_ _~~8

Nonideal Flow
in Reactors

8.1 I Introduction
In Chapter 3, steady-state, isothermal ideal reactors were described in the context
of their use to acquire kinetic data. In practice, conditions in a reactor can be quite
different than the ideal requirements used for defining reaction rates. For example,
a real reactor may have nonuniform flow patterns that do not conform to the ideal
PFR or CSTR mixing patterns because of comers, baffles, nonuniform catalyst pack-
ings, etc. Additionally, few real reactors are operated at isothermal conditions; rather
they may be adiabatic or nonisothermal. In this chapter, techniques to handle non-
ideal mixing patterns are outlined. Although most of the discussion will center
around common reactor types found in the petrochemicals industries, the analyses
presented can be employed to reacting systems in general (e.g., atmospheric chem-
istry, metabolic processes in living organisms, and chemical vapor deposition for
microelectronics fabrication). The following example illustrates how the flow pat-
tern within the same reaction vessel can influence the reaction behavior.

EXAMPLE 8.1.1 I
In order to approach ideal PFR behavior, the flow must be turbulent. For example, with an
open tube, the Reynolds number must be greater than 2100 for turbulence to occur. This flow
regime is attainable in many practical situations. However, for laboratory reactors conduct-
ing liquid-phase reactions, high flow rates may not be achievable. In this case, laminar flow
will occur. Calculate the mean outlet concentration of a species A undergoing a first-order
reaction in a tubular reactor with laminar flow and compare the value to that obtained in a
PFR when (kL)/u = I (u average linear flow velocity).

• Answer
The material balance on a PFR reactor accomplishing a first-order reaction at constant
density is:

260
 CHAPTER 8 Nonideal Flow in Beacto,.wrs~ --,,2..6.......
1

~ -
:

li(i',) = 0

\ V Parabolic velocity
distribution li(i')

\ I V
\ I I
,,~ v
V
li(O) = 2u

L-.. i -
Figure 8.1.1 I Schematic representation of laminar
velocity profile in a circular tube.

uA c dC A dC
---- u--A
Ac dz dz

Integration of this equation with CA = C~ at the entrance of the reactor (z 0) gives:

kz
U

For laminar flow:

u(1') = 2U[ I Ci1'}]

where 1', is the radius of the tubular reactor (see Figure 8.U).
The material balance on a laminar-flow reactor with negligible mass diffusion (discussed
later in this chapter) is:

__ C!C 4
u(r) - ' -kCA
C!Z

Since u(1') is not a function of z, this equation can be solved to give:

k::
CA (1')' = COA expr I
L

To obtain the mean concentration, must be integrated over the radial dimension as
follows:
262 CHAPTER 8 Nonideal Flow in Reactors

(I
·0
CA Cr)u(r)2r.rdr

u(r)2r.rdr

Thus, the mean outlet concentration of A, C~, can be obtained by evaluating CA at z L. For
(kL)/u I the outlet value of CA from the PFR, C~, is 0.368 C~ while for the laminar-flow re-
actor C~ 0.443 C~. Thus, the deviation from PFR behavior can be observed in the outlet con-
version of A: 63.2 percent for the PFR versus 55.7 percent for the laminar-flow reactor.

8.2 I Residence Time Distribution (RTD)

In Chapter 3, it was stated that the ideal PFR and CSTR are the theoretical limits
of fluid mixing in that they have no mixing and complete mixing, respectively. Al-
though these two flow behaviors can be easily described, flow fields that deviate
from these limits are extremely complex and become impractical to completely
model. However, it is often not necessary to know the details of the entire flow field
but rather only how long fluid elements reside in the reactor (i.e., the distribution
of residence times). This information can be used as a diagnostic tool to ascertain
flow characteristics of a particular reactor.
The "age" of a fluid element is defined as the time it has resided within the reac-
tor. The concept of a fluid element being a small volume relative to the size of the re-
actor yet sufficiently large to exhibit continuous properties such as density and con-
centration was first put forth by Danckwerts in 1953. Consider the following experiment:
a tracer (could be a particular chemical or radioactive species) is injected into a reac-
tor, and the outlet stream is monitored as a function of time. The results of these ex-
periments for an ideal PFR and CSTR are illustrated in Figure 8.2.1. If an impulse is
injected into a PFR, an impulse will appear in the outlet because there is no fluid mix-
ing. The pulse will appear at a time t1 = to + T, where T is the space time (T = Vjv).
However, with the CSTR, the pulse emerges as an exponential decay in tracer con-
centration, since there is an exponential distribution in residence times [see Equation
(3.3.11)]. For all nonideal reactors, the results must lie between these two limiting cases.
In order to analyze the residence time distribution of the fluid in a reactor the
following relationships have been developed. Fluid elements may require differing
lengths of time to travel through the reactor. The distribution of the exit times, de-
fined as the E(t) curve, is the residence time distribution (RTD) of the fluid. The
exit concentration of a tracer species C(t) can be used to define E(t). That is:

E(t) = coC(t) (8.2.1 )

fo
C(t)df

such that:

(8.2.2)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _~C~H~A~P~T.LOOEoJRI_>8~JJNQDidealFlow in Beac1oJjr"'s'-- "'2..
6t>;t3

Inlet Outlet

.g" .g"
S "
!:J
" g"
OJ OJ
U

U"
o o
U

Time
(a)

Inlet Outlet

.g" .g"
S E
§ "
OJ
U

"
U
o §
U

Time Time
(b)

Inlet Outlet

~I

JI~ '----r'1' - - - - - -
to
Time Time
(c)

Figure 8.2.1 I
Concentrations of tracer species using an impulse input.
(a) PFR (tl = to + r). (b) CSTR. (c) Nonideal reactor.
264 CHAPTER 8 Nonideal Flow in Reactors

With this definition, the fraction of the exit stream that has residence time (age)
between t and t + dt is:

E(t)dt (8.2.3)

while the fraction of fluid in the exit stream with age less than t] is:
Ii

I o
E(t)dt (8.2.4)

EXAMPLE 8.2.1 I
Calculate the RTD of a perfectly mixed reactor using an impulse of n moles of a tracer.

• Answer
The impulse can be described by the Dirac delta function, that is:

o(t

such that:

The unsteady-state mass balance for a CSTR is:

dC
V =no(t)-vC
dt
accumulation input output

where to in the Dirac delta is set to zero and:

fOO o(t)dt = 1
-00

Integration of this differential equation with C (0) 0 gives:

(a) r dC
dt
+C = (n) -
v
o(t)

,- (n)Ilo(t) -
-j -
(b)
Io dlCexp(tjr)J = ~ -r-exp(t/r)dt 0

(c) C exp(t/r) I' (n)Ilo(t) exp(t/r)dt

= -
10 v 0 r
 CHAPTER 8 Nonideal Flow in Reactors 265

(d) C(t) exp(tI7) - C(O) exp(O) = (~)r5~) exp(1/7)d1

(e) C(t) = f
(~) exp( -tI7) o~) exp(l/7)d1
(f) another property of the Dirac delta function is:

f')
-00
o(t to)f(t)dt f(t o)

C(t)
n) exp( -tl7) exp(O/7)
(g) ( -v 7

n) exp( - t 17)
(h) C(t) = ( -
V 7

(i) E(t) = C(t)

(~) looexp (; 1/7) d1

~rJ-t/T)
7
(j) E(t)=----
-exp( -tI7) I;;"

exp( -tI7)
(k) E(t)
7

Thus, for a perfectly mixed reactor (or often called completely backmixed), the RTD is an
exponential curve.

VIGNETTE 8.2.1
266 CHAPTER 8 Nonideal Flow in Reactors

Two types of tracer experiments are commonly employed and they are the in-
put of a pulse or a step function. Figure 8.2.1 illustrates the exit concentration
curves and thus the shape of the E(t)-curves (same shape as exit concentration
curve) for an impulse input. Figure 8.2.2 shows the exit concentration for a step
input of tracer. The E(t)-curve for this case is related to the time derivative of the
exit concentration.
By knowing the E(t)-curve, the mean residence time can be obtained and is:

fXltE(l)dt
o
(t) = --:-:-::---- (8.2.5)
fXlE(t)dt
o

EXAMPLE 8.2.2 I
Calculate the mean residence time for a CSTR.

• Answer
The exit concentration profile from a step decrease in the inlet concentration is provided in
Equation (3.3.11) and using this function to calculate the E(t)-curve gives:

exp( -tiT)
E(t)
T

Therefore application of Equation (8.2.5) to this E(t)-curve yields the following expression:

(t) =.! (OOtexp(-tIT)dt

do
Since:

1
o
00

x exp( - x)dx = 1

oo

(t) = ~ r T2 (tjT)exp(-tIT)d(tjT) = T
. Jo

As was shown in Chapter 3, the mean exit time of any reactor is the space time, T.

