Kinetics and Mechanism of Propylene to 4-Methyl-1-pentene Catalytic Dimerization

Lucio Forni" and Renzo lnvernizzil Istituto d i Chimica Fisica, Universitlc d i Milano, Milano, Italy

The kinetics of catalytic propylene dimerization to 4-methyl-1 -pentene was investigated by use of both dif- ferential and integral reactor techniques from 130 to 150" and from 18 to 80 kg/cm2 over supported and promoted metallic sodium. It was found that the rate-determining step i s probably the adsorption of propyl- ene, with formiation of an allylic intermediate. A general correlation of a11 involved kinetic parameters was ob- tained, together with a system of kinetic equations, which, by employing such parameters, may constitute a basis for reactor design. The influence of the presence of transition metals in the catalyst was also studied, and its effect on catalytic activity evaluated, Some hypotheses were also made for the interpretation of the last mentioned effect.

T l i e practical interest of the propylene dimerization reaction to 4-metmhyl-l-pentene (4511P) lies in the fact that poly- 41IlP is a tliermoplaijtic resin possessing interesting prop- erties, such as high softening point (about 200°), low specific gravity (0.83), and very high trarisparence (Xrnaud, 1968; Hambling and Northcott, 1968). The highest selectivities toward 4hIlP, in propylene dimerizatioii, are obtained with catalysts based on variously supported alkaline metals. The design, formulation, preparation, aiid employment of such catalysts constitut'e the object of a large number of industrial patents. -1s outlined hy Hanibling (1968, 1969), the kinetic study of propylene dimerization reaction over such catalysts is c.omplicat,ed by various experimental difficulties, which can be overcome only by nieaiis of a careful choice and a rigorous control of reaction para meters.

Experimental

Materials. Propylene was a Gerling Holz & Co. (Ham- burg) product, whose purit'y, tested gas chromat'ographically, resulted to be 98.01 mol 70; the remainiiig part was essentially propane, with very small traces of ethane, ethylene, and iso- butene. Kit'rogeii was ;? 99.999% pure.

Catalysts. Tlie main catalyst was metallic sodium, sup- ported on 20--30 mesh anhydrous and promoted with Hg (Forrii, 1972). Tlile chemical analysis of such a catalyst gave &C03, 86.03 (wt %); S a , 6.48; Hg, 5.72; SazO, 1.76. The catalyst BET (krypton) surface area was found to b: 0.147 ni2,'g, the pore volume 0.31 ml/g, for 75-400,000-A pores, and the pore radii distribution curves, measured by the mercury penetratioii 1,eclinique on the bare support, on the main catalyst and on all other catalysts, showed th3t' over 95% of the pore volume was always due to 2 10,000-A pores. The catalysts were prepared as follows: a thermounstable salt (e.g., acetate, oxalate, hydroxide, etc.) or metal comples (e.g., l't(XH3)d(OH)t, etc.) of the transition metal was added to predried &cos and dispersed by gently heating with stir- ring in a n inert atmosphere. Na was theii added after cooling and the mixture reheated with stirring in a n inert atmosphere until the transition metal compound was decomposed and reduced to metal (+;<azo) and S a finely dispersed on the

Present address, Societi Italisna Resine, Milano, Italy

catalyst. For all the catalysts t'he (Ka/support) aiid (Ka/ t,ransition metal) molar ratios were kept rigorously constant.

Apparatus and Procedure. T h e propylene dimerization reaction was performed in a fixed-bed st'ainless steel (316 XISI) reactor, 18 mm i d . , 400 mm in length. The sketch of the entire apparatus is given in Figure 1. Above and below the catalyst bed the reactor tube was filled with quartz chips. The temperature of the catalyst bed was coiistant within =t 1" and no appreciable temperature fluctuations were re- corded along the catalyst bed during all the esperimental runs. The propylene, dried over Type 3A Davison molecular sieves, was fed as liquid by means of a membrane micro- metric metering pump and vaporized in the preheater portion of the reaction tube. The pressure in the reactor was kept constant (f 1% of the chosen value) by means of a micro- metric discharge valve and was read onto two different cali- brated manometers. The connection between the reactor out- let and the gas chromatograph arid the sampling valve of the same were suitably heated in order to prevent any coiidensa- tion of liquids. The nitrogen for flushing the apparatus and for pressurizing the 1)ropylene feeding reservoir was dried over the same Type 3.1 sieves. AA typical run analysis is given in Table I.

