MIXING EFFECT ON HOMOGENEOUS REACTION

IN BATCH STIRRED TANK REACTOR

Yasuhiro MURAKAMI,Masaharu TAKAO,
Osamu NOMOTO and Kazuhiro NAKAYAMA
Department of Chemical Engineering,
Kyushu University, Fukuoka 812

To represent the macro-mixing rate process in a batch reactor, a combined model of the ex-
tremes for mixing is proposed by extending the treatment of Miyawaki's model. This model is
characterized by the segregation function and has the simplest mixing structure : that composed of
both complete segregation and perfect mixing. By equating the intensity of segregation derived
for the proposed model to that for the batch mass exchange-type model, the model parameter can
be related as a function of macro-mixing time with the mixing parameter of the mass exchange-
type model. This treatment of the model is experimentally confirmed by using three kinds of
reactions having different rate constants. The mixing parameter of this model calculated from
the experimental data of irreversible second-order reactions with different rate constants in a wide
range of operational conditions is more successfully correlated to the Reynolds number of the
reaction system. Further, the overall reaction performance is found to be characterized by a
dimensionless number which is given from both the reaction rate and the mixing parameter.

balance have been made to develop a simple and

Introduction
The mixing characteristic in a continuous stirred
tank reactor has been primarily represented by a
residence time distribution. However, the residence
time distribution alone cannot describe either the
mixing rate process of two reactive fluids evaluated
by the intensity of segregation Is2) and the mixing
rate process of fluids having different ages related to
the degree of segregation J2jl3). To avoid this limi-
tation of residence time distribution, many authors
have devoted their efforts to the study of so-called
"micromixing". In the previous papers8'10), the
authors confirmed that Is is just equal / calculated
for the two-environment model for micromixing7},
and the mass exchange-type model4] and random
coalescence redispersion model1>3). The mixing rate
process evaluated in terms of Is is the most basic
mixing phenomenon which is indirectly measured
from the degree of completion of chemical reaction.
For a batch stirred tank reactor where Is can be
most sensitively measured, only a few studies have
been performed. Miyawaki5) proposed a combined
model of the extremes for mixing by defining the degree
of micromixing aMand analyzed experimental data.
His model, however, was constructed at the local
liquid lump of micro-scale. Recently, a few attempts
applying the concept of macroscopic population
Received July 2, 1980. Correspondence concerning this article should be
addressed to M. Takao. O. Nomoto is now with Kyoto Ceramics, Co., Ltd.,
Sendai 895-02 and K. Nakayama is now with Kurale Co., Ltd.,Okayama 712.
196
practical model for batch mixing accompanied by
a second-order chemical reaction. In other papers9>n),
the mass exchange-type model in a continuous stirred
tank reactor4} has been applied to the overall batch
mixing rate process to interpret the reaction data
under incomplete mixing effect. The experimental
conditions in these works5>9) were limited to a narrow
range of Reynolds numbers due to the use of small
tank vessels.
In this paper, to describe the overall mixing rate
process of miscible reactive fluids in macro-scale, a
combined model of the extremes for mixing wherein
is introduced the segregation function s(t) is developed
by extending Miyawaki's study5\ The correlation of
intensity of segregation in the combined model of the
extremes proposed here and the batch mass exchange-
type model proposed previously is discussed. The
validity of the proposed model is examined by carrying
out three kinds of reactions having different rate
constants. Fromthe experimental result of irrever-
sible second-order reactions, a correlation between
the model parameter and Reynolds numbercovered in
a wide range of operational conditions is determined.
The dimensionless number q=krCb0/km9) proposed as
a correlating factor of reaction data is experimentally
investigated.
1. Combined Model of the Extremes for Mixing
Nishimura7) formulated a two-environment model
for micromixing by defining the segregation function
