- Email: [email protected]
Hydrogen back spillover phenomena in the aromatization of light alkanes over hybrid catalysts. Mechanistic considerations
T. Inui et al. (Editors), N e w Aspects of Spillover Effect in Catalysis 0 1993 Elsevier Science Publishers B.V. All rights resewed.
143
H drogen back spillover phenomena in the aromatization of light al anes over hybrid catalysts. Mechanistic considerations
1
Raymond Le Van Mao '*I1, Vittorio Ragain?
Jianhua Yao', Louise Dufresne', Riccardo Cadi2 and
Concordia University, Department of Chemistry i 3 Biochemistry, Laboratories for Inorganic Materials and Catalysis Research Laboratory, 1455 De Maisonneuve Blvd. W., Montreal (Quebec) H3G 1M8 Canada. University of Milan, Department of Physical Chemistry and Electrochemistry, Via Golgi 19, 2013 3 Milano, Italy
Abstract
By using irreducible oxides as cocatalysts or cocatalyst supports for the preparation of hybrid catalysts, it was possible t o show that the final steps of the light alkane aromatization involved a significant dehydrogenation activity by hydride abstraction over the zeolite acidic sites.
1. INTRODUCTION Aromatization of light alkanes (propane, butanes, liquefied petroleum gas) is an important industrial process and an exciting topic of research in the field of I is a gallium ZSMcatalysis. The catalyst used in the commercial Cyclar process [l 5 zeolite which is prepared according to the classical bifunctional catalysis, e.g. that gallium species are incorporated into the zeolite pore system in such a manner that each of the metal sites is adjacent t o an acid site. Such a catalyst can achieve high yields of aromatics (mostly, BTX, which are namely benzene, toluene and xylenes)[2,31. Some studies were specifically carried out in order t o elucidate the reaction mechanism on the pure acidic ZSM-5 zeolite [41. Recently, a new explanation based on the concept of hydrogen back (or reverse) spillover (HBS) was devised for the enhancement of aromatization of light alkenes and alkanes over hybrid catalysts 15-131. Such catalysts were prepared basically by intimately mixing the ZSM-5 zeolite particles with the Zn or Ga containing cocatalyst aggregates, and the resulting solid mixture was embedded in a clay matrix. Irreducible oxides such as quartz, which did not contain any Zn, Ga or Pt, were also capable of enhancing the aromatizing properties of the ZSM-5 zeolite [9,121. According t o the HBS concept, the cocatalyst acts as a hydrogen sink for the entire aromatization reaction, and such an action is exerted at the pore openings of the zeolite particles, e.g. at a fairly large distance from the zeolite acid sites [l 1,131. Recently, it was found that the the state of gallium species located at the "zeolite-cocatalyst particles" interface within the gallium containing hybrid catalysts had a great influence on the aromatizing performance [14-161. Thus, this paper aims at providing (a critical report of the) experimental evidence of the effects that the cocatalyst could have on the reaction mechanism.
I44
2. EXPERIMENTAL 2.1 PREPARATION AND CHARACTERIZATION OF CATALYSTS The parent ZSM-5 zeolites used for the preparation of t w o series of catalysts were synthesized according to the known technique. The powder acid forms were obtained by ion-exchange with ammonium ions and subsequent activation in air at elevated temperature 151. The Si/AI atomic ratios of the resulting zeolites, HZ and HZ , were 36 and 34, respectively. Final catalyst forms of the parent zeolites were obtained by extruding the zeolite powder with bentonite 151. i) First series of catalysts prepared from Hi!: Hybrid catalysts which comprised a ZSM-5 zeolite component (HZ) and a cocatalyst were obtained b y mixing mechanically these t w o powders and then extruding with bentonite clay as described in refs. 19-12,141. The cocatalysts used were Quartz (Qz), evaporated Ludox silica (LuSi) and Ga/Ludox silica coevaporation (Ga/LuSi), as mentioned in refs. 19,171. The resulting hybrid catalysts were called HZ//Qz, HZ//LuSi and HZ//Ga/LuSi, respectively. The Cyclar-type bifunctional catalyst was prepared by refluxing the HZ powder with a solution of gallium nitrate, and then evaporating the liquid, as described in ref. 151. The final catalyst, also called HZ/Ga(R,E) was obtained by extruding the resulting solid with bentonite. The Ga,O, content of such a catalyst was 3 wt %. ii) Two hybrid catalysts containing a Zn oxide precipitate and a Zn-Al oxide (1/ I coprecipitate, respectively, were also used. The preparation of such catalysts, also called HZ'//Zn and HZ'//Zn-AI, respectively, was fully described in refs.[8,181. 2.2 CATALYST TESTING The experimental set-up, the procedures for catalytic testing and analysis of products were similar t o those described in refs. 15,8-121. In particular, the reaction temperatures were = 500 "C for propylene and 540 "C for n-butane. The total conversion of n-butane or propylene, C,, the selectivity in product hydrocarbon i, Si, and the hydrogen yield, Y, were defined in refs. 114, 15, 17, 181. ,Y, is the ratio of the moles of hydrogen produced t o the number of carbon atoms of aromatics produced, e. g. Y ,, = YS /, , (Ar for aromatics). C ,, is the average number of carbon atoms in the product aromatics, the determination of which was done on the basis of the distribution of such hydrocarbons provided by gas chromatography (GC equipped with ionization and mass detectors)., ,Y , is the ratio of the moles of hydrogen produced t o the moles of aromatics produced.
