Chemical Engineering Science, 1965, Vol. 20, pp. 247-254. Pergamon Press Ltd., Oxford. Printed in Great Britain.

A Monte Carlo treatment for reacting and coalescing dispersed phase systems

by L. A. SPIELMANand 0. LEVENSPIEL
Illinois Institute of Technology, Chicago

(Received 20 January 1964; in revised form 29 April 1964)

Abstract-The Monte Carlo method of digital computer simulation is used to study the influence of
coalescence on the progress of reactions occurring in the dispersed phase of two-phase systems in
backmix reactors. Three different types of chemical reaction of both one and two reactants are
studied and performance charts are prepared and discussed. It is shown that depending on the sys-
tem, coalescence can either increase or decrease the extent of conversion of reactants. The Monte
Carlo method, because of its simplicity, is found to be especially suited to studying discrete systems
of this type since analytical solution is not yet possible and other numerical methods are very cum-
bersome. Possible extensions are discussed and the possibility of using the dispersed phase model
for studying mixing and reaction in homogeneous systems is pointed out.

CONSIJXRtwo fluids, a continuous and a dispersed
phase, flowing at a steady rate through a backmix
(ideal stirred tank) reactor. Let reaction of one or
two components occur in the dispersed phase and
let the continuous phase act either as carrier mate-
rial, or provide reactant by mass transfer, or cata-
lyse the dispersed phase reaction. In such systems,
the continual coalescence and breakup of dispersed
phase droplets redistributes reactant material in
these drops and thereby can influence the course of
the reaction.
CURL [l] was the first to study this interrela-
tionship quantitatively. His model was extremely
simple, nevertheless, the expression found for the
distribution of concentration of reactant in the
drops turned out to be the very special nonlinear
integro-differential equation
p aYe9
- 0 = 6(x - 1) - y(x, t) +
at
+4R
s x
tant by a numerical procedure using the finite
difference form of equation (1).
An alternate approach to problems of this type,
a Monte Carlo procedure, is presented here. This
method of numerically simulating the physical
situation is conceptually simple and bypasses com-
pletely expressions such as equation (l), hence
avoids the troublesome mathematical convergence
and stability problems of the tinite difference pro-
cedure. The Monte Carlo method is also adaptable
to different reaction kinetics and to relaxation of
some of the assumptions of the model, and can be
used to study the effect of coalescence on the ap-
proach to steady state conditions.
This Monte Carlo procedure is used here to
prepare performance charts for the following three
classes of reactions occurring in the dispersed
phase :
(0 A single reactant disappearing in accordance

0
y(x + ~1,t)y(x - a, t) da
with a zero order rate law

-r=k (2)

- Ry(x, t) + s & [x”y(x, t)] . . . (ii) A single reactant disappearing in accord-

ance with a second order rate law
. . . P, R, S constant (1) -r=kc2 (3)
for which no analytical solution is known. He then (iii) Two reactants A and B which enter in
solved this for zero order reaction of a single reac- identical fashion in separate dispersed phase