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
s(a,X). Let's consider a batch mixing rate process of
reactive fluids A and B in the present work. Here
s{t) takes the place ofs(a,X). The content in a batch
reactor is assumed to be composed of three parts, i.e.,
twomixed,
parts asare schematically
completely segregated and one is ideally
shown in Fig. 1 (a). The
mass of fluid transferred from each completely segre-
gated (C.S.) region is thoroughly mixed in the maxi-
mummixedness (M.M.) region.
Miyawaki5) has proposed a model of the same type
and defined the degree ofmicromixing aM. The treat- Fig. 1 Schematic illustration of mixing models
ment of the model, however, is limited to the liquid
lump at a local point. When it is extended to the secutive-type modeland the parallel-type model, and
whole reactor content, aMbecomes exactly equal to the exponential decay-type model, respectively, have
1-*(*).
been presented by Weinstein12), and Ng6) in a con-
When reactive fluids obey an irreversible second- tinuous stirred tank reactor. Amongthese models,
order reaction of A+B-+C,the reaction occurs only the exponential decay-type model denoted by s(t)=
in M.M. region. Take into account the increase in exp(-pi) is considered to be most reasonable to
moles of species due to the inflow from C.S. region describe the batch mixing rate process of miscible
and the decrease by the chemical reaction. The rate fluids because of the similar performance of Is and
of change in moles of A and B species is expressed as reaction to the batch mass exchange-type model.
follows. Equating the *Jls derived for the exponential decay-
dNaJdt+krNamNbJVm=(VaCa0+ Vb -0)f (1) type model to that for the batch mass exchange-type
model9\ mixing parameters for these models are
dNoJdt+krNbmNam/Vá"=(Va 'O+ VbCb0)f (2)
correlated with one another and with the macro-
where Nam=CamVm, Vm=V[l-s(t)]9 V=Va+Vb and
f=-ds(t)/dt, and the initial conditions at £=0 are mixing time9) as follows.
given as Nam=Nbm=0and Vm=0. Then mean con- rnr=p/2=km = [ln(l /3)]/t^ (7)
centrations are calculated as where 8=Vis means the degree of unmixedness at
Ca=[a - CaO+b -O]s(t)+NaJV (3) the measured mixing time and mr implies a generalized
Cb=[a -O+b ' Cbo]s(t)+NbJV (4) mixing rate coefficient. From this relation, the dimen-
sionless numbercorrespinding to the ratio of reaction
where a=VJV, b=Vb/Vand a+b=l. rate to the rate of mixing can be determined as
The intensity of segregation defined by Danckwerts2) q=krC0/mr==krC0/(p/2)=krC0/km for an irreversible
can be derived as follows. second-order reaction.
Is^(a~ayi(a ' b)=[{a(l -af+b(O~ay}s(t) 2. Experiments
+(a-~ay{l -s(t)}]/(a - b)=s(t)
[= tt fts(a,k)g(a,X)dadXW] (5) To examine the validity of the theoretical treatment
of the proposed model, three classes of reaction having
It is equal to the volume fraction of the C.S. part in different rate constants were carried out.
the reactor. This result coincides with the conclusion The chemical formula for an instantaneous reaction
obtained in a flow system10}. Thus the combined is
model of the extremes characterized by s(t) can provide CH3COOH(A) +NH4OH(B) ->
a clearer relationship between batch and flow systems. CH3COONH4+H2O (8)
The normalized concentration difference from the
mean can also be derived as dC/JC0=s(t). Through This instantaneous reaction was employedas an in-
these analyses of mixing characteristic for the model direct method to measure */Is and the model parame-
and the relation of unreacted fraction of instantaneous ter could be calculated from Eq. (7).
reaction given by Miyawaki5), the model is found to A rapid reaction is
satisfy the relation as CH2ClCOOC2H5(A) + NaOH(B) -^>
Is=l-F=AC/dC0=s(t) [=e-** 6>] (6)
CH3ClCOONa+ C2H5OH (9)
while \/Is= l -F=dC/dC0=exp (-kmt) was satisfied The reaction rate constant was kr=23.0 /-mol"1 ^"1
in the batch mass exchange type model9}, which is at 20°C. It was performed to determine the mixing
schematically shown in Fig. 1 (b). The three micro- parameter from the trial-and-error calculation of
mixing models of two-environment type, i.e., the con- Eqs. (l)-(4). In the previous study9}, both reactions
VOL 14 NO. 3 1981 197
 Fig. 2 Schematic drawing of stirred tank reactor and Fig. 4 Analysis of rapid reaction data under
apparatus for electrical conductivity measurement incomplete mixing effect by the exponential
decay-type model

