I n d . Eng. Chem. Res. 1987,26, 2245-2250
Appendix The reaction which proceeds during the period from tl to t 2 in Figure 9, plot 2, can be expressed by the conventional equation for pseudo-nth order reaction as
2245
atn=1 In C4/C, = k(tz - tl) atn # 1 c41-n
- C31-n = k(t,
- tl)
(A7)
Literature Cited As for the reactions during the heating and the cooling steps shown in Figure 9, the following equations can be applied:
Beuther, H.; Schmid, B. K. 6th World Petroleum Congress, Frankfurt, Germany, June 1963, Section 111, Paper 20, p 297. Chang, C. D.; Silvestri, A. J. Ind. Eng. Chem. Process Des. Deu. 1974, 13, 315.
Kosugi, M.; Yoshizawa, T. J. J p n . Pet. Inst. 1978,21, 199. Newson, E. J. Prep.-Am. Chem. Soc., Diu. Pet. Chem. 1970,15(4), A141.
Mears, D. E. In Chemical Reaction Engineering II; Hulburt, H. M., Ed.; ACS Monograph Series 133; American Chemical Society: Washington, DC, 1974; p 218. Ohtsuka, T.; Hasegawa, Y.; Koizumi, M. Bull. J p n . Pet. Inst. 1967,
-Lf"C-. dC = x t 13 k l 3dt
9, 1.
Oleck, S. M.; Sherry, H. S. Ind. Eng. Chem. Process Des. Deu. 1977, 16, 525.
-Lf'C-" dC = xzt4kz4dt Since the rate constants in eq A2-A4 are the functions of temperature alone and the rate of heating is equal to that of cooling, all of the right sides in eq A2-A4 are equal. Then, we can get the concentrations, C1 and Cz,which cannot be measured from experiments, as in the following: atn=l
Paraskos, J. A.; Frayer, J. A.; Shah, Y. T. Ind. Eng. Chem. Process Des. Deu. 1975, 14, 315. Riley, K. L. Prey.-Am. Chem. SOC.,Diu. Pet. Chem. 1978, 23(3), 1104.
Sato, M.; Takayama, N.; Kurita, S.; Kwan, T. Nippon Kagaku Zasshi 1971, 92, 834. Shimizu, Y.; Inoue, K.; Nishikata, H.; Koinuma, Y.; Takemura, Y.; Aizawa, R.: Kobayashi, S.; Egi, K.; Matumoto, K.; Wakao, N. Bull. Jpn. Pet. Inst. 1970, 12, 10: Shimura. M.: Shiroto. Y.: Takeuchi. C. Prem-Am. Chem. Soc.. Diu. Colloid Surf. Chehz.,'Las Vegas Meet., March 1982, 30. Ta", P. W.; Harnsberger, H. F.; Bridge, A. G. Ind. Eng. Chem. Process Des. Deu. 1981, 20, 262. Todo, N.; Kabe, T.; Ogawa, K.; Kurita, M.; Sato, T.; Shimada, K.; Kuriki, Y.; Ohshima, T.; Takematsu, T.; Kodera, Y. Kogyo Kagaku Zasshi 1971, 74(4), 563. Wheeler, A. In Advances in Catalysis; Academic: New York, 1951; '
C1 = (CoC4)"2
C2 = C3(Co/C4)1'z
(A5)
atn # 1 cll-n
=
(C41-n
C21-. = ( 2 C p +
+ Co1-")/2 c01-n - C*1-")/2
646)
Substituting (A5) and (A6) in (Al) derives
p 249.
