Applied C’cttalwis, 29 (1987) 175-184

Elsevier Science Publishers B.V., Amsterdam Printed in The Netherlnnds 175

CATALYTIC OLIGOMERIZATION OF ETHENE OVER NICKEL-EXCHANGED AMORPHOUS SILICA-ALUMINA;

EFFECT OF THE REACTION CONDITIONS AND MODELLING OF THE REACTION.

R L ESPINOZA*, C J KORF, C P NICOLAIDES and R SNEL

Chemical Engineering Research Group, Council for Scientific and Industrial Research,

P.O. Box 395, Pretoria 0001, Republic of South Africa.

(Received 28 July 1986, accepted 11 September 1986)

ABSTRACT

In the oligomerization of ethene on partially nickel-exchanged silica-alumina higher conversions were obtained with an increase in temperature, an increase in pressure and a decrease in space velocity. The deactivation of the catalyst was virtually independent of space velocity but its rate increased with pressure and with temperature. The effect of both pressure and space velocity indicated that the rate of ethene conversion is approximately first order in ethene. Increasing the temperature or pressure and particularly decreasing the space velocity shifted the product spectrum to higher products.

A random dimerization and co-dimerization model for all the alkenes present in the gas phase, adjusted for major kinetic differences between alkenes and predicting the product distribution in terms of the percentage conversion of the feed, fitted the experimental data rather well, although several simplifying approximations were made to describe the complicated system of consecutive reactions.

INTRODUCTION

We have previously shown [l] that the oligomerization rate of ethene on partially

nickel-exhanged silica-aluminas is high enough to make these catalysts eligible for

development for industrial application. In that work the oligomerization activity of

the supported nickel was found to be proportional to the acid strength of the

support. This made it possible to select promising supports for particular

applications.

It was clear, however, that attention had to be given to decreasing the

deactivation rate of the catalyst and to optimizing the reaction. We therefore

decided to investigate the effect of the reaction conditions on catalyst

performance. In addition we derived a mathematical model for the reaction [in terms

of the product distribution as a function of the conversion of the feed), for use in

optimization studies.

--__ -.--__

* Present address: SASOL Technology, P 0 Box 1, Sasolburg, RSA.

01669834/87/$03.50 01987 Elsevier Science Publishers B.V.

176

EXPERIMENTAL

The catalyst selected for this study was a partially nickel-exchanged

silica-alumina of low acid strength (LASA) with a silica/alumina mol ratio of 50 and

containing 0.27 96 nickel by mass, After exchange and washing, the catalyst was dried

in air at 80 "C before loading into the reactor. Its monovalent ion-exchange -1

capacity was 0.5 mm01 g , the surface area was 557 m2 g -1

and the mean pore

radius was 1.7 nm. Reactions were carried out in a tubular reactor with an internal

diameter of 15 mm and a catalyst volume of about 4.5 cm3. Details of the catalyst

and support preparation and characterization and of the techniques used in the

catalysis investigation were given before [1,23.

RESULTS AND DISCUSSION

Effect of the reaction conditions on the activity of the catalyst

The reaction parameters were systematically varied and the results are given in

Table 1. Gas samples were taken at one hour on stream using catalyst samples which

in each case had been regenerated once by direct contact with air at 773 K for

16 hours Cl].

TABLE 1

Effect of the reaction conditions on the activity.

Pressure = 1150 kPa MHSV = , 6

Temperature, K 393 513 573 653

Activity* 38 484 1280 1670

Pressure = 1150 kPa Temperature = , 573 K

MHSV, h-l 0.5 3 6 12

Activity* 160 720 1280 1645 -.--

Temperature = 573 K

Pressure, kPa 160 1150 1150 2140

MHSV, h-l 6 6 12 12

Activity* 130 1280 1645 3260

* Activity is expressed in terms of grams of ethene reacted per gram of nickel and per hour.

The results show that at constant space velocity and pressure the rate constant,

k, follows the Arrhenius equation (Fig. 1) with an apparent activation energy of

37 kJ mol-'. At constant temperature and pressure, the activity increased with

increasing space velocity to approach a plateau (Fig. Z), in line with normal space

velocity behaviour. At high space velocities the dimerization of ethene is virtually

the only reaction taking place (vide infra) and the asymptote of about 12 g g -1 h-1

indicates the rate of that reaction alone.

177

I

1.8 2.2 2.6

500

0 3 6 9 12

103/T, K-l MHSV, ge,hene !& h-l

FIGURE 1 Arrhenius plot for

ethene conversion.

