Ind. Eng. Chem. Prod. Res. Dev. 1980, 19, 205-208
205
Isomerization of rn-Xylene without Side Reactions, over a Perfluorinated Polymer Sulfonic Acid. A Kinetic Study Paolo Beltrame, Pler Luigl Beltrame, Paolo Carniti, and Guido Nespoli Isfituto di Chimica fisica, Universifi, 20 733 Milano, M y
A solid superacidic catalyst was obtained from the corresponding potassium salt (Nafion-K)and employed for studying the rn-xylene isomerization in a fixed bed flow reactor. Temperature ranged from 143 to 173 "C. A mixture of m-xylene with toluene, trimethylbenzenes, and 1,2,4,5-tetramethyIbenzene,in ratios closely approaching the overall equilibrium between polymethylbenzene fractions, was fed in a nitrogen stream. Disproportionation reactions were thus avoided, a clean isomerization of m- to 0-and p-xylene observed, and its kinetics measured. Since the catalyst deactivates, checks of its decreasing activity were performed at intervals and a correction of kinetic measurements devised, in order to evaluate the rate coefficients for a fresh catalyst.
Perfluorinated polymer sulfonic acids (Nafion-H) obtained from commercially known polymeric materials (Nafion perfluorosulfonic ion-exchange membranes) have been employed as catalysts for alkylations, rearrangements, toluene disproportionation, and alkane isomerizations (Kapura and Gates, l973; McClure, 1977; McClure and Brandenberger. 1977; Olah et al., 1977; Kaspi and Olah, 1978; Kaspi et al., 1!378; Onoda and Wada, 1976). We tested a catalyst of this type for isomerization of m- to 0and p-xylene (Beltrame et al., 19781, due to the practical interest of the latter isomers. However, also secondary reactions (mainly disproportionations) were found to take place, giving rise to toluene, trimethylbenzenes, and small amounts of benzene and tetramethylbenzenes. The present work was undertaken with the aim of avoiding side reaction;$by feeding an appropriate mixture of alkylbenzene fractions, close to overall equilibrium (1)
?l
=
1, 2, 3, 4, 5
apart from the presence of meta isomer alone in the xylene fraction. This is also an illustration of a general technique that can be used for studying kinetics of useful reactions accompanied by undesired side reactions.
Experimental Section GC Analysis. A flame-ionization apparatus was employed. The stainless steel column (3.5 m; i.d. 2 mm) contained 10% Bentone 34 diisodecyl phthalate (1:l) on Chromosorb P (80--100mesh). Nitrogen (20 mL/min) was used as carrier. Column temperature was programmed from 80 to 110 "C (heating rate 16 "C min-'). Calibration factors were determined by using toluene as internal standard. The relative amounts (mol) of analyzed products were normalized. Catalyst's Acidity 'Titration. A 0.01 N NaOH aqueous solution and ca. 0.05 g of catalyst dispersed in a 5% NaCl aqueous solution (100 mL) were used, according to a reported procedure (Beltrame et al., 1978; Grenall, 1949), for a potentiometric titration with an automatic buret at slow speed (0.15 mL/min). Apparatus and Runs. Kinetic runs were performed by a procedure previciusly described in detail (Beltrame et al., 1978). The tubular flow reactor, with internal diameter of 0.51 cm and 75 cm long, was obtained by con-
+
0196-4321/80/1219-0205$01.00/0
necting five parallel AIS1 304 stainless steel tubes. The reactor, containing a weighed amount (13.0 g = 17.4 mL) of solid catalyst, was assembled in an oven with a forced circulation of hot air and kept a t constant temperature f l . O "C. The feed of alkyl aromatics (commercial analytical grade products) was varied in the range 0.015-0.073 mL min-l by a syringe micropump. The liquid organic fraction was vaporized and mixed with nitrogen, in a molar ratio 1:4, preheated, and sent to the reactor. Contact time 0 was evaluated by the ratio of volume of the catalyst to the total volumetric feed measured at the temperature and pressure conditions of the reactor. Values of 0 ranged from 10 to 53 s and were varied by modifying the feed rate. Products, collected in a freeze trap (acetone/dry ice) during 15-20 min, were analyzed by GLC. At least 1 h was left to elapse before starting measurements to ensure