On the dynamics of nonequilibrium simple batch distillation

processes
J.M.F. Silva
a,
*, A. Knoechelmann
b
, Antonio J.A. Meirelles
c
, M.R. Wolf-Maciel
a,1
,
C.E. Lopes
b
a
Laboratory of Separation Process Development, Faculty of Chemical Engineering, State University of Campinas (UNICAMP), Cidade Universita´ria
Zeferino Vaz, CP-6066 Campinas, Sa˜o Paulo CEP-13081-970, Brazil
b
Laboratory of Fermentation Processes, Department of Antibiotics, Federal University of Pernambuco (UFPE), Cep-50670-901 Recife, PE, Brazil
c
Physical Separation Laboratory (LASEFI), Faculty of Food Engineering, State University of Campinas (UNICAMP), PO Box 6121, Campinas, Sa˜o
Paulo 13083-970, Brazil
Received 20 March 2002; received in revised form 31 May 2002; accepted 31 May 2002
Abstract
The study of the dynamics of a simple batch distillation usually assumes the equilibrium between the liquid and vapour phases. In
contrast to that approach, this work considers the analysis of the dynamic behaviour of a nonequilibrium simple batch distillation.
A simple batch distillation device is similar to an evapouration unit because no reflux is employed. A mathematical model and their
singular points are analysed through the usual technique of linearization followed by the eigenvalues analysis. The stability nature of
those points is presented for ideal solutions. Phase plane plots are made to show the occurrence of some unexpected behaviours such
as ‘dynamic azeotropic points’ and ‘inversion of the distillation order’. The dependence of these phenomena on the vapour
withdrawal flowrate is established for simple batch distillation.
# 2002 Elsevier Science B.V. All rights reserved.
Keywords: Separations process; Dynamic azeotrope; Nonequilibrium distillation; Dynamic simulation; Stability; Mass transfer
1. Introduction
An adequate dynamic description of distillation
devices is important, both in terms of steady state
analysis and in terms of the way the equipment behaves
in a processing situation of dynamic behaviour. This
area of study has attracted the attention of the
researchers since Rosembrock [1] proved that nonideal,
binary, constant molar overflow distillation columns
with nontheoretical plates and neglected vapour hold-
up, have a unique global asymptote stable steady state.
But a more recent research indicates that, for batch
distillation in a column with a middle vessel, additional
steady states can be found, which do not correspond to
pure components [2]. Doherty and Perkins [3,4] started
the dynamic analysis applied to simple batch distillation
processes. Investigations on dynamic analysis applied to
more complex situations are still underway [5Á/7].
Distillations are called equilibrium operations because
they are studied by assuming thermodynamic equili-
brium between phases, i.e. the bulk properties of the
vapour and liquid phases (temperatures, pressures and
compositions) are such that the equations of thermo-
dynamic equilibrium are satisfied. These equations are:
T
V
0T
L
P
V
0P
L
m
V
i
0m
L
i
(i 01; 2; 3; . . . ; c)
where T is the bulk temperature, P the bulk pressure
and m is the bulk chemical potential. The superscripts L
* Corresponding author. Fax: '/55-19-2894717
E-mail address: jmfs@lopca.feq.unicamp.br (J.M.F. Silva).
1
Fax: '/55-19-2894717.
Chemical Engineering and Processing 42 (2003) 475Á/485
www.elsevier.com/locate/cep
0255-2701/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 2 5 5 - 2 7 0 1 ( 0 2 ) 0 0 0 6 4 - 8
and V refer to the liquid and vapour phases, respec-
tively, and the subscript i refers to the components of
the phases. The chemical potential is function of the
system temperature, pressure and phase compositions.
An essential phenomenum in simple batch distillation
devices is the net mass transfer of the component across
the interface. It is impossible for such devices to work
without net mass transfer from liquid to vapour allow-
ing the liquid to vapourise. In this case it would be better
to assume that equilibrium prevails only at the interface
of the system, but not between the phase bulk proper-
ties. The equilibrium between the bulk properties is
closely linked to the process slowness, requiring quasi-
reversibility state conditions to occur.
When a batch distillation is performed under irrever-
sible conditions, equilibrium becomes a local property as
stated by the microscopic reversibility principle and the
bulk properties of the phases are not related by
thermodynamic equilibrium equations [8]. The equili-
brium, and consequently, the thermodynamic equations
for multiphase equilibrium prevail only at the interfacial
region.
Krishnamurty and Taylor [9] picked up this point and
developed a nonequilibrium model for the steady state
simulation of distillation columns, which was later
applied to an industrial column operation [10]. Sec-
ond-generation models, which incorporated the pressure
as a variable, were also developed [11,12]. Koojiman and
Taylor [13] considered the use of this model for the
dynamic simulation of distillation columns and showed
that the back-computed Murphree tray efficiencies are
time dependent and, as a consequence, the equilibrium
model is not suitable for dynamic simulations. In a
recent paper, Luna and Martı ´nez [14] analysed the
stability of the drying process in a multicomponent
nonideal mixture and they found interesting results
concerning to the existence of dynamic azeotropes and
the multiplicity of steady states (bifurcation).
The purpose of the present paper is to develop a
qualitative dynamic analysis of a simple batch distilla-
tion by assuming interfacial equilibrium coupled with
mass transfer between the phases. The resulting model
equations, as usual, will be linearized around the
equilibrium points and the asymptote stability analysed.
The results are compared with those obtained by
Doherty and Perkins [3,4].
2. The governing equations
The isothermal multicomponent simple batch distilla-
tions are governed by the following set of material
balance equations [3,4] (see Appendix A):
dx
i
do
0x
i
(y
i
(x) i 01; 2; . . . ; c(1 (1)
where x
i
and y
i
are mole fractions of the i component in
the liquid and vapour phases, respectively, and o is the
modified time variable defined by the integral
o 0
g
t
0
q
N
dt (2)
where q is the molar flow rate at which the vapour is
withdrawn from the equipment and N is the molar
amount of the residual liquid within the equipment.
The two-film theory will be used to correlate the bulk
concentrations of the solute in both phases. The liquid
and vapour compositions at interface are given by x
i
1
and y
i
1
, respectively. The compositions of the liquid and
vapour in the bulk of the phases are given by x
i
and y
i
.
The theory assumes also that all the resistance to the
mass transfer concentrates in the film region. The
schematic representation of the two-film theory is shown
in Fig. 1. As usual, the compositions in the liquid and
vapour sides of the interface are assumed to be in
equilibrium, as required by the thermodynamic of
irreversible processes which imposes stricter requirement
than a continuum theory does. It requires, in addition,
that the infinitesimal subsystems are in equilibrium at
every instant [8].
The material balance equation for the batch distilla-
tion device requires that the mole flow of vapour leaving
the equipment must be equal to the mole flow crossing
the liquid and vapour films. Considering component i ,
for the vapour which is withdrawn from the still, the
material balance equation becomes:
qy
i
0
(P1y
i
1(Py
i
)
RT=Ak
gi
0
(Cx
i
(C1x
i
1)
RT=Ak
li
(3)
where P* is the total pressure at interface, C* the mole
concentration of the liquid at the interface, P the total
pressure in the bulk of the vapour phase, C the bulk
mole concentration of the liquid phase, A the interfacial
area, k
gi
and k
li
are the mass transfer coefficients of
component i in the vapour and liquid films, respectively,
R the gas constant and T is the temperature of the
system.
Coulson and Richardson [15] pointed out that when
the components of the vapour are totally soluble in the
liquid, the main resistance to the mass transfer is located
in the vapour film and Eq. (3) becomes:
Fig. 1. Schematic representation of the two film theory.
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 476
qy
i
0
P1y
i
1(Py
i
RT=Ak
gi
(4)
or
y
i
0
Ak
gi
P1y
i
1
RTq 'Ak
gi
P
(5)
and
y
i
0a
i
y
i
1 (6)
where
a
i
0
Ak
gi
P1
RTq 'Ak
gi
P
(7)
Inserting Eq. (7) into the material balance equation
set, it gives
dx
i
do
0x
i
(a
i
y
i
1(x); i 01; 2; . . . ; c(1 (8)
where y
i
1
is the mole fraction of the component i at the
vapour interface, which is in equilibrium with the liquid
interface. The above set of equations is similar to the
Fig. 2. Phase space for the following conditions: T0/280 K, q0/0 gmol/s; (k
g1
0/0.001 m/s; k
g2
0/0.03 m/s k
g3
0/0.05 m/s).
Fig. 3. Phase space for the following conditions: T0/280 K; q0/0.02 gmol/s, (k
g1
0/0.01 m/s; k
g2
0/0.01 m/s; k
g3
0/0.01 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 477
one which is used by Doherty and Perkins [3,4], except
by the presence of the a
i
parameters, which consider the
mass transfer phenomena.
Under the Doherty and Perkins [3,4] assumptions, as
P0/P* and q0/0, the value of this parameter is unity.
The reversibility of the process implies absence of net
mass transfer. As P "/P* and q "/0, the irreversibility is
closely linked to the existence of the net mass transfer
and the a
i
?
s are usually different from 1.
It is very interesting to point out that a
i
becomes also
unity when k
g
is independent of i , i.e. k
g1
0/k
g2
0/. . .
k
gc(1
. In this case, an irreversible system is similar to a
reversible one. As the energy equation is not considered
here, the above set of equations was developed for
isothermal processes. Nevertheless, as indicated by
Doherty and Perkins [4], the results also apply to
isobaric distillation, since the analysis is essentially the
same.
3. Simple batch the distillation of ideal solutions
The ideal solutions obey the following set of equili-
brium relations:
y
i
10
P
0
i
P1(x)
x
i
(9)
This set of equations assumes that the vapour phase is
governed by the ideal gas law. Inserting the above
equilibrium equations into Eq. (8), it gives:
dx
i
do
0x
i
(1(a
i
(x)b
i
(x)); i 01; 2; . . . ; c(1 (10)
where:
b
i
(x)0
P
0
i
P1(x)
; i 01; 2; . . . ; c(1 (11)
P
i
0
is the vapour pressure of component i . According
to Doherty and Perkins [3,4], b
i
has the following
properties:
a) b
i
/(x)/ /0 for all feasible x
i
i 0/1. . .c(/1;
b) db/dt /0 i 0/1, 2, . . ., c(/1;
c) b
i
(e
i
)0/1 i 0/1, 2, . . ., c(/1 e
1
0/(1, 0, . . ., 0), e
2
0/(0,
1, . . ., 0), etc.;
d) b
c
(0) 0/1.
The steady state are:
1) /x00;/
2) /x
k
0e
k
;/
3) /x0x
s
when
b
i
(x
s
)0
1
a
i
(x
s
)
Depending on the process conditions, the nodes can
be stable, unstable or saddle points. The stability
characteristics of the nodes will depend on the thermo-
dynamic and mass transfer interactions of the compo-
nents. Owing to this dependence, nodes different from
those of pure components are called dynamic azeo-
tropes, as proposed by Schluender [16]. The third steady
state condition, mathematically, admits the existence of
a dynamic azeotropic point as a singular point of the
model. The model of Doherty and Perkins [3,4] does not
predict this occurrence. The mathematical existence of
Fig. 4. Phase space for the following conditions: T0/280 K; q0/0.02gmol/s, (k
g1
0/0.01m/s; k
g2
0/0.03m/s; k
g3
0/0.05 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 478
this steady state shows an interesting behaviour that can
happen with an ideal mixture. The ideal mixture serves
as a limit case for the study of this phenomenon.
The Liapounov first theorem starts from the linear-
ized equation:
dx
do
0J(x)(x(x
s
) (12)
where:
J(x
s
)0
n
Á Á Á

