Catalysis Communications 5 (2004) 583–590 www.elsevier.com/locate/catcom
A comparative study of Pd/SiO2 and Pd/MCM-41 catalysts in liquid-phase hydrogenation Joongjai Panpranot a
a,*
, Kanda Pattamakomsan a, James G. Goodwin Jr. b, Piyasan Praserthdam a
Center of Excellence on Catalysis and Catalytic Reaction Engineering, Department of Chemical Engineering, Chulalongkorn University, Bangkok 10330, Thailand b Department of Chemical Engineering, Clemson University, Clemson, SC 29634, USA Received 1 March 2004; revised 30 June 2004; accepted 9 July 2004 Available online 23 August 2004
Abstract The characteristics and catalytic properties of Pd/MCM-41 and Pd/SiO2 were investigated and compared in terms of Pd dispersion, catalytic activities for liquid-phase hydrogenation of 1-hexene, and deactivation of the catalysts. High Pd dispersion was observed on Pd/MCM-41-large pore catalyst while the other catalysts showed relatively low Pd dispersion due to significant amount of Pd being located out of the pores of the supports. Based on CO chemisorption results, the catalyst activities seemed to be primarily related to the Pd dispersion and not to diffusion limitation since TOFs were nearly identical for all the catalysts used in this study. In all cases, leaching and sintering of Pd caused catalyst deactivation after 5-h batch reaction. However, compared to Pd/SiO2 with a similar pore size, Pd/MCM-41 exhibited higher hydrogenation activity and lower amount of metal loss. 2004 Elsevier B.V. All rights reserved. Keywords: Liquid phase hydrogenation; Pd/MCM-41; Pd/SiO2; 1-Hexene hydrogenation
1. Introduction Supported Pd catalysts are widely used in liquidphase hydrogenation for many important organic transformations [1]. Common catalyst supports include activated carbons [2,3], silicas [4,5], aluminas [6,7], and to a lesser extent, polymers [8], and zeolites [9]. It is know that a support can affect catalyst activity, selectivity, recycling, refining, materials handling, and reproducibility. For examples, Shen et al. [10] reported the effect of the nature of the supports (Al2O3, SiO2, TiO2, and ZrO2) on both activity and selectivity of supported
*
Corresponding author. Tel.: +66 02 218 6859; fax: +66 02 218 6877. E-mail address:
[email protected] (J. Panpranot). 1566-7367/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.catcom.2004.07.008
Pd catalysts in catalytic hydrogenation of carbon monoxide. Pinna et al. [11] compared Pd on activated carbon, silica, and alumina in the selective hydrogenation of benzaldehyde to benzyl alcohol. Pd/Al2O3 was found to exhibit strong metal–support interaction while Pd/C showed the highest activity for benzaldehyde hydrogenation. Besides the nature of the support, support structure can also affect catalyst performance. Choudary et al. [9] reported that Pd/MCM-41 is more active in partial-hydrogenation of acetylenic compounds than Pd/Yzeolite or Pd/K-10 clay. Ordered mesoporous silicas such as MCM-41 [12,13], SBA-x [14], and HMS [15], have been shown to be suitable catalyst supports. These mesoporous materials possess high BET surface areas, large pore volumes, and highly ordered pore structures with narrow pore size distributions in the range of 2–50 nm, depending on synthesis chemicals and conditions. Pt/MCM-41 has been
584
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
reported to exhibit high activity for the hydrogenation of aromatics in diesel fuels [16]. Co–Mo/MCM-41 was found to be more active than the conventional Co– Mo/Al2O3 catalyst for hydrodesulfurization [17]. For liquid phase reaction, Shimazu et al. [18] recently reported that MCM-41-supported Pd catalysts prepared by grafting of a palladium complex on MCM-41 exhibited high activity and high regioselectivity in the hydrogenation of dienes with –OH groups. In this study, we compare the characteristics and catalytic properties of Pd/SiO2 and Pd/MCM-41 with 2 different pore sizes (small and large pore). The effects of silica-structure and pore size of the silica supports were investigated in terms of metal dispersion, catalytic activity in liquid-phase hydrogenation, and deactivation due to metal leaching. These results supply useful information on the role of support porosity and type of silica used in supported Pd catalysts for liquid phase hydrogenation reactions.
