Effect of Coke Formation on the Reactions of Dimethyl Ether on Acidic

Prod. Res. Dev. 1983, 22, 242-246. Effect of Coke Formation on the Reactions of Dimethyl Ether on. Acidic Oxide Catalysts. John W. Evans, David L. Trl...
0 downloads 0 Views 639KB Size
242

Ind. Eng. Chem. Prod. Res. Dev. 1983, 2 2 , 242-246

Effect of Coke Formation on the Reactions of Dimethyl Ether on Acidic Oxide Catalysts John W. Evans, David L. Trlmm,' and Mark S. Wainwrlght School of Chemical Engineering & Industrial Chemistry, University of New South Wales, Kensington 2033, NS W, Australia

The reactions of methanol and dimethyl ether over y-alumina and a range of silica-aluminas have been studied. Significant quantities of methane, ethane, ethylene, and higher hydrocarbons can be produced, but the catalysts are rapidly deactivated by coke formation. The influence of coking on conversion and selectivity and on ammonia adsorption on the catalyst has been investigated in some detail. For small pore catalysts, coke seals off pore mouths, and selectivity depends mainly on conversion in a given catalyst system. For large pore catalysts, coke deposits in the pores and affects acid site strength distributions, conversion, and selectivity.

Introduction Although the conversion of methanol to hydrocarbons has been known to be possible for many years (Adkins and Perkins, 1928; Den0 et al., 1964; Venuto and Landis, 1968), interest in the process has intensified as a result of the development of zeolites that catalyze the conversion of methanol to gasoline (Chang and Silvestri, 1977;Derouane et al., 1980; Kaeding and Butter, 1980; Chang et al., 1979, Ono and Mori, 1981). A range of hydrocarbons can be produced by contacting the alcohol with acids (Adkins and Perkins, 1928; Den0 et al., 1964) or acidic solids (Chang and Silvestri, 1977; Ono and Mori, 1981; Dejaifve et al., 1981; Cormerais et al., 1981),the main advantage of the newer ZSM catalysts resting with the desirable product spectrum and the relatively low formation of coke on the catalysts. This appears to arise as a result of optimal acidity in the zeolite coupled to a pore structure that favors the production of gasoline. The exact mechanism of the reactions is still open to some question, although there is general agreement that the production of dimethyl ether is the first step in the conversion and that Bronsted acid sites are important. Chang and Silvestri (1977) originally suggested that the methylene diradical (:CH2)was an important intermediate, but later workers have suggested that ionic species are the predominant intermediates; these may or may not contain oxygen (Derouane et al., 1980; Kaeding and Butter, 1980; Anderson et al., 1979). The product spectra obtained over different acidic solids are determined by the nature of these intermediates, by the surface acidity, and by the pore structure of the solid. Since the methanol conversion is acid catalyzed, it is not surprising that coke is laid down as an unwanted byproduct which, by affecting the acidity and the pore structure, may affect the activity and selectivity of the main reaction (Dejaifve et al., 1981; Langner, 1982). The present studies were initiated in order to study the effect of coke formation on the reactions of methanol and dimethyl ether on a range of silica-aluminas of varying acidity. Preliminary studies show that methanol reacted almost completely via the formation of dimethyl ether. As a result, attention was focused on the relation between activity, selectivity, and coking in oxides capable of catalyzing the conversion of the ether to hydrocarbons. Experimental Section Studies of the reactions of methanol and dimethyl ether were carried out in a conventional flow apparatus fitted with an on-line gas chromatograph. The reactor contained ca. 5 g of catalyst and was located in a molten salt bath controlled at a pre-set temperature (It1 "C). Nitrogen was

passed over the catalyst at reaction temperature prior to introducing the dimethyl ether feed. Gases emerging from the reactor were separated using a composite column of 3 m Porapak P S with 0.4 m Porapak T, the column being temperature programmed between 40 and 140 "C. This combination allowed the separation of all products up to C6 olefins, including resolution of the emergent water peak. Hydrocarbons of carbon number up to C,, were analyzed after passage through a column consisting of 1.5 m , 5% OV-101 supported on Chromosorb and maintained a t 65 "C. The catalysts used in this study were kindly donated by the Davison Chemical Division of W. R. Grace and Co. and their characteristics are reported in Table I. Measurements of catalyst acidity have been the subject of considerable debate (Deeba and Hall, 1979). In the present studies, some measure of total acid site number was obtained based on adsorption of ammonia from pulses of gas passed over the catalyst at room temperature (Benesi and Winquist, 1978;Forni, 1973). Measurements were obtained for fresh and equilibrated catalysts. In the following section the conversion (% ) is defined as mol of DME leaving reactor x 100 - mol of DME entering reactor

