Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract: In this paper, the analytical validation small dimensions, this approximation fails [3]
of Poisson-Boltzmann (PB) equation computed [7]. Despite the explosive growth of multiscale
with Comsol Multiphysics, in the case of a modeling for microfluidics, where the continuum
polarized surface in contact with the electrolyte is usually coupled to Molecular Dynamics
[1]-[2], is first presented. Comsol Multiphysics techniques, we investigated here the use of
algorithms easily handle the highly nonlinear coupled continuum models, based on the
aspect of the PB equation. The limitations of the Poisson-Boltzmann (PB) equation. For us, it is
PB model, that considers ions as pointlike interesting to represent the EDL using the
charges, are outlined. To account for the steric Comsol Multiphysics software application
effects of the ion crowding at the charged because its strong coupling to macroscopic
surface, the Modified Poisson-Boltzmann model, equations (Navier-Stokes in our case) is possible.
proposed by Kilic et al. [3], is analysed for
symmetric electrolytes. The MPB equation is 2. Theory
then coupled to the complex AC electrokinetic
and the Navier-Stokes equations to simulate the 2.1 The electrical double layer
AC electroosmosis flow observed inside an
interdigitated electrodes microsystem [4]-[6]. In this paper, we consider that electrodes are
ideally polarizable i.e. that no electron transfer
Keywords: numerical simulation, Poisson- (electrochemical) reactions occur at the
Boltzmann, Finite Element Method, AC electrode. The model which gave rise to the term
electrokinetics. 'electrical double layer' was first put forward in
the 1850's by Helmholtz. In order for the
1. Introduction interface to remain neutral, the charge held on a
polarized electrode is balanced by the
The Electrical Double Layer (EDL) represents redistribution of ions close to the electrode
the interface between a solid surface (polarized surface. In Helmholtz's view of this region, the
electrode) and an electrolyte. The charged attracted ions are assumed to approach the
surface attracts nearby counterions and repels electrode surface with a distance assumed to be
coions present in the solution. In microsystems, limited to the size of the ion: the overall result is
the same electrostatic phenomenon is also two layers of charge (the double layer) and a
present around charged nanoparticles linear potential drop which is confined to this
(biomolecules, latex beads…) immersed into an region only. A later model put forward by Gouy
electrolyte: they experience electrostatic and Chapman supposed that ions are able to
interactions which give rise to a counterion move in solution and so the electrostatic
cloud. The EDL or the counterion cloud is likely interactions are in competition with Brownian
to react to the applied electric fields and can motion. The result is still a region close to the
strongly influence various electrical phenomena electrode surface containing an excess of one
such as dielectrophoresis, electrophoresis of type of ion but now the potential drop is
polyelectrolytes (DNA, proteins,…) or AC exponential and occurs over the region called the
electrokinetic flows. diffuse layer:
RC circuit models are widely used by
electrochemists for representing the EDL.
However, in microsystems where applied electric
fields can be very strong because of the very
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble
surface
potential ψE
zeta
( n i being the number of ions i in the electrolyte
potential ζ
formula, c ∞ is the bulk concentration), T is the
temperature (K) and k the Boltzmann constant
(1.38 10-23 J/K). e is the proton charge (1.6 10-
19
C) and zi is the ion i charge number. For
convenience, concentrations can be expressed in
z
compact diffuse layer bulk (∞) Molar unit (M = mole/L): M i = 103 ci N A
layer where NA is the Avogadro’s number (6.022
1023).
Figure 2. The Stern representation of the EDL Each ion distribution corresponds to a
composed of the compact layer and the diffuse layer. volume free charge distribution qi such that:
The variation of the electrical potential V thru the
EDL (red line) is represented for the case of a qi = zie ci (2)
positively charged surface.
In return, the total free charge density
The potential at the interface between the
compact and the diffuse layer is called the zeta q = ∑ qi = ∑ z ie ci (3)
potential ζ which can be determined from i i
( )
r r z eψ This equation admits the following solution:
− i
At room temperature (298 K), for monovalent Tests are performed on the geometry of
Figure 3 for two different types of electrolytes
ions (z = 1) ψT ~ 26 mV, for divalent ions, ψT ~
(1:1 and 2:2), variable bulk concentrations,
13 mV. Under assumption (8), equation (10) can
be linearized: variable surface potentials ψE. In the two
following figures, the curves were drawn for
r
( r
∇. − ε∇ψ = −
kT
)
2 z2 e2 c∞
ψ (9)
surface potentials ψE of +50 mV and +1V and
bulk concentrations c ∞ of 0.01M and 0.1M.