3. RESULTS AND DISCUSSION Figure 1 shows the general reaction mechanism for the aromatization of light alkanes according t o the type of catalyst used. I)CASE (I)represents the sequential transformation of n-butane t o aromatics over the conventional bifunctional catalyst wherein the acid sites are adjacent to the metal (gallium or zinc oxide) sites. Therefore, as first step, n-butane undergoes dehydrogenation (DH) to the corresponding olefin which is then adsorbed on the protonic acid site providing thus the adsorbed carbenium ion species, (C4H9)+.An hydrogen molecule is liberated. Then, through a well-known sequence of reactions (oligomerization, cyclization/rearrangement), a cyclic carbenium ion was formed. At that stage, the aromatization occurs by means of an hydrogen transfer (TR) between the aromatizing species and a carbenium ion. This results in the formation of paraffins including n-butane. However, if strong dehydrogenating (Ga, Pt or Zn) sites are present in sufficient number, aromatics are now predominantly formed by dehydrogenation of the naphthenic intermediates, with thus more hydrogen
145
CASE (11)
CASE (I) Par
CqH10
Par
-.-.- t
'qH1O
I
i
13+
I
I
!
!
I
_._._. - !
C4H9 +
!
c
H2
(d
i2
AROMATICS
AROU~TICS
A2
Figure 1 . Reaction pathways according to the catalyst nature: (I)= bifunctional catalyst; (11) = pure zeolite catalyst; (Ill)= hybrid catalyst. (DH): dehydrogenation; (HAb1:hydride abstraction; (0C):oligomerizationlcyclization; (NA):transformation of naphthenic intermediates t o aromatics; (TR):hydrogen transfer; Par:paraffins other than n-butane; H,':(probably semi-sorbed) hydrogen species. produced (Figure 1 ) . The primary step is undoubtedly the dehydrogenation of nbutane t o butenes because in our previous work [lo], the hydrogen t o aromatic yield ratios (called Y ,, and YHM,) and the yield of butenes both went through a maximum when the reaction contact time tended t o zero. 11) CASE ( 1 1 ) represents the sequential transformation of n-butane over the pure ZSM-5 zeolite having only acid sites. As a first aromatization step, there is an adsorptionlhydride abstraction over the protonic site, resulting in the formation , at first, of a carbonium ion [ I 91 and then, by decomposition, an adsorbed (C,H,)+ species with the release of an hydrogen molecule. The latter is probably semiadsorbed. In fact, in our previous work [lo], there was no maximum observed for the yield of butenes (thus, butenes were not the primary products) and for the Y,, ratio, when the reaction contact time tended to zero. Because of such an absence of dehydrogenating sites, the paraffins (mainly, n-butane) formed during the aromatics-forming step will diffuse out without any further reaction. Thus, we ratio are much have the situation in which the n-butane conversion and the ,Y, lower than that obtained with the bifunctional catalyst. This is clearly shown in Table 1 wherein HZ is the parent ZSM-5 zeolite and HZ/Ga(R,E) represents the bifunctional Ga bearing catalyst. Table 1 Aromatization performance of the catalysts investigated CATALYST
ct
SA,
YH
HZ HZ//QZ HZ//LuSi HZ/Ga(R,E) HZlIGalLuSi
82.5 90.0 90.5 82.3 99.7
20.1 33.8 41.7 49.2 67.5
0.07 0.1 1 0.17 0.23 0.34
Y ,,
CA,
YHMA
0.34 0.36 0.41 0.47 0.50
7.19 7.13 7.01 6.98 6.94
2.44 2.53 2.87 3.28 3.47
146
Such detrimental effect of hydrogen transfer in the absence of strong dehydrogenating sites is also evidenced by the important formation of propane in the aromatization of propylene (no need of a hydride abstraction as initiation step: we are n o w focussed only on the last aromatization steps) as reported in Table 2. Table 2 Propylene aromatization over H-ZSM-5 zeolite and hybrid catalysts C, (C-atom %)
HZ'
95.3
Product selectivities (C-atom %) methane 2.7 ethylene 4.7 ethane 4.1 propane 30.5 butanes 8.4 butenes 2.5 C,+ aliphatics 4.1 aromatics 43.0
HZ'//Zn-AI
HZ'//Zn
3.1 2.9 5.5 10.5 3.4 1.6 4.4 68.6
4.3 1.4 12.0 14.0 2.1 0.9 0.8 64.5
96.2
96.5
Ill) For CASE (Ill),let us consider at first the catalytic properties of the hybrid catalyst, HZ//LuSi, whose cocatalyst was the evaporated colloidal Ludox silica. Such a cocatalyst could eventually intervene only at the level of the pore openings of the zeolite particles. If the spillover concept was effective in that case, the resulting more efficient removal of hydrogen (and maybe also other reaction products) induced by the presence of the cocatalyst, should result in a higher conversion of n-butane and a higher yield in aromatics. This was in effect obtained experimentally (Table 3). However, since there were not any dehydrogenating sites ratios (Table 1 ) in this hybrid catalyst, the significant increase in the ,Y, and Y, indicates that the hydrogen transfer (which was normally detrimental for the net hydrogen production) t o the (C,H,)+ ion or other olefinic carbenium ion was significantly reduced. Therefore, we have t o admit that some cyclic species undergo dehydrogenation t o aromatics by losing their hydrogen in the same manner as n-butane in its activation step, e.g. by hydride abstraction over the zeolite acid sites. All this is depicted in Figure 1 , CASE (111). Such an explanation is also supported b,y the data obtained with propylene as feed. In fact, when the hybrid catalyst HZ //Zn/Al or HZ //Zn was used, more aromatics were obtained (Table 2) and less propane was formed, indicating that, also in this case, the hydrogen transfer process was seriously reduced when the hydrogen removal became more efficient. Since the hydrogen back spillover action involves the "sink effect", e.g. the capacity of the cocatalyst surface to accept the spilt-over hydrogen species, the textural properties of the latter appear to be important factors. This was verified when we compared the catalytic properties of the HZ//Qz and HZ//LuSi hybrid catalysts. The latter one was more active because i t contained a cocatalyst having a larger surface area and a more homogeneous pore system than those of the quartz in the HZ//Qz hybrid catalyst [91. It is worth reminding that bentonite clay, quartz and the evaporated Ludox silica did not show any dehydrogenating properties with respect to n-butane [ I 71 and those of the Zn/Al co-precipitate were negligible under the reaction conditions investigated [ 181. Also in connection with CASE (Ill) of Figure 1 , let us consider n o w the
catalytic properties of the hybrid catalyst, HZIIGaILuSi, which contained the
147
