Fig. 1. The effect of cobalt particle size on reaction rate and TOF in the SR of ethanol reaction (H2O/ethanol molar ratio = 3.0, 773 K, 24 h TOS).

3.2. Catalyst activity Prior to testing the Co/CNF catalysts in the SR reaction, the ethanol conversion using CNFs only was studied. The ethanol conversion over CNF was low (5%) and approximately the same as the one observed in the experiment with SiC only (homogeneous reaction). This indicates that CNF does not exhibit significant activity for the SR reaction. Therefore, we may conclude that the SR over Co/CNF catalysts exclusively takes place on the cobalt particles. Taking into account the results obtained in the first series of experiments under the same initial ethanol conversion (iso-conversion, 20%), the respective reaction rate and TOF were calculated. In Fig. 1, the initial reaction rate and the respective TOF are plotted as a function of the cobalt particle size. From Fig. 1A, it can be concluded that the reaction rate increased with decreasing particle size. Moreover, as can be observed in Fig. 1B, the TOF increases as the Co particle size decreased. This indicates that the surface atoms of small particles are more active than those on

larger ones. This result suggests that the SR of ethanol is a structure sensitive reaction, which is considered in more detail in the general discussion.

3.3. Catalyst deactivation Ethanol conversion and product distribution as a function of TOS for all Co/CNF catalysts with different average initial cobalt particle size are plotted in Fig. 2. This second series of experiments enable us to study the effect of Co particle size on catalyst deactivation. From these figures, it was concluded that the initial ethanol conversion varied with cobalt loading as well as cobalt particle sizes. However, a similar product distribution was observed for all Co sizes, with H2 and CO2 as main products. Also small quantities of CO (up to 8%), methane (<5%) and acetaldehyde (3–6%) were

Table 1 Cobalt on carbon nanofiber catalysts with their preparation details, loadings, and particle sizes.

a b

ID

Solvent

Precursor

Co loading (wt%)

H2-chemisorption (nm)

TEMb dsw ± r (nm)

Co16 Co10 Co7 Co5 Co3 Co2

Water Water Water Ethanol Water Water

Cobalt Cobalt Cobalt Cobalt Cobalt Cobalt

22 9.7 6.0 3.5 0.9 1.0

16 9.9 6.9 4.8 2.9a 2.6a

16 9.7 ± 5.0 – – 2.8 ± 0.8 2.4 ± 0.7

nitrate nitrate nitrate nitrate acetate acetate

Co particle size determined by XPS. Surface-weight (dsw) average cobalt particle sizes after reduction and passivation, including the standard deviation (r (nm)).

70

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

B

Xethanol H2

90

CO2 ethene acetaldehyde

CH4

80

CO

100

Product distributions, Xethanol (%)

100

Products distributions, X ethanol (%)

A

70 60 50 40 30 20 10

90 80

5

10

15

20

CO2

CH4

ethene acetaldehyde

CO

70 60 50 40 30 20 10

0 0

Xethanol H2

0

25

0

5

10

90 80

D

Xethanol H2

CO2

CH4

ethene acetaldehyde

CO

70 60 50 40 30 20 10

Xethanol H2

100

Products distributions, X ethanol (%)

100

Product distributions, Xethanol (%)

C

0

90

25

80

CO2 ethene acetaldehyde

CH4 CO

70 60 50 40 30 20 10

5

10

15

20

0

25

5

10

Time (h) 100 90 80

F

Xethanol H2

CO2

CH4

ethene acetaldehyde

CO

70 60 50 40 30 20 10 0

15

20

25

Time (h)

Product distributions, Xethanol (%)

Product distributions, Xethanol (%)

20

0

0

E

15

Time (h)

Time (h)

100 90 80

Xethanol H2

CO2

CH4

ethene acetaldehyde

CO

70 60 50 40 30 20 10 0

0

5

10

15

20

25

Time (h)

0

5

10

15

20

25

Time (h)

Fig. 2. Ethanol conversion (XEtOH) and product distributions versus time-on-stream obtained for the ethanol SR reaction over (A) Co2; (B) Co3; (C) Co5; (D) Co7; (E) Co10; and (F) Co16. (H2O/ethanol molar ratio = 3.0; Treaction = 773 K and residence time = 0.02 g s/mL).

