MULTIPLICITY OF CONVERSION IN A CASCADE OF IMPERFECTLY STIRRED TANK REACTORS

S. N. LOt and A. CHOLElTE* Department of Chemical Engineering, Universite Laval, Qutbec, Canada
(Received 9 November 1981; accepted 8 July 1982)

Abstract-The multiple steady states of conversion in the model of non-ideal CSTRs in series have been studied for irreversible, homogeneous exothermic reactions of first order. The iuiluences of the parameters p and n and those of scune kinetic parameters on the occurrence of multiple conversions for a given residence time are presented. The existence of steady-state multiplicity of conversion in each CSTR is shown to be independent of n and depend largely on the values of EIRTo (activation energy), b/T (heat of reaction), and p. For certain parametric values, more than three steady states can exist in such a model.
INTRODUCTION

Since Van Heerden [l] published his work in 1953, the study of multiplicity of conversion in reactors has become more and more important to the design and control of chemical reactors. Numerous papers have been published since then on this subject, some of which deal with ideal stirred tank reactors [2-241, controlled cycletank reactor[25,26], and tubular reactors [27-33) in which axial heat and mass transfer was present. Others dealt with tubular reactors with recycle [34,35] and with MT reactor [36]. The particular case of tanksin-series reactor was treated by Gall and Aris [37], Berger and Perlmutter [38], Eguchi and Harada [39], and more recently by other investigators [40,41]. Multiplicity of conversion has also been shown by Lo and Cholette to exist in a non-ideal CSTR 1421, characterized by m (the level of mixing) and (I- n) (the degree of feed channeling), for exothermic reactions. Recently [43], these two authors studied the performance of a model composed of a cascade of p such non-ideal CSTRs , for simple isothermal and exothermic reactions, and showed how it could be optimized as a funcction of m, n and p and some kinetic parameters. As a result, it was interesting to observe that the existence of multiplicity of conversion in each CSTR was independent of n but depended rather on the values of E/RT,, b/To and p_ A more detailed study on the multiple steady states of conversion in such a model was thus undertaken for adiabatic conditions. Presented hereafter will be the results obtained for exothermic reactions of first order, as an example to show the complexity of the problem of multiplicity of conversion, for the parameters involved. It is believed that, for other exothermic or autocatalytic reactions, there should exist the same kind of complexity as that presented in this paper. For reference, the schematic representation of the model studied is shown again in Fig. I. As assumed previously [43-46], the model is composed of p non ideal CSTRs in series, the volume of each tank being V; and
*Author to whom correspondence should be addressed. ?Present address: Universitt de Quebec B Trois-Rivikres. TroisRivikres, P.Q., Canada, G9A SW.

each tank reactor, for any tank in the series, is characterized by the parameters m and n. The definitions of m and n, which can be found elsewhere [4346], are repeated here for convenience. The parameter m, the level of mixing, represents the volume fraction of a tank in which mixing is considered to be perfect, while the parameter n represents the fraction of the feed which enters the zone of perfect mixing. The remaining fraction of the feed, (l-n), is considered to be channeled directly to the tank outlet. As before, the numerical values of the parameters m and n applicable to one tank are also assumed to be applicable to the others in the series.
BASIC DIBIGN EQUATIONS FOR STEADY STATE OPERATION

Referring to the jth tank shown in Fig. 1, under the condition that the volumetric flow rate q is constant, a material balance around this jth tank leads to, for 1st order irreversible, homogeneous exothermic reactions of the type r = KC: I (x,-l - xi) +; K(l - x,) = 6

xj=(l-n)xj-,+nx;

(2)

where x,-, = (C,- C,-,)/C,, x;= (C,- C;)/C, and 6 = mV/q which is the residence time in each tank in the cascade. The heat energy balance for the jth tank is, after substitutions of eon (1) and b = (- AH)CO/gCo into this

hiTikJixJi&
n
L______________,

Fig.

I. Schematic

representation

of the model.

