SiO2-Al2O3 catalysts in the metathesis and in the partial oxidation of propene

Applied Catalysis A: General 391 (2011) 78–85

Contents lists available at ScienceDirect

Applied Catalysis A: General journal homepage: www.elsevier.com/locate/apcata

Opposite effect of Al on the performances of MoO3 /SiO2 -Al2 O3 catalysts in the metathesis and in the partial oxidation of propene Damien P. Debecker a,∗ , Damien Hauwaert a , Mariana Stoyanova b , Axel Barkschat b , Uwe Rodemerck b , Eric M. Gaigneaux a,∗ a Institute of Condensed Matter and Nanoscience – Molecules, Solids and reactiviTy (IMCN/MOST), Université catholique de Louvain, Croix du Sud 2/17, B-1348 Louvain-la-Neuve, Belgium1 b Leibniz-Institut für Katalyse e.V. an der Universität Rostock, Albert-Einstein-Str. 29a, 18059 Rostock, Germany

a r t i c l e

i n f o

Article history: Received 18 January 2010 Received in revised form 20 May 2010 Accepted 15 June 2010 Available online 23 June 2010 Keywords: Heterogeneous metathesis catalyst Olefin disproportionation C3 H6 Mesoporous mixed oxide support Allylic oxidation Propylene

a b s t r a c t The selective oxidation and the metathesis of light alkenes are two important catalytic reactions for the petrochemical industry. This paper highlights the contrasting effect of alumina on the catalytic behaviour of silica and silica-alumina supported molybdenum oxide catalysts in these two reactions. Model MoO3 /SiO2 -Al2 O3 catalysts with ca. 6% MoO3 weight loading are prepared via wet impregnation of ammonium heptamolybdate on a set of amorphous mesoporous silica-alumina supports with silica weight content comprised between 100% and 75%. The samples are characterized by ICP-AES, N2 physisorption, XRD, NH3 -chemisorption and XPS and are evaluated in the selective oxidation of propene to acrolein and in its self-metathesis to form butene and ethene. The addition of aluminium oxide into silicon oxide increases the acidity of the support and of the catalyst. It also affects the nature of the deposited species and the dispersion of Mo. Overall, the effect of Al is negative in the case of the partial oxidation of propene because it favours over-oxidation towards carbon oxides. On the contrary, the presence of Al is crucial for the metathesis reaction. It appears that the acidity, created by the presence of alumina in silica, is beneficial for the metathesis reaction at low temperature. An optimum of activity is found for the catalyst supported on the silica-alumina containing 15 wt% of Al2 O3 . © 2010 Elsevier B.V. All rights reserved.

1. Introduction Light alkenes are produced from oil at the refinery and constitute an important feedstock for the chemical industry. Propene – which is taken as model in this study – is highly demanded, mainly because it is the starting point for the production of polypropylene and of oxygenates (acrolein, acrylic acid, propylene oxide, etc.) themselves involved in the synthesis of polymers, resins, surfactants, etc. Taking into account current environmental concerns, the parsimonious use of oil as a natural resource is compulsory. Heterogeneous catalysis is certainly a privileged tool in that perspective [1]. As far as the catalytic upgrading of light alkenes is concerned, different kinds of chemical transformations may be envisaged. An important catalytic process is the selective oxidation of light hydro-

∗ Corresponding authors. Tel.: +32 10473665; fax: +32 10473649. E-mail addresses: [email protected] (D.P. Debecker), [email protected] (E.M. Gaigneaux). 1 IMCN and MOST are new research entities involving the group formerly known as “Unité de catalyse et chimie des matériaux divisés”. 0926-860X/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2010.06.021

carbons which is widely performed with Mo-based heterogeneous catalysts [2]. Very often, the catalysts are prepared by depositing Mo oxide onto another inorganic support: silica [3,4], magnesium oxide [5], alumina [6], etc. It is commonly considered that acidic supports should be avoided because acidic sites provoke coke formation and C–C bond cleavage which leads to catalyst deactivation and to poor selectivity respectively [7]. In silica-based materials, the introduction of Al atoms that substitutes Si atoms brings significant additional acidity [8–10]. Therefore, in selective oxidation reactions, pure silica supports are usually preferred to silica–alumina supports [11]. Another crucial reaction for the petrochemical industry is the metathesis of light alkenes. The cross-metathesis of ethene and butene is an economical mean to produce highly demanded propene from these two cheaper and more abundant alkenes [12–14]. The metathesis of light alkenes is preferentially performed with robust and cheap catalysts in which Mo, Re or W oxides are the active phase [15]. Mo-based catalysts are recognized as more active than W-based catalysts and more stable than Re-based catalysts. Classically, Mo oxide is deposited onto a support (silica [16,17], silica–alumina [18–20], alumina [21,22], titania [23,24], magnesia [25], zirconia [26] and even synthetic