The RTD curve can be used as a diagnostic tool for ascertaining features of
flow patterns in reactors. These include the possibilities of bypassing and/or re-
gions of stagnant fluid (i.e., dead space). Since these maldistributions can cause
unpredictable conversions in reactors, they are usually detrimental to reactor op-
eration. Thus, experiments to determine RTD curves can often point to problems
and suggest solutions, for example, adding or deleting baffles and repacking of
catalyst particles.
 CHAPTER 8 hJonideal Flow in Reactors 267

Inlet Outlet

c:
.'Eo" .8
g
c: c:
"u
§
"uc:
o
U U

Time
(a)

Inlet Outlet

Time Time
(b)

Inlet Outlet

J J
'ILL
r - - I-

.g
E
c:
u "u
8 §
U

Time Time
(c)

Figure 8.2.2 I
Concentrations of tracer species using a step input.
(a) PFR (tl to + 7). (b) CSTR. (c) Nonideal reactor.
 268 CHAPTER 8 Nonideal Flow in Reactors

EXAMPLE 8.2.3 I
In Section 3.5 recycle reactors and particularly a Berty reactor were described. At high im-
peller rotation speed, a Berty reactor should behave as a CSTR. Below are plotted the di-
mensionless exit concentrations, that is, c(t)1 Co, of cis-2-butene from a Berty reactor con-
taining alumina catalyst pellets that is operated at 4 atm pressure and 2000 rpm impeller
rotation speed at temperatures of 298 K and 427 K. At these temperatures, the cis-2-butene
is not isomerized over the catalyst pellets. At t = 0, the feed stream containing 2 vol % cis-
2-butene in helium is switched to a stream of pure helium at the same total flow rate. Reac-
tion rates for the isomerization of cis-2-butene into I-butene and trans-2-butene are to be
measured at higher temperatures in this reactor configuration. Can the CSTR material bal-
ance be used to ascertain the rate data?

• Answer
The exit concentrations from an ideal CSTR that has experienced a step decrease in feed
concentration are [from Equation (3.3.11)]:

clc O = exp[
If the RTD is that of an ideal CSTR (i.e., perfect mixing), then the decline in the exit con-
centration should be in the form of an exponential decay. Therefore, a plot of In(CI Co)
versus time should be linear with a slope of - 7 - 1. Using the data from the declining por-
tions of the concentration profiles shown in Figure 8.2.3, excellent linear fits to the data
are obtained (see Figure 8.2.4) at both temperatures indicating that the Berty reactor is
behaving as a CSTR at 298 K :=; T :=; 427 K. Since the complete backmixing is achieved
over such a large temperature range, it is most likely that the mixing behavior will also
occur at slightly higher temperatures where the isomerization reaction will occur over the
alumina catalyst.

c c
.8 .8
;;
.... ;;
1::: 0.8
E 0.8 c
"uc "uC
0 0
u 0.6 u 0.6
z; 'lJ

~
Y)

"2
0
0.4 "2" 0.4
.~ .~
c c
"
E 0.2 "E 0.2
is is
0 0
0 10 20 30 40 50 60 70 80 90 0 5 10 15 20 25 30 35 40 45 50
Time (min) Time (min)
(a) (b)

Figure 8.2.3 I Dimensionless concentration of cis-2-butene in exit stream of

Berty reactor as a function of time. See Example 8.2.3 for additional details.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _"'C'-'H...,AOOJ:p"'r"'E"'-lR"--'s"---"-'N'"-onideaLElow in Beacto~ 269

0.00 0.00

-0.50 -0.50

-1.00
~ -1.00 =0
\...)

~ Q -1.50
.s -1.50 .s
-2.00
-2.00 -2.50

-2.50 -3.00
10 20 30 40 50 60 70 80 90 10 15 20 25 30 35 40 45 50
Time (min) Time (min)
(a) (b)

Figure 8.2.4 I Logarithm of the dimensionless concentration of cis-2-butene in exit

stream of Berty reactor as a function of time. See Example 8.2.3 for additional details.

