Chemical Engineering and Processing 113 (2017) 20–34

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

General design methodology for reactive liquid–liquid extraction:

Application to dicarboxylic acid recovery in fermentation broth
Benoit Mizzia,* , Michel Meyera , Laurent Prata , Frédéric Augierb ,
Damien Leinekugel-Le-Cocqb
a
Laboratoire de Génie Chimique, Université de Toulouse, CNRS, INPT, UPS, Toulouse, France
b
IFP Energies nouvelles
Lyon, Rond-point de l’échangeur de Solaize
BP 3, 69360 Solaize, France

A R T I C L E I N F O A B S T R A C T

Article history:
Received 22 March 2016 A general design methodology for reactive liquid–liquid extraction is introduced in this paper. It is
Received in revised form 4 July 2016 composed of three different steps: feasibility analysis, pre-design determination and simulation
Accepted 7 October 2016 validation. This paper is focused on the first and the second step. This methodology leads to the design
Available online 13 October 2016 specifications of the units from the information concerning the physico-chemical behaviour of the
studied system, exploiting the equilibrium and material balance equations. The results of this
Keywords: methodology are a good starting point for an optimization study or for an investment calculation process.
Reactive liquid–liquid extraction This methodology has been applied to different case studies: two different strategies of extraction and
Liquid-liquid diagram
several solvents to recover succinic acid in fermentation broth.
Process design
ã 2016 Elsevier B.V. All rights reserved.
Process synthesis
Physical and chemical equilibrium
Succinic acid
Fermentation broth

1. Introduction The work of Samant and Ng [11] proposes a calculation

procedure for determining the design of a reactive liquid–liquid
The reactive separation involves the combination of reaction extraction column. They apply this procedure to a general pattern
and separation in a same unit. Esterification is mainly chosen as four reactive compounds at the maximum with an inert as an
case study in the literature. The transformation of the molecule in extracting agent.
its esterified form confers better physico-chemical and equilibrium The objective of this study is to adapt this procedure for real
characteristics which can be used for an easier separation. systems, and transcribe it in a design approachsimilar to the one
Furthermore, merging reaction and separation in a single introduced by Tuchlenski [14] for reactive distillation. A design
apparatus can lead to energy integration and process intensifica- methodology consists in determining the key structural param-
tion. eters of an operation unit. This study proposes a design
The efficiency of reactive separation strategies (especially methodology for reactive liquid–liquid reaction.
reactive distillation) has already been proved [13]. The advantages The design approach developed here is based on thermody-
of reactive distillation are multiple: increased conversion, in- namic equilibrium calculation to determine the design of the of
creased selectivity and energy integration or production quality reactive extraction process. The aim is to obtain the operating
can be cited. But distillation is energy consuming to maintain parameters such as possible raffinate and extract composition,
boiling and will be an expensive solution in the case of a very solvent rate and number of theoretical stages. The design approach
diluted system (especially in water). For this type of diluted includes three parts: the feasibility analysis, the pre-design and the
system, a cost effective way to recover valuable products especially validation step. This article is focused on the first and second step
from fermentation, such as acids or alcohols, is reactive liquid– of the approach
liquid extraction. In the first part, the general design methodology for reactive
liquid extraction is presented with the reactive liquid–liquid
diagram. It shows the location of the compositions of the organic
phase and the aqueous phase, which respect both chemical and
* Corresponding author. phase equilibrium. The second step of the methodology is also
E-mail addresses: benoit.mizzi@ensiacet.fr, benoit.mizzi@hotmail.fr (B. Mizzi).

http://dx.doi.org/10.1016/j.cep.2016.10.003
0255-2701/ã 2016 Elsevier B.V. All rights reserved.
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 21