gallium-silica cocatalyst. Such a catalyst was tested with n-butane and the following results were obtained. i)The behaviour of such a hybrid catalyst was similar to that of the parent ZSM-5 zeolite in the tests carried out a t low contact times (no maximum for the Y,, ratio and the yield of butenes), thus indicating that the n-butane aromatization reaction was not initiated by the dehydrogenation of n-butane as in the case of the conventional bifunctional catalyst 1201. It is worth mentioning that the galliumsilica cocatalyst alone did not show either any significant dehydrogenation activity with respect t o n-butane [ I 1,131; ii)The gallium transfer from the cocatalyst body to the zeolite particles, even upon drastic treatments with hydrogen, was fairly small, not exceeding 0 . 2 wt % 11 41. As reported in Table 1, the aromatizing properties of the HZ//Ga/LuSi hybrid catalyst were very high. The Y,, and Y, ratios increased with increasing aromatizing properties, independently from the presence or absence of gallium, according to the following sequence: parent zeolite c gallium-free hybrid catalysts C gallium-containing bifunctional catalysts c gallium-containing hybrid catalysts. Basically, such results strongly suggested the existence of dehydrogenating activity over acid sites in the very last steps of the aromatization. iii) Another experimental evidence for such promoting effect of the cocatalyst came from the following observation. In addition to undergoing aromatization according to the previously mentioned "direct" pathways, any alkane molecule when adsorbed on a relatively strong acid site is submitted to cracking reactions. The first one is the cracking of n-butane into methane and propylene. The latter may be converted rapidly to benzene or to other products (by further cracking, alkylation or hydrogen transfer). Table 3 reports the values of benzene and methane yields, Y,,, and Y,,, respectively. Let us make the assumption that benzene be predominantly produced by aromatization of propylene. Since the total yield of methane is always higher than the yield of methane in the cracking of nbutane t o propylene and methane, the ratio RAR = Y,EN/ 2 YM, may be considered as a parameter which is capable of giving an idea about the rate of transformation of naphthenic intermediates into aromatics. As shown in Table 3 and in agreement with the observations of Guisnet et al [211, the bifunctional catalyst was more efficient in the transformation of naphthenes into aromatics (N--A) than the pure zeolite. However, our hybrid catalyst HZ//LuSi which did not contain any gallium, had practically the same value of RAR as the bifunctional catalyst. Results of Table 1 (YHAand YHMA) and Table 3 (RAR) allows us t o say that the LuSi cocatalyst was actually activating such a transformation (N - A) by accelerating the removal of hydrogen and other products. Thus, this process liberated more acid sites for other naphthenic intermediates which could in turn undergo aromatization by hydride abstraction [(Ill in Figure 11. Table 3 Effets of the cocatalyst on the rate of transformation of the naphthenic intermediates into aromatics CATALYST HZ HZ//Qz H Z//Lu Si HZ/Ga (R, E) HZ//Ga/LuSi
',EN
3.91 8.20 11.81 13.36 25.73
YME
RAR
4.61 8.21 6.52 7.65 10.06
0.43 0.50 0.91 0.88 1.28
I48
Therefore, the higher stability and the higher dispersion of the gallium species incorporated into the cocatalyst [9,10,141 resulted in a higher activity, a much shorter induction period (to reach the maximum aromatizing activity without any reducing pretreatment) and a higher "on-stream" stability for the HZ//Ga/LuSi hybrid catalyst. This could not merely be ascribed t o the gallium migration into the zeolite particles which would create a situation similar t o that of a bifunctional catalyst. In fact, by using the technique of surface transfer 11 51, gallium oxide was deposited o n the external surface of the ZSM-5 zeolite particles. Although such gallium amounts were small (0.4-0.6 wt %), the "contaminated" zeolite particles showed an aromatizing activity as high as the best bifunctional catalysts. Since the required gallium content for a perfect bifunctional situation is 1.5 wt % and if w e admit that half of the amount of gallium deposited onto the zeolite external surface moved t o the inside (e.g. 0 . 2 wt %), would we be ready t o accept that 15 % of the bifunctional sites could behave like a 100 % bifunctional configuration ? By using X-ray photoelectron spectroscopy and hydrogen adsorption/temperature programmed desorption technique [161, w e have recently shown that there were t w o hydrogen chemisorption sites for the gallium component: the highest aromatizing performance corresponded t o the hybrid catalyst which exhibited the largest number of stronger chemisorption sites (namely S as opposed to the weaker sites, SHJ. These sites were associated with high 6 a dispersion. On the other hand, it was shown that the apparent activation energy was the same for the pure ZSM-5 zeolite and the gallium containing catalysts (both bifunctional and hybrid catalysts) [lo]. It means that all the three reaction pathways (I, I I and I l l ) probably have the same rate-determining step (RDS) as depicted in Figure 1. In the case of a bifunctional catalyst, the RDS does also include a dehydrogenation and an immediate protonation t o (C,H,)' ion, and not only the dehydrogenation process as accepted by reference [221. This thus implies the existence "in duo and at short distance" of these t w o acid and metal (as oxide) sites. Such a short distance requirement for a perfect bifunctional configuration may be the origin of observations made by several authors about possible interactions between the t w o active sites [23-251. Other experimental evidence of the beneficial effect of the cocatalyst on the rate of diffusion of hydrogen (and other reaction products) was provided by the following results. i) The amount and nature of the coke deposited on the pure zeolite HZ and the hybrid catalyst HZ//Ga/LuSi were also investigated [17al). It was found that under normal reaction conditions with n-butane as feed, on-stream aging the catalysts for 22 h led to the following carbonaceous species deposits, as analysed by thermal gravimetric method: 1.3 wt % for the HZ zeolite (average combustion temperature: 640 "C) and 0.8 wt YO for the hybrid catalyst (average combustion temperature : 530 "C). It was also observed by means of CP/MAS 13C NMR spectroscopy method, that the coke deposited on the pure zeolite was predominantly aromatic in nature (advanced coke, 100-16 0 ppm region) while that deposited o n the hybrid catalyst was predominantly paraffinic/olefinic (light coke, 0-50 ppm region) [17 a]. ii) The average number of carbon atoms related t o the aromatic products (C,J in the case of the hybrid catalysts was lower than that obtained with the parent zeolite (Table 2). Moreover, the selectivity to naphthenic products was higher for the hybrid catalyst than for the parent zeolite while the selectivity to C, and C,, aromatics was much lower [17a]. The only possible explanation stems on a faster removal of hydrogen and other products, thus preventing any excessive formation of heavy coke which normally occurs at the pore mouths of the pure zeolite.
