Professional Documents
Culture Documents
1 / 153
Rj
V
Q0 Q1
cj0 cj1
Conservation of mass
rate of (
) ( )
rate of inflow rate of outflow
accumulation = −
of component j of component j
of component j
rate of generation
+ of component j by (4.1)
chemical reactions
2 / 153
General Mole Balance
Rj
V
Q0 Q1
cj0 cj1
Conservation of mass
Z Z
d
cj dV = Q0 cj0 − Q1 cj1 + Rj dV (4.2)
dt V V
3 / 153
Z Z
d
cj dV = Q0 cj0 − Q1 cj1 + Rj dV
dt V V
4 / 153
Rate expressions
5 / 153
Rj
6 / 153
Batch Reactor
Z Z
d
cj dV = Q0 cj0 − Q1 cj1 + Rj dV
dt V V
Because the reactor is well stirred, the integrals in Equation 4.2 are simple to
evaluate, Z Z
cj dV = cj VR Rj dV = Rj VR
VR VR
7 / 153
Reactor Volume
dcj
= Rj (4.6)
dt
Use Equation 4.5 rather than Equation 4.6 if the reactor volume changes
significantly during the course of the reaction.
8 / 153
Analytical Solutions for Simple Rate Laws
9 / 153
First-order, irreversible
dcA
= −kcA (4.8)
dt
Watch the sign!
10 / 153
First-order, irreversible
cA (t) = cA0 , t =0
11 / 153
First-order, irreversible
0.8
0.6
cA
cA0
0.4 k = 0.5
k=1
0.2 k=2
k=5
0
0 1 2 3 4 5
t
12 / 153
First-order, irreversible
The A concentration decreases exponentially from its initial value to zero with
increasing time.
The rate constant determines the shape of this exponential decrease.
Rearranging Equation 4.9 gives
13 / 153
First-order, irreversible
k = 0.5
k=1
0.1
k=2
cA
cA0
0.01
k=5
0.001
0 1 2 3 4 5
One can get an approximate value of the rate constant from the slope of the
straight line.
This procedure is a poor way to determine a rate constant and should be
viewed only as a rough approximation (Chapter 9).
14 / 153
First-order, irreversible
dcB
= RB = kcA (4.10)
dt
with the initial condition for B, cB (0) = cB0
2 Note that the sum of cA and cB is constant.
d(cA + cB )
= RA + RB = 0
dt
Therefore, cA + cB is a constant.
15 / 153
First-order, reversible
cA + cB = cA0 + cB0
cB = cA0 + cB0 − cA
cB = cB0 + cA0 (1 − e−kt ) (4.11)
16 / 153
First-order, reversible
17 / 153
First-order, reversible
dcA
= −k1 cA + k−1 (cA0 + cB0 − cA ) (4.13)
dt
How do we want to solve this one?
Particular solution and homogeneous solution (see text)
Laplace transforms (control course)
Separation!
18 / 153
First-order, reversible
dcA
= acA + b
dt
Z cA Zt
dcA
= dt
cA0 acA + b 0
cA
1
ln(acA + b) =t
a cA0
b
cA = cA0 eat − (1 − eat )
a
Substitute a = −(k1 + k−1 ), b = k−1 (cA0 + cB0 )
19 / 153
First-order, reversible
k−1 h i
−(k1 +k−1 )t −(k1 +k−1 )t
cA = cA0 e + (cA0 + cB0 ) 1 − e (4.15)
k1 + k−1
k1 h i
−(k1 +k−1 )t −(k1 +k−1 )t
cB = cB0 e + (cA0 + cB0 ) 1 − e (4.16)
k1 + k−1
20 / 153
First-order, reversible
0.8 cA (t)
0.6 cBs
0.4 cAs
0.2 cB (t)
0
0 1 2 3 4 5
t
Figure 4.5: First-order, reversible kinetics in a batch reactor, k1 = 1, k−1 = 0.5, cA0 = 1,
cB0 = 0.
21 / 153
k−1
cAs = (cA0 + cB0 )
k1 + k−1
in which cAs is the steady-state concentration of A.
22 / 153
Nonzero steady state
23 / 153
Second-order, irreversible
24 / 153
Second-order, irreversible
Z cA Zt
dcA
2 = −k dt
cA0 cA 0
1 1
− = −kt
cA0 cA
Solving for cA gives
−1
1
cA = + kt (4.19)
cA0
Check that this solution satisfies the differential equation and initial condition
25 / 153
Second-order, irreversible
0.9
0.8
0.7
0.6
cA
0.5
cA0
0.4 second order
0.3
0.1
0
0 1 2 3 4 5
The second-order reaction decays more slowly to zero than the first-order
reaction.
