You are on page 1of 11

Analytica Chimica Acta 1158 (2021) 338414

Contents lists available at ScienceDirect

Analytica Chimica Acta


journal homepage: www.elsevier.com/locate/aca

A computational simulation of electromembrane extraction based on


Poisson - Nernst - Planck equations
Roshanak Dolatabadi a, Ali Mohammadi a, **, Mostafa Baghani b, *
a
Department of Drug and Food Control, Faculty of Pharmacy, Tehran University of Medical Sciences, Tehran, Iran
b
School of Mechanical Engineering, College of Engineering, University of Tehran, P.O. Box 11155-4563, Tehran, Iran

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 Developing a 2D numerical model for


electromembrane extraction.
 Modeling the electro-driven extrac-
tion process through PoissoneNernst
ePlanck equations.
 The finite-element approach is
employed to solve the governing
equations.
 The model is successful in prediction
of main features observed in experi-
ments for basic drugs.
 Analyte diffusivity, distribution co-
efficients and level of protonation are
major effective variables.

a r t i c l e i n f o a b s t r a c t

Article history: Electromembrane extraction (EME) has attracted a great deal of interest in researchers because of its
Received 26 December 2020 advantages. For analysis, design and optimization purposes, understanding the ion transport mecha-
Received in revised form nisms in the organic supported liquid membrane (SLM) is of prominent importance, where the interplay
3 March 2021
between the passive diffusion and electric-driven mass transport across SLM affects the mass transfer. In
Accepted 12 March 2021
Available online 16 March 2021
present work, a 2D numerical simulation is developed to examine the mass transfer behavior and the
analyte recovery in EME devices. The presented model is capable of describing the effect of different
parameters on the recovery of the EME setup. Initial analyte concentration in the sample solution, SLM
Keywords:
Electromembrane extraction
thickness, applied potential, permittivity, diffusion coefficient, and the reservoir pH within both the
Drug concentration sample and acceptor, can be considered as process variables. Predicted results revealed that the most
Computational model important factors playing key role in EME, are the analyte diffusivity, distribution coefficient of the
Nernst-Planck-Poisson equations analyte as well as the level of protonation in both the donor and acceptor solutions. The proposed model
Finite element method is helpful in predicting the mass transfer behavior of the EME process in practical applications.
© 2021 Elsevier B.V. All rights reserved.

1. Introduction

Scaling down the analytical instruments, such as those used in


sample preparation, has been continuously developing in the last
* Corresponding author.
decades [1,2]. Sample treatment should be simple, automated,
** Corresponding author.
E-mail addresses: alimohammadi@tums.ac.ir (A. Mohammadi), baghani@ut.ac.ir miniaturized, fast, inexpensive, and safe [3]. Solid phase [4e7],
(M. Baghani). single drop [8], and hollow fiber liquid phase microextraction [9,10]

https://doi.org/10.1016/j.aca.2021.338414
0003-2670/© 2021 Elsevier B.V. All rights reserved.
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