are explained and a graphical illustration of the method is

Nomenclature
proposed using the reactive liquid–liquid diagram.
In the second part of this article, the design methodology is
E Molar organic phase flow (mol.s
1)
applied to succinic acid recovery. The production of succinic acid
F Molar feed flow (mol.s
1)
by fermentation of biomass presents a viable alternative [19,20].
H Liquid retention of the one stage extraction process
The recovery of succinic acid in the fermentation medium is quite
(mol)
complicated because of the presence of a large amount of water.
K eq Chemical equilibrium constant
Several separation techniques have been studied such as
M Mixture of F and S flow or RN and E1 flow (mol.s
1)
crystallization [3–5], electrodialysis [2,16] or reactive extraction
m Variable for RECM defined in the variable change
with tertiary amines [1]. However, all these methods have a high
N0T Initial total amount of mole for reactive liquid
operation cost and this is a serious brake to the development of
diagram (mol)
alternative ways of production. Several reactive liquid–liquid
Nc Number of compound
extraction strategies will be considered, including the number of
Nr number of reactions
reactive or not reactive compounds.
Q fractional conversion
R Molar aqueous phase flow (mol.s
1)
2. General design methodology for reactive liquid–liquid
S Molar solvent feed flow
extraction
(S/F) Solvent rate
(S/F)min Minimum solvent rate
2.1. Strategy
t Time (s)
V Variance
Fig. 1 presents the principle of the design methodology that we
w Molar composition of the liquid retention H
proposed, based on the analogy with reactive distillation. It is
X Reactive molar composition of phase
composed of three steps [13]. This method allows us to determine
x Molar composition of phase
the design parameters of the extraction column with the
z Molar composition of feed flow F
introduction of more complexity along the method and from
Z Reactive molar composition of feed flow F
known data:
g Activity coefficient
b Fraction of phase I in the mixture
- Thermodynamic property of the system (thermodynamic
j Chemical extent: mol
models parameters, chemical reaction constant);
t Dimensionless time
- The chemical data (stoichiometric coefficient for each reaction);
# Stœchiometric coefficient
- The product purity, rate recovery or conversion rate, objectives
f Matrix of the reference compounds’ stœchiometric
to achieve.
coefficient
Step 1: Feasibility analysis
Exponents
The feasibility analysis allows to verify that the separation
I Aqueous phase
specifications are thermodynamically feasible. If applicable, the
II Organic phase
maximum attainable compositions of the organic and aqueous
phase respecting the separation specifications are determined. If
Subscript
these are not feasible separation objectives are revised. In the
i Component
feasibility step, two tools have been identified: the reactive liquid–
j Stage
liquid diagram (rLLD) and the reactive extractive curve map
L Limit
(RECM). These tools allow to determine the thermodynamically
N Last stage
feasible composition useful for the next step of the methodology.
r Reaction
The last part of the article on the application of the method shows
ref Reference
that in some cases, the application of the reactive liquid–liquid
diagram is sufficient and that in more complex cases, RECM are
Acronym
needed to determine thermodynamically feasible compositions.
DMS Di-Methyl succinate
That’s why the two tools are identified.
DOS Di-octyl succinate
The cited tools are respectively based on the work of [10] and
MeOH Methanol
[15]. The feasibility analysis involved some simplified assumptions
MMS Mono-Methyl succinate
on infinite solvent rate, infinite intern organic and aqueous phase
MOS Mono-Octyl succinate
flow. They allow to identify the feasible composition for the
NTS Number of theoritical stage
organic and the aqueous phase from a thermodynamic point of
OcOH Octanol
view and independently of material balances consideration.
P-X Para-Xylene
Step 2: Pre-design
RECM Reactive extraction curves map
In the pre-design step, the studied physical-chemical phenom-
rLLD Reactive liquid-Liquid diagram
ena are more complex, and this step allows the determination of a
rLLE Reactive liquid-Liquid extraction
column configuration required to achieve the separation goals by
SMILES Simplified molecular input line entry specification
using data from the previous step. For the pre-design step the
SA Succinic acid
structural parameters of the column are determined. To begin
with, the minimum and maximum solvent rates are calculated.
Then, the aqueous and organic phase compositions of the process
are determined. To complete this step, the design of the column is
detailed; it is the pre-design method. It allows to evaluate the made (number of theoretical stages, composition and flow profile).
minimum solvent flow rate and the number of theoretical stages The boundary value design method introduced by Wuithier and
based on this reactive liquid–liquid diagram. The model equations Giraud book, [17,18] for simple extraction process and expanded to
22 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

Fig. 1. Organigram of the conception methodology for reactive separation process.

reactive extraction by [11] is used exploiting the data of feasibility

analysis.
Step 3: validation and design
g Ii xIi
g IIi xIIi ¼ 0 i ¼ 1; . . . ; Nc ð2Þ
The validation steps allow to validate the previously deter-
mined configuration of the column. From the perspective of
industrialization and investment analysis, sensitivity analysis and X
Nr
bxIi þ ð1
bÞxIIi
zi þ #i;r Q r ¼ 0 i ¼ 1; . . . ; Nc ð3Þ
optimization of structural and operating parameters of the column r¼1
are performed to make the commercially viable and feasible
extraction column. Complex physical phenomena are introduced
like kinetics and mass transfer limitation. This validation step can X
xIi ¼ 1 ð4Þ
also include experiments to validate the accuracy of the
calculations performed throughout the pre-design step. The This system of 2Nc + Nr + 1 equations and 3Nc + 2Nr + 1 variables
experimental results may also be compared to the simulation admits Nc + Nr freedom degrees which are the composition xF and
results in order to prove the precision of the simulation model. the equilibrium constant K eq : The unknown variables are the molar
In the first part of this article, the necessary tools for the first composition of each phase (xI , xII ), the fraction of phase I in the
two steps, feasibility analysis and pre-design, are described.
mixture (b) and the fractional conversion (Q r ¼ jr0 ) for each
NT

2.2. Reactive liquid–liquid phase diagram reaction. The resolution of this non-linear algebraic system is done
with a Matlab algorithm with the Simulis1 toolbox. The algorithm
To consolidate and exploit thermodynamic information, liquid– uses the “fsolve” function and uses the trust-region dogleg
liquid diagrams are implemented. They represent the locus of algorithm. The algorithm is a variant of the Powell dogleg method
aqueous and organic phases which respected the chemical and described in the work of [9].
phase equilibrium. In this section, the calculation method and the Once the resolution is done, the reactive compositions are
representation of reactive phase equilibrium are presented. calculated by choosing successively the reference compounds and
Building a reactive liquid–liquid diagram (rLLD) requires to the independent variables (Appendix A). These variables are then
perform coupled chemical and phase equilibria calculations for a represented in the reactive composition domain.
given system with set pressure and temperature. Descriptive According to the variance of the system, the shape of the field of
equations of such a system with Nc component and Nr indepen- reactive compositions can have different forms. At fixed tempera-
dent reactions, in thermodynamic equilibrium at fixed pressure ture and pressure the expression of the variance is:
and temperature are described below. There are the Nr chemical
equilibrium equation, the Nc physical equilibrium, Nc partial V ¼ Nc
Nr
1 ð5Þ
material balance and sum to unity for the molar composition of
The variance represents the space dimension needed for the
phase I.
representation of the system. If the variance is equal to 2, the
Y
Nc  #i;r system is representable in two dimensions. For a variance equal to
g i xIi ¼ K eq;r r ¼ 1; . . . ; Nr ð1Þ 3 the representation of the LLD is in three dimensions. The example
i¼1
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 23

B C

Pc

Fig. 3. Dynamic extraction process.