26 / 153
Another second-order, irreversible
k
A + B -→ C r = kcA cB
dcA
= −r = −kcA cB
dt
dcB
= −r = −kcA cB
dt
Subtract B’s material balance from A’s to obtain
d(cA − cB )
=0
dt
27 / 153
dcA
= −kcA (cA − cA0 + cB0 )
dt
This equation also is separable and can be integrated to give (you should
work through these steps),
cB0 (cB0 −cA0 )kt −1
cA = (cA0 − cB0 ) 1 − e , cA0 ≠ cB0 (4.24)
cA0
28 / 153
Another second-order, irreversible
dcC
= kcA cB
dt
and therefore, d(cA + cC )/dt = 0. The concentration of C is given by
cC = cA0 − cA + cC 0
29 / 153
Notice that if cA0 > cB0 (Excess A), the steady state
cAs = 0
cBs = cB0 − cA0
cCs = cA0 + cC 0
30 / 153
nth-order, irreversible
4.5
4 n=3
3.5 2
r 2.5
1
2
1/2
1.5
0
1
0.5
0
0 0.5 1 1.5 2 2.5 3
cA
31 / 153
nth-order, irreversible
k
A -→ B r = kcAn
dcA
= −r = −kcAn
dt
dcA
= −kdt
cAn
32 / 153
nth-order, irreversible
33 / 153
nth-order, irreversible
0.8
0.6
cA
n=5
cA0
0.4 4
3
0.2 2
0
0 1 2 3 4 5
t
The larger the value of n, the more slowly the A concentration approaches
zero at large time.
34 / 153
nth-order, irreversible
0.8
0.6
cA
cA0
n=2
0.4
0.2 1/2
0
−2 −1 −1/2
0
0 0.5 1 1.5 2
t
35 / 153
For n < 0, the rate decreases with increasing reactant concentration; the
reactant inhibits the reaction.
5
4.5
3.5
r 2.5 n=−2
2 −1
1.5 −1/2
1
0
0.5
0
0 0.5 1 1.5 2 2.5 3
cA
36 / 153
Negative order, inhibition
Inhibition reactions are not uncommon, but watch out for small
concentrations. Notice the rate becomes unbounded as cA approaches zero,
which is not physically realistic.
When using an ODE solver we may modify the right-hand sides of the material
balance (
dcA −kcAn , cA > 0
=
dt 0, cA = 0
Examine the solution carefully if the concentration reaches zero.
37 / 153
r1 = k1 cA
r2 = k2 cB
38 / 153
Two reactions in series
dcA
= RA = −r1 = −k1 cA
dt
dcB
= RB = r1 − r2 = k1 cA − k2 cB
dt
dcC
= RC = r2 = k2 cB
dt
The material balance for component A can be solved immediately to give
cA = cA0 e−k1 t as before.
39 / 153
k1 h i
−k2 t −k1 t −k2 t
cB = cB0 e + cA0 e −e , k1 ≠ k2 (4.30)
k2 − k1
40 / 153
Two reactions in series – C
cC = cA0 + cB0 + cC 0 − cA − cB
41 / 153
0.8
cA (t) cC (t)
0.6
cB (t)
c
0.4
0.2
0
0 1 2 3 4 5
t
Figure 4.11: Two first-order reactions in series in a batch reactor, cA0 = 1, cB0 = cC 0 = 0,
k1 = 2, k2 = 1.
42 / 153
Two reactions in parallel
Assume the rates of the two irreversible reactions are given by r1 = k1 cA and
r 2 = k 2 cA .
43 / 153
dcA
= RA = −r1 − r2 = −k1 cA − k2 cA
dt
dcB
= RB = r1 = k 1 cA
dt
dcC
= RC = r2 = k 2 cA
dt
44 / 153
Two reactions in parallel
dcB
= k1 cA0 e−(k1 +k2 )t
dt
This equation is now separable and can be integrated directly to give
k1
−(k1 +k2 )t
cB = cB0 + cA0 1−e (4.34)
k1 + k2
45 / 153
k2
cC = cC 0 + cA0 1 − e−(k1 +k2 )t (4.35)
k1 + k2
46 / 153
Two reactions in parallel
0.8
cA (t) cC (t)
0.6
0.4
cB (t)
0.2
0
0 0.5 1 1.5 2 2.5 3
t
Figure 4.12: Two first-order reactions in parallel in a batch reactor, cA0 = 1, cB0 = cC 0 = 0,
k1 = 1, k2 = 2.