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

clays [27]) by impregnation or grafting to yield active metathesis catalysts. Silica-alumina and silica-supported catalysts are more frequently proposed as best supports [15]. As the metathesis of light alkenes is conducted at relatively low temperature and under anaerobic condition, coke formation and C–C bond cleavage would presumably not occur. On the other hand the process might demand reactive surface species that would allow the first activation step of the C C double bond and its breaking. The rare comparative studies concerning the effect of the support indeed seem to indicate the better performances of molybdenum oxide when deposited onto silica–alumina [18,19,28,29]. Similarly, a recent study reported interesting results obtained with MoO3 SiO2 -Al2 O3 mixed oxide catalysts obtained by a non-hydrolytic sol–gel route [30]. However, the actual role of Al is not fully identified. The influence of the environment of Mo active centres is sometimes interpreted in comparison with the situation encountered in organometallic metathesis catalysts for which the effect of each ligand surrounding the metal active centre is more straightforward. Indeed, one can compare the interactions (chemical bonds) between one Mo species and its support to the interaction between the metal centre of the homogeneous catalyst and its ligands. In that way, the superior activity of silica-alumina supported Mo oxide catalysts was attributed to the electron withdrawing effect created by acidic centres found at Si–O–Al bridges [19]. Also, attempts are made to describe the nature of the molybdenum active species by theoretical approaches [31,32]. Calculations show that the role of the support is to impose a constraint on the Ms –O–Mo bonds (where Ms is the metal of the support), that deforms the Mo environment and reduces the stability of the metallacyclobutane intermediate, leading to facilitated metallacycle opening and enhanced turnover frequencies [32]. Such approaches however need to start from simplified hypotheses (for example isolated Mo species deposited on silica or alumina but not polymolybdates on silica–alumina) and generally omit parameters like surface irregularities, defects, etc. So, as it is clear that the nature of the support can have a major – yet not fully understood – impact on the formation and eventually on the performance of the active phase, it is of primary interest to investigate the effect of Al incorporation experimentally in a systematic way. In this work, model catalysts are prepared by the wet impregnation of a molybdenum salt onto a series of preformed silica and silica-alumina supports. The catalysts and the supports are characterized in terms of texture, bulk and surface compositions, crystallinity and acidity. The catalytic performances are measured both in the metathesis and in the partial oxidation of propene. The purpose of this study is to identify what is the most suited support for each reaction of interest and try to explain in the concrete case why given support formulations outperform others. Two preliminary remarks have to be made. Firstly, the common idea that silica is a better support than silica-alumina for selective oxidation reactions will be verified by measuring the activity of our catalysts in the partial oxidation of propene. It is obvious, however, that the selective production of acrolein from propene is already a mature technology for which highly active and highly selective catalysts have been developed (e.g. (V)–Bi–Mo [2], Mo–V–Sb–Nb [33], Mo–V–Te–Nb [34] mixed oxides). Our results will merely be taken as a point of comparison for the metathesis reaction. Secondly, the main economical motivation of the research in heterogeneous olefin metathesis is the production of propene from butene and ethene. However, for practical reasons, the self-metathesis of propene is studied here. As the metathesis reaction is an equilibrated isothermal reaction, all qualitative conclusions are considered to be transferable to the reverse reaction.

79

2. Experimental 2.1. Preparation of the catalysts The supports are denoted SxA, where “x” represents the proportion of silica in the silica-alumina support (in wt.%). In this study, x ranges from 75 to 100. The supports have been prepared by co-precipitation, through hydrolysis of ethyl orthosilicate and aluminium isopropoxide, as described by Rouxhet and co-workers [10,35]. Prior to impregnation, the supports were calcined for 15 h at 500 ◦ C under static air. A classical wet impregnation technique was used for the transfer of the Mo precursor on the silica and silica-alumina supports. Ammonium heptamolybdate (Aldrich, 99.98%) was dissolved in distilled water to obtain a solution containing 0.666 g of Mo in 100 ml. The desired volume of this solution was introduced in a flask and diluted with distilled water to obtain 200 ml of impregnation solution. 4 g of calcined support were suspended in the impregnation solution for 2 h under magnetic stirring. Water was then evaporated under reduced pressure in a rotary evaporator at 40 ◦ C. The recovered solid was dried at 110 ◦ C for one night and calcined at 400 ◦ C for 2 h in a muffle furnace under static air. Catalysts are denoted SxAMo. 2.2. Characterization The weight percentages of Mo, Si and Al were measured by inductively coupled plasma-atomic emission spectroscopy (ICPAES) on an Iris Advantage apparatus from Jarrell Ash Corporation. Textural analysis was carried out in a Micromeritics Tristar 3000 equipment using the adsorption of nitrogen at −196 ◦ C, working with relative P/P0 pressures in the range of 10−2 to 1.0. For each analysis, 150 mg of sample were degassed under vacuum (50 mTorr) at 150 ◦ C for one night. The specific surface area (SSA) was calculated from the amount of gas adsorbed by using five points with relative P/P0 pressures ranging between 5 × 10−2 and 0.3 (theory of Brunauer, Emmet and Teller). The distribution of pores diameter was determined by the BJH (Barret, Joyner and Halenda) method and the total pore volume was assessed from the amount of nitrogen adsorbed at a relative pressure of about 0.98. Surface characterization was done with X-ray photoelectron spectroscopy (XPS) on a Kratos Axis Ultra spectrometer (Kratos Analytical, Manchester, UK) equipped with a monochromatized aluminum X-ray source (powered at 10 mA and 15 kV). The pressure in the analysis chamber was about 10−6 Pa. The analyzed area was 700 ␮m × 300 ␮m. The pass energy of the hemispherical analyzer was set at 160 eV for the wide scan and 40 eV for narrow scans. Charge stabilization was achieved by using the Kratos Axis device. The electron source was operated at a filament current of 1.8 A and a bias of −1.1 eV. The charge balance plate was set at −2.8 V. The sample powders were pressed into small stainless steel troughs mounted on a multi specimen holder. The following sequence of spectra was recorded: survey spectrum, C 1s, O 1s, Si 2p, Al 2p, Mo 3d, and C 1s again to check for charge stability as a function of time and for the absence of degradation of the sample during the analyses. The binding energy (BE) values were referred to the C–(C, H) contribution of the C 1s peak fixed at 284.8 eV. Molar fractions (%) were calculated using peak areas normalized on the basis of acquisition parameters after a linear background subtraction, experimental sensitivity factors and transmission factors provided by the manufacturer. Spectra were decomposed with the CasaXPS program (Casa Software Ltd., UK) with a Gaussian/Lorentzian (70/30) product function. X-ray diffraction (XRD) measurements were performed on the fresh catalysts with a Siemens D5000 diffractometer using the K␣ radiation of Cu ( = 1.5418 Å). The 2␪ range was recorded between