2. Experimental 2.1. Preparation of supported Pd catalysts Pure silica MCM-41 with 3 nm pore diameters was prepared in the same manner as that of Cho et al. [19] using the gel composition of CTABr:0.3 NH3:4 SiO2: Na2O:200 H2O, where CTABr denotes cetyltrimethyl ammonium bromide. Briefly, 20.03 g of colloidal silica Ludox AS 40% (Aldrich) was mixed with 22.67 g of 11.78% sodium hydroxide solution. Another mixture comprised of 12.15 g of CTABr (Aldrich) in 36.45 g of deionized water, and 0.4 g of an aqueous solution of 25% NH3. Both of these mixtures were transferred into a Teflon lined autoclave, stirred for 30 min, then heated statically at 100 C for 5 days. The obtained solid material was filtered, washed with water until no base was detected and then dried at 100 C. The sample was then calcined in flowing nitrogen up to 550 C (1–2 C/min), then in air at the same temperature for 5 h, and is referred to in this paper as small pore MCM-41. The larger pore MCM-41 was prepared by treating the MCM-41-small pore (before calcination) in an emulsion containing N,N-dimethyldecylamine (0.625 g in 37.5 g of water for each gram of MCM-41) for 3 days at 120 C. This was washed thoroughly, dried, and calcined in flowing nitrogen up to 550 C (1–2 C/min), then in air at the same temperature for 5 h. This support is referred to in this paper as large pore MCM-41. The amorphous SiO2small pore (chromatographic grade) and SiO2-large pore were obtained commercially from Grace Davison and Strem chemicals, respectively. Supported Pd catalysts were prepared by the incipient wetness impregnation of the supports with an aqueous
solution containing the desired amount of Pd nitrate dehydrate to yield a final loading of approximately 0.5 wt% Pd. The catalysts were dried overnight at 110 C and then calcined in air at 500 C for 2 h.
2.2. Catalyst characterization The bulk composition of palladium was determined using a Varian Spectra A800 atomic adsorption spectrometer. The BET surface area of the catalysts were determined by N2 physisorption using a Micromeritics ASAP 2000 automated system. Each sample was degassed in the Micromeritics ASAP 2000 at 150 C for 4 h prior to N2 physisorption. The XRD spectra of the catalysts were measured using a SIEMENS D5000 X-ray diffractometer and Cu Ka radiation with a Ni filter in the 2–10 or 20–80 2h angular regions. Catalyst particle morphology was obtained using a JEOL JSM-35CF scanning electron microscope operated at 20 kV. The palladium oxide particle size and distribution of palladium was observed using a JEOLTEM 200CX transmission electron microscope operated at 100 kV. The catalyst sample was first suspended in ethanol using ultrasonic agitation for 10 min. The suspension was dropped onto a thin Formvar film supported on a copper grid and dried at room temperature before TEM observation. Relative percentages of palladium dispersion were determined by pulsing carbon monoxide over the reduced catalyst. Approximately 0.2 g of catalyst was placed in a quartz tube, incorporated in a temperature-controlled oven and connected to a thermal conductivity detector (TCD). Prior to chemisorption, the catalyst was reduced in a flow of hydrogen (50 cc/min) at room temperature for 2 h. Then the sample was purged with helium for 1 h. Carbon monoxide was pulsed at room temperature over the reduced catalyst until the TCD signal from the pulse was constant.
2.3. Liquid-phase hydrogenation Liquid-phase hydrogenation reactions were carried out at 25 C and 1 atm in a stainless steel Parr autoclave. Approximately 1 g of supported Pd catalyst was placed into the autoclave. The system was purged with nitrogen to remove remaining air. The supported Pd catalyst was reduced with hydrogen at room temperature for 2 h. The reaction mixture composed of 15 ml of 1-hexene and 400 ethanol was first kept in a 600 ml feed column. The reaction mixture was introduced into the reactor with nitrogen to start the reaction. The content of hydrogen consumption was monitored every 5 min by noting the change in pressure of hydrogen.