have been introduced to address the issues that where unreachable between the total amount of ions in the donor and acceptor, while
with traditional methods. Regardless of their great characteristics, the major assumptions of the model remain the same. In other
some downsides were observed during their application. For attempts to model the EME, Gjelstad et al. [26] and Seip et al. [27]
instance, microfibers in solid phase microextraction are expensive, proposed a time-dependent model to describe the governing
organic drop may get unstable in single drop microextraction, and phenomena during EME. To account for the time-dependent dis-
in hollow fiber liquid phase microextraction the extraction is time- tribution of analytes, a simplified kinetic model was proposed,
consuming. Therefore, new techniques have been developed to where the main assumptions were: (a) considering a lag time for
resolve these problems [11,12]. the residence of the ions in the SLM, (b) the rate determining stage
Electromembrane extraction (EME) is a microextraction tech- of the process is the transport within the SLM, (c) the convection is
nique introduced for the first time in 2006 [13]. The dominating large enough so that mass transport in sample or acceptor solutions
principle for EME is depicted in Fig. 1. Target analytes are being do not limit the rate, and ultimately (d) the transport in the SLM is
extracted from a sample solution, via a thin supported liquid one-dimensional, and after a long time the entire analytes are
membrane (SLM) to an acceptor solution, where the SLM contains accumulated in the acceptor. Compared to the work of Gjelstad
an organic solvent (such as NOPE or 1-octanol) immobilized in a et al. [24], this model presents more details about the process, but
porous hollow fiber. The driving force for EME is an electric field still does not specify how the electric potential distribution should
applied over the SLM, generated by a power supply [13e16]. be evaluated or how the analytes are distributed along the SLM
Numerical simulations are helpful for understanding the SLM thickness or the acceptor solution. To consider the effect of applied
mass transport. Mass transport models discussing SLM processes electric potential, they have assumed that the distribution co-
are categorized into two main approaches. As the first approach, efficients are voltage-dependent. Whereas a voltage distribution is
some studies characterized the mass transfer across the SLM based being developed is the system, it should be pointed out that, which
on mass transfer resistances in donor, SLM and acceptor phases, voltages should be selected to show the dependence of the po-
where the mass transfer resistances are described by mass transfer tential to the distribution coefficients. In another work, Restan et al.
parameters for each phase such as the analyte diffusivity, pH, and [28] in 2020, introduced a simple theoretical model to discuss the
partitioning coefficients. This approach is simple, but it involves effect of pH for EME of hydrophobic basic analytes (logP > 2) under
calibrating all mass transfer parameters against experimental tests their acidebase equilibria together with partitioning of species to
[17e21]. To find the concentration distribution at different phases, the SLM. In 2020, Drobnjak et al. [29] made use of a numerical
the second approach requires solving a series of partial differential model to find electric potential distribution for EME with vials
equations [22,23]. However, both these approaches are appropriate made of conducting polymer. However, they did not solve the
for just one-dimensional description of the phenomena and do not diffusion problem and only reported the electric voltage across the
provide predictions where 2D or 3D distributions are desirable. SLM.
Experimental observations reveal that the applied electric po- These works provide good insight to the phenomena and are
tential as well as the organic solvent composition are key param- helpful, but they are zero-dimensional in space and cannot give
eters in EME [24]. To appropriately analyze, design, and optimize practical quantitative predictions [24,26,27,30]. Different parame-
the mass transport in EME, we need an accurate model to simulate ters such as electrodes distance and shape, setup geometry, applied
the whole process. In an early work to model EME, the analyte flux voltage distribution and pH of the solutions can affect the EME
across the SLM was formulated by Gjelstad et al. [24]. This relation recovery and should be taken into account in any attempt to model
was based on a simple reduced form of Nernst-Planck equation EME [16,24]. Although the available studies on EME experiments
borrowed from the literature, which was useful to understand are numerous, mathematical models are still scarce, despite being
some phenomena occurring in SLM. However, it did not determine are of high importance for the design and active control for the
how the applied electric potential on SLM should be computed. optimum use of these devices.
Besides, it did not provide any prediction for extraction times or In this paper, it is aimed to develop a numerical model to study
recoveries, since the time-dependent effects have not been dealt the effects of different parameters, i.e., applied potential, pH of
with in this model. In 2020, Hansen et al. [25] modified this model donor and acceptor phases, and distance of electrodes on EME. The
to consider the effect of ion-balance introduced as the ratio analyte’s concentration distribution in the donor, SLM and acceptor
phases over time are examined in detail. As discussed earlier, there
are few mathematical models available in literature, where none of
them considered the analytes distribution in the sample, SLM and
acceptor phases. More importantly, the Poisson Nernst Planck
equations as the main governing equations have not been solved
simultaneously in previous works. Solving these equations enables
us to better understand the diffusion patterns in different parts of
the EME system.

2. Governing equations

In this paper, the electric potential, is denoted by f. To find the


electric field, one may write [31,32]:

E ¼  Vf (1)

where E stands for the electric field intensity vector and V denotes
the gradient operator. Besides, the constitutive equation to intro-
Fig. 1. Schematic of the equipment for EME. duce the dielectric model can be given as [31,32]:
2
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

II The top and bottom walls of the setup ðU3 Þ are rigid, ion-
D ¼ ε0 εr E (2) impervious, and uncharged. The electric potential is contin-
uous at the SLM-electrolyte interfaces ðU4 Þ and ðU5 Þ.
in which ε0 ¼ 8:854  1012 F=m and εr are the permittivity of III No external pressure gradient is applied along the setup
vacuum and relative permittivity of the medium, respectively; and IV The ions partitioning occurs at the SLM-electrolyte in-
D expresses the electric displacement or electric flux density vector. terfaces. It could happen due to changes in the electrostatic
The Poisson equation is used to model the distribution of the and/or van der Waals interaction forces of the ions. One
electric flux density vector, as [31,32]: important phenomenon, for instance, is variation in the
electrostatic self-energies of ions arising from the permit-
V:D ¼ rv (3) tivity difference, called the Born energy. The distribution
coefficient, Pi , specifies the ratio of the interfacial ionic con-
P
where rv ¼ F N i¼1 ci zi . Besides, ci and zi are the i’th species con- centrations inside and outside the SLM, where applying the
centration and charge, respectively. (V: denotes the divergence Boltzmann distribution, is given as [33]:
operator.) Since the dimensions in this modeling are in the order of
microns and higher (much larger than the Debye length which is ci;M Dwi
below 50 nm), the electro-neutrality condition can be applied, Pi ¼ ¼ expð Þ (7)
ci kB T
P
which gives N i¼1 ci zi x0.
The well-known Nernst-Planck equation is used to predict the where kB denote the Boltzmann constant and Dwi stand for the
electro-diffusion of ion species in terms of ion concentration. In energy difference of the ith ionic species. ci;M denotes the analyte
addition to mass transfer due to the passive diffusion, the ionic concentration inside the membrane at its interfaces. This param-
species transfer by migration under an electric field should also be eter determines that how hydrophobic or hydrophilic the analyte is.
considered. Therefore, the mass balance becomes [31]: It should be highlighted that since the donor and acceptor solutions
may have different pH’s and the partition coefficient is highly
vci dependent to the pH, different partition coefficients for the left and
þ V:Ji þ u:Vci ¼ Ri (4)
vt right walls of the SLM should be assumed in the modeling. Thus, we
have [34]:
Ji ¼  Di Vci  zi um;i Fci Vf (5) 
Pi;s ¼ ci;M ci;s (8)
in which the flux vector of the ith ionic species is shown by Jj; ci
represents the concentration of species i, Di is its diffusion coeffi- where Pi;s denotes the partition coefficient at the interface of the
cient, zi denotes its ionic charge, and um,i is its ionic mobility and F sample and SLM, and ci;s shows the analyte concentration in the
refers to Faraday’s constant (96,485 [C/mol]). It is noted that u and sample. At the sample/SLM interface, the flux continuity should
Ri are used to consider the fluid velocity, and generation and also be satisfied in the form of [34]:
destruction of species i, respectively, which in this work have been
neglected. Moreover, for the sake of simplicity, the deformations of Di;s Vci;s þ zi;s um;i;s Fci;s Vfs ¼ Di;M Vci;M þ zi;M um;i;M Fci;M VfM
the SLM are ignored. (9)
The Nernst-Einstein equation in many cases is applied to asso-
ciate the species mobility to its diffusivity Di by Ref. [31]: Similar conditions are required for the SLM and acceptor solu-
tion interface as:
Di 
um;i ¼ (6)
RT Pi;a ¼ ci;M ci;a (10)

