Corrosion Science 51 (2009) 415–425

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

The effect of temperature on the corrosion of steel in concrete.

Part 1: Simulated polarization resistance tests and model development
M. Pour-Ghaz, O. Burkan Isgor *, P. Ghods
Carleton University, Department of Civil and Environmental Engineering, 1125 Colonel by Drive, Ottawa, Ontario, Canada K1S 5B6

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

Article history: The effect of temperature on the corrosion rate of steel corrosion in concrete is investigated through sim-
Received 25 October 2008 ulated polarization resistance experiments. The simulated experiments are based on the numerical solu-
Accepted 31 October 2008 tion of the Laplace’s equation with predefined boundary conditions of the problem and have been
Available online 7 November 2008
designed to establish independent correlations among corrosion rate, temperature, kinetic parameters,
concrete resistivity and limiting current density for a wide range of possible anode/cathode (A/C) distri-
Keywords: butions on the reinforcement. The results, which successfully capture the resistance and diffusion control
A. Steel reinforced concrete
mechanisms of corrosion as well as the effect of temperature on the kinetic parameters and concrete/pore
B. Modelling studies
B. Polarization
solution properties, have been used to develop a closed-form regression model for the prediction of the
C. Kinetic parameters corrosion rate of steel in concrete.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction in the surrounding environment, the control and monitoring of the

The effect of temperature on steel corrosion in concrete can be
quite complex. This complexity is partly due to the influence of
temperature on the kinetic parameters of corrosion (e.g. Tafel
slopes, exchange current densities and equilibrium potentials)
and partly due to associated changes in the properties of con-
crete/pore solution. Existing theoretical models that have been
developed for predicting the rate of steel corrosion in concrete,
including the one developed by the authors of this paper [1], do
not consider the explicit effect of temperature in their formula-
tions. These theoretical models generally consider the effect of
temperature on the corrosion of steel in concrete through concrete
resistivity. Although this indirect inclusion of temperature in the
modelling exercises can be justified for many cases since concrete
resistivity is known to be a significant factor for the corrosion of
steel in concrete, it does not cover all effects associated with it.
To address this issue, a number of researchers have developed
empirical models [2–4]; however, the applicability of these empir-
ical models is usually limited by the prescribed conditions of the
experiments that they were developed from. To expand the range
of applicability of the empirical models, they need to be based on
significantly large number of experiments that are well controlled
and free from instrumental and human errors.
The precise control of the test setup and environmental condi-
tions is a major challenge in corrosion studies, especially in the
investigation of steel corrosion in concrete. Since the complex ma-
trix of concrete can undergo major changes due to small variations
* Corresponding author. Tel.: +1 613 520 2600x2984; fax: +1 613 520 3951.
E-mail address: burkan_isgor@carleton.ca (O.B. Isgor).
test variables and environmental conditions can become hard, if
not impossible. In addition, due to the inherent difficulties, inaccu-
racies and uncertainties associated with the corrosion measure-
ment techniques in a material like concrete [4–7], the results of
experimental studies are generally subject to debate. Furthermore,
these experiments can be expensive and time consuming when a
subject such as corrosion of steel in concrete is concerned.
Numerical experiments, which are developed for simulating
physical experiments using robust models that are based on solid
theoretical foundations, can help significantly overcome associated
difficulties described above. A comprehensive experimental pro-
gram that may take years and cost large sums to carry out can
be completed in a relatively short period of time for only the cost
of computation. Furthermore, when a certain experiment requires
precisely controlled conditions, well-designed simulated experi-
ments can become the choice over physical tests since the control
of most conditions (e.g. environmental conditions, test setup, etc.)
can be challenging in the latter.
A preliminary study by authors have indicated that a compre-
hensive experimental study to investigate the effects of all neces-
sary parameters on the corrosion of steel in concrete and to
develop a comprehensive closed-form regression model simulating
the existing corrosion rate measurement techniques (such as the
ones that are based on polarization resistance method) is not fea-
sible within a realistic time and cost frame. Due to the large num-
ber of variables affecting the phenomenon, it is estimated that
hundreds of tests need to be carried out to obtain data that will
provide a closed-form regression model that is statistically reliable.
Considering the fact that there are different commercial instru-
ments that are based on the same theoretical basis but measure