observed. These results indicate that only small variations in product distribution were observed as a function of the Co particle size, which is in line with studies by Haga et al. [40]. There are small differences in product distribution at the beginning of the reaction. In order to analyze them more accurately, the selectivities to products containing carbon are presented in Fig. S1 from supplementary information. After 24 h TOS, the main product formed is CO2, indicating that the SR of ethanol is the main reaction taking place for all catalysts. Increasing Co content significantly increased the selectivity to CO2, whereas the selectivity to acetaldehyde decreased. The selectivity to methane and CO remained practically

unchanged. Co2 and Co3 catalysts exhibited the highest acetaldehyde selectivity, which was due to their lower ethanol conversion. Whereas the Co2 and Co3 catalysts showed a quite stable conversion in the SR reaction, the catalysts containing larger Co particle sizes (>4 nm) showed a significant decrease in conversion with TOS. In addition, the loss of activity occurred mainly at the beginning of reaction. In Fig. 3, the stability of the Co/CNF catalysts is plotted as a function of the Co particle size. The stability was defined as Eq. (3). Co2 and Co3 catalysts exhibited a lower initial ethanol conversion than the catalysts containing a higher Co content. Then,

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

the test with the Co3 catalyst was repeated under a high residence time in order to have similar initial ethanol conversion to compare deactivation of all catalysts. The stability calculated for this run is represented by a triangle. For the Co/CNF catalysts with the smallest Co sizes (<4 nm), a significantly higher stability was found as compared with catalysts with larger particles. It is important to stress that the stability of Co3 catalyst did not change significantly even at high initial ethanol conversion. This indicates a lower deactivation for smaller Co particles, and might be either due to a lower carbon deposition rate or less sintering. A TEM study was conducted to investigate the origin of the catalyst deactivation by studying the extent of particle sintering and the formation of carbon over Co/CNF catalysts during SR. TEM images of fresh and spent Co3 and Co16 catalysts are shown in Fig. 4. The TEM images of the reduced and passivated samples display the presence of cobalt particles decorating carbon fibers (Fig. 4A and C). Representative images of the spent catalysts used in the SR reaction (H2O/ethanol ratio of 3.0 at 773 K, residence time = 0.02 g s/mL, 24 h TOS) are shown in Fig. 4B and D for the Co3 and Co16 catalyst, respectively. These images suggest the presence of a significant amount of amorphous carbon covering the Co16 catalyst surface after the SR reaction. This might well explain the significant loss in activity for this catalyst. It is also observed that carbon filaments contained the Co particle at the end (images not shown). For the Co3 catalyst, however, amorphous carbon deposition was hardly detected. In addition, the images of the spent catalyst containing low cobalt content did not exhibit Co particles located at the ends of carbon filaments (vide supra). The formation of carbon on Co16 catalyst was studied in more detail by carrying out TG analysis under air after reaction at different TOS (3 and 24 h). In order to determine the contribution of carbon stemming from CNF to the weight loss curve, the TG analysis of the fresh sample was also performed. Then, this amount was subtracted from the weight loss obtained for both catalysts and the amount of carbon deposited during the reaction calculated. For the Co16 catalyst, the amount of carbon deposited after 3 and 24 h TOS was 0.29 and 0.34 mgcarbon/mgcatal, respectively, which were approximately the same. This result indicates that the main carbon deposition occurs at the beginning of reaction, which is in agreement with the quick loss of activity observed in the first 3 h TOS. In order to compare the amount of carbon formed on Co3 (test at high initial ethanol conversion) and Co16, the carbon accumulation rates were normalized by the number of moles

