MODULE

BIOCHEMICAL REACTORS
8

After studying this module, the student should be able to

Develop the dynamic modeling equations for a two~statc biochemical reactor

Understand the concept of "washout"
Objetivos:
Understand the different types of steady-state and dynamic behavior exhibited by
the MOl1od and Substrate Inhibition models
Find the numher of steady-state solutions and to determine the stability of each
steady-stale

The Inajor sections of this module arc:

MS.I Background
MS.2 Modeling l~quatiolls

MS.3 Steady-state Solution

MS.4 Dynamic Behavior
MS.) Linearization
MS.6 Phase-plane Analysis
MS.7 Understanding Multiple Steady-states
MS.S Bifurcation Behavior

M8.1 BACKGROUND

USOS: Biochemical reactors arc used to produce a large number of intermediate and final procl~
ucts, including pharmaceuticals, food, and beverages. Biochemical reactor models arc
similar to chemical reactor models, since the same type of material balances are per-

529
 530 Biochemical Reactors Module 8

F
X1l

x"
.-
X,
x2 F
V X,

X2 FIGUH.E MS.l Biochemical reactor.

formed. In the simplest reaclor we consider two components: biomass and substrate. The
biomass consists of cells that consume the substrate. Onc exmnplc would be a wastewater
treatment system, where the biomass is lIsed to "cat" waste chemicals (substrate). Another
example is fermentation, where cells consuille sugar and produce alcohol.
Consider the schematic of a biochemical reactor shown in Figure Mg.l.
Se asume In this module we aSSlllllC that the reactor is perfectly mixed and that the volume is
constant. We usc the following notation:
mass of cells
Xl = biomass concentration
volume

mass of substrate
X2 = substrate concentration
volume

mass ofc~II~_g?ncrated
1'1 = rate ofceH generation
vol lImc -time

mass of substrate consumed

"2 -_'::0 rate of substrate consumption
volume- time
F = volumetric f10wrate -= volume/timc
Now we can write the material btdances to describe the behavior of this systell1.

M8.2 MODELING EQUATIONS

The dynamic model is developed by writing material balances on the biomass (cells) and
the substrate (feed source for the cells). Biomass grows by reeding on the substrate.

M8.2.1 Biomass Material Balance

We write the biomass material balance as:

rate of accumulation:;:: in by flow -- out by flow + generation
dVx,
dl ,= ['\1/ - fit, + VI', (MS.I)
Sec. M8.2 Modeling Equations 531

where ,x lf is the concentration of biomass in the feed stream and F is the volumetric
flowratc.

M8.2.2 Substrate Material Balance

The substrate material balance is written:

rate of accumulation::;;: in by now - out by flow ~ consumption

dVx, . .
=~
dt , .} ~ [-<x".. - Vr.,
l'x)· (M8.2)

where x 2I is the concentration of substrate in the feed stream.

M8.2.3 Specific Growth Rate

The reaction ratc (mass of cells generatedlvolume time) is nonnally written in the follow-
ing form:

r1 ::::: fL<t 1. (M83)

where /.L is the specific growth rate coefficient. We can think of /-l as being sitnilar (0 a
first-order reaction rate constant; however, jJ. is not constanl-·-it is a fUllction of the sub-
strate concentration as shown in Section M8.2.6. The units of /L arc time·- I ,

M8.2.4 Yield

There is a relationship between the rate of generation of biomass and the rate of COnS1l1l1p~
tion of substrate. Define Yas the yield, that is, the mass of cells produced per mass of sub-
strate consumed:

y~
mass of cells produced 1',
(M8A)
mass of substrate consumed
I'"rom (MBA) we can write:

(M8.5)

and substituting (M8.3) into (M8.5). we find:

r = P'~~J (M8.6)
) Y

We assume in the subsequent analysis that Y is a constant.

M8.2.5 Dilution Rate

Assuming a constant volume reactor, we can write (M8. J) and (M8.2) as:
Asumiendo:
 532 Biochemical Reactors Module 8

dX I F
V Xjf - (MS.7)
dl

F F
VX2f- V XZ - r2 (MS.S)

Defining F/Vas D, the dilution rate, and using the rate expressions in (MS.J) and (M8.6),
we find:

dX I
(MS.9)
dl

dx]
D x_21 - D..~tJ
.\,-
. y (MS.IO)
dl

Se asume: Generally, it is assumed that there is no biomass in the fecd stream, so xlJ':::;: O. The biore-
actor modeling equations arc then norrnally written in the following form:

dX 1
(MS.II)
dl

(MS.12)

The dilution rate (D) is the same as the space velocity in the chemical reaction engineer-
ing literature. It is also the inver,",c of the reactor re.,;idence time and has units of time·-- l .
The expressions for jJ- (specific growth rate) are developed in the following section.

M8.2.6 Growth Rate Expressions

The growth rate coefficient is usually not constant. A numbcr of functional relationships
hetween the growth rate coefficient: and substrate concentration have been developed. The
lllost common are (i) MOJ/od and (ii) Substrate inhibition.

MONaD
The growth rate coefficient oftcn varies in a hyperholic fashion. The following forin
was proposed by MOl1od in 1942. Notice that jJ- is first-order at low x2 and zcro order at
high x2'

(MS.I3)

Notice lhat jJ- is firsl-order at low x 2 and zero order at: high x 2 . That is, when x2 is low:
Sec. M8.2 Modeling Equations 533

and when '-\"2 is high:

Since the reaction rate is:

rl /LY 1

this lncans that the Monod description is similar to a second-order (bimolecular) reaction
when x 2 is low, since

r ~_.

and to a first-order reaction when x 2 is high, since

1'1 = !Lilla;; Xl

[~quation (MB.l3) is the same form as the Langmuir (u.Jsorption isotherm and the sl2uHJ;Jrd
rate equation for enzyme-catalyzed reactions with a single substrate (Michaclis-l'viclllC'1l
kinetics).