Analysis. The analysis of the substances entering a n d leaving the reactor was performed by means of a FID gas chromatograph, equipped with a copper column, 2 mm i d . , 9 m long, packed with Bentone 34 + didecylphthalate (5% + 5%) on 60-80 mesh Chromosorb W ; column temperature 3 5 O , carrier gas nitrogen, 20 ml/miii. The propane-propylene separation required a second column, 2 mm i.d., 8 m long, packed with 21% propylene carbonate over 60-80 mesh Chromosorb P, operated a t 30"; carrier gas nitrogen, 25 ml/ min. h careful determination of analytical correction factors was performed to ensure precise analytical data.

Experimental Results

Preliminary Runs. Some preliminary runs were per- formed in order to determine the limitiiig values of experi- mental parameters, within which it was possible to obtain reproducible kinetic data. The highest obtainable pressure with the employed pump was 80 kg/cm2. The highest value

Ind. Eng. Chem. Process Des. Develop., Vol. 12, No. 4, 1973 455

Figure 1. Apparatus: PR, propylene reservoir; F, filters; P, pump; R, reactor; BD, bursting disk; TRC, temperature recorder controller; GC, gas chromatograph; GM, gas metering device

Table 1. Typical Run Analysis Run no. Pressure, kg/cm2 Temperature, "C Propylene feed rate, mol/hr Catalyst weight, g Product analysis, mol %

Propane Propylene 4-Methyl-1-pent ene 4-Methyl-2-pentene 1-Hexene 2-Methyl-1-pentene 2-Hexene

1 5-D-140 40 140 0 .29 1 .40

1 .97 97.21 0 .76

0 .06

for teniperature was 150' (absence of detectable catalyst de- activation) and, in order to avoid any condensat,ion of re- zicting gases a t the highest pressure, the safe lowest tempera- ture was 130'. During such preliminary runs i t was confirmed, 3s previously observed by Ilambling (1968, 1969) and Rilkes (196i), that the cat'alyst requires relatively long times, from 3-4 hr to 20-30 hr, to reach t.he perfect steady state. For this reason all the samples for t'lie analysis of reaction products were withdrawn a t least 2 hr after the steady state had been leached, Le., from 5-6 to 25-35 hr aft,er the run was started.

Experimental Data. Three series of experimental runs have been performed. The first series x a s made with the main catalyst, employing the differential reactor technique, a t the temperatures of 130, 140, and 150' and pressures ranging from 18 to 80 kp'cni? (gauge). In this first series catalyst \\-eiglits raiigiiig from 1.40 to 1.68 g, together iTith a propylene f e d flon rate of 0.29 mol;hr, w r e employed and total con- versions ranging froin 0.45 to 4.53 mol % were obtained. The results are giveii in Figure 2. The second series was also per- formed with the main catalyst, but employing t,he integral reactor technique, a t the same temperatures of 130, 140, and 150" and a t the constant pressure of 70 kg/'cm*. All the ex-

1 50 70 30 atm

Figure 2. Rate of 4M1 P formation vs. propylene partial pressure: (0) 130, (A) 140, (m) 150"; solid lines from eq 2; data from Table II

1. .

