CkmicaI Engineering Science, Vol. 42, No. 3, pp. 575-580. 1987. 000%2505’/87 $3.00 + 0.

00
Printed in Great Britain. Pergamon Journals Ltd.

DETERMINATION OF THE MICROMIXING LEVEL IN A CSTR:

TRANSIENT STEP RESPONSE OF REACTIVE TRACERS

J. BALDYGA’and SOHRAB ROHANIZ

Department of Chemical Engineering, University of Saskatchewan, Saskatoon, S7N OWO, Canada

(Received 16 January 1986; in revised form 21 July 1986; accepted for publication 18 August 1986)

Abstract-An experimental technique for the determination of the level of micromixing in a CSTR is
proposed. The technique involves the generation and the analysis of the transient step response of reactive
tracers with non-linear kinetics. Two existing micromixing models, a two-environment model based on the
spectral interpretation of turbulent mixing and Toor’s hypothesis, are used to develop the predictive
equations for the determination of the limits and the level of micromixing. The effects of the initial reactants
concentration ratio, the intensity of mixing and the Damkohler number on the level of micromixing,
predicted by the two models, are then compared.

INTRODUCTION experiments suffer from a changeable reacting mixture

The determination of the level of micromixing in a flow volume which complicates modelling and the analysis
reactor requires reactive tracers with non-linear kin- of the experimental data.
etics. The experimental methods for the evaluation of Here we suggest a new method for evaluation of the
the micromixing theories have recently been reviewed level of micromixing which involves the step response
by Baldyga and Bourne (1985). Among the existing analysis of unpremixed reactive tracers in a CSTR.
experimental techniques for the determination of the This technique does not suffer from the disadvantages
micromixing level, the two of prime importance are: associated with the other experimental methods. First,
(a) Premixed reactants in a CUR. The step re- the reactants are unpremixed, so instantaneous and
sponse of premixed reactants was first introduced by fast reactions can both be used as test reactions.
Nishimura and Matsubara (1970). Later Goto et al. Second, the experiments are performed at constant
(1975) used an almost premixed step response exper- volume by initially filling the reactor with one reactant
iment to determine the parameters of a generalized and subsequently introducing the other reactant. The
three-parameter two-environment model. Spencer et predictive equations for the determination of the
al. (1980) introduced the double-impulse response of micromixing level in a CSTR are obtained by using two
unpremixed reactants and the square-pulse response of existing micromixing models in the literature.
premixed reactants. Spencer and Lunt (1980) later
checked the latter technique experimentally and de- TWO-ENVIRONMENT MODEL
tected a significant departure from maximum mixed- The spectral interpretation of the turbulent mixing
ness. In the above techniques, premixed reactants are in an isotropic homogeneous turbulent field supports
used (except in the double-impulse method, which is the existence of a completely segregated zone and a
experimentally difficult to perform). Therefore slow zone of molecular dissipation of the concentration
reactions have to be used as test reactions which result fluctuations.
in smail micromixing effects. In non-viscous liquids, large eddies consisting of
(b) Unpremixed reactants in a semibatch pure reactants form completely segregated zones.
reactor. For instantaneous and fast single reactions, Pohorecki and Baldyga (1983a) have shown that in
A + B + product, the volume of the completely segre- such liquids and in the presence of single reactions with
gated zone, and its rate of decay which represents the non-linear kinetics the mixing time is determined by
rate of breakage and deformation of large turbulent the inertialsonvective destruction of these eddies, and
eddies of fresh reactants, was measured experimentally the viscous-diffusive process has practically no in-
(Pohorecki and Baldyga, 1983a, b). In the case of the fluence on the overall reactant conversion.
competitive consecutive reactions, A + B + R, R + B Consequently, the application of a two-environment
+ S, the selectivity of the reaction was measured at the model for the determination of micromixing in such
completion of the reaction and was related to the final systems is justifiable. The large blobs of fresh reactants
stage of micromixing; namely, small-scale engulfment, form the completely segregated zone and the rest of the
deformation and molecular diffusion in eddies smaller reacting mixture forms the reaction zone, which is well
than 12il, (Baldyga and Bourne, 1984a, b). Semibatch mixed to the molecular scale. The product distribution
of complex reactions in non-viscous fluids and the
conversion of single reactions in viscous media
are, however, significantly influenced by the
‘Dr. J. Baldyga is with the Instytut Inzynierii Chemicznej
Politechniki Warszawskiej, Poland. viscous-diffusive dissipation of the concentration fluc-
*To whom correspondence should be addressed. tuations, and the two-environment model cannot be

