Chemicoi Engineering Science, Vol. 43, No. 7, pp. 1489- 1498, 1988. 0009 2509/88 S3.00+0.

00
Printed in Great Britain. 0 1988 Pergamon Press plc

DYNAMIC MODELLING OF MASS TRANSFER

PHENOMENA WITH CHEMICAL REACTION IN
IMMOBILIZED-ENZYME BIOREACTORS

THANOS D. PAPATHANASIOU,+ NICOLAS KALOGERAKIS and LEO A. BEHIE*

Department of Chemical and Petroleum Engineering, University of Calgary, Calgary, Alberta,
Canada T2N lN4

(First received 10 August 1987; acceptedfor publicnrion 6 January 1988)

Abstract-This paper presents a dynamic model for a fixed or liquid fluid bed immobilized-enzyme
bioreactor, together with a novel method for the solution of the coupled partial differential equations in the
real-time domain. Both the tanks-in-series and the dispersion models are used to describe the non-ideal axial
mixing in the reactor. The solution, in its final form, comes in both cases as a system of simultaneous ordinary
differential equations that can be readily solved using commercially available software packages. Based on
this solution, a complete parametric analysis was performed. The analysis revealed the importance of
intraparticle and external mass transfer resistances, intraparticle chemical reaction and axial dispersion on
the transient behaviour of the reactor. Most important, the analysis revealed ways for parameter estimation
and system identification via simple dynamic experiments. The design and optimization implications are
demonstrated by using the derived solution to simulate the performance of an immobilized-urease fluidized-
bed bioreactor with a recycle loop. Such a configuration is characterized by a time-varying feed concentration
and can be used, as part of an extracorporeal artificial kidney device, for the treatment of uremic patients.

INTRODUCTION since they permit the simultaneous determination of

Classic two-phase fixed- or fluidized-bed bioreactors
with the biocatalyst immobilized inside porous solid
support particles (beads) have become increasingly
important in carrying out continuously many types of
biochemical reactions. Applications of immobilized
enzymes range from the production of syrups from
corn starch to artificial kidney devices, affinity chro-
matbgraphy or analytical biochemistry. Efficient use of
immobilized enzymes in fixed or liquid fluidized beds
requires quantitative knowledge of the dispersion and
mass transfer characteristics of the bioreactor together
with knowledge of the kinetic properties of the
immobilized enzyme. The prediction of these prop-
erties based on theoretical or empirical laws is subject
to large uncertainties and, thus, experimental tech-
niques are often employed for the “on-line” determi-
nation of the parameters of interest in fixed- or
fluidized-bed systems.
For this purpose steady-state or dynamic exper-
imental data can be utilized. At steady state, the
individual rates of the various mass transfer and
consumption mechanisms are identical and, thus,
steady-state experiments can only yield a global rate or
the rate of the controlling step. Recently, Davis and
Watson (1985) described a method for measuring
diffusivities of substrates into thick cell slurries in
hollow fiber factors, using steady-state data. In this
method, the consumption rate constant must be
known. Consequently, steady-state experiments are
limited in the amount of information that they can
reveal. Dynamic experiments are much more valuable,
‘Present address: Department of Chemical Engineering,
McGill University, Montreal, PQ, Canada H3A 2A7.
:To whom correspondence should be addressed.
1489
several mechanism constants (Deisler and Wilhelm,
1953). Ahn ef al. (1985) have shown how dynamic
experimental data can be used to obtain values for the
adsorption, reaction and desorption rate constants in a
catalytic slurry reactor. Steady-state experiments in
this case could only yield an “overall” consumption
rate constant, that would also include adsorption,
desorption and mass transfer effects. The same logic
applies to the determination of mass transfer rate
constants in that, when more than one mass transfer
mechanism is significant, its individual rate constants
can be obtained only from dynamic experiments, since
during the transient period these individual steps
proceed with unequal rates.
From the above it becomes apparent that it is
necessary to have a dynamic model of the process that
will take into account all the mechanisms that are to be
identified. Since the heat of reaction in immobilized-
enzyme systems is insignificant, .an immobilized-
enzyme bioreactor is, from the chemical engineering
point of view, mathematically equivalent to an iso-
thermal catalytic reactor and will be treated as such in
the model development that follows. The first attempts
to present a mathematical solution of the mode1
equations for an isothermal catalytic reactor go back
to the mid 1900s. Lapidus and Amundson (1952)
examined the effect of longitudinal dispersion in ion
exchange and chromatographic columns, assuming
pointwise local equilibrium and neglecting mass trans-
fer resistances. Rosen (1952) was the first to solve the
problem of transient mass transfer to particles in
fluid-particle systems, neglecting axial dispersion and
chemical reaction. He employed the method of Laplace
transforms and obtained the solution in the real-time
domain in the form of an infinite integral of an
 1490 THANOS D. PAPATHANASIOUet al.