Received for review August 14, 1986 Accepted June 26, 1987
Kinetic Study on the Hydrotreating of Heavy Oil. 2. Effect of Catalyst Pore Size Satoru Kobayashi,* Satoshi Kushiyama, Reiji Aizawa, Yutaka Koinuma, Keiichi Inoue, Yoshikazu Shimizu, and Kozo Egi National Research Institute for Pollution and Resources, 16-3, Onogawa, Yatabe-machi, Tsukuba, Ibaraki 305, Japan
The hydrodemetalation reactions of residual oil were carried out by use of a trickle bed reactor to investigate the relation between catalyst pore diameter and activity. The catalysts used were 3% molybdenum on alumina and silica-alumina. The results showed that the optimum pore diameter was located around 100-150 A at 400 "C. The optimum pore diameter increases with the elevation of reaction temperature, and the curve showing the change of metals removal with pore diameter becomes broader a t higher reaction temperature. The selectivity between two reactions, such as vanadium and nickel removal, passes through a maximum with varying the pore diameter. These phenomena were satisfactorily explained by a simple kinetic equation including the Thiele modulus. In our previous paper (Kobayashi et al., 1986), we investigated the effect of catalyst pellet size on hydrodemetalation (HDM) of residual oil by the use of commercial Co-Mo/A1,03 hydrodesulfurization (HDS) catalysts. The results showed that, above 1-mm pellet sizes for large pore catalysts and even at pellet sizes of less than 0.5 mm for catalysts whose pore diameter was as small as 70 A, HDM activity was significantly affected by pellet size. 0888-5885/87/2626-2245$01.50/0
This indicates that HDM reaction is influenced by intraparticle diffusion in the usual HDS catalysts. This diffusion limitation reduces the effectiveness factor of the catalysts, resulting in the decreased apparent activities of heavy metals removal. It is possible to solve this problem by extending the pore diameter of the catalysts. However, the surface area decreases with increasing pore diameter of the catalysts. This again decreases the apparent activity. 0 1987 American Chemical Society
2246 Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987 Table 11. Pore Size Distribution of Supports
Dore distribution, vol %
I
name
-70A
A-93 A-100 A-105 A-142 A-160 A-255 A-265 A-540 SA-65 SA-70 SA-79
16
SA-87
ii
Figure 1. Flow diagram of experimental apparatus: A, feed reservoir: B, feed pump; C, H2cylinder; D, low-pressure H, accumulator; E, H,compressor; F, high-pressure Hzaccumulator; G and I, pressure regulator; H, reactor; J, low-pressure separator; L, wet gas meter.
SA-93 SA-104
70-
100-
150-
1008, 66 38 32
1508,
3008,
30020008,
11
1
1
2
2
1
46 49 52 26 20 12
0
0
2
56 46 41 19 14
33 44 19 40 42 23
6 1 11 21 21 46
8 7 5 1
5
2 4
12 21
13
-20008, 4 4
2
1
8
28
4
61
7
6 4
18 16 18
36 46 62
11
2
1
2
1 19 2 2
2 9
7
2
12 12 10
3 18 1
3 3
5
Table I. Properties of Supports
PV,mL/e
A
PS and shape, mm, mesh
0.61 0.50 0.69 0.56 0.95 0.51 0.51 0.48 0.65 0.54 0.79 0.68 0.70 0.73
93 100 105 142 160 255 265 540 65 70 79 87 93 104
1.3 extrudate 16/25 sphere 16/25 sphere 16/25 sphere 1.6 extrudate 16/25 sphere 16/25 sphere 16/25 sphere 1.3 extrudate 16/25 sphere 16/25 sphere 16/25 sphere 16/25 sphere 16/25 sphere
PD,
dbt
name A-93 A-100 A-105 A-142 A-160 A-255 A-265 A-540 SA-65 SA-70 SA-79 SA-87 SA-93 SA-104
comDosition
e/mL
0.68 0.79 0.65 0.72 A183 0.52 A1203 0.83 A1203 0.79 A1203 0.92 A1203 10% Si02-A1203 0.67 5% Si02-A1203 0.84 10% Si02-A1,03 0.58 11% Si02-A1203 0.69 7% Si02-A1203 0.69 5% Si02-A1203 0.67 A183 A1203 A183
Therefore, it is considered that the optimum pore diameter may be a function of the effectiveness factor and surface area. The investigations on the relation between catalyst pore structure and HDM activity have been already carried out by many researchers (Hardin et al., 1978; Plumail et al., 1982; Riley, 1978; Shimura et al., 1982). However, there were few studies which interpreted the experimental results from the viewpoint of detailed kinetic analysis. Therefore, we attempt in this paper to analyze our experimental result on the relation between HDM activity and pore diameter by use of the Thiele modulus (Wheeler, 1951).