)

and the selectivity ( % ) is given by mol of component i carbon no. of i X mol of DME converted carbon no. of DME 100

Results Initial studies were focused on the reactions of methanol over the different catalysts. Conversion became significant (>ca. 30%) at temperatures between 250 and 300 O C , with dimethyl ether being the major product a t high space velocities. Other products appeared to be produced only as a result of the further reactions of the ether. Attention was then focused on the reactions of dimethyl ether, and a typical product spectrum obtained on an equilibrated catalyst (see below) is shown in Figure 1. Small amounts of methanol were produced by the reaction of dimethyl ether with the water produced during the formation of hydrocarbons. No products of carbon number >2 were formed using y-alumina as a catalyst. Silica did not catalyze hydrocarbon formation from either methanol or dimethyl ether. Catalyst 970-13 produced methanol, methane, and carbon oxides only. Methane, carbon monoxide and dioxide, ethylene, ethane, propylene, pro-

0196-4321/83/1222-0242$01.50/0@ 1983 American Chemical Society

Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 2, 1983

243

Table I. Properties of Catalysts surface area, m z g-' * 2 %

pore volume (N,), mL g-' ?5%

NH, ads, mL ads g"

sample

% Al,O,

fresh

equil

fresh

equil

fresh

equil

y-A120,

100 13 25 13

85 384 337 94

332 244 86

0.31 0.45 0.76 0.31

0.35 0.50 0.27

1.2 17.2 11.1 2.9

1.1 5.4 4.9 1.7

980-13 980-25 970-13

regenerated

16.8 10.5 2.8 a Silica-aluminas were commercial samples in which maximum impurity levels were quoted to be ( w t %): 0.05% Na,O, 0.3% SO,, 0.03% Fe, 0.05% CaO, 0.01% C1. r-A1,0, did not coke. Equilibrium conditions: temperature = 400 'C, dimethyl ether a t 0.05MHSV (mol of DME (mL of cat.)-' h-') for 3 h. Table 11. The Selectivity and Conversion of Dimethyl Ether to Products over Catalyst 980-13as a Function of Time on Streama time, min

conv, %

NH, ads, mL g"

CH,

C2H4

C,H,

c4-C6

CH,OH

17.2 92.9 9.1 8.1 16.1 22.1 46.2 4.2 78.0 7.8 16.4 15.2 18.9 36.6 8.9 46.0 6.9 12.3 15.6 19.5 31.6 16.6 6.4 23.8 12.1 14.4 37.0 27.4 17.6 18.4 5.9 19.8 13.7 20.2 18.5 22.5 12.4 5.6 19.1 15.3 10.8 18.6 30.2 Feed = dimethyl ether; space velocity = 0.05 MHSV (mol of DME (mL of cat.)-' h-'); temperature = 400 "C. 0 2.5 6.0 15.0 30 60 100

(%I

CONVERSION

H H S V-'

Figure 2. Selectivity as a function of conversion: feed, dimethyl T = 400 OC; MHSV = 0.05 mol of DME (mL ether; catalyst, 980-13; of cat.)-' h-'.

Figure 1. The conversion of dimethyl ether as a function of space velocity; units of MHSV: mol of DME (mL of cat.)-l h-I; catalyst: 980-13;T = 400 OC; (V)methanol; (A)dimethyl ether; (0) CH,; ( 0 ) CzH4; (0) C3H6;( 0 )C4 hydrocarbons.