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
3. Results and discussion 3.1. Catalytic activities for liquid phase hydrogenation of 1-hexene Liquid-phase hydrogenation of 1-hexene was carried out as a model reaction to compare the hydrogenation activity of the MCM-41- and SiO2-supported Pd catalysts. Mild conditions were chosen because rate of reaction can be studied more easily. In terms of kinetics study, the effects of 1-hexene concentration (0.1–0.3), hydrogen pressure (1–2 bar), and temperature (25–40 C) were investigated. The initial rate of hydrogenation of 1-hexene was found to be independent on both 1-hexene concentration and hydrogen pressure. The apparent activation energy calculated from the Arrhenius plot was found to be 41 kJ/mol. The kinetics study of liquid phase hydrogenation using rate of hydrogen consumption have also been reported in other research investigations [9,20]. The slope of the line at the origin represents the initial rate of the reactions and the rate constant assuming zero order dependence of reaction on hydrogen [9,20,21]. The rate constant of the different supported Pd catalysts in liquid phase hydrogenation of 1-hexene at 25 C and the corresponding turnover frequencies (TOFs) are reported in Table 1. The activity of the catalysts were found to be in the order of Pd/ MCM-41-large pore > Pd/MCM-41-small pore Pd/
585
SiO2-large pore > Pd/SiO2-small pore. The TOFs were calculated based on the CO chemisorption data (Table 2). Given that in all cases the support was silica, albeit in slightly different forms, it is not surprising that specific activities in the form of TOFs were so similar. Since the TOFs were similar but Pd particle locations so disparate: on the outside of the silica granules for Pd/ SiO2-small pore vs. inside mesopores for Pd/MCM-41large pore, etc. (based on the average particle sizes of PdO measured from XRD and TEM), it would appear that there were no pore diffusion effects on reaction rate. However, this does not mean that there might not have been some limitations in the mass transfer of hydrogen from the gas phase to the liquid phase, given that hydrogenation is such a fast reaction on noble metals. However, we did not detect any mass transfer limitations due to pore diffusion. This is contrary to the effect of pore size on the liquid-phase diffusion resistance reported in the literature for other catalyst systems [22,23] The structure of MCM-41, however, may promote the reaction by affecting the crystal structure of the supported Pd to form more active centers for Pd catalysts [24]. 3.2. Catalyst characterization before and after reaction The characteristics of the catalysts before and after reaction were obtained using various analysis tools such
Table 1 Catalyst characteristics and liquid phase hydrogenation activity Catalyst
Pd/SiO2-small pore Pd/SiO2-large pore Pd/MCM-41-small pore Pd/MCM-41-large pore a b c d
BET surface areaa (cm3/g)
PdO particle sizec (nm)
Support
Fresh
Spentb
Fresh
Spent
716 277 921 901
675 201 670 726
343 191 377 546
7.8 10.8 14.2 9.1
13.8 21.9 16.1 18.0
Rate constant · 103 (mol/min g cat.)
TOFsd (s 1)
2.45 3.40 3.41 4.75
19 29 28 23
Error of measurement was ±5%. After 5-h batch hydrogenation of 1-hexene at 25 C and 1 atm and re-calcination at 500 C for 2 h. Based on XRD results. Based on CO chemisorption results.
Table 2 Results from pulse CO chemisorption and atomic adsorption Catalyst
Pd/SiO2-small pore Pd/SiO2-large pore Pd/MCM-41-small pore Pd/MCM-41-large pore a b c d
CO chemisorptiona (·1018molecule CO/g cat.)
Pd dispersionb (%)
Pd0 particle sizec (nm)
Fresh
Spent
Fresh
Spent
Fresh
Spent
Fresh
Spent
1.29 1.22 1.20 2.10
0.79 0.61 0.77 1.34
6.5 7.3 4.4 12.0
3.8 4.1 2.3 7.7
17.2 15.3 24.9 9.4
29.2 30.1 45.3 14.7
0.35 0.29 0.41 0.33
0.15 0.18 0.32 0.31
Error of measurement was ± 5%. Based on the total palladium. Based on d = (1.12/D) nm [29]. Based on atomic adsorption (AA) results.