SUBSTRATE INHIBITION
Sometimes the growth ra.le coefficient increases at low substrate concentration. but de-
creases at high substrate concentratioll. The physical reason may be that the suhstrate has
a toxic effect on the biomass cells at a higher concentration. This effect is called suhstrate
inhibition and is represented by the t()llowing cquation:

0.6
Monad
0.5

0.4

E 03

0.2

0.1

0
0 2 3 4
x2

FIGURE IV18.2 COIllparison of Monod and substratc inhihition models for

growth rate.
 534 Biochemical Reactors Module 8

TABLE MS.l I'aramctcrs for lVIonod

and Substrate Inhibition lVlodcls

MOllOd Substrate Inhibition

0.53 hr l fL max 0.53 hI I

0.12 g/litcr kill 0.12 g/Jitcr
k, 0.4545 litcr/g
y 0.4 Y 0.4
4.0 g/Iitcr 4.0 g/li'c!

fL~ (MBI4)
kill + x 2 + 1<1
Notice that the Monod equation is a special case of (M8.14), with k 1 = O.

SPECIFIC GROWTH RATE RELATIONSHIPS

The characteristic relationships between substrate (x 2 ) and specific growth rate (f.-l) arc
quite different for Monnd and substrate inhibition. The curves for f.L as a function of _\:! for
both models arc compared in I<igurc M8.2. Notice that the substrate inhibition model ex-
hibits a maximum in the growth rate curve, while Monod becomes zcnH}rdcr al high sub-
strate concentrations.

M8.3 STEADY·STATE SOLUTION

In this section, the MATLAB function fsolve will be used to solve for the steady-slate
values or the biomass and substrale concentrations. The nUlllerical values used in our silll~
ulations are shown in Table MR.I.
We \vill study the following cases:

Case 1. Medium Dilution Rate, 1\::::: 0.3 hr"

Case 2. Low Dilution Rale, D s ::::: 0.15 he l
Case 3. High Dilution Rate, D, ~ 0.45 he'

EXAM PI ,E MS. I Case I Results (Il ~ 0.3)

The function file bio ss. m (Appendix 1) is set for Case! (D =: 0.3) and Ihe sub.'ilrale inhibi-
tion model (kl -co 0.4545). The l'vfATLAB function fsolve is used to solve for the steady-slatc
values by entering tbe following in the comfnand window (with an initial guess of x (1) =: 1
andx(2) 1):
0;:

» X = f~301ve( 'bio )3~:;' ,[1;1])

 Sec. M8.4 Dynamic Behavior 535

The sh:ady-statc solution obtained is:

x =
O.99 tj1
1.5122

Different initial guesses resuil in two other solutions for the substrate inhibition model. Also. the
MOJlod model has two steady-slate solutions. "fhe reader should find the following results using
[sol vc and bio_s~;, m, by entering different initial guesses.

Monod (2 steady-state solutions)

Equilibriulll I 0
Xis::;;: X2s::::; 4.0
Equilibriulll 2 'b = 1.5374 x 2s :;:;:; 0./565

Substrate Inhibition (3 steady-state solutions)

Equilibriulll I X
Is := 0 X21::::; 4.0
Equilibrium 2 Xis;;;:: 0.9951 .1 2.1'::::;1.5123
EquilibriullI J .\1.1-:;;;' 1.5302 '2, = 0.1745

Notice that EC]uilibriUl,11 3 on the Sf model is ahnosl identical to Equilibrium 2 for the
Monad model. In this section we have discussed case 1 results (D :;:: 0.3) only. Cases 2
and 3 will be discussed in Section M8.?
In the next section we will analyze the dynamic behavior of this systeln, and in Sec-
tion MS.7 we will show how multiple steady-state solutions arise.

M8.4 DYNAMIC BEHAVIOR

In the previous section we found that the Monod and substrate inhibition models had t\V()
and three steady-state solutions, respcctively, for the Case 1 paramcler values. In this sec-
tion we perform simulations or the dynamic behavior of this system. A function file
named bio. m is shown in Appendix 2.

M8.4.1 Case 1 (0 = 0.3), Substrate Inhibition Model

'rhe initial simulation is with the substrate inhibition parameters under Case I conditions
(D:;::0.3). The simulations for two different initial conditions arc shown ill f,'igurc M8.3.

,,[tl,xlJ odc,45( 'bio' ,0,30, [1;lJ);

" [t2, x2 J - ode45 ( 'bio' ,0,30, [0.75; 2 J) ;

 536 Biochemical Reactors Module 8

a
a 5 10 15 20 25 30
time

"I 2

o ~---, _ .... -

o 5 10 15 20 25 30
time

FIGlJIU~ iVI8.J Substrate inhibition, Case I. xO:::o II, I] (solid), xO 0:;;: [0.75,2]
(dashed),

Although both initial conditions arc reasonably close to the Equilibriufll 2 solution
found in section 3, one simulation converges to Equilibriulll I (dashed line) while the
other converges to Equilibriulll :3 (solid line). We find in the next section that F~quiJih­
rillm 2is unstable: Furtber simulations will be performed and analyzed in the phasc-
plane (section 6),

M8.5 LINEARIZATION

In this section we find the linear state-space and transfer function models. So that there is
no confusion in notation, we will usc the following fonn:

i Az+BII
y ~ Cz

where:

21 = XI-XIs
2 2 = Xl - x 2s
II, ccc f) - D,
il Z x 2I -- Xl/i-

The st<:ltc-spacc matrices arc:

Sec. M8.5 Linearization 537

where it is assumed that both states are output.s. The notation tL,; is used in the A matrix to
represent the derivative of growth rate with respect to substrate concentration, evaluated
at steady-state:

aILs
rb: 2s
For the MOl/od model:

dr<, IJomaJ(m
(MS. IS)
(}X;:'I (kill + x::"J 2
and for the substrate inhibitioH model:

aILs V'11HI_~~~h (1 + ~1~J__' (2s)

dX 2s (kllJ + x 2.\" +- {(I-ttY
P-_I~l,!X (km + + k1xis) - J.LmllX X2\' (1 + 2k 1x 2s ) (MS.16)
(kll/ + X2,y + k 1xiJ 2

~ klxI,·)
1- ,_')2
k 1-'-2.1

M8.5.1 Substrate Inhibition Model

Here we analyze the substrate inhibition model under Case I conditions. A MATLAB
m-file, bio..._jac.m (Appendix 1), is used to generate the A matrix and the eigenvectors
and eigenvalues.

EQUILIBRIUM POINT 1
The sleady-slale value (seclion 3) is (x".x,) = (0,4).
The following command is entered:

»(jac, evec, lambda] :::; bio jac ([ a i 4] )

538 Biochemical Reactors Module 8

where j ac is the Jacobian (A matrix), evec is the eigenvector matrix and lambda arc
the eigenvalues.

jac :; :;
-0.1139 o
-0.4652 -0.3000

evec ;::;;
o 0.3714
1.0000 -0.9285

lambda ~
-0.3000 o
o -0.1139

so,

A ~ r.~O.1l39
~0.4652 - OU.300]
"j ~ ~O.3 ~j ~ [~]
"2 ~ ~ 0.1139
"
~2 ~-O.9285
I
0.3714 "I

Since both eigenvalues are negative, the system is stahle at equilibrium point 1, verifying
the simulation results shown in Section M8A.

EQUILIBRIUM POINT 2
The steady-state value is (xb,x,) ~ (0.995 I, 1.5 I22).

»[jac,evec, lambda] ~ bio~jac([0.9951;1.5122])

jac :; :;
0.0000 -0.0679
-0.7500 -0.1302

evec ;: ;
0.3714 0.2209
-0.9285 0.9753

lambda ~
0.1698 o
o -0.3000

The positive eigenvalue (0.1698) indicates that equilibrium 2 is unstable.

 Sec. M8.6 Phase~Plane Analysis 539

EQUILIBRIUM POINT 3
The steady-state is (x b ,x2,) = (1.5302,0.1746).

»[jac,evec,lambda] = bio~jac([1.5302;0.1746])

jac :::0

0.0000 0.9048
-0.7500 -2.5619

evec =:::

0.9492 -0.3714
-0.3147 0.9285

lambda =
-0.3000 o
o -2.2619

Both eigenvalues arc negative, indicating that equilibrium point 3 is stable.

M8.6 PHASE-PLANE ANALYSIS

The m-filc bio~phas_gen.m (Appendix 2) was llsed to generate the following phase-
plane plot for the substrate inhibition model under Case I conditions (sec Figure M8.4).
Notice that all initial conditions converge to either the washout steady-state (trivial solu-

2
o

1 -

o
o 0.2 0.4 0.6 0.8 1.2 1.4 1.6

x1

FiG-URE M8.4 Phase-plane plot for suhstrate inhihitioll model, Case I con-
ditions (x:;;:: stable steady-state, 0 = unstable steady-state).
 540 Biochemical Reactors Module 8

lioll. equilibrium I) or equilibrium 3; while cqliilihriutn 2 is a saddle point (ullstable).

These results arc consistent with the stability analysis of Section M8.).
A phase-plane plot for the Monad model was shown in Chapter 13.

MS.7 UNDERSTANDING MULTIPLE STEADY-STATES*

In this section we find analytically the steady-state solutions for the biorcactor model and
determine their stability.
The stcady-slate solutions (djdt = d,,1dt = 0) of (MH.!!) and (MH.12) arc:

o= (fL,· OJ x" (MH.I7I

o= f).I' (.t2j.i' - x 2.1,) ---.y.--

fLx"
(MHIHI

where the subscript s indicates steady-state.

'fhcrc arc two different types of solutions to (MS.I?) and (MS. IX). One is kno\vll as
the trivial or ;'washout" solution. The other type is the nontrivial solution.

M8.7.1 Washout Condition

l:;'rorn (M8.I?) and (M8.IS) we can immediately sec one solution, llslli111y called llw trivial
solution.
x" 0
x2_~ = X2{I' (MHIl)1

This is also known as the washout condition, since the reactor concentrations arc equal to
the feed concentrations; that is, there is no "reaction." Since there is no biomass in the
feed stream, then there is no biomass in the reactor under these conditions; all of the cells
have been "washed out" of the reactor.

M8.7,2 Nontrivial Solutions

From (M8.17), assuming that .r ls *- 0, then:

fL., = D, (ivlH.20)

which indicates that the specific growth rate is equal to the dilution rate, at steady-state.
From (MH.IH) we find thai:

(ivlH.211
y
and from (MH.20) and (MH.21):

X 1.\ (M8.22J

"'This section contains a detailed analysis which the reader may wish to skip on a first reading.
Sec. MS.7 Understanding Multiple Steady-States 541

We can solve for x 2s ' by llsing the relationship for IJ. s as a function of x 2s (either Monod or
substrate inhibition), since we kllow that IJ.- s =- D s (from (M8.20». Let !J..Jx2.\') represent
this general functionality. Then, we tHust solve:

"(c)··D
rs - 2.\ S
(M8.2.1)

for x 2s ' then substitute this value into (M8.22) to solve for .1: 1.1" The specific cases of
Monoc! and substrate inhibition arc shown in the subsections below.