oscillatory function. In a later paper (Rosen, 1954), he

discussed a method for the numerical evaluation of this
immoblllzed
integral and presented breakthrough curves for certain urease reactor
values of the process parameters. Masamune and
Smith (1965) solved the model including pore surface /(fluidized bed)

adsorption, again neglecting axial dispersion in the

bulk. Babcock et al. (1966)and Pellett (1966) proposed
an analytical solution for the same problem including
axial dispersion. Rasmuson and Neretnieks (1981) and
Rasmuson (1981, 1982), extended the work of Rosen
and developed analytical solutions including axial
urea reservoir
dispersion and first-order intraparticle chemical reac-
tion, also in the form of an infinite integral of an .I< :
oscillatory decaying function. A method to obtain the _/ __
convergence of this integral has been discussed by ~
Rasmuson and Neretnieks (1981), who discussed the
Fig. 1. Schematic diagram of an immobilized-urease bioreac-
significance of longitudinal dispersion in a situation
tor in which the feed concentration changes continuously
where radionuclides migrate from a repository in the with time.
surrounding bedrock. All the above-mentioned ref-
erences used the Laplace transform to obtain a sol-
ution of the model equations. These solutions come reservoir changes continuously with time as a result of
invariably in the form of an infinite integral, whose the action of the immobilized enzyme. Such a reactor
numerical integration can often lead to convergence system can be used for the treatment of uremic patients
problems (Rasmuson and Neretnieks, 1981; Rosen, as part of an artificial kidney machine. Arbeloa and
1954). Nevertheless, the analytical solutions in the real- Neufeld (1986) have reported experimental data from a
time domain, even derived for a special case or in a system similar to that of Fig. 1. The absence of a model
difficult to use form, have the advantage of being for this type of reactor has hindered the authors from
suitable not only for parameter estimation or inter- fully exploring the results of their experiments.
pretation of experimental data, but also for simulation This work presents a novel computer-
and optimization of the performance of a fluid-solid implementablef.semi-analytical method for the sol-
reacting system operating continuously under trans- ution of the coupled partial differential equations that
ient conditions. Finally, assuming plug flow, Shyam et simulate the dynamic behaviour of an isothermal
al. (1975) presented a detailed study of the dynamics of catalytic reactor or equivalently an immobilized-
an immobilized-enzyme packed-bed reactor with em- enzyme bioreactor, according to the tanks-in-series
phasis on the kinetics of multi-reacting enzyme and dispersion models. Our model includes axial
systems. dispersion, mass transfer effects and a first-order intra-
Apart from the efforts to obtain a mathematical particle chemical reaction. The model is ideal for
solution in the real-time domain, significant work has computer simulation studies, since it comes in the form
been directed towards the interpretation of the chro- of a set of ordinary differential equations (ODES)
matographic peak and parameter estimation using the which can be easily solved by commercially available
method of statistical moments (Kubin, 1965; Kucera, software packages.
1965; Schneider and Smith, 1968). This theory has been
employed to calculate chemisorption rates and rate
constants, intraparticle diffusivities and surface dif- I’ROBLEM FORMULATION
fusivities from chromatographic data (Suzuki and From the modelling point of view, determination of
Smith, 1971; Ahn et al., 1985). According to this the dynamic behaviour requires the solution of the
method, the moments of the chromatographic re- governing differential equations which describe the
sponse are determined experimentally and using ex- mass accumulation in the bulk liquid as we11 as the
pressions which relate the parameters of the system to mass accumulation and consumption in the intra-
these moments one can easily perform parameter particle space. The pertinent independent variables in
estimation studies in fixed- or fluidized-bed systems. this problem are time, axial distance and the radial
Though relatively simple to apply, this approach is (intraparticle) coordinate. The model equations are
limited to linear systems. Furthermore, the solution in coupled through the boundary condition at the par-
the Laplace domain cannot be used for real-time ticle surface which expresses the exchange of mass
design-simulation studies of reactor systems with between the two phases. The phenomena that have
changing feed composition. Even though it is a been considered in the present analysis are:
common assumption that feed concentration can be
held constant, there may exist cases where the reactor (i) Mass accumulation in the bulk liquid.
operates continuously under a varying feed com- (ii) Mass diffusion across the stagnant film which
position. An example is the immobilized-urease reac- surrounds the catalyst particles (external mass
tor shown in Fig. 1, where the ureaconcentration in the transfer).
 Dynamic modelling of mass transfer phenomena 1491

(iii) Mass diffusion and accumulation into the void The partition coefficient (B) is used to express the
space of the catalyst particles (pore-volume partitioning of a substrate between the bulk and the
diffusion). intraparticle space (Papathanasiou et al., 1987). The
(iv) Axial dispersion in the bulk liquid. mass flux from the bulk liquid to the catalyst at any
(v) Intraparticle first-order chemical reaction. In axial position in the reactor is given by
most cases, enzymatic reactions follow com-
plicated non-linear expressions, like simple or
reversible Michaelis-Menten kinetics with or
without product or substrate inhibition.
The mass balance in the bulk liquid, according to the
However, at low substrate concentrations the dispersion model, in dimensionless form, is
enzymatic reactions tend to approach first-
order kinetics. Consequently, the proposed
2 = (;)(!?!?)+LQ*. (7)
model will be suitable for immobilized-enzyme
reactors at sufficiently low substrate
concentrations. The initial and boundary conditions for eq. (7) are:

c L(o’ x) = 0.0 (8)

