- Email: [email protected]
ZSM5 catalysts: Quantitative study by TPR and XAS
Microporous and Mesoporous Materials 93 (2006) 62–70 www.elsevier.com/locate/micromeso
Effect of water on cobalt speciation during solid-state synthesis of Co2+/ZSM5 catalysts: Quantitative study by TPR and XAS Hanene Ben Boubaker a, Mourad Mhamdi a, Eric Marceau b,*, Sihem Khaddar-Zine a, Abdelhamid Ghorbel a, Michel Che b,c, Youne`s Ben Taarit d, Franc¸oise Villain e,f a Laboratoire de Chimie des Mate´riaux et Catalyse, Universite´ Tunis II-El Manar, 1060 Tunis, Tunisia Laboratoire de Re´activite´ de Surface (UMR 7609 CNRS), Universite´ Pierre et Marie Curie, 4 place Jussieu, 75252 Paris Cedex 05, France c Institut Universitaire de France, France d Institut de Recherches sur la Catalyse (UPR 5401 CNRS), 2 Avenue Albert Einstein, 69626 Villeurbanne Cedex, France e Laboratoire de Chimie Inorganique et Mate´riaux Mole´culaires (UMR 7071 CNRS), Universite´ Pierre et Marie Curie, 4 place Jussieu, 75252 Paris Cedex 05, France Laboratoire pour l’Utilisation du Rayonnement Electromagne´tique (LURE), Centre Universitaire Paris-Sud, BP 34, 91898 Orsay Cedex, France b
f
Received 12 December 2005; received in revised form 28 January 2006; accepted 1 February 2006 Available online 20 March 2006
Abstract The effect of water on cobalt speciation in the synthesis of Co2+/ZSM5 catalysts is quantitatively studied by the combined use of macroscopic and local characterization techniques, TPR and X-ray absorption spectroscopies, respectively. The addition of a small quantity of water to cobalt acetate and zeolite during the preparation of Co2+/ZSM5 by solid-state reaction favors cobalt dispersion and migration into the zeolite. It also inhibits the formation of extraframework phases (Co3O4 particles, cobalt phyllosilicate). Dissociation of solid cobalt acetate upon contact with water is thought to be the driving force for enhanced cobalt migration into the zeolite. 2006 Elsevier Inc. All rights reserved. Keywords: Solid-state reaction; TPR; XAS; Speciation; Quantitative analysis
1. Introduction Among the various processes used to introduce metal ions into zeolites, solid-state reactions between a protonic zeolite and a solid precursor salt have gained interest for two reasons: they avoid the handling of large quantities of solutions and there is no waste of precursor in contrast to the method of aqueous ion exchange at equilibrium [1,2]. However, a non-negligible fraction of metal ions deposited during solid-state reaction do not exchange with protons, especially when the amount of precursor salt exceeds the stoichiometry of exchange: they react instead with zeolite to form extraframework phases. It becomes thus important
*
Corresponding author. Tel.: +33 1 44 27 60 04; fax: +33 1 44 27 60 33. E-mail address: [email protected] (E. Marceau).
1387-1811/$ - see front matter 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2006.02.004
to characterize such phases which have a different reactivity from that of exchanged ions, as shown for example in the cases of NOx reduction [3–6] and hydrocarbon ammoxidation [7–9]. These phases are usually oxidic particles, but it has been shown recently that the use of solid cobalt acetate to prepare Co2+/ZSM5 systems may lead to another type of phase, a well-crystallized phyllosilicate incorporating in its bulk the majority of cobalt ions, after dealumination of the zeolite by acetate anions during thermal treatment [10]. In order to modify this anion-support interaction leading to a result detrimental to catalysis [9], one can imagine to add another chemical partner, water, in a limited amount, as a way to change the type of contact between precursor salt and support before thermal treatments be involved. The purpose of this paper is to identify the phases formed during this ‘‘wet’’ solid-state procedure—hereafter
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
63
2. Experimental
110 C overnight. The resulting solid was heated in a helium flow (25 mL min1) up to 500 C (heating rate: 2 C min1) and kept at 500 C overnight (12 h) before being submitted to the oxygen activation treatment described above. It can be stressed here that during the preparation of catalysts I, the sample is contacted with water before the first thermal treatment, that is before reaction between zeolite and cobalt acetate; whereas the contact with water occurs after the first thermal treatment in series A, that is after the reaction between zeolite and cobalt acetate has taken place.
2.1. Catalysts preparation
2.2. Characterization
Two series of three samples were prepared, one by solidstate reaction and the other one by impregnation (Table 1). For the sake of homogeneity regarding earlier results [11], samples prepared by solid-state reactions are referred to as Ax, following the nominal Co/Al atomic ratio, x = 1/2, 1 or 3/2 (i.e. corresponding to one, two or three times the exchange stoichiometry). Samples prepared by impregnation are referred to as Ix. It must be noted that the purpose of the synthesis of samples from series I is primarily characterization, while reaching the exact nominal ratio x is not sought for; consequently, x in the denomination Ix was chosen mostly by analogy with the A series with increasing Co/Al content. Catalysts A were prepared by solid-state reaction between H-ZSM5 zeolite (Zeocat, Si/Al = 26, Al wt.% = 1.6, water content: 5 wt.%, pore volume: 0.54 mL g1) and cobalt(II) acetate [Co(CH3COO)2] Æ 4H2O (Prolabo). Powders were finely ground and mixed in a mortar for 15 min in ambient conditions. The resulting mixture was heated in a helium flow (25 mL min1) up to 500 C (heating rate: 2 C min1) and kept at 500 C overnight (12 h). After having been washed twice with deionized water and dried overnight at 110 C in a static oven, the powder was submitted to a final activation treatment in O2, up to 500 C (25 mL min1, heating rate: 5 C min1) and 1 h at that temperature. In the preparation procedure of catalysts I, a volume of deionized water (1.5 mL g1 of zeolite) was added to the mixture of ground powders (zeolite and precursor). The whole system was pugged for 10 min and dried in air at
Chemical analyses were performed by ICP at the Vernaison Center of Chemical Analysis of CNRS. Specific surface areas were determined by the BET method from nitrogen physisorption at 196 C, using an automatic Micromeritics ASAP 2000 instrument. Micropore volumes were assessed by application of the Dubinin equation between 0.01 < P/P0 < 0.1. Temperature-programmed reductions (TPR) were performed under 5% H2 in argon (25 mL min1); the hydrogen consumption was measured using a thermal conductivity detector, from room temperature to 1000 C with a heating rate of 7.5 C min1. UV–Vis–NIR spectra were recorded in the reflectance mode (1 nm resolution) on a Cary 5 spectrometer (Varian) equipped with an integration sphere, using Teflon as a reference. XANES and EXAFS spectra were recorded in the transmission mode at the Co K edge on the XAS 13 beamline of the DCI storage ring (LURE, Orsay, France). For XANES measurements, a double-crystal Si(311) monochromator was used and the energies were scanned in 0.3 eV steps from 7680 to 7830 eV. For EXAFS measurements, a channel-cut Si(111) monochromator was used, and the energies were scanned in 2 eV steps from 7600 to 8600 eV. The energy was calibrated using a Co metal foil reference. After background correction, the XANES spectra were normalized in the middle of the first EXAFS oscillation. EXAFS analyses were performed in the framework of single-scattering treatments with the package of programmes ‘‘EXAFS pour le Mac’’ [12]. The Fourier transforms (FT) were calculated on w(k)k3v(k), where w(k) is a Kaiser–Bessel window with a smoothness parameter equal ˚ 1. FT are presented to 2.5. The k limits were 2 and 11.7 A without phase correction. A detailed discussion of the single-scattering fits referred to in this paper can be found in Refs. [10,11]. For the sake of clarity, values of N, R and r are recalled in the text whenever necessary.
referred to as ‘‘impregnation’’ by analogy with the classical catalyst preparation mode—and to compare them with the phases formed during a ‘‘dry’’ solid-state procedure. Quantification is attempted by the combined use of macroscopic and local characterization techniques, respectively, TPR and X-ray absorption spectroscopies, which are sensitive to all types of cobalt sites present in the system, unlike techniques such as X-ray diffraction, silent with respect to dispersed species and isolated ions.
Table 1 Cobalt content and physicochemical characteristics of Co2+/ZSM5 catalysts A and I Samples
Co (wt.%)
BET surface area (m2 g1)
Micropore volume (cm3 g1)
H-ZSM5 A1/2 A1 A3/2 I1/2 I1 I3/2
– 1.5 3.5 5.0 1.3 2.3 5.3
356 360 350 360 341 271 299
0.15 0.10 0.10 0.10 0.10 0.09 0.09
3. Results 3.1. UV–Visible spectroscopy Before solid-state reaction in helium, solid mixtures A and I are investigated by UV–Visible spectroscopy at room temperature and exhibit different absorption spectra. In
64
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
I1/2 after drying 526 565
Absorbance (KM) (a.u.)