@f
i
@x
j

T;x?
Á Á Á
n

¸
¸
¸
¸
(13)
and
f
i
0x
i
(1(a
i
(x)b
i
(x)) i 01; 2; . . . ; c(1 (14)
@f
i
@x
J
0(x
i
a
i

@b
i
@x
j

T;x?
(x
i
b
i

@a
i
@x
j

T;x?
'd
ij
(1(a
i
b
i
) (15)
where:
d
ij
0
@x
i
@x
j
0
1 if i 0j
0 if i "j

@a
i
@x
j

T;x?
0
da
i
dP1

@P1
@x
j

T;x?
and

@b
i
@x
j

T;x?
0
db
i
dP1

@P1
@x
j

T;x?
and so
J(X
s
)0
u
11
: : : u
1C(1
: : : : :
: : : : :
: : : : :
u
(C(1)1
: : : u
(C(1)(C(1)

¸
¸
¸
¸
¸
¸
(16)
where:
u
ii
0(x
i
a
i
db
i
dP1

@P1
@x
i

T;x?
(a
i
b
i
'1(x
i
b
i
da
i
dP1
Â

@P1
@x
i

T;x?
(17)
and
u
ij
0(x
i
a
i
db
i
dP1

@P1
@x
j

T;x?
(x
i
b
i
da
i
dP1

@P1
@x
j

T;x?
(18)
Now, the properties of the Jacobean matrix will be
analysed at the points x
s
00 and x0x
s
; such as:
b
i
(x
s
)0
1
a
i
(x
s
)
to determine the stability behaviour of these points. The
properties of the Jacobean matrix at the point x
k
0e
k
are given in Appendix B.
Case 1: for the trivial solution, when x
s
00; the
Jacobean matrix becomes:
J(0)0
1(a
1
(0)b
1
(0) Á Á Á 0
Á Á Á 1(a
2
(0)b
2
(0) Á Á Á 0
:
:
:
1(a
(c(1)
(0)b
(c(1)
(0)

¸
¸
¸
¸
(19)
and the determinant of this matrix is:
Fig. 5. Phase space for the following conditions: T0/ 280 K; q0/0.02 gmol/s, (k
g1
0/0.001 m/s; k
g2
0/0.05 m/s; k
g3
0/0.1 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 479
det J(0)0
¸
c(1
i01
(1(a
i
b
i
) (20)
The topological nature of this singular point is
determined by the eigeinvalues of the Jacobean matrix
J(0). As it is a symmetric matrix, all the eigenvalues
must be real [17]. The trivial singular point, as a
consequence, is a node or a saddle point and the
eigenvalues are given by:
l
i
01(a
i
b
i
(21)
The corresponding possibilities are:
a) A stable node occurs if and only if a
i
b
i
/1, since in
this case all the eigeinvalues assume negative values.
As b
i
(0) obey the relationship:
b
i
(0)0
P
0
i
P1(0)
i 01 . . . c(1 (22)
and as, at the origin, P*(0) 0/P
c
0
, this implies that:
P
0
i

P
0
c
a
i
i 01 . . . c(1 (23)
where P
c
0
is the pure c-component vapour pressure
and, as usual, component c is the heavier one
(which has the smallest value of P
i
0
). The origin
will be a stable node, if the values of P
c
0
determines
the above inequality. Nevertheless, if the values of
a
i
are such that break the inequality and the origin
becomes an unstable node, another component, i.e.
one of the other vertices, becomes the stable node.
Otherwise, the dynamic azeotropic point becomes
the stable node, as we will see later. In batch
distillation, there is always at least one stable
node. The existence of a dynamic azeotropic point
does not occur when a
1
0/a
2
0/. . .a
(c(1)
, which
implies that all the k
gi
are equal, or when q0/0,
i.e. the distillation process is reversible. In both
cases a
i
0/1 for every component. Such situations
seem not likely, since each component would be
usually transferred to the vapour phase with a
different speed.
b) An unstable node occurs at x
s
00 only if all
a
i
(0)b
i
(0) B/1.
In these conditions, all the eigenvalues are
positive. Following a similar reasoning to the above
item, the condition for an unstable node is:
P
0
i
B
P
0
c
a
i
i 01 . . . c(1 (24)
and there is the possibility that the origin becomes
an unstable node.
c) Finally, it is possible to get a saddle point when one
or more components obey condition Eqs. (23) and
(24) is valid for the other components, simulta-
neously.
Case 2: when x0x
s
; such as b
i
(x
s
)01=a
i
(x
s
); then
(@P1=@x
j
)
T;x?
00 and the Jacobean matrix becomes:
Fig. 6. Phase space for the following conditions: T0/280 K; q0/0.02 gmol/s, (k
g1
0/0.005 m/s; k
g2
0/0.001 m/s; k
g3
0/10 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 480
J(x
s
)0
1(a
1
b
1
. . . 0
. . . 1(a
2
b
2
. . . 0
:
:
:
1(a
(c(1)
b
(c(1)

