3

Nonideal Flow

1.1 Introduction
Basic chemical reactors (plug flow reactor or PFR, and continuously stirred tank
reactor or CSTR) are studied considering their behavior is that of an ideal reactor.
Unfortunately, in practice, we often find behaviors that are far from that consid-
ered ideal. Consequently, working with them, the chemical engineer must be able
to handle and diagnose the behavior of these reactors. At the time of describing
the nonideal behavior of a reactor, three concepts are introduced: the residence
time distributions (RTDs), the quality of the mixture (not discussed in this book),
and the models that can be used to describe the reactor. These three concepts are
used to describe the deviations of the mixing assumed in the ideal models and
are considered as attributes of the mixture in nonideal reactors.
One way of approaching the study of nonideal reactors is to consider them, in
a first approximation, as if the flow model were the one corresponding to a CSTR
or a PFR. However, in real reactors, the nonideal flow model implies a minor con-
version, so a method that allows for this conversion loss to be considered must
be available. Therefore, a higher level of approximation implies the use of infor-
mation about the RTD.

1.2 Residence Time Distribution (RTD) Function

The idea of introducing the RTD in the analysis of the behavior of reactors
occurred thanks to MacMullin and Weber (in 1935), although it was Danckwerts
(later, in 1953) who structured this analysis and defined most of the distributions
of interest.
In an ideal PFR, all the particles (or units) of material that leave the reactor have
remained in it the same time. Analogously, in an ideal (well-mixed) batch reactor,
all particles are in the reactor the same period of time. The time that these units
have remained in the reactor is what we call the residence time of those particles
in that reactor.
The ideal reactor’s PFR and batch are the only ones in which all the portions
of reactants present in the reactor have the same residence time. In all other
reactors, the particles entering the reactor vessel remain inside the reactor for
different periods of time; that is, there exists a RTD inside the reactor.

Chemical Reactor Design: Mathematical Modeling and Applications,

First Edition. Juan A. Conesa.
© 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.
4 1 Nonideal Flow

For example, consider an ideal CSTR; the input flow that is introduced to the
reactor at a given moment mixes instantaneously and completely with the rest of
the material that already exists inside the reactor. In this way, some of the particles
that enter the reactor abandon this one almost immediately with the exit current,
whereas other atoms remain of almost indefinite form, since all the material is
never dragged. Of course, many of the particles leave the reactor after a period
close to the average residence time.
The RTD of a reactor is a feature of the mixture that is taking place inside the
reactor. Thus, in an ideal PFR there is no axial mixing, and this absence is reflected
in the RTD that this type of reactors exhibit. In contrast, in an ideal CSTR there
is a great degree of mixing, so the RTD that these reactors exhibit is very differ-
ent from that of the plug flow. However, not all RTDs are unique to one type of
reactor; reactors with marked differences can give identical RTDs. Despite this,
the RTD of a certain reactor presents distinctive keys with respect to the type of
mixture that is taking place inside it and is one of the ways to characterize the
reactor that provides more information.

1.2.1 Measurement of the RTD

The experimental measurement of the RTD is done by injecting a tracer into the
reactor at a definite time (t = 0). This tracer is a chemical, molecule, or an inert
atom. The tracer concentration is then measured at the outlet at different times.
The tracer must be inert, easily detectable, with physical properties similar to
those of the substances present in the reaction mixture and easily soluble in it.
In addition, it should not be adsorbed on walls or other surfaces on the reac-
tor. The objective is to reflect, as best as possible, the behavior of the substances
that are flowing through the reactor. The most commonly used tracers are dyes
and radioactive material, while the most commonly used injection methods are
pulse input (“Dirac delta function”) and step input (suddenly increase of the tracer
concentration).

1.2.1.1 Pulse Input

This kind of experiment consists of introducing into the current entering the reac-
tor, quickly and at once, an amount N 0 of tracer. The output concentration is
subsequently measured as a function of time. Concentration-time curve charac-
teristics for the input and exit of an arbitrary reactor can be observed in Figure 1.1.
The concentration-time curve corresponding to the effluent is called curve C in
RTD analysis.
Let us consider a system with a single input and a single output, in which a
tracer is injected in pulse and in which it is transported, exclusively because of
the flow (not because of dispersion), through the system. If a time increment Δt
small enough is chosen so that the concentration of tracer, C(t), which leaves
the system between the instants “t” and “t + Δt” is constant, we can express the
amount of tracer (in moles or grams), ΔM, that leaves the reactor between “t”
and “t + Δt” as
ΔM = C(t) ⋅ Q ⋅ Δt (1.1)
 1.2 Residence Time Distribution (RTD) Function 5

Tracer
input

Feed Output
Reactor

Tracer
Tracer detection
concentration Tracer
concentration
Pulse input
Pulse
response
0 Time
0 Time
Tracer Tracer
concentration concentration

Step input Step

response

0 Time 0 Time

Figure 1.1 Measurement of the RTD.

That is, ΔM is the amount of tracer (in moles or grams, for example) that has
remained in the reactor for a time interval comprising between “t” and “t + Δt.”
Dividing by the total amount of tracer injected into the reactor (M0 ), we obtain
the tracer fraction whose residence time in the reactor is between “t” and “t + Δt”:
ΔM C(t) ⋅ Q
= ⋅ Δt (1.2)
M0 M0
For a pulse injection, we define the RTD function, E(t), as
Q ⋅ C(t)
E(t) = (1.3)
M0
This expression describes in a quantitative way how long the different elements
of fluid have passed inside the reactor. In consequence:
ΔM
= E(t) ⋅ Δt (1.4)
M0
If M0 is not directly known, it may be obtained from the output concentrations,
adding the different amounts of the tracer that have exited the reactor between 0
and infinity. Expressing in differential form:
dM = Q ⋅ C(t) ⋅ dt
and integrating, we obtain
∞
M0 = Q ⋅ C(t) ⋅ dt (1.5)
∫0
Since the volumetric flow (Q) is generally constant, we will define E(t) as
C(t)
E(t) = ∞ (1.6)
∫0 C(t) ⋅ dt
6 1 Nonideal Flow

In that expression, the integral in the denominator is the area under the curve,
C(t). In this way, from the tracer concentration, C(t), it is possible to find the curve
E(t), as long as that curve is obtained from a perfect pulse of the input tracer.
Another way to interpret the function of residence times is in its integral form:
[ ] t2
Amount of tracer exiting the reactor after
= E(t) ⋅ dt (1.7)
passing inside it a time between t1 and t2 ∫t1
Now, the fraction of tracer that passed inside the reactor a time t, between 0
and infinity is equal to 1; therefore:
∞
E(t) ⋅ dt = 1 (1.8)
∫0
The main drawback of the use of the pulse technique lies in the difficulty in
achieving a tracer input to the reactor that is reasonably pulsed (as we explain
in Chapter 2, deconvolution of curves). The injection should take place in a very
short period compared to the residence times and the tracer dispersion between
the injection point and the reactor inlet should be negligible. If these conditions
are satisfied, the technique is a simple and direct way to obtain the RTD.