' g i h lo 15 - mol

Figure 3. Moles of outcoming substances per 1 mol of pure fed propylene vs. time factor (T = 1 30', P = 70 kg/cm2); (0) 4M1 P, (v) propylene, (A) other branched chain hexenes, ( w ) linear hexenes, (+) propane; solid lines from eq 5; data from Table I I

ooU axh '( 15 -

mol

Figure 4. See caption to Figure 3, T = 140"

periments of the second series escept one were performed with catalyst weight,s ranging from 0.21 to 4.97 g and propylene feed flow rates from 0.29 to 0.45 niolihr. For the last run (7 = 55.23) a t 150°, the feed flon rate of the reactant was 0.15 mol/hr. The results are slioivn in Figures 3-5. The third series was conducted \\-it11 all the different kinds of catalysts at rigorously equal values of the experimental parameters, iiamely, propylene feed flow rate 0.15 moll hr, catalyst weight 4.0 g, temperature 150', and pressure i 0 kg,'cm*. I t was found that the promoters affected quite markedly the reaction rate, but they did not significantly alter the product distribu- tion; the reaction selectivity toward 4M1P was aliyays about 85 mol %. The results of the third series are shoivn in Figure 6.

456 Ind. Eng. Chem. Process Des. Develop., Vol. 12, No. 4, 1973

J L 0 2 o gxh 30 40 50 60 - 10

mol

Figure 5 . See caption to Figure 3, T = 150"

Influence of Diffusion

In order to test if external diffusion viould affect our kinetic re.wlts. the nietliod suggested by Corrigaii (1955) has been folloiwd. Some of tlie mo.st significant runs 1%-ere then redone by kwpiiig constant^ the value of time factor T = TI7, F , but rcducing tlie reactant fced ratr b' (mol, hr), and consequently thc catalyst weight TT- ( g ) J until a .mall significant difference was recorded in coilversion for the same value of t'ime factor, but for different flow rates. ?\To influence of external diffusion n-a; rerordcd, eyeii for propylene flow rates as low as 0.145 mol lir.

Coiiceriiing iiiteriial diffysion, since the catalyst pore radii vrry large (2.10,OOO A) and tlie specific surface area is

very low (< 0.15 m2,g): n iiumerical control was coiisidered uniiecesary and it tvas assumed that internal diffusion did iiot alter our kinetic measurements.

Reaction Kinetics and Mechanism

The propylene dimerization reaction may be represent,ed schematically as suggested by Hambling (1969) (Figure 7 ) . Following the Laiigmuir-Hinshelnood theory, the reaction riiechaniam was investigated by aiialyziiig the influence of reactant partial preauure on 4 N 1 P rate of format'ion. Both simple I)eeudo-first- and second-order kinetic equations lvere discarded because the best straight lines drawn either for the 41111' rate of forniat'ioil r1 LS. propylene partial pressure Pp or for r l vs. (PP)? plot.; do iiot go to zero when Pp or ( P p ) ? , respectively, teiid to zero. On the other hand, the trend and thr relative scattering of esperiniental data do not alloIv ail imrnrdinte iiidividua1iz:ltioli of a given mechanism to be stated, so that a certain number of different mechaiiisnis hare been tested. The 41IlP formatioil (see Figure '7) could in fact be conditioned by tlie adsorption of propylene, with formatioil of the allylic uitermediat'e, by the surface reaction betweeii the allylic iiitcrniediate and a second molecule of propylene, coming from the gaseous phase, followilig a Ri- deal-t'yl)e nirchanisln, by the surface reaction between tlie allylic intermediate and a second molecule of propylene! ad- m h e d 011 a iioiimetallatiiig center, or by a trans metallatioil reactioii between the nic:tallate dimer aiid propylene. It was found that the mechanism in which the adsorptioii of propyl-

unied to be the rate-determinirig step allows the be>t interpretation of both differential and integral reactor tecliiiique data. This as:wiiption is riot in contrast with the sup1)ositioii made by Kilkes (196'7) about, the rate-determiii- ing stel) during the initial iiiductioii period of the propylene

Figure 6. Influence of the addition of the various transition metals on the catalyst activity (propylene mol % conversion): values for Na-K*C03 alone (absence of any promoter); 10.05