Experimental Section (1) Reaction Procedure. Experiments were carried out by trickle-bed reactors containing 100-200 cm3 of catalysts. The schematic flow diagram of the reaction apparatus is shown in Figure 1. Reaction conditions were temperature, 400 "C;hydrogen pressure, 100 kg/cm2; LHSV, 1.0 h-l; and H,/oil ratio, 1000 nL/L. The oil used for reaction was Khafji residual oil. Its properties have been described in the previous report. Analytical procedures of liquid products have also been described on the previous report. (2) Catalysts. Tables I and I1 summarize the main characteristics of the supports used to prepare the catalysts for this study. The catalysts used were prepared by wet impregnation according to the following procedure: (1) calcining the supports in air at 500 "C for 2 h; (2) wet impregnation with aqueous ammonium molybdate; (3) evaporating to dryness on a water bath; (4) calcining in
IO0
200
300
Pore Diameter,
400
500
A
Figure 2. Relationship between vanadium removal and pore diameter: (OAO) Si02-A1,0,; (.Am)
A1203.
air at 500 "C for 3 h. The fixed Mo content for all catalysts was 3.0 wt %. In order to find the exact optimum pore diameter, it would be ideal for all the support materials to have the following characteristics. (1) The chemical properties should be the same. (2) The pore size distributions should be monomodal and most pores should be concentrated within a narrow range of sizes. (3) The total pore volumes should be the same. (4) The pellet sizes should be the same. In practice, however, it is impossible to obtain a set of many supports of different pore sizes and maintain the above restriction. Therefore, in this study, we tried to conquer and compensate for the difficulties by using a sufficient number of supports which were selected so as to meet the above conditions as much as possible. The supports thus selected are listed in Table I. Since it was difficult to obtain sufficient kinds of alumina supports whose mean pore diameters are smaller than 90 A, several silica-alumina supports shown in Table I were used to deduce the pore size effect in that range. The discrepancy in the chemical and physical properties of the selected supports from the viewpoint of the above-mentioned ideality could make the estimate of the optimum pore size for HDM reactions somewhat inaccurate, but it would be possible to elucidate the overall tendency in terms of the relation between pore size and activity of catalysts for HDM reaction.
Results The relation between vanadium removal (DV reaction) and pore diameter of the catalysts is shown in Figure 2. Although there is some scattering in the data, it can be seen that the pore diameter which gives the highest re-
Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987 2247
s 601
B
' E
40-
-
3 c 3
m
20-
OL
100
2017
300
Pore Diameter,
400
500
1
8
200
300
Pore Diameter,
400
100
200
300
Pore Diameter,
F i g u r e 3. Relationship between nickel removal and pore diameter.
100
L
500
400
I
500
A
F i g u r e 5. Relationehip between sulfur removal and pore diameter.
1
8
F i g u r e 4. Relationship between asphaltene removal and pore diameter.
moval is located around 100 8,or larger. In addition, the optimum pore diameter increases with the elevation of the reaction temperature; that is, the optimum pore diameter seems to be present at about 100 8, at 380 "C, in a range of 100-150 8, at 400 O C , and around 200 8, at 420 "C. In the case of nickel removal (DNi reaction) (Figure 3), the optimum pore diameter is present at a position of less than 100 8, at a reaction temperature of 380 "C and at about 100 A at 400 "C. Thus, the optimum pore diameter for the DNi reaction seems to be smaller than that for the DV reaction. Also the phenomenon of the higher the reaction temperature, the larger the optimum pore diameter is not obviously observed compared with the case of the DV reaction. In Figure 4, there is shown the relation between asphaltene removal (DAs reaction) and pore diameter. The optimum pore diameter in this case seems to be present at positions of about 150 8, or larger at any reaction temperatures. Although the phenomenon that the position of the pore diameter does not vary so much with the reaction temperature is similar to that in the case of the DNi reaction, it is different in that the optimum pore diameter is higher than that in the DNi reaction. The relation between sulfur removal (DS reaction) and pore diameter is shown in Figure 5. In this case, there is no maximum in the range of pore diameter used in this experiment, and the optimum pore diameter, if there is one, seems to be present at a position less than 90 A. We previously reported the influence of pore diameter on the DS reaction using Co-Mo/A1203 catalysts and showed that pore diameters of 70-80 8, are optimum (Inoue et al., 1973). Thus, the results of the present study are not in-
40
60
80
1
Nickel Removal, %
F i g u r e 6. Selectivity between vanadium and nickel removal: (0) reaction a t 380 "C, (A)a t 400 "C, ( 0 )a t 420 "C; (0) 80 A under, (0' 80-100 A, (e) 100-150 A, ( 0 )150-300 A, ( 0 )300 A over.