1.2.1.2 Step Input

During a step input experiment, a tracer is added at a steady rate to the reactor
feed to give a steady input concentration of C 0 (see Figure 1.1). The concentration
of the tracer in the effluent is then monitored from the time of adding the tracer
until it reaches a concentration approximating that of C 0 .
As stated before, the RTD curve can be easily obtained by injecting a tracer
in a pulse input, but now we will formulate a more general relationship between
a tracer injection (not necessarily a pulse input) and the corresponding tracer
concentration in the effluent (current leaving the system).
Chapter 2 presents a more general equation establishing the relation of the con-
centration of a tracer leaving a reactor (C out ) and the input concentration (C in ).
This is presented as the convolution integral:
t
′ ′ ′
Cout (t) = Cin (t − t )E(t )dt (1.9)
∫0
In this equation, E(t) is the corresponding RTD for the reactor. The concentra-
tion at the input is evaluated at different times and the resulting signal C out is a
convolution of C in and the RTD in the vessel.
In this way, in the following chapters we are able to calculate the E(t) curve
given any function of concentration input C in and measuring C out . For now, we
continue with more simple techniques, considering that the inlet concentration
is introduced in the form of a step input (Figure 1.1).
Let us consider a system of constant volumetric flow (Q) in which a tracer input
is introduced in the form of a step, with a rate of addition of tracer to the constant
input flow and which begins to occur at time t = 0. Before that moment, it is
considered that no tracer enters with the inlet current. Under these conditions:
Cin (t) = 0 for t < 0
Cin (t) = constant = C0 for t ≥ 0
 1.2 Residence Time Distribution (RTD) Function 7

The tracer concentration in the feed should be maintained at this value until
the concentration of tracer in the effluent is practically C 0 , i.e. equal to that of the
feed, at which time the test can be interrupted. Figure 1.1 shows a typical output
concentration curve for this type of input.
Since the input concentration (C 0 ) remains constant over time, we can extract
it from the integral using Eq. (1.9):
t
′ ′
Cout (t) = C0 E(t )dt (1.10)
∫0
Obviating that t′ is mathematically equal to t, we see that the signal at the output
of this experiment is a cumulative function of E(t), as it evaluates the integral of
all E(t) from t = 0 to the instant “t.” This cumulative distribution is called the “F
curve” and can be directly determined from a step input.
[ ] t
C(t)
= E(t)dt = F(t) (1.11)
C0 step input ∫0
Differentiating this expression, we obtain the function of RTD, E(t):
[ ]
d C(t)
E(t) =
dt C0 step input
The determination of the E(t) using a step input is, in general, easier to carry
out experimentally than the pulse input. It has some other advantages, as in that
it is not necessary to know the total amount of tracer introduced during the
test period. In return, it has some disadvantages: (i) sometimes, it is not easy
to maintain a constant concentration of tracer in the feed; (ii) for calculating the
E(t) curve, this procedure implies the differentiation of the data, which can lead,
sometimes, to errors; (iii) as the feed should be maintained for a long period of
time, the amount of tracer required is usually very large, so if the tracer is expen-
sive, a pulse input is usually used.
Other techniques to introduce the tracer are possible: negative step (dilution of
the food), a periodic signal, or a random signal. However, they tend to be more
difficult to carry out and are discussed in the following chapters.

1.2.2 RTD Concept in Heterogeneous Systems

Just as in the homogeneous reactors, the concept of RTD is valid in heteroge-
neous systems, with the particularity that there will exist an RTD related to the
flow of each phase involved in the reaction. For example, let us think about the
flow produced in a catalytic reactor where a gaseous species is introduced on one
side, and, simultaneously, the solid catalyst is introduced on the other. Figure 1.2
presents such a situation.
The experimental techniques to measure these RTD functions are equivalent
to the ones used in one-phase reactors. The application of such measurements
is extremely important in the design of reactors. For example, let us think of a
catalyzed fluidized bed, where gas is flowing at its bottom and circulating up.
The flow of the solid is very similar to that of the ideal CSTR, as the fluidized
bed is supposed to be perfectly mixed due to the fact that the mass transfer is
8 1 Nonideal Flow

Gas out Solid E(t)

output
Gas

Gas in Solid

Solid
input Time

Figure 1.2 Two E(t) functions are needed to characterize a reactor with two moving phases.

extremely fast. Nevertheless, the gas will flow through this bed, and will present
a specific and complicated flow pattern, as we know. This pattern will depend on
different factors and usually is between the plug flow and the CSTR behavior. The
measurement of the precise RTD of both phases will give important information
for the design and scale-up of the reactors.

1.2.3 Characteristics of RTD

Going back to the use and measurement of RTD, it is important to know some
characteristics. If we consider the age of a reacting particle as the time it has
remained under the conditions in which the reaction is carried out, then E(t)
is related to the age distribution of the effluent. In fact, the E(t) is also called
function of distribution of exit ages.
The fraction of the output current that has remained in the reactor for a period
less than a given value, t, equals the sum over all times of the function E(t)⋅Δt, or
expressed continuously:
t [ ]
Fraction of output current that has remained
E(t) = = F(t)
∫0 in the reactor a time less than “t”
which represents the cumulative RTD function, F(t). Analogously, the integral of
the E(t) function from time “t” to infinity will represent the fraction of the output
current that has remained in the reactor a time higher than “t.”
This function F(t) (cumulative distribution function) can be calculated at var-
ious “t” values using the area of the curve of the representation of E(t) vs. time.
An example of curve F(t) for a tracer step input can be seen in Figure 1.3, where

Figure 1.3 Interpretation of the cumulative

distribution curve, F(t).
1.0

0.8

0.6 Step response

0.4

0.2

0
40
Time
 1.2 Residence Time Distribution (RTD) Function 9

it can be affirmed that 80% of the input tracer spends less than 40 units of time
in the reactor.

1.2.3.1 Mean Residence Time

Variables described by distribution functions are characterized by their moments.
Correspondingly, function E(t) can also be characterized by its moments. In this
sense, the mean value of the variable is equal to the first moment of the RTD
function, E(t). So, the mean residence time is defined as
∞
∫0 t ⋅ E(t)dt ∞
tm = = t ⋅ E(t)dt (1.12)
∫0
∞
E(t)dt ∫0
When studying the ideal reactors, the spatial time (𝜏 = mass of catalyst/Q) and
the average residence time (t = V ∕Q) are usually used.
We can demonstrate that, regardless of the RTD that exists in a given reactor,
ideal or not ideal, the nominal average residence time, t = V ∕Q is equal to the
mean residence time of the RTD distribution, t m , if the system is of constant den-
sity (where Q is constant). For doing this, let us consider a reactor completely
filled with a colored fluid (for example, red bromine gas); at an instant t = 0,
we start to inject yellow gas to replace the bromine filling the reactor. That is,
the reactor volume V is equal to the volume occupied by the bromine leaving
the reactor. In such a situation, during a certain time dt, the volume of bromine
that will leave the reactor is Q⋅dt, “Q” being the volumetric flow rate (considered
at constant temperature and pressure). In that case, [1 − F(t)] will represent the
fraction of the gas that has remained in the reactor for a longer time. Since only
this red gas has remained in the reactor more than “t,” the volume of bromine,
dV , leaving the reactor at time “dt” is
dV = (Q ⋅ dt)[1 − F(t)] (1.13)
If we add all the bromine that has left the reactor in the period between
0 < t < ∞, we have
∞
V = Q ⋅ [1 − F(t)] ⋅ dt (1.14)
∫0
For constant volumetric flow:
∞
V =Q [1 − F(t)] ⋅ dt
∫0
∞
V
= [1 − F(t)] ⋅ dt (1.15)
Q ∫0
For the integration of the right-hand side, we will use the integration by parts,
indicating that

u ⋅ dv = u ⋅ v − v ⋅ du
∫ ∫
Applying to Eq. (1.15):
1
V
= t[1 − F(t)]∞
0 +∫ t ⋅ dF (1.16)
Q 0
10 1 Nonideal Flow

At t = 0, the cumulative function is zero, F(t) = 0, and when t → ∞,

[1 − F(t)] = 0:
1
V
= ∞ ⋅ (1 − 1) − 0 ⋅ (1 − 0) + t ⋅ dF
Q ∫0
In this way, the first term on the right-hand side is zero, and we can write
1
V
=t= t ⋅ dF (1.17)
Q ∫0
On the other hand, since dF = E(t)dt:
∞
t= E(t) ⋅ dt (1.18)
∫0
Now, the right side of this equality is, precisely, the mean residence time, so we
can conclude that spatial time and mean residence time are equal in systems of
constant density, regardless of whether the flow is ideal or not:
t = tm (1.19)
This result is only valid for closed systems (that is, without dispersion). The
volume of the reactor can now be easily determined:
V = Q ⋅ tm (1.20)

1.2.3.2 Second and Third Moments of the RTD

When comparing RTDs, it is usual to use the moments associated with the distri-
bution instead of using the entire distribution. To this end, three are the moments
that are usually used. The first moment is the average residence time, already
defined. The second moment is calculated from the mean and is the variance
which is defined as
∞
𝜎2 = (t − tm )2 ⋅ E(t)dt (1.21)
∫0
This equation corresponds to the square of the standard deviation. The magni-
tude of this moment gives a measure of the dispersion of the distribution, in such
a way that the greater this moment, the greater the dispersion of the distribution
(see Figure 1.4).