%Pd loadingd (wt%)
586
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
Fig. 1. Pore size distribution of different silica supports.
as BET, XRD, TEM, SEM, atomic adsorption, and CO chemisorption. The spent catalysts were collected after one 5-h batch reaction. In order to avoid the influence of carbon deposits on the pore structure and the contrast of SEM and TEM images of the used catalysts. After being taken out from the reactor and dried at room temperature, the catalysts were re-calcined in air at 500 C for 2 h before characterization to remove any carbon deposits. The BET surface areas and pore volumes of the original supports and the catalysts before and after reaction are given in Table 1. Except for the low surface area commercial large pore silica, SiO2-small pore, MCM41-small pore, and MCM-41-large pore possessed high BET surface areas of 716–921 m2/g. After impregnation of palladium, the BET surface area and pore volume of the original supports decreased by approximately 6– 20%, suggesting that palladium was deposited in some of the pores of the supports. Fig. 1 shows the pore size distribution of the four different silica supports used in this study. A narrow pore size distribution was observed for all the catalyst supports except the commercial large pore silica. The BET surface areas of the spent catalysts ranged between 191 and 546 m2/g. Since most of the carbon deposits were removed by calcination at 500 C, the changes in BET surface areas after reaction was probably due to pore blocking by metal sintering and/or destruction of support pore structure after reaction. The X-ray diffraction patterns of the original MCM41-small pore and Pd/MCM-41-small pore (before and after reaction) are shown in Fig. 2. The ordered structure of small pore MCM-41 gave XRD peaks at 2.40, 3.96, and 4.53 2h. After impregnation of palladium the intensities of the XRD peaks for MCM-41 were decreased suggesting either that the structure of MCM-41 became less ordered upon impregnation with Pd or that Pd particles inside the pores caused significant X-ray
Fig. 2. XRD patterns of MCM-41-small pore and Pd/MCM-41-small pore before and after reaction.
scattering. In any case, the structure of MCM-41 was not destroyed, although the long-range order of MCM41 may have shrunk [25]. After reaction, the character-
Fig. 3. XRD patterns at high degrees 2h.
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
istic peaks of MCM-41-small pore decreased further, probably due to the instability of pure silica MCM-41 and/or the greater X-ray scattering ability of sintered metal particles [26]. The XRD patterns at higher diffraction angles of the different silica supported Pd are shown in Fig. 3. In the calcined state, diffraction peaks for palladium oxide (PdO) were detectable at 33.8 and less so at 42.0, 54.8, 60.7, and 71.4 2h. After reaction (before re-calcination), palladium metal (Pd0) was in evidence with diffraction peaks at 40.1 and less so at 46.6 2h. The recovered catalysts were then re-calcined in order to compare the PdO particle sizes before and after reaction. The PdO particle sizes were calculated from XRD line broadening of the peak at 33.8 2h using ScherrerÕs equation [27] and are reported in Table 2. In the calcined catalysts, the PdO particles/clusters were found to be ca.
587
8–14 nm in the order of Pd/MCM-41-small pore > Pd/ SiO2-large pore > Pd/MCM-41-large pore Pd/SiO2small pore. After reaction and re-calcination, it was found that the PdO particle sizes for all catalyst samples became larger, suggesting sintering of palladium metal particles [28]. Use of large pore supports resulted in the greatest amount of Pd sintering. It should be noted that the lower the surface area, the higher probability of a metal particle to be closed to other ones and therefore the sintering probability is also higher. Typical TEM micrographs of the fresh and spent supported Pd catalysts are shown in Fig. 4. TEM micrographs were taken in order to physically measure the size of the palladium oxide particles and/or palladium clusters. Since the metal loading is low and the surface area of the sample is very high, in principle it is quite difficult to observe metal dispersion by TEM. However, a
Fig. 4. TEM micrographs of (a) fresh and (b) re-calcined used supported Pd catalysts.
588
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
large number of pictures were collected from different portion of the samples in order to obtain further evidence about the dispersion of the palladium. The particle sizes of PdO particles measured from TEM images before and after reaction were found to be in accordance with the results from XRD. The TEM images of the used catalysts were taken after all the carbon deposits have been removed. The differences in PdO particle sizes on different supports then can be ascribed to differences induced by the support on the sintering of Pd during preparation and reaction [9]. Typical SEM micrographs of catalyst particles are shown in Fig. 5. The SEM micrographs show different catalyst particle sizes. For both small pore and large pore MCM-41 supported catalysts, the particle sizes were approximately 20–30 lm, whereas particle sizes
of the small and large pore SiO2 were around 130–150 and 400–500 lm, respectively. It would appear that the size and shape of the catalyst particles were not affected during the 5-h batch liquid-phase hydrogenation reaction suggesting that the different silica supports used in this study have reasonably high attrition resistance. The amounts of CO chemisorption on the catalysts, the percentages of Pd dispersion, and the Pd metal particle sizes are given in Table 2. The pulse CO chemisorption technique was based on the assumption that one carbon monoxide molecule adsorbs on one palladium site [29–34]. It should be noted that this is an approximation since 1-hexene may not reach all the adsorption sites available for CO adsorption. Before reaction, it was found that Pd/MCM-41-large pore exhibited the highest amount of CO chemisorption, whereas the other cata-
Fig. 5. SEM micrographs of (a) fresh and (b) re-calcined used supported Pd catalysts.