575
576 J. BALDYGA and SOHRAB ROHANI

used for the interpretation of the micromixing For t < t,, where t, is the time of disappearance of
phenomenon in such cases. reactant A, the component balances are:
For the system under investigation, the volume of
V(dC,/dt) = -qC, - VR,; C,(O) = C,, (6)
the completely segregated zone decays with time as a
result of turbulent mixing according to: V(dC,/dt) = qC,,-qC, - VR,; C,(O) = 0. (7)

dV,/dt = - VJt,, (1) With CB = 0 and dC,/dt = 0 (for instantaneous reac-

tions), eq. (7) results in
where
t ms= (,,&-(1/3)Cy)/(21-(5/6))k&2’3~-1’3 (2) R, = &,f V (8)

represents the time of the inertial-convective destruc- which is equal to R, according to the reaction
tion of large eddies (Rosenweig, 1964). For an agitated stoichiometry. Substitution of eq. (8) in eq. (6) and
tank, eq. (2) reduces to (Pohorecki and Baldyga, expressing the resulting equation in dimensionless
1983b): form
?*(e) = (l+&exp(-e)-_ (9)
t ms= (2.07 (z/L)~ (D/L)2)/(c#2’3 N(D/L
Pa(e) = 0. (10)
- 1)“3 (H/L)“J) (3)
For t > t,, reactant B behaves as an inert tracer.
where
F*(e) = 0 (11)
H---w/2, h < H/2
z= (3a)
( h + w/2, h 3 H/2. 5Qe) = i -exp[-(e-e8,)] (12)

The micromixing level in a macroscopically
mixed CSTR in terms oft,, can be defined as t,,/(t,,
+T), which lies between 0 and 1 for small and large
values of tmsr respectively.
(a) Instantaneous reactions
Let us consider an instantaneous
reaction scheme
A + 3 + product
well where or = t,/r = In (1 + l/p) is the dimensionless
time of disappearance of reagent A.
(ii) Partial segregation (PS): The level of micromix-
ing in a real CSTR lies between the MM and the CS
limits. The content of the reactor in this regime,
consists of the completely segregated blobs of fluid B
reaction with floating around a well mixed fluid A environment
(reaction zone). The volume of the completely segre-
gated zone is given by

taking place in a macroscopically well mixed CSTR, dVJdt = q.1 -q(VJV)-(V-,/t,,); V,(O) = 0
initially filled with reactant A. At t = 0, reactant B is (13)
introduced at a flow rate of q (see Fig. 1). where the first and second terms on the right-hand side
of eq. (13) indicate the rate of flow into and out of the
completely segregated zone, respectively, and the third
(i) Limits of micromixing. Complete segregation term represents the decay rate of this zone [eq. (l)].
(CS): Under this regime, there is no mixing at the Equation (13) in terms of dimensionless variables
molecular level and, hence, there is no reaction. The results in
concentrations of the reactants in the effluent stream in
dimensionless form are: V,(e)/V = T/(T+ l)[l -exp (-(T+ 1)0/T)] (14)

FA(0) = exp (-0) where

(4) T = t,,/T (19
i;,(0) = 1 -exp (-0) (5) is a measure of the intensity of mixing.
where pA((8) = eA/CAO, pB(0) = &,/C, and 8 = t/z. The component balances in the reaction zone are:
Maximum mixedness (MM): In this regime, the dV,C,/dt = -qAC, - RAVA (16)
reactants are in direct contact. The concentration of
reactant B therefore remains zero until the reactant A dV,C,/dt = -(dVs/dt)C,--(l - V,/V)C,
is fully diminished. -RBVA (17)
where VA is the volume of the reaction zone given by
v,= V-V, (18)
and
qA= qtl--dv) (1%
--9
is the rate of flow out of the reaction zone.
For t < tc, where t, is the time of disappearance of
TIME (~1 reactant A, eq. (17) results in
Fig. 1. Schematic diagram of the experimental procedure. Ra VA = - (dVJdr)C,. (20)
 Micromixing level in a CSTR 577

Substitution of V, and RAVA from eqs (14) and (20) 1.0

into eq. (16) and expressing the resulting equation in - cs
terms of dimensionless variables give: 0.8- --- PS