Figure 1.4 Effect of the second

E(t) moment of the RTD in the curve
shape. Both distributions are
Low variance centered at the same tm .
High variance

tm t
 1.3 RTD in Ideal Reactors 11

E(t) E(t)
Negative Positive
skew skew

t t

Figure 1.5 Form of the RTD curve for positive and negative skewness.

The third moment is also calculated around the average and is known as skew-
ness. It is defined as
∞
1
s3 = (t − tm )3 ⋅ E(t)dt (1.22)
𝜎 3∕2 ∫0
The greatness of this moment measures the extent to which the distribution
with respect to the mean is displaced in one direction or another. A positive value
of the skewness indicates that a tail to the right is expected in the distribution,
and to the left if the skewness is negative. Figure 1.5 shows a scheme.

1.3 RTD in Ideal Reactors

1.3.1 RTD of the Batch and PFR Reactors
The RTD of these two types of reactors is the simplest we can consider, as already
mentioned. All the particles (considered very small portions of the fluid, or solid
particles if there is a solid moving in the reactor) that leave these reactors have
stayed the same time inside the reactor. The output is then infinite at a particular
time, and zero otherwise. This will produce an output function similar to that of
an arrow of infinite height and zero width but presenting an area equal to one.
This arrow appears at t = V /Q = t. Mathematically, this arrow is represented by
the Dirac delta function, in such a way that the E(t) for an PFR is
E(t) = 𝛿(t − t) (1.23)
The delta function has the following properties:
𝛿(x) = 0 when x ≠ 0
𝛿(x) = ∞ when x = 0 (1.24)
∞
𝛿(x) dx = 1 (1.25)
∫−∞
b
g(x) ⋅ 𝛿(x − t) dx = g(t) if a < t < b (1.26)
∫a
b
g(x) ⋅ 𝛿(x − t) dx = 0 if t ∉ [a, b] (1.27)
∫a
12 1 Nonideal Flow

Figure 1.6 Dirac delta function, i.e. response of an ideal

PFR to a pulse tracer input when “x” is time and the
δ(x – x0) signal to a tracer is registered. In that case, x 0 would
represent the average residence time.

0 x0 x

In this way, the variance of this distribution is zero, by the fact that all values of
(t − t m ) are zero (all the signal is obtained at t = t m ). The function can be seen in
Figure 1.6.

1.3.2 RTD of an ideal CSTR

The main characteristic of a CSTR is that the concentration of all substances
inside the reactor are identical to those concentrations in the output stream.
Imagine that an amount M0 of inert tracer is injected into a CSTR. Once the
tracer in injected, we can do a material balance for inert tracer, and we obtain
Input − Output = Accumulation
dC
0−Q⋅C =V ⋅ (1.28)
dt
Note that the input is zero (we consider the balance when the tracer has been
already injected, i.e. at t > 0). Obviously, the generation term is also zero as the
tracer is inert.
Since the reactor is perfectly agitated, C is the tracer concentration at the outlet
and also at any point of the reactor. Separating variables and integrating, taking
into account that C = C 0 for t = 0, we obtain
C(t) = C0 ⋅ exp(−t ⋅ Q∕V ) = C0 ⋅ exp(−t∕t) (1.29)
an expression that allows to obtain the concentration of the tracer in the effluent
for any instant “t.”
If we take into account the definition of E(t) and introduce the value of C(t), we
obtain the RTD of an ideal CSTR:
C(t) C0 ⋅ exp(−t∕t)
E(t) = ∞ = ∞
(1.30)
∫0 C(t) ⋅ dt ∫0 C0 ⋅ exp(−t∕t) ⋅ dt
Evaluating the integral of the denominator:
∞
C0 ⋅ exp(−t∕t) ⋅ dt = C0 ⋅ [t ⋅ exp(−t∕t)]∞
∫0 0

= −C0 ⋅ t ⋅ [exp(−∞) − exp(0)] = C0 ⋅ t

 1.3 RTD in Ideal Reactors 13

Figure 1.7 Response of a CSTR to a pulse tracer E(t)

input.

t t

Thus, the RTD is obtained for an ideal CSTR (Figure 1.7):

1
E(t) = ⋅ exp(−t∕t) (1.31)
t
We have already seen that the average residence time in a reactor is given by
V /Q or t. This relationship can now be obtained in a simpler way by applying the
definition of the average residence time of an RTD for a CSTR:
∞ ∞ ( )
1 t
tm = t ⋅ E(t) ⋅ dt = t ⋅ ⋅ exp − ⋅ dt = t (1.32)
∫0 ∫0 t t
That is to say, the nominal residence time (spatial time) t = V /Q coincides with
the mean residence time that the material remains in the reactor, t m .
To know the degree of dispersion in the reactor, we will calculate the second
moment, the variance:
∞ ( )
(t − t)2 t 2
𝜎 =
2
⋅ exp − ⋅ dt = t (1.33)
∫0 t t
That is to say, 𝜎= t. Or in other words, the standard deviation of the RTD coin-
cides with the average.

1.3.3 RTD of PFR/CSTR in Series

In agitated tank reactors, there are zones, in the vicinity of the agitator, with a
high degree of agitation and where an ideal CSTR could be a valid model. Nev-
ertheless, depending on the location of the conductions, the reaction mixture
can follow a somewhat tortuous path when entering, or when leaving the area
perfectly agitated, or in both cases. This tortuous path can be modeled as if it
behaved like a PFR. Thus, the tank-type reactor can be modeled as if it were a
CSTR in series with a PFR (Figure 1.8), and the PFR can be before or after the
CSTR. Next, we study the RTD for this type of reactor.
Consider, first, the system constituted by a CSTR followed by a PFR (Figure 1.8).
The mean residence time in the CSTR will be called tt and the mean residence
time in the PFR, t p . If we inject a tracer pulse at the entrance of the CSTR, the con-
centration at the exit of the tank will vary with time according to the expression:
C(t) = C0 ⋅ exp(−t∕t t ) (from Eq.(1.29))
14 1 Nonideal Flow

Feed

CSTR PFR

CA, T
Product

Figure 1.8 Real reactor modeled as a CSTR and a PFR in series.