time in a single-stage reactive extraction process. This system

being still fed over time of pure solvent, the solvent ratio can be
A S regarded as infinite, and the amount of solute extracted maximum.
The extraction curves therefore represent a sequence of equilibri-
Fig. 2. Reactive liquid–liquid diagram for a system with 4 compounds and 1 um state that can be compared to a composition profile in a multi-
reaction V = 2. stage extraction column, operating at infinite solvent rate. So the
reactive extraction curves reflect the feasibility of the process. The
presented in Fig. 2 is for a system with four compounds and one RECM can be drawn for phases I and II.
reaction: We consider a system with Nc compounds and Nr independent
A þ S$B þ C reactions in a one stage extraction process presented in Fig. 3. The
operation is at fixed temperature and pressure and the following
In that case the variance is 2 and a planar representation of assumptions are formulated:
composition space is required.
The Curve A-Pc-D represents the phase envelope and point Pc – The total molar retention H is constant but HI and HII are variable.
represents the critical point. The blue line is the locus of aqueous – System at liquid–liquid and chemical equilibrium.
phase composition and the green line is the locus of organic
composition. Each composition is connected in pairs to a The input data of the system are H representing the constant
composition at the chemical and physical equilibrium by a tie liquid retention system, the feed flow F and the composition z of
line represented by the black dotted line. the Nc compounds in the feed flow.
This general example shows that the number of chemical The unknown variables are:
reactions reduces the space compositions accessible by the system.
When the variance (Eq. (5)) is strictly greater than 3, the – w, the global composition of the retention H
composition range is no longer limited to a curve but it is a – xI , the composition of the aqueous phase (I)
surface. The handling and use of such a diagram with three- – xII , the composition of the organic phase (II)
dimensional representation for the feasibility analysis is more – b, the fraction of phase I in the mixture
complex. The tool introduced in the next paragraph, RECM, is
– j the chemical extent of each reaction
introduced to overcome this difficulty in determining thermody-
namically feasible compositions.
The following equation system is expressed to describe the
dynamic extraction process (Fig. 3).
2.3. Reactive extraction curve map (RECM)
The Nc partial material balances on each component are
expressed here:
For this study a new conceptual tool has been developed by
analogy with the reactive distillation to determine thermodynam- dwi XNr
dj
ically feasible compositions, the reactive extractive curves map H
Fzi þ ExIIi þ H#i;r r ¼ 0 i ¼ 1; . . . ; Nc ð6Þ
dt i¼1
dt
(RECM). The RECM are the transposition to liquid–liquid reactive
extraction of Residue Curve Map (RCM) in reactive distillation. The
RCM is a very useful tool for the feasibility analysis of a reactive
wi ¼ xIi b þ xIIi ð1
bÞ i ¼ 1; . . . ; Nc ð7Þ
distillation process. It allows to identify the stable unstable nodes
and the distillation regions [13]. The RCM is obtained from the The developed algorithm uses the ode function which
simulation of a simple distillation process “Rayleigh distillation” integrates the differential equations system y0 = f(t,y). It is easier
[8,15]. A reactive residue curve is the locus of points compositions to solve an equation with only one differential term, so variable
of the liquid phase in equilibrium with the vapor phase in a single changes are performed to simplify the system of equation.
distillation operation and involving a system, the seat of one or The following changes of variables are being made:
more chemical reaction. A residue curve is a good approximation of !
liquid composition at infinite solvent ratio. XNr
mi ¼ H wi
#i;r jr i ¼ 1; . . . ; Nc ð8Þ
The objective of this section is to extend this concept to the i¼1
reactive liquid–liquid extraction. Fig. 3 illustrates the one-stage
extraction process. It is a decanter with a total constant retention, a This material balance is expressed in function of the variables
free interface and fed by the light phase (phase II). The liquid is at mi described previously:
chemical and phase equilibrium. The decanter is fed with pure dmi
solvent flow and a flow of organic phase is recovered. The reactive H
Fzi þ ExIIi ¼ 0 i ¼ 1; . . . ; Nc ð9Þ
dt
extractive curves represent the evolution of the composition over
24 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

A new time step t is defined which is non-dimensional:

Ft Fdt
t¼ dt ¼ ð10Þ
H H
The material balance can now be written:

dmi
F
Fzi þ ExIIi ¼ 0 i ¼ 1; . . . ; Nc ð11Þ
dt

dmi E

zi þ xIIi ¼ 0 i ¼ 1; . . . ; Nc ð12Þ
dt F
The Nc liquid–liquid equilibrium equations:

xIi g Ii
xIIi g IIi ¼ 0 i ¼ 1; . . . ; Nc ð13Þ

The Nr chemical equilibrium equations:

Y
Nc  #i;r
Keq;r
xIi g Ii ¼0 r ¼ 1; . . . ; Nr ð14Þ
i¼1

The sum equation on the variablexI :

X
Nc
1
xIi ¼ 0 ð15Þ
i¼1

The sum equation on the variablem:

!
X
Nc Nc X
X Nr
mi
H 1
#i;r jr ¼0 ð16Þ
i¼1 i¼1 r¼1

This system of 2Nc + Nr + 2 equations and of 2Nc + Nr + 2

variables (mi , xIIi ,jr ,E =F , bÞ is solved thanks to a Matlab algorithm
with the Simulis1 toolbox. The algorithm uses the “ode15s” solver
which is a multistep solver. It is a variable order solver based on the
numerical differentiation formulas (NDFs).
The results of the calculation are represented on the rLLD. From
the initial state with the global composition w of the retention, the
Matlab integration is performed in the forward and backward Fig. 4. Schema of a counter-current multistage reactive extraction.
direction from this point. This figure enables to choose a feasible
separation specification (xIN ) for the next step of the design
methodology, the synthesis and design step, particularly for the
flow rate can be calculated. The material balances are written with
steady state analysis.
the reactive composition (Appendix A) to avoid a reactive term and
remove the fractional conversion variable. Fig. 5 is the graphical
2.4. Feasibility analysis: minimum solvent rate determination, column
representation of Eq. (17), the crossing between the feed-solvent
static analysis
line (line F-S) and the organic-aqueous line (RN-El) defines point ML
Fig. 4 shows the reactive extraction column. This step of the
feasibility analysis allows to determine the flow and the
composition of all the input and output streams E1, F, S and Rn.
Static analysis provides the accessible composition at the
boundary of the process. The available data are the feed flow rate,
the feed composition (F, z) and the solvent composition (xIINþ1 ). The
separation objectives set by at the feasibility analysis with the
RECM allow to determine the composition of the aqueous phase RN
(xIN ).
First, the minimum solvent flow rate is calculated; the design
calculation requires the solvent ratio to be higher than the
minimum solvent ratio so the corresponding solvent rate is taken
between 1.3 and 1.6 times the minimum solvent rate [12]. The
determination of the minimum solvent rate is analogous to the
method used for simple liquid–liquid extraction. The tie line
through the feed point is the richest organic phase composition
that can be obtained with a counter-current extraction (EL ).
Knowing the composition of the limit extract, partial material
balances can be written on the column and the minimum solvent Fig. 5. Graphical representation of the minimum solvent rate determination.
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 25

which allows to calculate the minimum solvent rate.

F þ S ¼ ML X
c

FZ i þ SX i ¼ ML X ML ;i xIIi;j
1 ¼ 0 ð27Þ
FZ i þ SX i ¼ ðF þ SÞX ML ;i i¼a

FðZ
 i
X ML ;i Þ ¼ SðX ML ;i
X i Þ The set of Eqs. (23)–(27) allows to perform the design method
S ðZ i
X ML ;i Þ ð17Þ
¼ from the known flows and compositions of F and E1 stream if the
F
 min ðX
Xi Þ
L ;i 
M   method starts at the top of the column. The boundary value design
S S S method is based on this property. From the top or the bottom of the
¼ 1:6  $S ¼ F  1:6 
F F min F min column, the design of the column is defined with a succession of
equilibrium lines, which is the solution of the equilibrium
The last unknown variables of the system are calculated from
equation, and operating line, which is the solution of the material
the resolution of the following equation system which is composed
balance and chemical equilibrium equations. This is for graphical
of the overall material balance and partial material balances on the
representation and can be easily translated in equation for the
column written with the reactive compositions (equation 18 and
resolution algorithm on Fig. 6.
19) and the physical and chemical equilibrium equations (Eqs. (20)
and (21)).
3. Application to succinic acid recovery
S
RN
E1 þ F ¼ 0 ð18Þ
The previous parts present the general method for the design
determination of a reactive-extractive column. Now this method is
SX IINþ1
RN X IN
E1 X II1 þ F Z i ¼ 0 i applied to different liquid–liquid systems involving Succinic acid,
¼ 1; . . . ; Nc
Nr ð19Þ different heavier or lighter alcohols, water and the corresponding
mono-ester and di-ester. Selected n-alcohols are ethanol (C2), 1-
butanol (C4) and 1-octanol (C8). This choice allows to identify the
Y
Nc  effect of alcohol carbon length on the biphasic zone and on the
#i;r
Keq;r
xIi g Ii r ¼ 1; . . . ; Nr ð20Þ column design. Another system is studied with a non-reactive
i¼1 compound (para-xylene) used as a solvent. In that case the number
of compounds is Nc = 6 and in these conditions the graphical
representation is tridimensional.
xIi;1 g Ii
xIIi;1 g IIi ¼ 0 i ¼ 1; . . . ; Nc ð21Þ
3.1. Thermodynamic model

X
Nc In this article the reactions of succinic acid with different
xIIi;1
1 ¼ 0 ð22Þ alcohols (methanol, ethanol, butanol or octanol) and with or
i¼1

This system of 2(Nc + 1) variables and equations is solved with a

Matlab procedure using the “fsolve” function. The result can be
plotted on the reactive liquid–liquid diagram and the data are used
to determine the number of theoretical stages in the following
step.

2.5. Pre-design step: number of theoretical stage determination

The pre-design step and the determination of the number of

theoretical stagescan be done with the data determined in the
feasibility analysis. It allows to determine the number of
theoretical stages and the compositions profiles inside the column.
The phase equilibrium equation, the material balance and
chemical equilibrium equation can be expressed over j different
stages.
- for the jth stage j 6¼ 1 and j 6¼ N:

xIi;j g Ii
xIIi;j g IIi ¼ 0 i ¼ 1; . . . ; Nc ð23Þ

Ej X IIi;jþ1
Rj
1 X Ii;j
1
E1 X IIi;1 þ F Z i

¼0 i ¼ 1; . . . ; Nc
Nr ð24Þ

Y
Nc  #i;r
Kr
xIi g Ii r ¼ 1; :::; Nr ð25Þ
i¼1

X
C
xIi;j
1 ¼ 0 ð26Þ
i¼A
Fig. 6. Algorithm for the pre-design step, determination of the NTS.
26 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