149
4. CONCLUSION We have shown that the reaction pathway followed by the hybrid catalyst, is mainly based on the hydride abstraction process not only in the n-butane activation step but also in the transformation of naphthenes into aromatics. The latter is considerably enhanced by the hydrogen back spillover action exerted by the cocatalyst itself. In addition, when gallium cocatalyst is used, the configuration of gallium sites on the "zeolite / cocatalyst particles" interface appears t o be the key factor for the high aromatization performance of the hybrid catalysts.
5. ACKNOWLEDGEMENTS We would like t o thank NSERC of Canada and Quebec's Actions Structurantes Program for their financial support. We also thank Dr. Georges Denes for helpful discussions and Mr. Bernard Sjiariel for technical assistance.
6. REFERENCES 1 2 3 4
P.C. Doolan and P.R. Pujado, Hydroc. Process., (Sep. 1989) 72. Y. Ono, Catal. Rev.-Sci. Eng., 34(3). (1992) 179. M. Guisnet and N.S. Gnep, Appl. Catal. A, 89 (1992) 1. M. Guisnet, N.S. Gnep, D. Aittaleb and Y.J. Doyemet, Appl. Catal. A, 87 (1992) 255 and references therein. 5 J. Yao, R. Le Van Mao and L. Dufresne, Appl. Catal. 65 (1990) 175. 6 T. Inui, "Successful1 Design of Catalysts", T. lnui Ed., Studies in Surface Science and Catalysis 44, Elsevier, Amsterdam (1989) 189. 7 K. Fujimoto, I. Nakamura and Y. Yokota, Proc. Second International Conference on Spillover, Leipzig (Germany), June 1989, 176. 8 R. Le Van Mao and L. Dufresne, Appl. Catal. 52 (1989) 1. 9 R. Le Van Mao, J. Yao and B. Sjiariel, Catal. Lett. (1990) 23. 10 R. Le Van Mao and J. Yao, Appl. Catal. 79 (1) (1991) 77. 11 L. Dufresne, J. Yao and R. Le Van Mao, "Novel Production Methods for Ethylene, Light Hydrocarbons, and Aromatics", L.F. Albright, B.L. Crynes and S.Nowak Eds., Chem. Ind. 46, Marcel Dekker Inc., New York (1992) 509. 12 J. Yao and R. Le Van Mao, Catal. Lett. 11 (1991) 191. 13 R. Le Van Mao, L. Dufresne and J. Yao, Appl. Catal. 65 (1990) 143. 14 R. Le Van Mao, J. Yao and R. Carli, Appl. Catal. A: 86 (19921, 127. 15 R. Le Van Mao, R. Carli, J. Yao and V. Ragaini, Catal. Lett, 16 (1992) 43. 16 R. Carli, R. Le Van Mao, C. Bianchi and V. Ragaini, Catal. Lett., submitted. 17 a) J. Yao, Ph.D. Thesis, Concordia University, August 1992; b) R. Le Van Mao and J. Yao, U.S. Patent 5 135 898 (1992). 18 a) L. Dufresne, Ph. D. Thesis, Concordia University, April 1992; b) R. Le Van Mao and L. Dufresne, U.S. Patent 4 975 402 (1990). 19 G.A. Olah, J. Amer. Chem. SOC.94 (1972) 808. 20 N.S. Gnep, J.Y. Doyemet, A.M. Seco, F.R. Ribeiro and M. Guisnet, Appl. Catal. 35 (1987) 93. 21 M. Guisnet, D. Aittaleb, J.Y. Doyemet and N.S. Gnep, ACS Symp. on Alkvlation, Aromatization. Oliaomerization of Short Chain Hvdrocarbons, New York (199.11 668. 22 J.L. Harris, N. Krisko and X.M. Wang, Appl. Catal. A: 83 (1992) 59 and refs. 23 P. Meriaudeau and C. Naccache, J. Mol. Catal. 59 (1990) L31. 24 C.R. Bayense, A.J.H.P. van der Pol and J.H.C. van Hoof, Appl. Catal. 72 (1991) 81. 25'G. Buckles, G.H. Hutching and C.D. Williams, Catal. Lett. 11. (1991) 89. .