E(t) Figure 1.9 RTD for a CSTR and a PFR in

series.

tp t

This output concentration will exit the PFR in series delayed by a time tp . There-
fore, the RTD of the series reactor system will be given by (Figure 1.9)
E(t) = 0 if t < t p
( )
1 (t − t p )
E(t) = ⋅ exp − if t ≥ t p (1.34)
tt tt
Note that Eq. (1.34) is the same as Eq. (1.31), but the time scaling has been
moved tp units.
Let us see now the case of a CSTR preceded by a PFR. If a signal tracer is intro-
duced into the input pulse PFR, the same signal appears at the input of the CSTR,
but with a delay of tp seconds, so the system’s RTD will be the same as when the
CSTR is the first reactor and is followed by a PFR. That is, the order in which both
reactors are placed is not important, and the resulting RTD is the same provided
that the sum of residence times in the two sections is the same.
However, this is not the only thing that should be considered; in case the reac-
tion that takes place in the system is of the second order, the conversion that
would be obtained with both dispositions would be different, as the conversion
depends on the concentration. Contrarily, a first-order reaction will produce the
same conversion both for PFR + CSTR and CSTR + PFR. This means that the
RTD is not a complete description of what happens in a reactor or reactor sys-
tem. The RTD is unique for a particular reactor; however, the reactor or reactor
system is not unique to a particular RTD. In this way, in nonideal reactors, the
RTD gives information that is not enough to characterize their behavior, and we
need more information.
 1.4 Modeling the Reactor with the RTD 15

1.4 Modeling the Reactor with the RTD

On many occasions, the flow inside a reactor is not adjusted either to the com-
plete mixture or to the plug flow, so when trying to use the RTD to predict the
conversion that we are going to obtain, we find that the ideal models so far no
longer serve us. There will, therefore, be a need to model the real reactor with
some type of combination of ideal reactors or introduce new models. In these
models, the adjustable parameter is usually evaluated on the basis of the RTD
analysis obtained with a tracer test. We classify the models according to the num-
ber of adjustable parameters that are extracted from the information provided by
the RTD.

1.4.1 Models with One Parameter: Tanks-in-series and Dispersion

Models
In these models, we use a single parameter to bear in mind the nonideal behavior
of a particular reactor. The parameter is determined by analyzing the measured
RTD in a tracer test, as mentioned before.
For modeling the nonideal behavior of CSTRs, usually a dead zone volume (V d ),
where the reaction does not take place, is used. Also, it is usual to consider the
existence of a part of the fluid that passes through the reactor in short circuit and
therefore does not react. In the case of tubular reactors, there are two models that
usually represent the flow: the tanks-in-series model and the dispersion model.
Both are one-parameter models, being the number of tanks in the first one, nt ,
and the dispersion coefficient, De , in the last one.
For both types of distributions, once the value of the parameter is known, we
will be able to calculate the conversion and or the concentrations at the output of
the reactor.
In the case of a nonideal tubular reactor, usually it is assumed that the fluid
moves in a plug flow through the reactor, so that each atom passes through the
reactor the same time, and that the velocity profile is flat and there is no axial
mixing. Both statements are false, to a greater or lesser extent, in all tubular reac-
tors. Two approaches are often used to compensate for failures in the two ideal
assumptions. In one case, the real reactor is modeled as a series of CSTRs of the
same size. In the other (dispersion model), an axial dispersion is superimposed
on the piston flow.

1.4.1.1 Tanks-in-series Model

In this model, we analyze the RTD of a particular tubular reactor to determine
the number of CSTRs in series that will present an RTD approximately the
same as the actual RTD. Next, we apply the balance equations valid for the ideal
CSTR in order to calculate the conversion. We first consider the case of three
tanks (Figure 1.10), developing the equations for the expression of the RTD, and
then generalize it for “nt ” reactors connected in series. In this way, we obtain an
equation that permits to calculate the number of tanks that best correlates the
data of the actual RTD.
16 1 Nonideal Flow

Feed

CA0

…..

CA1, V1

CAn, Vn

CA2, V2

Figure 1.10 Tanks-in-series model.

If an impulse tracer signal is injected into the first tank, the tracer fraction that
leaves the third reactor system after remaining in the system for a time between
t and t + dt is given by E(t) ⋅ dt, which can be estimated from the concentration
obtained in a pulse tracer experiment:
CA3 (t)
E(t) = ∞ (1.35)
∫0 CA3 (t) ⋅ dt
In this expression, C A3 (t) is the concentration of the tracer at the outlet of the
third reactor. Now, we must obtain how this concentration varies with time. For
a unique CSTR, the mass balance will be
V1 ⋅ dCA1
= −Q ⋅ CA1 (1.36)
dt
Integrating, we obtain the expression of the tracer concentration at the exit of
that reactor:
( ) ( )
t t
CA1 = CA0 exp −Q ⋅ = CA0 exp − (1.37)
V1 t1
Since the volumetric flow is constant Q = Q0 and that the volume of all the
reactors is the same (V 1 = V 2 = V 3 ), the average times will be identical (t1 = t2 =
t3 = t𝜄 ), t𝜄 being the residence time in each one of the reactors, not in the whole
system. Posing a balance of the tracer in the second reactor:
V2 dCA2
= Q ⋅ CA1 − Q ⋅ CA2 (1.38)
dt
 1.4 Modeling the Reactor with the RTD 17

Taking into account the expression that we have previously obtained for C A1 ,
we arrive at the differential equation:
( )
dCA2 CA2 t
+ = CA0 exp − (1.39)
dt t2 t1
which can be solved using an integration factor together with the initial condition
C A2 = 0 for t = 0:
( )
CA0 t t
CA2 = ⋅ exp − (1.40)
ti ti
Using the same procedure for the third reactor, we obtain the expression for the
tracer concentration at the exit of the third tank (and, therefore, of the system):
( )
CA0 t 2 t
CA3 = 2
⋅ exp − (1.41)
2t i ti
Substituting in the equation for the curve E(t):
( )
C3 t2 t
E(t) = = 2 ⋅ exp − (1.42)
C0 2t i ti
If we generalize for nt equal tanks in series:
( )
t nt −1 t
E(t) = nt ⋅ exp − (1.43)
(nt − 1)!t i ti
since t i = t/nt , where t is the quotient of the total volume of the system by the
volumetric flow rate Q.
Figure 1.11 shows the RTD for different CSTR numbers in series. As nt
increases, the behavior is closer to piston flow.
The number of reactors in series can be calculated from the dimensionless
variance 𝜎 2 :
∞ 2
t
𝜎2 = (t − t)2 ⋅ E(t)dt = … = (1.44)
∫0 nt

Figure 1.11 Response to a pulse nt = ∞

tracer input in function of the E(t)
nt = 10
number from tanks according to the
tanks-in-series model. In the figure, t
represents the average residence nt = 4
time of the whole system. nt = 2

t t
18 1 Nonideal Flow

That is an expression that gives us the number of tanks needed to model the
nonideal reactor as a series of nt CSTR connected in series.
Let us now calculate the conversion of a reaction in the tanks in series. If the
reaction is of the first order:
CA0
1st reactor ∶ CA1 = (1.45)
1 + t1 k
CA1 CA0
2nd reactor ∶ CA2 = = (1.46)
1 + t2 k (1 + t 1 k)(1 + t 2 k)
As all residence times are equal, t i , and the temperature is constant (k 1 = k 2 = k),
we can generalize the expression as
CA0
CAn = (1.47)
(1 + t i k)nt
Therefore, we can express the conversion as
1
XA = 1 − (1.48)
(1 + t i ⋅ k)nt
In general, the value of nt obtained from the variance is considered as a non-
integer number when calculating the conversion. In this sense, equations of the
model can be applied to fractional number of tanks. Nevertheless, if the reaction
is not of the first order, sequential molar balances must be made in each reactor
(see Example 1.3 in the following sections).