8.3 I Application of RTD Functions to the

Prediction of Reactor Conversion
The application of the RTD to the prediction of reactor behavior is based on the as-
sumption that each fluid element (assume constant density) behaves as a batch re-
actor, and that the total reactor conversion is then the average conversion of all the
fluid elements. That is to say:

mean concentratIOn]
o

of reactant 111
[ reactor outlet
o

=
2: [concentration of
reactant remammg m
0 •
] [fraction of exit stream]
0

th at consists
0 0 . 0 f fl UI°d
0
(8.3.1)
a flUId element of age elements of age
between t and t + elt between t and t elt

where the summation is over all fluid elements in the reactor exit stream. This equa-
tion can be written analytically as:

(CAl = j''''CA(t)E(t)dt (8.3.2)

o

where CA(t) depends on the residence time of the element and is obtained from:

dC A
(8.3.3)
dt

with
270 eM APTER 8 NonideaL8o.wJnBea'-.Lcl'-'-'o"-'[s'-- _

For a first-order reaction:

(8.3.4)

or

C A = C~ exp[ -kt] (8.3.5)

Insertion of Equation (8.3.5) into Equation (8.3.2) gives:

\CA ) = fXJ C~ exp[ -kt]E(t)dl (8.3.6)

°
Take for example the ideal CSTR. If the E(t)-curve for the ideal CSTR is used
in Equation (8.3.6) the result is:

\CA )
COfoo exp( -kl) exp( -1/7)dl
= -:-

°
or

\~~) ~ = 100

exp [ -(k + ~)lJdl

that gives after integration:

\CA) =~[_(_1)
7 k+ 1T
eXP[-(k + ~)tJIOO]
7 °
=_1
kT+l
(8.3.7)

Notice that the result shown in Equation (8.3.7) is precisely that obtained from the
material balance for an ideal CSTR accomplishing a first-order reaction. That is:

or

(8.3.8)
c1 kT + 1

Unfortunately, if the reaction rate is not first-order, the RTD cannot be used so di-
rectly to obtain the conversion. To illustrate why this is so, consider the two reac-
tor schemes shown in Figure 8.3.1.
Froment and Bischoff analyze this problem as follows (G. F. Froment &
K. B. Bischoff, Chemical Reactor Analysis and Design, Wiley, 1979). Let the
PFR and CSTR have space times of 71 and 72, respectively. The overall RTD for
either system will be that of the CSTR but with a delay caused by the PFR. Thus,
a tracer experiment cannot distinguish configuration (I) from (II) in Figure 8.3.1.
 CHAPTER 8 Nonideal Flow in Reactors 271

c~

(I) (II)

Figure 8.3.1 I
PFR and CSTR in series. (I) PFR follows the CSTR, (II) CSTR follows the PFR.

A first-order reaction occurring in either reactor configuration will give for the
two-reactor network:

C~ e- krj
(8.3.9)
C~ 1 + kT 2

This is easy to see; for configuration (I);

C~
CA = ------ (CSTR)
I + kT 2
C*
~ = exp[ -kT I ] (PFR)
CA
or

C~ e- krj
C~ 1 + kT 2
and for configuration (II);
CA
C~ = exp[ -kTd (PFR)

C~ 1
(CSTR)
CA 1 + kT 2

or

C~ 1 + kT 2

Now with second-order reaction rates, configuration (1) gives:

1 + \/
1(1 + 4(kC~'T2))
~ 1+
(8.3.10)
272 CHAPTER 8 Nonideal Flow in Reactors

while configuration (II) yields:

C~ -1 +V 1 + 4kC~T2
(8.3.11)
C~ 2kC~T2 + kC~T]( -1 + V 1 + 4kC~T2)
If kC~ = 1 and TdT] = 4, configurations (I) and (II) give outlet dimensionless
concentrations (C~/C~) of 0.25 and 0.28, respectively. Thus, while first-order
kinetics (linear) yield the same outlet concentrations from reactor configurations
(1) and (II), the second-order kinetics (nonlinear) do not. The reasons for these dif-
ferences are as follows. First-order processes depend on the length of time the mol-
ecules reside in the reactors but not on exactly where they are located during their
trajectory through the reactors. Nonlinear processes depend on the encounter of
more than one set of molecules (fluid elements), so they depend both on residence
time and also what they experience at each time. The RTD measures only the time
that fluid elements reside in the reactor but provides no information on the details
of the mixing. The terms macromixing and micromixing are used for the RTD and
mixing details, respectively. For a given state of perfect macromixing, two extremes
in micromixing can occur: complete segregation and perfect micromixing. These
types of mixing schemes can be used to further refine the reactor analysis. These
methods will not be described here because they lack the generality of the proce-
dure discussed in the next section.
In addition to the problems of using the RTD to predict reactor conversions, the
analysis provided above is only strictly applicable to isothermal, single-phase sys-
tems. Extensions to more complicated behaviors are not straightforward. Therefore,
other techniques are required for more general predictive and design purposes, and
some of these are discussed in the following section.

