Chemical Engineering Science 61 (2006) 3850 – 3857

www.elsevier.com/locate/ces

On the multiple solutions of the flash equations

Miguel Vaca a,b , Arturo Jiménez a,∗ , Rosendo Monroy-Loperena c
a Departamento de Ingeniería Química, Instituto Tecnológico de Celaya, Av. Tecnológico y A. García Cubas S/N, A.P. 57, Celaya, Gto. 38010, México
b División de Ingeniería Química y Bioquímica, Tecnológico de Estudios Superiores de Ecatepec, Av. Tecnológico S/N, Ecatepec 55210, México
c ROMON, Paseo de los Pirules 124, México 04250, México

Received 3 April 2004; received in revised form 19 May 2005; accepted 29 June 2005
Available online 9 March 2006

Abstract
Input multiplicity is an important feature of industrial processes. In this paper, we extend the work by Tiscareño et al. [1998. Multiplicity
of the solutions of the flash equations. Chemical Engineering Science 53, 671–677] to analyze multiple solutions of the flash equations when
a product mole fraction of one component is specified and the vapor fraction or the temperature is calculated. For this type of multiplicity to
arise, at least one stationary point for the mole fractions with respect to vapor fraction or temperature must exist. By using a constant relative
volatility model, we demonstrate the uniqueness of the stationary point, and show the necessary and sufficient conditions for the existence of
such a stationary point. From these results one can state that systems with a constant relative volatility assumption can only present second-order
input-multiplicities.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Distillation; Modeling; Flash; Multiplicity; Multiple solutions

1. Introduction solutions in isobaric flash separations and showed that the

solution involving a multicomponent homogeneous mixture is
The existence of steady-state multiplicity (SSM) has an im- unique. The monotonic behavior of the flash equations was
portant role in the design, simulation and control of industrial demonstrated for some isobaric separations. However, the
processes. Güttinger and Morari (1996) state that the general uniqueness of the solution with respect to product composition
notion of multiplicity is that a system with specified variables was not explored. Tiscareño et al. (1998) studied the SSM of
that meet its degrees of freedom exhibits different solutions at the solutions when a product mole fraction is specified and
steady-state. The multiplicity is called input multiplicity when the vapor fraction or the temperature is calculated. Follow-
two or more sets of input variables (flows) provide the same ing the results of Tiscareño et al. (1998), Monroy-Loperena
output condition. Jacobsen (1994) discussed the implications (2001) suggested a fast method to check steady-state input
of input multiplicity for the system dynamics. Koppel (1982) multiplicities in vapor–liquid flash separations.
gives an excellent discussion on why the input multiplicity im- Rodríguez et al. (2001) studied the steady-state behavior of
poses more practical control problems than output multiplicity. an isobaric, adiabatic reactive flash for binary mixtures. They
Several studies have been done on SSM and stability in found that the existence of multiple steady states and their sta-
distillation columns (see for example the works by Tiscareño bility properties are related to a dimensionless quantity that de-
et al. (1998) and Zheng et al. (1998) for nonreactive cases, and pends on the heat of reaction, vaporization rate and the mixture
by Güttinger and Morari (1999) for the reactive case). How- composition. The same group extended their work to study the
ever, only a few studies have been conducted on multiplicity parametric dependence of multiplicities for reactive flash sys-
in flash separations. Lucia (1986) studied the uniqueness of tems (Rodríguez et al., 2004).
This work deals with input-multiplicities in vapor–liquid
∗ Corresponding author. Tel.: +52 461 611 7575; fax: +52 461 611 7744. flash distillation (FD) when a product mole fraction is specified.
E-mail address: arturo@iqcelaya.itc.mx (A. Jiménez). As shown in Fig. 1, for this type of multiplicity to exist, one or
0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.06.036
 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857 3851