617
519 468
decrease in the BET surface area. No contribution from mesopores is found. Speciation of cobalt is first studied by analysis of temperature-programmed reduction data. According to previous results from the literature, TPR thermograms of cobalt-containing systems prepared by the solid-state procedure can be divided into three zones of H2 consumption, corresponding each to a reduction to the metallic state:
Cobalt acetate 637
400
450
500 550 600 650 Wavelength (nm)
700
Fig. 1. UV–Visible spectra recorded at room temperature of reference cobalt acetate and sample I1/2 before solid-state reaction.
mixtures A (ground zeolite and cobalt acetate), the absorption in the 400–700 nm region is characteristic of Co2+ in the unaffected precursor salt, i.e. coordinated to CH3COO ions and H2O molecules. In contrast, for all mixtures I (ground zeolite and cobalt acetate, contacted with water and dried), the coordination sphere of cobalt ions has been modified by wetting and drying, leading to absorptions at higher wavelengths (set of bands at 526, 565 and 617 nm), that is at lower energies. Fig. 1 gives the spectrum of sample I1/2 where the position of the central band at 565 nm could arise from ligands which create a lower crystal field than oxygens in [Co(CH3COO)2] Æ 4H2O, possibly oxygens from the zeolite structure. A bathochromic shift corresponding to the grafting of Co2+ in octahedral symmetry on an alumina surface has also been documented by Vakros et al. [13]. According to the classification by Verberckmoes et al. [14], 565 nm rather falls within the region corresponding to six-coordinated Co2+ ions, since the main band for tetrahedral Co2+ in zeolite is usually seen around 590 nm. 3.2. N2 physisorption and temperature-programmed reduction A comparative study by thermogravimetry and mass spectrometry of the transformations induced by the thermal treatments in helium of mixtures A and I does not reveal blatant differences, except that the weight loss steps (release of H2O, CH4 and CO2) are more progressive for samples I. As a consequence, the study will focus on the final catalysts after activation in oxygen. N2 physisorption measurements (Table 1) show that micropore volumes of all the samples are lower than that of the pristine support, which can be accounted for by a partial clogging of the pores. The further decrease in volume measured on samples I1 and I3/2 may be considered as significant, since it is correlated with a corresponding
• below 400 C, reduction of extraframework bulk-like Co3O4 particles with peaks usually intense and narrow [11,15] (extraframework being understood as: located outside the zeolite channels); • between 400 and 700 C, reduction of cobalt ions, reported to be present in the form of small, poorly organized intraframework (CoOx)n oligomers [16] (intraframework being understood as: located inside the zeolite channels); • above 700 C, reduction of extraframework cobalt phyllosilicate [10]. Sample A1/2 contains bulk-like Co3O4 particles (Tred = 360 C) as major H2-consuming species, as well as intraframework (CoOx)n oligomers (Tred = 680 C) (Fig. 2). No contribution from phyllosilicate is observed at high temperature. In contrast, phyllosilicate is the major reducible phase both for sample A1 (Tred = 780 C; Co3O4 as minor phase, Tred = 360 C) and even more for sample A3/2. The low quantity of acetates (sample A1/2) is thus not sufficient to react significantly with the zeolite during helium treatment, with subsequent formation of phyllosilicate as is the case of A1 and A3/2. For samples I, the dominant feature is that there is no fast reduction of bulk-like Co3O4 below 500 C, but only a series of small peaks difficult to link to one single phenomenon. They may be assigned to a multiple-step reduction of dispersed oxide particles, outside the zeolite grains if a low Tred means a low interaction of cobalt with the support. Peaks around 670 C are attributed to the reduction of intraframework (CoOx)n oligomers for I1/2 and I1. I1 does not contain phyllosilicate, unlike A1. The main peak for I3/2 (Tred = 770 C) is attributed to the reduction of cobalt phyllosilicate, formed even without washing between the two thermal treatments. Since the reduction of phyllosilicate has always come out as a symmetrical peak for the other catalysts, the clearly visible shoulder around 670 C is attributed to the reduction of intraframework (CoOx)n oligomers rather than that of another type of phyllosilicate differing by its crystallinity [17]. It is difficult to rely only on TPR for any accurate quantification. Except for bulk-like Co3O4 and cobalt phyllosilicate, the average oxidation state of the phases revealed by TPR is not known and needs confirmation by another technique. Isolated Co2+ ions exchanged in zeolites are reported to be reduced at temperatures above 900 C [3,15,18]; their quantity can be calculated only by difference between the total content in cobalt and the amount of
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
65
340
675
I1/2
H consumption (µmol/(g cat . s)) 2
470
x6
670 680
A1/2
340
405 490
I1
780
770 360 775
A1
I 3/2 0.5
325 360
A3/2
505
300 400 500 600 700 800 900 T (°C)
0.3
425
300 400 500 600 700 800 900 T (°C)
Fig. 2. TPR thermograms of catalysts A and I. Table 2 Quantification of Co species (%) based on analysis of TPR thermograms Samples
Cox+ in Co3O4
Co2+ in (CoOx)n
Isolated Co2+
Co2+ in phyllosilicate
A1/2 A1 A3/2
49 4 1
20 – 14
31 48 15
– 48 70
I1/2 I1 I3/2
15 18 5
23 36 4
62 46 29
– – 62
cobalt quantified on the thermograms—provided that the latter have been quantified accurately. Finally, for samples I1/2 and I1, the reduction peaks partially overlap, making hydrogen consumptions difficult to determine. Table 2 presents the results of quantification obtained by integration of TPR peaks, assuming that • peaks below 500 C are due to the reduction of bulk-like or dispersed extraframework Co3O4 species (reduction stoichiometry H2/Co = 1.33); • the peaks between 500 and 700 C are due to the reduction of intraframework (CoOx)n oligomers containing Co2+ (in which case x = 1); • peaks above 700 C are due to the reduction of extraframework cobalt(II) phyllosilicate; • isolated exchanged Co2+ ions may be quantified by difference with the total cobalt content. 3.3. X-ray absorption spectroscopies: qualitative study X-ray absorption spectroscopies allow deeper investigation into the analysis of cobalt speciation, which is first
considered qualitatively. The literature concerning EXAFS fitting for pure Co3O4 [19–21] evidences large discrepancies in the values of N, sometimes exceeding by 50% the values expected from Co3O4 structure, which was confirmed by our own attempts to obtain satisfactory fits for that compound. Indeed, EXAFS analysis relative to Co3O4 requires four sets of parameters for cobalt nearest and next-nearest neighbours in the oxide (two Co–O and two Co–Co distances) making the overall fitting complex in terms of choice for N and r in each shell. The introduction of extra parameters associated with a second compound adding more complexity and resulting in data difficult to interpret in term of speciation, it was preferred to examine the Fourier transforms on a qualitative ground, on the assumption that the only compounds involved were the compounds identified by TPR. Other compounds such as CoO or Co(OH)2 were rejected on a chemical basis, given that if present initially, they should be transformed into Co3O4 upon thermal treatment in oxygen. Fourier transforms of EXAFS signals are presented in Fig. 3, for samples A and I and for two reference compounds. The first reference is Co3O4, the FT of which exhibiting three distinct peaks from three shells of backscatterers: the first peak includes the two types of oxygen neighbours mentioned above and corresponding to the environment of cobalt ions in octahedral and tetrahedral sites of the oxide; the next two peaks come from two distinct shells of second-neighbours cobalt atoms [22]. The second reference (Co2+/ZSM5 prepared by aqueous ion exchange from cobalt nitrate [10]) corresponds to Co2+ ions in an octahedral environment of six oxygen atoms at ˚ , without any detectable second neighbours, and rep2.07 A resents what will be globally called ‘‘intraframework ions’’,
66
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
Co 3O 4
I 1/2
Co 2+ ions / ZSM5
F(R)
I1 A1/2
A1
I 3/2 A3/2 5 10 0
2
4
6
8
10
R (Å)
0
2
4
6
8
10
R (Å)
Fig. 3. Fourier transforms calculated from experimental EXAFS signals, for references Co3O4 and Co2+ ions/ZSM5, catalysts A and catalysts I.