I
143
H drogen back spillover phenomena in the aromatization of light al anes over hybrid catalysts. Mechanistic considerations
1
Raymond Le Van Mao '*I1, Vittorio Ragain?
Jianhua Yao', Louise Dufresne', Riccardo Cadi2 and
Concordia University, Department of Chemistry i 3 Biochemistry, Laboratories for Inorganic Materials and Catalysis Research Laboratory, 1455 De Maisonneuve Blvd. W., Montreal (Quebec) H3G 1M8 Canada. University of Milan, Department of Physical Chemistry and Electrochemistry, Via Golgi 19, 2013 3 Milano, Italy
Abstract
By using irreducible oxides as cocatalysts or cocatalyst supports for the preparation of hybrid catalysts, it was possible t o show that the final steps of the light alkane aromatization involved a significant dehydrogenation activity by hydride abstraction over the zeolite acidic sites.
1. INTRODUCTION Aromatization of light alkanes (propane, butanes, liquefied petroleum gas) is an important industrial process and an exciting topic of research in the field of I is a gallium ZSMcatalysis. The catalyst used in the commercial Cyclar process [l 5 zeolite which is prepared according to the classical bifunctional catalysis, e.g. that gallium species are incorporated into the zeolite pore system in such a manner that each of the metal sites is adjacent t o an acid site. Such a catalyst can achieve high yields of aromatics (mostly, BTX, which are namely benzene, toluene and xylenes)[2,31. Some studies were specifically carried out in order t o elucidate the reaction mechanism on the pure acidic ZSM-5 zeolite [41. Recently, a new explanation based on the concept of hydrogen back (or reverse) spillover (HBS) was devised for the enhancement of aromatization of light alkenes and alkanes over hybrid catalysts 15-131. Such catalysts were prepared basically by intimately mixing the ZSM-5 zeolite particles with the Zn or Ga containing cocatalyst aggregates, and the resulting solid mixture was embedded in a clay matrix. Irreducible oxides such as quartz, which did not contain any Zn, Ga or Pt, were also capable of enhancing the aromatizing properties of the ZSM-5 zeolite [9,121. According t o the HBS concept, the cocatalyst acts as a hydrogen sink for the entire aromatization reaction, and such an action is exerted at the pore openings of the zeolite particles, e.g. at a fairly large distance from the zeolite acid sites [l 1,131. Recently, it was found that the the state of gallium species located at the "zeolite-cocatalyst particles" interface within the gallium containing hybrid catalysts had a great influence on the aromatizing performance [14-161. Thus, this paper aims at providing (a critical report of the) experimental evidence of the effects that the cocatalyst could have on the reaction mechanism.
I44
2. EXPERIMENTAL 2.1 PREPARATION AND CHARACTERIZATION OF CATALYSTS The parent ZSM-5 zeolites used for the preparation of t w o series of catalysts were synthesized according to the known technique. The powder acid forms were obtained by ion-exchange with ammonium ions and subsequent activation in air at elevated temperature 151. The Si/AI atomic ratios of the resulting zeolites, HZ and HZ , were 36 and 34, respectively. Final catalyst forms of the parent zeolites were obtained by extruding the zeolite powder with bentonite 151. i) First series of catalysts prepared from Hi!: Hybrid catalysts which comprised a ZSM-5 zeolite component (HZ) and a cocatalyst were obtained b y mixing mechanically these t w o powders and then extruding with bentonite clay as described in refs. 19-12,141. The cocatalysts used were Quartz (Qz), evaporated Ludox silica (LuSi) and Ga/Ludox silica coevaporation (Ga/LuSi), as mentioned in refs. 19,171. The resulting hybrid catalysts were called HZ//Qz, HZ//LuSi and HZ//Ga/LuSi, respectively. The Cyclar-type bifunctional catalyst was prepared by refluxing the HZ powder with a solution of gallium nitrate, and then evaporating the liquid, as described in ref. 151. The final catalyst, also called HZ/Ga(R,E) was obtained by extruding the resulting solid with bentonite. The Ga,O, content of such a catalyst was 3 wt %. ii) Two hybrid catalysts containing a Zn oxide precipitate and a Zn-Al oxide (1/ I coprecipitate, respectively, were also used. The preparation of such catalysts, also called HZ'//Zn and HZ'//Zn-AI, respectively, was fully described in refs.[8,181. 2.2 CATALYST TESTING The experimental set-up, the procedures for catalytic testing and analysis of products were similar t o those described in refs. 15,8-121. In particular, the reaction temperatures were = 500 "C for propylene and 540 "C for n-butane. The total conversion of n-butane or propylene, C,, the selectivity in product hydrocarbon i, Si, and the hydrogen yield, Y, were defined in refs. 114, 15, 17, 181. ,Y, is the ratio of the moles of hydrogen produced t o the number of carbon atoms of aromatics produced, e. g. Y ,, = YS /, , (Ar for aromatics). C ,, is the average number of carbon atoms in the product aromatics, the determination of which was done on the basis of the distribution of such hydrocarbons provided by gas chromatography (GC equipped with ionization and mass detectors)., ,Y , is the ratio of the moles of hydrogen produced t o the moles of aromatics produced.