8.4 I Dispersion Models for

Nonideal Reactors
There are numerous models that have been formulated to describe nonideal flow in
vessels. Here, the axial dispersion or axially-dispersed plug flow model is described,
since it is widely used. Consider the situation illustrated in Figure 8.4.1. (The steady-
state PFR is described in Chapter 3 and the RTD for a PFR discussed in Section 8.2.)
The transient material balance for flow in a PFR where no reaction is occur-
ring can be written as:

ac
Acaz a/ = uACC i - [uACC i + a(uAcCJ] (8.4.1 )
(accumulation) (in) (out)

or

(8.4.2)
 CHAPTER 8 Nonideal Flow in Reactors 273

1
1

----: 1 I
Fi~ :~Fi+dFi
----1_-1-----
1 I
1 dz I
I I

"'AXiallY-diSpersed
~PFR
de
Fi = uACCi - ACDa df
Figure 8.4.1 I
Descriptions for the molar flow rate of species i in a
PFR and an axially-dispersed PFR. A c : cross-sectional
diameter of tube, u: linear velocity, D a : axial dispersion
coefficient.

where A c is the cross-sectional area of the tube. If u == constant, then:

ac = - uaC- i
(8.4.3)
at az
Now if diffusion/dispersion processes that mix fluid elements are superimposed on
the convective flow in the axial direction (z direction), then the total flow rate can
be written as:

dC i
Fi = uACCi - AcDa - (8.4.4)
dz
(convection) (dispersion)

Note that D a is called the axial-dispersion coefficient, and that the dispersion term
of the molar flow rate is formulated by analogy to molecular diffusion. Fick's First
Law states that the flux of species A (moles/area/time) can be formulated as:

dC A
N 4 = -D AB - , -
.. az + uCA (8.4.5)

for a binary mixture, where DAB is the molecular diffusion coefficient. Since axial
dispersion processes will occur by molecular diffusion during laminar flow, at this
condition the dispersion coefficient will be the molecular diffusion coefficient. How-
ever, with turbulent flow, the processes are different and D a must be obtained from
correlations. Since D a is the molecular diffusion coefficient during laminar flow, it
is appropriate to write the form of the dispersion relationship as in Equation (8.4.4)
and then obtain D a from correlations assuming this form of the molar flow rate ex-
pression. Using Equation (8.4.4) to develop the transient material balance relation-
ship for the axially-dispersed PFR gives:
274 CHAPTER 8 Nonideal FiQllLirlBeactors-

aCi ac)
A az-
at AeDa - '
e az
(accumulation) lin) (out)

ac'l
+ a(uAeC; - AeDa - . - , ) I (8.4.6)
, dz J
(out)

or for constant u and D a :

aCi +u-'
ac
(8.4.7)
at az
If 8 = t/(t) = (tu)/L (L: length of the reactor), Z z/L and Pea = (Lu)/D a (axial
Peclet number), then Equation (8.4.7) can be written as:

ac; + ac; = _1_ a2c; (8.4.8)

a8 az Pea az 2
The solution of Equation (8.4.8) when the input (i.e., Cat t = 0) is an impulse is:

(8.4.9)

Thus, for the axially-dispersed PFR the RTD is:

E (8) =
47T8
r_
(~Pc~?a)1 exp _--,-(l_8-,-)_2P~ea ]
48
(8.4.10)

A plot of E(8) versus 8 is shown in Figure 8.4.2 for various amounts of dispersion.
Notice that as Pea -+ CXJ (no dispersion), the behavior is that of a PFR while as
Pea -+ 0 (maximum dispersion), it is that of a CSTR. Thus, the axially-dispersed
reactor can simulate all types of behaviors between the ideal limits of no back-
mixing (PFR) and complete backmixing (CSTR).
The dimensionless group Pea is a ratio of convective to dispersive flow:

convective flow
- - - - - - in the axial direction (8.4.11 )
dispersive flow

The Peclet number is normally obtained via correlations, and Figure 8.4.3 illus-
trates data from Wilhelm that are plotted as a function of the Reynolds number for
packed beds (i.e.. tubes packed with catalyst particles). Notice that both the Pea
and Re numbers use the particle diameter, dp , as the characteristic length:

dpu dpup
Pea Re
Da ' M
 CHAPTER 8 Nonideal Flow in Reactors 275

o
I

Impulse
Jil
Measuring
point
(a)

2.0

1.5

§
~ 1.0
-<

0.5

o
o 0.5 1.0 1.5 2.0
8 =1/<1>
(b)

Figure 8.4.2 I
(a) Configuration illustrating pulse input to an axially-dispersed PFR. (b) Results observed
at measuring point.