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
lysts showed similar amounts of CO chemisorption. This was probably due to the dispersion of palladium being better on the large pore MCM-41. The percentages of Pd dispersion before and after reaction were calculated to be in the range of 4.4–12.0% and 2.3–7.7%, respectively, in the descending order of Pd/MCM-41-large pore > Pd/SiO2-large pore > Pd/SiO2-small pore > Pd/ MCM-41-small pore. Average particle size for reduced Pd0 was calculated to be in the range of 9.4–24.9 nm before reaction and 14.7–45.3 nm after reaction. These average Pd metal particle sizes calculated based on CO chemisorption are not identical of course to the calculated PdO particle sizes obtained by XRD and TEM. The MCM-41-large pore supported catalysts appeared to have smaller average Pd particle sizes than those on other catalysts, resulting in their relatively high Pd dispersion. After reaction, the amount of CO chemisorption and % Pd dispersion decreased significantly for all catalyst samples by approximately 35–50% of that of the fresh catalysts. The actual amounts of palladium loading before and after reaction were also determined by atomic adsorption spectroscopy and are given in Table 2. Before reaction, palladium loading on the catalyst samples was approximately 0.29–0.48 wt%. After one 5-h batch hydrogenation reactions of 1-hexene, palladium loading had decreased to 0.15–0.32 wt%. This indicates that leaching of palladium occurred during reaction. The order of the percentages of the amount of palladium leaching was Pd/SiO2-small pore > Pd/SiO2-large pore > Pd/ MCM-41-small pore Pd/MCM-41-large pore. Pd/ MCM-41-large pore showed almost no leaching of palladium into the reaction media within experimental error. The results suggest that the structure of MCM-41 may modify bulk and surface properties of the Pd since Pd metal particles on MCM-41 exhibited different characteristics and catalytic properties compared to those on amorphous silica. An alternative explanation is that Pd/ MCM-41-large pore catalyst is likely to have small palladium particles located mostly in the pores of MCM41, consequently Pd was not prone to leaching compare to the other catalysts. Smaller particles can also interact more with the supports than larger ones. A peculiar metal–support interaction between metal and MCM-41 support has been reported in the literature for CoRu/ MCM-41 catalysts, Co was found to interact more with MCM-41 than SiO2 [35].
4. Conclusions Type of silica, pore size and pore structure were found to affect characteristics and catalytic properties of the silica supported Pd catalysts in liquid-phase hydrogenation. The catalyst activities were found to be merely dependent on the Pd dispersion, which as itself
589
a function of the support pore structure. There was no evidence for pore diffusion limitations on any of the catalysts since TOFs were nearly identical. Among the four types of the supported Pd catalysts used in this study, Pd/MCM-41-large pore showed the highest Pd dispersion and the highest hydrogenation rate with the lowest amount of metal loss.
Acknowledgements The financial supports of the Thailand Research Fund (TRF) and TJTTP-JBIC are gratefully acknowledged. The authors also thank Professor Abdel Sayari, University of Ottawa, Canada and Dr. Aticha Chaisuwan, Chulalongkorn University, Thailand for the assistance with MCM-41 preparation.