3. RESULTS AND DISCUSSION Figure 1 shows the general reaction mechanism for the aromatization of light alkanes according t o the type of catalyst used. I)CASE (I)represents the sequential transformation of n-butane t o aromatics over the conventional bifunctional catalyst wherein the acid sites are adjacent to the metal (gallium or zinc oxide) sites. Therefore, as first step, n-butane undergoes dehydrogenation (DH) to the corresponding olefin which is then adsorbed on the protonic acid site providing thus the adsorbed carbenium ion species, (C4H9)+.An hydrogen molecule is liberated. Then, through a well-known sequence of reactions (oligomerization, cyclization/rearrangement), a cyclic carbenium ion was formed. At that stage, the aromatization occurs by means of an hydrogen transfer (TR) between the aromatizing species and a carbenium ion. This results in the formation of paraffins including n-butane. However, if strong dehydrogenating (Ga, Pt or Zn) sites are present in sufficient number, aromatics are now predominantly formed by dehydrogenation of the naphthenic intermediates, with thus more hydrogen
145
CASE (11)
CASE (I) Par
CqH10
Par
-.-.- t
'qH1O
I
i
13+
I
I
!
!
I
_._._. - !
C4H9 +
!
c
H2
(d
i2
AROMATICS
AROU~TICS
A2
Figure 1 . Reaction pathways according to the catalyst nature: (I)= bifunctional catalyst; (11) = pure zeolite catalyst; (Ill)= hybrid catalyst. (DH): dehydrogenation; (HAb1:hydride abstraction; (0C):oligomerizationlcyclization; (NA):transformation of naphthenic intermediates t o aromatics; (TR):hydrogen transfer; Par:paraffins other than n-butane; H,':(probably semi-sorbed) hydrogen species. produced (Figure 1 ) . The primary step is undoubtedly the dehydrogenation of nbutane t o butenes because in our previous work [lo], the hydrogen t o aromatic yield ratios (called Y ,, and YHM,) and the yield of butenes both went through a maximum when the reaction contact time tended t o zero. 11) CASE ( 1 1 ) represents the sequential transformation of n-butane over the pure ZSM-5 zeolite having only acid sites. As a first aromatization step, there is an adsorptionlhydride abstraction over the protonic site, resulting in the formation , at first, of a carbonium ion [ I 91 and then, by decomposition, an adsorbed (C,H,)+ species with the release of an hydrogen molecule. The latter is probably semiadsorbed. In fact, in our previous work [lo], there was no maximum observed for the yield of butenes (thus, butenes were not the primary products) and for the Y,, ratio, when the reaction contact time tended to zero. Because of such an absence of dehydrogenating sites, the paraffins (mainly, n-butane) formed during the aromatics-forming step will diffuse out without any further reaction. Thus, we ratio are much have the situation in which the n-butane conversion and the ,Y, lower than that obtained with the bifunctional catalyst. This is clearly shown in Table 1 wherein HZ is the parent ZSM-5 zeolite and HZ/Ga(R,E) represents the bifunctional Ga bearing catalyst. Table 1 Aromatization performance of the catalysts investigated CATALYST
ct
SA,
YH
HZ HZ//QZ HZ//LuSi HZ/Ga(R,E) HZlIGalLuSi
82.5 90.0 90.5 82.3 99.7
20.1 33.8 41.7 49.2 67.5
0.07 0.1 1 0.17 0.23 0.34
Y ,,
CA,
YHMA
0.34 0.36 0.41 0.47 0.50
7.19 7.13 7.01 6.98 6.94
2.44 2.53 2.87 3.28 3.47
146
Such detrimental effect of hydrogen transfer in the absence of strong dehydrogenating sites is also evidenced by the important formation of propane in the aromatization of propylene (no need of a hydride abstraction as initiation step: we are n o w focussed only on the last aromatization steps) as reported in Table 2. Table 2 Propylene aromatization over H-ZSM-5 zeolite and hybrid catalysts C, (C-atom %)
HZ'
95.3
Product selectivities (C-atom %) methane 2.7 ethylene 4.7 ethane 4.1 propane 30.5 butanes 8.4 butenes 2.5 C,+ aliphatics 4.1 aromatics 43.0
HZ'//Zn-AI
HZ'//Zn
3.1 2.9 5.5 10.5 3.4 1.6 4.4 68.6
4.3 1.4 12.0 14.0 2.1 0.9 0.8 64.5
96.2
96.5
Ill) For CASE (Ill),let us consider at first the catalytic properties of the hybrid catalyst, HZ//LuSi, whose cocatalyst was the evaporated colloidal Ludox silica. Such a cocatalyst could eventually intervene only at the level of the pore openings of the zeolite particles. If the spillover concept was effective in that case, the resulting more efficient removal of hydrogen (and maybe also other reaction products) induced by the presence of the cocatalyst, should result in a higher conversion of n-butane and a higher yield in aromatics. This was in effect obtained experimentally (Table 3). However, since there were not any dehydrogenating sites ratios (Table 1 ) in this hybrid catalyst, the significant increase in the ,Y, and Y, indicates that the hydrogen transfer (which was normally detrimental for the net hydrogen production) t o the (C,H,)+ ion or other olefinic carbenium ion was significantly reduced. Therefore, we have t o admit that some cyclic species undergo dehydrogenation t o aromatics by losing their hydrogen in the same manner as n-butane in its activation step, e.g. by hydride abstraction over the zeolite acid sites. All this is depicted in Figure 1 , CASE (111). Such an explanation is also supported b,y the data obtained with propylene as feed. In fact, when the hybrid catalyst HZ //Zn/Al or HZ //Zn was used, more aromatics were obtained (Table 2) and less propane was formed, indicating that, also in this case, the hydrogen transfer process was seriously reduced when the hydrogen removal became more efficient. Since the hydrogen back spillover action involves the "sink effect", e.g. the capacity of the cocatalyst surface to accept the spilt-over hydrogen species, the textural properties of the latter appear to be important factors. This was verified when we compared the catalytic properties of the HZ//Qz and HZ//LuSi hybrid catalysts. The latter one was more active because i t contained a cocatalyst having a larger surface area and a more homogeneous pore system than those of the quartz in the HZ//Qz hybrid catalyst [91. It is worth reminding that bentonite clay, quartz and the evaporated Ludox silica did not show any dehydrogenating properties with respect to n-butane [ I 71 and those of the Zn/Al co-precipitate were negligible under the reaction conditions investigated [ 181. Also in connection with CASE (Ill) of Figure 1 , let us consider n o w the
catalytic properties of the hybrid catalyst, HZIIGaILuSi, which contained the
147