0

980-13 9lC-25 970-13

I

A '42%

pane, and traces of higher hydrocarbons (carbon number less than or equal to 10) were produced at longer residence times over catalyst 980-13 and 980-25. Measurements of selectivity were carried out as a function of a conversion and typical results are shown in Figure 2 for catalyst 980-13. The activity of all silicaalumina catalysts declined over the first 100 min to a constant value (equilibrated catalyst) (Figure 3) and this was accompanied by increased production of methane at the expense of higher hydrocarbons (Figure 4). Deactivation was caused by coke, since regeneration of the catalyst by heating at 400-450 "C for 4 h under oxygen led to restoration of the original acidity (Table I), surface area, porosity, and activity of the catalyst. Traces of hydrogen could be found among the products when the catalysts were coking and this presumably originates from the polymerization/ dehydrogenation reactions leading to carbon. The surface areas and pore volumes of fresh and equilibrated catalysts are compared in Table I. A typical pore size distribution curve is shown in Figure 5. In no case was the pore size distribution changed by coking, with the maxima in the curves being 40 A (970-13), 17 A (980131, and 30-35 A (980-25). The decrease in pore volumes observed on coking are reported in Table I. The effect of coking was then studied in detail using catalyst 980-13. In a series of experiments dimethyl ether was passed over fresh samples of catalyst for different

T IM

E

Inin)

Figure 3. The decrease in conversion of dimethyl ether with time on line: feed, dimethyl ether; T = 400 OC; MHSV = 0.05 mol of DME (mL of cat.)-' h-'.

T IM E

lminl

Figure 4. Selectivity as a function of time on line: feed, dimethyl T = 400 "C; MHSV = 0.05 mol of DME (mL ether; catalyst, 980-13; of cat.)-' h-'; (0) CH,; ( 0 )CzH4; (0) C3H,;).( C4 hydrocarbons; (A) C6 hydrocarbons; (A)c6 hydrocarbons.

244

Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 2, 1983

Table 111. Selectivity and Conversion of Dimethyl Ether on Samples of Catalyst 980-13Which have been Subjected to Different Levels of Coking. Measurements Made at Low and Approximately Constant Conversion NH3 ads, mL g' conv, % MHSVa 9.1 5.7 0.62 0.49 7.8 6.1 0.18 6.9 10.1 0.17 6.4 9.8 5.6 6.5 0.12 a Units = mol of DME (mL of cat.).' h-'. 400 "C.

selectivity, %

init rateb 3.52 2.99 1.80 1.67 0.77

CH, 15.6 14.4 15.9 19.9 11.9

C,H, 10.5 16.6 13.8 11.0 10.3

C3H6 16.3 17.4 18.2 14.6 15.9

Units = mol of DME converted h-' mL cat:'.

CH,OH 18.5 31.2 15.9 25.7 18.0 31.2 23.2 26.2 18.1 33.8 Feed = dimethyl ether, T C446

""

-

80 -

z - 60YI = % 40Y

Y

20

POR E

RADIUS

0

[dl

Figure 5. Pore size distributions of fresh and equilibrated 980-13 catalysts (equilibration conditions: pure dimethyl ether passed over catalyst at 400 "C for 3 h): (A) fresh catalyst; ( 0 ) equilibrated catalyst. 100 6

AMMONIA ADSORPTION

1

-10 TIME

ON

LINE

0'

(mini

Figure 6. Adsorption of ammonia, conversion, and rate as a function of time on line. Conversion from Table 11; ammonia adsorption and initial rate from Table 111;feed, dimethyl ether; catalyst, 980-13; T = 400 "C; MHSV = 0.05 mol of DME (mL cat.)-' h-l.

periods of time. In a given experiment, the conversion and selectivity were measured after a given time (Table 11) and the flow rate was then increased to a point where the conversion was reduced to 8 f 2%, deactivation being very slow at this flow rate. Selectivity and conversion were then measured and the rate of reaction was calculated on the basis of the differential reactor approximation ("initial rate"). These values are summarized in Table 111. The catalyst was then removed from the reactor and the amount of ammonia adsorbed on the used catalyst was measured (Table 111). A fresh catalyst was then introduced into the reactor and the experiment repeated for a different period of time. Results are summarized graphically in Figures 6 and 7 . Experiments to determine the selectivity of dimethyl ether conversion to hydrocarbons on catalysts that had been equilibrated to constant activity were performed at a constant conversion of ca. 10%. The selectivities and the quantities of ammonia adsorbed by the catalysts used are summarised in Table IV. Discussion The observations that acidic oxides catalyze the conversion of methanol to dimethyl ether and to higher hydrocarbons and are deactivated by coke formation are not unexpected (On0 and Mori, 1981; Cormerais et d.,1981). However, the interaction between conversion and selectivity, catalyst acidity, and deactivation is of greater in-