where T is the absolute temperature. The boundary conditions


associated with Eqs. (1)e(6) are tabulated in Table 1 based on the Di;M Vci;M þ zi;M um;i;M Fci;M VfM ¼ Di;a Vci;a þ zi;a um;i;a Fci;a Vfa
nomenclature given in Fig. 2. To determine the boundary condi- (11)
tions, the following assumptions are made:
in which Pi;a stands for the partition coefficient at the interface of
I The left wall of the donor reservoir (sample), ðU1 Þ is exposed the SLM and acceptor, and ci;s is the analyte concentration in the
to an electric potential of Vapp , while the right wall of the acceptor solution.
acceptor reservoir ðU2 Þ is grounded. Recovery for each analyte is computed according to:

Table 1
The boundary conditions associated with the computational domain in Fig. 2b. The parameter n is the outward normal vector.

Surface Electric potential Ionic conservation

U1 constant voltage bias ion-impenetrable


4 ¼ Vapp n,Nj ¼ 0
U2 Grounded ion-impenetrable
4 ¼0 n,Nj ¼ 0
U3 Uncharged ion-impenetrable
n,V4 ¼ 0 n,Nj ¼ 0
U4 Pi;s ¼ ci;M =ci;s
Di;s Vci;s þ zi;s um;i;s Fci;s Vfs ¼ Di;M Vci;M þ zi;M um;i;M Fci;M VfM
U5 Pi;a ¼ ci;M =ci;a
Di;M Vci;M þ zi;M um;i;M Fci;M VfM ¼ Di;a Vci;a þ zi;a um;i;a Fci;a Vfa

3
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

Fig. 2. Electromembrane extraction; (a) schematic of the equivalent setup of the system under consideration; (b) boundary conditions used in the system. The setup has the length
LN , the sample width wd , the SLM width ws , and the acceptor width wa and connects two identical large reservoirs filled with aqueous HCl solutions.

function for u. This relation allows us to properly address the


H
ca;final dV nonlinear diffusion phenomenon occurring across the SLM in
na;final V practical problems. However, in this work, it is assumed that the
R¼  100 ¼ H a  100 (12)
ns;initial cs;initial dV capability of the SLM remains the same along the extraction pro-
Vs cess. It is also noted that the pH in both the donor and acceptor
phases may change during the extraction. However, in most cases
where ns;initial and na;final stand for the number of analyte moles this pH change is not that significant and buffers are used to avoid
initially available in the sample phase and finally in the acceptor this effect. Thus, it is assumed that pH remains almost fixed during
phase, respectively. Va is the volume of acceptor phase, Vs the the process.
sample volume, ca;final the final concentration of analyte in the
acceptor solution and cd;initial is the initial analyte concentration in
the donor solution. 3. Solution procedure
The effective diffusion coefficient of the SLM is found through
the following relation in terms of its porosity ~ε and tortuosity t Equations (1)e(6) are a set of coupled nonlinear partial differ-
[35e37]: ential equations to which no analytical solution can be found for
different geometries and boundary conditions. In this work, the
Dm;i ¼ Di u ¼ Di~ε=t (13) governing equations are numerically solved subject to the associ-
ated boundary conditions in COMSOL Multiphysics 5.6, a software
As this ratio decreases, the diffusivity in SLM and therefore its working based on finite element method. The coupled Nernst-
transport capability decreases; i.e., a lower extraction efficiency. Planck-Poisson (PNP) equations were employed to compute the
It is worth noting that increasing the analyte ion concentration ionic concentration, as well as the electric potential distributions. A
lowers the diffusion coefficients in the SLM. In other words, the total number of ~32,000 type free triangular mesh elements was
efficiency of the membrane decreases, as the process goes on. To found to be adequate to arrive at mesh-independent outcomes.
consider this effect on the diffusion coefficients of the ions across Extremely fine meshes were used near the SLM walls to arrive at
the membrane, one may assume a monotonically-decreasing the correct solution. The mesh network used in all simulations

Fig. 3. The mesh network used in all simulations.