In addition it is assumed that:
ac
(9 The system is isothermal.
(ii) The effective intraparticle diffusivity is indepen-
!->ax X=0 =
- Pe(C, - cJ (9)

dent of time and also constant throughout

catalyst particle.
the cs\
\ dx 1, = I
=o.
(iii) Beads are spherical with uniform and constant In eq. (7) the term ( l/Pe)(~2C/~x2) is the mass flux in
radius and uniform pore size distribution. the bulk liquid due to axial diffusion (diffusion term),
(iv) The enzyme activity is uniform throughout the (K/ax) is the flux due to convection (convection term)
particle. and Q* is the mass flux from the bulk liquid to the
(v) Pressure drop across the reactor, radial concen- catalyst (i.e. the beads) at any position in the reactor
tration gradients in the bulk, catalyst wear and and is given by eq. (6).
other mechanical effects are negligible. When the tanks-in-series model is used, the ODE
(vi) The particles remain stationary in the bed. This which gives the concentration in the ith CSTR in the
is a very reasonable assumption, since most cascade is given by
systems where the fluidizing fluid is liquid are
characterized by particulate fluidization and dC-
2 = (C,_, -C,)N-QQ: (11)
for many practical purposes behave like ex- dr,
panded fixed beds, with a relatively small
with initial condition
tendency to form aggregates or leave voids
(Hetsroni, 1982; Carlos and Richardson, 1968). C,(O) = 0, i = 1, 2, . . . , N. (12)

Cj and C,_ , are the reactant concentrations in the ith

and (i - 1)th CSTR in the cascade, respectively, and Q:
is given by eq. (6) evaluated at the ith CSTR.
MODEL EQUATIONS
The interparticle and intraparticle mass balance
The equations which describes the mass transfer,
equations are coupled through the flux term Q* given
accumulation and consumption in the intraparticle
by eq. (6). In this approach we apply Duhamel’s
space, along with the corresponding initial and bound-
theorem to obtain the solution of eq. (1) for the time-
ary conditions are:
dependent boundary condition (Ck), using the sol-
ution of the same equation for a constant boundary
D&[g$+(:>%
]-kc&,=sp(??) (1)
condition, as given by Crank (1975). Using this sol-
ution, the mass flux term (Q*) can be evaluated in a
Cb(r, 0) = 0.0 closed form, avoiding unnecessary and time-
consuming intraparticle spatial discretization. After
= k, (Ck-C,) substitution in eqs (7) and/or (1 l), the response of the
r=R reactor can be found. Equation (7) has been discretized
acb for the axial distance variable (x) using central finite

c-1
ar #==CJ
= 0. (4) differences, and was thus converted into an ODE, with
time as the independent variable. Details on the
Ck is the concentration at the inside of the stagnant solution procedure are given in the Appendix.
film which surrounds the bead and is related to the
concentration on its surface by
Bulk liquid concentration
When the tanks-in-series model is used, the ODE for
c,(t) = y
the bulk concentration in the ith CSTR in the cascade
 1492 THANOS D. PAPATHANASIOU et al.
is given by eq. (13): 100 I

dC. Dispersion Model

2 = (Ci_ 1 - C,)N.- Q:. (13) 80 --
dt,
If the dispersion model is to be used, the appropriate
equation is (in central finite difference form)

X, 1 c,,, +c,_, -2cj c,,, -c,_,

_=- -Q*.
at, Pe 6x2 - 26x
(14)
The effect of the type of finite differences used in eq.
(14) on the stability and accuracy of the numerical
solution can be found in standard numerical methods +R
textbooks (Roache, 1972).
Fig. 2. Breakthrough curves as a function of the axial Peclet
number, for a reacting system (dispersion model).
Concentration on the particle surface
The concentration on the particle surface (Ci), as
shown in the Appendix, is given by the solution of 100 I

Tanks-in-Series Model
80 ‘1

= gCr.(ta)-~wR).(15)
P 1
60

40
The functions $.(r,) in eq. (15) are solutions of

2 +&D,[(mt)’
R
+ 9#* Jl&“(tJ = z R
20

0
(16)
1
+t#*eD,C,(t,), n = 1, 2,
I I I I
. . . , no. 0 0.5 1 1.5 2 2.5 3

+R

PARAMETRIC ANALYSIS
Fig. 3. Breakthrough curves of an immobilized-enzyme reac-
Simulation runs were performed for parameter tor according to the tanks-in-series model.
values that extended over a range reflecting practical
values of the process parameters (D,, k,, k, R and
F/V,). The computations were carried out on an (dispersion model). The response curves of the same
175 CDC computer with double precision arithmetics, reactor according to the tanks-in-series model are
using the IMSL subroutine DGEAR to solve the given in Fig. 3. Apart from determining the axial
resulting system of ODES. The ranges of the dimen- dispersion in the reactor using dynamic response data,
sionless parameters were as follows: these curves show how the proposed model can be used
to predict the overall conversion in the reactor as a
9: between 0 and 10 function of the process parameters (mass transfer
D,: between 10d4 and 50 rates, intrinsic reaction rates, bead size, liquid mean
B,: between 0.01 and 100. residence time etc.).
Another interesting observation is that a fixed- or
Very low values of the mass Biot number can be fluid-bed immobilized-enzyme bioreactor, where mass
found in fixed or liquid fluidized beds, as a result of the is transferred from the bulk to the immobilization
anomalously low mass transfer coefficients often en- matrix, exhibits substantially different transient be-
countered in such systems at very low Reynolds haviour than a reactor of the same type filled with non-
numbers. For these cases, the available data deviate porous beads. Figure 4 gives the response of a reactor
from the predictions of the Froesling type equations full of non-porous beads (dotted lines) and the re-
and follow other empirical forms, with the property of sponse of one filled with porous beads (solid lines), for
converging to zero as the liquid Reynolds number various degrees of axial dispersion in the bulk, as
approaches zero. An excellent discussion of this matter described by the axial Peclet number. It is obvious that
can be found in Hetsroni (1982). mass transfer from the bulk to the beads results in an
intense tailing of the response curve. This has also been
7’he shape of the breakthrough curve reported for step change experiments in ion-exchange
Figure 2 shows the breakthrough curves of an columns (Denbigh and Turner, 1971), and has been,
immobilized-enzyme bioreactor when fluid-to-particle qualitatively, attributed to factors such as dead spaces
mass transfer and intraparticle reaction are both into the reactor, adsorption on the bead surface and
significant, as a function of the axial Peclet number axial diffusion. In this work the actual contribution of
 Dynamic modelling of mass transfer phenomena 1493

