JID: AMC

ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

Applied Mathematics and Computation 0 0 0 (2017) 1–11

Contents lists available at ScienceDirect

Applied Mathematics and Computation

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

Simulation of chloride migration in reinforced concrete

J. Němeček∗, J. Kruis, T. Koudelka, T. Krejčí
Faculty of Civil Engineering, Department of Mechanics, Czech Technical University in Prague,Thákurova 7, Prague 6 166 29, Czech
Republic

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

Article history: In a large extent, durability and service life of reinforced concrete structures is influenced
Available online xxx by penetration of deicing salts or seawater salts to the steel reinforcement. The steel corro-
sion is induced by a raising chloride concentration and possibly can lead to concrete cover
Keywords:
spalling due to volumetric changes of oxides formed at the steel surface. Therefore, a crit-
Concrete
ical concentration of chlorides is monitored and investigated by numerical simulations in
Reinforcement
Chloride extraction service life assessments.
Diffusion This contribution is devoted to modeling and numerical solution of the diffusion–
Convection convection problem applied to chloride migration in concrete. The theoretical background
Electromigration is reviewed with some of the numerical consequences and strategies that need to be used
for a successful problem solution. The solution is implemented into an in-house open-
source FEM software accounting for non-constant electric field and chloride binding exist-
ing in a real reinforced concrete element. The three stage approach is applied and demon-
strated on an example of a reinforced concrete beam. First, an increased concentration of
chlorides in the cover layer of the beam is simulated with pure diffusion over a long time
period (10 years). Second, the two-dimensional distribution of an externally applied elec-
tric field is calculated in the beam. Third, an electrochemical extraction process in which
the migration of ions is accelerated by an application of the electric field is simulated. The
short time (2 days) treatment leads to rapid decrease of chloride concentration in the con-
crete cover but leaves elevated concentrations in deeper areas of the beam cross section as
a direct consequence of the real action of two-dimensional and non-constant electric field.
© 2017 Elsevier Inc. All rights reserved.

1. Introduction

The durability of reinforced concrete structures is very much related to the transport of water and ionic species in a
porous microstructure of concrete. Major factor influencing the service life of a structure is the state of the load-bearing
steel reinforcement and its possible corrosion. Depending on exposure conditions, concrete type, geometrical configuration
and other structural and environmental parameters, the corrosion is initiated usually after several tens of years in normal
conditions but can be significantly shorten (to years) if the concrete cover is not sufficient, carbonated or salt exposure is
high as in the case of e.g. sea shore structures. Due to volumetric expansion of corrosion products, the corrosion process
ultimatively leads to spalling of the concrete cover and further ingress of harmful species to the reinforcement. One of
the most common and most deteriorating mechanisms causing steel damage is chloride induced corrosion. Critical chloride

∗
Corresponding author.
E-mail addresses: jiri.nemecek@fsv.cvut.cz (J. Němeček), jaroslav.kruis@fsv.cvut.cz (J. Kruis), tomas.koudelka@fsv.cvut.cz (T. Koudelka), tomas.krejci@
fsv.cvut.cz (T. Krejčí).

http://dx.doi.org/10.1016/j.amc.2017.07.029
0 096-30 03/© 2017 Elsevier Inc. All rights reserved.

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

2 J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11

concentration level at which a considerable corrosion damage starts is about 0.2 − 0.4% Cl− per cement weight and it was
studied by many researchers in the past (e.g. [1,2]).
Concrete, as a heterogeneous composite, is in a real situation characterized with fully or partially saturated pores. Water
in pores allows ionic species (chlorides and other ions) to be dissolved and transported in the concrete due to different
mechanisms. One of them is the ionic diffusion. It presents a relatively slow process governed by water saturation level (e.g.
[3,4]), concentration gradient (e.g. [5]), porosity and pore tortuosity of the material (e.g. [6]). Convection of ions appears
in the case of water flow (which is more pronounced in cracked concrete) or in the case of application of an external
electric field which governs the flux of charged ions (e.g. [7,8]). An external electric field is used for performing accelerated
penetration tests on concrete specimens as well as for repairing a real structure [9,10]. In the later case the method is called
electrochemical chloride extraction (ECE) and has been successfully used for mitigation of chloride attack on e.g. concrete
bridges [11].
Although the topic of electromigration and ECE has been widely covered in the literature from the experimental point
of view [12–16] as well as from theoretical point of view [7,8,17,18], contributions respecting the real rebar configuration in
a reinforced concrete structure and the real electric potential gradient distribution are rather rare. The potential gradient is
widely assumed to be constant [7] or the variation of the potential gradient is derived only from the multi-ionic species
transport [8] or a simple domain geometry is solved to simulate experimental results [18]. In contrast, application of the
ECE treatment to the reinforced concrete element leads to an uneven distribution of chloride concentration which is caused
by a non-constant electric potential gradient [19].
Therefore, the purpose of this paper is twofold. First, the theory for chloride migration in concrete is presented and
implemented in an open-source finite element code and typical situations such as chloride penetration and extraction on a
reinforced concrete beam with real reinforcement arrangement are calculated using experimental characteristics obtained in
the previous research [20,21]. The uneven distribution of an electric potential around the rebars is calculated based on the
Gauss law of electrostatics and used for extraction of chlorides in a numerical example.