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble
0.035 GC
PB 150
PSI (V)
0.03 GC
100
0.025
analytic Debye Huckel
0.02
0.01 M 50
0.015 0 −5 −4 −3 −2 −1 0
10 10 10 10 10 10
0.01
bulk concentration (M)
0.1 M
0.005
0
0 1 2 3 4 5 6 7 8 9 10
z (nm) Figure 6. The EDL width L EDL according to the bulk
concentration for KCl, at +0.1V: from GC solution
Figure 4. Comparison of numerical (black = PB, blue (blue), PB solution (black) and Debye length formula
= GC) and analytical (red) electric potentials for a 1:1 (11) (red).
electrolyte at bulk concentrations of 0.01 M and 0.1
M. ψE = +50mV. z (nm) is the distance from the
electrode surface. 3.2 Limitations of the PB equation validity
1
One could expect that the PB equation (and
0.9 1 V − 0.1 M
0.8
the GC equation for binary symmetric
0.7
electrolytes), when fully solved in the non linear
0.6
PB regime (ψ > ψT), would give a good estimation
PSI (V)
GC
0.5 analytic of the EDL. However, even at large applied
0.4 potentials, the PB and the CG equations have
0.3 limited applicability.
0.2
One of the assumptions made in the PB equation
0.1
is that ions are pointlike charges. This means that
0
0 0.1 0.2 0.3 0.4 0.5
z (nm)
0.6 0.7 0.8 0.9 1 the ions are considered to have no size. The
consequence is that the PB equation can predict
Figure 5. Comparison of numerical (black = PB, blue an infinite concentration of counter-ions near the
= GC) and analytical (red) electric potentials for a 1:1 charged surface, which is not realistic. For
electrolyte at bulk concentration of 0.01 M and ψE = example, for the aqueous electrolyte (Na+, Cl-),
+1V. at a bulk concentration of 1 mM, ambient
Figure 6 compares the Debye length with the temperature and ψE = +1V, the surface charge
EDL width ( L EDL ) computed from the PB and calculated from expression (6) corresponds to a
concentration of spherical counter-ions (Cl-) of 5
the GC solutions according to the bulk
1040 ions/m3 hence 8.3 1016 M! This would mean
concentration of a KCl electrolyte, for the
that the chloride ion radius is 1.5 10-14m which is
applied voltage +0.1V. As expected, the Debye
10 000 times smaller than the real value. Here
length underestimates the EDL width and the
we use the numerical PB model to determine the
error committed when using (11) is quite
area in which the PB equation is valid, the limit
important. It increases with the bulk
being given by the steric effect which
concentration.
corresponds to a maximum concentration
reached at the charged surface due to the
hydrated ion crowding.
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble
200
−200
K+ in KCl
−300 Mg2+ in MgCl2
−400
−500 −5 −4 −3 −2 −1 0
10 10 10 10 10 10
bulk concentration (M)
zoom
200
surface concentration (% of max)
0.01M
1
PB 1V, Cl− 0.01M
150 GC
0.9
0.8
steric limit MPB
100
0.7
PSI (V)
0.6
50
0.5
0.4 MPB
0
0 0.2 0.4 0.6 0.8 1
PSI (V) 0.3
0.05
120
0.04 PB
1V, 0.01M
0.03 100
Cl concentration (M)
MPB
0.02
80
0.01
60
0
0 1 2 3 4 5 6 7 8 9 10
z (nm) 40
−
20
Figure 9. Validation of the MPB equation on a quasi-
linear case (KCl electrolyte, +0.1V, 0.01 M) by 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
comparison with PB and GC solutions. z (nm)
For higher voltages (+1V, see Figure 10), the Figure 11. The Cl- concentration profile given by the
MPB and PB solutions do not overlap anymore MPB equation for a KCl electrolyte of bulk
concentration 0.01M and a high applied surface
because the PB equation validity fails. The MPB
potential (+1V).