Effect of B, on Response

Porous Beads _

Fig. 4. Combined etFect of mass transfer and axial dispersion

on the dynamic response of an immobilized-enzyme bioreac-
tor. No reacting system.
0 0.5 1 1.5 2 2.5

mass transfer and axial dispersion on the shape of the

+,
response curve has been illuminated, and, thus, it is Fig. 5. Dynamic response of an immobilized-enzyme reactor
possible to extract the actual Peclet number from step to a step change in the input concentration of a non-reacting
change experiments. Since the effect of mass transfer tracer, as function of the mass Biot number.
can be important, it is essential to know the mass
transfer characteristics in the reactor when we try to the upper and the lower limiting curves (solid and
determine the amount of axial dispersion in the bulk dotted lines, respectively). The upper limiting curve
liquid using dynamic tracer experiments. corresponds to very large external mass transfer
resistance and, consequently, to negligible mass flux to
The importance of the internal and external resistances the beads. It is worthwhile to mention that the
The mass Biot number (B,) is defined as the ratio of particular value of B, that corresponds to this curve
the internal to external mass transfer resistances. In depends on the value ofD,. Furthermore, it seems that
operating immobilized-enzyme bioreactor, with the this limiting curve corresponds to parameter values
diffusing substance and the bead pore structure given, such that D,B, -C 0.01.
the intraparticle diffusivity normally has a fixed value, Alternatively, the values of B, that correspond to
which is unaffected by operating parameters of the negligible external mass transfer resistances depend on
system, such as the dilution rate or the intensity of the the particular value of Da. It can be seen that for large
turbulence in the bulk. On the other hand, these B, (typical values: B, > 100 when DR = 0.05 and B,
parameters highly affect the external mass transfer > 5 when D,= 1) the response curves are identical
coefficient and subsequently the mass Biot number. It and independent of B,. They depend only on the value
is therefore of interest to examine the effect of B, on of D,, as can be concluded by comparing Fig. 5(a) and
the dynamic response of the bioreactor. This will reveal (b). This implies that, in this case, the intraparticle
which of the two mass transfer resistances controls the resistance controls the fluid-to-particle mass transfer.
transfer of nutrients or substrate from the bulk to the For intermediate values of B,, the dynamic response
immobilization matrix and will also dictate ways to depends on both, the particular value of B, and the
overcome mass transfer problems. For example, if the particular value of D,, indicating a combined effect of
fluid-to-particle mass transfer is limited by the ex- internal and external mass transfer resistances.
ternal resistance, increasing the turbulence in the bulk Plots like the ones in Fig. 5(a) and (b) can be used to
could be a way to overcome the problem. If intra- extract the mass Biot number from simple dynamic
particle resistance is the limiting factor, smaller beads experiments, provided that the parameter D,is known.
or beads with a different pore structure should be used Furthermore, they can yield information about the
and no effort should be spent to modify the flow controlling mass transfer step, by simply observing the
pattern in the reactor. position of the experimental response relative to
Figure 5(a) and (b) show the response of a reactor to the two limiting curves. It is noted that the effective
a step change in the input concentration of a non- diffusivity D,, and consequently D,, can be de-
reacting tracer (4 = 0). B, is the varying parameter. A termined by performing a simple batch experiment
first observation is that both Dsand B, affect the shape (Tanaka et al., 1984; Papathanasiou et al., 1987).
and the position of the breakthrough curve. In Fig. 5, Figure 6 gives the shape of the breakthrough curves
one can distinguish two characteristic response curves; of a fixed or liquid fluidized bed immobilized-enzyme
 1494 THANOS D. PAPATHANASIOU et al.

Cascade of 5 CSTRs - flux to Beads

100 , , 1 I I 1
a

B, = 50
D, = 0.1 _

cl 1 2 4 5 6

Fig. 6. Breakthrough curves of an immobilized-enzyme bio-

b
reactor as a function of the Thiele modulus. Note the
transition from reaction controlling (ib < 0.1) to mass trans-
fer controlling ($ z 10).