47 / 153
Notice that the two parallel reactions compete for the same reactant, A
The rate constants determine which product is favored
Large values of k1 /k2 favor the formation of component B compared to C and
vice versa
48 / 153
Conversion, Yield, Selectivity
There are several ways to define selectivity, yield and conversion, so be clear
about the definition you choose.
Point selectivity: The point (or instantaneous) selectivity is the ratio of the production rate
of one component to the production rate of another component.
Overall selectivity: The overall selectivity is the ratio of the amount of one component
produced to the amount of another component produced.
Yield: The yield of component j is the fraction of a reactant that is converted into
component j.
Conversion: Conversion is normally defined to be the fraction of a component that has
been converted to products by the reaction network. Conversion has
several definitions and conventions. It is best to state the definition in the
context of the problem being solved.
49 / 153
Qf Q
cjf Rj cj
50 / 153
CSTR – Constant Density
If the reactor volume is constant and the volumetric flowrates of the inflow
and outflow streams are the same, Equation 4.36 reduces to
dcj 1
= (cjf − cj ) + Rj (4.37)
dt τ
The parameter
τ = VR /Qf
is called the mean residence time of the CSTR.
We refer to this balance as the constant-density case. It is often a good
approximation for liquid-phase reactions.
51 / 153
The steady state of the CSTR is described by setting the time derivative in
Equation 4.36 to zero,
0 = Qf cjf − Qcj + Rj VR (4.38)
Conversion of reactant j is defined for a steady-state CSTR as follows
Qf cjf − Qcj
xj = (steady state) (4.39)
Qf cjf
One can divide Equation 4.38 through by VR to obtain for the constant-density
case
cj = cjf + Rj τ (steady state, constant density) (4.40)
52 / 153
Transient behavior of the CSTR
the feed concentration of A is cAf = 2 mol/L, the residence time of the reactor
is τ = 100 min, and the rate constant is k = 0.1 min−1 .
1 Find the steady-state concentration of A in the effluent for the given feed.
2 Plot the concentration of A versus time for constant feed concentration
cAf = 2 mol/L if the reactor is initially filled with an inert so cA0 = 0 mol/L.
3 Plot the concentration of A versus time for constant feed concentration cAf = 2
mol/L if the reactor is initially filled with feed so cA0 = 2 mol/L.
53 / 153
cA = cAf + RA τ
Substituting the production rate RA = −kcA and solving for cA gives the
steady-state concentration
cAf
cAs =
1 + kτ
Substituting in the numerical values gives
2 mol/L
cAs = = 0.182 mol/L
1 + (0.1 min−1 )(100 min)
54 / 153
Parts 2 and 3
dcA 1
= cAf − cA − kcA (4.41)
dt τ
cA (0) = cA0
55 / 153
Parts 2 and 3
1.5
cA (t) (mol/L)
1 cA0 = 2
0.5
cAs = 0.182
cA0 = 0
0
0 10 20 30 40 50 60 70 80
t (min)
Both solutions converge to the same steady-state even though the starting
conditions are quite different.
56 / 153
Phenol production in a CSTR
r = kcCHP
Find the reactor volume to achieve 85% conversion of CHP at steady state.
The flowrate into the reactor is Qf = 26.9 m3 /hr and k = 4.12 hr−1 .
57 / 153
Phenol production
cA = cAf + RA τ
58 / 153
Phenol production
59 / 153
Phenol production
Qf xA
VR =
k(1 − xA )
Substituting in the known values gives the required CSTR volume
(26.9 m3 /hr)(0.85)
VR = −1 = 37 m3
(4.12 hr )(0.15)
60 / 153
The Semi-Batch Reactor
The semi-batch reactor is a cross between the batch reactor and CSTR.
The semi-batch reactor is initially charged with reactant, like the batch reactor,
but allows a feed addition policy while the reaction takes place, like the CSTR.
Normally there is no outflow stream.
61 / 153
62 / 153
Volume Change Upon Reaction
d cj V R
= Qf cjf − Qcj + Rj VR (4.45)
dt
Equation 4.45 covers both the batch, CSTR and semi-batch reactors,
depending on how we specify Qf and Q.
