Deactivation of MgyAlOx mixed oxides during aldol condensation reactions of ketones

Studies in Surface Science and Catalysis, Vol. 139 J.J. Spivey, G.W. Roberts and B.H. Davis (Editors) 9 2001 Elsevier Science B.V. All rights reserved.

303

Deactivation of MgyAlOx mixed oxides during aldol condensation reactions of ketones V.K Diez, C.R. Apesteguia and J.I. Di Cosimo Instituto de Investigaciones en Catfilisis y Petroquimica-INCAPE- (UNL-CONICET). Santiago del Estero 2654, (3000) Santa Fe, Argentina

Abstract The deactivation of Mg-AI mixed oxides during the gas-phase self-condensation of acetone was studied. Main aldol condensation reaction and secondary coke-forming reactions take place on both basic and acidic sites. Although deactivation is caused by carbon deposition, it was found that coke formed on basic sites rather than on acidic sites is responsible for the activity lost. 1. INTRODUCTION Mg-A1 mixed oxides obtained by thermal decomposition of anionic clays of hydrotalcite structure, present acidic or basic surface properties depending on their chemical composition [1]. These materials contain the metal components in close interaction thereby promoting bifunctional reactions that are catalyzed by Br6nsted base-Lewis acid pairs. Among others, hydrotalcite-derived mixed oxides promote aldol condensations [2], alkylations [3] and alcohol eliminations reactions [1]. In particular, we have reported that Mg-A1 mixed oxides efficiently catalyze the gas-phase self-condensation of acetone to ot,J3unsaturated ketones such as mesityl oxides and isophorone [4]. Unfortunately, in coupling reactions like aldol condensations, basic catalysts are often deactivated either by the presence of byproducts such as water in the gas phase or by coke build up through secondary side reactions. Deactivation has traditionally limited the potential of solid basic catalysts to replace environmentally problematic and corrosive liquid bases. However, few works in the literature deal with the deactivation of solid bases under reaction conditions. Studies relating the concerted and sequential pathways required in the deactivation mechanism with the acidbase properties of the catalyst surface are specially lacking. In this work, we studied the deactivation of Mg-Al mixed oxides in the gas phase oligomerization of acetone. We prepared and characterized calcined Mg-A1 hydrotalcites with Mg/Al atomic ratios of 1-9. The effect of composition on both the surface and catalytic properties and the catalyst deactivation was investigated by combining several characterization methods with catalytic data. 2. EXPERIMENTAL Mg/AI hydrotalcite with Mg/A1 atomic ratios (y) of 1, 3, 5 and 9 were prepared by coprecipitation ofMg(NO3)2.6 H20 and Al(NO3)3.9 H20 at 333 K and pH = 10. After drying