80

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

5◦ and 75◦ at a rate of 0.02◦ s−1 . The ICDD-JCPDS database was used to identify the crystalline phases. The acidity of the samples has been probed by NH3 chemisorption on the Micromeritics ASAP2010Chemi apparatus. The U-shaped sample tube containing a precisely weighted amount of catalyst (in the 150–200 mg range) was first flushed under a He flow at 300 ◦ C for 2 h and then evacuated at 50 ◦ C for 2 h down to a residual pressure lower than 5 ␮m Hg. The first NH3 adsorption isotherm was measured at 50 ◦ C. Then the sample was evacuated at the same temperature and down to <5 ␮m Hg again. Subsequently, a second NH3 adsorption isotherm was taken. The difference between the two isotherms represents the amount of NH3 still chemisorbed on the sample after evacuation at 50 ◦ C and under vacuum (<5 ␮m Hg). The intercept of a linear interpolation built on the difference points measured between 200 and 700 mm Hg is taken as a measure for the acidity of the sample, expressed in terms of cm3 of chemisorbed NH3 per gram of catalyst. This value is then normalized by the specific surface area of the sample, to give its “total surface acidity”. The same experiment is carried out a second time (on a fresh sample), also taking the two consecutive isotherms at 50 ◦ C but the intermediate evacuation down to <5 ␮m Hg pressure is run at 150 ◦ C. The difference between the two isotherms gives the amount of NH3 still chemisorbed on the sample after evacuation at 150 ◦ C and under vacuum, which is also normalized by the specific surface area to give the “strong surface acidity” of the sample. The standard deviation for each reported value is lower than 10% in relative. 2.3. Catalytic tests 2.3.1. Propene partial oxidation All catalysts were sieved and selected in the 200–315 ␮m size range. The catalytic test was performed in a 50-channel parallel fixed-bed reactor. The quartz reactor tubes (5 mm internal diameter) are located inside an electrically heated furnace. The catalytic bed was deposited on a quartz wool plug and composed of 100 mg of catalysts diluted in 1000 mg of inert silicon carbide. Additional silicon carbide was put onto the catalyst bed for pre-heating the feed gases before entering the catalyst bed. The temperature was measured by a thermocouple located inside the furnace. The reactor is fed with a pre-mixed flow composed of 62.5 vol% of N2 , 25 vol% O2 and 12.5 vol% of propene. The reactor is first heated up (10 ◦ C min−1 ) to the reaction temperature (425 ◦ C) in an air flow. After flushing with nitrogen the feed gas flow is adjusted and after stabilization for 1 h, the product distribution is measured by on-line gas chromatography. For analysis of the feed gas composition a bypass line is used. A 7890 gas chromatograph from Agilent equipped with one TCD and one FID detectors was used to quantify hydrogen, oxygen, nitrogen, CO (HP-Molsieve column, TCD), CO2 , water (HP-PLOT Q column, TCD), acetaldehyde, acetone, propanal, acrolein 2-propanol, 1-propanol, acetic acid, propionic acid, acrylic acid (FFAP column, FID), methane, ethane, ethylene, propane and propene (HP-Al2 O3 column, FID). All reported data are averaged value for three activity measurements. The conversions of propene and oxygen are defined as the ratio “moles converted/moles in the feed × 100%”. The selectivity for each product X is defined as the ratio “moles of X produced/(moles of propene converted × CX )100%”, where CX is the number of carbon atoms in propene divided by the number of carbon atoms in the product X. 2.3.2. Propene metathesis The catalytic tests were carried out at atmospheric pressure in a multi-channel apparatus with a capacity of treating simultaneously up to 15 samples under identical conditions [36]. The whole

Fig. 1. SiO2 content in the support as determined ( of the support, (

) from ICP-AES and (

) from the nominal composition

) from XPS measurements.

design allows fully automated control of gas flows and of three temperature zones (gas pre-heating, reactor, and post reactor lines with 16-port valve) along with reactor switching and product sampling. All catalysts were sieved and the fraction selected in the 200–315 ␮m size range was used. 200 mg of each catalyst were introduced in quartz straight reactors (5 mm i.d.). In each experiment, several samples were pre-treated together by heating up to 550 ◦ C (temperature ramp of 5 ◦ C min−1 ) in N2 (14 ml min−1 flow in each reactor) and keeping this temperature constant for 2 h. Afterwards the system was cooled down to the reaction temperature (40 ◦ C) under the same N2 flow. A pure propene flow (8 ml min−1 ) was admitted for 1 h sequentially in each reactor in order to measure the initial metathesis activity of each sample. These conditions were chosen in order to keep the conversion moderate (below 15%) and ensure that propene is supplied in excess along the catalytic bed. Also, under these conditions, the selectivity towards metathesis products (ethene, cis- and trans-butene) is close to 100%. During activity measurement in a selected reactor, the other reactors are kept under the same N2 flow. Propene and nitrogen were purified over Molsieve 3A (Roth) filters. N2 was additionally purified by an oxygen filter (Oxysorb, Linde). The composition of the reaction gas was analyzed by an Agilent 6890 GC. Product analysis took about 6.5 min for each injection including back flushing of the column. The separation of hydrocarbons was performed on a HPAL/M column (30 m length, 0.53 mm i.d., 0.15 ␮m film thickness) applying a temperature ramp between 90 and 140 ◦ C and FID detection. The activity is calculated on the basis of the addition of the three metathesis products (ethene, cis-butene and trans-butene). The results are expressed in specific activity per Mo atom, namely the number of propene molecules converted to metathesis products per atom of Mo introduced in the reactor and per second. This value is thus not a turn over frequency for each active site but is an average activity for all Mo atoms. The standard deviation is about 3% in relative. 3. Results 3.1. Catalyst characterization The nominal composition of the supports has been verified by ICP-AES and the surface of these supports was also studied by XPS (Fig. 1). The proportion of SiO2 both in the bulk of the materials and at their surface corresponds well with the ratio calculated from their nominal compositions. It is important to note that the texture of the supports varied along the range of compo-

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

81

Table 1 Specific surface area of the supports and of the catalysts (N2 -physisorption), experimental MoO3 loading (ICP-AES) and surface Mo/(Si + Al) atomic ratio (XPS).