consistent with the previous ones. Since the sulfur compounds are relatively uniformly distributed in all fractions of residual oil compared with heavy metals, it can be considered that also the average molecular size of sulfur compounds is small and, as a result, they are not as influenced by intraparticle diffusion resistance, resulting in the optimum pore diameter becoming small. When vanadium (V) removal and nickel (Ni) removal are compared (Figure 6), all experimental values for V are higher. That is, V is more easily removed than Ni. Since V and Ni are concentrated in the same component, asphaltene (As), the diffusion rates of both are thought to be nearly the same. Accordingly, the results suggest that the contribution of the diffusion resistance to the DV reaction is larger than that to the DNi reaction. This supports the difference between optimum diameters of DV and DNi reactions mentioned above. In Figure 6, there can also be seen a tendency that the selectivity for V removal relative to Ni removal continuously decreases with a decrease in the pore diameter. Such a variation of the selectivity with pore diameter appears particularly remarkably in the relation between DV and DS (Figure 7 ) . Similar phenomenon can be seen also in the relation between DNi and DS. What can be said commonly on these relations is that they are closely related to the extent of contribution of the diffusion resistance. That is, with large pore catalysts, a reaction which is more strongly influenced by diffusion resistance proceeds relatively more easily than a reaction in which the contribution of diffusion resistance is small, and the reverse holds with small pore catalysts. Although a simple interpretation has to be avoided because
2248 Ind. Eng. Chem. Res., Vol. 26, No. 11, 1987 I 90
E 2
43
U 0
C > 0
20, lo'
Oc---. 2c
40
Sulfur
80
60
Removal,
100
130
200
Pore
Yo
Figure 7. Selectivity between vanadium and sulfur removal. Legends are the same as in Figure 6.
the initial concentration of the reactant or the reaction mechanism is different in each reaction, the relation mentioned above can be explained with a simple model of intraparticle diffusion as will be developed in the next paragraph.
Discussion In our previous paper (Kobayashi et al., 1986), we note that, where the contribution of thermal cracking is small in the hydrodemetalation of heavy oils, kinetic analyses by use of the Thiele modulus (eq 1) are possible. Ac-
300
De = K(1- MD/PD)4
(2)
500
Figure 8. Relationship between theoretical conversion and pore diameter.
where K is a proportional constant, MD is the diameter of the reactant molecule, and PD is the pore diameter. In eq 2, when MD of V compounds is assumed to be 50 A (Pollack and Yen, 1969; Egi et al., 1978) and the value of De a t a PD of 175 A (1-10 X cm2/s) is applied, the proportional constant ( K ) becomes 4-40 X lo-' cm2/s. Therefore, the relation between De and PD of the V compounds at 400 "C can be represented by De = 4-40
X
10-$(1- 50/PD)4
(3)
( c ) Surface Area: S . The surface area ( S ) was calculated, assuming that the pore volume (PV) is 0.6, according to
S = 4PV/PD
cordingly, some considerations on the optimum pore diameter in the hydrodemetalation reaction of Khafji residual oil using the Thiele modulus will be made. (1) Pore Diameter and Activity. The case of DV reaction will be presented. Each variable in eq 1 is estimated as described below. (a) Intrinsic Activity: ki (V). Assuming that the DV reaction proceeds as first order wit1 respect to the vanadium concentration, the apparent rate constant for the catalysts which demonstrate DV percentages of 60-6570 at 400 "C in Figure 2 can be calculated to give about 3 X cm/s. Also the catalyst effectiveness factors (Ef) of these catalysts are estimated to be in the range 0.7-0.2 based upon the results in our previous paper. Accordingly, the rough value of the intrinsic rate constant (hi) in this reaction was estimated to be 4-20 X cm/s. (b) Diffusion Coefficient: De (V). In our previous paper, we estimated the value of De of V compounds in cm2/s a t 400 "C Khafji residual oil to be about 7.8 x in the case of the catalyst having a pore diameter of 175 A. Newson (1970) and Shimura et al. (1982) also obtained a value on the order of cm2/s for De of metal compounds in a residual oil at about 400 "C. This value is of the same order as ours. Consequently, De of V compounds at 400 "C in the case of a pore diameter of 175 is assumed to be in the range 1-10 x cm2/s. Though the value of De varies depending upon the pore diameter, Shimura et al. (1982) reported that the relation between De and pore diameter can be represented by
400
Diameter,
(4)