Stosnont Film Bulk

I1 1I I , i
0 1 2 4 5 6

Fig. 8. Fluid-to-particle mass transfer in each CSTR in a

cascade of five CSTRs in series, as percentage of the mass
flowing into the system.
0
0 0.2 0.4 0.6 0.0 1
(r/R) CSTR in a cascade of five CSTRs, as percentage of the
mass entering the cascade at any time, for various
Fig. 7. Effect of the Thiele modulus on the steady-state values of the Thiele modulus. Comparing Fig. S(a) and
intraparticle concentration profile for beads in the last CSTR (b), it can be seen that the initial mass transfer rate is
in a cascade of three. The reaction controlled regime is clearly higher for higher values of the Thiele modulus. This
indicated (& < 0.1).
reflects an enhancing effect of the chemical reaction on
the overall mass transfer process in accordance with
bioreactor for various values of the Thiele modulus. what is suggested by theoretical analysis of combined
The mass transfer parameters (DR and 8,) have reaction and diffusion. When 4 = 1, the amount of
constant values. For $ < 0.1, the conversion process is substrate transferred at steady state to the immobiliz-
reaction rate controlled, since the reaction rate is very ation beads in each CSTR is approximately the same.
slow. Increasing +, we move to a region where the This implies that the immobilized enzyme is more or
combined effect of intraparticle reaction and mass less equally utilized in each CSTR. By increasing the
transfer determines the response. Finally, for cp > 10, Thiele modulus to 1.5, it becomes obvious that a
the process is mass transfer limited. Figure 7 shows the smaller amount of substrate is consumed in CSTRs
effect of the Thiele modulus on the steady-state towards the end of the cascade. This is an indication for
intraparticle concentration profiles, for beads in the ineflicient use of the biocatalyst, possibly because of
third CSTR in a cascade of three (tanks-in-series substrate depletion along the reactor’s length. The
model). For low I$S, the concentration profiles co- design implication is clear: introduction of side feed
incide, indicating that the process is under kinetic streams could enhance the overall productivity of the
control. As expected, increasing 0 results in steeper cascade.
concentration profiles. The peak observed in Fig. 8(a) and (b) indicates that,
in the initial stages, the mass transfer to the particle is
Substrate consumption in a cascade of CSTRs much faster than the intraparticle consumption; this
When a cascade of perfectly mixed tanks in series is results in a fast “saturation” of the intraparticle space.
used for a particular bioconversion, it is often useful to Afterwards. the mass transferred to the particle drops
know the amount of substrate consumed in each to the corresponding steady-state value.
CSTR. This will give valuable information about the
degree of enzyme utilization in each CSTR and will be CASE STUDY: IMMOBILIZED-UREASE REACTOR WITH
a first step in an effort to solve problems of substrate A RECYCLE LOOP
depIetion along the length of the reactor. Figure 8(a) In order to demonstrate the applicability of the
and (b) show the fluid-to-particle mass flux in each proposed model for the simulation of fixed or liquid
 Dynamic modelling of mass transfer phenomena 1495

fluidized reacting systems with time-varying feed com- major factors that should be considered for optimum
position, we consider the urea degradation reaction reactor performance: (a) the magnitude of the external
which is catalyzed by immobilized urease: mass transfer resistance, and (b) the amount of the
enzyme present in the reactor. A high external mass
(NH,),CO + H,O 4 CO, + 2NHa. transfer coefficient will facilitate the process of sub-
The physical system is depicted in Fig. 1. It is assumed strate transfer to the immobilized enzyme, while a large
that mass transfer inside the support matrix can be amount of enzyme provides the driving force for fast
described by a constant effective diffusivity (D,,,). Urea urea conversion. The superficial velocity (u) affects
solution is pumped at a constant rate from a perfectly both factors, but in a different way. Increasing u
mixed reservoir, passes through the reactor and is reduces the external mass transfer resistance and
recycled back into the reservoir. Such a device can be facilitates the process of urea conversion. On the other
part of an artificial kidney machine, used by chronic hand, higher u results in larger bed voidage and
renal failure patients for the removal of urea from consequently in a smaller amount ofenzyme present in
blood plasma. The reactor volume was taken equal to a the reactor. Figure 9 shows how the part of the reactor
full-scale artificial kidney*device and was lOS< of the volume occupied by beads (1 -6) changes as the liquid
reservoir volume. The simulation experiments that superficial velocity changes. By increasing the liquid
follow focus on the performance of the system at superficial velocity from 0.02 to 0.25 cm/s the bed
various operating conditions, illuminating the effect of voidage increases from 0.52 to 0.9. This represents a
the liquid recirculation rate. five-fold decrease in the reactor volume occupied by
It is assumed that the immobilized urease is charac- beads containing enzyme, even though the velocity
terized by the same kinetic parameters as the soluble (0.25 cm/s) is only 65 “a of the particle terminal vel-
urease. The kinetics of the soluble urease have been ocity. Figure 9 also shows the effect of the liquid
found to be of a Michaelis-Menten form with the superficial velocity on the mass Biot number (B,). As
ammonium anion acting as a non-competitive inhibi- expected, the mass Biot number increases substantially
tor (Ollis and Carter, 1974). Nevertheless, it has been as u increases, since B, is proportional to the external
shown that at low urea concentration levels (around mass transfer coefficient. From Fig. 9 there exists a
2 g/l), a first-order kinetic expression is quite satisfac- range of liquid superficial velocities for which the
tory (Lewis and Middleman, 1974). This kinetic ex- performance of the reactor is optimum.
pression is adopted here and, thus, the urea concen- A criterion for optimum reactor performance will be
tration in the reservoir is assumed to be below 2 g/l. taken as the minimization of the time required for 50 %
The first-order rate constant used is 0.01125 s-l. To urea conversion in the reservoir. Figure 10 shows the
simulate conditions similar to those in an artificial percentage urea degradation in the reservoir, as a
kidney machine, the viscosity of the urea solution was function of time. It can be seen that the conversion
taken the same as the viscosity of human blood plasma becomes faster as the liquid velocity increases from
(1.24 x 10e6 m’is). In the simulations that follow, the 0.01 to 0.02 and 0.1 cm/s. For this range of velocities,
effective intraparticle diffusivity of urea in the im- the improvement in the external mass transfer coef-
mobilization beads was taken as equal to its diffusivity ficient is sufficiently high to overbalance the loss in
in water, which is 1.2 x tom5 cm’/s. The particle utilized enzyme due to bed expansion. As u increases
diameter was 0.03 cm and the minimum-Ruidization further to 0.2,0.25 and 0.3 cm/s, the performance of the
velocity of the particles was taken equal to that reactor deteriorates. In this case, the improvement in
experimentally determined by Arbeloa and Neufeld mass transfer caused by the increase in u is not
(1985) (7.7 x lo-’ cm/s). The liquid superficial velocity sufficient to balance the decline in the amount of the
in the reactor varied between 0.01 and 0.3 cm/s, and the enzyme present in the reactor due to the increase in E.
height-to-diameter ratio was 2.
For the determination of the external mass transfer
coefficient, the Froessling equation is not applicable
because of the very low Reynolds numbers in the 0.6 , , 4.5