without inert compounds are considered. The main compounds reactive alcohol, hence, the heavier the alcohol will be, the more
are listed in Table A1 of Appendix B. non-polar the mono-ester and di-ester will be. A high composition
The esterification reaction of succinic acid with an alcohol is in alcohol widens the two phase area.
composed of two consecutive reactions: initially, the formation of Fig. 7 shows this phenomenon, it presented the different bi-
a mono-ester and then the formation of a di-ester. The assumption phasic zones for a system with different carbon length alcohols. As
that there is no secondary reaction is formulated. discussed previously, for the heavier alcohol: octanol (C8), the bi-
phasic zone will be larger than for the lighter one: butanol (C4) and
SA þ ROH$H2 O þ MonoEster
ethanol (C2).
Then, different types of reactive liquid–liquid equilibria can be
identified as for liquid ternary diagrams. In accordance with a
MonoEster þ ROH$H2 O þ DiEster
ternary diagram, the immiscibility or partial immiscibility
The data on chemical reaction are the results of the work of between different compounds can be deduced from reactive
Orjuela et al. [6] on the esterification of succinic acid. These liquid diagrams. If a binary mixture considered separately does not
chemical data are considered the same regardless of the alcohol present total or partial immiscibility, the diagram is type 0 (see the
chosen, in accordance with the literature [7]; the same order of example of reactive mixture with ethanol). If the diagram presents
magnitude can be found for the chemical equilibrium constant an immiscible binary, it will be type I (see the example of reactive
(Table 1). mixture with butanol and octanol), and a diagram with two or
The UNIFAC thermodynamic model is chosen for the represen- more immiscible binary can be expected, in this case then the
tation of the liquid–liquid equilibrium. The major difficulty is the diagram will be type II or higher.
lack of some molecules in the data base, especially for mono-ester The design methodology for reactive extractive column is
and for long carbon chain on the alkoxy group. The unknown applied to these different systems. The shape of the two-phase area
molecules are created on Simulis1 with the regression tools and has an impact on the design of the column and according to the
from the Simplified Molecular Input Line Entry Specification system, particularly the molecular weight of the alcohol, the
(SMILES), the UNIFAC decomposition. separation may be more or less easy. For butanol and octanol, the
diagrams are type I, but for octanol the two phase’s area is larger.
3.2. Strategy A: solvent is the reactant Therefore reactive extraction process is possible with both alcohols
but it is most favourable with octanol. This part of the article deals
3.2.1. Choice of solvent with the design method for a reactive liquid–liquid extraction
The main component of the aqueous phase is water while in the column which is illustrated by the extraction of succinic acid with
organic phase it is non-polar or slightly polar molecules. The octanol.
higher the molecular mass of the reactive alcohol will be, the more
non-polar mono-ester and di-ester will be. As a consequence the 3.2.2. Design determination of the reactive liquid–liquid extraction
composition in the organic phase will also increase. The envelope column
of the two phases area depends on the initial composition in
3.2.2.1. Feasibility analysis. Fig. 8 represents the reactive liquid–
liquid diagram which shows the bi phasic zone and the mono
Table 1
Chemical reaction constant for the two consecutives reactions.
phasic zone. If the composition of the system is in the bi phasic
zone, the mixture splits in two phases with compositions
Keq(
) corresponding to the end point of the tie line. For the system
Succinic esterification 48.9 chosen, Fig. 8 presents the reactive composition domain composed
Mono-ester esterification 10.14 of five vertexes which represent the pure component. The blue and
green curves represent respectively the aqueous and organic phase

Fig. 7. Reactive liquid–liquid diagram for the examples chosen.

 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 27

Fig. 8. Reactive liquid–liquid diagram for system Nc = 5 Nr = 2 succinic acid, octanol, water, mono-ester, di-ester represented in reactive coordinates.

and the black dotted lines represent the tie-line. As for a simple rate. This example however is confined to the strategy and the
liquid–liquid diagram, the tie-lines tend to approach a critical composition set above.
point marked by the intersection of the aqueous and organic phase.
The conservative appearance of balances and summation 3.2.2.2. Pre-design. With the aqueous composition as input data
equations allows us to use the same tool and properties as the determined at the last step (feasibility analysis), the design
lever rule or design methodologies for the simple extraction methodology can be pursued. A Matlab algorithm was developed
process. Therefore, the liquid–liquid reactive diagram is an to resolve the equation described in the previous section synthesis
essential tool for the design method that must be developed; it and design. All the results obtained here are calculated with this
will be the core of the design methodology. algorithm.
The following task of the feasibility analysis is to determine the Fig. 9 is the representation of the steady states analysis. From
separation specifications with the rLLD and the RECM. The RECM the chosen separation specification, the boundary values of the
for the system can be presented on the reactive liquid–liquid process (useful for the next step) are determined. Here the material
diagram. The aqueous composition is located on the blue aqueous balance on the entire column is used to calculate the solvent rate,
curves of the reactive liquid–liquid diagram. Each point represents the aqueous and organic phase flows and composition.
a thermodynamically feasible point. The RECM starts at the critical The steady states analyses allow to determine the boundary
point which can be considered as the unstable node and tend to value thanks to separation specification. The method is as follows;
end at the binary water-alcohol mixture on liquid–liquid first point M representing the mixing of feed F and solvent S
equilibrium which is the stable node. Since this system has a (alcohol) is placed with the minimum solvent ratio. Secondly,
variance of 2, the phase envelope is a curve, and the RECM thanks to the steady states analyses the compositions of E1 are
correspond exactly to this equilibrium curve. Thus, we can resolved.
conclude that the RECM analysis does not give any more The next step is the determination of the column design with
information than the equilibrium diagram for system with a the resolution of Eqs. (23)–(27). Fig. 10 is the graphical
variance less or equal to 2. representation of the results with the boundary value of columns
The separation objective is set on the aqueous composition, the F, E1 RN and S depicted by the vertex alcohol. For this example the
higher the composition of the water compound in the aqueous column has four theoretical stages.
flow, the higher the C4 compounds are recovered (C4 represent all Table 3 summarizes all the design results of the reactive liquid–
the molecules with 4 carbons, succinic acid, and the two ester). liquid extraction column for the recovery of succinic acid with
This example is to determine the design of a column with the alcohol as reactant and extractant.
objective, the best yield of C4 molecules recovered in the extract It may be necessary when the selected alcohol cannot act as
flow. The compositions of the raffinate phase at the boundary of both reagent and solvent (examples of ethanol and butanol) to use
the process can be set directly on the reactive liquid–liquid an inert compound that only plays the role of solvent. However, the
equilibrium diagram. The following table gives this chosen use of an inert in a system with already five compounds makes it
composition (Table 2). more difficult to represent the reactive phase equilibrium as
Other strategies to fix the separation specifications are possible, explained in the next section.
such as choosing the highest possible composition of the extract
flow or by formulating the separation objectives with recovery 3.3. Strategy B: solvent is an inert