¸
¸
¸
¸
¸
(25)
As this matrix is null, det J(x
s
)00; this is a none-
lementary singular point or a degenerate singular point.
As it is widely known, degenerate points are not
amenable to mathematical and physical analysis. This
is a necessary but not enough condition for the multi-
plicity of steady states (bifurcation), and a similar result
was also reported by Luna and Martı ´nez [14]. So, we
will perform numerical simulations of the phase space to
reach some conclusions on this question.
4. Numerical simulations
To perform the computational experiments, the dis-
tillation of a three-component mixture will be investi-
gated. Component 1 is assumed to be the most volatile
and component 3 the least one. For equilibrium
calculation purposes, it is assumed that component 1
has the properties of propane, and components 2 and 3
of n-butane and n-pentane, respectively. The mass
transfer coefficients were varied in a large range of
values to test all possible situations. The values were
selected in such a way that it is possible to pick up the
important dynamic behaviour, and the exact values are
presented in the captions of the figures. For each
selected set of mass transfer coefficients the bulk
pressure P was calculated using Eq. (26) below. It
should be noted that this equation corresponds to the
summation over all components, taken into account Eq.
(5) given above.
¸
3
i01
Ak
gi
P1y
i
1
RTq 'Ak
gi
P
01 (26)
The results of the stability studies for all steady states
considered in the present work are reported in Appendix
C. The experiments start by considering the equilibrium
conditions where q0/0 and a
i
0/1, Ö/
i
. This corresponds
to the situation analysed by Doherty and Perkins [4]. As
expected, the phase plane curves converged to the origin,
which corresponds to the least volatile component, as
shown in Fig. 2. In this figure, three different phase
plane trajectories were represented because in the
numerical experiments three different initial points
were selected for describing the system over time. It is
important to note that the mass transfer coefficient
values have no influence because the mass transfer is
very slow.
In a second set of computational experiments, the q
value was assumed to be different from zero and all
components were assumed to have the same mass
transfer coefficient. It is usual in chemical engineering
process design to assume the equality of the mass
transfer coefficients, despite of the fact that this
assumption is not likely to occur. Under these circum-
stances, again a
i
0/1, Ö/
i
, and we obtained a similar
behaviour to that observed in the first case. It is worth
to point out that the Doherty and Perkins’s model [4] is
still valid at nonequilibrium situation, as is this case and
Fig. 7. Phase space for the following conditions: T0/280 K; q0/0.02 gmol/s, (k
g1
0/0.001 m/s; k
g2
0/0.0015 m/s; k
g3
0/10 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 481
can be seen in the curves presented in Fig. 3. The third
set of computational experiments were carried out
assuming that q is different from zero and that all
mass transfer coefficients have different values, and
under these circumstances, a
i
"/1, Ö/
i
. Similarly to the
previous situation, the origin remained as a stable point.
The phase plane trajectory, however, became different,
as shown in Fig. 4.
An interesting phenomena occurred when component
3 mass transfer coefficient was assumed to be 2 times
higher than that corresponding to component 2 and 100
times higher than component 1 mass transfer coefficient.
The origin still remains as a stable node, but a
modification of the distillation path occurs, reflecting
the fact that component 2 distillates before component
1, as it is easy to see in Fig. 5. In spite of the large
volatility of component 1, its lower mass transfer
coefficient reduces the transfer velocity to the vapour
phase. This is the reason for the inversion of the
distillation order. In the literature, there are already
reports of a similar inversion of the distillation order.
For example, this behaviour happens in the distillation
of the mixture chloroform, acetone and benzene, in
which the azeotrope acetone-chloroform (boiling point
of 64.5 8C) distils before pure chloroform (61.2 8C) [3].
Obviously, in the majority of the experimental situa-
tions, the higher mass transfer coefficient correspond to
the most volatile component.
Exploring the inversion phenomena through numer-
ical analysis, a very interesting result was obtained when
the mass transfer coefficient of the least volatile
component was assumed to be thousands of times larger
than the mass transfer coefficient of the other two
components. In this case, the origin still remains a stable
node, but component 2 also becomes a stable node. In
Fig. 6, it is shown a situation where the phase plane
trajectories converge to the component with intermedi-
ate volatility (component 2). This fact indicates the
multiplicity of the steady-state of an ideal solution, as it
is suggested by the theory exposed before. In the present
case (ideal solutions) such behaviour was obtained by
using very different mass transfer coefficients for all
components. It might occur that for real solution a
similar behaviour can be obtained without assuming
such a difference between the mass transfer coefficient
values. For example, the consideration of solutions
nonideality would mean the introduction of component
i activity coefficient multiplying the irreversibility factor
a
i
in the Eq. (23). Activity coefficients lower than 1 in
such equation willl increase the possibility of inverting
the distillation order. It should be noted that the mixture
acetone-chloroform mentioned above exhibits signifi-
cant negative deviation from ideality.
Another phenomena predicted by the model and
obtained in the numerical experiments is the occurrence
of a dynamic binary azeotropic stable point for ideal
mixtures, as shown in Fig. 7. In this case, the theory
shows that the origin is still a stable node and the
vertices, components 1 and 2, are saddle point and
unstable node, respectively, what also suggests the
existence of multiple steady states. The captions of the
above mentioned figures show the values of the para-
meters which induce the above result. In the present
work there is no mathematical restriction to the
Fig. 8. Phase space for the following conditions: T0/280 K; q0/0.02 gmol/s, q0/0.21 gmol/s; (k
g1
0/0.003 m/s; k
g2
0/0.01 m/s; k
g3
0/0.1 m/s).
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 482
existence of a dynamic azeotropic point as occurs in the
Doherty and Perkins [4] equilibrium model. According
to Doherty and Perkins [4], if a binary ideal mixture
possesses an azeotropic point, it would not be an isolate
singular point because it belongs to the straight line y
i
0/
x
i ,
and this violates the hypothesis of an isolated point.
Such restriction does not exist in our model, where a
dynamic azeotropic point appears without any doubt.
It was verified that the Jacobean matrix of the
dynamic azeotrope is null, and this is a necessary
condition to occur multiplicity of steady states. For
this reason, some operational parameters were disturbed
in order to confirm the existence of bifurcation. As a
i
is
a function of several operational variables, it was tried
to perturb those variables to test the behaviour of the
model. The parameters were the interfacial area, dis-
tillation temperature and flow rate. When we change the
vapour flow rate ten times its initial value, a change in
the phase plane trajectory occurred, as shown in Fig. 8.
We started for a given initial point and in the case of the
lower flow rate (q0/0.02 gmol/s), the system converged
to the origin, and for the higher value (q0/0.21 gmol/s),
the system converged to the dynamic ternary azeotropic
point. These results show that the final steady state of
batch distillation is dependent on vapour flow rate. This
confirms the existence of steady state multiplicity
(bifurcation) in the simple batch distillation. A similar
result was also reported by Luna and Martı ´nez [14] in
their analysis of stability in multicomponent drying of
homogeneous nonideal liquid mixtures.
5. Conclusion
This work considered the distillation of ideal liquid
mixtures in a batch equipment. The mathematical
analysis was carried out for a general n-component
mixture. The numerical experiments were performed
considering a three-component mixture.
The mass transfer was included in the mathematical
model by equating the rate of the vapour that is
withdrawn to the overall mass transfer across the
vapourÁ/liquid interface. The resulting equation allows
one to relate y
i
*, the equilibrium concentration of
component i , at interface, to the bulk vapour concen-
tration of this component y
i
. The insertion of this
relationship in to the main balance equation leads to
the following conclusions:
1) batch distillation equipment are devices that re-
quires that the mass transfer phenomena at the
interface be considered;
2) the nonequilibrium model reproduces the Doherty
and Perkins [4] model when the following condi-
tions are fulfilled:
a) the rate of vapour withdrawn is zero, i.e. the
distillation is performed under reversible con-
ditions (very slow distillation) or
b) the mass transfer coefficients of the compo-
nents have the same values.
Condition (b) is the only one where the Doherty and
Perkins [4] model prevails under irreversible conditions.
In the general case, the nonequilibrium model predicts
that the following system behaviours for ideal mixture
are also possible:
1) The occurrence of a situation in which the residual
liquid curve converges to the component with
intermediate volatility (see Fig. 6).
2) The occurrence of stable dynamic azeotropic points
(see Figs. 7 and 8).
3) A change in the vapour withdrawn rate may result
in a change on the stability of the equilibrium point.
This work considers ideal vapour and liquid phases,
and shows the wealth of phenomena occurring in a
simple batch distillation when mass transfer is taken into
account.
Silva [18] considered also regular quadratic solutions.
This discussion will appear in a future paper.
Acknowledgements
This work was supported financially by research
grants from Fundac¸a˜o de Amparo a` Pesquisa do Estado
de Sa˜o Paulo (FAPESP-99/09944-2'/98-12302-0) and
Conselho Nacional de Desenvolvimento Cientı ´fico e
Tecnolo´ gico (CNPq-521011/95-97 '/301309/1996-5).
Appendix A: Balance mass equation
Considering the diagram representation of a batch
distillation shown in Fig. 1A below and performing a
material balance of the component i, it gives:
d(Nx
i
)
dt
0(qy
i
i 01; 2; . . . c(1 (1:A)
so
x
i
d(N)
dt
'
Nd(x
i
)
dt
0(qy
i
i 01; 2; . . . c(1 (2:A)
From an overall material balance it gives:
d(N)
dt
0(q (3:A)
so
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 483
d(x
i
)
dt
0
q(t)
N(t)
(x
i
(y
i
) i 01; 2; . . . c(1 (4:A)
which is not an autonomous system. To convert the
above equation in an autonomous system, a new time
variable is defined by the equation:
d(o)
dt
0
q(t)
N(t)
(5:A)
Inserting this equation into the above nonautono-
mous system, the following system can be obtained:
d(x
i
)
do
0(x
i
(y
i
) i 01; 2; . . . c(1 (6:A)
which is the autonomous form of the material balance
equation.
Appendix B: Analysis on the vertices
in the vertices