since no distinction can be made between poorly organized (CoOx)n oligomers or truly isolated Co2+ ions. Samples A1 and A3/2 have been characterized previously by EXAFS and the fits presented in Refs. [10,11]. Analysis allows identification mostly of well-crystallized extraframework cobalt phyllosilicate in A3/2, with a first ˚, shell of six oxygen atoms around Co2+ ions at 2.08 A and a second shell of neighbours with six cobalt and four silicon atoms giving a unique peak on the FT (R = 3.13 ˚ , respectively) [10]. Because of the good agreeand 3.20 A ment with structural parameters and spectroscopic characteristics of bulk phyllosilicates, the EXAFS signal of sample A3/2 is chosen as reference for supported phyllosilicate in our series of catalysts. The two peaks observed in the FT of sample A1 at the same positions as those of sample A3/2 confirm that this sample also contains cobalt phyllosilicate, though poorly crystallized (see for example the lesser intensity of the second peak). N values for O, Co and Si were found to be 5.3, 3.5 and 3.6, respectively [11]. In contrast, the FT obtained for sample A1/2 exhibits the peaks characteristic of Co3O4, confirming the conclusions drawn from TPR. The same analysis can be carried out for sample I1/2, thus corroborating the assumption that the small peaks seen on TPR thermograms correspond to the reduction of Co3O4-like species. For sample I1, the peaks characteristic of the second shell of Co3O4 seem to be present but are less intense. Because of the lack of high-temperature peaks in TPR for these three samples, the presence of cobalt silicate seems unlikely. We will
consider that they all contain both intraframework species, as well as extraframework Co3O4 particles. Finally, the FT of I3/2 is the only one from series I to exhibit the second shell peak of cobalt phyllosilicate, in agreement with TPR. It has a low intensity compared with the peak of the first shell, which shows that the contribution from intraframework ions is high. 3.4. X-ray absorption spectroscopies: quantitative study Fig. 4 presents the experimental XANES absorption spectra of sample A1/2 and samples from series I, along with: (i) spectra recorded for reference compounds; (ii) calculated simulations (sim) based on a linear combination of the reference spectra (a procedure formerly applied to surface species, for instance in Refs. [23–25]). As mentioned above, sample A3/2 is used as a reference for supported cobalt phyllosilicate. It can be added that the XANES spectrum of A1 (not shown) is identical to that of A3/2, with a white line only 3% less intense. A decrease in the preedge intensity and a shift toward lower energies are noted along the series A1/2, I1/2, I1 and I3/2. They are due to lower proportions of ions in tetrahedral symmetry and of Co3+, respectively, which is related to a decreasing content in Co3O4. This complements TPR quantification, in which the quantification of Co3O4 for samples I was difficult to perform accurately. If one compares the XANES spectra of A1/2, I1/2 and I1 with the experimental spectra of Co3O4 and intraframework ions, one can note the presence of isosbestic points,
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
x
2
67
2
x
(1) Normalized intensity
Normalized intensity
x
x
x
1
(1)
x
x
x
(2)
(1)
(2)
x
x
x
x
1
x
(2)
(2)
(1)
I1/2
A1/2 0 7700
7720
7740 7760 Energy (eV)
7780
0 7700
7800
7720
7740 7760 Energy (eV)
7780
7800
+
2
x
(1)
(3)
x
x
x
1
x
(2)
(1)
x +
+ x
x
Normalized intensity
Normalized intensity
2
(2)
1
xx
+ x
(3)
(2)
(2)
I1 0 7700
7720
7740 7760 Energy (eV)
7780
x
+
I3/2 0 7700
7800
7720
7740 7760 Energy (eV)
7780
7800
Fig. 4. XANES spectra of samples A1/2 and I, and their simulations by linear combination from reference spectra. Experimental spectra are presented in full lines and simulations based on the values given in Table 2 in dotted lines. Reference spectra are numbered as follows: (1) Co2+ ions/ZSM5 (s); (2) Co3O4 (·); (3) cobalt phyllosilicate (A3/2) (+).
It can be observed that linear combinations of reference EXAFS signals [26] using the proportions found by study of the XANES spectra give a reasonable simulation of A1/2 and I1/2 experimental signals (Fig. 5). However, there is no evidence that the phases present in the catalysts have exactly the same characteristics as the reference compounds in terms of structural organization. Such comparison should thus be taken with caution and simulations are not presented for I1 and I3/2, for which the choice of references is not straightforward, due to the incertitudes on
which shows that a quantitative analysis based only on these two compounds is realistic. I3/2 contains at least one more compound, cobalt phyllosilicate, which accounts for the absence of clear isosbestic points. Table 3 sums up the respective contribution of the different components in each linear combination, which has been checked not to depend on the choice of the normalization point. To those results are added the assumed compositions of A1 and A3/2 based on EXAFS analysis, as described in the previous section. Table 3 Quantification of Co species (%) as determined by XAS Samples
Cox+ in Co3O4
Intraframework ions (Co2+ in (CoOx)n + isolated Co2+)
Co2+ in phyllosilicate
A1/2 A1
XANES EXAFS
52 –
48 –
A3/2
EXAFS
–
–
– 100 (low crystallinity?) 100
I1/2 I1 I3/2
XANES XANES XANES
33 25 25
67 75 –
– – 75
68
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
Co 3O 4
k*khi(k)
Co 2+ ions / ZSM5
I 1/2 I 1/2 sim
A1/2 A1/2 sim
1
4
6
8 k (Å -1)
10
12
Fig. 5. EXAFS signals of reference compounds and samples A1/2 and I1/ 2, and their simulations by linear combination from reference spectra. Simulations (‘‘sim’’) correspond to the proportions mentioned in Table 3.
Co3O4 crystallinity for the former, and on the precision of a XANES simulation based on two components only for the latter. 4. Discussion Quantification of species both by TPR and X-ray absorption spectroscopies brings complementary results, which is necessary because each technique is more sensitive to some types of species or to some characteristics, as already stressed in the case of Fe/ZSM5 systems [5,27]. XAS make possible the identification of the compounds through their structural characteristics and the determination of their oxidation state. TPR allows ranking the supported compounds, according to the decreasing proximity of cobalt ions and to their increasing interaction with the support along the series: • weakly interacting oxidic condensed phases; • oligomeric species of smaller nuclearity located inside the zeolite framework; • cobalt ions in neighbouring octahedral sites, embedded in a silicic extraframework matrix; • and isolated exchanged ions (absent from TPR thermograms). The combination of TPR and XAS leads to a quantification of extra- and intraframework species, further discrim-
inated upon their dispersion and/or nuclearity. It must be stressed though that there is no consensus in the literature on whether oligomeric species are located inside or outside the zeolite channels. We have followed the interpretations by da Cruz et al. [16], recalled by Wang et al. [3] and supported by results according to which oligomeric species containing 4–6 metal ions are stable inside the framework [28,29]. But it should be mentioned that Li et al. consider the TPR intermediate peak as characteristic of extraframework (CoOx)n oligomers [15], while Tang et al. consider temperature peaks at the lowest temperatures as characteristic of intraframework organized oxidic species [30]. There is an overall good agreement between the results presented in Tables 2 and 3, which will be discussed in terms of sensitivity of the techniques, since it is still difficult to derive a strategy for precise error calculations for these six complex systems. First, the XAS signals related to phyllosilicate are sensitive to the compound crystallinity, as shown by the study of A1 and A3/2. The fact that XAS overestimate the contribution of phyllosilicate (A3/2) comes from the uncertainties inherent to EXAFS analysis, for which a minor phase such as a small proportion of isolated ions cannot be discriminated against a major phase and only acts as perturbation through higher values of the Debye–Waller parameters (r = 0.10 for A1 and A3/2, compared with 0.08 for the simulation of oxygens first shell around isolated Co2+ ions [10,11]). This may explain the discrepancies between the results given by TPR and XAS for A1 and I3/ 2, with TPR appearing as more precise for the contribution of Co3O4, even when present in very small quantities. For samples containing phyllosilicate, TPR is more reliable and should be preferred. The second point to be discussed is the proportion of Cox+ in Co3O4 found by TPR for I1/2. This proportion is much lower than that found by XAS, which is probably the more reliable technique, since a separate integration of the peaks corresponding to dispersed Co3O4 and (CoOx)n oligomers is difficult to carry out because of peak overlap. Besides, it is not sure that (CoOx)n oligomers contains only Co2+ ions as supposed in the TPR section. Sample I1/2 is a typical example for which drawing a clear limit between dispersed extraframework Co3O4 entity and an intraframework (CoOx)n group would not be straightforward. Fig. 6a represents the quantity of intraframework and extraframework species per gram of catalyst as a function of cobalt loading, and Fig. 6b and c the quantities of each class of compound for the two series of catalysts. Quantification for series A and I3/2 is based on TPR results, because XAS lacks precision when phyllosilicate is present; moreover, the agreement between XAS and TPR is found to be good for A1/2. For I1/2 and I1, quantification is based on XAS results as far as the ratio oxide/intraframework ions is concerned, but since this technique is silent on the proportions of (CoOx)n oligomers and isolated Co2+ ions, the quantities of these species are calculated according to their relative proportions found by TPR. It was checked
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
600
Co (µmol.g -1 cat)
Cox+ in oxide (CoOx)n isolated Co2+ Co2+ in phyllosilicate
extrafr. A intrafr. A extrafr. I intrafr. I
700
69
Cox+ in oxide (CoOx)n isolated Co2+ Co2+ in phyllosilicate
500 400 300 200 100 0 1
(a)
2
3 4 Co wt%
5
6
1 (b)
2
3 4 Co wt%
5
6
1 (c)
2
3 4 Co wt%
5
6
Fig. 6. (a) Overall quantification of extra- and intraframework cobalt species as a function of Co weight content. (b) Detailed quantification of cobalt species as a function of Co weight content in catalysts A. (c) Detailed quantification of cobalt species as a function of Co weight content in catalysts I.
that a calculation based on TPR only would not change the trends described below. The quantity of intraframework cobalt is higher in series I than in series A, which is consistent with porosity measurements. A plateau seems to be reached for I around 2 wt.% of cobalt, showing that cobalt added in excess cannot be introduced into the zeolite. Moreover, it may be extrapolated that, for series I, intraframework cobalt exceeds extraframework cobalt up to a higher loading compared with series A, for which phyllosilicate formation is easier. Cobalt oxide, present in a more dispersed form in series I, represents less than 130 lmol Co g1 of catalyst in all samples, the quantity of cobalt in Co3O4 decreasing with increasing Co content. Finally, (CoOx)n oligomers seem to be present in higher quantity in series I than in series A, indicating that water helps the migration of cobalt into the porosity and the formation of smaller entities, but cannot guarantee that intraframework species will be isolated ions. These results show that the stage at which water is introduced is most important. Samples A and I can be compared along two points: • When water is added to the solid mixture prior to solidstate reaction, the subsequent formation of bulky Co3O4 particles is inhibited, as observed by Park et al. [6]. For A1/2, cobalt remains as oxide outside zeolite which has not been much damaged by the attack of acetates. In contrast, I1/2, like all samples I, contain smaller Co3O4 particles and entities located inside the zeolite porosity. • Though neither in dry nor wet processes, chemical species—in particular acetates—have been eliminated before heating, the formation of phyllosilicate occurs for a higher acetate/zeolite ratio in the wet process. When the contact with water has occurred prior to solid-state reaction, acetate anions are not chemically aggressive toward the zeolite surface, in contrast to what is observed in the solid-state procedure.