reactor (between 0.01 and 1). A formula proposed by

Galloway and Nelson was used instead (Hetsroni,
1982). This formula is suitable for very low Reynolds
numbers and has the property of giving external mass
transfer coefficients that converge to zero as the liquid
velocity approaches zero. For infinite dilution, the
predictions approach those of the Froessling equation.
The bed expansion was determined using the corre-
lation by Richardson and Zaki (Hetsroni, 1982).
The liquid superficial velocity in the reactor is a 0.00 0.05 0.10 0.15 0.20 0.25 0.30

design parameter of primary importance. Its absolute Liquid Velocity (cm/s)

minimum and maximum values are dictated by the
Fig. 9. Effect of the liquid superficial velocity on the reactor
minimum-fluidization velocity (u,,,~) and the particle volume occupied by enzyme (1 -E) and on the mass Biot
terminal velocity (u,), respectively. There are still two number (B,).
 1496 THANOSD. PAPATHANASIOU~~ al.

revealed ways for parameter estimation and system

identification via simple dynamic experiments.
Simulation of an immobilized-urease reactor operat-
ing in a recycle loop for the removal of urea from a
reservoir revealed operating conditions for maximum
productivity and showed the applicability of the
proposed method of solution for the computer simu-
lation of fixed or liquid fluidized beds operating under
a continuously changing feed composition.

O! I
0 100 200 300 400 500
Recirculation Time (min) Acknowledgements-This work was supported by the Natural
Science and Engineering Research Council of Canada. The
Fig. 10. Percentage urea degradation in the reservoir as computations were performed with the facilities of the
function of time, for various liquid superficial velocities. University of Calgary Computer Centre.

NOTATION
kR
& dimensionless number defined as B, = D
en-
CL concentration in the bulk liquid, mol/l
CR time-dependent concentration at the particle
surface, mol/l
Cii concentration in the inner side of the stagnant
film, mol/l

Ofi
. . .u (cm/s)
(Cd, concentration
rium, mol/l
in the bulk liquid at equilib-

(C& concentration in the intraparticle pore volume

Fig. 11. Time required for 50% urea degradation in the liquid at equilibrium, mol/l
reservoir, at various liquid superficial velocities. The observed Cb intraparticle concentration in the pore volume
minimum indicates the existence of optimum operating
liquid, mol/l
conditions.
CO concentration in the feed stream, mol/l
D Fli effective intraparticle diffusivity, cm*/s
DL axial dispersion coefficient, cm2/s

Figure 11 summarizes the results of several simu-

D
DR dimensionless diffusivity, D, =
lation experiments. A minimum observed in the time R2(F;?,) ~
required for 50 oA urea conversion suggests that a range F volumetric flowrate, I/s
of optimum liquid superficial velocities is between 0.02 k first-order reaction constant, l/s
and 0.15 cm/s. This range can be narrowed further by kc external mass transfer coefficient, cm/s
2
introducing enzyme cost considerations. When
k parameter defined as k, = D,
u = 0.15 cm/s, the bed voidage is 0.81. while for
u = 0.02 cm/s it is 0.44. This means that for approxi- L reactor length
mately the same performance, one has to use almost 3 N number of CSTRs in series that simulate a
times as much enzyme when the reactor is operated at fixed or liquid duidized bed bioreactor
the lower velocity than when it is operated at the higher n order of the functions $,(t)
one. This suggests that the range of optimum liquid Pe Peclet number, defined as Pe = (uL/DJ,
superficial velocities is around 0.15 cm/s. dimensionless
fluid-to-particle mass flux [given by eq. (6)]
CONCLUSIONS fluid-to-particle mass flux evaluated at the ith
A dynamic model for a fixed or liquid fluidized CSTR
bed immobilized-enzyme reactor has been presented R bead radius, cm
together with a novel method for the solution of the r intraparticle radial distance, cm
model coupled partial differential equations. The s, constant evaluated by eq. (A22)
solution comes as a system of simultaneous ordinary r time. s
differential equations and therefore can be easily
R
‘
dimensionless time, tR = f
solved by commercially available software packages.
Parametric simulation analysis revealed the signifi- u axial superficial fluid velocity in the reactor
cance of axial dispersion, fluid-to-particle mass trans- vT total reactor volume, 1
fer resistances and intraparticle chemical reaction on vi volume of each CSTR in a cascade, Vi = VJN
the dynamic characteristics of the reactor, The analysis x dimensionless axial distance, x = r/L
 Dynamic modelling of mass transfer phenomena 1497