1,0

Stability

0,8

71

of ethanol converted during 24 h TOS. The amount of carbon formed on Co3 catalyst (0.71 mgcarbon/mgcatal/moles ethanol converted) was significantly lower than that on Co16 catalyst (9.76 mgcarbon/mgcatal/moles ethanol converted). The cobalt particle sizes of the fresh and spent catalysts were determined from the TEM images. For the Co/CNF catalyst with an initial cobalt size of 2.9 nm an average particle size of 3.4 nm was found for the spent catalysts from the SR reaction. This result reveals that minor sintering was observed during SR for the catalyst containing small Co particle sizes. Also for the Co16 catalyst only moderate sintering was observed as an increase in cobalt size from 16 nm to 19 nm was found. Since these sintering effects are moderate (for both catalysts a 17% decrease in cobalt surface area was observed), the catalyst deactivation of the Co/CNF systems with large cobalt particles has to be ascribed to carbon deposition mainly. The stability of carbon nanofibers at temperatures as high as 773 K and very high water content was also investigated by TPSR of Co3 catalyst. Only traces of CO and CO2 were detected above 650 K when the catalyst was heated under the mixture H2O/He. The amount of carbon oxidized by water during TPSR was less than 0.5% of the amount of sample. Therefore, carbon oxidation/gasification at such high temperatures and in the presence of high water concentrations does not play an important role in the long-term stability of the catalysts. The effect of the H2O/ethanol molar ratio on deactivation was investigated for the Co10 catalyst, which showed a relatively low stability in the SR reaction under H2O/ethanol molar ratio of 3.0. In this case, the SR was also carried out under a H2O/ethanol molar ratio of 10.0 (Fig. 5). From the ethanol conversion and product distributions as a function of TOS, it was concluded that an increase in H2O/ethanol molar ratio of 3.0 (Fig. 2e) to 10.0 (Fig. 5) caused an increase in ethanol conversion. However, catalyst deactivation is also observed for the high H2O/ethanol ratio, though at a lower rate as compared with the SR experiments with a low H2O/ethanol ratio. This result has been previously reported in the literature [41,42]. According to the reaction mechanism proposed for SR of ethanol, CHx species are generated from the decomposition of dehydrogenated and acetate species [43]. This CHx species can further decompose to H and carbon, which in turn can result in catalyst deactivation. Increasing the steam-to-ethanol molar ratio of the feed increases the rate of the carbon gasification reaction, and this, in turn, decreases the catalyst deactivation rate through continuous carbon removal. However, this was still not enough to achieve long-term catalyst stability. Furthermore, the selectivity to CO2 increased whereas the formation of CO decreased as the H2O/ethanol molar ratio increased. This was likely due to an enhanced water gas-shift reaction. 3.4. General discussion

0,6

0,4

0,2

0,0 0

2

4

6

8

10

12

14

16

18

Co particle size (nm) Fig. 3. Stability of Co/CNF catalysts as a function of the cobalt particle size after 24 h TOS SR reaction. (d) Residence time = 0.02 g s/mL; (N) residence time = 0.15 g s/mL (H2O/ethanol = 3.0, Treaction = 773 K).

In order to provide a possible explanation for the increase in TOF and catalyst stability with decreasing Co size, the fraction of cobalt surface atoms with a coordination number (CN) of 4–7 (corner and edge atoms) and 8–11 (terrace atoms) was calculated as a function of the Co particle size, based on a geometrical, cuboctahedral model from van Hardeveld and Hartog [24,44]. After that, the fraction of edge and corner atoms and the fraction of terrace atoms were plotted (Fig. 6) versus TOF and stability, respectively. In Fig. 6A, an increase in activity with increasing fraction of corner and edge atoms is found. This in fact suggests that the edge and corner atoms are active sites for the SR of ethanol reaction. Che and Bennett [1] and van Santen [4] showed that the surface sensitivity observed for specific catalytic reactions is related to the type of chemical bond to be activated. For example, in the Fischer– Tropsch reaction, the p-bond of CO has to be activated, requiring

72

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

Fig. 4. TEM images of the fresh (A and C) and spent (B and D) of Co3 and Co16 catalysts, respectively.