1.4.1.2 The Dispersion Model

In this model, it is considered that there is a dispersion of the material overim-
posed to the flow, and that this dispersion is governed by an expression analogous
to that of Fick for diffusion, which is superimposed on the plug flow. Thus, in
addition to the term (u⋅S) due to the flow of the mass of fluid, each component
of the mixture will be transported through any section of the reactor with an
additional rate [De ⋅S⋅(dC/dz)] due to molecular and turbulent diffusion. At first
glance, this simple model could only serve the effects of axial mixing. However,
it can be seen that it serves to compensate also for the effects of radial mixing
and those due to non-flat velocity profiles. These variations in concentration may
be due to different speeds and flow paths, as well as to molecular and turbulent
diffusion.
Let us imagine a pulse injection of tracer to a tubular reactor of section “S.” Dur-
ing the movement of the fluid through the reactor, the pulse widens and becomes
more diluted. The molar flow rate of the tracer (nT ), both by dispersion and con-
vection, is
( )
𝜕C
nT = u ⋅ S ⋅ CT + −De T ⋅ S (1.49)
𝜕z
In this expression, “z” is the spatial dimension where the fluid is moving, De
is the effective dispersion coefficient (m2 /s) and “u” is the superficial velocity.
 1.4 Modeling the Reactor with the RTD 19

Note that the term corresponding to the dispersion of the component “A” is based
on Fick’s law for diffusion. If we do an inert tracer balance in a differential volume:
In − Out = Accumulation
dCT dCT
− dnT = dV ⋅ = S ⋅ dz ⋅ (1.50)
dt dt
and, using partial derivatives:
𝜕nT 𝜕CT
− =S⋅ (1.51)
𝜕z 𝜕t
Substituting for nT (Eq. (1.49)) and dividing by the cross-section S:
𝜕 2 CT 𝜕(u ⋅ CT) 𝜕CT
De − = (1.52)
𝜕z2 𝜕z 𝜕t
If we divide by (u⋅L):
De 𝜕 2 CT 𝜕CT 𝜕CT
− = (1.53)
u ⋅ L 𝜕z 2 L ⋅ 𝜕z u ⋅ L ⋅ 𝜕t
The parameter De /uL is the so-called recipient dispersion module or the
Peclet–Bodenstein module (Bo = De /uL) that measures the degree of axial
dispersion. When this module tends to zero, the system is close to piston flow,
and when it tends to infinity (large dispersion), we have complete mixing. In the
case of a packed bed, the module would be (𝜀⋅De /u⋅dp ), where dp is the particle
diameter and 𝜀 the porosity of the bed.
Equation (1.52) is only solvable for small values of Bo number, usually Bo < 0.01.
In that case, the dispersion modifies the input signal in the reactor, but the tracer
widening does not vary in the measuring point with time, in such a way that the
boundary conditions are well known. In this situation, the expression for the RTD
curve of the reactor is
( )2
⎡ t ⎤
⎢ 1 − ⎥
1 t
E= √ ⋅ exp ⎢− ⎥ (1.54)
t ⋅ 4 ⋅ 𝜋 ⋅ Bo ⎢ 4 ⋅ Bo ⎥
⎣ ⎦
V
tm = t = (1.55)
Q
𝜎 2 = 2 ⋅ Bo (1.56)
Nevertheless, if the value of the Bodenstein module Bo is higher than 0.01, the
response of the tracer to the impulse is wide and passes through the measuring
point so slowly that it can change its form during the time of measuring. This
produces an asymmetrical E curve that, on some occasions, does not have an
analytical expression for the E curve.
In this case, the E curve also depends on what happens in the input and output
sections of the reactor vessels. We consider two cases: closed boundary condi-
tions (where there is plug flow behavior outside of the system), and open bound-
ary conditions (where the flow is not affected when passing through the system).
20 1 Nonideal Flow

Closed recipient Open recipient

– + – + – + – +
z=0 z=0 z=L z=L z=0 z=0 z=L z=L

De = 0 De>0 De = 0 De>0 (same value of De>0)

Plug flow

z=0 z=L z=0 z=L

Flat profile of velocity Fluctuations due to dispersion

Figure 1.12 Effect of the dispersion on the velocity profile.

For simplifying all possibilities, let us consider only two cases: the closed–closed
containers in which there is neither dispersion nor radial variation of the con-
centration, both upstream and downstream of the reaction zone; and open–open
containers in which there is dispersion both before and after the reaction zone.
Both cases are shown in Figure 1.12, where it is observed that the fluctuations
of the concentration due to dispersion overlap the piston flow velocity profile. A
closed–open container would have no dispersion at the entrance but only at the
exit of the reaction zone.

Boundary Conditions for a Closed–Closed Vessel (Bo > 0.01) In this case, immediately
before the reactor entrance zone (z = 0− ) and immediately after the exit zone
(z = L+ ), we have piston flow (without dispersion). However, between z = 0+ and
z = L− , there is dispersion and convection by the flow movement. The corre-
sponding boundary condition at the input is
nT (0− , t) = nT (0+ , t) (1.57)
Substituting the value nT of at each side, we obtain
[ ]
𝜕CT
uSCT (0 , t) = −SDe
−
+ uSCT (0+ , t) (1.58)
𝜕z z=0+
Taking into account that at the input C T (0− , t) = C T0 (known concentration):
[ ]
De 𝜕CT
CT0 = − + CT (0+ , t) (1.59)
u 𝜕z z=0+
Also, at the exit of the reactor considered, we can write C T (L− , t) = C T (L+ , t),
this being later the measured exit concentration. In this way, when z = L:
CT (L− , t) = CT (L+ , t)
( )
𝜕CT C (L− , t) − CT (L+ , t)
= T =0 (1.60)
𝜕z z=L 𝜕z
The combination of Eqs. (1.59) and (1.60) are known as the Danckwerts’ bound-
ary conditions.
On the other hand, the initial condition of the reactor at t = 0 is
t = 0, z > 0 CT (0+ , 0) = 0 (1.61)
 1.4 Modeling the Reactor with the RTD 21

The injected tracer mass (M0 ) is given by

∞
M0 = u ⋅ S ⋅ CT (0− , t) ⋅ dt (1.62)
∫0
In the closed–closed case, we do not have an analytical expression for the
E curve, but the curve can be calculated by numerical methods and, also, we
can calculate exactly its mean and variance. Bischoff and Levenspiel (in 1963)
found the following relationships for the mean residence time and variance for
this case:
t = tm
( )2 [ ( )]
𝜎 1
= 2 ⋅ Bo − 2 ⋅ Bo2 ⋅ 1 − exp − (1.63)
tm Bo
Experimentally, the dispersion module can be calculated from the values of t m
and 𝜎 2 obtained from the RTD, substituting in the previous expression.

Boundary Conditions for an Open–Open Container (Bo > 0.01) These conditions
would be applicable in the case of a packed bed in which the tracer was injected
at a point downstream of the inlet, a distance around two to three times the
diameter, and whose concentration was measured at a certain distance before
the exit. A solution of the differential equation (Eq. (1.52)) could be obtained in
the case of a pulse injection.
For an open–open system, the boundary condition at the input is
nT (0− , t) = nT (0+ , t) (1.64)
Note that the expression is the same as that obtained in the previous case. If
the dispersion coefficient is the same at the entrance as in the reaction zone, we
will have
[ ] [ ]
𝜕CT 𝜕CT
−De + u ⋅ CT (0− , t) = −De + u ⋅ CT (0+ , t) (1.65)
𝜕z z=0− 𝜕z z=0+
As we can imagine, the derivatives at z = 0+ and z = 0− are the same, as no
discontinuity is included in the model, so:
CT (0− , t) = CT (0+ , t) (1.66)
while on the exit:
[ ] [ ]
𝜕CT 𝜕CT
−De + u ⋅ CT (L , t) = −De
−
+ u ⋅ CT (L+ , t) (1.67)
𝜕z z=L− 𝜕z z=L+
CT (L− , t) = CT (L+ , t) (1.68)
In addition to these boundary conditions, many other modifications may occur.
For example, the dispersion coefficient can have different values in each of the
three regions (before the entrance, in the reaction zone, and after the output)
and/or the tracer can be injected at a point other than z = 0. However, we consider
only the case that the dispersion coefficient is the same for any value of z and that
the pulse tracer is injected at the point z = 0 at time t = 0.
22 1 Nonideal Flow