@P1
@x
i

T;x0e
k
00 i 01 . . . c(1 (8A)
Then
J(e
k
)00 (9A)
This is a nonelementary singular point or degenerate
singular point. This result will be investigated by
numerical simulations presented in the paper
Appendix C: Condition of stability
In agreement with the theory exposed in this paper,
the stability is given for:
P
0
i
a
i
P
0
c
(10:A)
The temperature of the simulated systems is T0/280
K, so that the pressure of saturation of each component
is:
P
0
i
05:83 bar P
0
2
01:33 bar P
0
3
00:33 bar
Then, the conditions of stability given in the figures
are;
Figure 2;
5.83)/1/0.33 0.33)/1B/1.33 0.33)/1B/5.83
1.33)/1/0.33 5.83)/1/1.33 1.33)/1B/5.83
Stable node Saddle point Unstable node
Figure 3;
5.83)/1/0.33 0.33)/1B/1.33 0.33)/1B/5.83
1.33)/1/0.33 5.83)/1/1.33 1.33)/1B/5.83
Stable node Saddle point Unstable node
Figure 4;
5.83)/0.74/0.33 0.33)/1B/1.33 0.33)/1B/5.83
1.33)/0.944/
0.33
5.83)/0.74/
1.33
1.33)/0.944B/
5.83
Stable node Saddle point Unstable node
Figure 5;
5.83)/0.18/
0.33
0.33)/1.003B/
1.33
0.33)/1.003B/
5.83
1.33)/0.95/
0.33
5.83)/0.18/1.33 1.33)/0.95B/5.83
Stable node Saddle point Unstable node
Figure 6;
5.83)/4.515/
0.33
0.33)/732/1.33 0.33)/732B/5.83
1.33)/0.907/
0.33
5.83)/4.515/
1.33
1.33)/0.907B/
5.83
J(e
k
)0
1(a
1
(e
k
)b
1
(e
k
) Á Á Á
:
:
:
(a
k
db
k
dP1
¸
@P1
@x
1

T;X
0
(b
k
da
k
dP1
¸
@P1
@x
1

Á Á Á (a
k
db
k
dP1
¸
@P1
@x
k

T;X?
(b
k
da
k
dP1
¸
@P1
@x
k

T;X?
Á Á Á
:
:
:
1(a
c(1
(e
k
)b
c(1
(e
k
)

¸
¸
¸
¸
¸
¸
¸
¸
(7:A)
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 484
Stable node Saddle point Unstable node
Figure 7;
5.83)/0.99/
0.33
0.33)/2.93B/
1.33
0.33)/2.93B/
5.83
1.33)/0.55/
0.33
5.83)/0.39/
1.33
1.33)/0.55B/
5.83
Stable node Saddle point Unstable node
Figure 8;
5.83)/0.41/
0.33
0.33)/1B/1.33 0.33)/1B/5.83
1.33)/0.72/
0.33
5.83)/0.41/
1.33
1.33)/0.72B/
5.83
Stable node Saddle point Unstable node
Appendix D: Notation
C number of components in a multicomponent
mixture
E
I
unit vector in i th orthogonal direction
J Jacobean matrix of first partial derivatives
P total pressure (bar)
P
i
0
saturated vapour pressure of pure component i
(bar)
Q flow rate (gmol/s)
R gas constant (m
3
bar/g mol K)
T temperature (K)
/x/ vector of mole fraction in the liquid phase
/x1/ vector of liquid mole fraction in equilibrium with
vapour at interface
/y/ vector of mole fraction in the vapour phase
/y1/ vector of vapour mole fraction in equilibrium
with liquid at interface
Greek symbols
a
i
irreversibility factor
d
ij
kronecker delta
l eingeinvalue
u
ij
component of the Jacobean matrix
Subscripts
i , j , c components
T temperature
x? vector of x
1
. . .x
j(1
, x
j'1
,. . .x
c(1
hold constant
Superscripts
O pure component quantity
s singular point value
V, L vapour and liquid phases, respectively
References
[1] H.H. Rosembrock, Approach to a stable steady state far from
equilibrium, Proc. Phy. Soc. London 80 (1962) 962.
[2] A.G. Davidyan, V.N. Kiva, G.A. Meski, M. Morari, Batch
distillation in column with a middle vessel, Chem. Eng. Sci. 49
(1994) 3033Á/3051.
[3] M.F. Doherty, J.D. Perkins, On the dynamics of distillation
processes */I. The simple distillation of multicomponent non-
reacting homogeneous liquid mixtures, Chem. Eng. Sci. 33
(1978a) 281Á/301.
[4] M.F. Doherty, J.D. Perkins, On the dynamics of distillation
processes */II. The simple distillation of model solutions, Chem.
Eng. Sci. 33 (1978b) 569Á/578.
[5] P. Rouchon, Y. Creff, Geometry of the flash dynamics, Chem.
Eng. Sci. 48 (18) (1993) 3141.
[6] M. Barolo, G.B. Guarise, S.A. Rienzi, A. Trotta, Understanding
the dynamics of batch distillation column with a middle vessel,
Comput. Chem. Eng. 22 (Supl. S) (1998) S37Á/S44.
[7] L.U. Kreul, A. Gorak, P.I. Barton, Dynamic rate-based for
multicomponent batch distillation, AIChE J. 45 (9) (1999) 1953Á/
1962.
[8] G.D.C. Kuiken, Thermodynamics of Irreversible Processes:
Application to Diffusion and Reology, Wiley, Chichester, 1994.
[9] R. Krishnamurty, R. Taylor, A nonequilibrium stage model of
multicomponent separation process part */I: model description
and method solution, AIChE J. 31 (3) (1985) 449Á/455.
[10] R. Taylor, M.F. Powersa, A. Vickery, A. Arehole, A none-
quilibrium stage model of multicomponent separation process */
V: computational methods for solving the model equations,
Comput. Chem. Eng. 12 (12) (1988) 1229Á/1241.
[11] R. Taylor, H.A. Kooijman, J.S. Hung, A second generation
nonequilibrium model for computer simulation of multicompo-
nent separation process, Comput. Chem. Eng. 18 (3) (1994) 205Á/
217.
[12] M.H. Pescarini, A.A.C. Barros, M.R. Wolf-Maciel, Development
of a software for simulating separation processes using a none-
quilibrium stage model, Comput. Chem. Eng. 20 (1996) S279Á/
284.
[13] H.A. Kooijman, R. Taylor, A nonequilibrium model for dynamic
simulation of tray distillation collumns, AIChE J. 41 (8) (1995)
1852Á/1863.
[14] F. Luna, J. Martı ´nez, Stability analysis in multicomponent drying
of homogeneous liquid mixtures, Chem. Eng. Sci. 54 (1999) 5823Á/
5837.
[15] A. Coulson, Q. Richardson, Unit Operations, vol. 2, Pergamon
Press, Oxford, UK, 1968.
[16] E.U. Schluender, Einfu¨ hrung in die Stoffu¨ bertragung, Verlag,
Stuttgart, 1984, pp. 56Á/77.
[17] M. Marcus, H. Mink, A Survey of Matriz Theory and Matriz
Inequalities, Allyn & Bacon, New York, 1964.
[18] J.M.F. Silva, Dynamic of an irreversible simple batch distillation
(in portuguese), M.Sc. thesis, Department of Chemical Engineer-
ing, Federal University of Pernambuco, Recife, Pernambuco,
Brazil, 1996.
J.M.F. Silva et al. / Chemical Engineering and Processing 42 (2003) 475Á/485 485