gallium-silica cocatalyst. Such a catalyst was tested with n-butane and the following results were obtained. i)The behaviour of such a hybrid catalyst was similar to that of the parent ZSM-5 zeolite in the tests carried out a t low contact times (no maximum for the Y,, ratio and the yield of butenes), thus indicating that the n-butane aromatization reaction was not initiated by the dehydrogenation of n-butane as in the case of the conventional bifunctional catalyst 1201. It is worth mentioning that the galliumsilica cocatalyst alone did not show either any significant dehydrogenation activity with respect t o n-butane [ I 1,131; ii)The gallium transfer from the cocatalyst body to the zeolite particles, even upon drastic treatments with hydrogen, was fairly small, not exceeding 0 . 2 wt % 11 41. As reported in Table 1, the aromatizing properties of the HZ//Ga/LuSi hybrid catalyst were very high. The Y,, and Y, ratios increased with increasing aromatizing properties, independently from the presence or absence of gallium, according to the following sequence: parent zeolite c gallium-free hybrid catalysts C gallium-containing bifunctional catalysts c gallium-containing hybrid catalysts. Basically, such results strongly suggested the existence of dehydrogenating activity over acid sites in the very last steps of the aromatization. iii) Another experimental evidence for such promoting effect of the cocatalyst came from the following observation. In addition to undergoing aromatization according to the previously mentioned "direct" pathways, any alkane molecule when adsorbed on a relatively strong acid site is submitted to cracking reactions. The first one is the cracking of n-butane into methane and propylene. The latter may be converted rapidly to benzene or to other products (by further cracking, alkylation or hydrogen transfer). Table 3 reports the values of benzene and methane yields, Y,,, and Y,,, respectively. Let us make the assumption that benzene be predominantly produced by aromatization of propylene. Since the total yield of methane is always higher than the yield of methane in the cracking of nbutane t o propylene and methane, the ratio RAR = Y,EN/ 2 YM, may be considered as a parameter which is capable of giving an idea about the rate of transformation of naphthenic intermediates into aromatics. As shown in Table 3 and in agreement with the observations of Guisnet et al [211, the bifunctional catalyst was more efficient in the transformation of naphthenes into aromatics (N--A) than the pure zeolite. However, our hybrid catalyst HZ//LuSi which did not contain any gallium, had practically the same value of RAR as the bifunctional catalyst. Results of Table 1 (YHAand YHMA) and Table 3 (RAR) allows us t o say that the LuSi cocatalyst was actually activating such a transformation (N - A) by accelerating the removal of hydrogen and other products. Thus, this process liberated more acid sites for other naphthenic intermediates which could in turn undergo aromatization by hydride abstraction [(Ill in Figure 11. Table 3 Effets of the cocatalyst on the rate of transformation of the naphthenic intermediates into aromatics CATALYST HZ HZ//Qz H Z//Lu Si HZ/Ga (R, E) HZ//Ga/LuSi
',EN
3.91 8.20 11.81 13.36 25.73
YME
RAR
4.61 8.21 6.52 7.65 10.06
0.43 0.50 0.91 0.88 1.28
I48
Therefore, the higher stability and the higher dispersion of the gallium species incorporated into the cocatalyst [9,10,141 resulted in a higher activity, a much shorter induction period (to reach the maximum aromatizing activity without any reducing pretreatment) and a higher "on-stream" stability for the HZ//Ga/LuSi hybrid catalyst. This could not merely be ascribed t o the gallium migration into the zeolite particles which would create a situation similar t o that of a bifunctional catalyst. In fact, by using the technique of surface transfer 11 51, gallium oxide was deposited o n the external surface of the ZSM-5 zeolite particles. Although such gallium amounts were small (0.4-0.6 wt %), the "contaminated" zeolite particles showed an aromatizing activity as high as the best bifunctional catalysts. Since the required gallium content for a perfect bifunctional situation is 1.5 wt % and if w e admit that half of the amount of gallium deposited onto the zeolite external surface moved t o the inside (e.g. 0 . 2 wt %), would we be ready t o accept that 15 % of the bifunctional sites could behave like a 100 % bifunctional configuration ? By using X-ray photoelectron spectroscopy and hydrogen adsorption/temperature programmed desorption technique [161, w e have recently shown that there were t w o hydrogen chemisorption sites for the gallium component: the highest aromatizing performance corresponded t o the hybrid catalyst which exhibited the largest number of stronger chemisorption sites (namely S as opposed to the weaker sites, SHJ. These sites were associated with high 6 a dispersion. On the other hand, it was shown that the apparent activation energy was the same for the pure ZSM-5 zeolite and the gallium containing catalysts (both bifunctional and hybrid catalysts) [lo]. It means that all the three reaction pathways (I, I I and I l l ) probably have the same rate-determining step (RDS) as depicted in Figure 1. In the case of a bifunctional catalyst, the RDS does also include a dehydrogenation and an immediate protonation t o (C,H,)' ion, and not only the dehydrogenation process as accepted by reference [221. This thus implies the existence "in duo and at short distance" of these t w o acid and metal (as oxide) sites. Such a short distance requirement for a perfect bifunctional configuration may be the origin of observations made by several authors about possible interactions between the t w o active sites [23-251. Other experimental evidence of the beneficial effect of the cocatalyst on the rate of diffusion of hydrogen (and other reaction products) was provided by the following results. i) The amount and nature of the coke deposited on the pure zeolite HZ and the hybrid catalyst HZ//Ga/LuSi were also investigated [17al). It was found that under normal reaction conditions with n-butane as feed, on-stream aging the catalysts for 22 h led to the following carbonaceous species deposits, as analysed by thermal gravimetric method: 1.3 wt % for the HZ zeolite (average combustion temperature: 640 "C) and 0.8 wt YO for the hybrid catalyst (average combustion temperature : 530 "C). It was also observed by means of CP/MAS 13C NMR spectroscopy method, that the coke deposited on the pure zeolite was predominantly aromatic in nature (advanced coke, 100-16 0 ppm region) while that deposited o n the hybrid catalyst was predominantly paraffinic/olefinic (light coke, 0-50 ppm region) [17 a]. ii) The average number of carbon atoms related t o the aromatic products (C,J in the case of the hybrid catalysts was lower than that obtained with the parent zeolite (Table 2). Moreover, the selectivity to naphthenic products was higher for the hybrid catalyst than for the parent zeolite while the selectivity to C, and C,, aromatics was much lower [17a]. The only possible explanation stems on a faster removal of hydrogen and other products, thus preventing any excessive formation of heavy coke which normally occurs at the pore mouths of the pure zeolite.