2. A flash distillation model based on constant relative

volatilities

SP Consider the isobaric FD shown in Fig. 2. The typical situ-

ation for a two-phase vapor–liquid equilibrium problem is to
solve the system of material balance, equilibrium, and consis-
(i.e. product composition of component i)

tency equations given by

C
F zi = V y i + Lx i , (1)
Output Variable

yi = Ki xi , (2)
NC


zi = 1, (3)
i=1

NC


xi = 1, (4)
i=1

NC


yi = 1, (5)
i=1

where F, V and L are molar flow rate of the feed, vapor and
A B liquid streams, respectively; xi , yi and zi are the mole fractions
Input Variable of component i (i = 1, . . . , NC ) in the liquid, vapor, and feed
(i.e. vapor fraction or temperature) stream, respectively; and Ki is the vapor–liquid equilibrium
Fig. 1. Input multiplicity.
coefficient of component i (Seader and Henley, 1998). When a
thermodynamic model with constant relative volatility is used,
the liquid and vapor mole fractions of component i are related
by
more stationary points SP must exist if two (or more) inputs A
i xi
and B (vapor fraction or temperature) provide the same output yi = N , (6)
C
j =1
j x j
condition C (outlet composition of one component). Because
input variables are those that can be manipulated by controllers where
i is the relative volatility of component i. Then,
to keep output variables at their specified values, potential prob- vapor–liquid equilibrium coefficients are only functions of
lems during process operation can arise. liquid mole fractions,
Tiscareño et al. (1998) analyzed the mass balance and phase

i
equilibrium equations that describe the isobaric FD of a non- Ki = N . (7)
j =1
j xj
C
azeotropic system well below the critical states of the mixture;
they showed analytically that at least one stationary point
(maximum) for mole fractions with respect to temperature Another useful, although not commonly used, equilibrium
or vapor fraction may exist in the interval from the boiling relationship for constant relative volatility systems can be de-
temperature of the most volatile component to the boiling rived as follows. Dividing Eq. (2) for component i by the same
temperature of the least volatile component. By analyzing equation for a base component b, one can obtain
the solution of the Rachford–Rice function, they determined yi Ki xi xi x b yi
the necessary and sufficient conditions under which multiple = =
i → xi = . (8)
yb K b xb xb y b
i
solutions can exist.
This work can be seen as an extension of that by Tiscareño Using the consistency Eq. (4), it follows that
et al., since we seek to find the order of the input multiplicity, NC
that is, how many solutions can be found for a component for xb
yj
= 1.
a given mixture. By using a constant relative volatility model, yb
j
j =1
we demonstrate that a second-order input-multiplicity can exist;
the conclusion arises because of the uniqueness of a stationary After obtaining the quotient xb /yb and substituting into Eq. (8),
point that defines the SSM in FD. Although the results obtained one finally gets,
using constant relative volatilities are only approximate, they
yi /
i
provide a fast and useful insight for the analysis of ideal or xi = N (9)
j =1 yj /
j
C
nearly ideal systems.
3852 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857

V 2.2. Solution for the F, zi ,
i ,  specification

yi Let us take the case when the degrees of freedom are met by
specifying the set F, zi ,
i , . The problem is to solve Eq. (11)
coupled with Eq. (7) to find the component mole fractions in
the liquid product. Alternately, we can solve Eq. (12) coupled
with Eq. (10) for the vapor mole fractions. Using Eq. (11), we
have
T, P zi
fi (x) = xi − = 0, i = 1, . . . , NC , (13)
F
1 + (Ki − 1)

zi where x = (x1 , . . . , xNC )T is the vector of liquid mole frac-

tions. The system of non-linear equations (13) can be solved
with a Newton–Raphson iterative method (see for instance
Constantinides and Mostoufi, 1999). In such case, the elements
of the Jacobian matrix are given by
jfi zi  dKi
= ij + ,
jxj [1 + (Ki − 1)] dxj
2

where ij is the Kronecker delta and

dKi
i
j
= − N = −Ki Kj .
dxj ( C

m x m ) 2
m=1
L
xi Typically, three to five iterations were needed in this work to
get convergence for a Euclidean norm of the vector of functions
Fig. 2. Schematic representation of a flash distillation. lower than 10−6 , with x = z as initial estimate, for any value
of the vapor fraction lower than 1. We applied this procedure
to analyze how the product mole fractions change with respect
from which we obtain the Ki values as a function of the vapor
to the vapor fraction in the interval 0