. The results are compared with those obtained by Doherty and Perkins [3. Considering component i. C the bulk mole concentration of the liquid phase. The resulting model equations. An essential phenomenum in simple batch distillation devices is the net mass transfer of the component across the interface. .476 J. The liquid and vapour compositions at interface are given by x1 i and y1. which incorporated the pressure as a variable. as a consequence. for the vapour which is withdrawn from the still. as required by the thermodynamic of irreversible processes which imposes stricter requirement than a continuum theory does. . and consequently. The theory assumes also that all the resistance to the mass transfer concentrates in the film region. Silva et al. The schematic representation of the two-film theory is shown in Fig. the compositions in the liquid and vapour sides of the interface are assumed to be in equilibrium. equilibrium becomes a local property as stated by the microscopic reversibility principle and the bulk properties of the phases are not related by thermodynamic equilibrium equations [8]. / Chemical Engineering and Processing 42 (2003) 475 Á/485 and V refer to the liquid and vapour phases. respectively. The equilibrium. that the infinitesimal subsystems are in equilibrium at every instant [8]. In this case it would be better to assume that equilibrium prevails only at the interface of the system. The governing equations The isothermal multicomponent simple batch distillations are governed by the following set of material balance equations [3. Schematic representation of the two film theory. were also developed [11. the material balance equation becomes: qyi 0 (P1yi1( Pyi ) RT =Akgi 0 (Cxi ( C1xi1 ) RT=Akli (3) where P * is the total pressure at interface. The compositions of the liquid and i vapour in the bulk of the phases are given by xi and yi . respectively. The purpose of the present paper is to develop a qualitative dynamic analysis of a simple batch distillation by assuming interfacial equilibrium coupled with mass transfer between the phases. The equilibrium between the bulk properties is closely linked to the process slowness. Second-generation models. Koojiman and Taylor [13] considered the use of this model for the dynamic simulation of distillation columns and showed that the back-computed Murphree tray efficiencies are time dependent and. and o is the modified time variable defined by the integral t o0 g N dt 0 q (2) where q is the molar flow rate at which the vapour is withdrawn from the equipment and N is the molar amount of the residual liquid within the equipment. the thermodynamic equations for multiphase equilibrium prevail only at the interfacial region. It is impossible for such devices to work without net mass transfer from liquid to vapour allowing the liquid to vapourise. 1. The material balance equation for the batch distillation device requires that the mole flow of vapour leaving the equipment must be equal to the mole flow crossing the liquid and vapour films. 2. the main resistance to the mass transfer is located in the vapour film and Eq. When a batch distillation is performed under irreversible conditions. but not between the phase bulk properties. Krishnamurty and Taylor [9] picked up this point and developed a nonequilibrium model for the steady state simulation of distillation columns. (3) becomes: 2. 1. A the interfacial area. the equilibrium model is not suitable for dynamic simulations. Coulson and Richardson [15] pointed out that when the components of the vapour are totally soluble in the liquid. respectively.M. The chemical potential is function of the system temperature.F. kgi and kli are the mass transfer coefficients of component i in the vapour and liquid films. respectively. requiring quasireversibility state conditions to occur. P the total pressure in the bulk of the vapour phase.4] (see Appendix A): dxi do 0xi (yi (x) i 01. as usual. . will be linearized around the equilibrium points and the asymptote stability analysed. c(1 (1) Fig.12]. In a recent paper. It requires.4]. where xi and yi are mole fractions of the i component in the liquid and vapour phases. As usual. which was later applied to an industrial column operation [10]. C * the mole concentration of the liquid at the interface. R the gas constant and T is the temperature of the system. and the subscript i refers to the components of the phases. pressure and phase compositions. in addition. Luna and Martınez [14] analysed the ´ stability of the drying process in a multicomponent nonideal mixture and they found interesting results concerning to the existence of dynamic azeotropes and the multiplicity of steady states (bifurcation). The two-film theory will be used to correlate the bulk concentrations of the solute in both phases. .

05 m/s). .M.F. 2. kg2 0/0. (kg1 0/0. 3. Silva et al.02 gmol/s. .01 m/s.001 m/s. i i 01. (kg1 0/0.J.01 m/s. / Chemical Engineering and Processing 42 (2003) 475 Á/485 477 Fig. q0/0.01 m/s). Phase space for the following conditions: T0/280 K. kg3 0/0. kg2 0/0.03 m/s kg3 0/0. q0/0 gmol/s. . which is in equilibrium with the liquid interface. Phase space for the following conditions: T0/280 K. qyi 0 or yi 0 and P1yi1( Pyi RT =Akgi (4) ai 0 Akgi P1 RTq ' Akgi P (7) Akgi P1yi1 RTq ' Akgi P Inserting Eq. . The above set of equations is similar to the Fig. 2. it gives (5) dxi do (6) 0xi (ai y1(x). . (7) into the material balance equation set. c(1 (8) yi 0 ai y1 i where 1 where yi is the mole fraction of the component i at the vapour interface.