304 overnight at 348 K, precursors were decomposed in N2 at 673 K in order to obtain the MgyA1Ox mixed oxides. More experimental details are given elsewhere [1]. Reference samples of A1203 and MgO were prepared following the same procedure. The crystalline structure of the samples was determined by powder X-ray diffraction (XRD) methods. Acid site densities were measured by temperature-programmed desorption (TPD) of NH3 preadsorbed at room temperature. The structure of CO2 chemisorbed on hydrotalcite-derived samples was determined by infrared spectroscopy (IR). CO2 adsorption site densities were obtained from TPD of CO2 preadsorbed at 298 K. Total surface areas (Sg) were measured by N2 physisorption at 77 K. Self-condensation of acetone was carried out at 473 K and 100 kPa in a flow system with a differential fixed-bed reactor. Acetone was vaporized in 1-12(H2/acetone = 12) before entering the reaction zone. The standard contact time (0c) was 0.84 g of cat h/g of acetone. Main reaction products were mesityl oxides (MO's), isophorone (IP) and mesitylene (MES). Traces of phorone and light hydrocarbons were also identified. The coke formed on the catalysts was characterized after reaction, ex-situ, in a temperature-programmed oxidation (TPO) unit. The TPO experiments were carried out in a microreactor loaded with 50 mg of catalyst and using a 3 % O2/N2 cartier gas. Sample temperature was increased linearly from room temperature to 973 K at 10 K/min. The reactor exit gases were fed into a methanator operating at 673 K to convert COx in methane and then analyzed by flame ionization detector. 3. RESULTS AND DISCUSSION 3.1. Catalyst characterization The hydroxycarbonate precursors showed diffraction patterns consistent with the layered double hydroxide structure proposed for hydrotalcite-like compounds, i.e., brucitelike sheets formed by OH groups and cations (Mg2+ and A13+) in octahedral coordination. Water molecules and charge-compensating CO32 anions are located in the interlayers. The hydrotalcite-like precursor composition is [Mgl_rAlr(OH)2]r+(co3)~.mn20 with r = A1/(AI+Mg); the stoichiometric hydrotalcite compound, Mg6AI2(OH)16CO3.4H20, is obtained for r = 0.25. Upon thermal decomposition at 673 K, the hydroxycarbonate precursors lose water and CO2. This gaseous evolution Table 1 develops a significant Chemical composition, Crystalline Structure, Surface Area porous structure and gives rise to relatively and Acid-base Properties ofMgyAIOx, MgO and A1203 Oxides. high surface area mixed Site Density Catalyst r = A1/(AI+Mg) Crystalline Sg oxides as shown in (molar) ~ Phase b (m2/g) (~tmol/m2) Table 1. In the nb c naa compositional range of MgO 0.00 MgO 143 3.13 0.51 this study, the analysis Mg9A1Ox 0.11 MgO 114 1.17 0.81 by X R of the MgyA1Ox Mg5A1Ox 0.18 MgO 184 0.46 0.85 oxides showed only one Mg3A1Ox 0.24 MgO 238 0.59 1.02 crystalline phase, MgO 1.57 MglA1Ox 0.47 MgO 231 0.83 periclase (ASTM 4A1203 1.00 Amorphous 230 0.34 1.34 0829). No traces of "by AAS; b by XR ; Cby TPD of C O 2 ; d by TPD of NH3 hydrotalcite phases were

305 detected after the thermal decomposition at 673 K. Table 1 presents the acid and base site densities measured for all the samples by TPD of preadsorbed NH3 and CO2, respectively. The base site density (nb) of the MgyA1Ox oxides was always between the values measured for MgO and A1203. Basic sites of different chemical nature were observed by FTIR of CO2 and deconvolution of the CO2 TPD traces: isolated low-coordination 02. ions, 02. in metal-oxygen pairs and OH groups. The acid site density (na), on the other hand, increases with increasing A1 content. By deconvolution of the NH3 TPD traces it was found that the MgyAIOx oxides contain both BrOnsted (OH groups) and Lewis (metal cations) acid sites.

3.2. Sample composition, surface acid-base properties and catalytic activity Acetone aldol condensation proceeds on either acidic or basic catalysts. On basic catalysts, the reaction products are mainly ot,J3-unsaturated 10 ketones [5] whereas on acidic materials formation of aromatics and olefins is ~"~' favored [6]. In our catalytic tests, the main reaction products were mesityl oxides (MO's) and isophorone (IP). MO ~ 1 is formed from the initial selfcondensation of acetone whereas IP is a ~" "~~~/Po secondary product arising from the consecutive aldol condensation between MO and acetone. Over all the samples the .] . . . . . . . . . 9 reaction rate diminished as a function of 10 0 2 4 6 8 time-on-stream as shown in Fig. 1 for the Time (h) MggA1Ox sample which lost about 60 % of its initial activity aRer 10 h-run. Initial Fig. 1: Time dependence of the acetone reaction rates (r~ and product selectivities conversion and product formation rates on (S ~) were calculated by extrapolating the MggA10,, (T = 473 K, P = 100 kPa) reaction rates vs. time curves to zero. In Fig. 2 we have represented both the r0Acetonevalues and the total site density (nT) as a function of catalyst composition. Qualitatively, the variation of r~ e with increasing A1 content is similar to that followed by nT thereby suggesting that acetone conversion depends on both acid and base sites. Pure MgO was the most active catalyst whereas A1203 showed the lowest activity. This is because A1-O pairs are much less active than Mg-O pairs for promoting the proton abstraction and carbanion stabilization steps involved in aldol condensation reactions. We have showed [1] that the acetone aldolization rate is controlled on basic catalysts by the number of metal cation-oxygen anion surface pairs. Mg-rich MgyA1Ox oxides are less active than MgO because they exhibit a lower base site density and also poor acidic properties. In contrast, Al-rich MgyA1Ox oxides are more active than A1203 due to a proper combination of acid and base sites. In Fig. 3 the initial product selectivities measured at isoconversion are represented as a function of catalyst composition. It is observed that the initial selectivities toward IP and MES increase with increasing the AI content whereas high selectivity toward MO's was obtained on Mg-rich catalysts. The presence of MES in low concentrations on Al-rich