equation predicts an EDL width much bigger (~
0.2 nm) than the one given by the PB solution 3.4 Getting convergence during the PB and
(<< 0.01 nm). The crowding effect at the charged the MPB equation computation
surface repels the counterions into the diffuse
The previous numerical results show that the
layer and provides a much larger EDL width
higher the electrode potential ψ E is, the higher
than what the PB equation is predicting.
the non linearity of the problem is. To obtain a
1
−
good convergence of the solution, several ‘tricks’
0.9 1V, Cl 0.01M
0.8
are used. First, the mesh is highly refined near
0.7
the electrode surface where gradients are very
0.6 GC steep. The one dimensional character of the
PSI (V)
leading to a weak coupling between the electrical where σ is the bulk conductivity (S/m) , V* the
stress and the fluid flow. complex electrical potential of real part Re(V*)
In most papers which deal with ac = V and σ* the complex conductivity.
electroosmosis modeling, the thin double layer On insulated surfaces in contact with the
approximation under the linear regime is used for electrolyte, the boundary condition associated to
the EDL [4]-[6]. The EDL is estimated from the (16) is of Neumann homogeneous type:
Debye-Hückel theory and is not included inside
the computation domain. The electric field inside ∂V *
Re − σ * = 0 (17)
the bulk is computed with the ac electrokinetic
∂n
equation (see (16)) connected to the EDL thru a
Neumman boundary condition (18). The fluid with n being the outer normal.
motion is obtained from the Navier-Stokes Above the charged electrodes, which are
equation (20) where the electrical stress acts as a assumed perfectly polarizable (no
slip velocity imposed as a boundary condition on electrochemical reactions), the bulk is in contact
the electrode surface. This slip velocity is with the EDL. Equation (16) is connected to the
estimated from empirical parameters (the MPB equation at this interface where the
capacitance of the compact and the diffuse conservation of the normal current density
layers) and the tangential component of the gives:
electric field given by (16).
Our goal here is to take off these empirical ∂V *
Re − σ * = iω C EDL ∆ψ EDL
∂n
(18)
parameters from the numerical model. This
supposes that the EDL is fully represented inside
the computation domain for the fluid motion. where ∆ψ EDL is the potential drop across the
The electrical volume force acting inside the
EDL:
EDL on the fluid is not transformed into a slip
velocity. The numerical difficulty here is the ∆ψ EDL = ψ E − V (19)
multiscale coupling that has to be performed: the
EDL, which is tens of nm wide, has to be and C EDL is given by (15).
included in a microsystem which size reaches 1 The time-averaged fluid flow is obtained from
mm. the Navier-Stokes equation where effects from
the Joule heating are supposed to be negligible:
4.1 Ac electroomosis equations
equation to estimate more precisely the EDL ρ m is the mass density of the fluid (1000 kg/m3
features: the EDL width for the electric field
calculation inside the bulk and the free charge for water) and η is its dynamic viscosity (10-3
density for the velocity field. kg/m/s for water). Because Reynolds numbers
The EDL is assumed to be a capacitance per unit are very low in microsystems [12], the inertia
term is generally very low in (20).
area C EDL (C/m2) such that:
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble
r
FE is the time-averaged electrical force due to
the interaction of the ac electric field with the
free charges of the EDL. Under the assumption (18)
that the fluid permittivity is uniform (which is
not generally the case for high voltages [7]): AC complex electrokinetic
equation (17)
r 1 r
FE = q Re( E * ) (21) (19) using LEDL
2
where q is the free charge density inside the EDL
defined by (3) and computed from the source
term of the MPB equation (14). As the MPB +ψE -ψE
equation involves only the diffuse layer of the half electrodes
EDL, the boundary above the electrodes for
equation (20) represents the interface between LEDL q+
the compact layer and the diffuse layer: the MPB equation (14)
boundary condition of type ‘slip/symmetry’, +ψE -ψE
which is equivalent to the nonpermeability
q-
condition, is used at this interface (see figure 12):
rr
v.n = 0 (22)