etc. c(/1 e1 0/(1. the results also apply to isobaric distillation. The model of Doherty and Perkins [3. i 01. c(/1. . The third steady state condition. 0). The steady state are: 1) /x00.03m/s. Simple batch the distillation of ideal solutions The ideal solutions obey the following set of equilibrium relations: yi10 P0 i P1(x) xi (9) This set of equations assumes that the vapour phase is governed by the ideal gas law. 2. since the analysis is essentially the same. According i to Doherty and Perkins [3. .4] does not predict this occurrence. . the above set of equations was developed for isothermal processes.478 J. i 01. as indicated by Doherty and Perkins [4]. i.e. 4. Under the Doherty and Perkins [3./ 2) /xk 0ek . nodes different from those of pure components are called dynamic azeotropes. it gives: dxi do 0xi (1(ai (x)bi (x)). kgc(1.. ./ 3) /x0xs when bi (xs )0 1 ai (xs ) 3. . d) bc (0) 0/1. . except by the presence of the ai parameters. Nevertheless. where: bi (x)0 P0 i P1(x) . . . As P "/P * and q "/0. (8). The reversibility of the process implies absence of net mass transfer. admits the existence of a dynamic azeotropic point as a singular point of the model.. (kg1 0/0. kg3 0/0.. Owing to this dependence. as proposed by Schluender [16]..4] assumptions. as P 0/P * and q0/0. . .05 m/s). kg2 0/0. . It is very interesting to point out that ai becomes also unity when kg is independent of i. . . the nodes can be stable. As the energy equation is not considered here.02gmol/s. 0). .4]. c(1 (11) P0 is the vapour pressure of component i. the value of this parameter is unity. The stability characteristics of the nodes will depend on the thermodynamic and mass transfer interactions of the components. bi has the following properties: a) bi/(x)/ /0 for all feasible xi i 0/1. . q 0/0. . which consider the mass transfer phenomena. . kg1 0/kg2 0/. c(1 (10) Depending on the process conditions.F. 2. / Chemical Engineering and Processing 42 (2003) 475 Á/485 Fig. Phase space for the following conditions: T0/280 K.. . .c(/1. 2. . unstable or saddle points. e2 0/(0. . c) bi (ei )0/1 i0/1. . . one which is used by Doherty and Perkins [3. The mathematical existence of . the irreversibility is closely linked to the existence of the net mass transfer and the ai? s are usually different from 1. an irreversible system is similar to a reversible one. Inserting the above equilibrium equations into Eq.01m/s. 2. mathematically. . 0.M. Silva et al. 1.4]. b) db/dt /0 i0/1. In this case.

kg3 0/0.05 m/s.x? u(C(1)1  dbi @P1 @xi : : : : : : : : : :  : : : : : u1C(1 : : : u(C(1)(C(1) 3 7 7 7 7 5 dai dP1 (17) (16) 3 7 Á Á Á7 5 (13) and 6 J(xs )0 6Á Á Á 4 and  dP1   @P1 @xi (ai bi '1(xi bi T. / Chemical Engineering and Processing 42 (2003) 475 Á/485 479 Fig.001 m/s.x? @xj n uij 0(xi ai (14) dbi dP1  @P1 @xj  T. 5.x? dP1 @xj T.J. The ideal mixture serves as a limit case for the study of this phenomenon. the properties of the Jacobean matrix will be analysed at the points xs 0 0 and x 0xs . 2. kg2 0/0.x? (xi bi dai dP1   @P1 @xj T. the Jacobean matrix becomes: 2 6 J(0)06 4 1(a1 (0)b1 (0) ÁÁÁ 1(a2 (0)b2 (0) ÁÁÁ ÁÁÁ :: : 0 0 1(a(c(1) (0)b(c(1) (0) 3 7 7 5 (19) and the determinant of this matrix is: . . q 0/0. when xs 00. (kg1 0/0. Silva et al. Phase space for the following conditions: T 0/ 280 K. c(1     @fi @bi @ai 0(xi ai (xi bi @xJ @xj T.x? dij 0 0 and     @bi db @P1 0 i @xj T. The Liapounov first theorem starts from the linearized equation: dx do 0J(x)(x(xs ) (12) and so 2 6 6 J(X )0 6 6 4 s u11 : : : where: where: a uii 0(xi i 2  n  @fi T.x? @xi  1 0 Now.M.x? (18) fi 0xi (1(ai (x)bi (x)) i 01. such as: (15) bi (xs )0 1 ai (xs ) to determine the stability behaviour of these points.x? dP1 @xj T.x? 'dij (1(ai bi ) where: if i 0j if i "j @xj     @ai dai @P1 0 @xj T. . this steady state shows an interesting behaviour that can happen with an ideal mixture.x? @xj T.1 m/s). Case 1: for the trivial solution.F. The properties of the Jacobean matrix at the point xk 0ek are given in Appendix B. .02 gmol/s. .x? T.

001 m/s. . there is always at least one stable node. if the values of ai are such that break the inequality and the origin becomes an unstable node. The existence of a dynamic azeotropic point does not occur when a1 0/a2 0/. if the values of P0 determines c the above inequality. or when q 0/0. this implies that: c P0  i P0 c ai i 01 . the dynamic azeotropic point becomes the stable node. the condition for an unstable node is: and as.M. then (@P1=@xj )T. As bi (0) obey the relationship: bi (0)0 P0 i P1(0) i 01 .x? 0 0 and the Jacobean matrix becomes: . . 6.480 J. In batch distillation. (23) and (24) is valid for the other components. c(1 (23) P0 B i P0 c ai i 0 1 . c) and there is the possibility that the origin becomes an unstable node. . all the eigenvalues must be real [17]. as a consequence. Silva et al. which implies that all the kgi are equal. . Finally. as we will see later. c(1 (22) (21) Otherwise. it is possible to get a saddle point when one or more components obey condition Eqs. c(1 (24) where P0 is the pure c -component vapour pressure c and. b) An unstable node occurs at xs 00 only if all ai (0)bi (0) B/1. Following a similar reasoning to the above item. The origin i will be a stable node. the distillation process is reversible. at the origin. such as bi (xs )0 1=ai (xs ). i. . as usual. is a node or a saddle point and the eigenvalues are given by: li 01(ai bi The corresponding possibilities are: a) A stable node occurs if and only if ai bi /1.005 m/s. / Chemical Engineering and Processing 42 (2003) 475 Á/485 Fig. kg3 0/10 m/s). . kg2 0/0. . In both cases ai 0/1 for every component. In these conditions. one of the other vertices. P *(0) 0/P0 .e.e. Phase space for the following conditions: T 0/280 K. Case 2: when x0xs . simultaneously. since each component would be usually transferred to the vapour phase with a different speed. The trivial singular point. As it is a symmetric matrix. i.a(c(1). becomes the stable node. Nevertheless. Such situations seem not likely. all the eigenvalues are positive. q0/0. . component c is the heavier one (which has the smallest value of P0). (kg1 0/0.F. det J(0)0 c(1 Y i01 (1(ai bi ) (20) The topological nature of this singular point is determined by the eigeinvalues of the Jacobean matrix J (0). another component.02 gmol/s. since in this case all the eigeinvalues assume negative values.

0015 m/s. the phase plane curves converged to the origin. For equilibrium calculation purposes. Component 1 is assumed to be the most volatile and component 3 the least one. Akgi P1y1 i 01 RTq ' Akgi P (26) 4. For each selected set of mass transfer coefficients the bulk The results of the stability studies for all steady states considered in the present work are reported in Appendix C. It is important to note that the mass transfer coefficient values have no influence because the mass transfer is very slow. (26) below. It is worth to point out that the Doherty and Perkins’s model [4] is still valid at nonequilibrium situation. So. three different phase plane trajectories were represented because in the numerical experiments three different initial points were selected for describing the system over time.. kg3 0/10 m/s). It is usual in chemical engineering process design to assume the equality of the mass transfer coefficients. As it is widely known. Phase space for the following conditions: T 0/280 K. Silva et al. It should be noted that this equation corresponds to the summation over all components. The experiments start by considering the equilibrium conditions where q0/0 and ai 0/1. / Chemical Engineering and Processing 42 (2003) 475 Á/485 481 Fig. as is this case and . Ö/i .M. 6 J(xs )0 6 4 1(a2 b2 :: : . . In this figure. and a similar result was also reported by Luna and Martınez [14]. we ´ will perform numerical simulations of the phase space to reach some conclusions on this question. degenerate points are not amenable to mathematical and physical analysis. 7. again ai 0/1. the q value was assumed to be different from zero and all components were assumed to have the same mass transfer coefficient. it is assumed that component 1 has the properties of propane... Ö/i . despite of the fact that this assumption is not likely to occur. This corresponds to the situation analysed by Doherty and Perkins [4].. respectively. Under these circumstances. 2. this is a nonelementary singular point or a degenerate singular point. The values were selected in such a way that it is possible to pick up the important dynamic behaviour. and the exact values are presented in the captions of the figures. This is a necessary but not enough condition for the multiplicity of steady states (bifurcation). taken into account Eq. det J(xs )00. and we obtained a similar behaviour to that observed in the first case. q0/0. and components 2 and 3 of n -butane and n -pentane.F.001 m/s. which corresponds to the least volatile component.02 gmol/s. (kg1 0/0.J. kg2 0/0.. The mass transfer coefficients were varied in a large range of values to test all possible situations. Numerical simulations To perform the computational experiments.. as shown in Fig. 0 0 1(a(c(1) b(c(1) 3 7 7 7 5 (25) pressure P was calculated using Eq. the distillation of a three-component mixture will be investigated. As expected. In a second set of computational experiments. 2 1(a1 b1 6 . (5) given above. 3 X i01 As this matrix is null.