2. Diffusion and convection of ions in concrete

2.1. Ionic transport in saturated environment

In this paper, transport of ions in saturated environment is assumed to be governed by three main driving forces. First,
the classical diffusion based on Fick’s law leads to the ion flux density in the form

j d = −D∇c (1)
where D is the diffusion coefficient (m2 /s) and ϱc is the partial density of chlorides (kg/m3 ). Generally, in the flux density
(1) the diffusion coefficient should be replaced by a second order tensor but with respect to the assumption of isotropy, a
single coefficient is sufficient. In reality, the D is a function of different variables such as concrete age, porosity, degree of
hydration, aggregate size, temperature, humidity and local chloride concentration (e.g. [6,22]).
It is more common to use concentration instead of the partial density of chlorides. The concentration, c, is the ratio of
the weight of chlorides in an unit volume and total weight of the unit volume. Therefore, it has the form
c
c= (2)

where ϱ is the total density of concrete (kg/m3 ). With assumption of constant total density, the diffusion flux density can
be rewritten into the form

j d = −D∇ c . (3)
The second part of the ion flux density is driven by the intensity of applied electric field. The flux density is expressed
by the Nernst–Planck equation
DF zc DF zc DF zc
je = E=− ∇φ = − ∇φ (4)
RT RT RT
where D is the diffusion coefficient (m2 /s), F = 96, 487 C/mol is the Faraday constant, z is the valence of ions, ϱc is the partial
density of ions (kg/m3 ) R = 8.314 J/K/mol is the molar gas constant, T is the temperature (K), E is the intensity of electric
field (V/m), φ is the electric potential (V). The diffusion coefficient, D, is sometimes called the migration coefficient. More
details can be found in [23].
Finally, ion convection can be caused by the net water flow. The ion flux density has the form

j c = c v =  c v . (5)
Another possible driving force, the ion chemical activity, is sometimes assumed in the literature [7,24] and the flux den-
sity has the form

j chem = −Dc∇ (lnγ ) . (6)

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11 3

In (6), γ is the chemical activity coefficient. However, the chemical activity of ions can be considered negligible compared
to the applied electric field and is not taken into account in this paper.
The total flux density of ions has the form
 DF zc

j = j d + j e + j c =  −D∇ c − ∇ φ + cv . (7)
RT
The mass balance equation for ions has the form
∂ c
= −div j + s (8)
∂t
where s is the source of ions which is assumed to be zero in this paper. Substitution of (7) and (2) to the mass balance
Eq. (8) results, after cancelation of the total density, in the form
∂c  DF zc

= div D∇ c + ∇ φ − cv . (9)
∂t RT

2.2. Chloride binding

A chemical interaction of chloride ions with pore walls exists in concrete. Chloride ions may be present either in the
pore solution (free chlorides), chemically bound to the cement hydration products, or physically held to the surface of the
hydration products (i.e. chemically and physically bound chlorides) [25]. Chemical binding mainly refers to the reaction be-
tween chlorides and unhydrated cement components of C3 A and C4 AF to form a Friedel’s salt [26,27]. Chlorides can also be
physically bound to the main hydration products, i.e. C-S-H gels in cement [28]. Mechanisms of chloride binding in cement
are very complicated and influenced by many factors like cement type, salt type, chloride concentration, temperature, sup-
plementary materials in concrete, carbonation level and is still under examination by many authors [25,26,29,30]. Transport
models usually decompose the total chloride concentration, c, into free cf and bound cb concentrations (e.g. [6,8,31]) as

c = c f + cb . (10)
Chloride binding is often described at thermodynamic equilibrium by chloride binding isotherms which are known as the
relationships between free and bound chloride ions over a range of chloride concentrations at a given temperature. Four dif-
ferent types of binding isotherms are used in the literature, namely linear, Langmuir, Freundlich and BET binding isotherms.
The mostly used are probably the Freundlich and Langmuir isotherms [6,8,31,32]. None of the isotherms describes the re-
lation in the full range of concentrations. For smaller ion concentrations Langmuir isotherm seems to fit well experiments
whereas for free chloride ion concentrations larger than 0.01 mol/l Freundlich isotherm is more appropriate [28]. The Fre-
undlich isotherm is defined as
β
cb = α c f (11)

whereas Langmuir isotherm can be expressed as

αˆ c f
cb = (12)
1 + βˆ c f

in which α , β , αˆ and βˆ are binding constants. The rate of the total chloride concentration can be described as
 