stationary conditions. Additional runs with the more finely powdered catalyst (80-120 mesh) were carried out in a shortened reactor (1 or 3 tubes). Catalyst. The polymer potassium salt Nafion-K 1200 E.W. (Du Pont) was ground in a centrifugal mill after freezing at liquid nitrogen temperature. The 30-80 mesh fraction was acidified (McClure, 1977; McClure and Brandenberger, 1977) by a threefold treatment with H2S04 30% under stirring during more than 5 h. The resulting acid form Nafion-H was filtered and washed with water up to pH 5 and negative test for sulfate ion. It was finally dried in a rotary desiccator at 100 "C during 16 h under vacuum (1 mmHg). Its apparent bulk density, obtained by weighing a known volume of the solid packed as in the reactor, was found to be 0.75 g/mL. The measured acidity was 0.781 rnequiv/g. An analogous procedure was followed in the case of the 80-120 mesh fraction; its acidity and apparent density were found to be 0.763 mequiv/g and 0.72 g/mL, respectively. The indication Nafion 2 refers to the catalyst prepared as described above; Nafion 1 is the catalyst sample employed during the previous research and partially deactivated (Reltrame et al., 1978). Acid Treatment of Used Catalyst. Attempts to regenerate partially exhausted catalyst (Nafion 1) were carried out with chromic/sulfuric acid (50 "C; 4 h), HNOB 50-70% (50-70 "C: 2-4 h), and H2S0496% (20 "C; 43 d). These treatments were practically ineffective since the acidity of catalyst as well as its activity, in terms of kinetic coefficient of m-xylene isomerization, resulted unaltered, when not slightly depressed, after the treatment. No re8 1980 Ameiican Chemical Society
206
Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 2, 1980
Table I. Preliminary Run a t 153 “ C (e = 45 s). Molar Fractions in Reagent and Product Mixtures. Catalyst: Nafion 1 feed benzene toluene xylenes (0) (m)
(P)
trimethylbenzenes (1,3,5) (1,2,4) (1,2,3) tetramethylbenzenes
0.500
0.159 0.317 0.024 1.000
very good approximation, the only reaction in the system m-xylene
product
k, k-8
0,p-xylene
(2)
p-xylene were always obtained in ratios around 0.8, close to the calculated equilibrium value. Therefore, they were treated as a single product. A reversible process, first order in both directions, was considered to be at work. Results were expressed as molar fractions within the condensed organic phase. The fraction of m-xylene in the mixture of the three isomers was obtained as f, = x,/(x, + x, + xp). The corresponding fraction at equilibrium (4,) was calculated for each temperature value from thermodynamic data (Stull et al., 1969). A fraction fo,, may be defined as (1 - f m ) , and analogously an equilibrium fraction 4,,, as (1- 4,). Values of &, were in the range from 0.4352 (143 “C) to 0.4416 (173 “ C ) . The degree of approach of the system to the isomeric equilibrium was evaluated as y = fo,p/4Jo,p. The integrated rate equation for the case under examination (Levenspiel, 1972) is -do,, In (1- y) = k,6’ (3) 0-and
0.002 0.450 0.043 0.084 0.043 0.116 0.190 0.016 0.056 1.000
generation of catalyst Nafion 2 was attempted.
Results and Discussion Evidence that the disproportionation of m-xylene can be depressed by the intervention of the reverse reaction was obtained by carrying out a preliminary run with a feed of toluene and trimethylbenzenes. Reaction products (Table I) prove that mainly two reactions of type 1took place, that is the forward reaction with n = 2 and the backward reaction with n = 3. The two processes had about the same weight, since converted trimethylbenzenes (Ax = -0.178) are about three times as much as converted toluene (Ax = -0.050), while, in the product, xylenes (Ax = 0.170) amount to about three times the tetramethylbenzene fraction (Ax = 0.056). In a series of preliminary trials, the composition of the ‘‘overall equilibrium” feed was progressively adjusted, until a satisfactory mixture was found, In fact, its composition compares well (Table 11, method A) with the average composition of the products of 20 kinetic runs, when mxylene isomerization is not considered. A dependence of this products composition