If we multiply Equation 4.45 by the molecular weight of species j and sum
over all species we obtain,
P
d( j cj Mj VR ) X X X
= Qf cjf Mj − Q cj Mj + Rj Mj VR (4.46)
dt j j j
P
The term j cj Mj is simply the mass density of the reactor contents, which we
denote ρ
ns
X
ρ= cj Mj (4.47)
j=1
63 / 153
P
The term j cjf Mj is the mass density of the feedstream, ρf .
P
We know that conservation of mass in chemical reactions implies j Rj Mj = 0
(see Chapter 2). Substitution into Equation 4.46 leads to
d(ρVR )
= Qf ρf − Qρ (4.48)
dt
Equation 4.48 is clearly a total mass balance, in which the total mass in the
reactor changes in time due to the inflow and outflow of mass.
Notice that chemical reactions play no role in the total mass balance.
64 / 153
Equation of state for the mixture
ρ = fe (T , P, c1 , c2 , . . . , cns )
f (c1 , c2 , . . . , cns ) = 0
65 / 153
For example, we could express the equation of state in terms of the partial
molar volumes as X
cj V j = 1
j
66 / 153
Equation of state for the mixture — Ideal mixture
67 / 153
Constant density
ρ = ρf
68 / 153
Constant density
dVR
= Qf − Q
dt
69 / 153
Nonconstant density
Unknowns.
In the general case, consider the following variables to fully
determine the state of the reactor: T , P, nj , VR .
We also require the value of Q to specify the right-hand sides
of the material balances.
The set of unknowns is nj , VR , Q.
We therefore have ns + 2 unknowns.
Equations.
We have the ns equations from the component mole balances.
The equation of state provides one additional equation.
The final equation is provided by a statement of reactor
operation.
70 / 153
Nonconstant density – reactor operation
71 / 153
in which fj is
∂f
fj =
∂cj
and ∆fi is
X X ∂f
∆fi = νij fj = νij
j j
∂cj
72 / 153
Nonconstant density – idea mixture
P
For the ideal mixture we have f (cj ) = j cj Vj◦ − 1 = 0.
fj = Vj◦
the pure component specific volumes
The ∆fi are given by
∆fi = ∆Vi◦
the change in specific volume upon reaction i.
So the reactor volume can be calculated from
dVR X
= Qf − Q + ∆Vi◦ ri VR
dt i
73 / 153
Nonconstant density
dnj
Component balances: = Qf cjf − Qcj + Rj VR , j = 1, . . . , ns
dt
P P
Defined quantities: nj = cj VR ρ= j cj Mj ∆Vi◦ = ◦
j νij Vj
P ◦
(i) constant density: ρ = ρ0 (ii) ideal mixture: j cj Vj = 1
P ◦
1. vol VR = VR0 Q = Qf VR = VR0 Q = Qf + i ∆Vi ri VR
dVR P
2. mass VR = VR0 Q = Qf = Qf (1 − ρf /ρ) + i ∆Vi◦ ri VR Q = Qf ρf /ρ
dt
dVR dVR P
3. Q = Qf − Q Q specified = Qf − Q + i ∆Vi◦ ri VR Q specified
dt dt
74 / 153
Nonconstant density
dnj
Component balances: = Qf cjf − Qcj + Rj VR , j = 1, . . . , ns
dt
P P ∂f
Defined quantities: nj = cj VR ρ= j cj Mj ∆fi = j νij ∂c
j
DAEs ODEs
P P
j fj cjf ∆f r V
1. vol VR = VR0 f (cj ) = 0 VR = VR0 Q = Qf P + iP i i R
j fj cj j fj cj
P P
dVR j fj cjf ∆f r V
2. mass ρVR = ρ0 VR0 f (cj ) = 0 = Qf P − Q + iP i i R Q = Qf ρf /ρ
dt j fj cj j fj cj
P P
dVR f c
j j jf ∆f r V
3. Q Q specified f (cj ) = 0 = Qf P − Q + iP i i R Q specified
dt j fj cj j fj cj
75 / 153
Semi-batch polymerization
76 / 153
Semi-batch polymerization
77 / 153
78 / 153
Semi-batch polymerization
d(cM VR )
= Qf 0 cMf − kcM VR
dt
in which cMf = ρM /MM is given, and Qf = Qf 0 is constant during the filling
operation.