-

J

=

Ind. Eng. Chem. Prod. Res. Dev., Vol. 22, No. 2, 1983 245

conversion increased the selectivity changed, with the production of higher hydrocarbons increasing and the production of methane and methanol (from dimethyl ether) decreasing (Figure 2). As a result, it would seem that conversion is dependent on catalyst acidity but that selectivity depends on conversion and not on acidity. Similar trends were observed on other silica-aluminas, but comparison of selectivities at the same conversion gave different results on different catalysts (Table IV). The selectivity for the production of C,-C, hydrocarbons is not the same, and the trends for production of methane and methanol are different. These results show that selectivity apparently does not depend on acidity for one particular catalyst (Table 111) but does vary with acidity between different catalysts (Table IV). For acidic oxides, it is known that conversion and selectivity are controlled by the number and strength of surface acid sites and that these sites may be deactivated by coking. If the distribution of acid site strengths is not affected by coking (as would be the case for unselective blocking of pore mouths (see below)) then conversion will decrease as the number of sites is reduced, but the conversion/selectivity relationship will remain the same. However, this relationship will differ for different silicaaluminas and, as a result, the selectivities observed at given conversions will also differ (Table IV). This could be confiimed by measurement of the acid site strength distributions for individual catalysts during coking but, regrettably, the ammonia adsorption technique is not selective. However, support for the concept comes from examination of fresh and equilibrated catalysts (Table I, Figure 5). It is obvious from Figure 7 that hydrocarbons are produced only on catalysts that are at least moderately acidic, and these catalysts are known to favor coke formation by a polymerization/growth mechanism (Walsh and Rollman, 1977). The pore size distribution in catalysts is unchanged on partial coking (Figure 5) and the amount of carbon deposited is insufficient (with the exception of catalyst 970-13) to explain the observed decrease in gas adsorption (Table I) in terms of reduction in pore volume by deposited coke. At least in catalysts 980-13 and 980-25, carbon must be blocking off pores in a random fashion (the pore size distribution is unchanged) by growth over the pore entrance. This would occur if coke deposition were nucleated at a given site and coke grew by polymerization from that nucleation centre (Rollman and Walsh, 1979). The pore size distribution and acid strength distribution would then not change and the selectivity (controlled by the uncovered acid sites) would be independent of the coke deposited. Pore mouth bridging can only be expected if the pores are small and this may not be the case with catalyst 970-13. The decrease in pore volume can be explained in terms of displacement of gas by deposited carbon (based on the lattice dimensions of graphite) and it could be that, in this case, the pore mouth is too large to be bridged and carbon grows into the pore. Comparisons can be drawn between these results and those reported by Dejaifve et al. (1981) and Langner (1982). In the present systems, pore size distribution measurements showed coke to bridge pore mouths of radius less than ca. 35 A and probably to form inside larger pores (Table I). Dejaifve et al. (1981) observed a similar phenomenon but found coke to deposit inside much smaller pores (ca. 7 A dimension). These observations were based only on comparisons of volumes of coke deposited as measured by changes in pore volume. However, changes in pore size distributions on coking were not reported.

Dejaifve et al. also showed that the acid site distribution was altered by coking H-mordenite and H-offretite, and the number of acid sites was reduced by coking both these catalysts and H-ZSM-5. If pore volume is lost by blockage of the pore mouths, as is the case for H-ZSM5,980-13,and 980-25, then the number (but not necessarily the acid strength distribution) of acid sites would be reduced. If coke deposits over all the surface (H-mordenite, H-offretite, 970-13) then both the number and the strength of the acid sites would change. In these terms, selectivity over H-ZSM5,980-13, and 980-25 should be apparently independent of acid strength distribution (since this does not change on coking) but should vary as conversion, which reduces with the number of acid sites remaining. For the other catalysts, selectivity should change with the changing acid site number and acid strength distribution as coking proceeds. Langner (1982) is less concerned with the effect of coking on selectivity, but it is interesting to note that the amount of methane produced increases with coking: direct comparisons with results summarized in Figures 2, 4,and 6 would suggest that a similar explanation is tenable for both studies. The change in selectivity to favor methane at low conversions can be explained tentatively in terms of previously suggested reaction mechanisms (Ono and Mori, 1981; Dejaifve et al., 1980). It is generally agreed that Bronsted acid sites are important in the reaction and the formation of a “methoxy” group on the surface has been suggested (Ono and Mori, 1981).