4
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

Table 2 presenting different numerical simulations. The sample analyte


Geometric values of the EME setup. concentration, the SLM thickness, the applied voltage, the permit-
Parameter Description Value tivity, as well as the diffusion coefficient, the reservoir pH in both
the sample and acceptor phases can be considered for parametric
CD Drug concentration in donor phase  2:6  103 mM
LN Length of setup 31 mm study. It is attempted to simulate a real EME setup; for this purpose,
wa Acceptor thickness 600mm geometrical parameters are assumed as listed in Table 2, while the
wd Donor thickness 4:5 mm material model parameters used in the numerical simulations are
ws SLM thickness 200mm shown in Table 3. In this paper, basic drugs are considered, implying
that the EME setup is cation-selective. It is worthwhile to mention
that in EME experiments, a stirrer mixed agitates the whole setup
Table 3
to accelerate the extraction process and ensure efficient convection
Parameters required for simulation of EME.
in the donor solution. This means the ion transport in the sample is
Parameter Description Value much faster than without stirring, and the time-consuming part of
DD Diffusivity of drug ions 1:0  1011 m2 =s the process is the ion transport across the organic SLM, acceptor
u Porosity to tortuosity ratio of the SLM x0:1 solution and only its neighboring media in the sample solution.
R Universal gas constant 8:3145 J=ðmol:KÞ This issue has been implemented in the modeling by applying a
F Faraday constant 96500 C=mol
larger diffusion coefficient for the analyte in the donor phase,
KB Boltzmann constant 1:381  1023 J=K
which makes the sample have a monotonous ions concentration
T Absolute temperature 298:15 K
ε=ε0 Relative permittivity of solutions 80 distribution all over the donor solution.
εm =ε0 Relative permittivity of SLM 10 [38] In the following since we are interested in real-world drugs,
Dropridol as a basic drug is used in the donor solution. In Fig. 4,
different pairs of pH, (pHs, pHa), for the sample and acceptor are
which led to convergent results is illustrated in Fig. 3. The direct examined. Fig. 4a illustrates the effect of acceptor pH and applied
MUMPS solver with a maximum relative tolerance of 108 was voltage on the EME recovery, while pHs ¼ 2. As expected, higher
applied. electric potentials give higher recoveries. For this drug, at pH ¼ 2,
logD ¼ 0.46 and z ¼ 1; at pH ¼ 5, logD ¼ 1.26 and z ¼ 0.98; at
4. Results and discussions pH ¼ 7, logD ¼ 2.82 and z ¼ 0.36 [39]. Considering these distri-
bution coefficients, one may realize that the drug is hydrophilic at
In this section, the proposed model results are reported by pH ¼ 2. At pH ¼ 5, it is hydrophobic to some extent, while, at

Fig. 4. Predicted recoveries at different voltages and different acceptor pH’s for a) pHs ¼ 2, b) pHs ¼ 5, c) pHs ¼ 7. d) Predicted recovery in sample, SLM and acceptor (solid lines
stand for the case with pHs ¼ pHa ¼ 2, V ¼ 50v, while dash-dot lines denote the case with pHs ¼ 5, pHa ¼ 2, V ¼ 20v).

5
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

50
Sample SLM Acceptor
40
Drug concentration [ ppm ]

SLM pHs=5, pHa=2, V=50v


30 t = 30 s
t = 3 min
t = 6 min
20 t = 8 min

10

0
4 4.2 4.4 4.6 4.8 5 5.2
x-position [ mm ]
Fig. 5. The drug concentration distribution at different times across donor, SLM and
acceptor. It should be noted that only a portion of the donor phase which has a
gradient in distribution has been shown. Fig. 7. Effect of smaller diffusion coefficient at different sample and acceptor pH’s
where D0 ¼ 1011 m2/s.

pH ¼ 7, its hydrophobicity gets very high. Thus, apparently it seems


acting as a better solution compared to (2, 5). Since at pHa ¼ 5 the
that the best candidate for the sample and acceptor pH’s are (7, 2).
ion charge is almost the same, a smaller distribution coefficient at
However, the analyte ion charge dependence to pH, which is
pHa ¼ 2 allows us to reach a higher recovery. Fig. 4b examines the
playing a key role in reaching to high recoveries in EME, necessarily
impact of the same parameters at pHa ¼ 5. The pH pair of (5, 2)
does not follow the same trend as distribution coefficient. It should
makes the extraction faster than (2, 2). This is stemming from the
be noted that there are two competing phenomena that are of high
fact that pHs ¼ 5 corresponds to a higher distribution coefficient,
importance in selection of the best donor and acceptor solutions.
which means the analyte is more likely to migrate to SLM. At pH
Now, looking at Fig. 4a, it is revealed that the pH pair of (2, 2) is
condition of (5, 2), we also have the best pHa because a low logD
leads to a fast ion migration from the SLM to the acceptor phase,
while its ion charge has been preserved. In all cases, where pHa ¼ 7,
the drug is so hydrophobic that it would not simply leave the SLM.
On the other hand, its ionic charge diminishes to 0.35. Hence, the
recovery drops to smallest values.
For pHs ¼ 7, in spite of its great distribution coefficient, the drug
loses a large part of its charge which makes the EME more slowly.
Possessing a higher charge along with a large logD can potentially
be a good donor. This means, although at pHs ¼ 7 we have a high
distribution coefficient, due to its small charge, it fails to be the
most desirable pH for the sample. Arranging the recoveries based
on the sample and acceptor solutions pH, one may find the
following order:

Fig. 6. The drug concentration contours (pHs ¼ 2, pHa ¼ 2, and V ¼ 50v) it is note-
worthy that only a portion of the donor phase which has a meaningful gradient in
distribution has been shown. Fig. 8. Electric potential distribution across the donor, SLM and acceptor.

6
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

Fig. 9. Variation of Ionic charge and logD of some different drugs versus pH. Data are taken from Ref. [39].

observed in the donor during extraction, with a steady rise in the


amount accumulated in the acceptor phase. Despite the lower
ð5; 2Þ > ð2; 2Þ > ð5; 5Þ > ð2; 5Þ > ð7; 2Þ > ð7; 5Þ > ð5; 7Þ > ð2; 7Þ > ð7; 7Þ applied voltage, because of a larger logD for pHs ¼ 5, the depletion
(14) rate is higher. For the case of (5, 2) under 20 v, a large portion of the
ions are accumulated in the SLM, while the remaining ions go into
This type of trend which implies a best pair of pH exists for each
the acceptor solution. This behavior acts similar to a capacitance
analyte, is in agreement with experimental observations reported
that releases the drug ions in a decaying exponential format. This
in literature [40]. It should be noted that this order is not solely
type of behavior has been experimentally observed and reported
based on the distribution coefficient difference at SLM interfaces, or
for different drugs by Seip et al. [27]. On the other hand, this
just the ionic charge. A combination of these parameters along with
capacitance-like response is faster for the case of (2, 2) under 50 v. A
the SLM properties and applied potential affect the pH selection
higher applied electric potential together with small values of
procedure and therefore EME recovery.
logD’s at both interfaces of the SLM make the SLM release faster the
To arrive at a better understanding of the ions’ migration, the
accumulated ions.
relevant recoveries in the sample, SLM and acceptor are depicted
The analyte concentration distribution along the ion migration
for the pH pair of (5, 2) under applied voltage of 20 v, and pH pair of
direction across the donor, SLM and acceptor solution at different
(2, 2) under 50 v in Fig. 4d. The analyte ions are depleted from the
times is shown in Fig. 5. As an example, the case with pHs ¼ 5,
sample in a short time. Gjelstad et al. [26] and Seip et al. [27] in
pHa ¼ 2, and V ¼ 50v has been reported in this Fig, where it can
their experiments also mentioned that a rapid depletion was
7
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

Fig. 10. Recovery of the EME system at extraction time of 15 min and voltage of 20 V for a) Ciprofloxacin, logP ¼ 0.81, (basic), (1,5); b) Ciprofloxacin, logP ¼ 0.81, (acidic), (11,11);
c) Topiramate, logP ¼ 0.13, (acidic), (12,12); d) Amisulpride, logP ¼ 0.25, (basic), (7,7); e) Sitagliptin, logP ¼ 1.26, (basic), (8,6); f) Buspirone, logP ¼ 1.78, (basic), (7,5); g) Warfarin,