Now, from values in the above-mentioned range and eq 1,3, and 4, the relation between apparent activity at 400 "C and pore diameter was calculated. The result showed that the pore diameter a t which the highest activity was obtained was in the range 100-200 A regardless of the values of ki and De. An example is shown in Figure 8 (the curve at 400 "C is the one in the case where PV = 0.6, d, = 1.0, ki = 6 X 10-lo,and K = 2 X lo4). It is apparent that the relation between DV and PD in Figure 8 is in comparatively good accordance with the experimental results in Figure 2. From this, it is considered that eq 1 is applicable to the DV reaction of heavy oil and the estimated values of Ki and De are reasonable ones. In Figure 8, there are also shown patterns a t reaction temperatures of 420 and 380 OC which were obtained by assuming that the intrinsic activation energies of the reaction and diffusion are 40 and 5 kcal/mol, respectively. These patterns show a tendency that the optimum pore diameter is shifted to the larger side with the elevation of the reaction temperature, and this is again in accordance with the experimental results shown in Figure 2. In addition, there is seen a tendency in Figure 2 that the shape of the curve becomes broad with the elevation of the reaction temperature. A similar tendency exists in Figure 8. This phenomenon can be interpreted as follows. Since the values of vanadium removal (fv) in Figure 8 are those calculated from the equation In 1/(1- fv) = kiSEft, they depend upon the product of Ef and S. The variation of Ef and S with pore diameter is illustrated in Figure 9. In this figure, two cases of Ef are plotted, one for the lower reaction temperature (380 "C) and the other for the higher temperature (420 "C). As known from the figure, in the case of the lower reaction temperature, Ef
Ind. Eng. Chem. Res., 1701. 26, No. 11, 1987 2249
cn
-200
1: 100
9 0
z
a
5 1
0
50
Nickel 100
200
300
400
500
600
700
Pore Diameter,
Figure 9. Effectiveness factor and surface area.
already begins to be constant in the range where the surface area variation is still remarkable. As a result, the activity rapidly decreases through maximum with an increase in the pore diameter. On the other hand, at the higher reaction temperature, in the range where the variation of the surface area becomes mild, Ef still increases with the pore diameter, to result in a broad change in activity. In summary, for the DV reaction, the experimental results on the relation between the activity and the pore diameter can be satisfactorily explained by the simple equation (eq 1) of Thiele modulus. (2) Selectivity of Reactions. In the above section, it was shown that the selectivity of two kinds of reactions varied with the pore diameter. For example, in the relation between DV and DNi removal (Figure 6), it is recognized that relative reactivity of vanadium removal to nickel removal increases with an increase of pore diameter. In this section, we attempt to analyze this phenomenon by the equation including the Thiele modulus (eq 1). The De value of vanadium compounds is assumed to be equal to that of nickel compounds, and it can be expressed by eq 2. On the other hand, the kiof both reactions may be different; this is, it is considered that the ki of nickel removal is less than that of vanadium removal. Therefore, in this analysis, the values of ki (V) and ki (Ni) at 400 "C are estimated to be 6 X and 3 X cm/s, respectively. From the above-mentioned values and eq 1,the DV and DNi percentages are calculated at various pore diameters. These results are shown in Figure 10 (curve A), in which the value of the pore diameter is marked beside each point. It can be clearly seen from the figure that the curve of selectivity with pore diameter makes a shape. In addition, when the selectivity in the range of large pore diameter (200-700 A) is compared with that of small pore diameter (60-90 A), it is apparent that the former is higher than the latter with respect to the selectivity toward vanadium removal. Curve A in Figure 10 is the one obtained in such a case where the De values for the two kinds of reactions (i.e., DV and DNi reactions) are the same and the kiof one reaction is twice the other. In practical situations, however, the difference of De and ki between two reactions is influenced in a various manner according to the reaction conditions, the type of feeds, and, of course, the kind of reactions, so that the selectivity pattern may be also influenced by such factors. Therefore, we made calculations at various combinations of De and hi for two reactions, and as a result, it was revealed that the activity curve was fundamentally U-shaped in almost all combinations. For instance, in the case where the difference of De and ki is large, the curve shows a large opened U-shape,
100
Removal, %
Figure 10. Theoretical curve of selectivity. (A) kiof reactant A (e.g., and of reactant B (e.g., Ni) is 3 X Proportional V) is 6 X constant ( K ) in eq 2 for De of reactants A and B is equal and its and of value is 2 X lo4. (B) k, of reactant A (e.g., V) is 6 X reactant B (e.g., S) is 3 X Proportional constant (K) for De of reactant A is 2 X lo4 and of reactant B is 6 X lo4.