MONOD
From (MS.13), the dilution rate at steady-state is

I-LJ!laxXh'
(M8.24)
kllJ + x 2s

Solving (M8.24) for x2s' we find:

(M8.25)

and since i-L s ;;;:;; D.I,

k ll!_ D.
I,
(M8.26)
iJ-max -- D,,_
For (M8.26) to be feasible, we note that D,I' < /L ma :c Actually, there is a more rigid require-
ment than that. 1'1'0111 (M8.22) we note that the highest value thal x2.1' can be is x 2f.i' other-
wise Xis will he less than zero. Thc maximum /).1' in reality is thcn f.1,Jx 2 ,..,J, (Jr (from
(M8.22), letting x2s ;;::; x 2JJ: .
ILJn'IX ..\:2D:
(Mollod) (M8.27)
kill + x2(\'
We also see from (M8.26) that there is a single solution for x2.\' as a function of D,I. 'rhis
means that there is a total of two steady-state solutions for the Monod model, since there
is also the washout (trivial) steady-state.

SUBSTRATE INHIBITION
We found in the previous subsection that there are two possible steady-stales for the
Monod model, for a given dilution rate. [11 this subsection we find the number or possible
steady-states for the substrate inhibition mode1.
From (M8.14) at steady-state:

/-Lmilx
j..L.\' (M8.28)
kill + x2. + 1, k JX~.I'

ESCOLA Dc E~:C;'!I'li\IW\
BIBLlQI60/\
542 Biochemical Reactors Module 8

l'rom (M8.28), we find thai:

k,xl, + (I - fL'''",) x" + k", ~0 (M8.2'J)

J.Ls

Since fL., = U, (M8.20), we snhslilutc into (M8.2'J) to find:

, , +
{(lXiI' (1 fL"'''')
,,-,
D,
+ I(m
X 21' - = () (M8.30)

Since (M8.30) is a quadratic equation, there will he two solutions for x2S" This means that
there arc three stcady~state solutions for substrate inhibition, since there is also the
washout (trivial) steady-state.
We sec from (M8.30) that for positive values of X2s the coefficient multiplying x 2.\
must be negative. The implication is that l.1 max must be greater than D s (the same result as
the Monod equation). This implication can be seen more clearly frolll the solution of the
quadratic formula for (M8.30):

_ fL"'''')'
D,
_ 4k I k III

(M8.31)
2k,
So, for solutiolls with physical significance:

D, > 4k 11/1
( 1 _fL""")' k (M8.32)

and (M833)

Because of (M8.33), \vc know that the tefm inside the hrackets in (M8.32) is negative. For
(M8.J2) to be satisfied, then we know:

( 1 _ fL'''''') < V4k 1/11

k (M8.34)
lJ,
which implies that:

_J-ll.!!<lX
Ds < 1 (substrate inhibition) (M8.3S)
+ 2'v(,:k
1\ 1/11

We could have found the samc result from viewing Figure M8.2. Notice that there is
a peak in the J..Ls curve, and again recall lhat D.I , ;::;: f.L r The steady-Slate dilution rate, D s
canuot be abovc the peak in the x2s versus J..Ls curve. We can find the peak by finding
ilfL,Idx2' ~ n. From (M8.16):
d/-1 s
(M8.36)
dX 2s
Sec. M8.7 Understanding Multiple Steady-States 543

We sec from (MS.36) that dfL./dX2' ~ () if:

(MS.37)

We can substitute this result into (M8,28) to find:

ik;"
fLlll(lx l 1
\X
(MS.3S)
km + !em k + 1 +- k m
'"

__ }l..l!1~1-"' __
fL., ~I (MS.3,)
+ 2 VX,k,,,
sO the maximum dilution rate (for the nontrivial steady-state) is:

f.111l<lX
D, 1+2'r.-;-
Vkkl/(m

which is the same result as eMS.3)).

M8.7,3 Summary of Steady-State-Monod and Substrate Inhibition

WASHOUT (BOTH MONOD AND SUBSTRATE INHIBITION)

Both Monod and substrate inhibition models have a washout (trivial) steady-state:
XIs :~~ 0 (MS. I')

NONTRIVIAL STEADY·STATE FOR MONOD

The nontrivial steady-state solutions for substrate and biomass arc:

kI/lD,\.
(MS.26)
f.Lm<lx - '[)",

Y (x 2I, - -"') (MS.22)

with the requirement that D s < IJ. max 'Y~f.Jklll + x~/.i' (that is D.I, < f.L.lr2j.J)

NONTRIVIAL STEADY-STATES FOR SUBSTRATE INHIBITION

The two nontrivial steady-state solutions for substrate arc;
 544 Biochemical Reactors Module 8

1(
\ .1 _ D 1:'',' ' )' - 41\1 1(/II
(M8.31)
2/(,
unci the associated biomass concentration is:

x" CC" Y (X 2!, - x,,) (M8.2S)

with the requirement for dilution rate:

(MX.32)

M8. 7.4 Stability of the Steady-States

The stability of each steady-stale solution is determined from the eigenvalues of the Jaco-
bian matrix (matrix A in the state-space form), For a two-state system we know that the
eigenvalues are found by:

dCl(Al- A) X' - lr(A) X + del(A) ~ () (M8AO)

j::"wm Chapter 13 we know that the following conditions mllst be satisfied for stability of a
second-order system:

lr(il) < 0 (MXAI)

del(A) > 0 (MXA2)
That is, the eigenvalues (A) will be negative if conditions (M8.41) and (M8.42) arc satis-
fied. The Jacobian of the biorcactor modeling equations (M8.11 andM8.12) is:

A C" [1:'.' - D, x "I:': J (M8A3)