100

Product distributions, Xethanol (%)

90 80 70 60 Xethanol H2

50

CO2 ethene acetaldehyde

CH4

40

CO

30 20 10 0

0

5

10

15

20

25

Time (h) Fig. 5. Ethanol conversion (XEtOH) and product distributions versus time-on-stream obtained for the ethanol SR reaction over Co10 under H2O/ethanol molar ratio of 10.0 (773 K).

an ensemble of surface atoms. However, with decreasing metal particle size (<4 nm) a decrease in the number of ensemble sites is expected, and hence a lower activity is observed e.g., for small

cobalt particles [7]. Moreover, for those small Co particles, a significant number of coordinatively unsaturated site (cus) atoms are present, which in fact hampers the Fischer–Tropsch reaction among others by bonding CO molecules irreversibly [24]. For the reforming reactions of alkanes, the rate of cleavage of r C–C bonds increase with an increasing amount of coordinative unsaturated metal surface atoms [14–16], whose fraction increases with decreasing particle size. Cortright et al. [45] calculated the activation energies for C–C bond dissociation of various C2Hx(ads) species. The stronger binding of various species on the step edge of Pt(2 1 1) compared to (1 1 1) terrace of Pt was used to explain the reported structure sensitivity and relatively higher activity of the stepped surface in ethane hydrogenolysis. The SR of ethanol reaction involves O–H, C–H, C–C and C–O bond cleavage reactions. DFT calculations have been used to study the nature and stability of surface intermediates formed in the ethanol hydrogenolysis reaction over Pt surfaces [45,46] and various transition metals [30]. The results predict that the cleavage of the C–C bond of the CH–CO intermediate is rate determining, which is facilitated by the presence of cus atoms. This is due to the fact that cus atoms generally stabilize intermediates and transition states more strongly as compared with terrace sites [30]. Hence, an increase in TOF is observed with decreasing cobalt particle size in the SR of ethanol reaction. The comparison with the Fischer–Tropsch reaction shows that size–activity relationships strongly depend on type of bond activation and the type of surface sites needed for high activity [1,4]. The non-linear behavior in Fig. 6A might indicate that the presence of a significant amount of coordinatively unsaturated site (cus) atoms on small Co particles influences the electronic

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

73

Fig. 6. (A) Surface-specific activity (TOF) of Co/CNF catalysts plotted as function of the fraction of surface atoms with coordination number 4–7. (B) Stability of Co/CNF catalysts plotted as function of the fraction of surface atoms with coordination number 8–11.

properties of neighboring terrace atoms, thereby increasing the activity of the latter atoms. The stability of ethanol steam reforming catalysts is an important issue. In literature two main causes for SR catalyst deactivation have been shown, viz. carbon deposition and metal particle sintering [43,47–49]. From the current study, we concluded that the main effect of deactivation of Co/CNF catalysts with large (>10 nm) cobalt particles is due to carbon deposition. In order to provide a possible explanation of the higher amount of carbon deposition on large Co particles, the catalyst stability was plotted versus the number of terrace sites in Fig. 6B. Based on the obtained linear relationship, we propose that the terrace atoms are involved in the catalyst deactivation. This might be due to the fact that nucleation of coke and carbon sheets require relative large domains of flat terraces or larger ensemble sizes [34,50–55], as was for example suggested for hydrocarbon reforming reactions [56–59]. Although formation of coke seems to start at terrace sites (Fig. 6B), the coke may ‘overflow’ to corner and edge sites and thereby affect the overall TOF that is dominated by the activity of corner and edge sites. The higher stability of steps and corner atoms could be also associated with their capacity to dissociate water. Recently, Sun et al. [23] studied the SR of acetone and ethanol over unsupported Co and Co nanoparticles supported on activated carbon (AC). Unsupported Co significantly deactivated mainly due to the formation of carbon filaments, whereas carbon deposits were not detected over Co/AC catalyst. Using DFT calculations, they proposed that water dissociation is promoted on Co nanoparticles, suggesting that water is easily converted into atomic O on the Co nanoparticles, releasing H2. Then, the surface O atoms react with CHx species produced from the decomposition of acetone or ethanol, leading to the formation of CO and inhibiting catalyst deactivation. On the other hand, water dissociation does not occur on large Co particles, indicating that terrace sites are not active for dissociation of water. This could also explain the higher stability of our catalysts with smaller Co particle size. Anyway, minimizing the fraction of terrace sites by tuning the metal particle size is a successful strategy to minimize coke lay-down and thereby to maximize catalyst stability.