247
 L. A. SPIELMANand 0. LEVENSPIEL
streams and disappearing in accordance
with a second order rate law
-r, = -rB = kc,c, (4)
First order reactions are unaffected by the coa-
lescence rate [2] hence are not of concern here.
THE MONTE CARLO METHOD OF SIMULATING THE
PHYSICAL SYSTEM
The physical model used is the same as CURL’S
and has the following assumptions:
(i) In the stirred tank reactor, the dispersed
phase consists of a very large number of
droplets, equal in size and constant in num-
ber.
(ii) These droplets coalesce two at a time and
redisperse instantly to form two identical
droplets.
(iii) Droplets coalesce at a constant rate, how-
ever, the probability of coalescing is the
same for all drops.
(iv) The concentration is uniform throughout
each droplet and reaction occurs in the
droplets alone.
(v) The continuous phase is homogeneous and
is chemically inert.
In the Monte Carlo simulation, droplets are
pictured as entering and leaving the reactor at a
steady rate and one at a time for the cases where a
single reactant is disappearing in the dispersed
phase. The average time between entering droplets
is
(5)
where d is the number of droplets entering and
leaving the reactor per unit time. From the assump-
tion of equal droplet size the mean residence time
of droplets in the reactor is
Vf N
-=- (6) in the reactor (a random subscript is generated)
==Q d
where Vf is the volume of dispersed phase fluid,
Q is the volumetric flow rate of dispersed phase in
and out of the reactor, and N is the total number
248
of droplets in the reactor. Combining equations (5)
and (6), the time between entering droplets is
Vf =
T=eN=N
For two reactants, the droplets are pictured as
entering in pairs, one droplet containing A and the
other containing B. The time T between entering
pairs of droplets is calculated in a manner similar
to that used for a single reactant, thus
where Q is the total volumetric flow rate of both
dispersed phase streams to the reactor.
Computer calculations can now be based upon
the time T. Computer arrays are set up, one cor-
responding to each reactant. A particular subscript
of the array corresponds to a particular droplet
in the reactor. The number in each of the array
locations corresponds to the dimensionless con-
centration of reactant in the individual droplet.
The dimensionless concentration of a single reac-
tant is defined as
where cO is the initial reactant concentration in the
dispersed phase inlet stream.
For two reactants entering in identical fashion,
the dimensionless concentrations are given by
(10)
where c,, and cBOare obtained by calculating the
initial moles of each reactant per unit volume of
combined feed.
Flow simulation
A droplet is selected at random from the droplets
and replaced by a fresh droplet having initial con-
centration of the particular reactant entering (the
dimensionless concentration in the entering droplets
is substituted in the subscripted location of the
 A Monte Carlo treatment for reacting and coalescing dispersed phase systems

array designating the entering reactant and a “zero”

is placed in the same subscripted location of the
array designating the other reactant if there is more
than one reactant).

Simulation of dispersed phase coalescence.

(c),,, = 0 [S 2 mt]
0

Two droplets are selected at random (two differ- For second order decomposition of a single reac-
ent random subscripts are generated) and are tant,
allowed to coalesce and redisperse, forming two
new droplets having equal concentrations of each (C)t+= = (O (15)
reactant (the numbers in the two subscripted loca- 1+ $%C)*
tions of the respective arrays are averaged and the
average replaces each of the two original numbers). For second order decomposition of two reactants A
Define a dimensionless coalescence parameter I and B,
such that on the average each droplet coalesces I
times during its stay in the reactor. Then (CAL, T - (CA),

= (GA+ T - CC,), (16)

(11) (cAM_(cAIt - GM
= (c,), _ (c,), eZI(C~)t-(CA)tlCkc~o~/NI - (‘h
where u is the number of coalescences occurring in
the reactor per unit time. Combining equations (7) Calculation of the chemical conversion
and (11) for a single reactant, the number of
The mean dimensionless concentration of reac-
coalescences occurring in the time interval T be-
tant in the system at any instant is given by
tween entering droplets is
I C=*j$
apUZ=_ (12) (17)
N 2
where the average is taken over all the droplets in
Combining equations (8) and (11) for two reactants,
the system.
the number of coalescences occurring during the
The procedures for stimulating flow, coalescence
time interval T between entering pairs of droplets
and reaction are consecutively repeated until the
is
value of i5 is observed to reach a steady value indi-
cating approach to steady state. Actually, the value
(13)
of C is seen to fluctuate with the addition of drop-
lets, even at steady state. Thus a time average
Equations (12) and (13) showthat Z/2corresponds to
concentration is needed and is given by
the number of coalescences occurring in the reactor
in the time interval between entering droplets.
F=F; (18)
Because of this I/2 is used as the counting para-
meter in the Monte Carlo procedure even though
t
Z is chosen as the basic coalescence parameter. where Nt is the number of values of c averaged with
time at steady state.
Simulation of dispersed phase chemical reaction For a single reactant the conversion is
The concentration of reactant in each of the N X=1-‘c (19)
drops changes during the time interval T. This is
For two reactants A and B entering the reactor at
evaluated according to the integrated dimension-
equal rates and concentrations,
less forms of equations (2), (3) and (4). For zero
order decomposition of a single reactant, x = 1 - _t(FA + FB) (20)