Greek letfers dispersion in packed beds. A.Z.Ch.E. J. 27, 1032.

Rasmuson, A., 1982, Transport processes and conversion in
partition coefficient, defined as /I =
(Cd,
__ an isothermal fixed-bed catalytic reactor. Chem. Engng Sci.
P
(G), 37, 411.
& bed voidage Rasmuson, A. and Neretnieks, I., 1981, Migration of
bead voidage radionuclides in fissured rock: the influence of micropore
&P diffusion and longitudinal dispersion. J. geophys. Res. 86,
i integration variable
3749.
r time constant, s Roache, P. J., 1972, Computational Fluid Dynamics. Hermosa
0.5
Publishers, Albuquerque, NM.
(P(r) Thiele modulus, (b = 4 i Rosen, J. B., 1952, Kinetics of a fixed bed system for solid
(1 St-r diffusion into spherical particles. J. them. Phys. 20, 387.
*L. functions defined by eq. (A13)
Rosen, J. B., 1954, General numerical solution for solid
diflusion in fixed beds. Znd. Engng Chem. 46. 1590.
REFERENCES Schneider, P. and Smith, J. M., 1968, Adsorption rate
constants from chromatography. A.I.Ch.E. J. 14, 762.
Ahn, B. J., McCoy, B. J. and Smith, J. M., 1985, Separation of
Shyam, R., Davinson, B. and Vieth, W. R., 1975, Mass transfer
adsorption and reaction rates: dynamic studies in a cata- and biochemical reaction in enzyme membrane reactor
lytic slurry reactor. A.Z.Ch.E. J. 31, 541. systems--II. Chem. Engng Sci. 30, 669.
Arbeloa, M. and Neufeld, R. J., 1986, Analysis of a microen-
Suzuki, M. and Smith, J. M., 1971, Kinetic studies by
capsulated urease fluidized bed reactor under steady state
chromatography. Chem. Engng Sci. 26, 221.
and transient operating conditions. J. Membrane Sci. 29, Tanaka, H., Matsumura, M. and Veliky, I. A., 1984, Diffusion
321. characteristics of substrates in Ca-alginate beads.
Babcock, R. E., Green, D. W. and Perry, R. H., 1966,
Biotechnol. Bioengng 13, 53.
Longitudinal dispersion mechanisms in packed beds.
A.1.Ch.E. J. 12, 922.
Carlos, C. R. and Richardson, J. R., 1968, Solids movement in APPENDIX: METHOD OF SOLUTION
liauid Ruidized beds 1: particle velocity distribution. Chem. The solution is based on the analytical solutions of the heat
Engng Sci. 23, 813. _ transfer equations derived by Carslaw and Jaeger (1959) and
Carslaw. H. S. aad Jaeger, J. C., 1958, Conduction of Heat in later adapted to mass transfer problems by Crank (1975).
Solids, 2nd edition, p. 230. Clarendon Press, Oxford. First we apply the following lemma (Crank, 1975):
Churchill, R. V., 1975, Operationa/ Mathematics, 3rd edition,
p. 309. McGraw-Hill, New York. Lemma 1: Let U, be the solution of the equation
Crank, J., 1983, The Mathematics of Diffusion, 3rd edition.
Oxford University Press, London. !TKDazu (Al)
Davis, M. E. and Watson, L. T., 1985, Analysis of a diffusion at ax2
limited hollow fiber reactor for the measurement of
effective substrate diffusivities. Biotechnol. Bioengng 17, for constant surface boundary condition. Then, the solution of
the equation
182.
Diesler, P. F. and Wilhelm, R. H., 1953, Diffusion in beds of
porous solids. Znd. Engng Chem. 45, 1220.
‘?I!= ,,??%_kU (A21
Denbigh, K. G. and Turner, J. C. R., 1971, Chemical Reactor
at ax2
Theory, 2nd edition, p. 90. Oxford University Press, is given by the following integral:
London. -,
Hetsroni, G., 1982, Handbook of Multiphase Systems. U=k Ule-“dt + U le -kt . (A3)
Hermosa Publishers, Albuquerque, NM. J0
Kubin, M., 1965, Beitrag zur Theorie der Chromatographie.
It is easy to see that the modified intraparticle mass balance
II. Einfluss der Diffusion Ausserhalb und der Adsorption
can be brought to a form similar to that of eq. (1) by dividing
Innerhalb des Sorbens-Korns. Colln Czech. them. Commun.
by E,, and applying the transformation U = Cr. The solution
30, 2900.
of eq. (Al) for constant surface boundary condition (C,) is
Kucera, E., 1965, Contribution to the theory of chromato-
known (Crank, 1975). Substituting U, by C,r, we obtain
graphy: linear non-equilibrium elution chromatography.
J. Chromat. 19, 237. Cl@, 1)
-= (A4)
Lapidus, L. and Amundson, N. R., 1952, Mathematics of
CO
adsorption in beds. VI. The effect of longitudinal disper-
sion in ion exchange and chromatographic columns. J. This is the intraparticle concentration profile for no
Whys. Chem. 56, 984. chemical reaction and constant surface concentration C,.
Lewis, W. and Middleman, S., 1974, Conversion in hollow- Evaluating the integral in eq. (A3) with C, from eq. (A4) we
fiber membrane enzyme reactor. A.Z.Ch.E. J. 20, 1012. obtain the solution of the intraparticle transient mass balance
Masamune, S. and Smith, J. M., 1965, Adsorption rate for constant surface boundary condition:

1.
studies-interaction of diffusion and surface processes. k + kme- t(k+ k-1
A.Z.Ch.E. J. 11, 34.
Ollis, D. F. and Carter, R. Jr., 1974, Kinetic analysis of urease k+k.
electrode. In Enzyme Engineering, Vol. 2. Plenum Press, (A5)
New York. To proceed further and obtain the solution of eq. (A2) for a
Papathanasiou, T., Kalogerakis, N., Behie, L. A., Gaucher, G. time-varying surface condition we need to use the following
M. and Thibault, J., 1987, Modelling the dynamic behavior lemma (Churchill, 1975).
of immobilized cell/enzyme bioreactor: the tanks-in-series
model. In Biotechnology Processes: Scale-up and Mixing
Lemma 2 (Duhamel’s theorem): Consider the boundary
(Edited by C. S. Ho and J. Oldshue), p. 238. value problem
Pellet, G. L., 1966, Longitudinal dispersion, intraparticle
diffusion, and liquid-phase mass transfer during flow W(x, Y. 2, t) = D(W) (A6)
through multiparticle systems. Tappi 49.
where D(W) is a linear differentialform. The initial condition is
Rasmuson, A., 198 1, Exact solution of a model for diffusion
and transient adsorption in particles and longitudinal wtx, y, 2, 0) = 0 t.47)
1498 THANOS D. PAPATHANASIOUet al.
and the boundary condition is
W(S, t) = F(t) (Ag)
where S indicates the boundary ofthe system. IfZ(x, y, z, t) is
the solution ofthe problem when F(t) = 1, then the solution for
the same problem when F(t) varies with time is

aZ~x,Y,z,t--.)d.,
W(x, y, 2, t) = F(i.)p (A91
.Jt-- L
Using the above property, one can easily extract the
solution of eq. (1) for the time-variable surface condition
C,(t), namely

(AlO)

Substituting eq. (A5) into eq. (AlO) and performing the

integration we obtain

9.(t). (Al 1)

The functions (L_(t) are given by

0 I I 1I 1 1 r I
0 1 2 t: 4 5 6

or, in differential dimensionless form

Fig. Al. The form of the various-order )(1. functions in the

1
W. presence of intraparticle chemical reaction.
dt+~D, (n1-~)~+9@ $.(tR) = d>+9+‘eD,C,(t,)
R R
(A13)
with initial condition dangerous, since these terms do not always converge to zero,
as can be seen in Fig. Al. With the above considerations, the
rlr.(O)= 0. (A14) 7.
Using eq. (Al l), the partial derivative (X,/&), = R appear- infinite summation 2$“(f) can be split as follows:
ing in the flux term [eq. (6)] can be easily evaluated by
1
differentiation. The result is remarkabiy simpte:

r$),-R=(;)fQ.(r,).
Evaluation of the inJinite summation x tin(t) The concentration Ck(tR) is related to the concentration in
the bulk (C,) via eq. (11) which can be written as
In eq. (Al 3), the dynamics of the t,Q,,
iunctions depend of the
value of the coefficient T = l/.sDg [(nx)’ + 9@‘]; small values
of r imply that the dynamics of the JI. functionsare fast, which
in turn means that the IL,,(t) functions follow the forcing
function on the right-hand side of eq. (A13). It is therefore or, using eq. (A15):
reasonable to assume that for r smaller than a certain value,
rmin, the #,, functions can be evaluated with sufficient
accuracy using the quasi-steady-state approximation:
cii=CL- ( ; )pb”.\Dm/ 1
(A19)

Equations (A17) and (A19) can be combined to give a

differential equation for the concentration Ck, as a function
of the process parameters and the reactant concentration in
the bulk. This equation follows:
The form of the I//, functions, evaluated by numerical
solution of eq. (AL3), can be seen in Fig. Al. The errors
associated with the quasi-steady-state approximation [eq.
(A16)] become really insignificant if one considers how small
is the magnitude of the higher-order &,s. For 4 = 1, = $C,(r,)- &&) tA20)
rL1cw = 33, while (Llb(0.5) is about 0.5. A 3 % error in the P I

approximation of (cllb(0.5) willactually befess than 0.04 “/0 of where Si is the summation:
the value of t&,(0.5) and about 0.02 Y0 of x$“_ On the other
k-1)
hand, simply neglecting the higher-ordei terms could be