-ij; ,- f)" _ ..lJv'r
,I
y,),"
where we have used the notation /-L,; to represent the derivative of growth rate with re-
spect to substrate concenlration, evaluated al steady-stale:

i!/-L,\ (M8A4)
c)x:2y

The trace and determinant of Ii are:

triA) ,c (I:' - D) _ J) _ I:';x" (M8A5)

" ., .\ Y

det(A) "" - (I:' _ D) (D

.\ .\ .'
+ /L; x,,) + X,dL;I:',
Y Y (MXA6)

We will usc (M8AS). (M8A6), and the eouditions shown in (M8AI) and (M8A2) 10 dc-
termine the stability of each steady-state.
Sec. MS.7 Understanding Multiple Steady·States 545

STABILITY OF WASHOUT STEADY·STATE

Under washollt conditions: X2.\':::: '\.2/.'1' and xis::::: o.
For stability of the washout steady-state, the following criteria then must be met.
r"irsl, from the requirement that tr(A) < 0:

fL,-D,-D,<O (MXA7)

and from the requirement that dct(A) > 0:

.. (fL, - DJ(D') > 0 (MXAX)

F"rom (M8.47) we sec that the requirement for stability is thell:

D > P'.I' (MXA9)

.\ 2

while from (M8.48) the requirement for stability is:

D s ::> /-Ls (MX.50)

Notice that f..L s is evaluated at the substrate feed concentration for the washollt condition.
Perhaps the expression I-L,lx2/) should be llsed to designate this relationship. Comparing
(M8.49) and (MS.50), we sec that (MS.50) is the more rigorous requirement for stability
of the washout steady-state.
'fhe growth rate expression for Monod kinetics is:

(MX.51)

while for substrate inhibition kinetics:

(MX.52)

Notice that IJ-.lr 2j:) is simply a shorthand expression for the specific growth rate evaluated
at the suhstrate feed concentration. We must use (MS.50) along with either (M8.51) or
(MS.52) to determinc the stability or the washout steady~state. Notice that the washout
steady~state will only bc stable if D.I , is high enough. Wc can think of D.I , as a dynamic bi~
rurcation parameter, becausc the stability of the washout steady-state will depcnd on the
valuc of the dilution ratc.

Stability of Washout Steady-State for Monad. From (MX.50) and (M8.51 l. Iht
washout stcady-statc will be stable if:

and unstable if:

(MX.54)
546 Biochemical Reactors Module 8

Stability of Washout Steady-State for Substrate Inhibition. From (MS.50)

and (MS.52), the washout sleady~state will he stable if:

and unstable if:

(MS.56)

NONTRIVIAL STEADY~STATES

For the nontrivial steady-states, D s = fL.\.. The stability requirerncnts for the nontrivial
steady-states arc then:
II f t:
- D _~·\~L>: <0 (MS57)
s Y

from the trCA) specification, and

(MS.5S)

from the det(A) specification. Since D.I. (and therefore 1J..), xis' and Yare positive, (MS.57)
and (MS.58) rcdu~c to the requirement that:

fL.: > 0 (MS. 59)

for slability.

Stability of Monod at the Nontrivial Steady-State. From (MS.24):

[1; -_ ...• ILm<lxklll

.. . ._, (MS.60)
. (k m + x 2 \)-
We see ilnmediatcly that fL,: is always positive for the Monod model at the non-
trivial steady-state; therefore, the nontrivial steady-state is always stable. Recall that
D.\. < fL.1.(X2!1.) for a nontrivial steady-state solution.

Stability of Substrate Inhibition at the Nontrivial Steady-States. We can

tcll from the substrate inhibition eurvc in Figure M8.2 that a steady-state that is on the
left side of the peak will be slable (since fL.:: > 0), while a steady-state on the right side
will be unstable (since f.<
< 0).
Numerically, from (MS.36):

~_!lF1X _(k'II=!~J~~3,~:2. _ (MS.61 )

(kll/ + .T2I. + k]xisf
Sec. MS.7 Understanding Multiple Steady-States 547

The x2s that is on the left side of the peak, and is therefore slable, is (from (MS.31»:

(fLO'' ' _1) - ~(I _iL",,,,r - 41< ,I<",

(M8.62)
21<,
The x 2s that is on the right side of the peak, and is therefore unvtahle, is (from (MS.31 »:

(fL'D,' ' ' _I) + _fLO'

D ' ' )' _41< I<
., L III
X2~ = . (M8.63)

Also, recall that Ds < fJ-ma/1 + 2Vk--I)~::: (which is equivalent to requiring a real nontrivial
solution).

M8.7.5 Case 1 (Ds = 0.3)

The reader should find the following results:

M0I10d

Equilibrium I-washout Xis::::::0 X2,1'::::: 4.0 unstable

Equilibrium 2-nontrivial xl.> = 1.5374 x,., = 0.1565 stable

Substrate Inhibition
Equilibrium i-washout .t Is =0 X2, = 4.0 stable
Equilibrium 2-l1ol1trivial xL, = 0.9951 x2, = 1.5123 unstable (saddle point)
Equilibrium 3-nontrivial xl.> = 1.5302 x2, = 0.1745 stahle

M8.7.6 Case 2

For a steady-state dilution rate of D.).::::: 0.15, the reader should find the following results:

Monod
Equilibrium l~washout XLI' =0 X2.\' = 4.0 unstable
Equilihrium L~nontrivial Xli = 1.5811 x" = 0.0474 stable

Snbstrate Inhihition
Equilibrium I-~washout XL, =0 4,0
X2.\':::;: unstable
Equilihrium 2~nontrivial Xli = -0.6104 x2, = 5.5261 not feasible
Equilibrium 3-nontrivial XL, = 1.5809 x" = 0,()478 stahle

Although there is a mathematical solution for cquilihrium 2, it is not physically feasible,

since it corresponds to a negative biomass concentration.
548 Biochemical Reactors Module 8

M8.7.7 Case3

r'or a steady~statc dilution rate of D s ::::: 0.6, the student should find the follmving results:

Mouod
Equilibrium l~~washout XIs:::: 0 X2, ~ 4.0 stable
Equilibriulll 2-nontrivial x",~2.0114 xl., ~ ~ I J)286 not feasible

StC;:HJy~state 2 is not feasible. because it corresponds to a negative substrate concentration.