1.
mole fractions
NC

3. The flash multiplicity problem
yj
Ki =
i . (10)

j Let us consider the case when one product mole fraction, say
j =1
xr or yr , is specified and it is desired to find the vapor fraction
Eqs. (1) and (2) can be written as or temperature. For this case, Tiscareño et al. (1998) stated that
zi more than one physical solution could exist. Therefore, one or
xi = , (11) more stationary points must exist for the product mole fractions
1 + (Ki − 1)
with respect to the vapor fraction or temperature. Following the
z i Ki uniqueness analysis of Lucia (1986) that states that only one
yi = , (12)
1 + (Ki − 1) equilibrium temperature exists for an equilibrium vapor fraction
and vice versa, we can focus our analysis in the vapor fraction
where  = V /F is the vapor fraction. Notice that if we use domain without loss of generality. The FD model based on
Eqs. (11) and (12) instead of Eqs. (1) and (2), the consistency constant relative volatilities is used for that purpose. It should be
equations (4) and (5) are dependent equations. noted that, instead of the Rachford–Rice equation, component
mole balances are used. Also, the analysis is carried out only
2.1. Degrees of freedom analysis for the flash interval, i.e. from the bubble point to the dew point
of the feed mixture.
Several alternative sets of independent equations can be Using Eq. (11) the derivatives of liquid mole fractions with
considered from Eqs. (1)–(12). One of them is formed by respect to the vapor fraction are given by (see Appendix A)
Eqs. (1)–(4) and (7) with 2NC + 1 degrees of freedom, which
dxi x2
can be met by specifying F,
i , NC − 1 values of zi , and either = i fi (), (14)
the molar flow rate of one of the products or one product mole d zi
fraction (either in the vapor or the liquid phase). For the use of where fi () is defined as
such formulation, one must include mean relative volatilities
applicable to the flash interval, evaluated at the flash pressure dKi
fi () = 1 − Ki −  (15)
and mean temperature, to describe appropriately the system. d
 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857 3853

or Proof. The auxiliary functions fi () or gi () are monotone-

decreasing, so at the most one zero of fi () or gi () can exist
fi () = 1 − Ki  (16) in the interval 0

1. Thus, from Lemma 1 the components
with  given by (see Appendix A) of the mixture may have one and only one stationary point
in the liquid or the vapor mole fractions in the flash interval
NC 2
x /zm 0

1, that is, the stationary point is unique. 
 = N m=1 m ,  1. (17)
m=1 (xm /zm )Km
C 2
From Theorem 1 we can state that at the most two solu-
In a similar fashion, for the vapor phase: tions can exist for the vapor fraction if we specify the prod-
uct mole fraction of one component. Notice that the stationary
dyi y2 point could exist outside the flash interval 0

1, as shown
= i gi (), (18)
d zi by Tiscareño et al. (1998); in such work, the interval of anal-
  ysis ranged from the boiling temperature of the most volatile
1 dKi 
gi () = 2 (1 − ) − Ki (Ki − 1) = − 1, (19) component to the boiling temperature of the least volatile com-
Ki d Ki ponent.
NC 2 In addition, Eqs. (16) and (17) for the liquid case, or (19)
y /zm
 = N m=1 m ,  1. (20) and (20) for the vapor case, provide the following theorems.
m=1 (ym /zm )(1/Km )
C 2