persed form. This can be the consequence of a partial dissolution of cobalt acetate in water. Dissolution is followed by dissociation of the complex, because cobalt acetate has a low global stability constant in water (b = 1–2 in the 20–50 C range [31]): 2þ
½CoðCH3 COOÞ2 4H2 Oaq þ 2H2 O ¢ ½CoðH2 OÞ6 aq þ 2CH3 COO aq
and UV–Visible spectroscopy shows that after drying, some cobalt ions have been able to interact with the zeolite. On the other hand, dissociated acetates which diffuse in the zeolite channels are allowed to react with inner zeolitic protons to which they have now access through the solvent: þ CH3 COO aq þ Hzeo ! CH3 COOHaq
It can be assumed that acetic acid CH3COOH remaining after drying decomposes easily into CH4 and CO2 during thermal treatment in helium, without needing any reaction with zeolite. Similarly, the thermal decomposition of iron acetate during solid-state exchange has been shown to be favored in the presence of water [32]. In contrast, unprotonated CH3COO ions must react with the zeolite surface during heating to catch protons, which seems to contribute to dealumination eventually leading to the phyllosilicate phase [10,11]. The more CH3COO ions are left unprotonated after wetting and drying, the more dealumination is expected later. The mass action law predicts that when the cobalt complex concentration increases in the wetting solution (which is the case along the series I1/2, I1 and I3/2), its degree of dissociation into cobalt ion and CH3COO decreases. A higher quantity of acetate ions is thus left unprotonated after drying on I3/2 and their release upon further heating may explain the formation of phyllosilicate in this sample, though not on catalysts prepared with a lower cobalt concentration. 5. Conclusions
Water thus promotes the transport of the cobalt salt inside the zeolite before thermal treatment, with the fraction remaining outside the framework appearing in a dis-
The effect of water on cobalt speciation during the preparation of Co2+/ZSM5 catalysts by solid-state reaction
70
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
between cobalt acetate and H-ZSM5 has been investigated quantitatively by the combined use of TPR and XAS. The introduction of water before solid-state reaction occurs has several effects on cobalt speciation in the final catalyst, compared with a dry procedure. At low cobalt content, water favors the formation of intraframework phases, such as (CoOx)n oligomers and isolated Co2+ ions, by facilitating the dissociation of cobalt acetate and the transport of cobalt ions inside the zeolite porosity; the dispersion of extraframework Co3O4 particles may also be increased for the same reason. The formation of extraframework cobalt phyllosilicate linked to the presence of acetates is inhibited, possibly as a consequence of the decomposition upon heating of acetic acid produced by reaction of acetates with zeolite protons during wetting. Finally, for high cobalt content, the use of a large excess of cobalt acetate above the stoichiometry of exchange does not help in reaching a higher deposition of cobalt inside the zeolite, most of the ions in excess contributing to the formation of extraframework phases, mostly phyllosilicate. Acknowledgments This work received the financial support of the FrancoTunisian CMCU (action 00F1205) which is gratefully acknowledged. Prof. Franc¸ois Bozon-Verduraz (ITODYS, Universite´ Denis Diderot, Paris) is thanked for providing access to the UV–Vis–NIR spectrometer. References [1] H.G. Karge, H.K. Beyer, in: P.A. Jacobs (Ed.), Zeolite Chemistry and Catalysis, Studies in Surface Science and Catalysis, vol. 69, Elsevier, Amsterdam, 1991, p. 43. [2] A.V. Kucherov, A.A. Slinkin, J. Mol. Catal. 90 (1994) 323. [3] X. Wang, H.Y. Chen, W.M.H. Sachtler, Appl. Catal. B 26 (2000) L227. [4] V. Schwartz, R. Prins, X. Wang, W.M.H. Sachtler, J. Phys. Chem. B 106 (2002) 7210. [5] F. Heinrich, C. Schmidt, E. Lo¨ffler, M. Menzel, W. Gru¨nert, J. Catal. 212 (2002) 157. [6] J.H. Park, C.H. Park, I.S. Nam, Appl. Catal. A 277 (2004) 271. [7] Z. Sobalı´k, A.A. Belhekar, Z. Tvaruzkova, B. Wichterlova´, Appl. Catal. A 188 (1999) 175. [8] H. Ben Boubaker, S. Fessi, A. Ghorbel, E. Marceau, M. Che, in: E. van Steen, L.H. Callanan, M. Claeys (Eds.), Recent Advances in the
[9]
[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]
[26] [27]
[28] [29] [30] [31] [32]
Science and Technology of Zeolites and Related Materials, Studies in Surface Science and Catalysis, vol. 154, Elsevier, Amsterdam, 2004, p. 2655, Part C. M. Mhamdi, S. Khaddar-Zine, A. Ghorbel, E. Marceau, M. Che, Y. Ben Taarit, F. Villain, in: J. Cejka, N. Zilkova´, P. Nachtigall (Eds.), Molecular Sieves: from Basic Research to Industrial Applications, Studies in Surface Science and Catalysis, vol. 158, Elsevier, Amsterdam, 2005, p. 748. M. Mhamdi, E. Marceau, S. Khaddar-Zine, A. Ghorbel, M. Che, Y. Ben Taarit, F. Villain, Catal. Lett. 98 (2004) 135. M. Mhamdi, E. Marceau, S. Khaddar-Zine, A. Ghorbel, M. Che, Y. Ben Taarit, F. Villain, Z. Phys. Chem. 219 (2005) 963. A. Michalowicz, EXAFS pour le Mac, Logiciels pour la Chimie, Soc. Fr. Chim., Paris, 1991, p. 102. J. Vakros, K. Bourikas, S. Perlepes, C. Kordulis, A. Lycourghiotis, Langmuir 20 (2004) 10542. A.A. Verberckmoes, B.M. Weckhuysen, R.A. Schoonheydt, Micropor. Mesopor. Mater. 22 (1998) 165. W. Li, S.Y. Yu, G.D. Meitzner, E. Iglesia, J. Phys. Chem. B 105 (2001) 1176. R.S. da Cruz, A.J.S. Mascarenhas, H.M.C. Andrade, Appl. Catal. B 18 (1998) 223. J.Y. Carriat, M. Che, M. Kermarec, A. Decarreau, Catal. Lett. 25 (1994) 127. J.D. Han, S.I. Woo, J. Chem. Soc., Faraday Trans. 88 (1992) 145. A.Y. Khodakov, J. Lynch, D. Bazin, B. Rebours, N. Zanier, B. Moisson, P. Chaumette, J. Catal. 168 (1997) 16. G.C. Shen, T. Shido, M. Ichikawa, J. Phys. Chem. 100 (1996) 16947. M. Shirai, K. Asakura, Y. Iwasawa, Catal. Lett. 15 (1992) 247. G.P. Huffman, N. Shah, J. Zhao, F.E. Huggins, T.E. Hoost, S. Halvorsen, J.G. Goodwin Jr., J. Catal. 151 (1995) 17. D.T. On, S. Kaliaguine, L. Bonneviot, J. Catal. 157 (1995) 235. G. Jacobs, P.M. Patterson, Y. Zhang, T. Das, J. Li, B.H. Davis, Appl. Catal. A 233 (2002) 215. F. Morales, F.M.F. de Groot, P. Glatzel, E. Kleimenov, H. Bluhm, M. Ha¨vecker, A. Knop-Gericke, B.M. Weckhuysen, J. Phys. Chem. B 108 (2004) 16201. G. Redden, J. Bargar, R. Bencheikh-Latmani, J. Colloid Interf. Sci. 244 (2001) 211. L. Capek, V. Kreibich, J. Dedecek, T. Grygar, B. Wichterlova´, Z. Sobalı´k, J.A. Martens, R. Brosius, V. Tokarova´, Micropor. Mesopor. Mater. 80 (2005) 279. A. Jentys, K. Kleestorfer, H. Vinek, Catal. Lett. 39 (1996) 119. R. Joyner, M. Stockenhuber, J. Phys. Chem. B 103 (1999) 5963. Q. Tang, Q. Zhang, P. Wang, Y. Wang, H. Wan, Chem. Mater. 16 (2004) 1967. P.H. Tedesco, V.B. de Rumi, J. Inorg. Nucl. Chem. 33 (1971) 969. K. La´za´r, G. Pa´l-Borbe´ly, H.K. Beyer, H.G. Karge, in: G. Poncelet, P.A. Jacobs, P. Grange, B. Delmon, J. Martens (Eds.), Preparation of Catalysts VI: Scientific Bases for the Preparation of Catalysts, Studies in Surface Science and Catalysis, vol. 91, Elsevier, Amsterdam, 1995, p. 551.