∂ c ∂ ( c f + cb ) ∂ c f d cb ∂ c f dc ∂cf
= = + = 1+ b (13)
∂t ∂t ∂t dc f ∂ t dc f ∂ t

dcb
where the ratio is called the binding capacity and is determined from the chloride isotherm. For the Freundlich
dc f
isotherm, the ratio is given by
d cb β −1
= αβ c f (14)
dc f
and for Langmuir isotherm by
d cb αˆ
= . (15)
dc f (1 + βˆ c f )2
Besides the aforementioned effects, electrical field also affects chloride binding as reported by [25,27,33]. Castellote et al.
[33] found that no binding occurs during accelerated migration experiments as long as the total chloride concentration
remains below 0.14%/concrete. Above this level the amount of bound chlorides is still less than in a natural diffusion ex-
periment depending on the experiment duration. The situation is further altered in concretes containing supplementary
materials or in carbonated concrete [26]. The difference between the free chloride concentration in migration and diffusion

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

4 J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11

tests results mainly from the different test durations. In diffusion experiments, the time is long enough to allow establishing
of a chemical equilibrium whereas in short migration tests the equilibrium need not to be achieved.
Taking chloride binding (Eq. (13)) into account, the mass balance equation Eq. (9) must be modified as
   
∂cf d cb DF zc f
1+ = div D∇ c f + ∇φ − cf v . (16)
∂t dc f RT

2.3. Modeling of an electro-static field

In this paper, we focus our attention to modeling of an electric field. As mentioned before, electric field is applied in
order to speed up extraction of ions from concrete in repair applications because pure diffusion would be very slow. It
means, the electric field is given and can be effectively assumed to be constant during the extraction procedure. Therefore,
the last two terms in the balance Eq. (16) could be aggregated into a single term
DF zc f
c f vˆ = ∇φ − c f v . (17)
RT
Special attention must be devoted to the gradient of the electric potential, φ . In many papers on the ion penetration or
extraction, the gradient of the electric potential is assumed to be constant which is generally not true. The electric potential
has to satisfy the Gauss law of electrostatics in the form
εr ε0 divE = εr ε0 φ = ρ (18)
where ρ is the density of charge εr is the relative permitivity (-) and ε0 =
(C/m3 ), 8.854 · 10−12
F/m is the permitivity of the
vacuum. Eq. (18) has to be solved for appropriate domain and boundary conditions first. The electric potential obtained is
then used in flux density (17).
With the help of (17) one has the mass balance Eq. (16) in the form of
∂c
= div(D∇ c f ) + div(c f vˆ ). (19)
∂t
The term div(D∇ cf ) represents the diffusion and the term div(c f vˆ ) represents the convection. The convection term can
be further modified into the form
∂c T
= div(D∇ c f ) + vˆ ∇ c f (20)
∂t
because the mass balance equation for water is described by divv = 0 and the charge density, ρ , in (18) is zero. The mass
balance Eq. (20) is defined on a domain  with boundary  which is split into two disjoint parts  D and  N . The concen-
tration is prescribed on  D (Dirichlet boundary condition) while ion flux density is prescribed on the part  N (Neumann
boundary condition). The problem is solved in time interval (0, tmax ). Because of the time dependent problem, initial condi-
tion has to be defined. The boundary and initial conditions can be written in the form
∀x ∈ D , t ∈ (0, tmax ) : c(x, t ) = cD (x, t ) (21)

∀x ∈ N , t ∈ (0, tmax ) : nT j (x, t ) = jN (x, t ) (22)

∀ x ∈  : c ( x, 0 ) = c 0 ( x ) (23)
where cD is the prescribed concentration, n is the unit normal vector to  N , jN is the prescribed ion flux density in the
normal direction to the boundary  N and c0 is the initial concentration.

3. Numerical solutions

Traditionally, the mass balance Eq. (20) for ion concentrations was solved by the finite difference method. We recall
this method in the following Section 3.1 with emphasis to appropriate finite differences that need to be used for stable
calculations. Then, we focus the attention to the finite element solution (FEM) in Section 3.2.

3.1. Finite difference method

It is known that not all finite differences can be used in connection with the convection–diffusion problem. In the case
j
of one space and one time dimension, the difference lengths are x and t and ci denotes the concentration in the point
j
x = i x at time t = j t, i.e. ci = c (i x, j t ).
For positive ion flux, the first space derivative has to be assumed backward in the form

∂ cij cij − cij−1

≈ (24)
∂x x
Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11 5

Fig. 1. Simulation of electromigration in a one-dimensional concrete element with positive ion flux using different finite difference schemes. Note, that
values for forward scheme go out of scale.

and the second derivative in the central form as

∂ 2 cij cij−1 − 2cij + cij+1

≈ . (25)
∂ x2 ( x )2
By substitution of the finite differences into the mass balance equation

∂c ∂ 2c ∂c
=D 2 +v (26)
∂t ∂x ∂x
with Dirichlet boundary conditions c (0, t ) = cl (t ) and c (L, t ) = cr (t ) (L is the length of one dimensional domain, cl and cr are
concentrations on the left and right end of the domain, respectively), the following system of ordinary differential equations
is obtained