Theorem 2. The most volatile component of a multicomponent

Several observations can be made from Eqs. (14), (16) and
mixture cannot have a stationary point in either the liquid or
(17) for the liquid phase and from Eqs. (18)–(20) for the vapor
the vapor phase.
phase, which are conveniently organized as lemmas and theo-
rems. The functions fi  and gi () are called auxiliary func-
Proof. For the most volatile component the quantities K1  and
tions. In addition, it is assumed that the NC component (i.e.
/K1 are given as
the heaviest component) is the reference component for relative
 C 2 NC
volatilities. K1 N m=1 xm /zm (x 2 /zm )K1
K1  = N = Nm=1 m
m=1 (xm /zm )Km m=1 (xm /zm )Km
C 2 C 2
Lemma 1. The sign and the zeros of the mole fraction deriva-
NC
tives dxi /d and dyi /d are the sign and the zeros of the aux- (x 2 /zm )
1
iliary functions fi () and gi (), respectively. = Nm=1 m ,
m=1 (xm /zm )
m
C 2

Proof. From the RHS of Eqs. (14) and (18), one may notice 1 NC NC 2
m=1 ym /zm
2
 K1 y /zm
that the first term is always positive. Then, the sign and the = N = N m=1 m .
zeros of the mole fraction derivatives dxi /d and dyi /d are K1 C
(ym /zm )(1/Km )
2 C
(ym /zm )(
1 /
m )
2
m=1 m=1
the sign and the zeros of fi () and gi (), respectively. 
A term-by-term comparison of the numerator and the denom-
inator of these equations shows that, since
1 /
m 1, then
Lemma 2. The functions fi () and gi () are monotone-
K1  > 1 and /K1 < 1. From Eqs. (16) and (19), both func-
decreasing functions in the interval 0

1.
tions f1 () and g1 () are always negative. From Lemma 1, it
follows that both dx1 /d and dy1 /d are also negative and,
Proof. The derivatives of fi () and gi () with respect to the
consequently, component 1 cannot have a stationary point in
vapor fraction are given by (see Appendix B).
either the liquid or the vapor phase. 
NC


dfi () 2
i x3
= − N (1 − Ki )2 2m , Theorem 3. The least volatile component of a multicomponent
zm
m=1 (xm /zm )
m m=1
d C 2
mixture cannot have a stationary point in either the liquid or
NC  2 3 the vapor phase.
dgi () 1 2
 ym
= − N −1 .

i Km zm
2
m=1 (ym /zm )(1/
m ) m=1
d C 2 Proof. Similar to Theorem 2. One can obtain expressions for
KNC  and /KNC ; from Lemma 1 both dxNC /d and dy1 /d
Notice that both derivatives are always negative in the flash are positive, so that component NC cannot have stationary point
interval 0

1; therefore, the auxiliary functions are in either the liquid or the vapor phase. 
monotone-decreasing. 
Theorems 2 and 3 show that one can specify the mole fraction
Lemmas 1 and 2 provide the theorem for the uniqueness of of the most or the least volatile component without the possi-
the stationary point. bility of input multiplicity (vapor fraction multiplicity). This in
turn implies that no multiplicity problems can arise when one
Theorem 1. When it exists, the stationary point for the mole controls the purity of the most or the least volatile components
fractions is unique in the flash interval 0