Effect of water on cobalt speciation during solid-state synthesis of Co2+/ZSM5 catalysts: Quantitative study by TPR and XAS Hanene Ben Boubaker a, Mourad Mhamdi a, Eric Marceau b,*, Sihem Khaddar-Zine a, Abdelhamid Ghorbel a, Michel Che b,c, Youne`s Ben Taarit d, Franc¸oise Villain e,f a Laboratoire de Chimie des Mate´riaux et Catalyse, Universite´ Tunis II-El Manar, 1060 Tunis, Tunisia Laboratoire de Re´activite´ de Surface (UMR 7609 CNRS), Universite´ Pierre et Marie Curie, 4 place Jussieu, 75252 Paris Cedex 05, France c Institut Universitaire de France, France d Institut de Recherches sur la Catalyse (UPR 5401 CNRS), 2 Avenue Albert Einstein, 69626 Villeurbanne Cedex, France e Laboratoire de Chimie Inorganique et Mate´riaux Mole´culaires (UMR 7071 CNRS), Universite´ Pierre et Marie Curie, 4 place Jussieu, 75252 Paris Cedex 05, France Laboratoire pour l’Utilisation du Rayonnement Electromagne´tique (LURE), Centre Universitaire Paris-Sud, BP 34, 91898 Orsay Cedex, France b
f
Received 12 December 2005; received in revised form 28 January 2006; accepted 1 February 2006 Available online 20 March 2006
Abstract The effect of water on cobalt speciation in the synthesis of Co2+/ZSM5 catalysts is quantitatively studied by the combined use of macroscopic and local characterization techniques, TPR and X-ray absorption spectroscopies, respectively. The addition of a small quantity of water to cobalt acetate and zeolite during the preparation of Co2+/ZSM5 by solid-state reaction favors cobalt dispersion and migration into the zeolite. It also inhibits the formation of extraframework phases (Co3O4 particles, cobalt phyllosilicate). Dissociation of solid cobalt acetate upon contact with water is thought to be the driving force for enhanced cobalt migration into the zeolite. 2006 Elsevier Inc. All rights reserved. Keywords: Solid-state reaction; TPR; XAS; Speciation; Quantitative analysis
1. Introduction Among the various processes used to introduce metal ions into zeolites, solid-state reactions between a protonic zeolite and a solid precursor salt have gained interest for two reasons: they avoid the handling of large quantities of solutions and there is no waste of precursor in contrast to the method of aqueous ion exchange at equilibrium [1,2]. However, a non-negligible fraction of metal ions deposited during solid-state reaction do not exchange with protons, especially when the amount of precursor salt exceeds the stoichiometry of exchange: they react instead with zeolite to form extraframework phases. It becomes thus important
*
Corresponding author. Tel.: +33 1 44 27 60 04; fax: +33 1 44 27 60 33. E-mail address: [email protected] (E. Marceau).
1387-1811/$ - see front matter 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2006.02.004
to characterize such phases which have a different reactivity from that of exchanged ions, as shown for example in the cases of NOx reduction [3–6] and hydrocarbon ammoxidation [7–9]. These phases are usually oxidic particles, but it has been shown recently that the use of solid cobalt acetate to prepare Co2+/ZSM5 systems may lead to another type of phase, a well-crystallized phyllosilicate incorporating in its bulk the majority of cobalt ions, after dealumination of the zeolite by acetate anions during thermal treatment [10]. In order to modify this anion-support interaction leading to a result detrimental to catalysis [9], one can imagine to add another chemical partner, water, in a limited amount, as a way to change the type of contact between precursor salt and support before thermal treatments be involved. The purpose of this paper is to identify the phases formed during this ‘‘wet’’ solid-state procedure—hereafter
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
63
2. Experimental
110 C overnight. The resulting solid was heated in a helium flow (25 mL min1) up to 500 C (heating rate: 2 C min1) and kept at 500 C overnight (12 h) before being submitted to the oxygen activation treatment described above. It can be stressed here that during the preparation of catalysts I, the sample is contacted with water before the first thermal treatment, that is before reaction between zeolite and cobalt acetate; whereas the contact with water occurs after the first thermal treatment in series A, that is after the reaction between zeolite and cobalt acetate has taken place.
2.1. Catalysts preparation
2.2. Characterization
Two series of three samples were prepared, one by solidstate reaction and the other one by impregnation (Table 1). For the sake of homogeneity regarding earlier results [11], samples prepared by solid-state reactions are referred to as Ax, following the nominal Co/Al atomic ratio, x = 1/2, 1 or 3/2 (i.e. corresponding to one, two or three times the exchange stoichiometry). Samples prepared by impregnation are referred to as Ix. It must be noted that the purpose of the synthesis of samples from series I is primarily characterization, while reaching the exact nominal ratio x is not sought for; consequently, x in the denomination Ix was chosen mostly by analogy with the A series with increasing Co/Al content. Catalysts A were prepared by solid-state reaction between H-ZSM5 zeolite (Zeocat, Si/Al = 26, Al wt.% = 1.6, water content: 5 wt.%, pore volume: 0.54 mL g1) and cobalt(II) acetate [Co(CH3COO)2] Æ 4H2O (Prolabo). Powders were finely ground and mixed in a mortar for 15 min in ambient conditions. The resulting mixture was heated in a helium flow (25 mL min1) up to 500 C (heating rate: 2 C min1) and kept at 500 C overnight (12 h). After having been washed twice with deionized water and dried overnight at 110 C in a static oven, the powder was submitted to a final activation treatment in O2, up to 500 C (25 mL min1, heating rate: 5 C min1) and 1 h at that temperature. In the preparation procedure of catalysts I, a volume of deionized water (1.5 mL g1 of zeolite) was added to the mixture of ground powders (zeolite and precursor). The whole system was pugged for 10 min and dried in air at
Chemical analyses were performed by ICP at the Vernaison Center of Chemical Analysis of CNRS. Specific surface areas were determined by the BET method from nitrogen physisorption at 196 C, using an automatic Micromeritics ASAP 2000 instrument. Micropore volumes were assessed by application of the Dubinin equation between 0.01 < P/P0 < 0.1. Temperature-programmed reductions (TPR) were performed under 5% H2 in argon (25 mL min1); the hydrogen consumption was measured using a thermal conductivity detector, from room temperature to 1000 C with a heating rate of 7.5 C min1. UV–Vis–NIR spectra were recorded in the reflectance mode (1 nm resolution) on a Cary 5 spectrometer (Varian) equipped with an integration sphere, using Teflon as a reference. XANES and EXAFS spectra were recorded in the transmission mode at the Co K edge on the XAS 13 beamline of the DCI storage ring (LURE, Orsay, France). For XANES measurements, a double-crystal Si(311) monochromator was used and the energies were scanned in 0.3 eV steps from 7680 to 7830 eV. For EXAFS measurements, a channel-cut Si(111) monochromator was used, and the energies were scanned in 2 eV steps from 7600 to 8600 eV. The energy was calibrated using a Co metal foil reference. After background correction, the XANES spectra were normalized in the middle of the first EXAFS oscillation. EXAFS analyses were performed in the framework of single-scattering treatments with the package of programmes ‘‘EXAFS pour le Mac’’ [12]. The Fourier transforms (FT) were calculated on w(k)k3v(k), where w(k) is a Kaiser–Bessel window with a smoothness parameter equal ˚ 1. FT are presented to 2.5. The k limits were 2 and 11.7 A without phase correction. A detailed discussion of the single-scattering fits referred to in this paper can be found in Refs. [10,11]. For the sake of clarity, values of N, R and r are recalled in the text whenever necessary.
referred to as ‘‘impregnation’’ by analogy with the classical catalyst preparation mode—and to compare them with the phases formed during a ‘‘dry’’ solid-state procedure. Quantification is attempted by the combined use of macroscopic and local characterization techniques, respectively, TPR and X-ray absorption spectroscopies, which are sensitive to all types of cobalt sites present in the system, unlike techniques such as X-ray diffraction, silent with respect to dispersed species and isolated ions.
Table 1 Cobalt content and physicochemical characteristics of Co2+/ZSM5 catalysts A and I Samples
Co (wt.%)
BET surface area (m2 g1)
Micropore volume (cm3 g1)
H-ZSM5 A1/2 A1 A3/2 I1/2 I1 I3/2
– 1.5 3.5 5.0 1.3 2.3 5.3
356 360 350 360 341 271 299
0.15 0.10 0.10 0.10 0.10 0.09 0.09
3. Results 3.1. UV–Visible spectroscopy Before solid-state reaction in helium, solid mixtures A and I are investigated by UV–Visible spectroscopy at room temperature and exhibit different absorption spectra. In
64
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
I1/2 after drying 526 565
Absorbance (KM) (a.u.)