Substrate Inhihition
EquilihriuI11 l,,--,washout XIs::::: 0 X2., ~ 4.0 stable
Equilibrium 2-11ontl"ivial x'" ~ 4.13 -
0.20j x2, ~-0.13 + 0.50j not feasible
Equilibrium 3-nontrivial x,-,~4.13 +0.2(lj x,., ~ -0.13 - 0.50j not feasible

The second and third steady-states (Irc not feasible because the concentrations for both the
biomass and the StJbslratc are complex.
There <Irc some vcry interesting changes ill the dynamic behavior of these lnodcls as
we vary the dilution rate (again, we can think of dilution rate as a bifurcation parameter).
Let us discuss this in order of the lowest dilution rate to the highest dilution rate.

LOW DILUTION HATE

Case 2 had the lowest dilution rate (D,I' ::;:; 0.15). The Monod model has two steady-
slates-the washout steady-state is unstable and the other (nontrivial) steady-state is sta-
ble. This means that any set of initial conditions will eventually converge to the nontrivial
steady~state, for the Monod model. The substrate inhibition model has only two feasible
steady-states-- the washout steady-state is unstable and the high conversion steady-state is
stable.
The interesting result is that at low dilution rates, the substrate inhibition model be-
haves like the Monod model.

MEDIUM DILUTION RATE

Case I had the next highest dilution rate (D,I' ::::: 0.30). The Monod model has two steady-
states----the washout steady-state is unstable and the other (nontrivial) steady-state is sta-
ble. 'fhis means that any set of initial conditions will eventually converge to the nontrivial
steady-state, for the Monod model. I'he substrate inhibition model has three feas-
ible steady··states. The w,-1shout (no conversion) steady-state is stable, the medium conver-
sion steady-state is unstable and the high conversion (low x 2) sleady~state is stable. This
means that any set of initial conditions will converge to one of the two stable steady-
states. A phase-plane mllst he drawn to determine if a particular set of initial conditions
will lead to washoul.
 Sec M8.8 Bifurcation Behavior 549

HIGH DILUTION RATE

Case 3 had the highest dilution rate (D\, ::: 0.(0). Both models had only one feasible
steady-slate, the \vashout steady-state, and it was stahle. The student should he able to
"sketch" phase planes for all three conditions for each of the models (Monad and sub-
strate inhibition).

M8.8 BIFURCATION BEHAVIOR

The conditions for stahility developed in Section MS.7 can be used to develop stcady-
state input-output diagrams for the numerical example presented in the previous sections.

M8.8.1 Diagram for the Monad Model

The diagram for the Monod model is shown in Figure M8.5. As calculated, the Monod
model has two steady-states for dilution rates that arc less than the specific growth rate
under the Iced conditions, D.I, < l-L.Jx2J-). The nontrivial steady-state is stable under those
conditions, while the washout steady-slate is unstable. For Ds > fL/:t2f.J there is a single
steady-stale, lhe washoul steady-state, and it is stable. '

M8.8.2 Diagram for the Substrate Inhibition Model

The diagram for the substrate inhibition model is shown in Figure M8.6. At low dilu-
tion rates, where D s < fL,JX 2f.i.) , there are two steady-slales (like the Monod model). The
nontrivial steady-slate is stable under those conditions, while the washout steady~_~t~~!C
is unstable. Por tbe intermediate dilution rate range, 1J-./or2j) < D s < /--L rna Jl + 2Vk 1k/ll'

Monad Model
5
unstable stable
4r----------------....--j
3

o-------~
o 0.1 0.2 0.3 0.4 0.5 0.6
dilution rate

FIGURE M8.5 Input-output diagram for the Monad model,

550 Biochemical Reactors Module 8

Substrate Inhibition Model

5
stablo

\
,,
3 ,,
2 ,
'\. unstable

" "-
"-
"-
0'L- stable - .->
o 0.1 0.2 0.3 0.4 0.5 0.6
dilution rate

FIGUHE M8.6 Inplll~outpnt diagram for the substrate inhibition model.

there arc three stcady~statcs. Two of these arc stable, while one is unstable. The stable
steady-slate that is attained will depend upon the initial conditions of the concentra~
lions, or on the way that the process is started lip. When the dilution rate meets the con-
dition that D,I_ > f.1.Jx 2j ), there is a single steady-state, the washout steady-state, and it is
stable.

M8.8.3 Hysteresis Behavior for the Substrate Inhibition Model

It is interesting to note that the way that the biorcactor is started up will determine the
steady-state concentrations that the reactor achieves.L,ook at Figllre M8.6. NotiC\.,~ that if
we start at a very low dilution rate we will have only one stable steady-state, so the reactor
IIltl:;;! operate at that condition. If we slowly increasc tJ.l~~tilution ratc, we relnain on the
lowcr curve of r'igure M8.6. When V,I' > ~ll1a/1 + XVk1kl/i (1\ : ;: 0.36126 for this exam-
ple), the stable solution suddenly "leaps" to the upper stahle steady~state (washout cOIl(Ii~
tions). As we increase Ds Luther, we remain on the washout curve.
Now, assume that we arc starting out at a high dilution rate along the upper curve,
the washout conditions. As we slowly decrease the dilution rate, we remain on the
washout curve until D,I' ::::: ~,Jx2Jj')' which is D,I' ::::: 0,186] for this example. The stahle
steady-state then "jumps" down to the lower curve. As we continue to deCl'case the dilu-
tion rate further. we remain on thc lower curve.
The type of behavior shown in Figure M8.6 is known as hysteresis and is exhibited
by a number of processes, including exothermic chemical reactors and valves that "stick."
The chemical reactor example is discussed further in Module 9,
The student should be able to show how the phase-plane behavior changes as a
function of dilution rate, for the example shown in Figure M8.6.
Student Exercises 551