1. of a mixture.
3854 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857

Finally, the necessary and sufficient conditions for the exis- of pressure and temperature, this means that any component
tence of a stationary point can be established. that reaches its boiling point in the flash interval must have
input multiplicity.
Theorem 4. The liquid mole fraction of an intermediate com- The sufficient conditions for a stationary point can alternately
ponent can reach a stationary point only if Ki
1. be established through the detection of a change of sign for the
derivatives dxi /d and dyi /d given in Appendix A.
Proof. For the liquid phase, the unique zero ∗ of dxi /d or
fi () is found by solving fi (∗ ) = 1 − Ki∗ ∗ = 0. Then, Ki∗ = 4. A numerical example
1/∗ . Since ∗ 1, the liquid mole fraction of an intermediate
component can reach a maximum only if Ki
1.  We analyze the example reported by Tiscareño et al. (1998)
with the simplified model. The polynomial expressions given
Theorem 5. The vapor mole fraction of an intermediate com- by Holland (1981) for the K values as a function of tempera-
ponent can reach a stationary point only if Ki 1. ture were used. Table 1 provides a description of the problem
and the relative volatilities for each component at the operating
Proof. For the vapor phase, the unique zero ∗ of dyi /d or pressure of 2068 kPa. Table 2 shows the vapor–liquid equilib-
gi () is found by solving gi (∗ ) = (∗ /Ki∗ ) − 1 = 0, which rium coefficients as well as the liquid and vapor mole fraction
leads to Ki∗ = ∗ . Since ∗ 1, then the vapor mole fraction derivatives at the bubble point ( = 0) and at the dew point
of an intermediate component can reach a maximum only if ( = 1) (determined as suggested in this work, see Appendix
Ki 1.  A). By applying the sufficiency condition for the existence of a
stationary point in mole fractions, one can notice that, in both
One may notice that these conditions are necessary but not phases, components iC 4 and nC 4 will show steady-state input
sufficient for a stationary point to occur (the extreme compo- multiplicities, since the equilibrium coefficients reach a value
nents satisfy one of the two, but they cannot have a stationary of 1.0 within the interval 0

1. In addition, by applying the
point). From these facts, the following sufficiency theorems can necessary conditions, we note that components C2 and C3 will
be established. not have a stationary point in the liquid phase and that iC 5 and
nC 5 will not exhibit a stationary point in the vapor phase. The
Theorem 6. If the value of Ki for an intermediate component same observation can be made with respect to the values of the
equals one at some (1) in the interval 0 <  < 1, such a com- derivatives of component mole fractions. The derivatives test
ponent will have its unique stationary point in the liquid phase also indicates that C3 will show steady-state input multiplicity
in the interval 0 <  < (1) . in the vapor phase, but C2 will not.
Fig. 3 shows the liquid and the vapor mole fractions of the
Proof. If Ki = 1 at  = (1) > 0, then from Eq. (16) components as a function of the vapor fraction, obtained from

dKi  the procedure outlined in Section 2. We observe that only iC 4
fi ((1) ) = −(1) < 0, and nC 4 have multiple solutions in the liquid phase, while C3 ,
d (1)
iC 4 and nC 4 have multiplicity in the vapor phase, as predicted.
since dKi /d is always positive. For fi (0) one obtains fi (0) = Fig. 4 gives the dependence of the mole fraction derivatives
1 − Ki (0) > 0. From Lemmas 2 and 1, only a zero of fi () with the vapor fraction. We can verify that only one zero exists
and dxi /d must exist in the interval 0 <  < (1) .  in the feasible interval for the components with multiplicity.
We also observe that the derivatives of liquid mole fractions
Theorem 7. If the value of Ki for an intermediate component for iC 4 and nC 4 present a minimum value close to  = 0.45
equals one at some (1) in the interval 0 <  < 1, such a com- and  = 0.65, respectively, and that in the vapor phase the
ponent will have its unique stationary point in the vapor phase derivative of mole fraction for C3 exhibits a minimum close
in the interval (1) <  < 1. to  = 0.4. This observation is relevant to select numerical
methods if one wishes to find the vapor fraction at the stationary
Proof. If Ki = 1 at  = (1) > 0, then from Eq. (21) gi ((1) ) =
i
(1 − (1) ) dK
d  (1)
. For gi (1) one may obtain
 >0

1  Table 1
gi (1) = −1 + < 0. Data for the examplea
Ki =1
Component zi
i
From Lemmas 2 and 1, the only zero of gi () and dyi /d must
C2 0.1 12.02
exist in the interval (1) <  < 1.  C3 0.1 5.186
iC 4 0.1 2.843
Theorems 6 and 7 show that it is sufficient that the K value of nC 4 0.1 2.265
an intermediate component equals one in the interval 0 <  < 1 iC 5 0.3 1.184
for the unique maximum in liquid and vapor mole fraction to nC 5 0.3 1.000
occur. For ideal solutions, where Ki values are only dependent a Relative volatilities evaluated at 2068 kPa and 394 K.
 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857 3855