fH3

t CH3

CH3\ 0

H+

This ‘“ethoxyn species could then act as a hydrogen transfer agent to produce methane (Dejaifve et al., 1980)

3

t

! CH3

0

*I

f

r

CH2

\0/cH3

-

t

CH20 CH3 I

A

CH4

P

c

or could produce methylcarbonium ions which could react with olefins to build up the carbon chain. Olefins are only produced at high conversions and, as a result, higher hydrocarbons would only be favored as conversion increases. Methane, on the other hand, would be expected, and is found, at lower conversions. Thus it would seem that conversion and selectivity are controlled by the number and strength distribution of acid sites in the silica-aluminas and that pores and acid sites may be closed off unselectively as a result of coking of small pore catalysts. As a result, the acid site strength distribution (which controls selectivity) and the pore size distribution will not change during a reaction on one particular silica-alumina but will be different for different silica-aluminas. For larger pore catalysts, coke formation can occur inside the pores, affecting acid site strength distribution and pore size distribution and, as a result, affecting conversion and selectivity. Acknowledgment The authors acknowledge, with gratitude, financial support from the Australian Research Grants Committee. Registry No. Carbon, 7440-44-0; alumina, 1344-28-1; silica, 7631-86-9; methanol, 67-56-1; methyl ether, 115-10-6;methane,

Ind. Eng. Chem. Prod. Res. Dev. 1983, 22, 246-250

248

74-82-8; ethane, 74-84-0; ethylene, 74-85-1.

Deno, N. C.; Boyd, D. 8.; Hcdge, J. D.; Pittman, C. U.; Turner, J. 0. J. Am. Chem. SOC. 1964, 86, 1745. Derouane, E. G.; Nagy, J. 6.; Dejaitve, P.; van Hoof, J. H. C.; Spekman, B. P.; Vedrine, J. C.; Naccache, C. J. Catal. 1980, 67, 155. Forni, L. Catal. Rev. Scl. Eng. 1973, 8 , 65. Kaeding, W.; Bulter, S. J. Catal. 1980, 67, 155. Langner, B. E. J. Appl. Cat. 1982, 2 , 289. Ono, Y.; Mori, T. J . Chem. SOC.Faradey Trans. 1981, 77, 2209. Roilman, L. D.; Walsh, D. E. J. Catal. 1979, 56, 139, 195. Venuto, P. 6.; Landis, P. S . Adv. Catal. 1968, 78, 309. Walsh, D. E.; Roilman, L. D. J. Catal. 1977, 49, 369.

Literature Cited Adkins, H.; Perkins, P. D. J. Phys. Chem. 1928, 32, 221. Anderson, J. R.; Foger, K.; Mole, T.; Rajadhyaksha, R. A,; Sanders, J. V. J. Catal. 1979, 58, 114. Benesi, H. A.; Winquist, B. H. C. A&. Catal. 1978, 2 7 , 97. Chang, C. D.; Silvestri, A. J. J. Catal. 1977, 47, 249. Chang, C. D.; Lang, W. H.; Smith, R. L. J. Catal. 1979, 56, 169. Cormerais, F. X.; Chen, Y.; Kern, M.; Gnep, N. S.;Perot, G.; Guisnet, M. J. Chem. Res. 1981, 290. Deeba, M.; Hall, W. J. Catal. 1979, 6 0 , 417. Dejaifve, P.; Auroux, A.; Gravelle, P. C.; Vedrine, J. C.; Gabellca, Derouane, E. G. J. Catal. 1981, 70, 123. Dejaifve, P.; Vedrine, J.; Bolls, V.; Derouane, E. G. J. Catal. 1980, 6 3 , 331.