S100AMo S95AMo S90AMo S85AMo S80AMo S75AMo a

SSAsupport (m2 g−1 )

SSAcatalyst (m2 g−1 )

Pore Vsupport (cm3 g−1 )

Pore Vcatalyst (cm3 g−1 )

Pore Dsupport (nm)

Pore Dcatalyst (nm)

MoO3 loading (wt.%)

Normalized surface atomic Mo/(Si + Al) ratioa

481 397 263 171 169 152

413 341 246 155 164 141

1.05 0.84 0.71 0.68 0.63 0.68

0.91 0.73 0.65 0.57 0.62 0.61

8.8 8.5 10.8 16.0 15.0 17.8

7.9 8.6 10.6 14.8 15.3 17.3

7.3 5.8 5.5 4.3 4.3 7.6

0.0034 0.0033 0.0030 0.0034 0.0031 0.0029

Ratio calculated from the atomic Mo, Si and Al surface concentrations measured in XPS and divided by the experimental MoO3 loading (in wt.%) determined from ICP-AES.

sition investigated. Table 1 shows that the SSA decreases when the proportion of alumina increases. This decrease seems to be inherent to the preparation method [10,35]. Similar trends were also observed in independent studies [8,37]. The SSA of each catalyst was always slightly lower than that of the corresponding support. Upon impregnation, both the pore volume and the average pore size tend to decrease. The MoO3 loading – verified by ICP-AES – was always in the same range for all catalysts (Table 1). The surface Mo/(Si + Al) atomic ratio was calculated from XPS quantification and normalized by the experimental MoO3 loading for each catalyst (Table 1). This parameter – which can be taken as an indirect indication of the dispersion of Mo at the surface of the supports – is fairly constant for all catalysts. In XRD experiments all samples were characterized by a broad band around 20–25◦ typical of amorphous solids (Fig. 2). In addition, traces of crystalline orthorhombic MoO3 were sometimes found as evidenced by the small diffraction lines mainly found around 27.3◦ and 23.3◦ (JCPDS 05-0508). The catalyst with the highest alumina content (S75AMo) is the only sample which does not exhibit diffraction lines at all. The acidity of the catalysts and of the supports was probed by NH3 -chemisorption. The amount of ammonia chemisorbed per gram of sample was normalized by the specific surface area of each

Fig. 2. XRD patterns obtained with (a) S100AMo, (b) S95AMo, (c) S90AMo, (d) S85AMo, (e) S80AMo and (f) S75AMo. Following JCPDS 05-0508, the peaks at 27.3◦ and 23.3◦ correspond respectively to the (0 2 1) and to the (1 1 0) crystallographic planes of orthorhombic MoO3 (marked with *).

sample (Fig. 3). For the set of support compositions investigated, and with the exception the S75A sample, the higher the Al content, the higher the total acidity. On the basis of IR spectroscopy coupled with pyridine adsorption, a similar trend was found by Rouxhet and co-workers [35]. In the latter study, the total acidity was shown to increase when the alumina content was increased from 0 to 25 and to start to decrease smoothly when the Al2 O3 content is further increased (up to ca. 75 wt.%). The total acidity of S80A was however not measured, so that the maximum measured here could not be observed. It is a well-known fact that silicaalumina materials are more acidic than silica powders [8,10,35]. From NH3 -chemisorption measurements, it appears clearly that the silica support is virtually not acidic. Only a small amount of weak acidic sites is detected, while the strong acidity equals zero for the silica support. As soon as a small amount of Al is introduced (5 wt.% of Al2 O3 in S95A), the total acidity increases significantly and a relatively high proportion of strong acid sites is present. This strong acidity is even more pronounced when the alumina content is raised up to 10 wt.% and then it remains approximately constant in the supports with a higher Al2 O3 content.

Fig. 3. Surface acidity of (a) the SxA supports and (b) of the SxAMo catalysts probed by NH3 -chemisorption. The total acidity (amount of chemisorbed NH3 per square meter of sample at 50 ◦ C) is represented by the light grey columns and the strong acidity (amount of chemisorbed NH3 per square meter of sample at 150 ◦ C) is given in the dark grey columns. The weak acidity can be visualized as the difference between both values.

Others (%)a

3.2 2.8 4.5 4.5 3.7 4.3 3.8

S (propanal) (%)

15.4 12.6 3.9 5.2 2.1 3.0 3.4

c

1-Propanol, 2-propanol, methane, ethane, ethylene, acetic, propionic and acrylic acids. Undetected compounds (calculated from carbon balance). Peak could not be integrated, acetaldehyde is part of undetected compounds. a

b

S (acetaldehyde) (%)

18.2 20.4 n.d.c 7.4 3.7 n.d.c 5.0

Fig. 4. Evolution of the metathesis activity with time-on-stream with () S100AMo, () S95AMo, (×) S90AMo, (♦) S85AMo, () S80AMo and () S75AMo. The activity is expressed in specific activity per Mo atoms (mole of propene converted to metathesis product per mole of Mo atoms introduced in the reactor and per second).

0.0 0.0 2.4 2.5 1.8 2.0 2.2 21.2 25.1 28.7 29.9 34.7 31.4 30.7

S (acetone) (%) S (acrolein) (%)

19.4 16.5 1.9 2.0 1.1 1.6 1.8

S (CO) (%) S (CO2 ) (%)

18.7 18.7 52.5 49.5 52.9 55.4 53.1 6.1 19.0 19.7 16.9 22.1 16.5 15.8

X (propene) (%) GHSVL h−1 g(cat) −1

12.3 3.1 12.3 12.3 12.3 12.3 12.3 S100AMo S100AMo S95AMo S90AMo S85AMo S80AMo S75AMo

Table 2 Conversion and selectivity measured in the oxidation of propene at 425 ◦ C under a 20 and 5 ml min−1 flow (N2 /O2 /C3 H6 = 5/2/1) with 100 mg of SxAMo catalysts.