[ 1 -exp (-$%)]CY,(@+fil~
0.6-
dY,(B)/d0 = - ^v
8 -
l+Texp(-TH)

Y,(O) = 1. (21)
The reactant concentrations in the effluent stream are:

m) = Y*(Q)Cl - v,(e)/ V] (22)

U
_
1.0
1
2.0
P&J) = v,(e)/ v. (23) 8

For t>tcr there is no A and fluid B undergoes Fig. 3. Influenceof T: instantaneous reactions, fi = 1.0, two-
passive mixing, i.e. environment model.

d%‘a(8)/d0 = 1 - $@) (24)

with
k
%(%) = V,(O,)lV (24a) A+B+ product. The concentration profiles under

as the initial condition. The reactant B concentration
profiles for different values of p and Tare shown in
Figs 2 and 3. These curves can be used for the
evaluation of T and hence determination of the
micromixing level for an existing system and a given j?
value, by experimental measurement of the exit con-
centrations. The reactant A concentration profiles are
not presented in order to save space.
CS remain unchanged. However, the profiles cor-
responding to the MM regime move closer to the CS
regime, as the reaction rate decreases. This reduces the
region of observable micromixing effects. Equations
representing the component balances in the reaction
zone are similar to eqs (16) and (17). However, the
reaction rates are now given by:
R,= R,=kC,C,. (25)

(b) Fast reactions Substitution of eqs (13) (14) and (18) for dV,/dt, V,
The test reaction is chosen as a single fast reaction, and V, into eqs (16) and (17) result in

d&(e)
-= _
[l --exp(
-~Q)ly,W+[l +TeXp(
-~s)]pD.y,(e)y,(e) (26)

de
[l+.exp(-?*)I

[1 -exp( -yS)] IYs(+tl+[t+Texp( -ye)]BoY,(H)Ya(t8 (27)

dY,(@
----= _
de
r1 l+Texp(-To)]

I”

I I with Y+,(O)= 1 and Ya(0) = 0, and Da = kCA,t.

The exit concentrations are given by
0.8 - --- PS
---MM g.(e) = Y,(e)[l - v,(e)/ VI (28)
0.6- %m = y,w f1 - ww vi + v,w y. (29)
Figure 4 shows the effect of Da on the exit concen-
tration of reagent A for fl = 1 and T = 1. Figure 5
indicates the effect of Da on the MM limit of
micromixing (T + 0). Similar curves are available for
reagent B.

MICROMIXING VIA TOOR’S HYPOTHESIS

Fig. 2. Influence of b: instantaneous reactions, T = 1.0, two- Let us now consider the step response technique by
environment model. the hypothesis of Mao and Toor (1971). The com-
578 J. BALDYGA and SOHRABROHANI

for the decay of a scalar contaminant, in accordance

- cs with Torr’s hypothesis.
---PS I’(dc;c’a /dt) = - qc,(C,, - ca) - c;\c;/t,, - qc’_+de;
--MM
c;c’e (0) = 0 (32)
where t msis the mixing time constant given by eq. (2).

(a) Instantaneous reactions

For instantaneous reactions

CA c, = -CIA c’e. (33)

By subtracting eq. (31) from eq. (30) and solving the
0 I.0
e
2.0 resulting equation with eqs (32) and (33), the three
unknowns (CA, c, and COCK) are obtained. Figure 6
Fig. 4. Influence of Da:fast reactions, p = 1.0, T = 1.0, two- shows the exit concentration profiles for p = 1 at
environment model.
different mixing intensities.

(b) Fast reactions

Equations (30)-(32) remain valid, but the reaction
ponent balances for the mean concentrations of re- rate is now finite and is given by
agents A and 3 are:
I’(dC, /dt) = - qc, - i? v; C,(O) = 1 (30) E = k,C,C, = k,[C,CB+ c;c’8]. (34)
The mean exit concentrations and the covariance of
V(dC, /dt) = q(C, - e,) - Ev; C,(O) = 0. (31)
concentration fluctuations in dimensionless form are
obtained from:

dY,(@ldB = -yA(0) -Daj?[YA(0)7B(8) +yX(B)y’,(0)]; YA(0) = 1 (35)

-
dY,(B)/de = 1 -ya(@ -Da[yA(0)yB(O) +y’,+(@yb(@]; -Y&O) = 0 (36)

dy;(Q)Y’a(S)lde = -[l --&,@)]~A(@ -Y;(@Y)B(@/T[(T+ 1)/T]; Y’,(o)Y’,(o) = 0 (37)

where
Y;(e)Yb(8) = C~(8)C’8(e)/CAOCBO-

At the reactor inlet, the temporary initial variance of
fluid B is taken with respect to the existing mean
concentration inside the reactor. Subsequent evalu-
ation of the covariance of the concentration fluctu-
ations is based on Corrsin’s equation (Corrsin, 1957)
Exit concentration profiles at p = 1, T = 1 and for
different Damkohler numbers are plotted in Fig. 7. The
CS profile remains the same as for instantaneous
reactions; however, the MM profiles are closer to the
CS profiles, especially at lower Da (see Fig. 8).