Fig. 3 Mixing rate process measured by in-

stantaneous reaction in batch stirred tank reactor Fig. 5 Alkaline hydrolysis reaction of ethyl
acetate
were undertaken only in water solution. Here they
were carried out also in methyl cellulose solution to reactive solution was determined by a back-titration
obtain experimental data in a low Reynolds number method or by the electrical conductivity method. As
range. shown in Fig. 2, two sets of platinum plane electrodes
A moderately slow reaction is were placed in parallel at upper and lower locations
CH3COOC2H5(A) +NaOH(B) -^-> in the vessel. The meanconcentration thus could be
more accurately measured than by one set in the upper
CH3COONa+ C2H5OH (10) position9}.
This hydrolysis reaction has been frequently used to
examine the micromixingeffect on the outlet conver- 3. Experimental Results and Discussion
sion in a continuous stirred tank reactor. However, 3. 1 Determination of the model parameter
the reaction rate constant of fcr=0.073 /-mol"1 ^"1 at 1) Instantaneous reaction The data denoted by the
20°C was about one-320th of that for the rapid reac- solid squares and triangles in Fig. 3 were obtained in
tion of Eq. (9). the large vessel and those denoted by the other symbols
Figure 2 shows a schematic drawing of the stirred in the small one. The data in semi-log plot lie well
tank reactor used. Here a 50-/ large vessel was used on a straight line, though a slightly periodic fluctuation
to examine scale-up effect as well as a 1-/ small vessel. is observed due to circulation. This figure clearly
Through a mechanical device9} 100 (or 4300) ml of indicates that the mixing rate process can be approxi-
A reactive solution was quickly charged into 900 mated by a lst-order rate process. The mixing
(or 1000 or 38,700)ml of B. The concentration of parameter is determined by its slope and/or from Eq.(7)
198 JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
 Fig. 6 Correlation of mixing parameter
determined as a function of Reynolds
number
at 8=</ls(=l-F)=0.02.
2) Rapid reaction Figure 4 shows a 2nd-order reac-
tion plot of the experimental data obtained in about
40 cP of solution mixed with methyl cellulose. Each
solid line is computed by the exponential decay type
(E.D.T.) model so as to fit the data. Insofar as one
treats an overall batch mixing rate process accom-
panied by reaction, it is also as useful for practical
engineering use as is the batch mass exchange-type
model9K Further, the numerical calculation
proposed model is simpler.
3) Moderately slow reaction The open symbols in
Fig. 5 show the data in 100m/ of A solution-1000 m/
of B solution system. The change in impeller speeds
from 0 to 5.0 s"1 has no effect on the course ofreac-
tion or reaction rate constant. Therefore the model
parameter cannot be determined.
3. 2 Correlation of experimental data
1) Correlation of the mixing parameter determined
The data at the upper part ofFig. 6 are those measured
by instantaneous reaction. As denoted by the open
stars, the specific mixing rate per unit revolution
mr/n=(p/2)/n=km/n in the large vessel matches that
in the small one. The data determined by rapid
reaction are plotted at the lower part in the figure.
Not only the trends of (p/2)/n versus Re but also the
absolute values obtained for both reactions are agree-
able for the exponential decay-type model. Speaking
of the mass exchange-type (M.E.T.) model, the cor-
related line denoted by dotted line is about 30%
higher than given by the exponential decay-type model.
The mixing parameter of the exponential decay-type
model is seen to be more successfully correlated to
a wide range of Reynolds number than the mass
exchange-type model. Consequently, the proposed
model, though the calculated performance of \ -F=Is
VOL. 14 NO. 3 1981
of the
Fig. 7 Correlation of conversion data as a function
of dimensionless numberq
differs from \-F=^/ls of the mass exchange-type
model, can evaluate reaction performance by use of
the relation of Eq. (7).
2) Correlation of conversion as a function of dimen-
sionless number q The conversion data obtained by
using the rapid reaction of Eq. (9) at various levels of
impeller speed and viscosity at 20 and 35°C are plotted
against the dimensionless number q in Fig. 7. The
open stars in the figure mean the data obtained by
using the large vessel. The mixing rate coefficients
necessary for the calculation of q are estimated from
Eq. (7) with the mixing time observed in the case of
instantaneous reaction. The observed data are located
more nearly along the dotted line for the exponential
decay-type model than the solid line for the batch mass
exchange-type model.
All the q values calculated for the reaction of
Eq. (10) are less than 0.1 for Re>W. For example,
(see Fig. 5) the value of q is calculated as (0.072)
(0.035)/(0.1)=0.025 even at the minimum impeller
speed of w=0.3 s"1, where mr=p/2=km=0.l s"1 is
199
 estimated from Fig. 6. It is found also from this
kr = ind-order reaction rate constant [/-mol us[s"1]x]
calculation that the condition of ideal mixing is mr
N
= generalized mixing rate coefficient [mol]
= mole of reactive species [s-1]
achieved in the experiment using the hydrolysis reac- n = impeller speed
tion of ethyl acetate of Eq. (10). P = micromixing parameter for exponential [s"1]
decay-type model6)
As a result, the dimensionless number q is concluded dimensionless number ( =&rC0/mr) [-]
to be useful in correlating the reaction data of irrever- Reynolds number (=pnd2jv) [-]
sible second-order reactions. segregation function in batch system
time [-]
mixing time at the unmixedness 5 [s]
Conclusion volume of total reactor content [s]
width of baffle
The performance of the overall batch mixing rate dimensionless concentration ( = CIC0) [cm3]
[cm]
process of reactive fluids was studied from both theory [-]
and experiment. The following results were obtained
for irreversible second-order reactions. age of fluid at a point [s]
1) A combined model of the extremes for mixing [-]
degree of micromixing5) [-]