3.9 3.9 5.8 0 0.1 2.0 0

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

Undetected (%)b

82

The total surface acidity of the SxAMo catalysts is clearly shown to be dependant on the support composition. The same trend is found as for the supports: the total acidity increases with the amount of Al2 O3 (except for S75AMo). For most catalysts, the amount of strong acid sites does not seem to be affected by the impregnation with MoO3 (SxAMo where x = 100, 90, 85) or only increases slightly (S95AMo, S80AMo). The only notable exception is S75AMo which exhibits an unexpected high proportion of strong acid sites. As it will be discussed later, the particularity of the S75A support seems to affect the properties of the deposited Mo oxide phase. From the comparison between the supports and the catalysts, three important observations can be made: (i) in SxAMo and for x comprised between 100 and 80, the impregnation of MoO3 mainly generates additional weak acid sites, (ii) in the S75AMo sample in contrast, the impregnation generates strong acidity and (iii) the silica-based catalysts (S100AMo) exhibits only weak acid sites. 3.2. Catalytic activity The SxAMo catalysts were tested in the partial oxidation of propene at 425 ◦ C. Table 2 shows that, at a given contact time, silica-alumina supported Mo oxide is much more active (propene conversion ranging from 16% to 22%) than silica-supported Mo oxide (conversion of ca. 6%). Acrolein, acetaldehyde, propanal and acetone were the main partial oxidation products. Acrylic acid, acetic acid, propionic acids, 1-propanol, 2-propanol, methane, ethane, ethylene were detected in small quantities (not quantified). Only the MoO3 /SiO2 catalyst exhibited a relatively high selectivity for acrolein and other selective oxidation products. The selectivity towards partial oxidation products like acrolein is very low for alumina-containing catalysts. In a dedicated experiment on S100AMo the flow was reduced (higher contact time) in order to reach a similar conversion level as those obtained with Al-containing supports. At approximately iso-conversion, the selectivity of the different samples can be compared. It appears clearly that silica-supported catalyst remains much more selective towards partial oxidation products. Alumina-containing formulations on the other hand promote the formation of total oxidation products. The selectivity towards CO and CO2 is very high in comparison with the MoO3 /SiO2 system.

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

The same set of catalysts was tested in the self-metathesis of propene at low temperature (40 ◦ C). Under these conditions, primary metathesis products (ethene, cis- and trans-butene) were detected with almost 100% selectivity. The supports were all checked to be inactive. The evolution of the catalyst activity with time-on-stream is showed in Fig. 4. All catalysts are active in the metathesis reaction (the specific activity reported in Fig. 4 correspond to a propene conversion in the 8–12% range for all alumina-containing catalysts). The silica-based catalyst is clearly weaker than the catalysts based on silica-alumina. Moreover, fast deactivation takes place with S100AMo during the first 20 min of the reaction. Al-containing catalysts deactivate slower. It appears that increasing the Al content in the support results in higher stability of the catalyst. For S85AMo, S80AMo and S75AMo a maximum in activity is even noticed after about 14 min time-on-stream (Fig. 4). This behaviour correlates well with the observation of a previous study [30] where the activity of MoO3 /SiO2 -Al2 O3 catalysts (13 wt.% of Al2 O3 in the support) also increased in the first stages of the reaction to reach a maximum at approximately the same moment. Note also that the level of activity compares well with reports of the literature for similar systems [19,30,38–41].

4. Discussion Silica–alumina prepared both industrially or at the laboratory scale usually cannot be seen as a homogeneously mixed oxide materials as they exhibit heterogeneities at the nanoscale range [42]. However, both surface and bulk characterizations of the supports used here showed that their composition is close to the nominal composition indicating that the composition of the supports was well controlled during the preparation and that alumina is homogeneously dispersed in the bulk and at the surface. Textural analysis shows that the SSA of the silica–alumina systems tends to decrease with the increase in Al content. As expected, a slight further decrease is resulting from the impregnation of Mo oxide. The variation of the SSA with respect to the composition of the support affects the Mo surface density, which might affect the morphology of the MoO3 layer that is deposited on the support (e.g., formation of crystals). However, considering the loadings and the SSAs encountered here, all catalysts are in the sub-monolayer range (as a guideline, a loading of 10 wt.% spread as a monolayer of Mo oxide would need a surface of about 110 m2 ). Additionally, XRD results show that the tendency to form MoO3 crystallites is not correlated to the decrease in SSA in the set of catalysts investigated here. The Mo oxide deposit appears as totally amorphous in the S75AMo catalyst and only small diffraction lines appear in the other formulations, revealing traces of MoO3 crystallites. In S100AMo small crystallites are also detected, even though this catalyst has the highest SSA. This must be correlated to the well-known [43] poor interaction between silica and MoO3 that tends to favour the agglomeration of Mo oxide during calcination. The results reported in Fig. 3a unambiguously show the effect of Al on the acidity of the supports. While silica exhibits only a small amount of weak and no strong acidic sites, all silica-alumina materials exhibit higher total acidity and a significant proportion of strong acidic sites. In fact, the range of composition of the silica–alumina supports studied here was selected on purpose. The superior acidity of silica-alumina is indeed reported to be particularly marked when the proportion of Al2 O3 is comprised between 5 and 25 wt.% [35]. A mixed SiAlOx phase is mainly formed in that case. The SiAlOx phase bears acidic hydroxyl groups attached both to a Si and an Al atom (Si–Al bridging hydroxyl groups, also written Si–OH–Al), these sites being directly created by the substitution of Al3+ for Si4+ in a tetrahedral arrangement [10]. In contrast, in silica-alumina having 25 wt.% alumina or more, a highly dispersed alumina phase