SUMMARY

The modeling equations for a biochemical reactor were developed for Monod and sub~
strate inhibition kinetics. We found that the Monod lnodel normally has two stcady~state
solutions, while the substrate inhibition model normally has three steady-state solutions.
At low dilution rates the substrate inhibition model behaves similarly to the Monod
mode], with a single stable steady~state. Washout will not be a problem at the low dilution
rates.
At medium dilution rates the substrate inhibition model behaves quite differently
from the Monod model. Depending on the initial conditions, the reactor will either con~
verge to a high conversion or to washout conditions for the substrate inhibition model. II
has not been discussed thus far, but if we wish to operate at an intermediate (unstable)
conversion level, then feedback control must be llsed. Notice that the Monod model still
has only one stable point, and there is no danger of wash-out.
At high dilution rates, both reactor models have only one feasible solution---
washout. The flow is simply too high (residence tinle too low) for any cell growth.

FURTHER READING

An excellent source for an introduction to biochemical engineering is:

Bailey, .I.E., & D.F, Ollis. (1986), Biochemical Engineering Fundarnentals, 2nd ed,
New York: McGraw-Hill.

STUDENT EXERCISES

1. In this module we developed the modeling equations assuming that no biomass is

fed to the reactor, Analyze the system studied for the case where the biomass feed
concentration is 2.5% of the substrate feed concentration (so xII::::: 0.1 for the nu-
merical values used ill this module).
Is there still the possibility of a washout steady-state?
2. Modify bio, phas_gen.rn and bio.rn to perform a phase-plane analysis for
cases 2 and 3 with the substrate inhibition model.
3. Data for specific growth rate coefficient as a function substrate concentration for a
biochemical reactor are shown below:

fh, hrA-1

o o
0.1 0.38
0.25 0.54
552 Biochemical Reactors Module 8

0.5 0.63
0.75 0.66
I 0.68
1.5 0.70
3 0.73
5 0.74

a. Estimate the parameter values for a Monod model (kill' ~max)

b. The production rate of cells (biomass) is DBxl. Find the steady-slale value of
the dilution rate that maximi7.es the production rate of cells. 'fhe substrate fecd
concentration is 5 gllitcr.
c. Find the steady-state concentration of biomass and substrate at this dilution rate.
d. Find the linear state-space model at this dilution rale, with dilution rate and sub-
strate feed concentration as the input variables. Also find the transfer function
relating dilution rate to biomass concentration.
e. Simulate the responses (using the nonlinear dynamic model) or the conccnln.l"
lions or biomass and substrate to step increases and decreases of IWit;, in the di~
lution rate (changes are from the dilution rate found in b.). Compare these re~
suits with those of the linear system (remelnher to convert deviation variables
back to physical variables).
4. In this module we have analyzed how the biomass and suhstrate concentrations
change depending on the dilution rate. If the purpose of a particular biochemical re~
actor is to produce cells, then we arc more concerned with the production rate of
cells. The' production rate is mass of cells produced per unit time:

steady-slale production rate of cells = D/'xls

For both the Monod and substrate inhibition models present(~din this modale, find
the dilution rate that maximizes the production rate of cells. Analyze the stability of
the reactor under this condition.
5. Consider a biochemical reactor where the consumption of substrate (\2) promotes
the growth of biomass (Xl) and formation of product (x 3 ). The three modeling eqlla~
lions arc:

dX I

lit

lil
where the specific growth rate is a function of both the biomass concentration and
the product concentration:
 Appendixes 553

with the following parameter values:]

Variable Value Variable Value

y 0.4 gig 2.2 gIg
0.2 hr-] "!J. OAg lu·-- J
f' max
Pm 50 g/liter kill 1.2 g/litcr
k,
IJ 0.202 hI'
,
0.04545Iitcr/g x21
xJ
20 gil iter
6 gIl iter
-"1'-2 5 glliter J.-:; 19.14 g/liter

H. Compare and contrast this rnode! with that of the two-state model with substrate
inhibition kinetics presented in this module.
h. Verify that the steady-state values for Xl' x 2 , and x] presented in the table above
are correct. For a steady-state input of f):::;: 0.202 (and all of the other parameters
constant), arc there (my additional solutions for the states (for example, the triv-
ial solution?). Analyze the stability of all steady~state solutions obtained.
c. Perform dynamic simulations of the nonlinear model, with step changes of
± 10(#) in the dilution rate. Discuss the resuIts or your step changes (Le., docs an
increase or ~lecrease in f) have a greater effect on the biomass concentralion?).
Compare your results with linear simulations.