306 samples is indicative of an acid-catalyzed reaction pathway [6]. 45 40 ~= ~ 35 ~

o~

2.0

25 ~

~ 40

1.6

20 ~

~-~ 1.2

15 ~

0.8

0.0

012

014

,

016

018

lO 1.0

Al/(A1 + Mg)

IP

80

3"61 3.2 2.8

~

60

~ 20 r~

MO's

"'~

04-~~,~, 0.0 0.2

MES

~--~-------T----------,------,---q 0.4 0.6 0.8 1.0 A1/(AI+Mg)

Fig. 2" Initial acetone conversion rate and total number of active sites as a function of composition of MgyA1Ox oxides (T = 473 K, 0c = 0.84 h)

Fig. 3" Initial product selectivity as a function of composition on MgyA1Ox catalysts (T = 473 K, X~ = 12-14 %)

3.3. Coke formation and characterization In a previous work [7], we found that MgO-based catalysts are deactivated during acetone oligomerization because of the formation of carbon deposits that block pores and surface active sites. Here, the coke formed on the samples after the 10-h catalytic runs was characterized by TPO technique. The obtained TPO curves are shown in Fig. 4 and quantitative results are summarized in Table 2. The amount of carbon measured on the samples was between 12 and 22 wt. % and, as a general trend, MgyA1Ox and A1203 formed more coke than pure MgO. The TPO curves of Fig. 4 show the existence of two different carbon deposits; the main one gives a broad oxidation peak centered at 650-700 K (lowtemperature peak), and the other one arises as a shoulder at temperatures above 720 K (hightemperature peak). The low-temperature peak maximum shifts to higher temperatures with increasing A1 content in the samples (Table 2). In order to obtain more insight on the nature of the carbon deposits, stripping experiments in pure N2 were carried out. In these experiments, coked samples were first treated in N2 from room temperature up to 973 K at 10 K/min before performing standard TPO runs. The TPO curves obtained after the N2 treatment are identified here as TPO(N) and are displayed in dotted lines in Fig. 4. The amount of carbon oxidized in TPO(N) experiments was between 4 and 8 wt % (Table 2). The shape and temperature maximum of the TPO(N) curves were similar to those of the high-temperature peaks in the corresponding TPO curves. These results show that the nitrogen treatment removes almost completely the carbon deposits that account for low-temperature peaks in TPO traces of Fig. 4. Both kind of carbon deposits seem to be related to the catalyst acid-base properties since both peak maxima gradually shifted to higher temperatures at increasing A1 contents as presented in Table 2. In previous work [7] we postulated that the coke precursor species on MgO-based catalysts are highly unsaturated linear compounds, such as 2,6-dimethylhepta-2,5-dien-4-one

307 (phorone), 4,6-dimethylhepta-3,5-dien-2-one and 4,4'-dimethylhepta-2,6-dione which are trimeric compounds formed from MO and acetone by secondary aldol condensation reactions. These linear trimers are non-detectable in the gas phase during reaction because they remain bound to [ ' TPO TPO(N) the surface due to the strong interaction between the unsaturated chain and the basic centers. All of them are able to give IP by consecutive internal aldol or Michael rearrangements [8]. Selective formation of one trimer in preference to the others depends on the catalyst acid-base properties. Phorone is the preferred intermediate on basic catalysts whereas formation of MES on acid sites (Fig. 3) takes place > only from 4,6-dimethylhepta-3,5-dien-2-one [8]. To obtain more information on the heavy compounds present in trace amounts in the gaseous products, we ed) 9 i,,,~ performed a GC-Mass Spectroscopy analysis of the reactor effluent collected in a condenser during acetone aldolization on MglA1Ox catalyst. Besides acetone, MO's, IP and MES, we detected the trimer 4,6-dimethylhepta-3,5-dien-2-one (m/z = 138), tetramers like isoxylitones (m/z = 178) and also tetramethyltetralones (rn/z = 202) and other heavy polymers of m/z = 218 and 240. Results can be explained by considering that the catalyst surface acid-base properties determine the preferential formation of a given trimeric 400 500 600 700 800 900 intermediate which in turn defines both the Temperature(K) formation of the final product released to the gas Fig. 4: Coke burning experiments phase (MO, IP or MES) and the nature of the carbon atter acetone aldol condensation at deposit. This interpretation is depicted in Fig. 5. Coke formed from the 4,6-dimethylhepta-3,5-dien-2473 K on MgyA1Ox catalysts. one intermediate on acidic catalysts will be r = A1/(A1 + Mg) ......