83

builds up at the expense of the mixed SiAlOx phase and this was associated to the appearance of Lewis acidity as confirmed by Rajagopal et al. [8]. S75A is thus situated at the border of the two domains described above. Thus it should be considered that the nature of acidic sites may be different in S75A and in the other silica–alumina supports studied. The fact that the continuous and linear tendency found in Fig. 3 is broken at the level of this catalyst thus can easily find its explanation. Finally, correlating our acidity measurements with previous investigations, it can be stated that (i) the acidity of the silica support is very low, (ii) the acidity of the SxA support increases when the SiO2 content decreases from 100 to 80, as the mixed SiAlOx phases builds up, thereby creating bridging OH acidic groups and (iii) the S75A sample seems to exhibit a particular character, most probably due to the appearance of an alumina phase at the expense of the mixed SiAlOx phase which results in a lower density of weak acid sites. Among the SxAMo catalysts (Fig. 3b), even if Mo oxide partially covers the support surface, the same trend is noticed as in the SxA series. In other words, the nature of the support determines strongly the acidic properties of the catalyst. The impregnated MoO3 phase appears to bring additional weak acidity to the SxAMo catalysts with x comprised between 100 and 80. The extent of this increase in weak acidity is roughly the same for all catalysts. Such increase in acidity provoked by the impregnation of Mo oxide is a classical observation, also noticed in the case of titania-supported catalysts [44]. The peculiar acidity of the S75A support translates into peculiar acidity of the S75AMo catalyst. Here, the impregnation of Mo oxide provokes an increase in the strong acidity. Reports indicate that the spreading of MoO3 is more effective on a support exhibiting higher number of Lewis sites [45]. XRD results (Fig. 2) also suggested that the spreading was better in S75AMo. The model of Rouxhet and co-workers [35] implies that an alumina phase starts to develop in the S75A support, which decreases the proportion of Si–Al bridging hydroxyl groups and creates additional Lewis acidity. Like described by Rajagopal et al. the heterogeneity of such support gives rise to a variety of Mo oxide species which results in different acidic characteristics [8]. In line with these authors, it is here believed that the additional acidic sites found in S75A are responsible for a particular mode of deposition of the MoO3 phase which favours the development of a stronger acidity in S75AMo. Raman characterization was tentatively used to step further in the description of the different MoOx species that are stabilized at the surface of each catalyst. However, the strong fluorescence signal originating from all Al-containing supports prevents the detection of these supported species (not shown). Table 2 shows that silica-based catalysts are the most appropriate for carrying out the partial oxidation of propene. Indeed, even if the overall conversion reached by silica-alumina based catalysts is higher, only the MoO3 /SiO2 catalyst exhibited an interesting acrolein yield. On the other hand, all Al-containing formulations preferentially oxidize propene to CO and CO2 . The depicted evolution of the metathesis activity with timeon-stream (Fig. 4) is commonly observed in the metathesis of propene with Mo oxide-based catalysts [19,30]. The induction period observed from the very beginning of the reaction corresponds to the formation of surface carbene species by reaction of propene with activated Mo sites. The nature of these activated sites is still under debate as results indicate that isolated [15,30,46–49], dimeric [49] or polymeric [48,50] MoOx species are the actual active sites. It is however known that a thermal activation is needed to remove traces of molecular oxygen and water from the catalyst [51]. Surface carbenes, produced from the reaction of Mo sites with the olefin at the very beginning of the reaction, then act as the catalytic active sites. It is thought that deactivation occurs when an oxidation reaction occurs instead of the metathesis reaction, leaving a reduced Mo site [40,51,52] or when the fragile carbene active

84

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

Fig. 5. Correlation between the maximum specific activity per Mo atoms (measured after 14 min time-on-stream) and the surface density of acid sites (total).

centre reacts with trace impurities (O2 or H2 O) possibly formed in situ [53]. In our experimental conditions (close to room temperature) only alumina-containing catalysts exhibit high activity in the metathesis of propene. The activity of the molybdena-silica catalyst is comparatively very low. Other studies focussing on MoO3 /SiO2 metathesis catalysts were published and the results are not in contradiction with ours since (i) the activity was low [19], (ii) a special activation procedure based on UV or gamma irradiation had to be used [17,54,55] or (iii) much higher reaction temperature had to be used [19]. Choosing (arbitrarily) to compare the catalysts on the basis of the activity reached after 14 min time-on-stream, the following ranking can be established: S100AMo < S95AMo < S75AMo < S90AMo < S80AMo < S85AMo. Thus, 15 wt.% content of alumina in silica-alumina is identified as the best support composition for preparing Mo-based metathesis catalysts, even though the supports with 10- and 20 wt.% of alumina yield similar results. The activity results discussed here should be understood in the light of acidity measurements (NH3 -chemisorption in Fig. 3). The presence of Al in the formulation provides a marked acidic character. This acidity is either essential – in the case of the metathesis reaction – or deleterious – in the case of the partial oxidation reaction. On the one hand, the presence of Al “switches off” the ability of the catalyst to perform relatively well in the partial oxidation reaction. This expected effect is strikingly confirmed in Table 2. The occurrence of strong acid sites resulting from the presence of Al gives rise to high CO2 selectivity. This tendency can be understood as follows: increasing strength of acidic sites results in a stronger adsorption of propene, which, under the high reaction temperature (425 ◦ C) and in the presence of O2 , leads to the over-oxidation of the hydrocarbon to CO and CO2 . On the other hand, the presence of Al “switches on” the ability of the catalyst to perform well in the metathesis reaction. The presence of strong acidic sites may be regarded as a key factor dictating the activity of the systems. Indeed only the samples exhibiting such strong acidic sites do perform well in the metathesis reaction. However, the total amount of acidic sites follows better the trend observed in the activity. In the set of catalysts investigated here, the catalysts exhibiting the highest density of total acid sites (S80AMo and S85AMo; Fig. 3) are those who exhibit the maximum activity (Fig. 4). There is a reasonable correlation between acidity and activity results, as depicted in Fig. 5. In the end, an optimum in the alumina content (15–20%)

