Studies in Surface Science and Catalysis 130 A. Corma, F.V. Melo, S.Mendioroz and J.L.G.

Fierro (Editors) 9 2000 Elsevier Science B.V. All rights reserved.

1691

Ammoxidation of propylene to acrylonitrile catalyzed by multimetal molybdate- and iron antimoniate-based active compounds dispersed in oxidic matrixes: the effect of the dispersing oxide on the catalytic performance
R. Catani b, F. Cavani a, U. Cornaro b, A. Del Bianco a, E. Frontani a, D. Ghisletti b, A. Tasso b, F. Trifir6 a aDipartimento di Chimica Industriale e dei Materiali, Viale Risorgimento 4, 40136 Bologna, Italy bSnamprogetti, Via Maritano 26, 20097 S. Donato Milanese, MI, Italy The effect of three different dispersing agents, silica, y-alumina and tin oxide, on the catalytic performance of multimetal molybdate- and antimoniate-based catalysts was evaluated in the reaction of propylene ammoxidation to acrylonitrile. Silica provided the better dispersing effect on the active components, due to both a negligible direct contribution of silica to the reaction, and a weak chemical interaction with the active components. 1. INTRODUCTION The ammoxidation of propylene to acrylonitrile is carried out in fluidized-bed reactors, catalyzed by heterogeneous multicomponent systems which consist of either multimetal molybdates, or antimoniates of U or Fe, or mixed molybdates/antimoniates [1]. In all cases the active phase is dispersed inside a silica matrix, to achieve a resistant and fluidizable catalytic material. Silica is considered the most suitable dispersing material, since it is the least reactive towards organic substrates, and gives the final catalyst the best mechanical and chemical-physical properties. In this work we compare the effect of three dispersing oxides: i) silica, ii) ~/alumina and iii) cassiterite SnO 2, on the catalytic performance of multimetal molybdates and of iron antimoniate in the ammoxidation of propylene to acrylonitrile. This comparison makes it possible to evaluate the use of oxidic materials with different surface properties as dispersing agents for mixed oxides in complex catalytic systems used for several applications in the field of selective oxidation of organic substrates.

2. EXPERIMENTAL
Three different dispersing oxides were prepared:

1692 1) silica, from a SiO2 sol Ludox AS 40 Du Pont; after spray-drying and calcination at 500~ the surface area was 125 m2/g; 2) t-alumina, from a A1OOH sol AL-20 Nyacol; surface area of calcined material was 188 m2/g; 3) cassiterite, from a SnO2 sol SN15CG Nyacol; surface area 50 m2/g. Two different mixed oxides were prepared as active compounds: 1) Fe/Sb 1.0/1.5 (atomic ratio); a slurry was prepared which contained fused Fe(NO3)3.9H20, Sb203 and HNO3 in the right proportions (slurry-redox method). Drying of the slurry and calcination o-T~e solid at 500~ led to the development of the rutile-type FeSbO4 compound, as shown by means of DRIFT spectroscopy and XRD. Other compounds identified were Sb204 cervantite (27 wt.%), Sb203 senarmontite (9 wt. %) and Fe203 (9 wt. %). The same compounds were also identified in dispersed, spray-dried systems. Only in the case of tin oxide-dispersed catalyst was identification of crystalline compounds present in the active phase not possible, since SnO2 and FeSbO4 are isostructural. 2) Bi/Fe/Ni/Co/Mo 1.0/1.9/4.6/9.7/18.6 (atomic ratios); aqueous solutions of (NH4)6Mo7024.4H20 with NH4OH and Bi(NO3)3.5H20 with HNO3 were mixed in HNO3 under stirring (pH = 2, maintained through addition of NH4OH). A precipitate was obtained, which was filtered and washed, dried and calcined at 500~ Separately, aqueous solutions of (NH4)6Mo7024.4H20 with NH4OH and Fe(NO3)3.9H20 were mixed under reflux temperature while stirring. The resulting precipitate was filtered and washed, dried, and calcined at 500~ A third precipitate was obtained by mixing an aqueous solution of (NH4)6Mo7024.4H20 and NH4OH with a second solution containing Co(NO3)2.6H20 and Ni(NOg)2.6H20, under reflux while stirring. The precipitate was filtered and washed, dried, and calcined. The three precipitates were then mixed through redispersion in water, in the relative amounts necessary to obtain the desired composition; the resulting slurry was filtered and the solid was washed and dried. Calcination of this solid at 500~ led to the development of the following crystalline compounds: Bi2Mo3012 (34 wt.%), Fea(MoO4)3 (42 wt.%), [3-NiMoO4 + ~-CoMoO4 (24 wt.%). When dispersed in ~/-alumina, only crystalline Fe2(MoO4)3 and [3-(Ni,Co)MoO4 could be unambiguously identified (the presence of Bi3FeMo2012 however could not be excluded). When dispersed in silica, the following crystalline compounds were identified: ~-(Ni, Co)MoO4, BiaMo209, Fe2(MoO4)3, Bi3FeMo2012. Finally, dispersion in tin oxide led to a catalyst where it was not possible to identify any crystalline compound relative to the active phase. In all cases, the active phases were dispersed in the oxidic gels (60 wt.% of the dispersing oxide), and then the slurry was homogeneized and spray-dried at 300~ using a laboratory spray-drier ((Lab-Plant SD-04). The solid was finally calcined at 500~ The surface area of the dispersed active phase B i / F e / N i / C o / M o / O was 128 m2/g for the alumina-dispersed system, 89 m2/g for the silica-dispersed system and 81 m2/g for the tin oxide-dispersed system. Catalytic tests were carried out in a fixed-bed glass laboratory reactor, loading I g of catalyst, and operating at atmospheric pressure. The typical feedstock composition was the following: 7.5 vol.% propylene, 10.4 vol.% ammonia, 17.3 vol % oxygen,