Table 2 Coke Characterization Catalyst

MgO MggAIOx MgsA1Ox Mg3AIOx MglA1Ox A1203

TPO TPO(N) Carbon Low-Temperature Carbon Peak Amount Peak Amount Temperature (wt %) (K) (wt %) (K) 12.5 12.1 13.6 13.6 22.4 15.8

623 625 650 642 697 709

4.7 5.2 55 50 7.6 6.0

730 724 736 750 760 787

fundamentally aromatic in composition since MES can further react to xylenes and other C9+ aromatic hydrocarbons [6,9]. On the other hand, coke formed on more basic materials from phorone, will be mostly heavy a,13-unsaturated ketones such as isoxylitones, tetralones and other C~2+ oxygenates [9,10]. Then, the gradual shift from one intermediate

308

to another causes a concomitant change in coke composition and could explain the higher temperatures needed to burn the carbon deposits at increasing AI content in the samples, as observed in Fig. 4.

Heavy a,13-unsaturated ketones on basic site pairs

t

Tetramers (isoxylitones)

~3 c=~ / CH s Acetone

H3C CH3 \/ C + Acetone 9~ CH II

Michael l Addition + + Acetone Aldol Cond.

Linear Trimeric .~ ~ Intermediates "1 - H20 (C 9 ketones) C =O - H20

Aldol Cond. I Acidic orBasic CH 3 Sites

+ Acetone

Aldol Cond. Acidic or Basic Sites

O Michael Addition or Aldol Cond. m.. Acidic orBasic m,Sites

H3C H3C

IP

CH 3

i Aldol l Acidic Cond. ~ Sites

MO CH3

H3C

+CHs IVIES

Aromatic coke on acidic site pairs Fig. 5: Reaction network for acetone oligomerization and coke formation on MgyA1Ox

3.4. C a t a l y s t d e a c t i v a t i o n

The catalytic activity is defined as a = rt/ro where rt and ro are the reaction rates at time t and zero time respectively. Plots of the catalytic activity as a function of time on stream (Fig. 6) showed that deactivation of the Mg-A1 mixed oxides is faster than that of A1203 but slower than that of MgO. Pure MgO presented a rapid deactivation and its residual activity aider the 10-h run was very low. From the slope of the plots of Fig. 6, the initial deactivation rate was calculated as do = [- da/dt ]r--0. In Fig. 7, the do values obtained for all the samples are represented in open squares as a function of the carbon content measured after the catalytic runs (Table 2); clearly, it does not exist any correlation between do and the amount of coke. Initial deactivation is lower on A1203 (do - 0.14 h l ) than on MgO (do - 0.53 h l ) in spite that alumina forms more coke during reaction and that the coke is more difficult to oxidize as compared to MgO (Table 2 and Fig. 4). These results show that neither the coke amount nor its polymerization degree account for the catalyst deactivation order observed in Fig. 6. A better explanation is obtained by considering the nature of the surface sites that are responsible for the formation of coke precursors on pure A1203 or MgO. Alumina contains BrOnsted (OH groups) and Lewis (metal

309

1.0

AI20s

0.8 0.6 ~ 1,,,,,I

;>

"~ 0.4 < 0.2 0.0

~~ " r Mg_o_ . . . . . " ,. ~ ~ - - ~ , . v , 0 2 4 6 8 10 Time (h)

Fig. 6 Acetone condensation activity as a function of time over MgyA1Ox, MgO and A1203 catalysts (T = 473 K, 0c = 0.84 h) Carbon Amount (wt %) 0.6 '