Table 2 The system chosen for this part has six compounds with two
Compositions of the RN flow fixed with the feasibility analysis. reactions. Methanol, the selected alcohol is too miscible with water
SA OcOH H2O MMS DMS and the separation could be complex as seen previously in part
Compositions 0.0005 0.0005 0.999 0 0
2.2.1. It is necessary to use a compound as inert solvent for the
reaction. The design of a reactive extraction column appears as the
28 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

Fig. 9. Minimum and work Solvent rate determination.

Fig. 10. Number of theoretical stage determination.

Fig. 11 presents the reactive phase cover which is representative

Table 3
Design parameters for the reactive extractive column processing the succinic acid
of the reactive liquid–liquid equilibrium.
recovery. To understand the phase diagram, the reagent composition
range can be decomposed into a succession of triangular plans
Composition
around the axis Para-Xylene-Water. The first and last plans of this
Flow (mol.s
1) SA OcOH H2O MOS DOS series are represented by the unreactive ternary Water-Methanol-
F 2.000 0.200 0 0.800 0 0 Para-xylene system which is Type I and P-Xylene-water-acid Type
S 0.301 0 1.000 0 0 0 II (Figs. 12a and d). The reactive succession plan describing the
E1 0.658 0.224 0.0168 0.382 0.355 0,0227
reactive phase envelope is thus transformed from a type I into a
Rn 1.643 0.0004391 0.000195 0.999 0 0
(S/F)min = 0.100 (S/F) = 0.150 type II.
NTS = 4 This representation with the succession of reactive planes,
allows identifying a critical curve, formed with all the critical point
of the reactive liquid–liquid equilibrium lines of the different
real challenge for this system. Our design methodology is applied planes. This remarkable line starts at the critical point of the
in the same way but the representation is more complex because of ternary Water-methanol-Para xylene system and finishes at the
the pentahedron form of the reactive composition domain. critical point of the ternary Succinic acid-mono ester-Para xylene
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 29

Fig. 11. Reactive liquid–liquid diagram of the system succinic acid/methanol/water/mono-ester/di-ester/para-xylene.

Fig. 12. Decomposition of the phase envelope in a reagent triangular plane succession.

system (Fig. 12b and c).The design methodology is applied to this evolution of the retention on this surface. The reactive RECM of the
system. system succinic acid-methanol-water-mono-ester-di-ester-para-
xylene is represented in Fig. 13. The RECM is very useful in this case
3.3.1. Feasibility analysis to fix a thermodynamically feasible separation specification, that is
The RECM is calculated for this system with the proposed to say from the phase envelope. The RECM represents the
method. For this system the variance (Eq. (5)) is equal to 3. For this thermodynamic path for a reactive liquid extraction.
kind of system, the RECM allows to observe the composition

Fig. 13. Feasibility analysis, Reactive Extractive Curve Map of the system succinic acid/methanol/water/mono-ester/di-ester/para-xylene.
30 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

For the aqueous phase, the green curves, all the RECM starts at representing the pure compound para-xylene. The critical points
an unstable node represented by the critical points represented on and the vertex para-xylene can be respectively related to unstable
the red line, and ends at the stable node symbolized by the aqueous node and stable node.
phase of the unreactive alcohol-water-para-xylene system sym- The principal node is the common point to all the extractive
bolized by the green point. This is in accordance with the dynamic curves can be spotted near the vertex water. This composition is
reactive extraction process. Initially the system presents a succinic thermodynamically feasible and because it is near the vertex
acid quantity which will decrease over time. Ultimatly, all the water, it is the composition that is the richest in water and the
succinic acid would be consumed and the aqueous mixture poorest in organic molecules. These characteristic could be chosen
composed only by alcohol, water and para-xylene. as a separation specification since the aqueous flow of the process
The blue curves on Fig. 13 represent the evolution of the is poor in valuable molecules. As a matter of fact, the recovery rate
composition of the extract flow. The points of these curves obtained here was the best.
represent the composition at physical and chemical equilibrium The Table 4 presents the fixed composition for the raffinate as
with the composition of aqueous phase. The blue curves start at the the composition with the fewest valuable compounds.
corresponding critical point on the red curve and end at the vertex
3.3.2. Pre-design
Table 5 gathered the results of the static analysis on the entire
Table 4 column. All the data needed for the static analysis are presented in
Compositions of the RN flow fixed with the feasibility analysis.
this Table 5. They are the feed flow chosen to be processed with a
SA MeOH H2O MOS DOS P-X reactive liquid extraction column (F, z), the composition of solvent
Compositions 0.00564 0.00604 0.988 0.000467 0 0 flow (xIINþ1 ) and the separation specification as the composition of
the aqueous phase (xIN ). This data are input data for the static
analysis procedure described in part 1.4 and they are exploited to
calculate the last variables as xII1 composition RN, S and E1 flow. The
Table 5 results of the static analysis are presented on the reactive liquid
Results of the static analysis on the extraction column for the Nc = 6 and Nr = 2 case diagram.
study. Figs. 14 present the results of the static analysis plotted in three
Stream Flow Composition dimensions (Fig. 14a) and according to the two static analysis
plane, the green and red planes (Figs. 14b and c). For the first plane
SA MeOH W MMS DMS P-X
the point representing the composition of feed and solvent flow are
F 2 0.156 0.220 0.624 0 0 0
coplanar, whereas the points representing the compositions of
S 10 0 0 0 0 0 1
RN 1.67 0.00564 0.00604 0.988 0.000468 0 0 extract and raffinate are on either side of the plane. For Fig. 14c, it is
E1 10.33 0.00210 0.00111 0.00199 0.0138 0.0133 0.968 the contrary, the point representing the raffinate and extract
NTS = 4 composition are coplanar and on the phase envelope while the S
and F points are on either side of the green plane.