In the open–open case, there exists an analytical solution of the differential

equation, with an expression of the E curve that is not much complicated:
( )2
⎡ t ⎤
⎢ 1 − ⎥
1 t
E= √ ⋅ exp ⎢− ⎥ (1.69)
4 ⋅ 𝜋 ⋅ Bo ⋅ t∕t ⎢ 4 ⋅ t ⋅ Bo∕t ⎥
⎣ ⎦
The corresponding average residence time is
tm = t ⋅ (1 + 2 ⋅ Bo) (1.70)
where t is based on the volume (and flow rate) between z = 0 and z = L; that is,
the reactor volume measured with a calibrated apparatus. Note that, as a result of
the previous equations, the average residence time in the open system is longer
than the one corresponding to the closed system. The variance will be
( )2
𝜎
= 2 ⋅ Bo + 8 ⋅ Bo2 (1.71)
tm
In the next two chapters, we discuss the numerical solution to these partial
differential equations systems and we present the problem of calculating the con-
version in systems with dispersion.

1.4.2 Models with Two Parameters

1.4.2.1 Two CSTR with Exchange of Matter
Consider the case that there is a strongly agitated region near the agitator of
an CSTR. Nevertheless, outside this region, the agitation is lower (Figure 1.13).
Both regions have an important exchange of material. We also consider that the
entrance and exit pipes are connected to the zone of greatest agitation. Each zone
will be modeled as a CSTR, both being connected and there is material transfer
between them.
If we propose a molar balance on the tracer, with a pulse injected at t = 0, for
each of the tanks, we obtain
Accumulation = Input − Output (there is no tracer generation)
dCT1
V1 = Q1 CT2 − (Q0 CT1 + Q1 CT1 ) (in the first tank) (1.72)
dt
Feed
Q0

Q1
CSTR CSTR
CA2, V2
CA1, V1 Q1

Product CA1, Q0

Figure 1.13 Real reactor and modeling using two CSTR.

 1.4 Modeling the Reactor with the RTD 23

dCT2
V2 = Q1 CT1 − Q1 CT2 (in the second tank) (1.73)
dt
where C T1 and C T2 are, respectively, tracer concentrations in both reactors. These
two differential equations are coupled and should be solved simultaneously.
In this model, the two adjustable parameters are the flow rate exchanged (Q1 )
and the volume of the most agitated region (V 1 ). Remember that the measured
volume (V ) is the sum of V 1 and V 2 . We will call 𝛽 the fraction of the total flow
that is transferred between both reactors:
Q1 = 𝛽 ⋅ Q0 (1.74)
and 𝛼 the fraction of the total volume that corresponds to the most agitated
area:
V1 = 𝛼 ⋅ V → V2 = (1 − 𝛼) ⋅ V (1.75)
On the other hand, the average time (t) is given by the quotient V /Q0 .
The initial conditions (t = 0) for this model are (i) C T1 = (C T1 )0 , and (ii)
(C T2 )0 = 0
An analytical solution is possible in this case, and is as follows:
( ) ( )
[ ] (𝛼m1 + 𝛽 + 1) exp
m2 t
− (𝛼m2 + 𝛽 + 1) exp
m1 t
CT1 t t
=
(CT1 )0 pulse 𝛼(m1 − m2 )
(1.76)
being:

[]⎡ √
⎤
1−𝛼+𝛽 ⎢ 4𝛼𝛽(1 − 𝛼) ⎥
m1 , m2 = −1 ± 1 − (1.77)
2𝛼(1 − 𝛼) ⎢ (1 − 𝛼 + 𝛽)2 ⎥
⎣ ⎦
However, for more complicated models, an approximate solution would be nec-
essary.
Equation (1.76) shows that, if tank 1 is small compared to 2 (𝛼 small) and the
transfer speed between both reactors is small (𝛽 small), the second exponential
term tends to 1 during the first part of the response to an impulse injection. Dur-
ing the second part, the first exponential term tends to 0. If we represent the
logarithm of the tracer concentration vs. time, the response curve will tend to a
straight line at both ends of the curve and the parameters will be obtained from
the slopes (m1 for t → ∞ and m2 for t → 0) and the cut points of both lines (for
t → ∞, the cut point is – {𝛼m2 + 𝛽 + 1}/𝛼{m1 − m2 }).

1.4.2.2 CSTR with Dead Volume and Short Circuit

In this case, the real CSTR is modeled as the combination of an ideal CSTR of
volume V t , a dead zone of volume V d , and a bypass (short circuit) of volumetric
flow rate Qb (Figure 1.14). Using a tracer injection, we will calculate the parame-
ters of the model V t and Qt , since the total volume and the volumetric flow are
known.
24 1 Nonideal Flow

Feed
CT0, Q0 CT0, Qb

CT0, Qt

Bypass CSTR Vd
Dead volume CTs, Vt
Q0 = Qt + Qb
Product CTs, Qt
CT

Figure 1.14 Real reactor and modeling using a single CSTR with dead volume and short
circuit.

In this case, the derivation of the equations is simpler if we consider the injec-
tion of a tracer in positive step. Let us use the scheme in Figure 1.14; the balance
in nonstationary regime (at t > 0, some tracer is still entering the system) of non-
reactive tracer in the volume of reactor V t , is
dMTs dC
Qt ⋅ CT0 − Qt ⋅ CTs = = Vt Ts (1.78)
dt dt
Remembering that for a positive step entry it is fulfilled that:
t < 0 → CT = 0
t ≥ 0 → CT = CT0
The tracer balance at the point of union of both currents will be
Qb ⋅ CT0 + Qt ⋅ CTs
CT = (1.79)
Q0
If we define:
Vt = 𝛼V and Qb = 𝛽Q0 , with t = V ∕Q0
Integrating and replacing in the expression of tracer balance in the reactor,
we get
[ ( )]
CTs 1−𝛽 t
= 1 − exp − (1.80)
CT0 𝛼 t
So, the expression of the tracer concentration that leaves the system will be
[ ( )]
CT 1−𝛽 t
= 1 − (1 − 𝛽) ⋅ exp − (1.81)
CT0 𝛼 t
If we reorder the equation, we can obtain the parameters of the model
(Qt and V t or similarly 𝛼 and 𝛽) from a plot of the tracer concentration at the
output as a function of time. Representing ln[C T0 /(C T0 − C T )] vs. time, if the
model is correct, a straight line of slope (1 − 𝛽)/t should be obtained 𝛼 and an
ordinate in the origin of value ln[1/(1 − 𝛽)].
In order to have in mind the possible RTD curves obtained with this model,
Figure 1.15 shows three curves corresponding to three different cases: the ideal
 1.5 Other Models of Real Reactors Using CSTR and PFR 25

0.4 α = 0.5; β = 0.3 α = 0.5; β = 0.1 Ideal CSTR

0.3

0.2

0.1

0
0 5 10 15 20

Figure 1.15 E(t) predicted by this model with different values of the parameters, and the one
for ideal CSTR for comparison.

CSTR, a reactor where 𝛼 = 0.5 and 𝛽 = 0.3 and the other reactor with 𝛼 = 0.5 and
𝛽 = 0.1; all of them with the same average residence time of five minutes. As we
can see, the differences are quite small and, in practice, it is difficult to affirm with-
out a little more information if the system has or not a dead volume and/or bypass.