149
4. CONCLUSION We have shown that the reaction pathway followed by the hybrid catalyst, is mainly based on the hydride abstraction process not only in the n-butane activation step but also in the transformation of naphthenes into aromatics. The latter is considerably enhanced by the hydrogen back spillover action exerted by the cocatalyst itself. In addition, when gallium cocatalyst is used, the configuration of gallium sites on the "zeolite / cocatalyst particles" interface appears t o be the key factor for the high aromatization performance of the hybrid catalysts.
5. ACKNOWLEDGEMENTS We would like t o thank NSERC of Canada and Quebec's Actions Structurantes Program for their financial support. We also thank Dr. Georges Denes for helpful discussions and Mr. Bernard Sjiariel for technical assistance.
6. REFERENCES 1 2 3 4
P.C. Doolan and P.R. Pujado, Hydroc. Process., (Sep. 1989) 72. Y. Ono, Catal. Rev.-Sci. Eng., 34(3). (1992) 179. M. Guisnet and N.S. Gnep, Appl. Catal. A, 89 (1992) 1. M. Guisnet, N.S. Gnep, D. Aittaleb and Y.J. Doyemet, Appl. Catal. A, 87 (1992) 255 and references therein. 5 J. Yao, R. Le Van Mao and L. Dufresne, Appl. Catal. 65 (1990) 175. 6 T. Inui, "Successful1 Design of Catalysts", T. lnui Ed., Studies in Surface Science and Catalysis 44, Elsevier, Amsterdam (1989) 189. 7 K. Fujimoto, I. Nakamura and Y. Yokota, Proc. Second International Conference on Spillover, Leipzig (Germany), June 1989, 176. 8 R. Le Van Mao and L. Dufresne, Appl. Catal. 52 (1989) 1. 9 R. Le Van Mao, J. Yao and B. Sjiariel, Catal. Lett. (1990) 23. 10 R. Le Van Mao and J. Yao, Appl. Catal. 79 (1) (1991) 77. 11 L. Dufresne, J. Yao and R. Le Van Mao, "Novel Production Methods for Ethylene, Light Hydrocarbons, and Aromatics", L.F. Albright, B.L. Crynes and S.Nowak Eds., Chem. Ind. 46, Marcel Dekker Inc., New York (1992) 509. 12 J. Yao and R. Le Van Mao, Catal. Lett. 11 (1991) 191. 13 R. Le Van Mao, L. Dufresne and J. Yao, Appl. Catal. 65 (1990) 143. 14 R. Le Van Mao, J. Yao and R. Carli, Appl. Catal. A: 86 (19921, 127. 15 R. Le Van Mao, R. Carli, J. Yao and V. Ragaini, Catal. Lett, 16 (1992) 43. 16 R. Carli, R. Le Van Mao, C. Bianchi and V. Ragaini, Catal. Lett., submitted. 17 a) J. Yao, Ph.D. Thesis, Concordia University, August 1992; b) R. Le Van Mao and J. Yao, U.S. Patent 5 135 898 (1992). 18 a) L. Dufresne, Ph. D. Thesis, Concordia University, April 1992; b) R. Le Van Mao and L. Dufresne, U.S. Patent 4 975 402 (1990). 19 G.A. Olah, J. Amer. Chem. SOC.94 (1972) 808. 20 N.S. Gnep, J.Y. Doyemet, A.M. Seco, F.R. Ribeiro and M. Guisnet, Appl. Catal. 35 (1987) 93. 21 M. Guisnet, D. Aittaleb, J.Y. Doyemet and N.S. Gnep, ACS Symp. on Alkvlation, Aromatization. Oliaomerization of Short Chain Hvdrocarbons, New York (199.11 668. 22 J.L. Harris, N. Krisko and X.M. Wang, Appl. Catal. A: 83 (1992) 59 and refs. 23 P. Meriaudeau and C. Naccache, J. Mol. Catal. 59 (1990) L31. 24 C.R. Bayense, A.J.H.P. van der Pol and J.H.C. van Hoof, Appl. Catal. 72 (1991) 81. 25'G. Buckles, G.H. Hutching and C.D. Williams, Catal. Lett. 11. (1991) 89. .
I