APPENDIXES

1 Steady-State Biochemical Reactor Model, bio~ss. m

function f :::, bio~ss (x)
%
% b.w. bequette
% (e) 16 Nov 92
% revised 18 July 96
%
% find steady-states of bioreactor, using fsolve:
%
% x::::: fsolve( 'bio' ,xOl

IThis model is from Chapter 4 of the following Illonograph:Henson, M.A., & D.E. Seborg
(cd.). (1997). Nonlinear Process Conlrol. Upper Saddle River, NJ: Prentice-Hall.
 554 Biochemical Reactors Module 8

% where xO is a vector of initial guessef3

% x(l) biomas,;.>
~8 x (2) substrate
%
% biomass ("bugs") consumes the substrate

CI, D dilution rate (F/V, timeA-l)

% y yield biomass/substrate
96 mu specific DrowLh rate
'l, mumax param(:!Ler (both Monad and Substrate Tnhibi tion)
~f, km parameter (both Monad and Substral~e Inhibition)
"6 kl parameter (Substrate:; Inhibition only, kl : : : () for Monod)
'r, sf subf3trate f(x~d concentration
'?6
% the function vector consists of 2 equations
%
f ~ zeros (2, 1) ;

96 parameter values

D ~ 0.3;
mumax 0:: 0.53;
Y ~ 0.4;
km 0.12;
sf 4.0;
kl 0.4545;

% Substrate Inhibition expression for specific growth rate

%
InU ~ mumax k x(2)/(km+x(2)+kl*x(2)*x(2));
90

% steady-state equations
% [solve varies xll) and x(2) to drive f(l) and f(2) to Zero

f(l) (mu D)*x(l);

f(2) (sf - x(2) )'D - mu*x(l) /Y;

2 bio.m, Function File for Dynamic Simulation Using ode45

function xdot = bio(t,x}

% b.w. bequett.e
% (c) 18 ~July 96
%
5(, dynamic equations for biorcactor, integrated using ode4.5,
(i; usiuq the following command

% [t,x] = ode4.5( 'bio' ,t.O,tf,xO)

 Appendixes 555

% where to is the initial time (usually 0), tf is

% the final time and xO is the initial condition vector
% xO (1) biomass initial conclition
% xO(2) substrate initial condition
%
96 biomass ("bugs") consumeS the substrate
'6
% state variables
%
% x(l) biomass
't x(2} substrate
q.
"
90 D - dil.ution rate (P/V, timeA-l)
% y yield biomass/substrate
% mu specific growth rate
% mumax - parameter (both Monad and Substrate Inhibition)
% km parameter (both Manod and Substrate InhiJyit.1.on)
% kl parameter (Substrate Inhibition only, kl :::; 0 for Monad)
96 sf substrate feed concentration
%
% the function vector consists of 2 equations

I zeros(2,1);
%
% param(~t~(,~r values
%
D O . 3;
mumax 0.53;
Y = 0.4;
km 0.12;
sf - 4.0;
kl 0.4545;
%
% Substrate Inhibit_ion expression for specific growth rate

mu :::; mumax-k x(2) / (km+x(2) +k1-k x(2) *x(2));

1-s
90 dynami c equa t i ODS

xdOL (1) {mu D)'x(l);

xdot(2) (sI x(2))*0 - mu*x(l)jY;

3 Phase-Plane Plot for Biochemical Reactor, bio---phas_gen.m

:6
'A b.w. bequett.e
'6 (e) 19 July 96
%
% generates phase-plane plots for the bioreactor
 556 Biochemical Reactors Module 8

%
% set-up the axis limits
%
axis{[O 1.75 0 5]);
%
% stable and unstable points for substrate
% inhibition model
%
xlu [0.9951];
x2u [1.5122];

xls [0;1.'j302J;
x2 s [4; 0 . 1'146] ;
%
% place an 'x' on stable points
% place a 'a' on unstable points
%
plot{xlu,x2u, 'wo' ,xls,x2s, 'wx')

hold on
%
% select different initial conditions
% xl ranges from 0.1 to 1.5 (every 0.35)
:!, x2 ranges from a to 5 (every 1.25)
% total of 16 initial conditions
%
xlinit [0.1 0.45 0.8 1.15 1.5 0.1 0.45 0.8 1.15 1.5];
x2 ini t [0 0 0 0 0 5 5 5 5 5] ;
xlinita [1.5 1.5 1.5 0.1 0.1 O.lJ;
x2 lni ta = [1.25 2. 5 3.75 1.25 2. 5 3.75] i
%
xG = [xlinit xlinita;x2init x2inita] ;
%
% ncol = number of initial conditions
%
[mrow,ncol] size(xO) i

% run simulations for each initial condition

%
for i ;:c. 1:nc01;
%
[t,xJ "ode45('bio',0,30,[xO(:,i)]);
%
plot (x ( : ,1) ,x ( : ,2) I 'w' )
%
end
%
xlabel ( . xl ' )
ylabel ('x2')
hold off

______________.L !
Appendixes 557

4 Direct Calculation of the Eigenvalues

Here we calculate the eigenvalues of the nontrivial solution:

del(A/- A) = A2 - (r(A)A + dc(A) = 0 (MXAO)
where the trace and determinant arc

(MXA5)

(MXA6)

For the nontrivial solution, 11-.1'':;:::: f),I':

11-,;'X 1,1'
(r(A) c= -D, - Y (MX.AI)

X Is 11-,~. 11-.1'.
I (A) ,-
(cl (MX.A2)
Y
The rools of (MXAO) arc:
- 4 de( A
A- (MX.A3)
2

(MX.A4)

+ 2 4 .r,_ fL,;' /.1",

2
+ y Y
(MX.A5)
2
2 x.,_ /-L,;' /Ls
y
(MX.A6)

A=
± ~(=;L, - x';:r (MX.A7)
2

y ± (I~'_ x'p:)
A= (MX.AX)
2
and our roots are:
558 Biochemical Reactors Module 8

A.[ = --/1-)' and (MX.A9)

and since /1-.1' ::::: D,I" we are assured that one pole will always he negative. The second root
will only be positive if /1-.: is negative. Since /1-.: is positive for the Monod model, the
nontrivial solution is stahle as long as the solution is feasible (D.I, < /1-./x2f~.)). The fL; can
be either positive or negative for the substrate inhibition model, so a J1()[]trivial slcady-
state may either be slable or unstable.