is identified, which translates in an optimum in terms of surface density of acidic sites and in terms of activity. Following Handzlik et al. [19], our interpretation is that the acidic sites present in silica–alumina play the role of electron withdrawer on the Mo species deposited on them. It appears that the presence of a mixed SiAlOx phase exhibiting bridging Si–Al hydroxyl groups is beneficial for the genesis of active metathesis centres. The same conclusion was reached for Re2 O7 -based catalysts as reviewed by Mol [56]. This interesting mixed phase builds up when the alumina content increases from 5 to 20 wt.%. A MoOx species located on such acidic site displays a certain electron deficiency which, exactly like in the case of homogeneous catalysts where the electron deficiency on the Mo centre is provoked by electron withdrawer groups (e.g., imido alkilidene) [57], favours the completion of the metathesis reaction mechanism. The catalyst based on the support containing 25 wt.% of alumina exhibited a higher proportion of strong acidic sites and this was tentatively attributed to the appearance of a new alumina phase that governs a different mode of deposition of MoO3 . These strong sites appear less appropriate for the metathesis reaction. This is coherent with a recent report showing that excessive acidity is deleterious for the metathesis reaction: in MoO3 /HBetalAl2 O3 catalysts the addition of Mg resulted in temperance of the acidity of the support and in increased catalytic performances [58]. Let us finally note that further studies should be conducted concerning the nature of the acidic sites (Brönsted vs. Lewis) in order to go further in the understanding of the nature of the acidic sites that dictate the performances of MoO3 /SiO2 -Al2 O3 metathesis catalysts. 5. Conclusion MoO3 supported on silica–alumina with low Al2 O3 content (between 0 and 25 wt.%) has been studied as catalyst for two reactions dedicated to the upgrading of light alkenes. More precisely, the effect of the composition of the support on the performances of such catalysts in the metathesis and in the selective oxidation of propene was investigated. The main conclusion of this work is very pragmatic: MoO3 /SiO2 formulations should be used for the selective allylic oxidation of light olefins while MoO3 /SiO2 -Al2 O3 catalysts should be used for the metathesis reaction. This distinction arises from the marked acidic character of Al-containing materials as compared to Al-free samples. The nature of the support dictates the acidic properties of the catalyst. As commonly accepted, this acidity is deleterious for performing a partial oxidation reaction since it favours the overoxidation of the hydrocarbon, leading to an increase in CO and CO2 selectivity and to very low yields of partial oxidation products like acrolein. On the contrary, the acidity brought about by the presence of Al is beneficial for the metathesis reaction. All silica-alumina supported catalysts were significantly active in the metathesis of propene while the activity of the silica-based catalyst was very low. An optimal Al2 O3 content in the support was found around 15 wt.%. Acknowledgments The authors acknowledge the Université catholique de Louvain and the Fonds National de la Recherche Scientifique (FNRS) of Belgium. D.P. Debecker thanks the FNRS for his Research Fellow position. The authors from UCL are involved in the “Inanomat” IUAP network sustained by the Service public fédéral de programmation politique scientifique (Belgium). They are also involved in the Cost Action D41 sustained by the European Science Foundation and in the European Multifunctional Material Institute (EMMI) built on the basis of the former “FAME” Network of Excellence of the EU 6th FP.

D.P. Debecker et al. / Applied Catalysis A: General 391 (2011) 78–85

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.apcata.2010.06.021. References [1] J.N. Armor, Appl. Catal. A 194–195 (2000) 3–11. [2] M.M. Bettahar, G. Costentin, L. Savary, J.C. Lavalley, Appl. Catal. A 145 (1996) 1–48. [3] Z.X. Song, N. Mimura, J.J. Bravo-Suarez, T. Akita, S. Tsubota, S.T. Oyama, Appl. Catal. A 316 (2007) 142–151. [4] M. Ruszel, B. Grzybowska, M. Gasior, K. Samson, I. Gressel, J. Stoch, Catal. Today 99 (2005) 151–159. [5] G.E. Vrieland, C.B. Murchison, Appl. Catal. A 134 (1996) 101–121. [6] M.C. Abello, M.F. Gomez, O. Ferretti, Appl. Catal. A 207 (2001) 421–431. [7] C.H. Bartholomew, R.J. Farrauto, Fundamentals of Industrial Catalytic Processes, Jon Wiley &Sons, Inc., Hoboken, NJ, 2006. [8] S. Rajagopal, J.A. Marzari, R. Miranda, J. Catal. 151 (1995) 192–203. [9] D. Martin, D. Duprez, J. Mol. Catal. A 118 (1997) 113–128. [10] R.E. Sempels, P.G. Rouxhet, J. Colloid Interface Sci. 55 (1976) 263–273. [11] US 5,212,137. [12] J.C. Mol, J. Mol. Catal. A 213 (2004) 39–45. [13] US (2002) 1,97,190 (A1). [14] DE102006039904. [15] K.J. Ivin, J.C. Mol (Eds.), Olefin Metathesis and Metathesis Polymerization, A. Press, London, 1997. [16] T. Ookoshi, M. Onaka, Chem. Commun. (1998) 2399–2400. [17] I.R. Subbotina, B.N. Shelimov, V.B. Kazansky, Kinet. Catal. 38 (1997) 678–684. [18] H. Aritani, O. Fukuda, T. Yamamoto, T. Tanaka, S. Imamura, Chem. Lett. (2000) 66–67. [19] J. Handzlik, J. Ogonowski, J. Stoch, M. Mikolajczyk, P. Michorczyk, Appl. Catal. A 312 (2006) 213–219. [20] X. Li, W. Zhang, S. Liu, L. Xu, X. Han, X. Bao, J. Phys. Chem. C 112 (2008) 5955–5960. [21] J. Handzlik, J. Ogonowski, J. Stoch, M. Mikolajczyk, Appl. Catal. A 273 (2004) 99–104. [22] I. Rodriguezramos, A. Guerreroruiz, N. Homs, P.R. Delapiscina, J.L.G. Fierro, J. Mol. Catal. A 95 (1995) 147–154. [23] K. Segawa, T. Soeya, D.S. Kim, Res. Chem. Intermed. 15 (1991) 129–151. [24] K. Tanaka, K. Miyahara, K.I. Tanaka, J. Mol. Catal. 15 (1982) 133–146. [25] S. Hasegawa, T. Tanaka, M. Kudo, H. Mamada, H. Hattori, S. Yoshida, Catal. Lett. 12 (1992) 255–266. [26] V. Indovina, A. Cimino, D. Cordischi, S. Dellabella, S. Derossi, G. Ferraris, D. Gazzoli, M. Occhiuzzi, M. Valigi, W. Grunert, M. Ichikawa, J. Goldwasser, C. Li, J.C. Mol, I.E. Wachs, W.K. Hall, New Front. Catal. Pts a-C 75 (1993) 875–887. [27] A. Gil, M. Montes, Ind. Eng. Chem. Res. 36 (1997) 1431–1443.