368

S. N. Lo and A. CHOLETTE

heat equation

and simplitication: 7-i = To + bx; (3) as follows: (4)

The rate constant

K in eqn (1) can be written z em.4 ex;,cB+exi_,

K = * e-RT,=

where A = EIRT,,, B = b/T, and D = AB. For exothermic reactions, the value of D in eqn (4) must be positive.
MULTIPLE STEADY STATES OF CONVERSION

To study the multiplicity of conversion in the model for given values of A and B, eqns (1) and (2) together with eqn (4) were solved simultaneously for the outlet conversion x,, as a function of the total residence time &, also by the Newton-Raphson numerical method used in the previous work [43]. For reference as well as for convenience, the curve showing the variation of Co/r* with x, which have already been presented in Fig. 7 of our previuous paper [43], is also shown again in Fig. 2. (a) Multiplicity of conversion when p = 1 (a.1) Criterion for multplle steady strifes.. As seen in Fig. 2, the shape of the plot of Co/r* vs x becomes more and more concave as the values of A and B increase. As the degree of concavity reaches a certain limit, the variation of x with & then yields, for all values of n, curves displaying S shapes such as those illustrated in Fig. 3. The two points L and N in this Fig. 3 represent two stable steady states while the point M the metastable steady state. The criterion whereby the existence of multiple conversions in

Pi.

3. Plot of [X,.~,,lO, vs [e r,nslJo, for first order exothermic reactions, when p = 1,

the model can be ascertained, for a given 0, within a certain time interval, has been obtained by forcing the equation of d&ldx = 0 to have two real roots of x which must lie within the interval from 0 to 1. When p = 1, eqns (1). (2) and (4) were combined to give the following equation, after substitutions of x1 = +,,), xi = x,,,,,/n, W = A/B and Z = l/B: 2 e--nwYl(nz+x(n<l))t&.(1 - x+,,/n) The application of the condition this eqn (5) yields: x&,,,(l+ w)+x,,,,,(22-x,,, Ij = 0. (5)

IOOO~

of d@ddx(,,,,

= 0 for

Vv)n+n2Z2=0

(6)

the roots of which are:

[Xc<l,loi=

n{( w - 22) r [( w - 2.92 - 4(1 + w)z?].~} 2(1+ w) (7) = n[%= &i

i = 1 and 2

I .O&

0.4

0.8

where [x(.,& or [xc,,<& is the value of xc.,,, at point O1 or 0, on any of the curves where n < 1, whereas [x+_~,]~, are the values of xc,_,, at those two points on the curve for n = 1, as shown in Fig. 3. The condition for the above eqn (7) to have real roots is: W>4Z(l+Z)orD>4(1+B).
order

X
Fig. 2. Variation of _Co/r* with conversion x for first reactions, for different values of A and D.

(8)

It can be verified further

from eqn (7) that [x~,,,~,]~, also

Multiplicity of conversion in a cascade of imperfectly stirred tank reactors inequality of D> satisfies 0 C [xcn< ,Joi c 1. The 4(1+ B), which was also obtained by Douglas for n = 1 [4], bears no relationship with n, indicating that the existence of multiplicity in the model is independent of feed channeling. (a.2) Determination oj the region in which a state can be stable or metastable. The residence time [BT~.,~Jo~ corresponding to point Oi in Fig. 3, where i = 1 and 2, can be determined by the substitution of [x~.~,Joi = n[~,,~,,]~~ into eqn (5) for xc,,,): P&A, n) = (22 - w)n - 2nZrA - 2(1 - n)(l - XJXA - W[x* + (I- n)xn] I;;(X& n) = n2z2(1- XA) + ( w - ZZ)nx, + (1 - n)[(W+2Zn) +(I - n)(l -x,)1xX

369

(12)

(13)

For eqn (10) to have real roots, its discriminant must be greater than or equal to zero, which leads to the following necessary condition imposed upon x*: x,~(W-42-zz)/(W+4Z+4) 05x*51. (14)

(9) Obviously , the locus of the point 0, or 02 is a straight line, starting from the origin with a slope of r*{x = ,,_,,loi} = [~~,_,,l~~/le~,,-,,l~~. The area enclosed [XC by the loop 00,h4PQOz0, composed of the curve 0,hfPQ02 and the two loci 00, and OOz, is a region in which a state can be stable or metastable, depending on which curve this state belongs to. Taking the state defined at point P as an example, it would be stable with respect to the curve for n = 0.5 and metastable when point P belongs to the curve for n = 1. In experimental studies of stability in ideal adiabatic CSTRs, the results reported up to date showed that the metastable steady states defined by the curve 01MPQ02 were obtained indirectly either by feeding and dumping [253 or by temperature perturbation [47]. The situation that a state can be either stable or metastable, as discussed in the preceding paragraph, would make it possible to determine directly by experiments the conversions on the portion OIMPQ02 of the curve for n = 1. An experimental approach is suggested which consists in operating the reactor by varying n with 0, in the interval [BT(~=I)]~ i 0~ < [BT(~=,Jo~. In that time interval, a point such as P or Q is then determined with the conversion obtained at a pair of fixed values of 6, and n, where n < 1. Making use tof the method, a study on metastable steady states in a CSTR was carried out and the results obtained will be reported in a future publication. (b) Multiplicity of conversion when p > 1 (b.1) Criterion for multiple steady states. When p > 1, the existence of multiplicity of conversion in the pth tank depends on the conversion x,_, in the feed to this tank; the value of xp_, can be considered as a parameter in the expression for & as a function of x,,. In so doing, the criterion for the occurrence of multiple conversion at a given f& can be determined also by the preceding method used for p = 1. Letting x,_, = x~, then d&fdx,, =0 obtained from eqns (1) and (2), with the K given by eqn (4). for j = p yields the following:

Since W and Z are positive and .rA satisfies 0 5 xA 5 1, then the condition ( W - 42 - 422) 2 0 must hold. This again leads to the same inequality of D >4(1+ B) obtained previously for a single CSTR, indicating that eqn (8) applies to all values of p. Since Wr4Z+ 42, 2Z- W must be negative. So, in eqn (lo), one has F&f,, a) t0 and F,(x,+ n) 10. implying that the two roots of the equation are real and positive. Furthermore, it can be proved that these two roots satisfy 0 I x, 5 1, for 05 xA I 1. The value of &- corresponding to each of these two roots was obtained by solving eqns (l), (2) and (4) with the method of Regula Falsi. The points as 0t4 and P,, shown in Fig. 4 were plotted with the results thus obtained. Under the condition of D > 4(1+ B). the existence of more than three conversions for a given 8~ is possible in certain cases. An example is shown in Fig. 4 for R = 1 when D = 9 and A = 30 and zeTA = 0.1. As observed, the stable states corresponding to the portion 00, on the

0.6 XP
0.4

where F,(xa)= w+1-XA (11)

Fig. 4. Effect of p on the variation of the outlet conversion x with the residence time &, for tist order reactions when n =l.

370

S. N.

Lo and A. Cno~erre For a particular value of n, the value of pw should be a function of A and D. As an example, also for n = 1, the variation of pM with these two kinetic parameters is presented in Table 1. From this table p.,, is seen to increase non-proportionally with D or with B for a constant A. Because of the drastic increase in pM with a small change in B, it can be concluded that the number of CSTRs so required in the series as to approach an ideal tubular reactor is quite different, when compared with the results obtained previously for isothermal reactions [43]. (b.4) Influence of n on mukipicity oj conversion of constant p. It has been proved previously from eqn (10) that the existence of multiple conversions in the model is only a function of kinetic parameters: A and D. However, results obtained from the study for a set of given kinetic parametric values revealed that the value of pm would decrease with n and that the shape of the curve for x, as a function of 0~ can be modified considerably by the presence of feed channeling. Such an effect is exampIified in Table 2 for A = 30 and D = 6, which presents the various values of pM obtained for several values of n. This table shows that pM increases with (1 - n) but to a much lesser extent than that shown in Table 1 for the influence of kinetic parameters on pnr. Thus, compared to A and D, channeling of the feed has an extremely small effect on pM, the value at which the existence of multiple conversions is no longer possible and beyond which only one conversion will be obtained at any given residence time &. An example of the influence of feed channeling on the performance profile is shown in Fis. 5 and 6, also using A = 30 and D = 6. In general, for any value of n, the number of curve segments, such as OZPZ and Q& in Figs. 4, 5 and 6, is equal to p when p < pM. When n = 1, the segment QiRi, where i = 2, . . . . p. is situated around the neighbourhood of x, = 1, as shown in Figs 4 and 5. However, it is not

curve of p = 1 yield the states lying on the part OOzPzQz of the curve for p = 2; while those corresponding to OIPIQl yield such states of Q2R2S2. Except for the points on the loop QIOz .. * OpM .. . P,P,O,, other values on the curves shown in Fig. 4 and in other figures presented later were the numerical solutions of eqns (1). (2) and (4). by the Newton-Raphson method. (b.2) Determination of regions in which a steady state is metasfable. The residence time &oi and 0,, corresponding respectively to the points 0, and Pi, where i= I, 2,. _ . . _p. in Fig. 4, can be determined by means of eqns (l), (2), (4) and (10). For a particular value of p. as when p = 3, any state on the curve O,P, is metastable. The point OPM is the point at which one stable state becomes possible under the conditions D > 4(1+ I?) and obtained when the value of xA in eqn (14) is at its maximum, I.e.: (xJmax = XA = (W-422)/(W+42+4) (15)