It is always prudent to check the variables uscd in each dimensionless group prior
to their application. This is especially true with Peelet numbers, since they can have
many different characteristic lengths.
Notice that in packed beds, Pea 2 for gases with turbulent flow Re
(dpup)/Ji > 40, while for liquids Pea is below 1. Additionally, for unpacked tubes,
 3
.}IQ' 10
1\

10-4 10-3

Figure 8.4.3 I
Axial and radial Peclet numbers as a function of Reynolds number for packed-beds.
[Adapted from R. H. Wilhelm, Pure App. Chem., 5 (1962) 403, with pennission of the
International Union of Pure and Applied Chemistry.]

Pea (dtu/ D a) is about 10 with turbulent flow (Re = (dtup)/Ii greater than 2100)
(not shown). Thus, all real reactors will have some effects of dispersion. The ques-
tion is, how much? Consider again Equation (8.4.7) but now define Pea = deu/ Da
where de is an effective diameter and could be either d p for a packed bed or d t for
an open tube. Equation (8.4.7) can be then written as:

ac1+ aCt'
(8.4.12)
a8 az
If the flow rate is sufficiently high to create turbulent flow, then Pea is a constant and
the magnitude of the right-hand side of the equation is determined by the aspect ra-
tio, L/de- By solving Equation, (8.4.12) and comparing the results to the solutions
of the PFR [Equation (8.4.3)], it can be shown that for open tubes, L/dt > 20 is suf-
ficient to produce PFR behavior. Likewise, for packed beds, L/dp > 50 (isothermal)
and L/dp > 150 (nonisothermal) are typically sufficient to provide PFR character-
istics. Thus, the effects of axial dispersion are minimized by turbulent flow in long
reactors.
 CHAPTER 8 Nonldeal Flow in Reactors 277

VIGNETTE 8.4.11
B. G. Anderson et aL [Ind. Chern. Res.. 37 (1998) 815] obtained in situ of
pulses of IlC-Iabeled alkanes that were passing through packed beds of zeolites us-
ing positron emission tomography (PET). PET is a technique developed primarily for
nuclear medicine that is able to create three-dimensional images of gamma-ray emitting
species within various organs of the human body. By using PET, Anderson et aL could
obtain profiles of IIC-Iabeled alkanes as a function of time in a
pulse of the tracer alkane. Using analyses sim-
ing data were obtained:

y ,the
value calculated with the information presented in Figure 8.4.3 is in good agreement
with the findings.

8.5 I Prediction of Conversion with

an Axially-Dispersed PFR
Consider (so that an analytical solution can be obtained) an isothermal, axially-
dispersed PFR accomplishing a first-order reaction. The material balance for this
reactor can be written as:

dC A
u- kC4 = 0 (8.5.1)
dz

If y = CA/C~, Z = z/L, and Pea = uL IDa' then Equation (8.5.1) can be put into di-
mensionless form as:

(-kL)y
d 2y dy
2 U •
o (8.5.2)
Pea dZ dZ
The proper boundary conditions used to solve Equation (8.5.2) have been exhaus-
tively discussed in the literature. Consider the reactor schematically illustrated in
Figure 8.5.1. The conditions for the so-called" open" configuration are:
278 CHAPTER 8 Nooideal Flow io Bear,tors

0)
Pre-reaction Reaction zone Post-reaction
zone zone

2=0 2=1

Figure 8.5.1 I
Schematic of hypothetical reactor.

Z = -00, y= 1
Z = +00, Y = is finite
Z = 0,
Z = 1,
y(O_) = y(O+) = y(O)
y(L) = y(l+)
) (8.5.3)

Note that the use of these conditions specifies that the flux continuity:

gives:

dC A dC A
D -=D-
a, dz a2 dz

That is to say that if the dispersion coefficients in zones 1 and 2 are not the same,
then there will be a discontinuity in the concentration gradient at z = 0. Alterna-
tively, Danckwerts fonnulated conditions for the so-called "closed" configuration
that do not allow for dispersion in zones land 3 and they are:

uCAl o_ = [UCA - Da d~A Jlo+

(8.5.4)
dC
dz
A
I
IL_
= °
The Danckwerts boundary conditions are used most often and force discontinuities
in both concentration and its gradient at z = 0.