8
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

provide us useful information about the kinetics of the entire


process. It should be pointed out that only a portion of the donor
phase which has a gradient in ions distribution has been plotted in
Fig. 5. At the early stages of the EME, e.g., 30 s, due to the high value
of logD ¼ 1.26, at the donor/SLM interface, the analyte is more
willing to exit the donor and reside in SLM. After 6 min, this time at
the SLM/acceptor interface the small logD ¼ 0.46 acts as a positive
driver to expedite the extraction. At this time, a significant fraction
of the analyte ions moves more toward the other side of the SLM
which is in contact with the acceptor. In the meantime, some ions
pass the SLM, while the others gain more energy to overcome the
energy level required to get out of the SLM. Now, if we give enough
time to the process to keep on, one may observe that at 6 min,
almost all the analyte ions have migrated to the acceptor, where
most of them have resided next to the grounded electrode. To give a
more graphical insight to the mass transfer process, the analyte
concentration contours for the case with pHs ¼ 2, pHa ¼ 2, and
V ¼ 50v are plotted in Fig. 6. Fig. 11. Average recovery of the EME system for simultaneous extraction of Topiramate
Fig. 7 is devoted to examining the effect of the analyte diffusion and Warfarin.
coefficient on the predicted EME recovery. It is revealed that this
parameter is of high importance in the modeling. Smaller diffusion
containing for example acidic drugs of Topiramate and Warfarin,
coefficient significantly reduces the recovery in all cases. Since this
where their best pair of (pHs, pHa) was (12,12) and (7,9), respec-
parameter is active in both the non-electric and electric driven
tively (see Fig. 10). However, the average system recovery meets its
terms in Nernst-Planck equation, its amplification or decrease
maximum at (pHs, pHa) ¼ (13,13). The applied conditions are the
highly affects the entire extraction operation.
same as those reported for Fig. 10.
The distribution of electric potential across the sample solution,
The following items can be included in the model to more
organic SLM and acceptor solution is depicted in Fig. 8. Whereas the
realistically simulate this electro-driven diffusion process.
electro-neutrality condition stands accurate in all cases in this pa-
per, electric potential does not vary much during the simulation.
i. The real 3D setup should be simulated to arrive at a more
Thus, only one distribution has been given for each applied voltage.
reliable model. In this paper, only the space between the
The organic SLM has a much larger electric resistance and much
electrodes has been considered in a 2D model.
smaller permittivity compared to the donor and acceptor phases;
ii. The effect of fluid velocity can be added to the model through
hence, its voltage drop is more abrupt.
simultaneously solving the Navier-Stokes equations to ac-
To show the capability of the presented model for different basic
count for the convective term present in Nernst-Planck
and acidic drugs with various logP’s, the system recovery is opti-
diffusion equation.
mized for different pHs and pHa. The distribution of the ionic charge
iii. The analyte may deprotonate partly during passing through
as well as the logD of the drugs are illustrated in Fig. 9. It is inter-
the organic SLM. This can be due to the material and
esting that for Ciprofloxacin, depending on the pH, it can play role
geometrical properties of the SLM. The level of this depro-
as both acidic and basic analytes. However, due to its low
tonation as well as possible ion pairing can be applied to the
logP ¼ 0.81, as shown in Fig. 10a and 10b, the system recovery is
model. However, it involves specific experiments to have
very small. One may observe from Fig. 10 that generally by
necessary evidences.
increasing the logP of the drug, its recovery increases. However, this
iv. The analyte molecules could be trapped in the SLM pores and
rise depends on the ionic charge of the drug at both phases and its
reduce the efficiency of the SLM by lowering the analyte
relevant logD’s. It should be noted that for acidic drugs, in the
diffusion coefficient over time. This issue has not been
simulations the locations of electrodes have been switched. The
considered in present work.
best pair of pH of the system for each drug is shown on the caption
v. It is assumed that no electrolysis occurs around the
of Fig. 10.
electrodes.
In this paper, it was attempted to present a mathematical
vi. High voltages could lead to electrochemical degradation
modeling of the electromembrane extraction. Employing the
which has not been included in this model. This effect could
coupled Nernst-Planck-Poisson equations subjected to necessary
explain the smaller recoveries in experiments.
boundary conditions associated with the problem, the ionic con-
vii. The effect of double layers on the analyte recovery over the
centrations of the species, and the electric potential distributions
extraction can be considered in the modeling especially for
were computed in a 2D field consisting of the donor solution,
thin SLMs.
organic SLM and acceptor solution. The proposed model was suc-
viii. The possible effect of salts in the sample solution on the
cessful in qualitatively reproduction of the experimental observa-
analyte recovery should also be accounted for.
tions available in literature.
ix. It is assumed that the permittivity and electric resistance of
It is noted that employing this model, one can study the
the SLM remain the same during the extraction, while in
simultaneous extraction of different analytes and optimize the
practice it may lose a part of its resistance which leads to a
conditions for the best recovery of all analytes at the same time,
smaller electric potential drop across the SLM and conse-
which is obviously different from the optimized conditions for only
quently lowers the total recovery of the whole system.
each analyte. Fig. 11 depicts the average recovery for a solution

logP ¼ 2.74, (acidic), (7,9); h) Hydroxy chloroquine, logP ¼ 2.89, (basic), (8,6); i) Mycophenolic acid, logP ¼ 3.53, (acidic), (10,12); j) Haloperidol, logP ¼ 3.66, (basic), (7,4). The best
pair of pH of the system (pHs, pHa) for each drug is shown in parentheses.