Ac (t ) = c˙ (t ) + f (t ) (27)
where c(t) is the vector of ion concentrations in grid points, the vector f(t) includes prescribed boundary concentrations and
matrix A has the form
⎛ ⎞
−2D v D
+
⎜ ( x )2 x ( x )2 ⎟
⎜ D v −2D v D ⎟
⎜ − + ⎟
⎜ ( x )2 x ( x )2 x ( x )2 ⎟
⎜ v −2D v ⎟
A=⎜ ⎟.
D D
⎜ − + ⎟ (28)
⎜ ( x )2 x ( x )2 x ( x )2 ⎟
⎜ .. .. ⎟
⎜ . . ⎟
⎝ D v −2D
⎠
− + vx
( x )2 x ( x )2
The vectors in Eq. (26) have the form
⎛ ⎞ ⎛ −D v  ⎞
⎛ ⎞ c˙ 1 (t ) + cl (t )
c1 (t ) ⎜ ( x ) 2 x ⎟
⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ c˙ (t ) ⎟ ⎜ ⎟
⎜ c (t ) ⎟ ⎜ 2 ⎟ ⎜ 0
⎟
⎜ 2 ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ c˙ 3 (t ) ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟
c (t ) = ⎜ c3 (t ) ⎟. c˙ (t ) = ⎜ ⎟ ⎜
f (t ) = ⎜ ⎟
⎟ (29)
⎜ ⎟ ⎜ . ⎟ ⎜ ⎟
⎜ ⎟ ⎜ . ⎟ ⎜ .. ⎟
⎜ .. ⎟ ⎜ . ⎟ ⎜ . ⎟
⎜ . ⎟ ⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎜ ⎟ ⎜ ⎟
⎝c˙ (t )⎠ ⎝ −D ⎠
cn−1 (t ) n−1
c r (t )
( x )2

The system of ordinary differential Eq. (27) has to be further discretized in time. The backward difference scheme leads
to the form
1 
Ac (t j+1 ) = c (t j+1 ) − c (t j ) + f (t j+1 ) (30)
t

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

6 J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11

and the final system of algebraic equations has the form

(I − t A )c (t j+1 ) = c (t j ) − t f (t j+1 ) (31)
where I is the identity matrix.
The system (31) is nonsymmetric and its solution is connected with numerical difficulties. The finite difference scheme
is only conditionally stable for the diffusion–convection problem and not all finite difference schemes can be used [34]. For
the pure diffusion and for cases with the Peclet number less than 2 the problem is diffusion dominated and the central
difference instead of the first derivative is applicable. For convection or diffusion–convection problem either forward or
backward difference must be used instead of the first derivative based on the flow direction (so called upwind schemes,
[20,34]). Otherwise, spurious oscillations appear in the solution as illustrated in Fig. 1 which shows concentration profile in
a one-dimensional concrete element exposed to electrical migration of chlorides with positive ion flux. It can be seen that
only the backward scheme gives stable results compared to central scheme (small oscillations) and forward scheme (large
oscillations). In the case of negative ion flux the situation is opposite, i.e. the forward scheme must be used [20].

3.2. Finite element modeling

Solution of the convection–diffusion Eq. (20) based on the finite difference method suffers from difficulties in the case
of complicated shape of the domain solved. In such case, it is uneasy to define suitable finite differences. Therefore, the
Eq. (20) will be solved by the finite element method based on the Galerkin–Petrov approach [35,36]. Another reason for
using the finite element method is that the analysis of chloride distribution can be coupled with mechanical analysis which
is usually based on finite elements.
In FEM, unknown concentration is approximated in the form
c (x, t ) = N c (x )c (t ) (32)
where Nc is the matrix of approximation functions and c is the vector of nodal values of the concentration. The test function
has the form
γ = Nγ γ (33)
where Nγ is the matrix of test functions and γ is the vector of nodal values of the test functions. In contrary to the classical
Galerkin method, here the matrices Nc and Nγ contain different functions which will result in nonsymmetric matrices [34].
The approximation functions stored in the matrix Nc are classical linear, bi-linear or tri-linear functions. The test func-
tions in one-dimensional case are in the form
h dNi
ti (x ) = Ni (x ) + α sign(v ) (34)
2 dx
where h is the element length and α is an auxiliary parameter. More details can be found in [34]. In more dimensions, the
test functions have the form

h  ∂ Ni
d
vj
ti (x ) = Ni (x ) + α (35)
2 ∂xj
j=1
 v
where h is the characteristic size of the element, d is the dimension of the problem. More details can be found in [34].
Substitution of (32) to the mass balance Eq. (20) and multiplication by test functions (33) and integration leads to the
form
  
dc
γ T N Tγ N c d = γ T N Tγ DnT Bc cd − γ T BTγ DBc cd +
 dt  

+ γ T N Tγ vˆ T Bc cd. (36)