1693 remainder nitrogen. The residence time (W/F) was equal to 2 g m1-1 s. Analysis of products was carried out through on-line sampling of a defined volume of the effluents to gas-chromatographs equipped with TCD. 02, N2 and CO were analyzed using a MS 5A packed column, while all the other unconverted reactants and products formed were analyzed using a Hayesep-T packed column. /" 3. RESULTS AND DISCUSSION
3.1 The reactivity of the dispersing oxides First the reactivity of the dispersing oxides alone, calcined at 500~ was studied. Figure 1 reports the conversion of propylene, the conversion of ammonia and the selectivity to N2, which was formed by combustion of ammonia, as functions of the reaction temperature for the three oxides. Figure 2 reports the distribution of the Ccontaining products at 465~
70 60
E

100 80
7 or)

50 30 -

.o 40 u')
c- 20 o o 10
>

60 40 20
w--~ r r =

>
cO

420 440 460 480

400

400

420

440

460

480

temperature,~

temperature,~

Fig. 1. Conversion of propylene (left), and ammonia (right, full symbols) and selectivity to N2 (right, open symbols) as functions of temperature for the three dispersing agents: ?-alumina (e), silica (11), and tin oxide (&). The three dispersing oxides showed considerable differences in reactivity towards propylene. On the contrary, the interaction with ammonia and its transformation either to C,N-containing products (acrylonitrile, acetonitrile, cyanidric acid) or to the combustion product N2 (selectivity to N2 in Figure 1) was similar for all the materials. Silica and alumina showed quite different product distributions. In the case of ?alumina the prevailing products were CO and acetonitrile (no formation of C3 products was observed), while with silica the main product was HCN. Tin oxide yielded almost exclusively CO2. In the case of alumina, surface acidity can be responsible for the formation of an ionic intermediate, which is then cracked to yield C1 and Ca fragments, precursors of CO2 and acetonitrile formation, through unselective oxidative attack and ammonia insertion, respectively. It is worth mentioning that the molar productivity of CO2 and acetonitrile was the same, thus suggesting a common intermediate.

1694
100 80 -

Elacrylonitrile Ilacetonitrile BHCN Ilacrolein

E]CO ielCO2

>~
~>

60 40

(D

20 0 -------alumina silica tin oxide

Fig. 2. Selectivity to the different products obtained in propylene ammoxidation with the three dispersing oxides (calcined at 500~ Temperature 465~ However, the product which was formed to the greatest extent was CO, the formation of which probably arose from another type of reaction intermediate. It is reported [2] that the interaction between Lewis-type acid centers on coordinatively unsaturated A13+ sites and the double bond of propylene, generates an allylic radicalic intermediate via H. abstraction by the O atom bridging two A13+ions. In the case of silica, the quite different distribution of products suggests a reaction mechanism different than with alumina. It is known that silica is able to activate paraffins and oxidize them, even though in many cases this role has been attributed to the impurities of metals present in silica [3]. However, with extremely pure silica the formation of 02.- species occurring on defect surface sites has been demonstrated, which might be responsible for the transformation of propylene to C1 fragments containing N and O (HCN and COx) [4]. Finally, tin oxide was extremely active, and catalyzed the combustion of propylene to CO2. Tin oxide is a n-type, wide-gap semiconducting oxide, and it is known that it can catalyze redox-type reactions in the presence of 02 and of a reductant (an hydrocarbon), with involvment of the couple Sn4+/Sn 2 However, it is not able to promote selective O2--insertion onto activated hydrocarbons, but rather it catalyzes oxidehydrogenation reactions, such as the formation of ethylene from ethane or of formaldehyde from methanol [5]. In our case, in the absence of NH 2-- or of O2--insertion sites the intermediate formed undergoes unselective oxidative attack with formation of CO2. Notwithstanding the differences observed in propylene ammoxidation, all three oxides behaved similarly in the activation of ammonia, in terms of both conversion and selectivity; in all cases the main N-containing product was N2. This suggests that the surface sites responsible for propylene and ammonia activation are likely not the same. Alternatively, it is possible that in these cases the mechanism of ammonia activation and combustion is not heterogeneous, but rather mainly homogeneous.