as shown in Figure 10, curve B. Returning to Figures 6 and 7 , we can now recognize that the U-shaped selectivity pattern shown in Figure 10 is also observed in these figures when we trace each point at each reaction temperature. The selectivity pattern between DV and DNi is a narrow U-shape, whereas that between DV and DS is a largely opened U-shape. The latter may be the result of the relatively large difference in the diffusivity and the reactivity between DS and DV reactions.
Conclusion In this paper, we investigated the influence of the catalyst pore diameter on HDM activity by use of a tricklebed reactor. The results of the experiments and the kinetic analyses can be summarized as follows. (1)The optimum pore diameter of the catalyst for HDM reaction was estimated to be 100-150 A at 400 "C. (2) The optimum pore diameter increases with the elevation of reaction temperature, and the curve showing the change of metals removal with pore diameter becomes broader at higher reaction temperature. These phenomena can be successfully explained by a simple kinetic equation including the Thiele modulus. (3) The selectivity between two reactions was influenced by the pore diameter of the catalyst, and the U-shaped selectivity pattern with the change in pore diameter was observed, which was also satisfactorily interpreted by kinetic analysis by using the Thiele modulus. Nomenclature De = effective diffusion coefficient, cmz/s db = bulk density of catalyst, g/cm3 d = pellet density of catalyst, g/cm3 E% = effectiveness factor, dimensionless f = Thiele modulus, dimensionless f v = fraction of vanadium removed, dimensionless K = proportional constant in eq 2, cm2/s ki = intrinsic rate constant, cm/s MD = molecular diameter, A PD = pore diameter of catalyst, A or cm PS = pellet size of catalyst, mm or mesh P V = pore volume of catalyst, cm3/g R = particle radius of catalyst, cm S = internal surface area of catalyst, cm2/g t = reaction time, s Literature Cited Egi, K.; Koinuma, Y.; Kushiyama, S.; Kobayashi, S.; Aizawa, R.; Inoue, K.; Shimizu, Y. Kougai 1978, 13, 398. Hardin, A. H.; Packwood, R. H.; Ternan, M. Prepr.-Am. Chem. SOC.,Diu. Pet. Chem. 1978, 23, 1450.
2250
Ind. Eng. Chem. Res. 1987, 26, 2250-2258
Inoue, K.; Kobayashi, S.; Kushiyama, S.; Koinuma, Y.; Aizawa, R.; Shimizu, Y.; Egi, K. Prepr.-Jpn. Pet. Inst., 16th Anna. Meet., Sapporo 1973, 32. Kobayashi, S.; Kushiyama, S.; Inoue, K.; Aizawa, R.; Koinuma, Y.; Shimizu, Y.; Egi, K. Znd. Eng. Chem. Res. 1987, preceding paper in this issue. Newson, E. J. Prepr.-Am. Chem. SOC.,Diu. Pet. Chem. 1970,15(4), A141. Plumail, J. C.; Jacquin, Y.; Toulhoat, H. In Proceedings of the Cli-
Pollack, S. S.; Yen, T. F. Prep.-Am. Chem. SOC.,Diu. Pet. Chem. 1969, 14(3), B-118. Riley, K. L. Prepr.-Am. Chem. SOC.,Diu. Pet. Chem. 1978, 23(3), 1104. Shimura, M.; Shiroto, Y.; Takeuchi, C. Am. Chem. SOC.,Diu. Colloid Surf. Chem., Las Vegas Meet., March 1982, 30. Wheeler, A,, In Aduances in Catalysis; Academic: New York, 1951; p 249.
max Fourth International Conference on Chemistry and Uses of Molybdenum, Golden, CO, Aug 1982 Barry, H. F., Mitchell, P. C. H., Eds.; Climax Molybdenum: Ann Arbor, MI, 1982; p 389.