EXAMPLE 8.5.1 I
Consider an axially-dispersed PFR accomplishing a first-order reaction. Compute the di-
mensionless concentration profiles for L/dp 5 and 50 and show that at isothermal con-
ditions the values for = 50 are nearly those from a PFR. Assume Pe 2.
dp = 0.004 m and k/u = 25 m- I .
 CHAPTER 8 Nonideal Flow in Reactors 279

• Answer
The material balance for the axially-dispersed PFR is:

2
, -dy - (L
Pe (Lid) -kPe
a
-) y = 0
a P dZ udp

or

d 2y dy
- - (500L) - - (12500L 2 )y = 0
2
dZ dZ

The solution to this equation using the Danckwerts boundary conditions of:

1 (dp ) dy atZ = 0
1 = Y - Pea dZ L
dy
dZ = 0 atZ = 1

gives the desired form of y as a function of Z and the result is:

y = (Xl exp(524LZ) + (X2 exp( -23.9LZ)

where (Xl and (X2 vary with L/dp • The material balance equation for the PFR is:

or

dy = _(Lk)y
dZ u

with

y=l atZ=O

The solution to the PFR material balance gives:

YPF = exp[ -(kLZ)/u] exp[ -25LZ]

Note that the second-term of y is nearly (but not exactly) that of the expression for YPF and
that the first-term of y is a strong function of L. Therefore, it is clear that L will significantly
"*
affect the solution y to a much greater extent than YPF and that Y YPF even for very long
L. However, as shown in Figure 8.5.2, at L/dp = 50, Y = YPF for all practical matters. No-
tice that Y "*I at Z 0 because of the dispersion process. There is a forward movement of
species A because of the concentration gradient within the reaction zone. The dispersion al-
ways produces a lower conversion at the reactor outlet than that obtained with no mixing
(PFR)-recall conversion comparisons between PFR and CSTR.
"'2...
8"'O'-- -"C....H1Aou:P"-'TuE<.JRlL<8'"--i'N'l\.QlLOuuideaLEill'LinBeac.lauL _ . . .

0.95

0.9

0.85

~, 0.8

0.75

0.7

0.65

0.6
0.0 0.20 0.40 0.60 0.80 1.0
Z

0.8

0.6

0.4

0.2

o
0.0 0.20 0.40 0.60 0.80 1.0
z
Figure 8.5.2 I Dimensionless concentration profiles
for axially-dispersed (y) and plug flow reactors.

VIGNETTE 8.5.1
CHAPTER 8 Nonideal Flow in Reactors 281
282 CHAPTER 8 Nonideal Flow in ReactQ[~.~ _ _~. ~

8.6 I Radial Dispersion

Like axial dispersion, radial dispersion can also occur. Radial-dispersion effects nor-
mally arise from radial thermal gradients that can dramatically alter the reaction rate
across the diameter of the reactor. Radial dispersion can be described in an analo-
gous manner to axial dispersion. That is, there is a radial dispersion coefficient. A
complete material balance for a transient tubular reactor could look like:

2
ac ac (PC [a c 1 ac]
+u-=D +D - + - - (8.6.1 )
at az a az2 ,. ar2 r ar

e
If = t/(t) = (tu)/ L , Z z/ L, R r/ de (de is dp for packed beds, d t for unpacked
tubes), Pea = (ude )/ D a and Pe,. = (ude)/D,. (D,. is the radial-dispersion coefficient),
then Equation (8.6.1) can be written as:

(8.6.2)

The dimensionless group Pe,. is a ratio of convective to dispersive flow in the radial
direction:

convective flow
Pe,. ~----~ in radial direction (8.6.3)
dispersive flow

Referring to Figure 8.4.3, for packed beds with turbulent flow, Pe,. = 10 if de = dp-
For unpacked tubes, de = d t and Pe,. = 1000 with turbulent flow (not shown).
Solution of Equation (8.6.1) is beyond the level of this text.

8.7 I Dispersion Models for Nonideal

Flow in Reactors
As illustrated above, dispersion models can be used to described reactor behavior
over the entire range of mixing from PFR to CSTR. Additionally, the models are
not confined to single-phase, isothermal conditions or first-order, reaction-rate func-
tions. Thus, these models are very general and, as expected, have found widespread
use. What must be kept in mind is that as far as reactor performance is normally
concerned, radial dispersion is to be maximized while axial dispersion is minimized.
The analysis presented in this chapter can be used to describe reaction con-
tainers of any type~they need not be tubular reaetors. For example, consider the
situation where blood is flowing in a vessel and antibodies are binding to cells on
the vessel wall. The situation can be described by the following material balance:

ac
u (8.7.1)
az

with
 CHAPTER 8 Nonideal Flow in Reactors 283

ac
uCo = uC - D a az at z = 0, all r

ac
-=0 at z = L, all r
az
ac
ar
= ° at r = 0, all z

ac
-Dr _ = AB at r = rn all z
ar
where C is the concentration of antibody, rt is the radius of the blood vessel and AB
is the rate of antibody binding to the blood vessel cells. Thus, the use of the dis-
persion model approach to describing flowing reaction systems is quite robust.