617
519 468
decrease in the BET surface area. No contribution from mesopores is found. Speciation of cobalt is first studied by analysis of temperature-programmed reduction data. According to previous results from the literature, TPR thermograms of cobalt-containing systems prepared by the solid-state procedure can be divided into three zones of H2 consumption, corresponding each to a reduction to the metallic state:
Cobalt acetate 637
400
450
500 550 600 650 Wavelength (nm)
700
Fig. 1. UV–Visible spectra recorded at room temperature of reference cobalt acetate and sample I1/2 before solid-state reaction.
mixtures A (ground zeolite and cobalt acetate), the absorption in the 400–700 nm region is characteristic of Co2+ in the unaffected precursor salt, i.e. coordinated to CH3COO ions and H2O molecules. In contrast, for all mixtures I (ground zeolite and cobalt acetate, contacted with water and dried), the coordination sphere of cobalt ions has been modified by wetting and drying, leading to absorptions at higher wavelengths (set of bands at 526, 565 and 617 nm), that is at lower energies. Fig. 1 gives the spectrum of sample I1/2 where the position of the central band at 565 nm could arise from ligands which create a lower crystal field than oxygens in [Co(CH3COO)2] Æ 4H2O, possibly oxygens from the zeolite structure. A bathochromic shift corresponding to the grafting of Co2+ in octahedral symmetry on an alumina surface has also been documented by Vakros et al. [13]. According to the classification by Verberckmoes et al. [14], 565 nm rather falls within the region corresponding to six-coordinated Co2+ ions, since the main band for tetrahedral Co2+ in zeolite is usually seen around 590 nm. 3.2. N2 physisorption and temperature-programmed reduction A comparative study by thermogravimetry and mass spectrometry of the transformations induced by the thermal treatments in helium of mixtures A and I does not reveal blatant differences, except that the weight loss steps (release of H2O, CH4 and CO2) are more progressive for samples I. As a consequence, the study will focus on the final catalysts after activation in oxygen. N2 physisorption measurements (Table 1) show that micropore volumes of all the samples are lower than that of the pristine support, which can be accounted for by a partial clogging of the pores. The further decrease in volume measured on samples I1 and I3/2 may be considered as significant, since it is correlated with a corresponding
• below 400 C, reduction of extraframework bulk-like Co3O4 particles with peaks usually intense and narrow [11,15] (extraframework being understood as: located outside the zeolite channels); • between 400 and 700 C, reduction of cobalt ions, reported to be present in the form of small, poorly organized intraframework (CoOx)n oligomers [16] (intraframework being understood as: located inside the zeolite channels); • above 700 C, reduction of extraframework cobalt phyllosilicate [10]. Sample A1/2 contains bulk-like Co3O4 particles (Tred = 360 C) as major H2-consuming species, as well as intraframework (CoOx)n oligomers (Tred = 680 C) (Fig. 2). No contribution from phyllosilicate is observed at high temperature. In contrast, phyllosilicate is the major reducible phase both for sample A1 (Tred = 780 C; Co3O4 as minor phase, Tred = 360 C) and even more for sample A3/2. The low quantity of acetates (sample A1/2) is thus not sufficient to react significantly with the zeolite during helium treatment, with subsequent formation of phyllosilicate as is the case of A1 and A3/2. For samples I, the dominant feature is that there is no fast reduction of bulk-like Co3O4 below 500 C, but only a series of small peaks difficult to link to one single phenomenon. They may be assigned to a multiple-step reduction of dispersed oxide particles, outside the zeolite grains if a low Tred means a low interaction of cobalt with the support. Peaks around 670 C are attributed to the reduction of intraframework (CoOx)n oligomers for I1/2 and I1. I1 does not contain phyllosilicate, unlike A1. The main peak for I3/2 (Tred = 770 C) is attributed to the reduction of cobalt phyllosilicate, formed even without washing between the two thermal treatments. Since the reduction of phyllosilicate has always come out as a symmetrical peak for the other catalysts, the clearly visible shoulder around 670 C is attributed to the reduction of intraframework (CoOx)n oligomers rather than that of another type of phyllosilicate differing by its crystallinity [17]. It is difficult to rely only on TPR for any accurate quantification. Except for bulk-like Co3O4 and cobalt phyllosilicate, the average oxidation state of the phases revealed by TPR is not known and needs confirmation by another technique. Isolated Co2+ ions exchanged in zeolites are reported to be reduced at temperatures above 900 C [3,15,18]; their quantity can be calculated only by difference between the total content in cobalt and the amount of
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
65
340
675
I1/2
H consumption (µmol/(g cat . s)) 2
470
x6
670 680
A1/2
340
405 490
I1
780
770 360 775
A1
I 3/2 0.5
325 360
A3/2
505
300 400 500 600 700 800 900 T (°C)
0.3
425
300 400 500 600 700 800 900 T (°C)
Fig. 2. TPR thermograms of catalysts A and I. Table 2 Quantification of Co species (%) based on analysis of TPR thermograms Samples
Cox+ in Co3O4
Co2+ in (CoOx)n
Isolated Co2+
Co2+ in phyllosilicate
A1/2 A1 A3/2
49 4 1
20 – 14
31 48 15
– 48 70
I1/2 I1 I3/2
15 18 5
23 36 4
62 46 29
– – 62
cobalt quantified on the thermograms—provided that the latter have been quantified accurately. Finally, for samples I1/2 and I1, the reduction peaks partially overlap, making hydrogen consumptions difficult to determine. Table 2 presents the results of quantification obtained by integration of TPR peaks, assuming that • peaks below 500 C are due to the reduction of bulk-like or dispersed extraframework Co3O4 species (reduction stoichiometry H2/Co = 1.33); • the peaks between 500 and 700 C are due to the reduction of intraframework (CoOx)n oligomers containing Co2+ (in which case x = 1); • peaks above 700 C are due to the reduction of extraframework cobalt(II) phyllosilicate; • isolated exchanged Co2+ ions may be quantified by difference with the total cobalt content. 3.3. X-ray absorption spectroscopies: qualitative study X-ray absorption spectroscopies allow deeper investigation into the analysis of cobalt speciation, which is first
considered qualitatively. The literature concerning EXAFS fitting for pure Co3O4 [19–21] evidences large discrepancies in the values of N, sometimes exceeding by 50% the values expected from Co3O4 structure, which was confirmed by our own attempts to obtain satisfactory fits for that compound. Indeed, EXAFS analysis relative to Co3O4 requires four sets of parameters for cobalt nearest and next-nearest neighbours in the oxide (two Co–O and two Co–Co distances) making the overall fitting complex in terms of choice for N and r in each shell. The introduction of extra parameters associated with a second compound adding more complexity and resulting in data difficult to interpret in term of speciation, it was preferred to examine the Fourier transforms on a qualitative ground, on the assumption that the only compounds involved were the compounds identified by TPR. Other compounds such as CoO or Co(OH)2 were rejected on a chemical basis, given that if present initially, they should be transformed into Co3O4 upon thermal treatment in oxygen. Fourier transforms of EXAFS signals are presented in Fig. 3, for samples A and I and for two reference compounds. The first reference is Co3O4, the FT of which exhibiting three distinct peaks from three shells of backscatterers: the first peak includes the two types of oxygen neighbours mentioned above and corresponding to the environment of cobalt ions in octahedral and tetrahedral sites of the oxide; the next two peaks come from two distinct shells of second-neighbours cobalt atoms [22]. The second reference (Co2+/ZSM5 prepared by aqueous ion exchange from cobalt nitrate [10]) corresponds to Co2+ ions in an octahedral environment of six oxygen atoms at ˚ , without any detectable second neighbours, and rep2.07 A resents what will be globally called ‘‘intraframework ions’’,
66
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
Co 3O 4
I 1/2
Co 2+ ions / ZSM5
F(R)
I1 A1/2
A1
I 3/2 A3/2 5 10 0
2
4
6
8
10
R (Å)
0
2
4
6
8
10
R (Å)
Fig. 3. Fourier transforms calculated from experimental EXAFS signals, for references Co3O4 and Co2+ ions/ZSM5, catalysts A and catalysts I.