FIGURE 2 Effect of space velocity on

catalyst activity.

The effect of pressure (last section of Table 1) was ascertained at two space

velocities and at a constant temperature. The reaction rate correlated with pressure

according to r = k' P", with n = 1.16 and 1.09 for MHSV = 6 and 12 respectively.

This indicates that, under the reaction conditions used, the ethene oligomerization

reaction is approximately first order in ethene. In that case

r= _k'C=- k CC (1 - x)

J X

W/F0 = - (1 / r) dx = - (1 / k Co) ln(l - x) 0

II)

I21

where r is the reaction rate, C is the concentration, x is the extent of conversion,

k is the reaction rate constant, W is the mass of the catalyst, F is the feed rate

and subscript 0 refers to the reactor entrance condition. The ratio W/F 0

is the

reciprocal space velocity. The experimental ethene conversions obtained were plotted

as - In (1 - x) against l/MHSV (Fig. 3), after correcting the experimental ethene

conversion levels, obtained at different space velocities, for the occupation by

reactive ethene oligomers of active sites on which ethene could have reacted if

these oligomers had not been present. This was done by assuming site occupancy to be

proportional to gas composition and subtracting the sites occupied by non-ethene

species (vi& infra, model). The data are then represented by a straight line

passing through the origin, thus confirming the first-order nature of the reaction.

178

w’Fo’ (gethenel-l gcat h FIGURE 3 Demonstration of first-order kinetics.

Effect of reaction conditions on deactivation

In order to compare the deactivation rates at different reaction conditions (and

therefore at different conversion levels), the activity at one hour on stream was

assigned a value of 100 for every run. The activities after that time were expressed

as percentages of the value at one hour. The deactivation figures thus

- 0 0.5 -a 3.0

_A 6.0 Al2.0

- -_ 1 I , ,, , ,,

0 20 60 too 0 20 60 100 0 20 60 100

Time on stream, h

FIGURE 4 Effect of reaction conditions on deactivation.

100

50

179

arrived at are given in Fig. 4 for different reaction conditions.

The deactivation rate was found to be insensitive to the space velocity, whereas

high values of temperature or pressure resulted in increased deactivation. The data

in Table 1 show, however, that the activity increases with increasing temperature

and pressure.

Taking into account the results of Table 1 and Fig. 4, we suggest

satisfactory compromise for practical application (maximum sustained

conversion) may be reached by employing a pressure of 1150 kPa and a

of 573 K.

Effect of reaction conditions on product distribution -- - __-_-_-

that a

ethene

temperature

In Table 2 the effect of the process conditions on the product distribution is

compared by changing one variable at a time. As the product spectra shifted to

lighter products with time on stream each comparison was made at approximately the

same ethene conversion level, representing about the same state of catalyst

deactivation.

TABLE 2

Effect of pressure, temperature and space velocity on the product spectrum

Press, Temp. MHSV Convers. Product spectrum, mass %

kPa K h-l % C2* C3 C4 C5 C6 C7 CB C9 CIO+

1150 573 6 28.9 1.12 0.41 82.07 0.60 12.32 0.05 2.08 0.03 1.28

2600 573 6 32.0 0.40 0.28 73.74 0.80 18.02 0 4.38 0.04 2.30

1150 573 6 37.2 1.21 0.86 80.51 0.51 12.13 0.04 2.67 0.20 1.81

1150 653 6 38.0 3.20 1.57 59.64 2.53 17.29 0.14 6.31 0.57 8.64

1150 573 6 47.8 1.49 0.46 77.06 0.67 13.26 0.12 3.6U 0.23 3.03

1150 573 3 48.0 2.93 0.83 55.08 3.40 19.58 0.18 10.26 1.02 6.68

1150 573 0.5 51.8 7.46 0.94 34.05 12.92 20.17 3.11 9.75 2.82 8.46

* Ethane; ethene is not counted in the product spectrum.

The data show that an increase in either pressure or temperature resulted in

a slightly heavier product. Acid catalysis may have contributed to the

oligomerization of C2+ species, particularly at the higher temperatures; note that

the nickel content of 0.27 % and the monovalent ion-exchange capacity of

0.5 mmol g-1 imply that a minimum of 80 % of the ion-exchangeable sites of the

support remained in the proton form after the partial exchange with nickel. These

acid sites would be free to oligomerize the alkenes higher than ethene; ethene is

not reactive by acid catalysis on this support [l].