1695
3.2 The reactivity of the active compounds dispersed in oxidic matrixes Figure 3 (left) shows the propylene conversion as a function of temperature for Fe/Sb/O, dispersed in alumina, silica and tin oxide. Figure 3 (right) reports the distribution of the products at 439~ for the same catalysts. Figure 4 shows the propylene conversion and the distribution of the products for the multimetal molybdate, dispersed in the three oxides.
80 100
8O

~"
C) ~)

60 40
.4..J

eo
40 20
I I I I I

2o

cD CD {/)

410

420

430

440

450

460

470

temperature, ~

alumina

silica

tin oxide

Fig. 3. Conversion of propylene (full symbols) and ammonia (open symbols) (left) and selectivity to the different products at 439~ (right) for the active phase Fe/Sb/O dispersed in ~,-alumina (o), silica (m), and tin oxide (A). Captions for Fig. 3 (right) as in Figure 2. Figure 3 shows that in all cases the conversion was independent of the reaction temperature. This was because oxygen was the limiting reactant (in all cases conversion was already practically complete at the lowest temperature), while the conversion of ammonia was the highest for the silica-dispersed system. The distribution of the products was similar for all catalysts, and this indicates that in all cases the reactivity of the dispersing matrix was negligible with respect to the activity of the antimoniate (except maybe in the case of the tin oxide-dispersed system, for which the highest selectivity to CO2 was obtained). The very low selectivity to acrylonitrile for all catalysts was likely due to the presence of Fe203, as evidenced by XRD. Calcination of the Fe/Sb/O system at 700~ is likely necessary to obtain complete development of the rutile structure. In the case of the active phase B i / F e / N i / C o / M o / O , considerable differences in reactivity between the three systems could be observed, both in propylene and in ammonia activation (Figure 4). The conversion of the silica-dispersed system increased with increasing temperature; only in this case was total conversion of oxygen reached at the highest temperature. In the case of the alumina- and tin oxidedispersed systems the conversion of propylene was instead limited by oxygen. In the tin oxide-dispersed catalyst the support was clearly responsible for the higher formation of CO2 (with a consequent lower maximum conversion of propylene). Differences between alumina- and silica-dispersed systems (which had comparable

1696 behavior in the case of the Fe/Sb/O active phase, and which themselves were both characterized by low activity in propylene ammoxidation) can be attributed to a role of alumina either i) on propylene transformation (this however can be neglected, also taking into consideration the distribution of products of the support alone, Figure 2), or ii) in favouring the combustion of the acrylonitrile formed, or iii) in affecting the nature of the active components in the catalyst. The latter effect likely is the most important, since i) the non-dispersed B i / F e / N i / C o / M o / O also behaves similarly to the silica-dispersed catalyst, and ii) there are considerable differences in the characteristics of the silica- and alumina-dispersed catalysts: indeed in the aluminabased system it was not possible to detect the presence of any crystalline Bi/(Fe)/Mo/O compound.
100 80

80

eo

"~ 40

o
40 410 430 450 470 490

"~

20

0 alumina silica tin oxide

temperature, ~

Fig. 4. Conversion of propylene (full symbols) and ammonia (open symbols) (left) and selectivity to the different products at 439~ (right) for the active phase B i / F e / N i / C o / M o / O dispersed in ~,-alumina (e), silica (11), and tin oxide (&). Captions for Fig. 4 (right) as in Figure 2.
4. ACKNOWLEDGEMENTS

Snamprogetti is gratefully acknowledged for financial support.

REFERENCES

1. R.K. Grasselli, in "Handbook of Heterogeneous Catalysis", G. Ertl, H. Knozinger and J. Weitkamp (Eds.), 1997, p. 2302 2. T.A. Gordymora and A.A. Davidov, Kinet. Catal., 20 (1979) 604 3. A. Parmaliana, V. Sokolovskii, D. Miceli, F. Arena, N. Giordano, J. Catal., 148 (1994) 514 4. Y. Barbaux, D. Bouqueniaux, G. Fornasari, F. TrifirG Appl. Catal., A: General, 125 (1995) 303 5. R. Burch and E.M. Crabb, Appl. Catal., A: General, 97 (1993) 49