1.5 Other Models of Real Reactors Using CSTR and PFR

So far, we have discussed about various reactor models. All of them are based on
physical observations, which, in almost all the stirred tanks, are based on the exis-
tence of a well-mixed zone in the proximity of the agitator, usually represented
by a CSTR. The region out of this well-agitated area can be modeled in various
ways. The simplest form is using a model implying a CSTR connected to a dead
zone; if it is suspected that some of the feed to the reactor may short out, a bypass
current is added.
When the models do not satisfactorily represent the deviations of the ideal flow,
we have to try more complicated models. In these models, it is assumed that the
real reactor is constituted by a series of regions (flow in piston, flow dispersed in
piston, flow in complete mixture, dead volumes) interconnected with each other
in different ways (flow in bypass, with recirculation or cross-flow).
The simple types of these models are shown in Figure 1.16, where the form of
the tracer response, in terms of E or F curves, for different models can be seen.

Example 1.1 Tubular Reactor

A tubular reactor was designed in order to obtain a conversion of 98% and process
0.03 m3 /s. The reaction is a first-order irreversible isomerization. The reactor is
3 m long, with a cross-sectional area of 0.1 m2 . In the newly constructed reactor, a
tracer pulse test gave the following data: t m = 10 s and 𝜎 2 = 65 s2 . What conversion
can be expected in the real reactor?
26 1 Nonideal Flow

Q Vp Q E(t) F(t)

Vd
Vp t Vs t
t= t=
Q Q

Q Vs Q E(t) F(t)

Vd
Vs t Vs t
t= t=
Q Q

Q1
Q Vp Q E(t) F(t)
Q2
t expected

Vp t Vp t
t= t=
Q1 Q1

Q1
Q Vs Q E(t) F(t)
Q2
t expected

Vs t Vs t
t= t=
Q1 Q1

Q1 Vp1

E(t) F(t)
Q Q
t expected
Q2 Vp2

Vp2 Vp1 t Vp2 Vp1 t

t2 = Q2
t1 = Q1
t2 = Q2
t1 = Q1

(1 + r)Q
E(t) F(t)
Q Vp Q
rQ

Figure 1.16 Combinations of ideal reactors used to model real reactors.

 1.5 Other Models of Real Reactors Using CSTR and PFR 27

Solution
During the design of the reactor, a plug flow must be assumed. In this case, the
molar balance of the reacting species “A,” for a first-order reaction, gives
XA = 1 − exp(−kt)
In the present case:
V 0.1 ⋅ 3
t= = = 10 s
Q 0.03
Coinciding with the value of t m . From the previous equations: k = 0.39 s−1
Assuming a dispersion model for the real reactor, Eq. (1.63) is fulfilled, so:
( )2 [ ( )]
𝜎 1
= 2 ⋅ Bo − 2 ⋅ Bo2 ⋅ 1 − exp −
tm Bo
Iterating, we can find the value of the dispersion module: Bo = (De /u⋅L) = 0.667
We have not seen in the previous sections the equations for the balances in
reacting systems. Let us see this now. If we consider a tubular reactor in which
we simultaneously have dispersion and reaction and we can do a molar balance
of component A, in a range Δz of the reactor:
Input − Output + Generation = Accumulation

nA − (nA + dnA ) + rA dV = 0

1 dnA
− + rA = 0 (1.82)
S dz
S ⋅ dz = dV
Combining this expression with the molar flow of substance A:
𝛿C
nA = −De S A + uSCA (1.83)
𝛿z
we obtain a differential equation of the second order:
De d2 CA dCA rA
− + =0 (1.84)
u dz2 dz u
which is only linear when the reaction rate is of order 0 or 1.
When the kinetics is of the first order (rA = −kC A ), the following expression is
obtained:
De d2 CA dCA kCA
− − =0 (1.85)
u dz2 dz u
which considers the flow, dispersion, and reaction. The solution of this second-
order differential equation can be done analytically. If we consider a closed–
closed system, we will apply the Danckwerts’ boundary conditions at the input
and at the exit of the reactor:
• Taking into account that at the input C A (0− ,t) = C A0 (known concentration).
D 𝜕CA ||
CA0 = − e ⋅ + CA (0+ , t)
u 𝜕z ||z=0+
28 1 Nonideal Flow

• At the exit of the reactor considered, we can write C A (L− , t) = C A (L+ , t), this
being later the measured exit concentration. In this way, when z = L:
( )
𝜕CA
=0
𝜕z z=L

Finally, the solution for the conversion in the reactor can be expressed by
( )
4a ⋅ exp 21uL
De
1 − XA = ( ) ( ) (1.86)
(1 + a)2 ⋅ exp 2 D − (1 − a)2 ⋅ exp − 2auL
auL
D
e e

where a = [1 + 4(t ⋅ k) ⋅ (De ∕uL)] . This expression allows knowing the conver-
1∕2

sion that would be obtained for a first-order reaction to be carried out in a tubular
reactor or in a bed reactor packed with dispersion.
For the system of the example, the value of “a” results to be 3.376 and the con-
version in the real reactor is 0.88

Example 1.2 Reaction in a Complex System

The second-order reaction 2A → B is going to be carried out in a CSTR showing
both a short circuit and a stagnant region. The tracer concentration at the outlet
of this reactor is shown in the table, when a step input with initial concentration
of tracer equivalent to 10 mg/l was used.

Time (min) 0 2 4 6 7 8 10 12 14 16 18 20
C T (mol/l) 3.0 5.3 7.2 8.0 8.3 8.6 9.2 9.7 9.7 9.8 9.9 10

The measured volume of the reactor is 1 m3 and the flow rate to the reactor is
0.1 m3 /min. The reaction rate constant is 150 l/(kmol min). The feeding contains
a concentration of A at the input of 2 kmol/m3 . Calculate the conversion that can
be expected in this reactor.

Solution
Let us first have a look of the experimental results. The graph C T -time corre-
sponding o the data is shown in Figure 1.17a.
As we have seen, Eq. (1.81) shows that the tracer concentration for a step input
in this model can be expressed by
[ ( )]
CT 1−𝛽 t
= 1 − (1 − 𝛽) ⋅ exp − (1.87)
CT0 𝛼 t
From the data in the table, it is easy to calculate the slope and intercept
( of the)
representation ln(C T /(C T0 − C T )) vs. t. The slope would correspond to − 1−𝛽𝛼t
and the intercept to ln(1/(1 − 𝛽)). In Figure 1.17b, we can see the corresponding
straight line. Taking into account an intercept of −0.4549 and a slope of 0.2875,
as solution of the model we find 𝛼 = 0.3 and 𝛽 = 0.3 bearing in mind that the
1
experimental value of average residence time t = VQ = 0.1 = 10 minutes should
be used as it was defined for the whole system.
 1.5 Other Models of Real Reactors Using CSTR and PFR 29

12.0

10.0
Concentration (mg/l)

8.0

6.0
Correlation Experimental

4.0

2.0

0.0
0 5 10 15 20 25 30
(a) Time (min)

ln(CT /(CT0 – CT)) vs. time

5

3
y = 0.2875x – 0.4549
2 R2 = 0.9804

–1

–2
0 2 4 6 8 10 12 14 16 18 20
(b) Time (min)

Figure 1.17 (a) Concentration of tracer obtained in a step input experiment and (b)
calculation of the parameters.