Received for review August 2, 1982 Accepted December 13, 1982

Inhibition of Hydrodesulfurization by Nitrogen Compounds L. Charles Gutberlet' and Ralph J. Bertolaclnl Research 8 Development, Amoco 011 Company, A Subsidiary of Stanc4rd Oil Company (Indiana), Amoco Research Center, Naperville, Illinois 60566

The effects of organic nitrogen compounds on the hydrodesulfurization of a catalytic naphtha were examined under conditions where the nitrogen compounds were generally unreactive. The nitrogen compounds tested as inhibitors included several alkylpyridines, 2,5dimethylpyrrole, 4-methylaniline, and benzylamine. Significant differences in desulfurization were observed which could be related to nitrogen compound structure. Comparison of alkylpyridines clearly shows a steric contribution of ring substituents. The lack of inhibition when methyl groups were attached to ring carbons adjacent to the nitrogen atom suggests adsorption of the pyridine molecule on the active desulfurization site through the nitrogen atom. Electronic effects, such as resonance stabilization, may account for the relatively small inhibition effect of sterically unhindered 4-methyianiline while the larger inhibition effects of the hindered 2,5dimethylpyrrole and benzylamine can be attributed to their greater reactivity to produce more strongly adsorbed polymeric materials and ammonia, respectively.

Introduction The susceptibility of petroleum refining catalysts to inhibition by ammonia and organic nitrogen compounds has long been recognized. It was reported by Viland (1957), for example, that the addition of isoquinoline to a lownitrogen, virgin, paraffmic gas oil decreased conversion and the yield of gasoline obtained at a constant coke yield in catalytic cracking tests over silica-alumina at 530 OC. In contrast, no loss in conversion or gasoline yield was observed upon addition of the nonbasic nitrogen compound, indole, until the amount of added nitrogen exceeded 0.4 w t %. The greater inhibiting effect of isoquinoline is attributed to preferential adsorption on the acidic sites of the silica-alumina catalyst. Earlier studies by Mills et al. (1950) and Plank and Nace (1955) provide information on the relative inhibition effects of a variety of nitrogen compounds. By use of the dealkylation of cumene over silica-alumina at 425 OC as a model reaction, the following order of decreasing inhibitor strength was obtained: imidazole > quinaldine, quinoline > pyrrole, pyridine > piperidine, indole > decylamine, n-butylamine > aniline. Calculated equilibrium constants for adsorption vary over two orders of magnitude between imidazole and aniline. As pointed out by these investigators, basicity, as measured by pK, at ambient temperature, is unreliable as a predictor of inhibitor strength. Furthermore, under reaction conditions, piperidine and the alkylamines can partially decompose while indole and pyrrole can polymerize, thus obscuring the true inhibition effects of these compounds. The impact of nitrogen compound stability in response to the reaction environment is further demonstrated by the work of Meisel et al. (1957) on the isomerization of 0196-4321/83/1222-0246$01.50/0

cyclohexane to methylcyclopentane over platinum-acidic oxide catalysts. In the presence of a large excess of hydrogen at a total pressure of lo00 psig, equimolar amounts of 2-methylpyridineand ammonia had the same inhibition effect over a wide temperature range (315-482 "C). Acridine, quinoline, pyrrole, pyrrolidine, and 2-methylpiperidine were reported to behave similarly. This lack of discrimination among nitrogen compounds was attributed to rapid and quantitative conversion to ammonia under the reaction conditions used. Kirsch et al. (1959), reported that olefin hydrogenation was more strongly inhibited by nitrogen compounds than hydrodesufurization. Their experiments were carried out at 370 OC and 300 psig (total pressure) with a heptaneheptene feed containing about 0.4 w t % sulfur, as thiophene, to which pyridine (up to 0.2 wt % N) was added. The effect, which was ascribed to the existence of different catalytic sites for hydrogenation and desulfurization, was more pronounced for sulfided nickel-alumina than for sulfided cobalt molybdenum-alumina. Desulfurization is also proposed to occur on two types of catalytic sites. Satterfield et al. (1975) interpreted the hydrodesulfurization of thiophene over cobalt-molybdenum-alumina at 50 psig in terms of strong and weak catalytic sites with the strong sites being more sensitive to inhibition by pyridine. Pyrrole, although unstable, exhibited an inhibition effect similar to pyridine, a result which is consistent with the previously cited inhibition of cumene dealkylation. The purpose of the present work was to examine the effects of nitrogen compound structure on the inhibition of hydrodesulfurization. Impetus for such a study derived from our observation that the addition of 2,6-dimethyl0 1983 American Chemical Society