Table 2
K values and derivatives of mole fractions at  = 0 and  = 1
   
dxi  dxi  dyi  dyi 
Component Ki ( = 0) Ki ( = 1) d =0 d =1 d =0 d =1

C2 4.164 7.937 −0.316 −0.014 −0.776 −0.087

C3 1.797 3.424 −0.080 −0.028 +0.091 −0.071
iC 4 0.985 1.877 +0.002 −0.039 +0.130 −0.047
nC 4 0.785 1.496 +0.021 −0.040 +0.119 −0.033
iC 5 0.410 0.782 +0.177 +0.007 +0.233 +0.084
iC 5 0.346 0.660 +0.196 +0.115 +0.203 +0.154

0.7 0.2
0.101 C2 C3 iC4
0.6 0.100
iC4
nC4 iC5 nC5 0.1
xi

nC4
0.5 0.099
0.0

dxi /dφ
0.4 0.00 0.04 0.08 0.12 0.16
xi

0.3 -0.1

0.2 -0.2
0.1
-0.3
0.0
(a) 0.0 0.2 0.4 0.6 0.8 1.0 (a) 0.0 0.2 0.4 0.6 0.8 1.0

0.5 0.4

0.4 0.2

0.0
0.3
dyi /dφ

-0.2
yi

0.2
-0.4 C2 C3
iC4 nC4
0.1 -0.6
iC4 nC4
-0.8
0.0
(b) 0.0 0.2 0.4 0.6 0.8 1.0 (b) 0.0 0.2 0.4 0.6 0.8 1.0
φ φ

Fig. 3. (a) Liquid mole fractions as a function of the vapor fraction; (b) Vapor Fig. 4. (a) Derivatives of liquid mole fractions as a function of the vapor
mole fractions as a function of the vapor fraction. fractional (b) Derivatives of vapor mole fractions as a function of the vapor
fraction.

point. Table 3 summarizes the numerical values obtained for

the stationary point for components that exhibit multiplicity in
liquid or vapor phases; a bisection method was used for such
calculations (Chapra and Canale, 1998). A good agreement can Table 3
be observed between the approximate values obtained using Stationary point for components with multiplicity
the constant relative volatility model and the rigorous method
Component Liquid phase Vapor phase
that use the polynomial expressions for K values. Note that our
results are in good agreement with those presented by Tiscareño  xi,max  yi,max
et al. (1998).
(a) Approximate solution with constant relative volatilities
C3 0.101 0.184
5. Conclusions iC 4 0.006 0.100 0.377 0.123
nC 4 0.105 0.101 0.527 0.111

We have analyzed the FD by using a constant relative volatil- (b) Rigorous solution using the polynomial expressions for K values
ity model, a useful simplification for ideal or nearly ideal mix- C3 0.063 0.182
iC 4 0.005 0.100 0.342 0.121
tures. We have extended the work by Tiscareño et al. (1998)
nC 4 0.095 0.101 0.509 0.109
on the multiplicity of the solutions of the flash equations when
3856 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857

a product mole fraction of a component is specified, but we Addition of Eq. (A.4) for all component gives
have analyzed only the interval of physical significance of the ⎛ ⎞
NC

NC NC NC
flash operation to establish our results. Through Theorem 1, we dxm
xm
2
dxj
xm
2
demonstrate the existence of a unique stationary point for mole = − ⎝1 −  Kj ⎠ Km = 0.
d zm d zm
fractions of the liquid and vapor product phases, as a function m=1 m=1 j =1 m=1