For the system with reaction, let us see how to obtain the conversion with this
model in the case of the first-order reaction (not in this case, but will be useful).
For a first-order reaction, the molar balance of A in the reactor where the reaction
takes place (V t ) gives
Qt CA0 − Qt CAs − kCAs Vt = 0
CA0 (1 − 𝛽)Q0
CAs =
(1 − 𝛽)Q0 + 𝛼Vk
If we do a reagent balance A in the point where the short circuit current and
the output current of the reactor are mixed, we will obtain
CA0 Qb + CAs Qt = CA (Qb + Qt )
30 1 Nonideal Flow

Clearing C A , the concentration at the exit of the system:

Qb CA0 + CAs Qt
CA = (1.88)
Q0
So, finally:
CA (1 − 𝛽)2
= 1 − XA = 𝛽 + (1.89)
CA0 (1 − 𝛽) + 𝛼tk
the expression that allows to calculate the conversion based on the parameters of
the model.
As in Example 1.2, the reaction is second order, we can write:
2
Qt CA0 − Qt CAs − kCAs Vt = 0 (1.90)
Rearranging:
Vt ⋅ k
⋅ CAs
2
+ CAs − CA0 = 0 (1.91)
Qt
Let us be aware that the flow rate and volume in the previous equations are
those in the well-stirred part of the reactor, in such a way that:
Vt 𝛼V 𝛼
= = t
Qt (1 − 𝛽)Q0 (1 − 𝛽)
and clearing, the value of C As results to be
√
𝛼
−1 + 1 + 4 1−𝛽 tkCA0
CAs = 𝛼
(1.92)
2 1−𝛽 tk

Using the data in the example, the concentration at the exit of the CSTR is
0.927 kmol/m3 . The value of C A at the exit of the system can be calculated from
Eq. (1.80). Finally, in the example, C A = 1.249 kmol/m3

Example 1.3 Second-order Reaction in a Series of CSTRs

An irreversible second-order reaction occurs in an isothermal, but not ideal,
CSTR. The reactor volume is 1000 l and the flow velocity of the reagent stream
is 1 l/s. At reactor temperature, k = 0.005 l/(mol s). The concentration of A in the
feed stream is 1 mol/l. The DTR is obtained by a tracer test in this reactor at the
desired feed rate and reaction temperature. Calculate the conversion that can be
obtained with the tanks-in-series model.
RTD data obtained:

t (s) 0 5 10 25 40 70 100 175 250 325

E(t) 9.712 2.206 6.411 0.000 0.000 0.000 0.000 0.000 0.000 9.712
(s−1 ) 18⋅10−6 86⋅10−5 68⋅10−5 1090 4 199 787 286 936 476 181 619 435 719 564 18⋅10−6
t (s) 400 700 1000 2 500 4 000 7 000 10 000 15 000 20 000
E(t) 0.000 0.000 0.000 6.954 4.508 1.200 2.505 6.80 1.60
(s−1 ) 782 749 791 834 627 888 06⋅10− 5 28⋅10−6 36⋅10−8 25⋅10−11 646⋅10−16 549⋅10−20
 1.5 Other Models of Real Reactors Using CSTR and PFR 31

Solution
First of all, we can plot the data obtained with the tracer.

9,00E–04

8,00E–04

7,00E–04

6,00E–04

5,00E–04
E(t) s–1

4,00E–04

3,00E–04

2,00E–04

1,00E–04

0,00E+00
0 5000 10 000 15 000 20 000 25 000
Time (s)

As we can see, the curve is similar to a CSTR, but small differences can be
accounted with the use of the tanks-in-series model. From the data, it is easy to
calculate:
∞
tm = t ⋅ E(t) ⋅ dt = 1129 s
∫0
∞
𝜎2 = (t − tm )2 ⋅ E(t) ⋅ dt = 5.420 ⋅ 105 s2
∫0
We can calculate the number of tanks in the model, that is, nt = t m 2 /𝜎 2 = 2.221
tanks.
We have seen in the previous sections the equations for the balances in react-
ing systems or first order. In that case, Eq. (1.48) with nt = 2.221 would give us
the expected conversion. Let us see what occurs if the system is nth order. If
we consider a single tank reactor (of volume V 1 ) we can do a molar balance of
component A:
V1 C − CA C − CA1
= A0 = A0 (1.93)
nA0 CA0 (rA ) CA0 (k ⋅ CA1
n
)
If the system has a constant density:
V C −C
t1 = = A0 n A1 (1.94)
Q (k ⋅ CA1 )
CA0
CA1 = (1.95)
n−1
(1 + kt 1 CA1 )
As we can check, the concentration at the exit can be calculated from this non-
linear equation, and an approximation can be used.
32 1 Nonideal Flow

In a second reactor connected in series with this, we would have

CA0
CA2 = (1.96)
n−1 n−1
(1 + kt 1 CA1 )(1 + kt 2 CA2 )
For nt tanks in series, the product in the denominator would have nt factors
and the intermediate concentrations C A1 , C A2 , … are needed.
In the present example, a value of nt = 2.221 tanks would represent the system,
but it is not possible to do the balance in a part of a reactor. In such situations, the
best solution is to take the nearest integer to nt or take both limits, in the present
case nt = 2 and nt = 3.
Assuming two tanks t i = t∕2, i.e. all tanks of the same volume, we have
t i = 500 s
We can now calculate C A1 using (Eq. (1.95)) with the corresponding values of k
and n. Easily, the equation gives C A1 = 0.461 mol/l. If a second reactor is consid-
ered, (Eq. (1.96)) would give us C A2 = 0.274 mol/l, that is, a conversion of 0.723.
On the other hand, if three tanks are considered, the residence time in each
tank is 333.3 s and the solution C A1 = 0.533 mol/l, C A2 = 0.326 mol/l, and
C A3 = 0.245 mol/l. The final conversion would be 0.754.

Example 1.4 Nonideal CSTR

In an isothermal nonideal CSTR, an irreversible second-order reaction is carried
out. The values of flow rate, volume, and rate constant are known, and the RTDs
has been measured with a tracer. Indicate the equations necessary to determine
the concentration at the output of the reactor as a function of the input flow (or
average residence time), if the reactor is modeled: (a) as three equal CSTRs in
series, or (b) as three equal CSTRs in parallel.

Solution

(a) If the three reactors are connected in series, the equations just discussed in
previous example are valid, so the concentration at the exit of each reactor
will be given by Eqs. (1.95) and (1.96). The one corresponding to the third
reactor is equivalent.
(b) On the contrary, if the reactors are connected in parallel, each of them will
have a flow rate equal to 1/3 of the total. The concentration at the exit of the
reactors will be the same in all three. And the mix of three currents of the
same concentration in the same proportion, will give a current with exactly
the same concentration, calculated by Eq. (1.95) bearing in mind the volume
and the flow rate at each reactor.

Bibliography
Aris, R. (1969). Elementary Chemical Reactor Analysis. Englewood Cliffs: Prentice
Hall.
Bourne, J. (1999). Turbulent mixing and Chemical Reactions. New York: Wiley.
 Bibliography 33

Carberry, J.J. and Varma, J. (eds.) (1987). Chemical Reaction and Reactor
Engineering. New York: Marcel Dekker.
Coulson, J.M. and Richardson, J.F. (1979). Chemical Engineering III. Chemical
Reactor Design, 2a ed. Oxford: Pergamon Press.
Fogler, H.S. (1998). Elements of Chemical Reaction Engineering, 3a ed. New Jersey:
Prentice Hall.
Holland, C.D. and Anthony, R.G. (1992). Fundamentals of Chemical Reaction
Engineering, 2a ed. Englewood Cliffs: Prentice-Hall.
Kayode Coker, A. (2001). Modelling of Chemical Kinetics and Reactor Design.
Elsevier.
Levenspiel, O. (1999). Chemical Reaction Engineering, 3rde. New York: Wiley.
Levenspiel, O. (1962). Chemical Reaction Engineering: An Introduction to the Design
of Chemical Reactors. New York: Wiley.
Levenspiel, O. (1979). The Chemical Reactor Omnibook. OSU. Bookstores, Corvalis.
Westerterp, K.R., Van Swaaij, W.P.M., and Beenackers, A. (1984). Chemical Reaction
Design and Operation, 2nde. Chichester: Wiley.