9
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

It should be noted even though these modifications can improve endogenous substances - a review, Anal. Chim. Acta 1077 (2019) 67e86.
[7] M.S. Jagirani, M. Soylak, A review: recent advances in solid phase micro-
the validity of the proposed model; however, most of them involve
extraction of toxic pollutants using nanotechnology scenario, Microchem. J.
detailed experiments to appropriately calibrate the proposed 159 (2020).
model. [8] S. Tang, T. Qi, P.D. Ansah, J.C. Nalouzebi Fouemina, W. Shen, C. Basheer,
H.K. Lee, Single-drop microextraction, Trac. Trends Anal. Chem. 108 (2018)
306e313.
5. Conclusion [9] E. Carasek, J. Merib, Membrane-based microextraction techniques in analytical
chemistry: a review, Anal. Chim. Acta 880 (2015) 8e25.
In this paper, it was attempted to present a mathematical [10] A. Esrafili, M. Baharfar, M. Tajik, Y. Yamini, M. Ghambarian, Two-phase hollow
fiber liquid-phase microextraction, Trac. Trends Anal. Chem. 108 (2018)
modeling of the electromembrane extraction. Employing the 314e322.
coupled Nernst-Planck-Poisson equations subjected to necessary [11] N. Drouin, P. Kuba n , S. Rudaz, S. Pedersen-Bjergaard, J. Schappler, Electro-
boundary conditions associated with the problem, the ionic con- membrane extraction: overview of the last decade, Trac. Trends Anal. Chem.
113 (2019) 357e363.
centrations of the species, and the electric potential distributions [12] F. Zarghampour, Y. Yamini, M. Baharfar, G. Javadian, M. Faraji, On-chip elec-
were computed in a 2D field consisting of the donor solution, tromembrane extraction followed by sensitive digital image-based colorim-
organic SLM and acceptor solution. The proposed model was suc- etry for determination of trace amounts of Cr(vi), Anal. Methods 12 (4) (2020)
483e490.
cessful in qualitatively reproduction of the experimental observa- [13] C. Huang, Z. Chen, A. Gjelstad, S. Pedersen-Bjergaard, X. Shen, Electro-
tions available in literature. membrane extraction, Trac. Trends Anal. Chem. 95 (2017) 47e56.
In this paper for the first time, numerical methods (finite [14] R. Dolatabadi, A. Mohammadi, S. Nojavan, S. Yaripour, A. Tafakhori,
M. Shirangi, Electro-membrane extraction-high-performance liquid
element method) were used to solve the coupled governing equa-
chromatography-ultraviolet detection of phenobarbital and phenytoin in
tions (Nernst Planck and Poisson equations) in two dimensions, human plasma, saliva and urine, J. Chin. Chem. Soc. (2021). In press.
while in previous studies on modeling of EME only one- [15] S. Pedersen-Bjergaard, Electromembrane extractiondlooking into the future,
Anal. Bioanal. Chem. 411 (9) (2019) 1687e1693.
dimensional qualitative results were discussed. Moreover, this
[16] C. Huang, K.F. Seip, A. Gjelstad, S. Pedersen-Bjergaard, Electromembrane
model enables us to optimize different effective parameters of EME. extraction for pharmaceutical and biomedical analysis - Quo vadis,
For example, in Fig. 10, such optimization was carried out to find the J. Pharmaceut. Biomed. Anal. 113 (2015) 97e107.
best pHs and pHa in order to maximize the system recovery. Such [17] P.J. Harrington, G. Stevens, Steady-state mass transfer and modeling in hollow
fiber liquid membranes, J. Membr. Sci. 192 (2001) 83e98.
approach can be used for finding other parameters, for example one [18] O.N. Ata, S. Çolak, Modelling of zinc transport through a supported liquid
may change the electrode locations, shapes and distances and membrane, Hydrometallurgy 80 (3) (2005) 155e162.
examine their effect on the system recovery. To the best knowledge [19] Z. Ren, W. Zhang, Y. Liu, Y. Dai, C. Cui, New liquid membrane technology for
simultaneous extraction and stripping of copper(II) from wastewater, Chem.
of the authors, there is no other works available in literature with Eng. Sci. 62 (22) (2007) 6090e6101.
such features. Besides, with the aid of this model, one can inves- [20] S. Ammari Allahyari, A. Minuchehr, S.J. Ahmadi, A. Charkhi, Th(IV) transport
tigate the simultaneous extraction of different analytes and find the from nitrate media through hollow fiber renewal liquid membrane, J. Membr.
Sci. 520 (2016) 374e384.
conditions for the highest recovery of all analytes at the same time, [21] S. Panja, P.K. Mohapatra, P. Kandwal, S.C. Tripathi, Uranium(VI) pertraction
which may differ from those of each analyte. The presented model across a supported liquid membrane containing a branched diglycolamide
is a useful tool can be applied in analysis and design of different carrier extractant: Part III: mass transfer modeling, Desalination 285 (2012)
213e218.
analytes in EME. It also can help the researchers to arrive at a better [22] T. Wannachod, N. Leepipatpiboon, U. Pancharoen, K. Nootong, Separation and