249
 L. A. SPIELMAN
By using various N values from 20 to 2000 it
was found for values of 500 and larger that no
significant difference existed in the outlet conversion
throughout all regions of the parameters used in
the charts. Thus, a value of 500 was used in pre-
paration of the charts even though for low conver-
sions smaller values could safely be used. For
details see [3].
Calculations were done on the IBM 7090/1401
digital computer system of the Illinois Institute of
Technology Research Institute. Computation times
vary with values of the parameters and increase as
the value of N is increased. Estimates of the aver-
age calculation time per point are 1.2 min for the
zero order programme, 1.6 min for the single
reactant second order programme and 5.3 min for
the two reactant second order programme.
About thirty points were used for each of the
three performance charts.
FIGS 1, 2 and 3 compare the performance of a
reactor with a finite coalescence rate to that of one
with an infinite coalescence rate. The comparison
is made by determining the reactor volume ratio
necessary to effect the same specified conversion in
each reactor, all other conditions of operation
being the same. These charts show that this
volume ratio is dependent on the type of reacting
system, the reaction rate group, the conversion
level and the coalescence rate group I.
The behaviour for the extremes of coalescence
can be found analytically. Thus at infinite coa-
lescence rate all the droplets are identical in com-
position and the system corresponds to a micro-
fluid, an ordinary homogeneous fluid, in a backmix
reactor. The performance of this system is calcu-
lated readily by the usual material balances.
For no coalescence the dispersed phase acts as
a macrofluid in which case the conversion depends
on the reaction kinetics and the residence time
distribution of droplets in the backmix reactor.
RIETEMA[4] has shown that the outlet concentration
for zero order reaction with no coalescence is
kz kz
=C= 1 -z+-exp n=O (21)
CO
250
and 0. LEVENSPIEL
DANCKWERTS[5] has shown that the outlet con-
centration for second order decomposition of a
single reactant with no coalescence is given by
Z=&exp(-&---e(k) n=2 (23
Curves of constant dimensionless reaction para-
meter are also shown on the charts. For further
details of the preparation of the performance
charts see SPIELMAN[3].
DISCUSSIONOF PERFORMANCE
CHARTS
Zero order reaction in the dispersed phase
The results for zero order reaction show that
chemical conversion increases as coalescence rate
increases, all other conditions being kept constant.
CURL [l] obtained curves for this case by solving
the finite difference form of equation (1). Where
the same values of the coalescence parameter I
were chosen, his curves are in substantial agreement
with those of FIG. 1. The curves of FIG. 1 are more
complete than those of CURL [l] in that they include
a wider range of reaction rate group and coalescence
parameter.
CURL [l] also points out that this system repre-
sents reaction in the dispersed phase between con-
tinuous and dispersed phase reactants in which
mass transfer of reactant into the droplets is the
relatively slow and rate controlling step of the
process.
For zero order reaction, FIG. 1 shows that the size
ratio V/V, approaches the dimensionless reaction
parameter kz/c, at high conversion. This is because
the value of kT,/cO (for infinite coalescence rate)
increases to approach unity as conversion rises.
For further details see CURL [l J and SPIELMAN[3].
Second order decomposition of a single reactant in
the dispersed phase
The results for second order decomposition of a
single reactant, FIG. 2, show that chemical conver-
sion decreases as the coalescence rate increases.
Thus coalescence plays opposing roles in this and
the other systems studied in this paper.
 A Monte Carlo treatment for reacting and coalescing dispersed phase systems