of the vapor fraction. In addition, the necessary and sufficient For convenience, let us define  as
conditions for the existence of multiple solutions, previously
NC NC
found by Tiscareño et al. (1998), have been reproduced with x 2 /zm
dxj
the simplified model. Theorems 4 and 5 provide the necessary  = N m=1 m =1− Kj . (A.5)
m=1 (xm /zm )Km
C d
conditions for the liquid and vapor phases; Theorem 6 and 7 j =1
show that the sufficient condition for a component to have in-
put multiplicity is that its Ki value equals one within the phys- From Eq. (A.3),  satisfies
ical interval of the flash. From these results, the existence of at  1. (A.6)
most two steady-state values of the vapor fraction is guaranteed
when a product mole fraction is specified in a standard isobaric Finally, substitution of Eq. (A.5) into (A.4) gives
flash distillation process; in other words, systems modeled with
constant relative volatilities (i.e. ideal systems) can only show dxi x2
= i (1 − Ki ). (A.7)
second-order input-multiplicities. d zi
In a similar fashion, beginning with Eq. (12) we arrive at the
Acknowledgments following equations:
 
M. Vaca is grateful to CONACyT and the Instituto Mexicano dyi y2 1 dKi
= i 2 (1 − ) − Ki (Ki − 1)
del Petroleo for the financial support given for this project. d zi Ki d
 
y 2

Appendix A. Derivatives of mole fractions with respect to = i −1 , (A.8)
zi K i
the vapor fraction
NC 2
y /zm
Derivation of Eq. (11) with respect to the vapor fraction  = N m=1 m 1. (A.9)
m=1 (ym /zm )(1/Km )
C 2
yields,

dxi zi (Ki − 1) zi  dKi Appendix B. Derivatives of auxiliary functions fi (

) and
=− − , gi (
) with respect to the vapor fraction
d [1 + (Ki − 1)] 2 [1 + (Ki − 1)] d
2

  The function fi () is defined in Eq. (16). Derivation with

dxi z dKi
= 1 − Ki −  respect to the vapor fraction gives
d [1 + (Ki − 1)]2 d
 
xi2 dKi d d
= 1 − Ki −  (A.1) fi () = − (Ki ). (B.1)
zi d d d

with equilibrium coefficients given by Eq. (7). The derivative From Eq. (17), the product Ki  is given by
of equilibrium coefficients with respect to the vapor fraction is NC 2 NC 2
m=1 xm /zm x /zm
NC NC
Ki  = Ki N =
i N m=1 m
dKi
i
dxj
dxj m=1 (xm /zm )Km
C 2
m=1 (xm /zm )
m
C 2
= − N
j = −Ki Kj . (A.2)
d [ C

j xj ]2
j =1
d
j =1
d
j =1 Derivation with respect to the vapor fraction yields
⎛
dKi NC 2
d(Ki )
i x /z
Since is always strictly positive it follows that = N ⎝−  m=1 m m
d N
m=1 (xm /zm )
m m=1 (xm /zm )
m
d C 2 C 2

NC
⎞

NC

NC
Kj
dxj
< 0. (A.3) xm dxm
xm dxm ⎠
d × 2
m + 2 .
j =1 zm d zm d
m=1 m=1

Substitution of Eq. (A.2) into (A.1) yields After simplification one may obtain
⎡ ⎛ ⎞⎤ ⎛ ⎞
NC

NC


dxi xi2 dxj d(Ki ) 2
i x m dx m⎠
= ⎣1 − Ki ⎝1 −  Kj ⎠⎦ . (A.4) = N ⎝ (1 − KNC
m ) ,
d zi d d C
(x 2 /z )
z m d
j =1 m=1 m m m m=1
 M. Vaca et al. / Chemical Engineering Science 61 (2006) 3850 – 3857 3857

⎛ ⎞  
NC

d  1 2
d(Ki ) 2
i ⎝ x m dx m ⎠. = −
= N (1 − Km ) 
zm d d Ki
i NC (ym2 /z )(1/
)
d C
m=1 (x 2 /z )

m m m m=1 m=1 m m
N


C  
ym dym
Using Eq. (A.7), × −1 .
Km zm d
m=1
NC


d(Ki ) 2
i x3
= N (1 − Km )2 2m . With the use of Eq. (A.8) one obtains
zm
m=1 (xm /zm )
m m=1
d C 2
  NC 