on temperature was not observed, being probably covered, within the narrow experimental range (153-173 “C), by the analytical uncertainties. Therefore a single feed composition was used for all the relevant kinetic runs. Within the product average composition it should be considered that the trimethylbenzene fraction contains the three isomers in ratios (1,3,5:1,2,4:1,2,3 = 34:61:5) close to those calculated from thermodynamic data (Stull et al., 1969) (1,3,5:1,2,4:1,2,3 = 32:63:5 in the temperature range 150-170 “C). During the subsequent group of kinetic runs (Table 11, method B) a further adjustment of the “overall equilibrium” feed was performed: results confirmed the previous conclusions. Working with feed compositions close to “overall equilibrium” the isomerization of m-xylene (2) was, to a
where 0 is contact time. However, the time on stream of the catalyst ( t ) also proved to have an influence on the kinetics. When comparing the results of runs carried out in the same conditions and with the same 0 value but different t values (Table 111; see x and y values for runs 1, 4, 9, and 20), it is apparent that the catalyst activity is a decreasing function of t . This catalyst deactivation was not so fast as to affect the results of a single kinetic run, but fast enough to induce large errors in the comparison of different runs, if not avoided or accounted for. However, to avoid it would have required using a fresh catalyst sample for each kinetic run, that is an overall quantity of catalyst much larger than was available. So we chose to introduce a correction for deactivation. The catalyst aging was followed by a series of check runs, all at the same temperature (153 “C)and many of them at the same 6’ value (27.3 s). A relative activity was defined by F = [In (1 - y)/flI,/[ln (1 - y ) / ~ I 1 . ~ (4) that is by comparing the activity of the catalyst at 1.53 “ C after a given time on stream t (at whichever temperature) with the activity of the same catalyst after 1.5 h on stream at 153 “C, the latter being considered as the reference or “fresh” catalyst. The relative activity F,if known as a function of the operating time of the catalyst, allows one to correct a kinetic measurement performed on a partially deactivated catalyst, obtaining the value that would have been mea-
Table 11. Kinetic Runs a t 153-173 “ C (Method A: a Single Catalyst Sample for t h e Whole Set) and at 143-173 ” C (Method B: a Fresh Catalyst Sample for Each Temperature). Molar Fractions in Reagent and Product Mixtures. Catalyst: Nafion 2 (30-80Mesh) method B method A benzene toluene xylenes (0)
(m) (n) I \r
trimethylbenzenes (1,3,5)
(1,294) (1,2,3) tetramethyl benzenes a
feed
producta
0.262
0.001 i 0.0004 0.262 F 0.004
0.263
0.261
4
0.004
0,488i 0.004
0.489
(0.493
i
0.005
0.077
0.077 0.138 0.013 0.020 1.000
0.485 0.081 0.139 0.013 0.020 1.000
Average values with their standard deviations.
{
k 0.002 0.138 i 0.003 0.013 i 0.0005 0.021 i 0.001b 1.000
feed
producta 0.001 i 0.0003
0.078 i 0.002 0.135 I0.003 0.012 i- 0.0004 0.020 i 0.001 1.000
Besides traces of pentamethylbenzene (ca. 0.0001).
_.
Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 2, 1980 207 Table 111. Example o f Kinetic Measurements. Runs a t 153 "C (Method A; Catalyst: Nafion 2, 30-80 Mesh). (Determination of Components Other Than Xylenes Is Not Reported)
~ - - run no. 0 , s
x,
xo
Y
XP
a
11 10.6 0.0228 0.4428 0.0258 0.226 0.177 13 14.3 0.0283 0 4 2 4 3 0.0336 0.291 0.248 1 27.3 0.0602 0 3524 0.0812 0.654 0.465 4 27.3 0.0565 0 3614 0.0744 0.615 0.465 9 27.3 0.0490 0 3832 0.0602 0.507 0.465 20 27.3 0.0385 0 4 0 7 2 0.0459 0.393 0.465 1 2 36.8 0.0555 0 3 6 6 5 0.0680 0.577 0.611 10 51.0 0.0682 0 3 3 3 1 0.0835 0.716 0.853
1, h
17.5 20.3 1.5 6.1 14.4 29.7 19.3 16.4
- ( G ~ . ~ / In F )(1 - Y).
a
1,0r9 . ...... I \
..-.....
------
..-,
7 \ 1 -'-
(A)
Figure 2. Kinetic runs by methods A and B on Nafion 2 catalyst. -~ ,------Ink,,
081-
I
\
,
I
0.4L
_-__-__ 10
.~LLILL--d
~-L
2o
llh)
30
lo
l(h)
2.35
2.40
2o
2.25
2.30
1O3/T
2'20
Figure 1. Relative activity vs. time on stream of Nafion 2 catalyst during kinetic runs on 21 single (A) or different (B) samples.