12

16 |

I

20

//

,

I

"7 ,.~

0.5

9~

0.4

~

0.3

~

24

|

0.2 0.1

o

i

~

cations) acidic sites and probably coke forms preferentially on Br6nsted sites rather than on A1-O pair sites. Br6nsted acid sites are not active for gas-phase acetone aldolization and the reaction would proceed, therefore, at an appreciable rate on AI-O pairs even in the presence of a high carbon content. On MgO, the fast deactivation is due to the fact that aldolization and coke formation reactions compete for the basic Mg-O pairs. The aldol reaction stops almost completely even when the carbon deposition is less intense than on Al-containing samples. On MgyA1Ox oxides, coke will mainly form on metal-oxygen pairs (Mg-rich samples) or BrOnsted acid sites (AIrich samples) following the A1 content in the sample. In Fig. 7, the do values obtained for MgO, A1203 and MgyA1Ox oxides were plotted as a function of the density of basic sites, nb (Table 1). It is observed that the initial deactivation correlates linearly with the density of basic sites. No correlation was found between the NH3 site density (na, Table 1) and do. These results show that although MgyAIOx oxides promote the selfcondensation of acetone by both acid- and base-catalyzed pathways (Fig. 2), the initial deactivation rate is essentially related to the surface basic properties. This dependence of do with nb for MgyAIOx oxides probably reflects the fact that the acetone oligomerization rate decline is significantly higher when coke poisons very active basic Mg-O pairs than when eliminates moderately active acidic A1-O pairs.

Base site density, n b (i.tmol/m 2) 4. CONCLUSIONS Fig. 7: Initial catalyst deactivation of MgyAIOx oxides, MgO and Al203 as a function of the base site density and carbon deposit amount.

The activity and selectivity for acetone oligomerization on MgyA1Ox oxides depend on the catalyst acid-base properties, which in turn are determined by the aluminum content in the sample. Mg-rich catalysts selectively yield mesityl oxides whereas M-rich MgyAIOx oxides produce mainly isophorone. MgyAIOx samples are more active than alumina. Intermediate Mg/A1 compositions show unique catalytic properties because fulfil the reaction requirements for the

310 density and strength of basic and acid sites. The MgyAIOx activity declines in the acetone oligomerization reaction due to a blockage of both basic and acid active sites by a carbonaceous residue formed by secondary aldol condensation reactions. The key intermediate species for coke formation are highly unsaturated linear trimers that are formed by aldol condensation of mesityl oxide with acetone and remain strongly bound to the catalyst surface. The catalyst surface acid-base properties determine the preferential formation of a given trimeric intermediate, which in turn defines the chemical nature of the carbon deposit. Aromatic hydrocarbons are the main component of coke formed on acidic M-rich MgyA1Ox samples whereas heavy ~,13-unsaturated ketones preferentially form on basic Mg-rich catalysts. Alumina forms more and heavier coke than MgO but the latter deactivates faster. This is explained by the fact that on MgO primary aldol condensation reactions and formation of coke precursors species take place on the same basic sites. In contrast, alumina forms predominantly aromatic coke involving Br6nsted OH groups that are inactive for aldol condensation reactions. The initial deactivation rate of acetone oligomerization on MgyA1Ox oxides is essentially related to the surface basic properties. This is because the activity decline is significantly higher when coke poisons very active basic Mg-O pairs than when eliminates moderately active acidic A1-O pairs. ACKNOWLEDGEMENTS Support of this work by the Consejo Nacional de Investigaciones Cientificas y T6cnicas (CONICET), Argentina and by the Universidad Nacional del Litoral, grant CAI+D 94-0858-007-049, Santa Fe, Argentina, is gratefully acknowledged. REFERENCES 1. J.I. Di Cosimo, C.R. Apesteguia, M.J.L. Gin6s and E. Iglesia, J. Catal. 190 (2000) 261. 2. W.T. Reichle, J. Catal. 94 (1985) 547 3. S. Velu and C.S. Swamy, Appl. Catal., 145 (1996) 225. 4. J.I. Di Cosimo, V.K. Diez and C.R. Apesteguia, Appl. Clay Science 13 (1998) 433. 5. J.I. Di Cosimo, V.K. Diez and C.R. Apesteguia, Appl. Catal., 137 (1996) 149. 6. C.D. Chang and A.J. Silvestri, J. Catal. 47 (1977)249. 7. J.I. Di Cosimo and C.R. Apesteguia, J. Molec. Catal., 130 (1998) 177. 8. W.T. Reichle, J. Catal. 63 (1980) 295. 9. G.S. Salvapati, K.V. Ramanamurty and M. Janardanarao, J. Molec. Catal. 54 (1989) 9. 10. S. Lippert, W. Baumann and K. Thomke, J. Molec. Catal. 69 (1991) 199.