Second order decomposition of two reactants in the
dispersed phase
FIGURE 3 represents two reactants entering the’
reactor with equal flow rates and concentrations
and reacting in accordance with a second order
rate expression. In systems of this kind the follow-
ing limiting conditions may be recognized. First,
no reaction can occur unless some coalescence
occurs. Secondly, the coalescence rate is the limit-
ing factor in determining the maximum extent of
conversion possible in any reactor. This is shown
in FIG. 3 by the approach of the I = constant lines
to the vertical, thus representing ultimate conver-
sion at infinite reaction rate. These limiting con-
versions for infinite reaction rate are also shown in equation has to be derived and solved for each
FIG. 4.
No comparison with the literature for second
order reaction of a single reactant or two reactants
was possible because these situations have not been
studied previously except for the extremes of zero
and infinite coalescence rate [2].
The use of the performance charts for quanti-
tative predictions depends on empirical determi-
nation of coalescence rates in real systems. Experi-
mental work dealing with this subject has been
done by MADDEN and DAMERELL[5] and by
MILLER et al. [7].
COMMENTS ABOUTTHE MONTE CARLO METHOD
With the analytical approach, a different type of
system studied. Since with the very simplest of

Fraction unconverted : I-X

FIG. 1. Performance of a backmix reactor with some coalescence as compared with one with an infinite coalescence
rate for a zero order reaction occurring in the dispersed phase.

251
 L. A. SPIELMANand 0. LEVENSPUSL

Fraction unconverted: I-X

FIG. 2. Performance of a backmix reactor with some coalescence as compared with one with an infinite coalescence
rate for a second order reaction of a single reactant in the dispersed phase.

1.0
C

Fraction of A or B unconverted: I-X

FIG. 3. Performance of a backmix reactor with some coalescence- as compared with one with an infinite coalescence
rate for a second order reaction in the dispersed phase between two reactants entering in identical fashion in separate
dispersed phase streams.

252
 A Monte Carlo treatment for reacting and coalescingdispersedphase systems
I I I11111
‘0.0’ 2 1 . 56ll)90,, .? 3 .
Fraction of A or B unconverted:
FIG. 4. Maximum possible extent of conversion
as a function of coalescence rate for an infinitely
fast second order reaction in the dispersed phase
between two reactants entering in identical fashion
in separate dispersed phase streams.
cases (one reactant) analytical solution is not yet
possible [l], finite difference methods must be used,
and as a consequence one must expect to encounter
different convergence and stability problems for
each and every flow model and reaction kinetics
studied. The extension of this procedure to more
than one reactant is much more difficult and time
consuming. As opposed to this, the Monte Carlo
method simulates the physical system directly, hence,
extensions to different kinetics of flow models
involve only minor modifications of the basic
computer programme.
The Monte Carlo method has many possibilities
in the study of dispersed phase reacting systems.
For example, with minor modification it can be
used to study extensions of the simple physical
model used for stirred tank reactors, e.g. non-
instantaneous breakup of drops, unequal droplet
sizes, or non-random coalescences of the different-
sized drops. Of course only experiment can tell
what extensions should be used to represent real
systems. Extensions can also be made to account
for different types of reactions, in particular, to
E 253
study product distribution from multiple reactions
(consecutive and parallel).
The Monte Carlo method can be used to study
flow patterns other than backmix flow, say dispersed
phase systems with small deviations from plug flow.
This could be accomplished by moving droplets in
file through a computer array with local interaction
of the droplets.
Since the Monte Carlo method is a computer
simulation of the physical system, each calculation,
in effect, shows the change with time of the system
in its approach to steady state. This may be useful
in dynamic studies. The unsteady state results for
the reacting systems studied here have not been
published but are available in the form of the origi-
nal computer output [3].
From the above discussion, we conclude that the
Monte Carlo method, because of its generality,
flexibility and simplicity, is especially suited for
567119,
studies of mixing and reaction in discrete systems.
I-X
Finally, it should also be pointed out that these
dispersed phase mixing models studied by the
Monte Carlo method may have interesting possi-
bilities in related areas, to study solids mixing or
to provide a simple representation for the mixing
of homogeneous systems. The latter possibility
was mentioned by CURL [l] and would be an
extension of the work of DANCKWERTS [5] and
ZWETERING [8].
NOTATION
Concentration of reactant in droplet
d Dimensionless concentration of reactant in drop-
let
c Mean concentration of reactant in the dispersed
phase at any instant
E Time-average concentration of reactant in the
dispersed phase
(0 Dimensionless reactant concentration in a droplet
at time t
d Number of droplets entering reactor per unit time
f Fraction of total volume consisting of dispersed
phase
I Average number of coalescences experienced by
a drop in passing through the reactor. (I/Z is
the number of coalescences occurring in the
reactor in the time interval between entering
droplets.)
k Reaction rate constant
kr Dimensionsless reaction rate parameter for zero
-
co order decomposition of a single reactant
 L. A. SPIELMANand 0. bVENSPIEL