1.0

- cs
Q8- ---PS
--MM

0 1.0 2.0 0 1.0 1

e e

Fig. 5. Influence of Da on the limits of micromixing: T Fig. 6. Influence of T: instantaneous reactions, p = 1.0,
= 0.0001, two-environment model. Toor’s hypothesis.
 Micromixing level in a CSTR
- cs
--- PS
--MM
e reactions using the two-environment
Fig. 7. Influence of Do: fast reactions, T = 1.0, B = 1.O,Toor’s
hypothesis.
0 1.0
0
Fig. 8. Influence of Da on the limits of micromixing:
= 0.0001, Toor’s hypothesis, j3 = 1.0.
DISCUSSION
Determination of the level of micromixing in a
CSTR is facilitated by infinitely rapid test reactions at a
low initial reactant concentration ratio. This is evident
from Figs 2,5 and 8, which indicate a larger separation
of the MM profile from the CS profile and therefore a
broader zone of micromixing effect at low /?and high
Da values.
In Fig. 2, points a, band c correspond to the times of
disappearance of reagent A for the given intensity of
mixing, whereas points a’, b’ and c’ represent the
corresponding times under the MM regime. Let us call
the time differences (a -a’), (b -b’) and (c -c’) the
overtitration time. The overtitration time is exper-
imentally measurable and is related to tms, which is a
measure of the micromixing level (see Appendix). The
overtitration time increases as T increases (for a given
B). For T Q 1, mixing is intense and the overtitration
time is independent of the reactant concentration ratio,
whereas at T = 1, it increases with increasing B.
Measurement of the time of decolorization of the
HCl-NaOH reaction, for example, in the presence of
phenolphthalein, and comparison of this time to the
corresponding time under the MM regime determine
579
the level of micromixing. Suppose a CSTR is initially
filled with base (reagent A) which is coloured with the
indicator. At t = 0, acid (reagent B) is introduced at a
constant rate until the moment of decolorization.
Titration determines the excess acid
A& = C,, V,(f?,)/V = C,T/(T+ 1)
x [l -exp (-(T+ 1)0,/T)] (38)
which results in determination of T and the level of
micromixing.
Higher mixing intensities correspond to a smaller t,,
and hence a smaller T. Figure 3 shows the effect of Ton
the reactant B exit concentration for the instantaneous
model. Similar
results are shown in Fig. 6 for Toor’s hypothesis. The
influence of Da on the reactant A concentration
profiles are shown in Figs 4 and 7 using the two-
environment model and Toor’s hypothesis, respect-
ively. The concentration profiles approach the MM
profiles at low T and high Da values. The results
obtained using the two different micromixing models
do not agree very well, especially close to the region of
disappearance of reagent A. This region is character-
ized by a high reactant stoichiometric ratio.
CONCLUDING REMARKS
The proposed technique offers an easy approach for
the determination of the micromixing level in a CSTR.
The two micromixing models predict the effects of the
2.0 intensity of mixing, the initial reactant concentration
ratio and the reaction rates. The discrepancy between
T the two models is larger close to the region of
disappearance of reagent A, where the instantaneous
stoichiometric ratio is much larger than one. This can
probably be attributed to the shortcomings of Toor’s
hypothesis at a high reactant stoichiometric ratio.
Other micromixing models which include the
viscous-diffusive mixing subrange could be used in
conjunction with the proposed technique. However,
no consideration was given in this study to the diffusive
mixing time, corresponding to the viscous-diffusive
subrange of the spectral distribution of the concen-
tration fluctuations. This simplification has been
shown to be valid for single reactions in non-viscous
media (Pohorecki and Baldyga, 1983a).
Acknowledgement-The authors acknowledge the financial
assistance provided by the National Sciences and Engineering
Research Council of Canada.
NOTATION
C concentration mol me3
concentration fluctuation mol me3
: constant in eq. (2); dimensionless
D’ tank diameter, m
Da Damkohler number = kCAOr
H liquid height in tank, m
h position of impeller from the bottom of the
tank, m
580 J. BALDYGA and SOHRAB ROHANI