The origin still remains as a stable node. but component 2 also becomes a stable node. / Chemical Engineering and Processing 42 (2003) 475 Á/485 Fig. there are already reports of a similar inversion of the distillation order. The captions of the above mentioned figures show the values of the parameters which induce the above result.2 8C) [3].01 m/s.F. 5. An interesting phenomena occurred when component 3 mass transfer coefficient was assumed to be 2 times higher than that corresponding to component 2 and 100 times higher than component 1 mass transfer coefficient. This fact indicates the multiplicity of the steady-state of an ideal solution.02 gmol/s. It should be noted that the mixture acetone-chloroform mentioned above exhibits significant negative deviation from ideality. Obviously. it is shown a situation where the phase plane trajectories converge to the component with intermediate volatility (component 2). For example.003 m/s. 3. the theory shows that the origin is still a stable node and the vertices. It might occur that for real solution a similar behaviour can be obtained without assuming such a difference between the mass transfer coefficient values. In this case. as shown in Fig. Another phenomena predicted by the model and obtained in the numerical experiments is the occurrence of a dynamic binary azeotropic stable point for ideal mixtures. Phase space for the following conditions: T0/280 K. In the present case (ideal solutions) such behaviour was obtained by using very different mass transfer coefficients for all components. 4. In spite of the large volatility of component 1. The phase plane trajectory.482 J. can be seen in the curves presented in Fig. Exploring the inversion phenomena through numerical analysis. as it is suggested by the theory exposed before. In the literature. but a modification of the distillation path occurs. Ö/i . the origin still remains a stable node. in which the azeotrope acetone-chloroform (boiling point of 64. acetone and benzene. became different. (23). kg3 0/0.M. in the majority of the experimental situations.21 gmol/s. (kg1 0/0. as shown in Fig. 8. are saddle point and unstable node. its lower mass transfer coefficient reduces the transfer velocity to the vapour phase. In Fig. the consideration of solutions nonideality would mean the introduction of component i activity coefficient multiplying the irreversibility factor ai in the Eq. q0/0. what also suggests the existence of multiple steady states.5 8C) distils before pure chloroform (61. components 1 and 2. Activity coefficients lower than 1 in such equation willl increase the possibility of inverting the distillation order. as it is easy to see in Fig. Similarly to the previous situation. q 0/0. reflecting the fact that component 2 distillates before component 1. the higher mass transfer coefficient correspond to the most volatile component. Silva et al. The third set of computational experiments were carried out assuming that q is different from zero and that all mass transfer coefficients have different values. the origin remained as a stable point. however. a very interesting result was obtained when the mass transfer coefficient of the least volatile component was assumed to be thousands of times larger than the mass transfer coefficient of the other two components. respectively. In the present work there is no mathematical restriction to the . For example. this behaviour happens in the distillation of the mixture chloroform.1 m/s). kg2 0/0. and under these circumstances. 7. 6. ai "/1. This is the reason for the inversion of the distillation order. In this case.

The mathematical analysis was carried out for a general n-component mixture. Conclusion This work considered the distillation of ideal liquid mixtures in a batch equipment. 7 and 8).J. ´ 5. Silva et al. and this violates the hypothesis of an isolated point. and this is a necessary condition to occur multiplicity of steady states. it would not be an isolate singular point because it belongs to the straight line yi 0/ xi . As ai is a function of several operational variables. The parameters were the interfacial area. When we change the vapour flow rate ten times its initial value. This discussion will appear in a future paper. . at interface. 1A below and performing a material balance of the component i. if a binary ideal mixture possesses an azeotropic point. A similar result was also reported by Luna and Martınez [14] in ´ their analysis of stability in multicomponent drying of homogeneous nonideal liquid mixtures. . 2) The occurrence of stable dynamic azeotropic points (see Figs. 2. Silva [18] considered also regular quadratic solutions. These results show that the final steady state of batch distillation is dependent on vapour flow rate. / Chemical Engineering and Processing 42 (2003) 475 Á/485 483 existence of a dynamic azeotropic point as occurs in the Doherty and Perkins [4] equilibrium model. distillation temperature and flow rate. where a dynamic azeotropic point appears without any doubt. . Acknowledgements This work was supported financially by research grants from Fundacao de Amparo a Pesquisa do Estado ` ¸˜ de Sao Paulo (FAPESP-99/09944-2'/98-12302-0) and ˜ Conselho Nacional de Desenvolvimento Cientıfico e ´ Tecnologico (CNPq-521011/95-97 '/301309/1996-5). In the general case. It was verified that the Jacobean matrix of the dynamic azeotrope is null. . Such restriction does not exist in our model. and shows the wealth of phenomena occurring in a simple batch distillation when mass transfer is taken into account. 8. it was tried to perturb those variables to test the behaviour of the model.e. the nonequilibrium model predicts that the following system behaviours for ideal mixture are also possible: 1) The occurrence of a situation in which the residual liquid curve converges to the component with intermediate volatility (see Fig. For this reason. a change in the phase plane trajectory occurred. the equilibrium concentration of component i . We started for a given initial point and in the case of the lower flow rate (q0/0. The insertion of this relationship in to the main balance equation leads to the following conclusions: 1) batch distillation equipment are devices that requires that the mass transfer phenomena at the interface be considered. and for the higher value (q 0/0. i. the system converged to the origin. . the system converged to the dynamic ternary azeotropic point. a) b) the rate of vapour withdrawn is zero. This confirms the existence of steady state multiplicity (bifurcation) in the simple batch distillation. 3) A change in the vapour withdrawn rate may result in a change on the stability of the equilibrium point. 2. The mass transfer was included in the mathematical model by equating the rate of the vapour that is withdrawn to the overall mass transfer across the vapourÁ/liquid interface. According to Doherty and Perkins [4]. 2) the nonequilibrium model reproduces the Doherty and Perkins [4] model when the following conditions are fulfilled: Appendix A: Balance mass equation Considering the diagram representation of a batch distillation shown in Fig. Condition (b) is the only one where the Doherty and Perkins [4] model prevails under irreversible conditions. . the distillation is performed under reversible conditions (very slow distillation) or the mass transfer coefficients of the components have the same values. 6). The numerical experiments were performed considering a three-component mixture. to the bulk vapour concentration of this component yi . This work considers ideal vapour and liquid phases. it gives: d(Nxi ) dt so xi d(N) dt ' Nd(xi ) dt 0(qyi i 01. c(1 (2:A) 0(qyi i 01.21 gmol/s). c(1 (1:A) From an overall material balance it gives: d(N) dt so 0(q (3:A) . some operational parameters were disturbed in order to confirm the existence of bifurcation. as shown in Fig.02 gmol/s). The resulting equation allows one to relate yi *.F.M.