In the previous equation, gradients of approximation and test functions are assembled in the matrices Bc and Bγ . The
following matrices and a vector are defined as

C = N Tγ N c d (37)



K = BTγ DBc d (38)



T
H = N Tγ vˆ Bc d (39)



f = N Tγ jN d (40)
N

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11 7

Fig. 2. Cross section of the beam with total area of 24 × 40 cm and location of reinforcement bars (white squares).

where C is the capacity matrix, K is the diffusivity matrix, H is the conductivity matrix and the ion flux density normal to
the boundary  N has the form DnT Bc c = jN with respect to (22). With help of the matrices, the Eq. (36) takes the form
dc
C + ( K − H )c = f . (41)
dt
The numerical approaches presented here were implemented in an in-house open-source FEM software SIFEL [37].

4. Numerical examples

As mentioned above, many authors dealing with the electrically accelerated ion penetration or extraction assume con-
stant gradient of the electric potential which is generally not true, especially in two- or three-dimensional cases of ion
transport in reinforced concrete structural elements. Therefore, the numerical approach of chloride transport introduced in
this paper will be demonstrated by an example of a long-term ion penetration on a typical reinforced concrete beam. Subse-
quently, ion extraction driven by an electric field will be shown on the same structural member. One of the main advantages
of this approach stems from correct distribution of the electric potential which is determined by the Gauss law of electro-
statics. There are three main steps performed. First, the ion penetration is described by pure diffusion. In the second step,
the Gauss law is used to determine the electric potential. Finally, the electro-extraction is described by convection–diffusion
equation. The model takes into account chloride binding using Freundlich isotherm in the mass balance equation (Eq. (16)).
For comparison, the problem of no binding (Eq. (9)) is also solved and corresponding chloride concentrations calculated.
In the example, rectangular cross section of a reinforced concrete beam is assumed with dimensions 24 × 40 cm. The
reinforcement is represented by four bars with 20 mm in diameter. The beam geometry is depicted in Fig. 2.
As mentioned, penetration of ions into concrete is described by pure diffusion at first. Initial chloride concentration is
assumed to be zero (i.e. c0 = 0 in Eq. (23)). Dirichlet boundary conditions (Eq. (21)) with concentration of cD = 0.008 kg/kg
are prescribed on the left edge of the cross section. The exposition time is assumed to be 10 years and the diffusion coef-
ficient as D = 7.4 · 10−12 m2 /s. The values of surface concentrations and the diffusion coefficient are taken as typical normal
strength concrete characteristics inspired by real experimental studies [20,21]. The simplified assumption of a constant dif-
fusion coefficient is taken here since the calculation serves for the assessment of an initial chloride concentration profile in
the numerical example. For the real concentration profiles, one should use a time-dependent diffusion coefficient respecting
different variables, e.g. [6,22]. The choice of the diffusion coefficient, however, only influences the absolute values of the
concentrations and not their distribution.
During the diffusion phase, chloride binding was taken into account assuming Freundlich isotherm, Eqs. (11) and (14).
The binding constants corresponding to normal strength Portland cement concrete were taken from [25,38] and recomputed
for concentration units (kg/kg) as α = 0.013 and β = 0.36.
The final chloride concentration distributions after 10 years with and without considering binding are shown in Fig. 3a
and 3b. It can be seen that the penetration front of chlorides is uneven and reaches approximately the 40–60 mm depth
region, i.e. deeper than the rebar position. The difference is more pronounced in larger depths where the concentration are,
however, smaller. At rebar position, the difference is not as large for the 10 years period and given material parameters but
the trend is obvious from Fig. 3c where the penetration profile through the beam is shown. Also, the amount of free/bound
chlorides with respect to total chlorides is depicted in Fig. 3d which shows the proportions between the individual concen-
trations.
In order to describe extraction of ions from the concrete beam with the help of electric field, distribution of the electric
potential is needed. For such purpose, larger domain than the cross section itself need to be used in the model. Therefore,
the domain size was set to 50 × 44 cm. The situation is depicted in Fig. 4. The electric potential is applied between the in-
ner steel reinforcement and the outer concrete surface which is in the ECE method covered by a conductive (electrolyte

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

8 J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11

Fig. 3. Final chloride concentrations after 10 years of diffusion considering (a) no binding and (b) Freundlich binding, (c) free chloride concentration profiles
in 1, 5 and 10 years (n.b.= no binding; F.b.= Freundlich binding) and (d) concentration profiles of cf , cb and c in 10 years.