CH,=CH-CH, -% ICH,- CH-CHJM'

CH, I

~ CH,=CH-CH,-CH-CH; M'

~(CH,=CH-CH,-CH,-CH-CH,lM*

lCH,-Ckl-CHJM+ + CH,-CH=CH,

5% W2=CH-CH,--CH-CH;M'+ CH,-CH=CH,- 4MlP + (CH,-CH-CH1,l M'

1 branched-chain hexene isomers

l C H $ + - C ~ - C l t C H - C K ) M ' + CY-CH=Cb$ + l-hexene + ICH2-CH- CHJM'

linear-chain hexene isomers 1

Figure 7. Propylene anionic dimerization reaction scheme

dimerizatioii catalyzed by potassium su,.l)eiisioiis in a rocked autoclave.

The kinetic equatioii correspoiidiwg to such a niec1ialii;ni may be written as

(1)

where r l represents the rate of 41111' formation (mol, hr g), referred to 1 mol of pure fed propylelie, k l i z the reaction ratch constant (mol, hi- g atin). Pi and bi Lire tlic 1)arti:il ~)rcssiu~cs (atm) aiid the adsorption equilibrium coiistaiits ( : ~ t n - - I ) of the various substances, rehpectively! aiid tlic sul)>cril)ts 1') I), R, and L indicate propylene! 4111P, the branched chaiii, and the linear chain dimeric by-product., respectively. For very I O K (differential) conversions, the influence of the ad- sorl)tion of reaction product. can he neglected and eq 1 writteii as

~- ~~

k1PP 1 + b p P p + ~ D P D + ~ R P R + ~ I Y L

r l = -

r1 = k l P P , ' ( l + bPPP) ( 2 )

which can he linearized, obtaining

1 ,'P1 = (1 SkIPP) + (bp, h) 13)

In Figure 8 the esperimeiitnl values of (1 I r l ) us. (1, Pp) are reported, atid the best straight linesj evaluated by the least- squ:ires method, have been drawl . From tlie liarametrrs of such straight lilies the values of k 1 and bp, reported ill Table 11, hare been calculated. The reaction rate cur re? , c:ilcul:ttrd from such data and from eq 2, are dmwn as yolid lines ill

Figure 2 . For the interpretation of the iiitrgral reactor tecliiiique

data. since both t h r branched chain niid the liiiear chain

Ind. Eng. Chem. Process Des. Develop. , Vol. 12, No. 4, 1973 457

151

Figure 8. ( l / r ~ ) vs. ( l / P p ) : (0) 130, (A) 140, (m) 1 5 0 °

Table II. Reaction Rate and Adsorption Equilibrium Constants

10% 1 05kz, 1 06k3, mol/hr mol/hr mol/hr 1 03bp, 1 ozbD,

r , OC g atm g atm g atm atm-1 atm-1

130 4 .76 2.82 2 .38 12.28 9 . 2 140 1 0 . 5 4.14 5.8 8 .0 8 . 0 150 17 .5 7 . 4 8 1 2 . 1 6 . 3 7 . 0

dimeric by-products are always present in very low concen- tration, it was assumed that the influence of such substances could be neglected and that their rate of formation could be expressed by means of pseudo-first-order equations with respect to propylene (linear dimeric isomers) or to 4111P (branched chain dimeric isomers) partial pressures. I n fact, as suggested first by Pines and h'lark (1956)) Pines and Haag (1958), and Bush, et al. (1965), and then by Hambling (1968, 1969) (see Figure 7), linear dimeric isomers would form with a mechanism similar to that of 4M1P formation, while branched chain dimeric isomers would form from 4M1P by subsequent isomerization reactions. Finally no changes were ever detected in the propane quantity, with respect to that present in the fed propylene. The rate equation system, ex- pressing the kinetics of the overall process, may then be written as

(4)

where r2 and r3 (mol/hr g) are the rates of formation of branched chain and linear dimeric by-products, referred to 1 mol of pure fed propylene, respectively, and k2 and k3 (mol/hr g atm) are the corresponding reaction rate constants.