Received for review August 14, 1986 Accepted June 29, 1987
On Multicomponent Adsorption Equilibria of Xylene Mixtures on Zeolites Renato Paludetto, Giuseppe Storti, Giuseppe Gamba, Sergio Carrk, and Massimo Morbidelli* Dipartimento d i Chimica Fisica Applicata, Politecnico d i Milano, 20133 Milano, Italy
Adsorption equilibria of two ternary systems involving m-xylene, p-xylene, and either toluene or isopropylbenzene on zeolite K-Y have been studied. Due to nonideal behavior of the adsorbed phase, m-and p-xylene selectivity is strongly dependent upon composition. In particular, it is found that the addition of a third component can either enhance or depress such selectivity values. Ternary experimental data are well predicted by the developed equilibrium model, whose parameters can be estimated based only on experimental data relative to pure and binary mixtures. Finally, the role of these nonidealities in the equilibrium behavior on the dynamics of adsorption separation columns is discussed. The separation of fraction C8is one of the most classical separation problems in the petrochemical industry. The core of the process is constituted by the final separation of m- and p-xylene isomers, which is most frequently performed by adsorption on zeolites. Such a separation process is based on the principle of displacement chromatography, thereby involving the introduction of an extra component, so-called desorbent, which improves the process efficiency by displacing the adsorbed components along the adsorber. The most widely adopted process operates in the liquid state and involves a simulated countercurrent adsorption unit equipped with a quite efficient rotatory valve (Broughton et al., 1970). Recently, the possibility of operating in the gaseous state has been investigated (Morbidelli et al., 1986a). It was found that, in general, operation in the gaseous state is more efficient than operation in the liquid state, mainly because of the drastic reduction of the noneffective holdup of the separation unit. In this case, a multiport switching unit is more adequate for simulating countercurrent operation. A detailed study of the behavior of once-through adsorbers, packed with zeolite Y exchanged with zeolite K and fed by mixtures of various aromatics, has been previously reported (Morbidelli et al., 1985a; Storti et al., 1985). The aim was to predict the column behavior with a suitable mathematical model, based on simple independent measurements of adsorption equilibria. In particular, the multicomponent Langmuir isotherm was adopted. The aim of this work is twofold. Firstly, we develop an efficient procedure for fully characterizing adsorption equilibria of multicomponent mixtures, in order to accurately predict the adborber behavior. With this respect, it is worth noticing that pure-component equilibrium measurements can be quite difficult in practice, particu0888-5885/87/2626-2250$01.50/0
larly for compounds exhibiting very high affinity for the adsorbent, as in the case of aromatics on zeolite Y. Thus, in order to reach the Henry region of the isotherm, it is necessary either to reach extremely low values of pressure and concentration or to extrapolate the data from rather different temperature values. Secondly, we investigate the accuracy of the multicomponent Langmuir isotherm for systems of the type under examination here. In fact, previous studies on adsorption of chloroaromatics on zeolite X, exchanged with Ca, have indicated a strong influence of selectivity on mixture composition (Morbidelli et al., 198613; Paludetto et al., 1987). This feature cannot be accounted for by the Langmuir model which intrinsically predicts composition-independent selectivity values.
Equilibrium Model The equilibrium model is developed along the lines of the classical thermodynamic approach, i.e., the Ideal Adsorbed Solution theory originally developed by Myers and Prausnitz (1965), introducing suitable modifications to account for nonideal behavior of the adsorbed phase. In the following we will assume that the adsorbent operates at saturation conditions. More precisely, i t is assumed that operating pressure values are so large that all pure-component isotherms have reached their asymptotic region, so that a pressure increase does not produce any further uptake from the gas phase. As mentioned above, this is certainly the case for most aromatic-zeolite systems at atmospheric conditions. But, even more important, this can be safely assumed for any system operating on the principle of displacement chromatography, since the displacement process is really effective when the adsorbent is at saturation. Recalling that most bulk adsorption separation processes are based on this principle, it is apparent that this assumption does not limit the generality of the developed procedure. 0 1987 A m e r i c a n C h e m i c a l Society