180

The effect of space ve7ocity is shown in the third section of Table 2. The

differences in the conversion levels are so small that they cannot be used to

explain the marked differences in the product distribution. An increase in space

velocity resulted in an increased degree of ethene dimerization, with less further

oiigomerization. At the same time there were fewer products arising from cracking

(those having odd carbon numbers).

As indicated above, the optimum conditions for high sustained ethene conversion

at pressures attractive for industrial application appear to be in the temperature

range 573 K to 613 K and at about 1100 kPa. Further, the conversion of ethene can be

oriented to dimerization (high space velocity, temperature below 573 K), the petrol

range (low space velocity and therefore a relatively high proportion of cracking

products) or the diesel range (intermediate space velocity of about 6 and a

temperature of about 573 K, with a recycle of light products).

Relative reaction rates of small alkenes ~-----~---------~

In the oligomerization of alkenes by (non-acid) transition metal catalysis the

alkenes with only terminal double bonds (ethene and propene) have higher rates of

reaction than internal alkenes [3]. Under identical reaction conditions the rates of

ethene and propene conversion on the catalyst were found to be about the same; as

mentioned above, in propene conversion acid catalysis makes a significant

contribution to conversion. The rate of conversion of 1-butene was found to be about

half that of ethene when averaged over the useful time on stream.

It was observed that the butenes in the product of ethene oligomerization were

present in their thermodynamic equilibrium composition (75 % P-butenes [4]). This

demonstrates that the double-bond isomerization is faster than both the formation

and the further conversion of the butenes, with both reactions being selective with

respect to 1-butene. The presence of a large amount of acidic sites contributes to

the fast establishment of the isomer equilibrium.

The lower reactivity of butene (and larger alkenes) on the type of catalyst used

is probably due both to the size of the molecules [3,5] and to the presence of the

internal alkenes.

Modellinq of the ethene oliqomerization

According to the model oligomers are formed by a series of alkene-alkene

couplings. At low ethene conversion levels ethene-ethene couplings dominate and

mainly butene is formed. Butene then reacts either with ethene or with itself to

form higher oligomers, etc. The coupling is considered to take place between an

alkene in the gas phase and another one in an adsorbed state; this is in line with

the observed first-order kinetics. Therefore, in the reaction with the surface

species, the gaseous species are assumed to react according to their concentration

181

in the gas phase. However, the lower reactivity of C4+ alkenes, mentioned above,

was taken into account quantitatively, by means of a reactivity factor. As a first

approximation the same reactivity factor as that observed for butene was used for

all C4+ alkenes. The composition of the surface species is assumed to be the same

as that of the gas phase. In short, the model represents concentration-driven random

dimerization and co-dimerization of the alkenes present, adjusted for major kinetic

differences between the various alkenes.

The probability that a given adsorbate A will react and couple with a molecule B

from the gas phase is:

P(A t 6) = P(A) x P(B) (3)

where P(A) is the probability that a reaction will occur on a catalytic site

occupied by adsorbate A, and P(B) is the probability that the reaction will involve

a molecule 8.

For each species of carbon number X:

Number of molecules with carbon number X P(X) = xR (4)

Total number of molecules

where the reactivity factor R = 1, except for the C 4+ species, where R is taken to

be 0.55. According to Flory [S], if only one or two atoms separate the functional

groups [in this case the carbon atoms on either side of the double bonds) in the

monomer, then the reactivities of monomer and dimer may differ greatly, whereas the

difference between dimer and trimer will be much less, between trimer and tetramer

still less, etc. Our approximation of R for the C4+ species should therefore not

result in a large error.

The change in the number of the species with different carbon numbers in any

small increment of time is:

AN(Z) = -2 [P(2) x P(2) + P(2) x P(4) + P(2) x P(6) . . . . . . . . 1 = -2 E [P(2) x P(X)]

AN(4) = P(2) x P(2) - 2 z [P(4) x P(X)]

AN(6) = 2P(2) x P(4) - 2 1 [P(6) x P(X)]

AN(8) = ZP(2) x P(6) + P(4) x P(4) - 2 E [P(8) x P(X)] etc.

(6)

(6)

(7)

(81

These equations were transcribed in a computer algorithm that involves the following

steps:

1) Entering the composition of the feed as the number of molecules present in an

arbitrarily selected total number of initial molecules (for instance 10000).