in that case eqn (10) has two equal roots. Fig. 4 shows that the major region where all states are metastable is the loop 0, Oz. . Op, . . PzP, 0,. In other cases, one of which will be shown later, regions of this kind, more than one in number, can exist. At this moment no analytical solution concerning these regions has yet been arrived at. (b.3) Influence of p on mukiplicity of conversion al constant n. As seen in Fig. 4, the portion OiPi for which any state is metastable is a function of p when p is smaller than a certain value, say p,,,. When p = pM, the portion OiPi converges to point O,,, so that there will then be only one conversion corresponding to a given 0,. In other words, there will be no possibility for the model under study to have multiple conversions at any given &when p > pM.

Table 1. Effect of kinetic parameters on the value of pM,for first order exothermic reactions when n = I Kinefic pB=-feTB
PM D 6 9 12 15 A 30 30 30 30 B 0.2 0.3 0.4 0.5 3.3 19.6 81.5 275.4

Table 2. Effect of feed-channeling

on the value of p M. for first order exothermic reactions when

D =

6 and B = 0.2

Multiplicity of conversion in a cascade of imperfectly stirred tank reactors

371

0.8

the case when n < 1; the difference between the two cases can be seen by comparing Figs. 5 and 6. It can be concluded that in the presence of feed-channeling, it would be impossible to attain conversions close to 1 with the model of a cascade of CSTRs. The use of such a model might also introduce additional control problems due to the increase in the value of pM.
Acknowledgement-The authors express their gratitude to the Natural Sciences and Engineering Research Council of Canada for financial support of this research.
NOTATION

0.6
XP

0.4

Fig. 5. Effect of feed-channeling on the variation of the outlet conversion x0 with the residence time &, for tirst order exothermic reactions when p > 1 and n = 1.

0.8 t 0.6

n =0.5 D= 6

A=30

Fig. 6. Effect of feed-channeling on the variation of the outlet conversion x, with the residence time BT,for first order exothermic reactions when p > 1 and n = 0.5.

dimensionless activation energy ( - AH)C&iC~), maximum temperature rise for complete conversion of reactant, K B b/To, dimensionless heat of reaction reactant concentration at the outlet of the G jth tank, gram-mole/l. reactant concentration in the perfect mixing c; zone of the jth tank, gram-mole/l. initial reactant concentration in the feed, gram-mole/l. average specific heat of a reaction mixture, call&ram K) .D Eb/RT& dimensionless E activation energy, cal/(gram-mole) function, see eqn (11) F,(x,) Fzh, n) function, see eqn (12) F&A, n) function, see eqn (13) heat of reaction, caNgram-mole) -(AN) index r: rate of reaction, set- m level of mixing, dimensionless n fraction of the feed entering into a tank reactor, dimensionless P number of tanks in series PM the maximal value of p at which the existence of multiplicity of conversion begins to disappear volumetric flow rate, l./sec. r(f rate of reaction, gram-mole/l. sec.) R gas constant, (gram-mole)(K)/cal TO initial temperature of the feed, K temperature of the feed to the (j + 1)th tank, T K T: temperature in the perfect mixing zone of the jth tank, K V volume of each tank in the tanks-in-series, 1. VT total volume of a tanks-in-series reactor, 1. X conversion variable conversion parameter X.4 maximal value of X~ (Xdmax outlet conversion of reactant when p = 1 X(.~l, and n 5 1 conversion of reactant at the outlet of the X, jth tank X; conversion of reactant in the perfect mixing zone of the jth tank conversion of reactant at the outlet of the XD pth tank
E/RT,,

A b

S. N. Lo and A. CHOLE~~E

or

value of x,, _, , at the point 0, when n I 1 A/B frequency factor, I&ram-mole set) l/B residence time in each tank of a tanks-inseries reactor, set m VJq or rnpV/q, total residence time in a tanks-in-series reactor, set value of 0, at the point 0, when n 5 1, set density of a reactant mixture, gram/cm
REFERENCES

]lP] Cluanu H. C. and Calo J. M., Chem. Engng Sci 1979 34 285. tw Lin K. F., Chem. Engng Sci. 1980 35 1537. WI Huang D. T.-J. and Varma A., Chem. Engng Sci. 1980 35 1806. El Leib T. M. and Luss D., Chem. Engng Sci 1981 26 210. WI Lin K. F., Chem. Engng Sci. 1981 26 1447. v41 Kumar V. R. and Kulkarni B. D., Chem. Enang - - Sci 1982 37 476. ~251 Ausikaitis J. and Engel A. J., A.I.Ch.E.J. 197420 256. WI Lin K. F. and Wu L. L., Chem. Engng Sci 1981 36 435. P71 Raymond L. R. and Amundson N. R., Can. J. Chem. Engng
1964 42 173.