By indicating amounts of the various substances, present in the gaseous phase, as propylene (w), 4 M l P (x>, branched chain byproducts (y), and linear chain by-products ( z ) , and referring to 1 mol of pure fed propylene, since the feed con- tains u mol of propane per 1 mol of propylene, the moles of the various substances outcomirig from the reactor can be written as

propylene w = 1 - 2x - 2y - 22 propane u 4M1P x branched y linear z

total 1 - x - y - z + u

-3.5

-4.0

-4.5

x 0 - -5.0

- 5 5

- 6.0 5 2Ao 245 -

T

I*)

15

c C -

24

2 J

0

Figure 9. Arrhenius plot for kl(.), kz (m) , k3(A), bp(+), and bdr )

By expressing the partial pressures by means of molar frac- tions and total pressure P (atm), the kinetic equations of the overall process may be written as

(5)

dz d r _ - - ( k 3 w P ) / Z

since u is constant. I n eq 5, Z = 1 - x - y - z + u. The system (eq 5) of differential equations has been integrated numerically by the Runge-Kutta method, by substituting, for k l and b p , the values obtained from the interpretation of differential reactor technique data (see Table 11) and mini- mizing, for each temperature, by the steepest descent method, the objective function

with respect to the parameters k2 , k 3 , and bD. I n eq 6, Cn,m represents the calculated value of the moles of each substance a t the exit of the reactor and Sn,m the corresponding experi- mental value; the subscripts n and m indicate the considered substance and the experimental run, respectively; N and -1f are the maximum number of substances and of experimental data, obtained a t a given temperature, respectively. The results of this integration optimization procedure are reported in the Table 11.

By means of eq 5 and of all the constants of Table 11, the curves, reported as solid lines in Figures 3-5, have been calcu- lated. In Figure 9 the Arrhenius plots, relative to the kinetic parameters of Table 11, are shown. The values of the apparent activation energy of the various considered reactions and of propylene and 4h11P standard enthalpy and entropy of ad- sorption, calculated by means of the geometric parameters of

458 Ind. Eng. Chem. Process Des. Develop., Vol. 12, No. 4, 1973

Table 111. Apparent Activation Energies of Reaction and Adsorption Parameters

AE1 *, kcaltniol AE,*, kcal mol &E3 *) kcall niol

22 15 =t 1 63 16 54 i 1 54 27 64 i 0 7 5

A H p O , kcal/mol -11 36 f 1 14 A H D O , kcaliniol -4 64 z 0 01 A s p o , cal mol "E; -36 99 i 2 76 ASD', cal niol O K -16 26 + 0 01

,201 ,

f l L , , , . I

Fe- Co- Ni Ti-V-Cr

Zr---Nb--&lo W Ta-.-.W Pt i I

Figure 1 1. Decomposition temperature of some transition metals allylic complexes

Figure 10. Electron affinity of transition metals

such A \ r r h t ~ ~ i i ~ ~ . ; plots and of the well-known equation+

h l k i = I l i d i - (AEi* !RT)

111 b i = - - ( A H i o R T ) + (AS,", R ) ( 7 )

( 8 )

arc reportcti in Table 111. Conc'eriiiiig the iiiterpretation of the data reported in

Figure 6, diffei,ent liyliotheses can be made. For example, it is knoivn (Eberliardt, 1!366) that the polarity of tlie metal to carlioii-boiltied iiitertnetliates bond places a varying aniount of negative charge on the carbon. The presence of tlie various transition niet,als, with varying electron affinity, could then iiiflueiice the polarity of the alkaline metal t o organic inter- iiicdiatt. bond. with the coi i ,quei i t alteration of the negative charge oil the carbon at 'mi bonded to the alkaline metal. The forniation of tlie allylic caoniplex would then be eahier when tlic negative charge 011 the allylic complex i,5 lowered by the preseiice of a transition metal atom n-ith higher electroil nffiiiity. This 1iypotlie.ji:: seems to be partly coniirnied hy the behavior of the plot (Figure 10) of the electroii affinity of the various traiibitioii metals, takeii from the literature data (Zollwg, 1969), which is quite similar to that of catalytic activity reliorted in Figure 6.