Fig. 4. FE-mesh of the reinforced concrete beam and its vicinity (grey=air, blue=concrete, white=steel). (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.).

saturated) layer attached to the surface. In the simulation, the situation is modeled by boundary conditions prescribed to
the respective finite elements (φ = 20 V at the reinforcement and φ = 0 V at the surface). The relative permeability of the
concrete is εr,c = 4.5 and the relative permeability of the air is εr,a = 1. Distribution of the electric potential is shown in
Fig. 5. Finally, ion extraction caused by the electric field is described by the convection–diffusion equation. The distribution
of ion concentration after 10 years of penetration considering Freundlich binding (Fig. 3b) is assumed as the initial condition
(c0 ) for the extraction. The gradient of electric potential is determined with respect to its distribution (Fig. 5) and appropriate
values are read for every finite element. The convection–diffusion equation is complemented with the Dirichlet boundary
conditions (zero concentration cD = 0 are prescribed at the whole perimeter of the beam.) The duration of the extraction
procedure is prescribed to be 48 h. After two days of the treatment, the chloride concentration is depicted in Fig. 6. The
problem was solved with and without considering the binding effect during the short (2 days) extraction period. Negligible
differences have been found and thus only one results (without binding) are presented in Fig. 6. Evolution of the ion con-
centration in a point situated in between two reinforcement bars is shown in Fig. 7. It can be noted that the concentration

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11 9

Fig. 5. (a) Distribution of the electric potential in the cross section and its vicinity (red=20 V, dark blue=0 V). (b) Detail of the electric potential gradient
around rebars (the domain corresponds to the grey rectangle in the left figure). (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

Fig. 6. Chloride ion concentration after two days of extraction.

0.008
Concentration cf (kg/kg)

0.006

0.004

0.002

0.000
0 5 10 15 20
Time (h)

Fig. 7. Evolution of the free ion concentration during extraction in a point situated between two rebars.

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

10 J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11

slightly increases in the short period after application of the electric field. It is a consequence of the fact that local point
concentration increases due to the transport of chlorides originally located in deeper parts of the beam. But shortly, the
concentration decreases all the way down during extraction and is diminished to a negligible value.
It can be seen in Figs. 3–7, the extraction process is very efficient in the 0–40 mm concrete layer (i.e. between the surface
and steel rebars) in which the concentration is lowered almost to zero. On the other hand, slightly decreased but still high
concentration remains behind the steel rebars. This is a consequence of a very small or almost zero potential gradient at
this area (Fig. 5). It can be concluded that the extraction process can effectively decrease the chloride concentration provided
the initial penetration front remains within the concrete cover.

5. Conclusions

The paper shows successful implementation and solution of the diffusion–convection problem applied to chloride mi-
gration in reinforced concrete based on the Nernst–Planck equation and Gauss law of electrostatics. A numerical example
showing the effective accelerated extraction of chloride ions by externally applied electric field was solved. The following
conclusions can be drawn.

1. The convection–diffusion problem solved by traditional finite difference schemes suffers from numerical difficulties con-
nected with nonsymmetry and conditional stability. Also, the applicability of the finite difference method is limited to a
simple domain geometry.
2. For complex domains and cases where coupling with mechanical analysis is required, application of the finite element
method based on the Galerkin–Petrov approach is appropriate. Similarly to finite difference method, the system of equa-
tions with nonsymmetric matrices needs to be solved.
3. The chloride diffusion process is relatively slow and chlorides are bound to inner concrete pores. The final concentration
level is thus lowered by a certain amount depending on the chloride binding isotherm parameters. The concentration
decrease due to binding was found the be more pronounced for deeper cross section parts where the concentration are
smaller.
4. Long-term diffusion of ions can lead to increased concentrations also in areas at or behind the steel reinforcement.
5. Chlorides can be effectively removed from concrete by the application of the ECE method, i.e. by the application of an
external electric field. The assumption of a constant gradient of electric potential is, however, not valid in the reinforced
beam and an exact calculation using the Gauss law needs to be used. Uneven distribution of the electric potential and
its gradient is noticeable especially around the steel rebars.
6. Compared to diffusion the extraction process that takes the advantage of an electric field acceleration can decrease the
chloride concentration in a very short time. The extraction is very efficient in the region between the electrodes, i.e.
between the concrete surface and the steel rebars. The concentration can be diminished in this part including the region
in between the rebars.
7. On the other hand, the concentration is only partially lowered in the region behind the steel reinforcement where the
potential gradient is small. Also, uneven distribution of the concentration remains in this region after the treatment. As
a practical consequence of this statement, the ECE method can be effectively used for repairing structures where the
elevated salt concentration did not pass over the reinforcement.

Acknowledgment

Financial support of the Czech Science Foundation (project 16-11879S) is gratefully acknowledged.