2 3
d  1 2  ym
Then, the derivative of the function fi () with respect to the =− NC − 1 .
Ki
i Km zm
2
m=1 (ym /zm )(1/
m ) m=1
d 2
vapor fraction is
NC

Substitution into Eq. (B.3) yields
d 2
i x3
fi () = − N (1 − Km )2 2m . (B.2) NC  2 3
zm

m=1 (xm /zm )
m m=1
d C 2 d 1 2  ym
gi () = − N −1

i Km zm
2
m=1 (ym /zm )(1/
m ) m=1
d C 2
For the vapor phase, the definition of the function gi () is given
by Eq. (19) as (B.4)
 References
gi () = − 1.
Ki
Chapra, S.C., Canale, R.P., 1998. Numerical Methods for Engineers: With
Derivation with respect to the vapor fraction gives Programming and Software Applications. McGraw-Hill, New York.
  Constantinides, A., Mostoufi, N., 1999. Numerical Methods for Chemical
d d  Engineers with MATLAB Applications. Prentice-Hall PTR, Englewood
gi () = . (B.3)
d d Ki Cliffs, NJ.
Güttinger, T.E., Morari, M., 1996. Multiple steady states in homogeneous
From Eq. (20), the quotient /Ki is given by separation sequences. Industrial & Engineering Chemistry Research 35,
4597–4611.
NC 2 NC 2
 m=1 ym /zm m=1 ym /zm
1 1 Güttinger, T.E., Morari, M., 1999. Predicting multiple steady states in
= NC =  N
. equilibrium reactive distillation. 1. Analysis of nonhybrid systems.
Ki Ki
i
m=1 (ym /zm )(1/Km ) m=1 (ym /zm )(1/
m )
2 C 2
Industrial & Engineering Chemistry Research 38, 1633–1648.
Holland, C.D., 1981. Fundamentals of Multicomponent Distillation. McGraw-
Derivation with respect to the vapor fraction yields Hill, New York.
  Jacobsen, E.W., 1994. Dynamics of systems with steady-state input
d  1 1
= N multiplicity. A.I.Ch.E. Annual Meeting.
d Ki
i
m=1 (ym /zm )(1/
m )
C 2 Koppel, L.B., 1982. Input multiplicities in nonlinear, multivariable control
systems. A.I.Ch.E. Journal 28, 935–940.
⎛
NC 2 NC Lucia, A., 1986. Uniqueness of solutions to single-staged isobaric flash process
y /zm
ym 1 dym
× ⎝− N m=1 m 2 involving homogeneous mixtures. A.I.Ch.E. Journal 32, 1761–1766.
C
(ym
2 /z )(1/
)
m m zm
m d Monroy-Loperena, R., 2001. A fast method to check steady-state input
m=1 m=1 multiplicities in vapor–liquid flash separations. Industrial & Engineering
⎞ Chemistry Research 40, 3664–3669.
NC

ym dym ⎠ Rodríguez, I.E., Zheng, A., Malone, M.F., 2001. The stability of a reactive
+ 2 . flash. Chemical Engineering Science 56, 4737–4745.
zm d Rodríguez, I.E., Zheng, A., Malone, M.F., 2004. Parametric dependence of
m=1
solution multiplicity in reactive flashes. Chemical Engineering Science 59,
Simplification yields 1589–1600.
  Seader, J.D., Henley, E.J., 1998. Separation Process Principles. Wiley,
d  1 2 New York.
= N
d Ki
i Tiscareño, F., Gómez, A., Jiménez, A., Chávez, R., 1998. Multiplicity of
m=1 (ym /zm )(1/
m )
C 2
the solutions of the flash equations. Chemical Engineering Science 53,
NC   671–677.

1  ym dym Zheng, A., Grassi, V., Meski, G., 1998. On control of distillation columns
× 1− , with input multiplicity. Industrial & Engineering Chemistry Research 37,

m KNC zm d
m=1 1836–1840.