We denote the total number of moles of monomer by M = cM VR , and can write
the monomer balance as
dM
= Qf 0 cMf − kM (4.54)
dt
M(0) = 0
79 / 153
Semi-batch polymerization
dVR
= Qf 0 + ∆VkM (4.55)
dt
VR (0) = 10 m3
80 / 153
The polymer mass production
To compute the polymer mass, we note from the stoichiometry that the mass
e P is
production rate of polymer R
e P = −RM MM
R
d Pe
e p VR = kcM MM VR = (kMM )M
=R (4.56)
dt
81 / 153
Semi-batch polymerization
The text solves this problem analytically. Instead, let’s solve it numerically.
Let t1 be the time that the reactor fills.
We need an ODE solver that is smart enough to stop when the reactor fills,
because we do not know this time t1 . The ODE solver needs to find it for us.
dasrt is an ODE solver with the added capability to find the time at which
some event of interest occurs.
82 / 153
Finding the time for filling the reactor
t1 = 11.2 min
Note the reactor would have filled in 10 min if the density were constant.
The extra time reflects the available volume created by converting some of
the monomer to polymer during filling.
After t1 we consider the two operations.
83 / 153
Operation 1.
84 / 153
Semi-batch polymerization
2
20
15
10
VR (m3 )
0 10 20 30 40 50
time (min)
Figure 4.15: Semi-batch reactor volume for primary monomer addition (operation 1) and
primary plus secondary monomer additions (operation 2).
85 / 153
Operation 2.
1
0.8
0.6
Qf (m3 /min)
Because the reactor volume is constant, we can solve Equation 4.55 for the
feed flowrate
0.4 during the secondary monomer addition
0.2
Qf = −∆VkM
2
Operation 2 is also shown in the figures.
0
Notice the final polymer production
1 is larger in operation 2 because of the
extra monomer addition.
-0.2
0 10 20 30 40 50
time (min)
Figure 4.16: Semi-batch reactor feed flowrate for primary monomer addition (operation 1)
and primary plus secondary monomer additions (operation 2).
86 / 153
Polymer production rate
Pe 2 − Pe 1
× 100% = 22.5%
Pe 1
By using the volume of the reactor more efficiently, the total polymer
production increases 22.5%.
87 / 153
Plug flow in a tube is an ideal-flow assumption in which the fluid is well mixed
in the radial and angular directions.
The fluid velocity is assumed to be a function of only the axial position in the
tube.
Plug flow is often used to approximate fluid flow in tubes at high Reynolds
number. The turbulent flow mixes the fluid in the radial and angular
directions.
Also in turbulent flow, the velocity profile is expected to be reasonably flat in
the radial direction except near the tube wall.
88 / 153
Thin Disk Volume Element
Given the plug-flow assumption, it is natural to take a thin disk for the reactor
volume element
Qf Q
cjf cj
z }| {
z z + ∆z
89 / 153
90 / 153
Length or volume as independent variable
91 / 153
Steady-State Operation
d(cj Q)
= Rj (4.66)
dV
The product cj Q = Nj is the total molar flow of component j. One also can
express the PFR mole balance in terms of the molar flow,
dNj
= Rj (4.67)
dV
92 / 153
Volumetric Flowrate for Gas-Phase Reactions
To use Equation 4.67 for designing a gas-phase reactor, one has to be able to
relate the volumetric flowrate, Q, to the molar flows, Nj , j = 1, 2, . . . , ns .
The important piece of information tying these quantities together is, again,
the equation of state for the reaction mixture, f (T , P, cj ) = 0.
Because the molar flow and concentration are simply related,
N j = cj Q (4.68)
93 / 153
94 / 153
Ideal Gas Equation of State
To evaluate the concentrations for use with the reaction rate expressions, one
simply rearranges Equation 4.68 to obtain
Nj P Nj
cj = = P (4.70)
Q RT j Nj
95 / 153
dM
= 0, M(0) = Mf
dV
in which Mf is the feed mass flowrate.
The total mass flow in a PFR is constant.
96 / 153
Volumetric Flowrate for Liquid-Phase Reactions
and the volumetric flowrate is constant and equal to the feed value.
Equation 4.73 is used often for liquid-phase reactions.
97 / 153
dcj
= Rj , constant flowrate (4.74)
dτ
which is identical to the constant-volume batch reactor.
For the constant-flowrate case, the steady-state profile in a PFR starting from
a given feed condition is also the transient profile in a batch reactor starting
from the equivalent initial condition.