33 )/0.33 Stable node 0.003 B/ 1.33 )/1 B/5. .515 / 0.83 )/1 /0.83 1.x0ek Then J(ek )00 (9A) This is a nonelementary singular point or degenerate singular point.83 Saddle point Unstable node 0.33 )/1 B/5.33 )/732 B/5.33 5. .33 )/1 /0. the stability is given for: P0 ai P0 i c (10:A) 5.33 )/1 B/5.33 )/1 B/1. Appendix B: Analysis on the vertices ÁÁÁ :: 6 :         6 6 dbk @P1 dak @P1 db @P1 da @P1 6(ak (bk (bk k Á Á Á (ak k J(ek )0 6 dP1 @x1 T.95 / 0.33 0.33 )/0.83 Unstable node i 01 . . 2.33 1.33 1. 5. 5.33 5. .33 1.944 / 0.33 0. .83 )/0. / Chemical Engineering and Processing 42 (2003) 475 Á/485 d(xi ) dt 0 q(t) N(t) (xi (yi ) i 01.33 )/1.907 / 0.83 5. c(1 (4:A) K.33 )/0.33 Stable node Figure 3. To convert the above equation in an autonomous system. the conditions of stability given in the figures are.33 Saddle point 0.83 Unstable node Inserting this equation into the above nonautonomous system.18 /1.33 Saddle point 0.33 )/1 B/1. c(1 (8A) Appendix C: Condition of stability The temperature of the simulated systems is T 0/280 .33 )/0.X 0 dP1 @x1 dP1 @xk T.M.83 )/1 /0.944 B/ 5.33 5.83 )/1 /1.33 1.X ? dP1 @xk T.83 )/0.33 )/1 B/5. c(1 (6:A) which is the autonomous form of the material balance equation. Figure 2.33 )/0.33 Stable node Figure 6.33 )/1B/5.83 )/0.515 / 1.83 1.83 )/1 /1.F.33 5. Silva et al.33 1.33 Stable node Figure 5. 2. in the vertices   @P1 00 @xi T. This result will be investigated by numerical simulations presented in the paper 5.33 )/1. so that the pressure of saturation of each component is: P0 05:83 bar i P0 01:33 bar 2 P0 0 0:33 bar 3 which is not an autonomous system.83 )/0.33 )/732 /1.74 / 1.83 Unstable node 0. .83 )/4.33 1. the following system can be obtained: d(xi ) do 0(xi (yi ) i 0 1. .X ? 6 :: 4 : 1(ac(1 (ek )bc(1 (ek ) 1(a1 (ek )b1 (ek ) 2 3 7 7 7 Á Á Á7 7 7 5 (7:A) Figure 4.95 B/5.83 0.33 Saddle point 0.33 1.33 )/1 B/1.74 /0.484 J.83 5.83 )/4.33 )/0.83 1. a new time variable is defined by the equation: d(o) dt 0 q(t) N(t) (5:A) Then.33 )/1 /0. 5. .003 B/ 0.18 / 0. In agreement with the theory exposed in this paper.907 B/ 5.

. Vickery. [3] M. Taylor.B. Batch distillation in column with a middle vessel.A.83 )/0. respectively 0.A. Davidyan. Proc. Eng. A. J.H. A. Phy. / Chemical Engineering and Processing 42 (2003) 475 Á/485 485 Stable node Figure 7. Eng.83 )/0. London 80 (1962) 962. Chem. Luna. [17] M.U. [12] M. 33 (1978b) 569 Á/578. [4] M. . [2] A. Eng. Marcus. M. Chem.33 5. Guarise. Chem.33 )/2. Pernambuco. Hung.83 Unstable node References [1] H.F.33 )/0.33 )/0.xc(1 hold constant Superscripts O pure component quantity . [9] R. Oxford. Taylor. Meski. On the dynamics of distillation processes */I. Dynamic rate-based for multicomponent batch distillation. Doherty.33 5. Kiva. Dynamic of an irreversible simple batch distillation (in portuguese).33 Stable node Figure 8. 45 (9) (1999) 1953 Á/ 1962. Mink.83 Unstable node Appendix D: Notation C EI J P P0 i Q R T /x / /x1 / / / / y y1/ number of components in a multicomponent mixture unit vector in ith orthogonal direction Jacobean matrix of first partial derivatives total pressure (bar) saturated vapour pressure of pure component i (bar) flow rate (gmol/s) gas constant (m3 bar/g mol K) temperature (K) vector of mole fraction in the liquid phase vector of liquid mole fraction in equilibrium with vapour at interface vector of mole fraction in the vapour phase vector of vapour mole fraction in equilibrium with liquid at interface symbols irreversibility factor kronecker delta eingeinvalue component of the Jacobean matrix Greek ai dij l uij Subscripts i. M. Eng. A nonequilibrium stage model of multicomponent separation process part */I: model description and method solution. Silva.M. R. [11] R. 5.N. P. [14] F.D. [7] L. 1996. Chem. Pergamon Press. Martınez.33 Saddle point 0. Powersa.41 / 1.33 1.83 )/0.99 / 0. AIChE J. Soc. Wolf-Maciel.33 )/1B/5.U.C.S. A. Perkins. [13] H. Sci. [8] G. Q. Chichester. H. AIChE J.55 / 0. pp. G. A. Comput. Gorak. Taylor.F. Understanding the dynamics of batch distillation column with a middle vessel. Eng. Comput. Barros.D. 22 (Supl. Kooijman. Geometry of the flash dynamics. 5.39 / 1. Doherty.83 1.33 Stable node Saddle point Unstable node s singular point value V. Chem. L vapour and liquid phases. [10] R.Sc. Barton. Recife. Rosembrock. ¨ ¨ Stuttgart. 31 (3) (1985) 449 Á/455.A. 1964. xj'1. 41 (8) (1995) 1852 Á/1863. Thermodynamics of Irreversible Processes: Application to Diffusion and Reology. [6] M. . Federal University of Pernambuco. 56 Á/77. Kuiken. 49 (1994) 3033 Á/3051. Eng.A. Sci. Richardson. [15] A. Sci. 1984.72 / 0.93 B/ 1. Barolo. 12 (12) (1988) 1229 Á/1241. j. Pescarini. S) (1998) S37 Á/S44. Comput. Wiley.F. Kreul. 1994. AIChE J. Development of a software for simulating separation processes using a nonequilibrium stage model. New York. Kooijman. Verlag. [5] P. Taylor.C. Arehole. .33 )/0. Sci. Unit Operations.A.J.41 / 0. Eng. [18] J. c components T temperature x? vector of x1.F.M. Morari.33 )/1 B/1.G. Chem. Schluender. V.83 )/0. G. Eng. Comput.D.I. A. 48 (18) (1993) 3141. Chem. 33 (1978a) 281 Á/301.33 1. A Survey of Matriz Theory and Matriz Inequalities. S. 0. J. 54 (1999) 5823 Á/ 5837. Approach to a stable steady state far from equilibrium.33 Saddle point 0. [16] E. J. Krishnamurty. thesis. Creff. 20 (1996) S279 Á/ 284. H. The simple distillation of model solutions. Eng. Coulson. Y. The simple distillation of multicomponent nonreacting homogeneous liquid mixtures. Chem. 2. Einfuhrung in die Stoffubertragung. M.55 B/ 5.F. Rienzi. UK. A nonequilibrium stage model of multicomponent separation process */ V: computational methods for solving the model equations..93 B/ 5. A second generation nonequilibrium model for computer simulation of multicomponent separation process.72 B/ 5. Sci. 1968. vol.H. J.R. A nonequilibrium model for dynamic simulation of tray distillation collumns. Chem. On the dynamics of distillation processes */II.xj(1.33 )/2. Stability analysis in multicomponent drying ´ of homogeneous liquid mixtures. M. Allyn & Bacon. Brazil. Rouchon. Department of Chemical Engineering. Silva et al. Trotta. 18 (3) (1994) 205 Á/ 217. R. Perkins.83 1.33 )/0.