Van Heerden C., Ind. Engng Ckem. 1953 45 1242. Bilous 0. and Amundson N. R., A.I.Ch.E.J. 1955 1 513. Aris R. and Amundson N. R., Chem. Engng Sci. 1958 7 121. Douglas I. M., Chem. Engng Symp. Ser. 1964 60 1. Kermode R. I. and Steven W., Can. J. Chem. Engng 1965 43 68. [6] Regenass W. and Aris R., Chem. Engng Sci 1965 20 60. [7] Aris R., Chem. Engng Sci. 1%9 24 149. [El Furusawa T., Nishimnra H. and Miyauchi T., J. Chem. Engng Japon 1969 2 95. [9] Hlavifek V., Kubicek M. and Visnak K., Chem. Engng Sci. 1972 27 719. Cl01 Varma A. and Amundson N. R., Can. I. Ckem. Engna 1973 51206. fll] Uppal A., Ray W. H. and Poore A. B., Chem. Engng Sci. 1974 29 967. El21 Ding J. S. Y., Shama S. and Luss D., Ind Engng Chem. Fundis 1974 13 76. [13] Luss D. and Chen G. T., Chem. Engng Sci. 1975 29 1483. 1141 Cohen D. S. and Keener J. P., Chem. Engng Sci. 1976 31 115. [lS] Uppal A., Ray W. H. and Poore A. B., Chem. Engng Sci. 1976 31 205. [16] Bosch B. V. D. and Luss D.. Chem. Engng Sci. 1977 32 203. 1171 _ _ Kauschus W.. Demont J. and Hartmann K.. Chem. Enana _SC;. 1978 33 1283. [I81 Tsotsis T. T. and Schmitz R. A., Chart. Engng Sci. 1979 34 135.

[ll [2] [3] [4] ]Sl

PSI Amuodson N. R., Can. J. Chem. Engng 1965 43 49. ~291 Luss D. and Amundson N. R., Chem. Engng SCL 1%7 22 253. [301 Amundson N. R. and Luss D., Can. J. Chem. Engng 1968 46 424. L311Gupalo Y. P. and Ryazantsey Y. S., Chem. Engng Sci. 1%9 t4 1441. Kapila A. K. and Poore A. B., Chem. Engng Sci 1982 37 57. :;:j Jensen K. F. and Ray W. H., Ckem. Engng Sci. 1982 37 199. [341 Reilly M. J. and Schnitz R. A., A.LCk.E.I. 1966 12 153. [351 Pareja G. P. and Reilly M. I., Ind. Engng Chem. Eundls 1969 8 443. Lo S. N. and Cholette A., Can. .I. Chetn. Engng 1972 50 71. t:: Gall C. E. and Aris R., Can. L Chem. Engng 1%5 42 16. [381 Berger I. S. and Perlmutter D. D., Ind. Engng Ckem. Fund/s 1965 4 90. [391 Eguchi W. and Harada M.. Kagaku Kogaku 1966 4(l) 54 (English version]. V. and Sinkule J., 1401 Kubicek hf., Hofmann H., Hlav&k Chem. Engng Sci 1980 35 987. 1411 Svoronos S.. Aris R. and SteDhanODoUlOS G., Ckem. Engng Sci. 1982 31 357. Lo S. N. and Cholette A., Can. .I Chem. Engng 1972 50 66. 1:; Lo S. N. and Cholette A., Chem. Engng Sci 1978 33 1031. WI Chiang D. and Cholette A., Can. I. Chem. Engng 1970 48 286. [451 Chiana D. and Cholette A.. Can. 1 Ckem. Enana - _ 1971 49
484. -

[461 Vo V. B. and Cholette A., Can. J. Ckem. Engng 1973 51 149. [471 Vejtasa S. A. and Schmitz R. A., A.1Ck.E.J. 1970 16 410.