k reaction rate constant; mol-’ m3 s-l Baldyga, J. and Bourne J. R., 1985, Principles ofmicromixing.
characteristic wave number for large-scale Encyclopedia of Fluid Mechanics, Vol. 1. Gulf Publishing.
koc
eddies, m - ’ Corrsin, S., 1957, Simple theory of an idealized turbulent
mixer. A.I.Ch.E. .I. 3, 329-330.
AL. impeller diameter, m Goto, H., Goto, S. and Matsubara, M., 1975, Generalized two
N stirrer speed, s a environment model for micromixing in a continuous flow
reactant flow rate; m3 s-t reactor-II. Identification of the model. Chem. Engng Sci.
: reaction rate, mol rnp3 s-l 30, 71-77.
Mao, K. W. and Toor, H. L., 1971, Second-order chemical
T dimensionless mixing time = t,,/r
reactions with turbulent mixing. Ind. Engng Chem. Fundum.
t time, s 10, 162-167.
t Ills mixing time [see eq. (2)], s Nishimura, Y. and Matsubara, M., 1970, Micromixing theory
neutralization time under MM regime, s via the two environment model. Chem. Engng Sci. 25,
t,
1785-1797.
t, neutralization time under PS regime, s
Pohorecki, R. and Baldyga, J., 1983a, New model of micro-
V reactor volume, m3 mixing in chemical reactors. 1. General development and
W height of impeller blade, m application to a tubular reactor. fnd. En&g Chem. Fundam.
YA,B dimensionless concentration, CA/CA,; C,/C, 22, 392-397.
dimensionless concentration fluctuations, Pohorecki, R. and Baldyga, J., 198313, New model of micro-
J&B
mixing in chemical reactors. 2. Application to a stirred tank
CAf C.&O;ci3/CBS, reactor. Ind. Engng Chem. Fundam. 22, 398405.
z defined in eq. (3a) Rosensweig, R. E., 1964, Idealized theory for turbulent mixing
in vessels. A.I.Ch.E. J. 10, 91-97.
Greek letters Spencer, J. L. and Lunt, R. R. 1980, Experimental characteriz-
ation of mixing mechanisms in flow reactors using reactive
B cBO/cAQ
tracers. Ind. Engng Chem. Fundam. 19, 142-148.
e dimensionless time = t/z Spencer, J. L., Lunt. R. R. and Leshaw, S. A., 1980,
E rate of energy dissipation, W kg-’ Identification of micromixing mechanisms in flow reactors:
4 power number transient inputs of reactive tracers. Ind. Engng Chem.
l- Gamma function Fundam. 19, 135-141.

1, Kolmogoroff microscale
T mean residence time, s APPENDIX
Analytical solution of eq. (21) is

Subscripts Y,(e) = (1 +B)(I +T)exp (--@)/cl+ T)

A reactant A x exp (- (I+ T)8/T). (Al)
B reactant B
The dimensionless time of disappearance of reactant A
s segregated
under the PS regime, 0,, can be obtained from eq. (Al) by
letting YA(~) = 0. For large 8, (i.e. when the plateau on the
Superscripts concentration profile is reached), we have:
- time average
8, = In (1 + l/B) + In (1 + T). (A2)
fluctuation
Under the MM regime, the dimensionless time of disap-
exit concentration
pearance of reactnat A is

REFERENCES 8, = In (1 + l//3)_ (A3)

Baldyga, J. and Bourne, J. R., 1984a, A 0uid mechanical The overtitration time (a -a’ in Fig. 2) is obtained from eqs
approach to turbuient mixing and chemical reaction, Part (A2) and (A3):
11. Micromixing in the light of turbulent theory. Chem. At, = rAe, = r(e, -e,) = r In (1 + t&r). (A4)
Engng Commun. 28, 243-258.
Baldyga, J. and Bourne, J. R., 1984b, A fluid mechanical For T + t,,,
approach to turbulent mixing and chemical reaction, Part
III. Computational and experimental results for the new At, z lim T ln (1-t t&r) = t,,. b-4
micromixing model. Chem. Engng Commun. 28, 259-281. r-m