A
wherein was introduced the segregation function s(t) i-th minus the mean
degree of unmixedness (= Vis) [-]
was proposed to describe the macromixingrate
in the batch reactor. The intensity
process
of segregation life expectation [s]
[cP]
viscosity of reactive solution
derived for the model was equal to the segregation density of reactive solution
[g em""3]

function s(t).
<Subscripts and superscripts)
2) By equating the intensity of segregation for the a = reactive fluid A or species A
proposed model to that for the batch mass exchange- b = reactive fluid B or species B
type model, the mixing parameter of the proposed m = maximum
mixedness
p
model could be related as a function of the macro- = poor
r
mixing time with that of the batch mass exchange-type = rich
s
model. Applying this correlation, the reaction per- = complete segregation
0
= time zero or initial
formance could be similarly predicted by the proposed = space averaged (=mean)
model. Further, the treatment of this model was
simpler than that for the mass exchange-type model. Literature Cited
1) Curl, R. L.: AIChEJ., 9, 175 (1963).
3) The model parameter calculated from the experi-
mental data of irreversible second-order reactions with 2) Danckwerts, P. V.: Chem. Eng. Sci., 8, 93 (1958).
different rate constants in a wide range of operational 3) Evangelista, J. J., S. Katz and R. Shinnar: AIChE J., 15,
843 (1969).
conditions was more successfully correlated to
Reynolds number of the reaction system. Further, 4) Harada, M., K. Arima, W. Eguchi and S. Nagata: Memoirs
the overall reaction performance was found to be ofFaculty ofEng., Kyoto Univ., 24, 431 (1962).
5) Miyawaki, O., H. Tsujikawa and Y. Uraguchi: /. Chem.
Eng. Japan, 8, 63 (1975).
characterized by the dimensionless number q which 6) Ng, D.Y.C. and D.W.T.Rippin: Third European
was given from both the reaction rate and the mixing
parameter. Symposium on Chemical Reaction Engineering, Amster-
dam, p. 61, Pergamon Press, Oxford, 1965.
Nomenclature 7) Nishimura, Y. and M. Matsubara: Chem. Eng. Sci., 25,
1785 (1970).
a
volume fraction of fluid A(= CaICao) 8) Takao, M., O.Nomoto, Y.Murakami and Y.Sato:
b
b'
volume fraction of fluid B(= Cb,ICbQ) /. Chem. Eng. Japan, 12, 408 (1979).
C
width of impeller blade 9) Takao, M., T. Yamato, Y. Murakami and Y. Sato: ibid.,
c
concentration of solution ll, 481 (1978).
D
concentration of functional group 10) Takao, M. and Y. Murakami: ibid., 9, 336 (1976).
d
inner diameter of vessel ll) Takao, M., Y. Murakami and H.Inoue: ibid., ll, 496
F
diameter of impeller (1978).
reacted fraction for instantaneous reaction 12) Weinstein, H.and R.J.Adler: Chem. Eng. Sci., 22, 65
in stoichiometry (= 1 - CJCaQ) (1967).
g(a, A) = simultaneous frequency function of age and 13) Zwietering, Th. N.: ibid., ll, 1 (1959).
[s-2]
life expectation
H
liquid height in vessel [cm]
Hc
concentration ratio of A to J5(=Ca0/C&0) [-] (Presented at the 40th Annual Meeting at Nagoya, Apr. 2
Is [-] and at the 9th Autumn Meeting at Fukuoka, Oct. 19, 1975;
J
intensity of segregation at the 41st Annual Meeting at Sendai, Apr. 4, 1976; and at the
degree of segregation [-]
Kr [-] 12th AutumnMeeting of The Soc. of Chem. Engrs., Japan
km
dimensionless reaction time( =krCQt ) [s-1] at Okayama, Oct. 10, 1978.)
mass exchange coefficient

200 JOURNAL OF CHEMICAL ENGINEERING OF JAPAN