increasing surface-specific activity with decreasing Co particle size was observed, which was ascribed to the increasing number of unsaturated cobalt surface atoms. Moreover, the smallest Co sizes (<3 nm) showed the highest catalyst stability after 24 h time-onstream, which was due to a lower amount of carbon deposition. This was ascribed to a lower fraction of terrace sites on small Co particles. Acknowledgments The authors acknowledge CTENERG/FINEP-01.04.0525.00 and Shell Global Solutions for financial support. C. van der Spek is thanked for TEM analyses and Dr. Bezemer for some Co/CNF catalysts. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jcat.2014.07.020. References [1] [2] [3] [4] [5] [6] [7]

4. Conclusions

[8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]

Cobalt on carbon nanofiber catalysts with different Co particle sizes (2.6–16 nm) were investigated in the ethanol steam reforming reaction to hydrogen production. For this reaction, an

[21] [22] [23]

M. Che, C.O. Bennett, Adv. Catal. 36 (1989) 55. C.R. Henry, Surf. Sci. Rep. 31 (1998) 231. G. Somorjai, F. Tao, J. Park, Top. Catal. 47 (2008) 1. R. van Santen, Acc. Chem. Res. 42 (2009) 57. S.W. Lee, S. Chen, W. Sheng, N. Yabuuchi, Y.-T. Kim, T. Mitani, E. Vescovo, Y. Shao-Horn, J. Am. Chem. Soc. 131 (2009) 15669. K. Honkala, A. Hellman, I.N. Remediakis, A. Logadottir, A. Carlsson, S. Dahl, C.H. Christensen, J.K. Norskov, Science 307 (2005) 555. 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. A. Barbier, A. Tuel, I. Arcon, A. Kodre, G.A. Martin, J. Catal. 200 (2001) 106. G. Prieto, A. Martínez, P. Concepción, R. Moreno-Tost, J. Catal. 266 (2009) 129. S. Fuentes, F. Figueras, J. Catal. 61 (1980) 443. B. Coq, A. Bittar, F. Figueras, Appl. Catal. 40 (1990) 103. D.J.C. Yates, J.H. Sinfelt, J. Catal. 8 (1967) 348. D. Nazimek, J. Ryczkowski, React. Kin. Catal. Lett. 40 (1989) 145. J. Wei, E. Iglesia, J. Catal. 225 (2004) 116. J. Wei, E. Iglesia, J. Phys. Chem. B 108 (2004) 7253. J. Wei, E. Iglesia, J. Phys. Chem. B 108 (2004) 4094. S.L. Tait, Z. Dohnálek, C.T. Campbell, B.D. Kay, Surf. Sci. 591 (2005) 90. J. Wang, S. Shen, B. Li, H. Lin, Y. Yuan, Chem. Lett. 38 (2009) 572. J. Wei, E. Iglesia, J. Catal. 224 (2004) 370. G. Jones, J.G. Jakobsen, S.S. Shim, J. Kleis, M.P. Anderson, J. Rossmeisl, F. AbildPedersen, T. Bligaard, S. Helveg, B. Hinnemann, J.R. Rostrup-Nielsen, I. Chorkendorff, J. Sehested, J.K. Norskov, J. Catal. 259 (2008) 147. K.P. De Jong, J.W. Geus, Catal. Rev.: Sci. Eng. 42 (2000) 481. P. Serp, M. Corrias, P. Kalck, Appl. Catal. A: Gen. 253 (2003) 337. J. Sun, D. Mei, A.M. Karim, A.K. Datye, Y. Wang, ChemCatChem 5 (2013) 1299.