2) Calculation of P(X) according to equation (4) for every carbon number species.

182

Conversion, mass % of ethene

I 1 I 1 4

0.06 0.15 0.27 0.49 1.20

NOiND

FIGURE 5 Prediction of product spectrum

by model (cracking absent].

FI

20 40 60 80 100

Conversion, mass % ethene

GURE 6 Comparison of experimental data

with the oligomerization model.

31

41

51

61

71

Calculation of the new number of molecules of each carbon number species

according to the scheme given above, keeping track of both the total number of

ethene dimerizations, ND, and the total number of other coupling reactions

taking place, NO.

Calculation of the fraction of feed molecules that have been converted since the

the reactions started.

Printing out of the gas-phase composition at predetermined steps in the

conversion level of the feed as calculated in step (41.

Exiting at a predetermined level of conversion of the feed.

Returning to step 2 using the new numbers calculated in step (3).

The result is a tabulation of gas-phase composition versus percentage feed

conversion; this is presented graphically in Fig. 5, using the ratio NO/ND for

the horizontal scale.

The model predictions are compared with experimental data in Fig. 6. The

reaction conditions were 573 K, 1150 kPa and MHSV = 6. Note that the experimental

points do not add up to 100 as alkenes with odd carbon numbers have been excluded

here because the model does not take cracking into account. Although the model

contains much simplification with respect to the chemico-physical phenomena

occurring, the prediction of the experimental data is rather good. Oligomers in the

C lo+ range are predicted well by the model.

183

Even at high ethene conversion levels, only traces of C23+ hydrocarbons were

observed in the products, whereas the model (Fig. 5) predicts that 3.9 % of the

products will be between Cz4 and C5C at 90 % ethene conversion. The partial

pressure of these C23+ products in the reactor system was low enough for them to

report in the product if formed in any substantial amount. These products would have

reported in the collected liquid, but did not do so even after runs extending to ten

days. The absence of a measurable amount of Cz3+ oligomers is ascribed to the

selective cracking of the large oligomers [l] and to a gradual decrease of

reactivity as molecular size increases. However, the retention of very small amounts

of large oligomers on the surface of the catalyst cannot be excluded.

The model, to which cracking terms may be added, provides a useful tool for the

design of reactor systems In which partial conversions, physical separations and

recycles can take place.

CONCLUSIONS

Silica-alumina, partially exchanged with nickel, provides an active catalyst for

the oligomerization of alkenes; it was evaluated particularly for the conversion of

ethene.

Increasing the space velocity resulted in an increase in the products with low

carbon numbers; the l!ghter the product the less cracking occurred. The space

velocity did not have a marked effect on the deactivation rate of the catalyst.

Conversion rates increased with temperature; simultaneously the deactivation rate

increased markedly. There was a moderate shift to heavier products as temperature

was increased, together with an increase in the products of cracking. Pressure had

an effect similar to that of temperature with respect to deactivation and

conversion.

Optimum conditions for high sustained ethene conversion at pressures attractive

for industrial applicat>on would be reached at a temperature in the range 573 K to

613 K and a pressure of about 1100 kPa. The conversion of ethene can be oriented to

dimerization, the gasoline range or the diesel range.

The effect of pressure and of space velocity both indicate that the rate of

oligomerization of ethene is approximately a first-order reaction. A concentration-

driven random dimerizatlon and co-dimerization model for all the alkenes present in

the gas phase, adjusted for major kinetic differences between the alkenes,

adequately predicted the experimental oligomerization produc

and its relation with ethene conversion.

ACKNOWLEDGMENTS

The authors thank Dr M.S, Scurrell and Mr W.G.B. Manders

discussions.

t distribution

oot for fruitful

REFERENCES

1. R.L. Espinoza, R. Snel, C.J. Korf and C-P. Nicolaides, Appl. Catal. in press. 2. C.J. Korf and R.L. Espinoza, Report CENG 584, CSIR, Pretoria, 1986. 3. G. Henrici-Olive and S. Olive, Co-ordination and Cata7ysi5, Verlag Chemie,

Weinheim, 1977. 4. D.L. Stull, F. Westrum, Jr. and G.C. Sinke, The Chemica 1 Thermodynamics of

Organic Compounds, John Wiley, New York, 1969. 5. ---* SHOP, Linear alpha olefins - Reactions and application,

Shell Chemicals UK Ltd, London. 6. P.J. Flory, J. Am. Chem. Sot., 58 (1936) 1877.