since no distinction can be made between poorly organized (CoOx)n oligomers or truly isolated Co2+ ions. Samples A1 and A3/2 have been characterized previously by EXAFS and the fits presented in Refs. [10,11]. Analysis allows identification mostly of well-crystallized extraframework cobalt phyllosilicate in A3/2, with a first ˚, shell of six oxygen atoms around Co2+ ions at 2.08 A and a second shell of neighbours with six cobalt and four silicon atoms giving a unique peak on the FT (R = 3.13 ˚ , respectively) [10]. Because of the good agreeand 3.20 A ment with structural parameters and spectroscopic characteristics of bulk phyllosilicates, the EXAFS signal of sample A3/2 is chosen as reference for supported phyllosilicate in our series of catalysts. The two peaks observed in the FT of sample A1 at the same positions as those of sample A3/2 confirm that this sample also contains cobalt phyllosilicate, though poorly crystallized (see for example the lesser intensity of the second peak). N values for O, Co and Si were found to be 5.3, 3.5 and 3.6, respectively [11]. In contrast, the FT obtained for sample A1/2 exhibits the peaks characteristic of Co3O4, confirming the conclusions drawn from TPR. The same analysis can be carried out for sample I1/2, thus corroborating the assumption that the small peaks seen on TPR thermograms correspond to the reduction of Co3O4-like species. For sample I1, the peaks characteristic of the second shell of Co3O4 seem to be present but are less intense. Because of the lack of high-temperature peaks in TPR for these three samples, the presence of cobalt silicate seems unlikely. We will
consider that they all contain both intraframework species, as well as extraframework Co3O4 particles. Finally, the FT of I3/2 is the only one from series I to exhibit the second shell peak of cobalt phyllosilicate, in agreement with TPR. It has a low intensity compared with the peak of the first shell, which shows that the contribution from intraframework ions is high. 3.4. X-ray absorption spectroscopies: quantitative study Fig. 4 presents the experimental XANES absorption spectra of sample A1/2 and samples from series I, along with: (i) spectra recorded for reference compounds; (ii) calculated simulations (sim) based on a linear combination of the reference spectra (a procedure formerly applied to surface species, for instance in Refs. [23–25]). As mentioned above, sample A3/2 is used as a reference for supported cobalt phyllosilicate. It can be added that the XANES spectrum of A1 (not shown) is identical to that of A3/2, with a white line only 3% less intense. A decrease in the preedge intensity and a shift toward lower energies are noted along the series A1/2, I1/2, I1 and I3/2. They are due to lower proportions of ions in tetrahedral symmetry and of Co3+, respectively, which is related to a decreasing content in Co3O4. This complements TPR quantification, in which the quantification of Co3O4 for samples I was difficult to perform accurately. If one compares the XANES spectra of A1/2, I1/2 and I1 with the experimental spectra of Co3O4 and intraframework ions, one can note the presence of isosbestic points,
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
x
2
67
2
x
(1) Normalized intensity
Normalized intensity
x
x
x
1
(1)
x
x
x
(2)
(1)
(2)
x
x
x
x
1
x
(2)
(2)
(1)
I1/2
A1/2 0 7700
7720
7740 7760 Energy (eV)
7780
0 7700
7800
7720
7740 7760 Energy (eV)
7780
7800
+
2
x
(1)
(3)
x
x
x
1
x
(2)
(1)
x +
+ x
x
Normalized intensity
Normalized intensity
2
(2)
1
xx
+ x
(3)
(2)
(2)
I1 0 7700
7720
7740 7760 Energy (eV)
7780
x
+
I3/2 0 7700
7800
7720
7740 7760 Energy (eV)
7780
7800
Fig. 4. XANES spectra of samples A1/2 and I, and their simulations by linear combination from reference spectra. Experimental spectra are presented in full lines and simulations based on the values given in Table 2 in dotted lines. Reference spectra are numbered as follows: (1) Co2+ ions/ZSM5 (s); (2) Co3O4 (·); (3) cobalt phyllosilicate (A3/2) (+).
It can be observed that linear combinations of reference EXAFS signals [26] using the proportions found by study of the XANES spectra give a reasonable simulation of A1/2 and I1/2 experimental signals (Fig. 5). However, there is no evidence that the phases present in the catalysts have exactly the same characteristics as the reference compounds in terms of structural organization. Such comparison should thus be taken with caution and simulations are not presented for I1 and I3/2, for which the choice of references is not straightforward, due to the incertitudes on
which shows that a quantitative analysis based only on these two compounds is realistic. I3/2 contains at least one more compound, cobalt phyllosilicate, which accounts for the absence of clear isosbestic points. Table 3 sums up the respective contribution of the different components in each linear combination, which has been checked not to depend on the choice of the normalization point. To those results are added the assumed compositions of A1 and A3/2 based on EXAFS analysis, as described in the previous section. Table 3 Quantification of Co species (%) as determined by XAS Samples
Cox+ in Co3O4
Intraframework ions (Co2+ in (CoOx)n + isolated Co2+)
Co2+ in phyllosilicate
A1/2 A1
XANES EXAFS
52 –
48 –
A3/2
EXAFS
–
–
– 100 (low crystallinity?) 100
I1/2 I1 I3/2
XANES XANES XANES
33 25 25
67 75 –
– – 75
68
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
Co 3O 4
k*khi(k)
Co 2+ ions / ZSM5
I 1/2 I 1/2 sim
A1/2 A1/2 sim
1
4
6
8 k (Å -1)
10
12
Fig. 5. EXAFS signals of reference compounds and samples A1/2 and I1/ 2, and their simulations by linear combination from reference spectra. Simulations (‘‘sim’’) correspond to the proportions mentioned in Table 3.
Co3O4 crystallinity for the former, and on the precision of a XANES simulation based on two components only for the latter. 4. Discussion Quantification of species both by TPR and X-ray absorption spectroscopies brings complementary results, which is necessary because each technique is more sensitive to some types of species or to some characteristics, as already stressed in the case of Fe/ZSM5 systems [5,27]. XAS make possible the identification of the compounds through their structural characteristics and the determination of their oxidation state. TPR allows ranking the supported compounds, according to the decreasing proximity of cobalt ions and to their increasing interaction with the support along the series: • weakly interacting oxidic condensed phases; • oligomeric species of smaller nuclearity located inside the zeolite framework; • cobalt ions in neighbouring octahedral sites, embedded in a silicic extraframework matrix; • and isolated exchanged ions (absent from TPR thermograms). The combination of TPR and XAS leads to a quantification of extra- and intraframework species, further discrim-
inated upon their dispersion and/or nuclearity. It must be stressed though that there is no consensus in the literature on whether oligomeric species are located inside or outside the zeolite channels. We have followed the interpretations by da Cruz et al. [16], recalled by Wang et al. [3] and supported by results according to which oligomeric species containing 4–6 metal ions are stable inside the framework [28,29]. But it should be mentioned that Li et al. consider the TPR intermediate peak as characteristic of extraframework (CoOx)n oligomers [15], while Tang et al. consider temperature peaks at the lowest temperatures as characteristic of intraframework organized oxidic species [30]. There is an overall good agreement between the results presented in Tables 2 and 3, which will be discussed in terms of sensitivity of the techniques, since it is still difficult to derive a strategy for precise error calculations for these six complex systems. First, the XAS signals related to phyllosilicate are sensitive to the compound crystallinity, as shown by the study of A1 and A3/2. The fact that XAS overestimate the contribution of phyllosilicate (A3/2) comes from the uncertainties inherent to EXAFS analysis, for which a minor phase such as a small proportion of isolated ions cannot be discriminated against a major phase and only acts as perturbation through higher values of the Debye–Waller parameters (r = 0.10 for A1 and A3/2, compared with 0.08 for the simulation of oxygens first shell around isolated Co2+ ions [10,11]). This may explain the discrepancies between the results given by TPR and XAS for A1 and I3/ 2, with TPR appearing as more precise for the contribution of Co3O4, even when present in very small quantities. For samples containing phyllosilicate, TPR is more reliable and should be preferred. The second point to be discussed is the proportion of Cox+ in Co3O4 found by TPR for I1/2. This proportion is much lower than that found by XAS, which is probably the more reliable technique, since a separate integration of the peaks corresponding to dispersed Co3O4 and (CoOx)n oligomers is difficult to carry out because of peak overlap. Besides, it is not sure that (CoOx)n oligomers contains only Co2+ ions as supposed in the TPR section. Sample I1/2 is a typical example for which drawing a clear limit between dispersed extraframework Co3O4 entity and an intraframework (CoOx)n group would not be straightforward. Fig. 6a represents the quantity of intraframework and extraframework species per gram of catalyst as a function of cobalt loading, and Fig. 6b and c the quantities of each class of compound for the two series of catalysts. Quantification for series A and I3/2 is based on TPR results, because XAS lacks precision when phyllosilicate is present; moreover, the agreement between XAS and TPR is found to be good for A1/2. For I1/2 and I1, quantification is based on XAS results as far as the ratio oxide/intraframework ions is concerned, but since this technique is silent on the proportions of (CoOx)n oligomers and isolated Co2+ ions, the quantities of these species are calculated according to their relative proportions found by TPR. It was checked
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
600
Co (µmol.g -1 cat)
Cox+ in oxide (CoOx)n isolated Co2+ Co2+ in phyllosilicate
extrafr. A intrafr. A extrafr. I intrafr. I
700
69
Cox+ in oxide (CoOx)n isolated Co2+ Co2+ in phyllosilicate
500 400 300 200 100 0 1
(a)
2
3 4 Co wt%
5
6
1 (b)
2
3 4 Co wt%
5
6
1 (c)
2
3 4 Co wt%
5
6
Fig. 6. (a) Overall quantification of extra- and intraframework cobalt species as a function of Co weight content. (b) Detailed quantification of cobalt species as a function of Co weight content in catalysts A. (c) Detailed quantification of cobalt species as a function of Co weight content in catalysts I.
that a calculation based on TPR only would not change the trends described below. The quantity of intraframework cobalt is higher in series I than in series A, which is consistent with porosity measurements. A plateau seems to be reached for I around 2 wt.% of cobalt, showing that cobalt added in excess cannot be introduced into the zeolite. Moreover, it may be extrapolated that, for series I, intraframework cobalt exceeds extraframework cobalt up to a higher loading compared with series A, for which phyllosilicate formation is easier. Cobalt oxide, present in a more dispersed form in series I, represents less than 130 lmol Co g1 of catalyst in all samples, the quantity of cobalt in Co3O4 decreasing with increasing Co content. Finally, (CoOx)n oligomers seem to be present in higher quantity in series I than in series A, indicating that water helps the migration of cobalt into the porosity and the formation of smaller entities, but cannot guarantee that intraframework species will be isolated ions. These results show that the stage at which water is introduced is most important. Samples A and I can be compared along two points: • When water is added to the solid mixture prior to solidstate reaction, the subsequent formation of bulky Co3O4 particles is inhibited, as observed by Park et al. [6]. For A1/2, cobalt remains as oxide outside zeolite which has not been much damaged by the attack of acetates. In contrast, I1/2, like all samples I, contain smaller Co3O4 particles and entities located inside the zeolite porosity. • Though neither in dry nor wet processes, chemical species—in particular acetates—have been eliminated before heating, the formation of phyllosilicate occurs for a higher acetate/zeolite ratio in the wet process. When the contact with water has occurred prior to solid-state reaction, acetate anions are not chemically aggressive toward the zeolite surface, in contrast to what is observed in the solid-state procedure.