85

[28] J. Handzlik, J. Ogonowski, J. Stoch, M. Mikolajczyk, Catal. Lett. 101 (2005) 65–69. [29] J.C. Mol, Catal. Lett. 23 (1994) 113–118. [30] D.P. Debecker, K. Bouchmella, C. Poleunis, P. Eloy, P. Bertrand, E.M. Gaigneaux, P.H. Mutin, Chem. Mater. 21 (2009) 2817–2824. [31] J. Guan, G. Yang, D.H. Zhou, W.P. Zhang, X.C. Liu, X.W. Han, X.H. Bao, J. Mol. Catal. A 300 (2009) 41–47. [32] J. Handzlik, P. Sautet, J. Catal. 256 (2008) 1–14. [33] E.K. Novakova, J.C. Vedrine, E.G. Derouane, J. Catal. 211 (2002) 235–243. [34] P. Botella, J.M.L. Nieto, B. Solsona, A. Mifsud, F. Marquez, J. Catal. 209 (2002) 445–455. [35] P.O. Scokart, F.D. Declerck, R.E. Sempels, P.G. Rouxhet, J. Chem. Soc., Faraday Trans. 1 (1977) 359–371. [36] U. Rodemerck, P. Ignaszewski, M. Lucas, P. Claus, Chem. Eng. Technol. 23 (2000) 413–416. [37] E.J.M. Hensen, D.G. Poduval, P.C.M.M. Magusin, A.E. Coumans, J.A.R. van Veen, J. Catal. 269 (2010) 201–218. [38] J. Handzlik, J. Stoch, J. Ogonowski, M. Mikolajczyk, J. Mol. Catal. A 157 (2000) 237–243. [39] X. Li, W. Zhang, S. Liu, L. Xu, X. Han, X. Bao, J. Catal. 250 (2007) 55–66. [40] X.J. Li, W.P. Zhang, X. Li, S.L. Liu, H.J. Huang, X.W. Han, L.Y. Xu, X.H. Bao, J. Phys. Chem. C 113 (2009) 8228–8233. [41] S. Liu, S. Huang, W. Xin, J. Bai, S. Xie, L. Xu, Catal. Today 93–95 (2004) 471–476. [42] C. Sarbu, B. Delmon, Appl. Catal. A 185 (1999) 85–97. [43] O. Collart, P. Van der Voort, E.F. Vansant, E. Gustin, A. Bouwen, D. Schoemaker, R.R. Rao, B.M. Weckhuysen, R.A. Schoonheydt, Phys. Chem. Chem. Phys. 1 (1999) 4099–4104. [44] D.P. Debecker, F. Bertinchamps, N. Blangenois, P. Eloy, E.M. Gaigneaux, Appl. Catal. B 74 (2007) 223–232. [45] F. Bertinchamps, C. Gregoire, E.M. Gaigneaux, Appl. Catal. B 66 (2006) 10–22. [46] M. Anpa, M. Kondo, Y. Kubokawa, C. Louis, M. Che, J. Chem. Soc., Faraday Trans. 1 84 (1988) 2771–2782. [47] J.L.G. Fierro, J.C. Mol, in: J.L.G. Fierro (Ed.), Metal Oxides: Chemistry and Applications, Taylor & Francis, Boca Raton, 2006, p. 517. [48] P. Topka, H. Balcar, J. Rathousky, N. Zilkova, F. Verpoort, J. Cejka, Microporous Mesoporous Mater. 96 (2006) 44–54. [49] J. Handzlik, Surf. Sci. 601 (2007) 2054–2065. [50] H. Aritani, O. Fukuda, A. Miyaji, S. Hasegawa, Appl. Surf. Sci. 180 (2001) 261–269. [51] J. Handzlik, J. Ogonowski, R. Dula, J. Stoch, E.M. Serwicka, React. Kinet. Catal. Lett. 79 (2003) 135–141. [52] K.A. Vikulov, B.N. Shelimov, V.B. Kazansky, J.C. Mol, J. Mol. Catal. 90 (1994) 61–67. [53] M.N. Kwini, J.M. Botha, Appl. Catal. A 280 (2005) 199–208. [54] I.R. Subbotina, B.N. Shelimov, V.B. Kazansky, Kinet. Catal. 39 (1998) 80–85. [55] K.A. Vikulov, I.V. Elev, B.N. Shelimov, V.B. Kazansky, J. Mol. Catal. 55 (1989) 126–145. [56] J.C. Mol, Catal. Today 51 (1999) 289–299. [57] R.R. Schrock, Tetrahedron 55 (1999) 8141–8153. [58] X. Li, W. Zhang, S. Liu, S. Xie, X. Zhu, X. Bao, L. Xu, J. Mol. Catal. A 313 (2009) 38–43.