98 / 153
Single Reaction Systems – Changing flowrate in a PFR
99 / 153
dNA
= RA
dV
The production rate of A is RA = −r = −kcA .
Substituting the production rate into the above equation gives,
dNA
= −kNA /Q (4.75)
dV
The volumetric flowrate is not constant, so we use Equation 4.69, which
assumes an ideal-gas equation of state,
RT
Q= (NA + NB + NC ) (4.76)
P
100 / 153
Changing flowrate in a PFR
dNB dNC
= RB = r = RC = r
dV dV
Adding the mole balance for A to those of B and C gives
d (NA + NB ) d (NA + NC )
=0 =0
dV dV
The stoichiometry does not allow the molar flow NA + NB or NA + NC to change
with position in the tube.
101 / 153
Because NA + NB and NB + NC are known at the tube entrance, one can relate
NB and NC to NA ,
NA + NB = NAf + NBf
NA + NC = NAf + NCf
NB = NAf + NBf − NA
NC = NAf + NCf − NA
102 / 153
Changing flowrate in a PFR
RT
Q= 2NAf − NA
P
Substituting this expression in Equation 4.75 gives the final mole balance,
dNA P NA
= −k
dV RT 2NAf − NA
103 / 153
RT
V =− NAf [2 ln(1 − xA ) + xA ]
kP
104 / 153
Changing flowrate in a PFR
Because Qf = NAf RT /P is given in the problem statement and the tube length
is desired, it is convenient to rearrange the previous equation to obtain
Qf
z=− [2 ln(1 − xA ) + xA ]
kAc
Substituting in the known values gives
!
35 × 103 cm3 /min 4
z=− [2 ln(1 − .35) + .35]
0.05 sec−1 60 sec/min π (2.5 cm)2
z = 1216 cm = 12.2 m
105 / 153
Multiple-Reaction Systems
The modeler has some freedom in setting up the material balances for a
plug-flow reactor with several reactions.
The most straightforward method is to write the material balance relation for
every component,
dNj
= Rj , j = 1, 2, . . . , ns
dV
nr
X
Rj = νij ri , j = 1, 2, . . . , ns
i=1
106 / 153
Benzene pyrolysis in a PFR I
Hougen and Watson [3] analyzed the rate data for the pyrolysis of benzene by
the following two reactions.
Diphenyl is produced by the dehydrogenation of benzene,
k1
-*- C12 H10 + H2
2C6 H6 )
k−1
2B -*-
) D+H
107 / 153
The reactions are assumed to be elementary so that the rate expressions are
2 cD cH cT cH
r1 = k1 cB − r2 = k2 cB cD − (4.79)
K1 K2
108 / 153
Benzene pyrolysis in a PFR
Calculate the tube volume required to reach 50% total conversion of the
benzene for a 60 kmol/hr feed stream of pure benzene.
The reactor operates at 1033K and 1.0 atm.
Plot the mole fractions of the four components versus reactor volume.
109 / 153
The rate and equilibrium constants at T = 1033K and P = 1.0 atm are given
in Hougen and Watson,
k1 = 7 × 105 L/mol · hr
k2 = 4 × 105 L/mol · hr
K1 = 0.31
K2 = 0.48
110 / 153
Benzene pyrolysis in a PFR
The mole balances for the four components follow from the stoichiometry,
dNB
= −2r1 − r2
dV
dND
= r1 − r2
dV
dNH
= r1 + r2
dV
dNT
= r2
dV
The initial condition for the ODEs are NB (0) = NBf and
ND (0) = NH (0) = NT (0) = 0.
111 / 153
The total molar flux does not change with reactor volume.
RT
Q= NBf (4.80)
P
The rate expressions are substituted into the four ODEs and they are solved
numerically.
The total conversion of benzene, xB = (NBf − NB )/NBf , is plotted versus
reactor volume in Figure 4.20.
A reactor volume of 404 L is required to reach 50% conversion. The
composition of the reactor versus reactor volume is plotted in Figure 4.21.
112 / 153
0.6
0.5
0.4
xB 0.3
0.2
0.1
0
0 200 400 600 800 1000 1200 1400 1600
V (L)
113 / 153
0.8
0.6 yB
yj
0.4
yH
yD
0.2
yT
0
0 200 400 600 800 1000 1200 1400 1600
V (L)
114 / 153
Ethane pyrolysis in the presence of NO
115 / 153
We have two continuous reactors in this chapter: the CSTR and the PFR.