persed form. This can be the consequence of a partial dissolution of cobalt acetate in water. Dissolution is followed by dissociation of the complex, because cobalt acetate has a low global stability constant in water (b = 1–2 in the 20–50 C range [31]): 2þ
½CoðCH3 COOÞ2 4H2 Oaq þ 2H2 O ¢ ½CoðH2 OÞ6 aq þ 2CH3 COO aq
and UV–Visible spectroscopy shows that after drying, some cobalt ions have been able to interact with the zeolite. On the other hand, dissociated acetates which diffuse in the zeolite channels are allowed to react with inner zeolitic protons to which they have now access through the solvent: þ CH3 COO aq þ Hzeo ! CH3 COOHaq
It can be assumed that acetic acid CH3COOH remaining after drying decomposes easily into CH4 and CO2 during thermal treatment in helium, without needing any reaction with zeolite. Similarly, the thermal decomposition of iron acetate during solid-state exchange has been shown to be favored in the presence of water [32]. In contrast, unprotonated CH3COO ions must react with the zeolite surface during heating to catch protons, which seems to contribute to dealumination eventually leading to the phyllosilicate phase [10,11]. The more CH3COO ions are left unprotonated after wetting and drying, the more dealumination is expected later. The mass action law predicts that when the cobalt complex concentration increases in the wetting solution (which is the case along the series I1/2, I1 and I3/2), its degree of dissociation into cobalt ion and CH3COO decreases. A higher quantity of acetate ions is thus left unprotonated after drying on I3/2 and their release upon further heating may explain the formation of phyllosilicate in this sample, though not on catalysts prepared with a lower cobalt concentration. 5. Conclusions
Water thus promotes the transport of the cobalt salt inside the zeolite before thermal treatment, with the fraction remaining outside the framework appearing in a dis-
The effect of water on cobalt speciation during the preparation of Co2+/ZSM5 catalysts by solid-state reaction
70
H. Ben Boubaker et al. / Microporous and Mesoporous Materials 93 (2006) 62–70
between cobalt acetate and H-ZSM5 has been investigated quantitatively by the combined use of TPR and XAS. The introduction of water before solid-state reaction occurs has several effects on cobalt speciation in the final catalyst, compared with a dry procedure. At low cobalt content, water favors the formation of intraframework phases, such as (CoOx)n oligomers and isolated Co2+ ions, by facilitating the dissociation of cobalt acetate and the transport of cobalt ions inside the zeolite porosity; the dispersion of extraframework Co3O4 particles may also be increased for the same reason. The formation of extraframework cobalt phyllosilicate linked to the presence of acetates is inhibited, possibly as a consequence of the decomposition upon heating of acetic acid produced by reaction of acetates with zeolite protons during wetting. Finally, for high cobalt content, the use of a large excess of cobalt acetate above the stoichiometry of exchange does not help in reaching a higher deposition of cobalt inside the zeolite, most of the ions in excess contributing to the formation of extraframework phases, mostly phyllosilicate. Acknowledgments This work received the financial support of the FrancoTunisian CMCU (action 00F1205) which is gratefully acknowledged. Prof. Franc¸ois Bozon-Verduraz (ITODYS, Universite´ Denis Diderot, Paris) is thanked for providing access to the UV–Vis–NIR spectrometer. References [1] H.G. Karge, H.K. Beyer, in: P.A. Jacobs (Ed.), Zeolite Chemistry and Catalysis, Studies in Surface Science and Catalysis, vol. 69, Elsevier, Amsterdam, 1991, p. 43. [2] A.V. Kucherov, A.A. Slinkin, J. Mol. Catal. 90 (1994) 323. [3] X. Wang, H.Y. Chen, W.M.H. Sachtler, Appl. Catal. B 26 (2000) L227. [4] V. Schwartz, R. Prins, X. Wang, W.M.H. Sachtler, J. Phys. Chem. B 106 (2002) 7210. [5] F. Heinrich, C. Schmidt, E. Lo¨ffler, M. Menzel, W. Gru¨nert, J. Catal. 212 (2002) 157. [6] J.H. Park, C.H. Park, I.S. Nam, Appl. Catal. A 277 (2004) 271. [7] Z. Sobalı´k, A.A. Belhekar, Z. Tvaruzkova, B. Wichterlova´, Appl. Catal. A 188 (1999) 175. [8] H. Ben Boubaker, S. Fessi, A. Ghorbel, E. Marceau, M. Che, in: E. van Steen, L.H. Callanan, M. Claeys (Eds.), Recent Advances in the
[9]
[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]
[26] [27]
[28] [29] [30] [31] [32]
Science and Technology of Zeolites and Related Materials, Studies in Surface Science and Catalysis, vol. 154, Elsevier, Amsterdam, 2004, p. 2655, Part C. M. Mhamdi, S. Khaddar-Zine, A. Ghorbel, E. Marceau, M. Che, Y. Ben Taarit, F. Villain, in: J. Cejka, N. Zilkova´, P. Nachtigall (Eds.), Molecular Sieves: from Basic Research to Industrial Applications, Studies in Surface Science and Catalysis, vol. 158, Elsevier, Amsterdam, 2005, p. 748. M. Mhamdi, E. Marceau, S. Khaddar-Zine, A. Ghorbel, M. Che, Y. Ben Taarit, F. Villain, Catal. Lett. 98 (2004) 135. M. Mhamdi, E. Marceau, S. Khaddar-Zine, A. Ghorbel, M. Che, Y. Ben Taarit, F. Villain, Z. Phys. Chem. 219 (2005) 963. A. Michalowicz, EXAFS pour le Mac, Logiciels pour la Chimie, Soc. Fr. Chim., Paris, 1991, p. 102. J. Vakros, K. Bourikas, S. Perlepes, C. Kordulis, A. Lycourghiotis, Langmuir 20 (2004) 10542. A.A. Verberckmoes, B.M. Weckhuysen, R.A. Schoonheydt, Micropor. Mesopor. Mater. 22 (1998) 165. W. Li, S.Y. Yu, G.D. Meitzner, E. Iglesia, J. Phys. Chem. B 105 (2001) 1176. R.S. da Cruz, A.J.S. Mascarenhas, H.M.C. Andrade, Appl. Catal. B 18 (1998) 223. J.Y. Carriat, M. Che, M. Kermarec, A. Decarreau, Catal. Lett. 25 (1994) 127. J.D. Han, S.I. Woo, J. Chem. Soc., Faraday Trans. 88 (1992) 145. A.Y. Khodakov, J. Lynch, D. Bazin, B. Rebours, N. Zanier, B. Moisson, P. Chaumette, J. Catal. 168 (1997) 16. G.C. Shen, T. Shido, M. Ichikawa, J. Phys. Chem. 100 (1996) 16947. M. Shirai, K. Asakura, Y. Iwasawa, Catal. Lett. 15 (1992) 247. G.P. Huffman, N. Shah, J. Zhao, F.E. Huggins, T.E. Hoost, S. Halvorsen, J.G. Goodwin Jr., J. Catal. 151 (1995) 17. D.T. On, S. Kaliaguine, L. Bonneviot, J. Catal. 157 (1995) 235. G. Jacobs, P.M. Patterson, Y. Zhang, T. Das, J. Li, B.H. Davis, Appl. Catal. A 233 (2002) 215. F. Morales, F.M.F. de Groot, P. Glatzel, E. Kleimenov, H. Bluhm, M. Ha¨vecker, A. Knop-Gericke, B.M. Weckhuysen, J. Phys. Chem. B 108 (2004) 16201. G. Redden, J. Bargar, R. Bencheikh-Latmani, J. Colloid Interf. Sci. 244 (2001) 211. L. Capek, V. Kreibich, J. Dedecek, T. Grygar, B. Wichterlova´, Z. Sobalı´k, J.A. Martens, R. Brosius, V. Tokarova´, Micropor. Mesopor. Mater. 80 (2005) 279. A. Jentys, K. Kleestorfer, H. Vinek, Catal. Lett. 39 (1996) 119. R. Joyner, M. Stockenhuber, J. Phys. Chem. B 103 (1999) 5963. Q. Tang, Q. Zhang, P. Wang, Y. Wang, H. Wan, Chem. Mater. 16 (2004) 1967. P.H. Tedesco, V.B. de Rumi, J. Inorg. Nucl. Chem. 33 (1971) 969. K. La´za´r, G. Pa´l-Borbe´ly, H.K. Beyer, H.G. Karge, in: G. Poncelet, P.A. Jacobs, P. Grange, B. Delmon, J. Martens (Eds.), Preparation of Catalysts VI: Scientific Bases for the Preparation of Catalysts, Studies in Surface Science and Catalysis, vol. 91, Elsevier, Amsterdam, 1995, p. 551.