74

A.L.M. da Silva et al. / Journal of Catalysis 318 (2014) 67–74

[24] J.P. den Breejen, P.B. Radstake, G.L. Bezemer, J.H. Bitter, V. Frøseth, A. Holmen, K.P. de Jong, J. Am. Chem. Soc. 131 (2009) 7197. [25] A. Haryanto, S. Fernando, N. Murali, S. Adhikari, Energy Fuels 19 (2005) 2098. [26] P.D. Vaidya, A.E. Rodrigues, Chem. Eng. J. 117 (2006) 39. [27] P.R. de la Piscina, N. Homs, Chem. Soc. Rev. 37 (2008) 2459. [28] S.M. de Lima, I.O. da Cruz, G. Jacobs, B.H. Davis, L.V. Mattos, F.B. Noronha, J. Catal. 257 (2008) 356. [29] S.M. de Lima, A.M. da Silva, U.M. Graham, G. Jacobs, B.H. Davis, L.V. Mattos, F.B. Noronha, Appl. Catal. A: Gen. 352 (2009) 95. [30] P. Ferrin, D. Simonetti, S. Kandoi, E. Kunkes, J.A. Dumesic, J.K. Nørskov, M. Mavrikakis, J. Am. Chem. Soc. 131 (2009) 5809. [31] R.U. Ribeiro, J.W.C. Liberatori, H. Winnishofer, J.M.C. Bueno, D. Zanchet, Appl. Catal. B: Environ. 91 (2009) 670. [32] H. Song, U.S. Ozkan, J. Catal. 261 (2009) 66. [33] J.-H. Kim, D.J. Suh, T.-J. Park, K.L. Kim, Appl. Catal. A: Gen. 197 (2000) 191. [34] D. Chen, K.O. Christensen, E. Ochoa-Fernández, Z. Yu, B. Tøtdal, N. Latorre, A. Monzón, A. Holmen, J. Catal. 229 (2005) 82. [35] K.O. Christensen, D. Chen, R. Lødeng, A. Holmen, Appl. Catal. A: Gen. 314 (2006) 9. [36] M.K. van der Lee, A. van Jos Dillen, J.H. Bitter, K.P. de Jong, J. Am. Chem. Soc. 127 (2005) 13573. [37] R.C. Reuel, C.H. Bartholomew, J. Catal. 85 (1984) 63. [38] H.P.C.E. Kuipers, H.C.E. van Leuven, W.M. Visser, Surf. Interface Anal. 8 (1986) 235. [39] V. Palma, F. Castaldo, P. Ciambelli, G. Iaquaniello, Appl. Catal. B 145 (2014) 73.

[40] F. Haga, T. Nakajima, H. Miya, S. Mishima, Catal. Lett. 48 (1997) 223. [41] A.M. da Silva, K.R. Souza, G. Jacobs, U.M. Graham, B.H. Davis, L.V. Mattos, F.B. Noronha, Appl. Catal. B: Environ. 102 (2011) 94. [42] S.M. de Lima, A.M. da Silva, G. Jacobs, U.M. Graham, B.H. Davis, L.V. Mattos, F.B. Noronha, Appl. Catal. B: Environ. 96 (2010) 387. [43] L.V. Mattos, G. Jacobs, B.H. Davis, F.B. Noronha, Chem. Rev. 112 (2012) 4094. [44] R. van Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189. [45] R.D. Cortright, R.M. Watwe, J.A. Dumesic, J. Mol. Catal. A: Chem. 163 (2000) 91. [46] R.M. Watwe, R.D. Cortright, J.K. Norskov, J.A. Dumesic, J. Phys. Chem. B 104 (2000) 2299. [47] J.M. Guil, N. Homs, J. Llorca, P.R. de la Piscina, J. Phys. Chem. B 109 (2005) 10813. [48] J.C. Vargas, S. Libs, A.C. Roger, A. Kiennemann, Catal. Today 107 (2005) 417. [49] H. Wang, Y. Liu, L. Wang, Y.N. Qin, Chem. Eng. J. 145 (2008) 25. [50] I. Alstrup, J. Catal. 109 (1988) 241. [51] R.T. Yang, J.P. Chen, J. Catal. 115 (1989) 52. [52] S. Helveg, C. Lopez-Cartes, J. Sehested, P.L. Hansen, B.S. Clausen, J.R. RostrupNielsen, F. Abild-Pedersen, J.K. Norskov, Nature 427 (2004) 426. [53] D.L. Trimm, Catal. Today 37 (1997) 233. [54] J.R. Rostrup-Nielsen, J. Catal. 85 (1984) 31. [55] S.M. Davis, F. Zaera, G.A. Somorjai, J. Catal. 77 (1982) 439. [56] J. Beltramini, D.L. Trimm, Appl. Catal. 31 (1987) 113. [57] B.D. Chandler, A.B. Schabel, L.H. Pignolet, J. Catal. 193 (2000) 186. [58] V. Ponec, H.C. de Jongste, Stud. Surf. Sci. Catal. 7 (1980) 186. [59] J. Barbier, P. Marecot, J. Catal. 102 (1986) 21.