References

[1] U. Angst, B. Elsener, C.K. Larsen, ystein Vennesland, Critical chloride content in reinforced concrete a review, Cem. Concr. Res. 39 (12) (2009) 1122–1138.
http://dx.doi.org/10.1016/j.cemconres.20 09.08.0 06.
[2] K.Y. Ann, H.-W. Song, Chloride threshold level for corrosion of steel in concrete, Corros. Sci. 49 (11) (2007) 4113–4133. http://dx.doi.org/10.1016/j.corsci.
20 07.05.0 07.
[3] L. Homan, A.N. Ababneh, Y. Xi, The effect of moisture transport on chloride penetration in concrete, Constr. Build. Mater. 125 (2016) 1189–1195.
http://dx.doi.org/10.1016/j.conbuildmat.2016.08.124.
[4] Z. Pavlík, L. Fiala, J. Maděra, M. Pavlíková, R. Černý, Computational modelling of coupled water and salt transport in porous materials using diffu-
sion–advection model, J. Frankl. Inst. 348 (2011) 1574–1587.
[5] J. Crank, The mathematics of diffusion, Oxford Science Publications, second ed., Oxford University Press, 1975.
[6] Y. Xi, Z. Bažant, Modeling chloride penetration in saturated concrete, J. Mater. Civ. Eng. 11 (1) (1999) 51–57. http://dx.doi.org/10.1061/(ASCE)
0899-1561(1999)11:1(51).
[7] E. Samson, J. Marchand, Numerical solution of the extended Nernst–Planck model, J. Colloid Interface Sci. 215 (1) (1999) 1–8. http://dx.doi.org/10.1006/
jcis.1999.6145.
[8] L. Li, C. Page, Finite element modelling of chloride removal from concrete by an electrochemical method, Corros. Sci. 42 (12) (20 0 0) 2145–2165.
http://dx.doi.org/10.1016/S0010- 938X(00)00044- 5.
[9] R. Polder, 8 - electrochemical techniques for corrosion protection and maintenance, in: H. Bhni (Ed.), Corrosion in Reinforced Concrete Structures,
Woodhead Publishing Series in Civil and Structural Engineering, Woodhead Publishing, 2005, pp. 215–241. http://dx.doi.org/10.1533/9781845690434.
215.
[10] Y. Liu, X. Shi, Finite element modelling of chloride removal from concrete by an electrochemical method, Corros. Rev. 27 (1–2) (2011) 53–82. http:
//dx.doi.org/10.1515/CORRREV.2009.27.1-2.53.

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029
JID: AMC
ARTICLE IN PRESS [m3Gsc;July 29, 2017;1:29]