understanding of the phenomena occurring in EME. However, mass transport of Nd(III) from mixed rare earths via hollow fiber supported
some aspects should be added to this model to give more reliable liquid membrane: experiment and modeling, Chem. Eng. J. 248 (2014)
158e167.
predictions especially under conditions that have not been
[23] O.N. Ata, Mathematical modelling of unsteady-state transport of metal ions
assumed in this modeling. through supported liquid membrane, Hydrometallurgy 87 (3) (2007)
148e156.
[24] A. Gjelstad, K.E. Rasmussen, S. Pedersen-Bjergaard, Simulation of flux during
CRediT authorship contribution statement
electro-membrane extraction based on the Nernst-Planck equation,
J. Chromatogr. A 1174 (1e2) (2007) 104e111.
Roshanak Dolatabadi: Methodology, Software, Validation, [25] F. Hansen, F. Jaghl, E. Leere Øiestad, H. Jensen, S. Pedersen-Bjergaard,
Writing e original draft. Ali Mohammadi: Conceptualization, C. Huang, Impact of ion balance in electromembrane extraction, Anal. Chim.
Acta 1124 (2020) 129e136.
Methodology, Writing e review & editing, Supervision. Mostafa [26] A. Gjelstad, H. Jensen, K.E. Rasmussen, S. Pedersen-Bjergaard, Kinetic aspects
Baghani: Methodology, Software, Writing e review & editing, of hollow fiber liquid-phase microextraction and electromembrane extrac-
Supervision. tion, Anal. Chim. Acta 742 (2012) 10e16.
[27] K.F. Seip, H. Jensen, M.H. Sønsteby, A. Gjelstad, S. Pedersen-Bjergaard, Elec-
tromembrane extraction: distribution or electrophoresis? Electrophoresis 34
Declaration of competing interest (5) (2013) 792e799.
[28] M.S. Restan, S.B. Ramsrud, H. Jensen, S. Pedersen-Bjergaard, Influence of acid-
base dissociation equilibria during electromembrane extraction, J. Separ. Sci.
The authors declare that they have no known competing 43 (15) (2020) 3120e3128.
financial interests or personal relationships that could have [29] M. Drobnjak, F. Hansen, E.L. Øiestad, T. Løvli, R. Trones, Ø.G. Martinsen,
appeared to influence the work reported in this paper. S. Pedersen-Bjergaard, Electromembrane extraction with vials of conducting
polymer, LC-GC N. Am. 38 (8) (2020) 435e439.
[30] C. Huang, H. Jensen, K.F. Seip, A. Gjelstad, S. Pedersen-Bjergaard, Mass transfer
References in electromembrane extractiondthe link between theory and experiments,
J. Separ. Sci. 39 (1) (2015) 188e197.
[1] N. Drouin, S. Rudaz, J. Schappler, New trends in sample preparation for bio- [31] M. Khatibi, S.N. Ashrafizadeh, A. Sadeghi, Covering the conical nanochannels
analysis, Am. Pharmaceut. Rev. 19 (2016) 62e66. with dense polyelectrolyte layers significantly improves the ionic current
[2] L. Xia, J. Yang, R. Su, W. Zhou, Y. Zhang, Y. Zhong, S. Huang, Y. Chen, G. Li, rectification, Anal. Chim. Acta 1122 (2020) 48e60.
Recent progress in fast sample preparation techniques, Anal. Chem. 92 (1) [32] A. Ganjizade, S.N. Ashrafizadeh, A. Sadeghi, Effect of ion partitioning on
(2020) 34e48. electrophoresis of soft particles, Colloid Polym. Sci. 297 (2) (2019) 191e200.
[3] M. Valcarcel, S. C
ardenas, R. Lucena, Microextraction techniques, Anal. Bioanal. [33] A. Sadeghi, M. Azari, S. Hardt, Electroosmotic flow in soft microchannels at
Chem. 406 (8) (2014) 1999e2000. high grafting densities, Phys. Rev. Fluids 4 (6) (2019), 063701.
[4] C.L. Arthur, J. Pawliszyn, Solid phase microextraction with thermal desorption [34] T. Wang, T. Xie, C. Xu, Numerical investigations of micro-SLM extraction/
using fused silica optical fibers, Anal. Chem. 62 (19) (1990) 2145e2148. stripping in a spiral channel, Chem. Eng. Sci. 212 (2020) 115344.
[5] J. Pawliszyn, Solid Phase Microextraction : Theory and Practice, Wiley-VCH, [35] A. Gabelman, S.-T. Hwang, Hollow fiber membrane contactors, J. Membr. Sci.
New York; Toronto, 1997. 159 (1) (1999) 61e106.
[6] S. Huang, G. Chen, N. Ye, X. Kou, F. Zhu, J. Shen, G. Ouyang, Solid-phase [36] B. Ghanbarian, A.G. Hunt, R.P. Ewing, M. Sahimi, Tortuosity in porous media: a
microextraction: an appealing alternative for the determination of critical review, Soil Sci. Soc. Am. J. 77 (5) (2013) 1461e1477.

10
R. Dolatabadi, A. Mohammadi and M. Baghani Analytica Chimica Acta 1158 (2021) 338414

[37] F. Fadaei, S. Shirazian, S.N. Ashrafizadeh, Mass transfer simulation of solvent J. Chem. Eng. Data 45 (4) (2000) 636e641.
extraction in hollow-fiber membrane contactors, Desalination 275 (1) (2011) [39] V. MarvinSketch, 16.9. 26.0, Calculation Module Developed by ChemAxon,
126e132. 2020.
[38] G. Ritzoulis, D. Missopolinou, S. Doulami, C. Panayiotou, Relative permittiv- [40] S. Yaripour, A. Mohammadi, I. Esfanjani, R.B. Walker, S. Nojavan, Quantitation
ities, densities, refractive indices and ultrasound velocities of the binary sys- of zolpidem in biological fluids by electro-driven microextraction combined
tems of g-butyrolactone with methanol, ethanol, 1-butanol, and 1-octanol, with HPLC-UV analysis, EXCLI J. 17 (2018) 349.

11

You might also like