Exercises for Chapter 8

1. Find the residence time distribution, that is, the effluent concentration of tracer
A after an impulse input at t = 0, for the following system of equivolume CSTRs
with a volumetric flow rate of liquid into the system equal to v:

How does the RTD compare to that of a single CSTR with volume 2V?
2. Sketch the RTD curves for the sequence of plug flow and continuous stirred
tank reactors given in Figure 8.3.1.
3. Consider three identical CSTRs connected in series according to the diagram
below.

(a) Find the RTD for the system and plot the E curve as a function of time.
(b) How does the RTD compare to the result from an axially-dispersed PFR
(Figure 8.4.2)? Discuss you answer in terms of the axial Peelet number.
(c) Use the RTD to calculate the exit concentration of a reactant undergoing
first-order reaction in the series of reactors. Confirm that the RTD method
284 CHAPTER 8 Nonideal Flow in ReactillS>--- _

gives the same result as a material balance on the system (see Example
3.4.3).
4. Calculate the mean concentration of A at the outlet (z L) of a laminar flow,
tubular reactor (C ~) accomplishing a second-order reaction (kC 1), and
compare the result to that obtained from a PFR when [( C~kL)/ u] I. Referring
to Example 8.1.1, is the deviation from PFR behavior a strong function of the
reaction rate expression (i.e., compare results from first- and second-order
rates)?
5. Referring to Example 8.2.3, compute and plot the dimensionless exit
concentration from the Berty reactor as a function of time for decreasing
internal recycle ratio to the limit of PFR behavior.
6. Consider the axially-dispersed PFR described in Example 8.5.1. How do the
concentration profiles change from those illustrated in the Example if the
second boundary condition is changed from:

at Z = I

to

y=o at Z 00

7. Write down in dimensionless form the material balance equation for a laminar
flow tubular reactor accomplishing a first-order reaction and having both axial
and radial diffusion. State the necessary conditions for solution.
8. Falch and Gaden studied the flow characteristics of a continuous, multistage
fermentor by injecting an impulse of dye to the reactor [E. A. Falch and
E. L. Gaden, Jr., Biotech. Bioengr., 12 (1970) 465]. Given the following RTD
data from the four-stage fermentor, calculate the Pea that best describes the
data.

Impulse Response Curve from a Four-Stage Fermenter

1

.S== 0.8
:;
.0
.~
'8 0.6
<U
.§
<U
u
0.4
==
<U
"0
"Vi
<U 0.2
~

o
o 0.5 1.5 2 2.5
Dimensionless time, t/<t>
 CHAPTER 8 Nonideal Flow in Reactors 285

0.000 0.000 0.950 0.745

0.050 0.001 0.990 0.710
0.090 0.040 1.035 0.675
0.120 0.080 1.080 0.630
0.170 0.140 1.110 0.600
0.210 0.220 1.180 0.550
0.245 0.330 1.200 0.525
0.295 0.420 1.240 0.485
0.340 0.530 1.290 0.460
0.370 0.590 1.365 0.380
0.420 0.670 1.410 0.360
0.460 0.730 1.450 0.325
0.500 0.810 1.490 0.310
0.540 0.830 1.580 0.250
0.590 0.860 1.670 0.220
0.630 0.870 1.760 0.180
0.670 0.875 1.820 0.150
0.710 0.850 1.910 0.140
0.750 0.855 2.000 0.125
0.790 0.845 2.120 0.100
0.825 0.840 2.250 0.080
0.870 0.810 2.320 0.070
0.920 0.780

9. Using the value of the Pea determined in Exercise 8, compute the concentration
profile in the reactor for the reaction of catechol to L-dopa catalyzed by whole
cells according to the following rate of reaction:

where Cs is the concentration of substrate catechol. This reaction is discussed

in Example 4.2.4. The values of the various parameters are the same as those
determined by the nonlinear regression analysis in Example 4.2.5, that is,
cf = 0.027 mol L -1, rmax = 0.0168 mol L - I h- I , and Km = 0.00851 mol
L- I. Assume that the mean residence time of the reactor is 2 h.
10. Using the same rate expression and parameter values given in Exercise 9,
compute the concentration profile assuming PFR behavior and compare to
the results in Exercise 9.