References [1] P.N. Rylander, Hydrogenation Methods, Academic Press, New York, 1985. [2] P. Chou, M.A. Vannice, J. Catal. 107 (1987) 129. [3] S.D. Jackson, L.A. Shaw, Appl. Catal. A 134 (1996) 91. [4] V.M. Frolov, O.P. Parengo, A.V. Novikova, L.S. Kovaleva, React. Kinet. Catal. Lett. 25 (1984) 319. [5] G.M. Cherkashin, L.P. Shuikina, O.P. Parengo, V.M. Frilov, Kinet. Katal. 26 (1985) 1110. [6] K. Borszeky, T. Mallet, R. Eschiman, W.B. Schweizer, A. Baiker, J. Catal. 161 (1996) 451. [7] P.L. LÕArgentiere, D.A. Liprandi, E.A. Cagnola, N.S. Figoli, Catal. Lett. 44 (1997) 101. [8] J.P. Mathew, M. Srinivasan, Eur. Polym. J. 31 (1995) 835. [9] B.M. Choudary, M.L. Kantam, N.M. Reddy, K.K. Rao, Y. Haritha, V. Bhaskar, F. Figueras, A. Tuel, Appl. Catal. A 181 (1999) 139. [10] W.J. Shen, M. Okumura, Y. Matsumura, M. Haruta, Appl. Catal. A 213 (2001) 225. [11] F. Pinna, F. Menegazzo, M. Signoretto, P. Canton, G. Fagherazzi, N. Pernicone, Appl. Catal. A 219 (2001) 195. [12] J.S. Beck, J.C. Vartuli, W.J. Roth, M.E. Leonowicz, C.T. Kresge, K.D. Schmitt, C.T.-W. Chu, D.H. Olson, E.W. Sheppard, S.B. McCullen, J.B. Higgins, J.L. Schlenker, J. Am. Chem. Soc. 114 (1992) 10834. [13] C.T. Kresge, M.E. Leonowicz, W.J. Roth, J.C. Vartuli, J.S. Beck, Nature 359 (1992) 710. [14] D. Zhao, J. Feng, Q. Huo, N. Melosh, G.H. Frederickson, B.F. Chmelka, G.C. Stucky, Science 279 (1998) 548. [15] P.T. Tanev, T.J. Pinnavaia, Science 267 (1995) 865. [16] A. Corma, A. Martinez, V. Martinez-Soria, J. Catal. 169 (1997) 480. [17] C. Song, K.M. Reddy, Appl. Catal. A 176 (1999) 1. [18] S. Shimazu, N. Baba, N. Ichiku, T. Uematsu, J. Mol. Catal. 182 (2002) 343. [19] D.H. Cho, T.S. Chang, S.K. Ryu, Y.K. Lee, Catal. Lett 64 (2000) 227. [20] M.K. Dalal, R.N. Ram, Eur. Polym. J. 33 (1997) 1495. [21] E.A. Sales, M.D. Mendes, F. Bozon-Verduraz, J. Catal. 195 (2000) 96. [22] Y.Y. Huang, W.M.H. Sachtler, Appl. Catal. A 163 (1997) 245. [23] R. Takahashi, S. Sato, T. Sodesawa, H. Nishida, Phys. Chem. Chem. Phys. 4 (2002) 3800.
590
J. Panpranot et al. / Catalysis Communications 5 (2004) 583–590
[24] G. van de Lee, V. Ponec, Catal. Rev. Sci. Eng. 29 (1987) 183. [25] L. Pasqua, F. Testa, R. Aiello, F. Di Renzo, F. Fajula, Micropor. Mesopor. Mater. 44–45 (2001) 111. [26] B. Marler, U. Oberhagemann, S. Vartmann, H. Gies, Micropor. Mater. 6 (1996) 375. [27] H.P. Klug, L.E. Alexander, X-ray Diffraction Procedures For Polycrystalline Amorphous Materials, second ed., Wiley, New York, 1974. [28] O. Dominguez-Quintero, S. Martinez, Y. Henriquez, L. DÕOrnelas, H. Krentzien, J. Osuna, J. Mol. Catal. A 197 (2003) 185.
[29] N. Mahata, V. Vishwanathan, J. Catal. 196 (2000) 262. [30] S.H. Ali, J.G. Goodwin Jr., J. Catal. 176 (1998) 3. [31] E.A. Sales, G. Bugli, A. Ensuque, M.J. Mendes, F. BozonVerduraz, Phys. Chem. Chem. Phys. 1 (1999) 491. [32] A. Sarkany, Z. Zsoldos, B. Furlong, J.W. Hightower, L. Guczi, J. Catal. 141 (1993) 566. [33] M.A. Vannice, S.Y. Wang, S.H. Moon, J. Catal. 71 (1981) 152. [34] N.K. Nag, Catal. Lett. 24 (1994) 37. [35] J. Panpranot, J.G. Goodwin Jr., A. Sayari, Catal. Today 77 (2002) 269.