011 the otlier hand, ah a .;ecoiid hypothesis, the formatioil of tlic allylic complex coultl be niade easier also by the preseiice of a nietnl i\-liicli gi1-e a more stable allylic conililes with propyleiie. This second 1iypotliesi.s seems to be better con- firmed by esl~erinieiitai data, .iiice a plot (Figure 11) of the deconiiio4tioii tcmiieratureb of allylic complexes of transition nictuls, takeii from the data of Kilke, et a l . (1966), shon-s a lieliuvior very h i l a r to that of tlie relative catalystic activity,

with two maxima, corre,sponding to the triades of chromium and of nickel and minima corresponding to the triades of titanium and of vanadium and to iron.

The consistency of the behavior of the relative catalystic activity with that of both the electron affinity of transition metals and of the decomposition temperatures of transition metals allylic complexes seems to confirm t'he validity of the adopted kinetic model, implying the adsorption of the pro- pylene as rate-determining step of the overall reaction ki- netics.

Conclusions

The kinetic scheme employed in the present work proved to be useful for tlie interpretation of all the collected differ- ential and integral reactor experimental data with a unique set of kinetic parameters. Xlso, tlie approximations introduced during the formulation of the kinetic equations proved to be satisfactor>-. The given set of kinetic constants, together with the system of eq 5 , could then coiistitut'e a basis for reactor design. S o discrepancies were noted with respect to the rela- tively few data of the literature. The proposed interpretations of the effect of added transition metals to the catalyst are in fact incoiisistent neither n-ith previously proposed mechanisms nor with our kinetic and mechanistic data.

literature Cited

.%inaud, P., Bcig. Chem. Ind., 33, 1101 (1968). Buyh, \V. V., Holsman, G., Shaw, A. \Y,, J . Org. Chern., 30, 3290

( 1% ,i ) . Corrigan, T. E., C h o n . Eng. i.\-ezu Y o r k ) , 62, 199 (1955). Eberhardt, G . G. , Organornetal. Chem. Reo., 1, 491 (1966). Forni, L. ( t o Yoc. Ital. Itesine), Italian Patent 913037 (1972). Rambling, J . K., Xorthcott, R. P., Rubber Plast. Age, 48, 224

(1968). IIambling, J. K., Lecture presented to the Symposium on -4d-

vances in Chemiatrl-, Manchester. Institute of Sciences and Tpchnolonv. .June. 1968.

HanGing, J."&) ~ ih . Brit . , 5 , (81, 354 (1969). Pines, H., Mark, V., J . A r n c r . Chem. SOC., 78, 4316, 5946 (19.36). Pine;, I*., Haag, W. O., J . Org. Chem., 23, 328 (19%). Wilke, G., Bogdanovic, B., Hardt, P., Heimbach, P., Keim, W.,

Kroner, LI., Oherkirch, W., Tanaka. K.. Steinriicke. E. , Wal- , ,

ter . 11.. Zimmerrnann,' H., ' Anuew. e h e h . . Int. Ed.' E&/.. 5 . , , " " , I

2, 151 il966). \Tilkea, J. B., J . Org. Cheni., 32, 3231 (1967). Zollweg, 1:. J., J . Chern. P h y s . , 50, 4251 (1969).

RI:CP:IVI:D for review October 24, 1972 ACCEPTKD March 20, 1973

The financial support of Italian Consiglio Sazionale delle Iticerche i h acknowledged.

Ind. Eng. Chem. Process Des. Develop., Vol. 12, No. 4, 1973 459