Let’s compare their steady-state efficiencies in converting reactants to
products.
For simplicity, consider a constant-density, liquid
k
A -→ B r = kcAn
116 / 153
Some PFR-CSTR Comparisons
We rearrange and solve for the time required to change from the feed
condition cAf to some exit concentration cA
Z cA
1 0
τ= 0 dcA
cAf r (cA )
The area under the curve 1/r (cA0 ) is the total time required to achieve the
desired concentration change.
1 CSTR
r (cA0 )
PFR
cA cAf
cA0
117 / 153
1 CSTR
r (cA0 )
PFR
cA cAf
cA0
118 / 153
Some PFR-CSTR Comparisons
119 / 153
If the reaction order is positive (the usual case), the PFR is more efficient. If
the reaction order is negative, the CSTR is more efficient.
1 PFR
r (cA0 )
CSTR
cA cAf
cA0
120 / 153
The PFR versus CSTR with separation
The PFR achieves higher conversion than an equivalent volume CSTR for the
irreversible reaction with first-order kinetics
A -→ B r = kcA
121 / 153
122 / 153
The PFR versus CSTR with separation
NA0 NA
VR
VR
αNA2 αNA2
pure B
In the text, we show how to find the recycle flowrate so this system achieves
the PFR conversion.
123 / 153
This example was motivated by a recent result of Feinberg and Ellison called
the CSTR Equivalence Principle of Reactor-Separator Systems [2].
This surprising principle states:
For a given reaction network with ni linearly independent reactions, any
steady state that is achievable by any reactor-separator design with total
reactor volume V is achievable by a design with not more than ni +1 CSTRs,
also of total reactor volume V . Moreover the concentrations, temperatures
and pressures in the CSTRs are arbitrarily close to those occurring in the
reactors of the original design.
124 / 153
Stochastic Simulation of Chemical Reactions
125 / 153
126 / 153
Stochastic Simulation of Chemical Reactions
Assume we have only a hundred molecules moving randomly in the gas phase
k1
A -→ B
k2
B -→ C
r1 = k1 xA r2 = k2 xB
127 / 153
The basic idea of the Gillespie algorithm is to: (i) choose randomly the time at
which the next reaction occurs, and (ii) choose randomly which reactions occurs at
that time.
1 Initialize. Set integer counter n to zero. Set the initial species numbers,
xj (0), j = 1, . . . ns . Determine stoichiometric matrix ν and reaction probability
laws (rate expressions)
ri = ki h(xj )
for all reactions.
2 Compute reaction probabilities, ri = ki h(xj ). Compute total reaction
P
probability rtot = i ri .
128 / 153
Stochastic Simulation of Chemical Reactions
129 / 153
1 Determine reaction m to take place in this time interval. The idea here is to
partition the interval (0,1) by the relative sizes of each reaction probability
and then “throw a dart” at the interval to pick the reaction that occurs. In this
manner, all reactions are possible, but the reaction is selected in accord with
its probability.
p2
r1 r2
0 r1 +r2 r1 +r2 1
130 / 153
Stochastic Simulation of Chemical Reactions
1 Update the simulation time t(n + 1) = t(n) + et . Update the species numbers
for the single occurrence of the mth reaction via
xj (n + 1) = xj (n) + νmj , j = 1, . . . ns
131 / 153
If rtot is the total probability for reaction, e−rtot et is the probability that a
reaction has not occurred during time interval et .
We will derive this fact in Chapter 8 when we develop the residence-time
distribution for a CSTR.
The next figure shows the results of this algorithm when starting with
xA = 100 molecules.
132 / 153
Stochastic Simulation of Chemical Reactions
100
80
xC
xA
60
xB
xj
40
20
0
0 1 2 3 4 5
t
133 / 153
Notice the random aspect of the simulation gives a rough appearance to the
number of molecules versus time, which is quite unlike any of the
deterministic simulations.
Because the number of molecules is an integer, the simulation is actually
discontinuous with jumps between simulation times.
But in spite of the roughness, we already can make out the classic behavior of
the series reaction: loss of starting material A, appearance and then
disappearance of the intermediate species B, and slow increase in final
product C.