Fig. 14. a: 3D representation of the static analysis. b: plane representation of the static analysis plane with S-F. c: plane representation of the static analysis plane with E1-RN.
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 31

This representation is the graphic interpretation of the static Different representations presented on Figs. 16 and 17 allow to
analysis equation (Eqs. (18)–(22)). It illustrate the alignment of the determine the number of theoretical stages.
FMS point and of the E1MRN according to the global material Fig. 16 represents the evolution of the reactive organic
balance and the positioning of the E1 and RN points on the phase composition for the reactive compound, succinic acid and
envelope due to the reactive liquid–liquid equilibrium. methanol along the column. The horizontal lines represent the
The setcomposition of RN xIN and the objective composition of E1 separation objectiveX IISA;1 . When the profile curves cut the objective
xII1 are now exploited as starting data for the determination of the line, the number of theoretical stages is determined. For this
number of theoretical stages. For this, Eqs (23)–(27) are solved example the number of theoretical stages is five.
with the following stopping criterion. The role of the reactive The distribution diagram proposed on Fig. 17 presents the
extractive column is to recover the C4 molecules (succinic acid, equilibrium curve and the operating curves required to determine
mono-ester and di-ester), the stopping criterion is the reactive the number of theoretical stages. The composition of succinic acid
composition of succinic acid XIISA;1 since the latter is the addition of in the extract (xIISA ) is plotted as a function of the succinic acid in the
the three real compositions of valuable molecules. The value of this raffinate (xISA ). It is another representation of the number of
reactive composition is X IISA;1
¼ 0:0292. theoretical stages in a manner analogous to McCabe Thiele for
On Fig. 15, the succession of equilibrium lines and operating distillation. The number of points on the equilibrium curves
lines are represented on the three-dimension representation. For represents the number of theoretical stages, and the exploitation of
this representation it is difficult to determine the number of this method returns the same result, the number of theoretical
theoretical stages. stages is equal to five.

Fig. 15. Pre-design step for the Nc = 6 system.

Fig. 16. Reactive organic phase composition profile for Succinic Acid and Methanol. NTS determination, NTS = 5.
32 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

Fig. 17. Distribution diagram for the NTS determination.

4. Conclusion - vTi : row vector composed of the sums of stoichiometric

coefficients for each Nr reactions.
A method for determining the design of a liquid–liquid  
extraction column is proposed in this article. This method is vTt ¼ vt;1 ; vt;2 ; . . . ; vt;Nr
based on three stages: a feasibility analysis, a shortcut step also
1
- fðrÞ : square matrix of stoichiometric coefficients for the Nr
called a pre- column design and a final step of simulation.
Throughout this method, the complexities in the study of physical reference component in the Nr reaction equation.
phenomena are introduced. The developed method is based on 0 1
vðNc
Nrþ1Þ;1    vðNc
Nrþ1Þ;Nr
fðrÞ ¼ B C
work done on reactive separation processes such as reactive .. ..
@ . vi;Nr . A
distillation. The aim of this method is the determination of the
reactive extractive column design and its operating parameters. vNc;1  vNc;Nr
This method has been applied to different reactive systems with xðrÞ : column vector of the compositions for the Nr reference
different strategies and numbers of compounds, with succinic acid, component.
and leads to the design of a reactive extractive column. To complete
 T
this study, the simulation step must be performed to identify the xref ¼ xðc
rþ1Þ ; xðc
rþ2Þ ;    xc
kinetics and transfer effects on the design configuration and on the
operating parameters. The matrix fðrÞ must be invertible, so the choice of the reference
components was subject to some conditions:
Appendix A. Defination and benefits of reactive composition
(Barbosa and Doherty, 1987)  No rows or columns of the matrix can be equal to zero. That is to
say, each chemical reaction needs to match at least one reference
We consider a system with Nc reacting compounds and Nr component. An inert compound cannot be selected as a reference
independent reactions. If the composition of the compounds’ compound.
compositions are expressed according to the Nr reference  Two reference compounds cannot have the same stoichiometric
compounds, Nr composition variables can be removed. A set of coefficients in each reaction.
composition variables with the same dimension as the variance of
the system (Nc-Nr) is obtained. The expression of the reactive The main benefit of using reactive composition is that the final
composition variables is as follows: result does not depend on the reaction and the chemical extent j
(Eq. (29)), in addition to the fact that the sum of the molecular

1 composition is always equal to 1.
xi
vTi :fðrÞ :xðrÞ
Xi ¼
1
ð28Þ X i ð0Þ ¼ X i ðjÞ 8j i ¼ 1; 2;    ; Nc
Nr ð29Þ
1
vTt :f ðrÞ :xðrÞ

with,
- vTi : row vector of stoichiometric coefficients of component i in X
Nc
Nr X
Nc
Nr
Xi ¼ 1 Yi ¼ 1 ð30Þ
each Nr reactions. i¼1 i¼1

 
vTi ¼ vi;1 ; vi;2 vi;Nr
Appendix B.