J. Němeček et al. / Applied Mathematics and Computation 000 (2017) 1–11 11

[11] G. Clemena, D. Jackson, Trial Application of Electrochemical Chloride Extraction on Concrete Bridge Components in Virginia, Technical Report, Final
report FHWA/VTRC 00R18, Charlottesville, 2000.
[12] C. Arya, Q. Sa’id-Shawqi, P. Vassie, Factors influencing electrochemical removal of chloride from concrete, Cem. Concr. Res. 26 (6) (1996) 851–860,
doi:10.1016/0 0 08-8846(96)0 0 067-1.
[13] G. Fajardo, G. Escadeillas, G. Arliguie, Electrochemical chloride extraction (ECE) from steel-reinforced concrete specimens contaminated by ‘artificial’
sea-water, Corros. Sci. 48 (1) (2006) 110–125, doi:10.1016/j.corsci.2004.11.015.
[14] B. Elsener, U. Angst, Mechanism of electrochemical chloride removal, Corros. Sci. 49 (12) (2007) 4504–4522, doi:10.1016/j.corsci.2007.05.019.
[15] M. Sánchez, M. Alonso, Electrochemical chloride removal in reinforced concrete structures: improvement of effectiveness by simultaneous migration
of calcium nitrite, Constr. Build. Mater. 25 (2) (2011) 873–878, doi:10.1016/j.conbuildmat.2010.06.099.
[16] H. Shan, J. Xu, Z. Wang, L. Jiang, N. Xu, Electrochemical chloride removal in reinforced concrete structures: improvement of effectiveness by simulta-
neous migration of silicate ion, Constr. Build. Mater. 127 (2016) 344–352, doi:10.1016/j.conbuildmat.2016.09.137.
[17] Y. Wang, L. Li, C. Page, A two-dimensional model of electrochemical chloride removal from concrete, Comput. Mater. Sci. 20 (2) (2001) 196–212,
doi:10.1016/S0927-0256(0 0)0 0177-4.
[18] A. Toumi, R. Franois, O. Alvarado, Experimental and numerical study of electrochemical chloride removal from brick and concrete specimens, Cem.
Concr. Res. 37 (1) (2007) 54–62, doi:10.1016/j.cemconres.2006.09.012.
[19] W. Yeih, J. Chang, C. Chang, K. Chen, M. Chi, Electrochemical chloride removal for reinforced concrete with steel rebar cage using auxiliary electrodes,
Cem. Concr. Compos. 74 (2016) 136–146, doi:10.1016/j.cemconcomp.2016.08.002.
[20] J. Němeček, Y. Xi, Simulation of chloride extraction tests on concrete specimens (Paper 151), in: J. Kruis, Y. Tsompanakis, B. Topping (Eds.), Proceedings
of the Fifteenth International Conference on Civil, Structural and Environmental Engineering Computing, Civil-Comp Press, Stirlingshire, Scotland, 2015,
pp. 1–12.
[21] P. Hlaváček, J. Němeček, Accelerated chloride migration tests in concrete, in: I. Zolotarev, V. Radolf (Eds.), Engineering Mechanics 2016, Institute of
Thermomechanics, Academy of Sciences of the Czech Republic, Dolejskova 5, Prague 8, 182 00, Czech Republic, 2016, pp. 194–197.
[22] B.H. Oh, S.Y. Jang, Effects of material and environmental parameters on chloride penetration profiles in concrete structures, Cem. Concr. Res. 37 (1)
(2007) 47–53. http://dx.doi.org/10.1016/j.cemconres.2006.09.005.
[23] R. Černý, P. Rovnaníková, Transport Processes in Concrete, Spon Press, London, New York, 2002. ISBN 0-415-24264-9
[24] O. Truc, J.-P. Ollivier, L.-O. Nilsson, Numerical simulation of multi-species transport through saturated concrete during a migration test msdiff code,
Cem. Concr. Res. 30 (10) (20 0 0) 1581–1592. http://dx.doi.org/10.1016/S0 0 08-8846(0 0)0 0305-7.
[25] Q. Yuan, C. Shi, G. De Schutter, K. Audenaert, D. Deng, Chloride binding of cement-based materials subjected to external chloride environment – a
review, Constr. Build. Mater. 23 (1) (2009) 1–13, doi:10.1016/j.conbuildmat.2008.02.004.
[26] M. Saillio, V. Baroghel-Bouny, F. Barberon, Chloride binding in sound and carbonated cementitious materials with various types of binder, Constr. Build.
Mater. 68 (2014) 82–91. http://dx.doi.org/10.1016/j.conbuildmat.2014.05.049.
[27] Q. Yuan, D. Deng, C. Shi, G. De Schutter, Chloride binding isotherm from migration and diffusion tests, J. Wuhan Univ. Technol.-Mater. Sci. Ed. 28 (3)
(2013) 548–556, doi:10.1007/s11595- 013- 0729- y.
[28] L. Tang, L.-O. Nilsson, Chloride binding capacity and binding isotherms of OPC pastes and mortars, Cem. Concr. Res. 23 (2) (1993) 247–253. http:
//dx.doi.org/10.1016/0 0 08-8846(93)90 089-R.
[29] M. Florea, H. Brouwers, Chloride binding related to hydration products: part i: ordinary portland cement, Cem. Concr. Res. 42 (2) (2012) 282–290.
http://dx.doi.org/10.1016/j.cemconres.2011.09.016.
[30] M. Florea, H. Brouwers, Modelling of chloride binding related to hydration products in slag-blended cements, Constr. Build. Mater. 64 (2014) 421–430.
http://dx.doi.org/10.1016/j.conbuildmat.2014.04.038.
[31] N. Damrongwiriyanupap, L. Li, Y. Xi, Coupled diffusion of chloride and other ions in saturated concrete, Front. Archit. Civil Eng. China 5 (3) (2011) 267,
doi:10.1007/s11709-011-0112-z.
[32] V. Kočí, J. Maděra, R. Černý, Computational simulation of salt transport and crystallization in surface layers of building envelopes, in: T. Simos (Ed.),
Proceedings of the Conference on American Institute of Physics, 1479, American Institute of Physics, 2012, pp. 2054–2057.
[33] M. Castellote, C. Andrade, C. Alonso, Chloride-binding isotherms in concrete submitted to non-steady-state migration experiments, Cem. Concr. Res. 29
(11) (1999) 1799–1806. http://dx.doi.org/10.1016/S0 0 08-8846(99)0 0173-8.
[34] O.C. Zienkiewicz, R.L. Taylor, The Finite Element Method for Fluid Dynamics, 3, fifth, Butterworth–Heinemann, Oxford, UK, 20 0 0.
[35] Z. Bittnar, J. Šejnoha, Numerical Methods in Structural Mechanics, ASCE Press, New York, USA, 1996.
[36] T.J. Hughes, The Finite Element Method: Linear Static and Dynamic Finite Element Analysis, Prentice-Hall, Inc., Englewood Cliffs, New Jersey 07632,
1987. ISBN 0-13-317025-X 025
[37] J. Kruis, T. Koudelka, T. Krejčí, Computer code SIFEL 2001-2017, Czech Technical University, Prague, http://mech.fsv.cvut.cz/∼sifel/.
[38] B. Martín-Pérez, H. Zibara, R. Hooton, M. Thomas, A study of the effect of chloride binding on service life predictions, Cem. Concr. Res. 30 (8) (20 0 0)
1215–1223. http://dx.doi.org/10.1016/S0 0 08-8846(0 0)0 0339-2.

Please cite this article as: J. Němeček et al., Simulation of chloride migration in reinforced concrete, Applied Mathematics
and Computation (2017), http://dx.doi.org/10.1016/j.amc.2017.07.029