134 / 153
Stochastic Simulation of Chemical Reactions
135 / 153
1000
800
xC
xA
600
xB
xj
400
200
0
0 1 2 3 4 5
t
136 / 153
Stochastic Simulation of Chemical Reactions
4000
3200
xC
xA
2400
xB
xj
1600
800
0
0 1 2 3 4 5
t
137 / 153
We see the random fluctuations become less pronounced. Notice that even
with only 4000 starting molecules, the results compare very favorably with
the deterministic simulation shown previously.
Another striking feature of the stochastic approach is the trivial level of
programming effort required to make the simulations.
The biggest numerical challenge is producing the pseudorandom numbers
and many well-developed algorithms are available for that task.
The computational time required for performing the stochastic simulation
may, however, be large.
138 / 153
Hepatitis B virus modeling
139 / 153
cccDNA
nucleotides −−
−−
−→ rcDNA
nucleotides + rcDNA −−
−−
−→ cccDNA
cccDNA
amino acids −−
−−
−→ envelope
cccDNA −−
−−
−→ degraded
envelope −−
−−→ secreted or degraded
−
rcDNA + envelope −−
−−→ secreted virus
−
140 / 153
Hepatitis B virus modeling
The reaction rates and production rates for Reactions 4.86–4.91 are given by
r1 k 1 xA
r2 k2 xB
RA r2 − r4
r3 k3 xA
RB = r1 − r2 − r6 (4.94)
r = k x
4 4 A RC r3 − r5 − r6
r5 k5 xC
r6 k6 xB xC
141 / 153
Assume the systems starts with a single cccDNA molecule and no rcDNA and
no envelope protein, and the following rate constants
h iT h iT
xA xB xC = 1 0 0 (4.92)
h i
kT = 1 0.025 1000 0.25 2 7.5 × 10−6 (4.93)
142 / 153
Average stochastic is not deterministic. I
25
20
15
cccDNA
deterministic
10
average stochastic
5
0
0 50 100 150 200
t (days)
250
200
150
rcDNA
deterministic
100
average stochastic
50
0
0 50 100 150 200
t (days)
10000
8000
envelope
6000 deterministic
2000
0
0 50 100 150 200
t (days)
143 / 153
35
30 stochastic 1
25
20
cccDNA
15
10 deterministic
5
stochastic 2
0
Figure 4.35: Species cccDNA versus time for hepatitis B virus model; two representative
stochastic trajectories.
144 / 153
Hepatitis B virus modeling
145 / 153
In fact, it occurs for 125 of the 500 simulations. So the average stochastic
simulation consists of 75% trajectories that fluctuate about the deterministic
trajectory and 25% trajectories that go to zero.
25
20
15
cccDNA
deterministic
10
average stochastic
5
0
0 50 100 150 200
t (days)
146 / 153
Summary
147 / 153
Summary
d(cj VR )
BATCH = Rj VR (4.95)
dt
dcj
constant volume = Rj (4.96)
dt
d(cj VR )
CSTR = Qf cjf − Qcj + Rj VR (4.97)
dt
dcj 1
constant density = (c − cj ) + Rj (4.98)
dt τ jf
d(cj VR )
SEMI-BATCH = Qf cjf + Rj VR (4.100)
dt
∂cj ∂(cj Q)
PFR = − + Rj (4.101)
∂t ∂V
d(cj Q)
steady state = Rj (4.102)
dV
dcj
constant flowrate = Rj , τ = V /Qf (4.103)
dτ
148 / 153
Summary
149 / 153
Summary
150 / 153
Summary
We showed that the PFR achieves higher conversion than the CSTR of the
same volume if the reaction rate is an increasing function of a component
composition (n > 0 for an nth-order rate expression).
Conversely, the CSTR achieves higher conversion than the same-volume PFR if
the rate is a decreasing function of a component composition (n < 0).
Finally, we introduced stochastic simulation to model chemical reactions
occurring with small numbers of molecules.
151 / 153
Summary
The stochastic model uses basic probability to compute reaction rate. A given
reaction’s probability of occurrence is assumed proportional to the number of
possible combinations of reactants for the given stoichiometry.
Two pseudorandom numbers are chosen to determine: (i) the time of the next
reaction and (ii) the reaction that occurs at that time.
The smooth behavior of the macroscopic ODE models is recovered by the
random simulations in the limit of large numbers of reacting molecules.
With small numbers of molecules, however, the average of the stochastic
simulation does not have to be equal to the deterministic simulation. We
demonstrated this fact with the simple, nonlinear hepatitis B virus model.
152 / 153
References I
153 / 153