Table A1.
 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34 33

Table A1
Compounds studied in this application.

Name of the molecules Empirical chemical formula developed chemical formula Molecular weight (g.mol
1)
Succinic Succinic (SA) C 4 H 6 O4 118.09

Methanol (MeOH) CH3 OH 32.04

Ethanol (EtOH) C 2 H5 OH 46.07
Butanol (BuOH) C 4 H9 OH 74.12

Octanol (OcOH) C 8 H6 OH 130.23

Water H2 O H2 O 18.02
Mono-Methyl Succinate (MMS) C 5 H 8 O4 132.11

Dimethyl Succinate (DMS) C 6 H10 O4 146.14

Mono-Ethyl Succinate (MES) C 6 H10 O4 146.14

Diethyl Succinate (DES) C 8 H14 O4 174.19

Mono-Butyle Succinate (MMS) C 8 H14 O4 174.19

Dibutyle Succinate (DMS) C 12 H22 O4 230.30

Mono-Octyl Succinate (MOS) C 12 H22 O4 230.30

Dioctyl Succinate (DOS) C 20 H38 O4 342.51

Para-xylène C 8 H10 106.17

34 B. Mizzi et al. / Chemical Engineering and Processing 113 (2017) 20–34

References [10] K.D. Samant, K.M. Ng, Synthesis of extractive reaction processes, AIChE J. 44
(1998) 1363–1381.
[1] C.S.L. Garzón, A.J. Straathof, Recovery of carboxylic acids produced by [11] K.D. Samant, K.M. Ng, Design of multistage extractive reaction processes,
fermentation, Biotechnol. Adv. 32 (2014) 873–904. AIChE J. 44 (1998) 2689–2702.
[2] D.A. Glassner, P. Elankovan, D.R. Beacom, K.A. Berglund, Purification process [12] J.D. Seader, E.J. Henley, Liquid–Liquid extraction with ternary systems. In:
for succinic acid produced by fermentation, Appl. Biochem. Biotechnol. 51–52 separation process principles: chemical and biochemical operations 3rd
(1995) 73–82. edition john wiley and sons, Inc (2011) 299–358 (ISBN: 978-0-470-48183-7).
[3] Q. Li, D. Wang, Y. Wu, W. Li, Y. Zhang, J. Xing, Z. Su, One step recovery of succinic [13] R. Thery, X.M. Meyer, X. Joulia, M. Meyer, Preliminary design of reactive
acid from fermentation broths by crystallization, Sep. Purif. Technol. 72 (2010) distillation columns, Chem. Eng. Res. Des. 83 (2005) 379–400.
294–300. [14] A.B. Tuchlenski, Reactive distillation industrial applications: process design
[4] H. Lin, G.N. Bennett, K.-Y. San, Metabolic engineering of aerobic succinate and scale-Up, Chem. Eng. Sci. 56 (2001) 387–394.
production systems in Escherichia coli to improve process productivity and [15] S. Ung, M.F. Doherty, Calculation of residue curve maps for mixtures with
achieve the maximum theoretical succinate yield, Metababolic Eng. 7 (2005) multiple equilibrium chemical reactions, Ind. Eng. Chem. Res. 34 (1995) 3195–
116–127. 3202.
[5] R. Luque, C.S.K. Lin, C. Du, D.J. Macquarrie, A. Koutinas, R. Wang, C. Webb, J.H. [16] R.C. Wu, Y.Z. Xu, Y.Q. Song, J.A. Luo, D. Liu, A novel strategy for salts recovery
Clark, Chemical transformations of succinic acid recovered from fermentation from 1,3-propanediol fermentation broth by bipolar membrane
broths by a novel direct vacuum distillation-crystallisation method, Green electrodialysis, Sep. Purif. Technol. 83 (2011) 9–14.
Chem. 11 (2009) 193–200. [17] P. Wuithier, A. Giraud, Le Pétrole: Raffinage Et génie Chimique, Technip, Paris,
[6] A. Orjuela, A.J. Yanez, A. Santhanakrishnan, C.T. Lira, D.J. Miller, Kinetics of 1972 (Tome I. Editions).
mixed succinic acid/acetic acid esterification with Amberlyst 70 ion exchange [18] P. Wuithier, A. Giraud, Le Pétrole: Raffinage Et génie Chimique, Technip, Paris,
resin as catalyst, Chem. Eng. J. 188 (2012) 98–107. 1972 (Tome II. Editions).
[7] V.K.S. Pappu, V. Kanyi, A. Santhanakrishnan, C.T. Lira, D.J. Miller, Butyric acid [19] J.G. Zeikus, Chemical and fuel production by anaerobic bacteria, Annu. Rev.
esterification kinetics over Amberlyst solid acid catalysts: the effect of alcohol Microbiol. 34 (1980) 423–464.
carbon chain length, Bioresour. Technol. 130 (2013) 793–797. [20] J.G. Zeikus, M.K. Jain, P. Elankovan, Biotechnology of succinic acid production
[8] H.N. Pham, M.F. Doherty, Design and synthesis of heterogeneous azeotropic and markets for derived industrial products, Applied Microbiology
distillations—II: Residue curve maps, Chem. Eng. Sci. 45 (1990) 1837–1843. Biotechnology 51 (1999) 545–552.
[9] M.J.D. Powell, A fortran subroutine for solving systems of nonlinear algebraic
equations, in: P. Rabinowitz (Ed.), Numerical Methods for Nonlinear Algebraic
Equations, 1970 (Ch. 7).