Figure 3. Arrhenius plot. Measured coefficients are represented by circles (method A) and full points (method B).
sured in the same conditions, if a fresh catalyst sample had been used. The correction was effected by employing eq 5 instead of eq 3 --(&,,3/F)In (1 - u ) = k,B (5)
Table IV. Rate Coefficients for m-Xylene Isomerization. Catalyst: Nafion 2 (30-80 Mesh)
and substituting in it the appropriate value of F for each kinetic run. In principle, the knowledge of F requires, besides the few check runs that were actually carried out, a rate equation for the deactivation process. This is so because, on a plot of F vs. t, the points corresponding to the check runs have to be join'ed by lines whose shape depends on the kinetic order of the catalyst, deactivation. At constant temperature, a straight line corresponds to a zero-order deactivation, while 21 logarithmic curve corresponds to a first-order deactivation, etc. (Rase, 1977). In practice, the difference between the lines drawn under a zero-order or a first-order assumption is very small, provided a sufficient number of check runs has been made. Therefore we used straight lines, that is, a zero-order approximation (Figure 1). When working with a single catalyst sample for a whole set of runs (method A), the deactivation plot was found to contain, as expected, a zig-zag line, different slopes corresponding to different reaction temperatures (Figure 1A). When method B was used, that is each catalyst sample was used only for a limited series of runs at constant temperature (apart from the check runs at 153 "C), the plot had a different appearance (Figure lB), but the deactivation kinetics were again found to be dependent on temperature. If first-order deactivation kinetics are assumed, plots very similar to those of Figure 1 are obtained; the corresponding deactivation coefficients range between 0.01 h-' (143 "C) and 0.04 h - ' (173 "C); that is, the catalyst loses
method A.
method B. 102ki,a S-'
143 153 163 173
1.68 2.21 2.78
i
i ?
0.01 0.04 0.10
previous work. 102k,a S-'
1.14 1.70
i i
0.01 0.02
2.69
i
0.13
0.70 1.11 1.90 2.58
a Uncertainties are standard deviations. Average values of coefficients from runs o n rn-xylene and o n rnxylene -t toluene; use has been made of t h e Arrhenius equations (Beltrame et al., 1978).
from 1to 4% of its activity per hour on stream, according to temperature. Kinetic measurements were done by method A at 153, 163, and 173 "C; by method B at 143,153, and 173 "C. A 70-90% approach to isomeric equilibrium was reached in the different cases, depending on temperature. As an example, results of runs at 153 "C (method A) are given with some detail in Table 111. All kinetic runs are graphically presented in Figure 2, where the left side of eq 5 is plotted vs. 8. Coefficients kiwere obtained as slopes of the least-squares straight lines passing through the origin. These coefficients are given in Table IV. It can be appreciated that methods A and B gave the same results, within experimental error. As a consequence, all points, independently of the measurement method, were reported on a single Arrhenius plot, which is fairly linear (Figure 3). The following apparent activation parameters were obtained: A = 3.20 X lo3 s-l (log A = 3.51 f 0.34); E = 10.3 f 0.7 kcal mol-'. As shown in Table IV, kinetic coefficients of the same order of magnitude were obtained
208
Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 2, 1980
Table V. Dependence of Nafion 2 Catalytic Activity on Particle Size. Temperature 1 7 3 “C no.
feed flow rate mLs-l
catalyst size, mesh
of reactor runs min
30-80 80-120 80-120
5 tubes 3 tubes 1 tube
6
0.36 0.35 0.21
max
1OZki, s-’
1.71 2.78
ra
0.984 0.975 0.999
4 0.89 3.67 4 1.70 3.19 a Correlation coefficient of the least-squares straight line passing through the origin, corresponding to eq 5.
in this and in the previous work (Beltrame et al., 1978). The main difference is in the temperature dependence, that is in the Arrhenius parameters: on the average, previous measurements had given A = 2 X lo6 s-I (log A = 6.3); E = 16.0 kcal mol-l, for work in the 149-171 “ C range. Results of the two works should be comparable, since either the catalyst deactivation was slow (previous work; ca. 10% in 100 h) or a correction was applied (present work), so that in both cases kinetic coefficients measure the activity of a substantially “fresh” catalyst in isomerizing m-xylene. A lower apparent activation energy and a possible curvature of the Arrhenius plot in the present case (Figure 3) could be taken as evidence of mass transport phenomena interfering with the kinetics. However, typical tests of diffusional control gave negative results. First of all, feed flow rates were often varied by a factor around 5 during kinetic runs on the 30-80 mesh catalyst, with no evidence of systematic deviations from linearity in kinetic plots (Figure 2). Additional runs on the 80-120 mesh catalyst a t 173 O C (Table V) confirmed the previous observations, ruling out any influence of external diffusion on the kinetics. In the internal diffusion test a difference has been found between activities of 30-80 and 80-120 mesh catalysts (Table V). However, BET surface area measurements (nitrogen, 77 K) have shown that the Nafion catalyst has a very small area (