. kcv Dimensionless reactionrate parameter for second
order decomposition of a single reactant
(kcAo7) Dimensionsless reactionrateparameterforsecond
order decomposition of two reactants
Reaction order
1E: Total number of droplets in reactor
Number of values of C used to obtain time-
Nt
average reactant concentration at steady state
Q Volumetric flow rate of dispersed phase passing
through reactor
-r Rate of disappearance of reactant per unit VOW-
ume of dispersed phase
t Time
Mean residence time of dispersed phase
; Average time between entering droplets (or pairs 6(x - 1) Dirac delta function (pulse at x = 1)
of droplets for two reactant systems)
II Number of coalescences occurring in reactor per
unit time for two reactant systems
V Total reactor volume (total volume of both con-
tinuous and dispersed phases)
X Fractional conversion of limiting reactant
Subscripts
A Refers to reactant A
B Refers to reactant B
0 Refers to amount of reactant per unit volume of
dispersed phase in total feed
00 Refers to infinite coalescence rate
Functions
ei(x) Exponential integral function
exp(x) Exponential function

CURL R. L., Amer. Inst. Gem. Engrs J. 1963 9 175.

LEVENSPIEL O., Chemical Reaction Engineering, Chap 10. Wiley, New York 1962
SPIELMANL. A., Monte Carlo Methods for Coalescence of Reacting Dispersed Phase Fluids. M.S. thesis, Illinois Institute
of Technology, 1963.
RlETEMA K., Chem. Z&gng Sci. 1958 8 103.
DANCKWERTS P. V., Chem. Engng Sci. 1958 8 93.
MADDENA. J. and DAMERELL G. L., Amer. Inst. Chem. Engrs J. 1962 8 233.
MILLERR. S., RALPHJ. L., CURLR. L. and TOWELLG. D., Amer. Inst. Chem. Engrs J. 1963 9 196.
Z~IETERINOT. N., Chem. Engng Sci. 1959 11 1.

Zusamme&amng-Mit Hilfe der Monte Carlo-Methode wird der Einfluss der Koaleszenz auf den
Reaktionsablauf in dispersen Phasen in einem Reaktor mit Rilckmischung untersucht. Verschiedene
A&en von Reaktionen werden studiert und es wurde gefunden, dass das Zusammenfliessen den
Umsatz erhijhen oder herabsetzen kann. Die Monte Carlo-Methode ist wegen ihrer Einfachheit
speziell zur Untersuchung dieser Probleme peignet, fiir welche ncch keine analytischen Losungen
bekannt sind. Die Mijglichkeit der Anwendung zur Untersuchungdes Mischeffektes und desReaktions-
ablaufes in homogenen Systemen wird diskutiert.

254