0010-938X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2008.10.034
416 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425

corrosion rates with relatively large differences [4,7], to include the methods. The simulated experiments are based on the solution of
uncertainty associated with the device used, most of the test need the Laplace’s equation for determining the electric potential distri-
to be repeated using different instruments. bution and current densities in the domain of analysis. In this ap-
The main objective of this paper is to theoretically investigate proach, the electrochemical reactions occur at the anodic and
the effect of temperature on the corrosion rate of steel in concrete cathodic surfaces of the steel reinforcement represented by the
using simulated polarization resistance experiments. The develop- oxidation of iron and the reduction of oxygen, respectively. Poten-
ment of a comprehensive and practical mathematical model, tial distribution in the domain is governed by the Laplace’s equa-
which is based on the results of these numerical tests, for the pre- tion [8]:
diction of the rate of steel corrosion in concrete is also presented.
The experimental verification and parametric investigation of the
r2 / ¼ 0 ð1Þ
developed model are provided in a separate paper (Part 2: Model where / is the electric potential. The Dirichlet boundary conditions
verification and parametric study). at anodic (/a) and cathodic (/c) sites are prescribed by the polariza-
tion equations via Eqs. (2) and (3), respectively [9]:
2. Simulated polarization resistance experiments
ia
/a ¼ /oFe þ ba log ð2Þ
ioa
Because of the considerations presented in the introduction, in
ic 2:303RT iL
the present work, simulated experiments are chosen over a /c ¼ /oO2 þ bc log  log ð3Þ
comprehensive laboratory study to investigate the effect of ioc zc F iL  ic
temperature on the corrosion of steel in concrete and to develop where /oFe and /oO2 (volts) are anodic and cathodic equilibrium
a closed-form model simulating the existing corrosion rate mea- potentials respectively, ba and bc , respectively, are anodic and
surement techniques that are based on polarization resistance cathodic Tafel slopes, ia and ic (A/m2) are anodic and cathodic cur-

a Guard ring Reference

Controls electrode

Auxiliary
electrode Guard ring
Guard ring

Rebar
Polarized length of
rebar

b ∂f
=0
∂n

Anodic Cathodic
φ = φa φ = φc
Fig. 1. (a) Schematic illustration of polarization resistance test setup; (b) schematic illustration of domain and its boundary conditions.
 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
rent densities, respectively, ioa (A/m2) is the exchange current den-
sity of the anodic reaction, ioc (A/m2) is the exchange current den-
sity of the cathodic reaction, iL (A/m2) is the limiting current
density, R (8.314 J/(mole K)) is the universal gas constant, F
(96,500 C/mole) is the Faraday’s constant, T (°K) is temperature,
and zc is number of electrons that are involved in cathodic reaction.
As it can be observed from these equations, anodes are assumed to
polarize through activation polarization, and cathodic boundary
condition consists of both activation and concentration polarization
components. In concrete, the inclusion of the concentration polari-
zation in the cathodic boundary condition is essential if the effect of
oxygen diffusion through concrete cover is investigated. These
boundary conditions necessitate a non-linear solution of Eq. (1),
for which a customized finite element program was previously
developed and verified with experimental data [10]. Once the po-
tential distribution is obtained by solving Eq. (1), the corrosion cur-
rent density at any point on the surface of the rebar can be obtained
by locally applying the Ohm’s law [8]:
1 @/
i¼ ð4Þ
q @n
where i (A/m2) is the current density, q (X m) is the concrete/pore
solution resistivity and n is the direction normal to the bar surface.
The domain of the problem is chosen to simulate the common
polarization resistance tests which use an external electrode (e.g.
a guard ring) covering a representative area of the concrete surface
and measure the corrosion rate as an average current density over
the critical length of the reinforcement underneath, as illustrated
in Fig. 1a. The polarized critical length of the rebar in the current
investigation is conservatively taken as 300 mm. The domain of
the problem is discretized using 5  5 mm rectangular elements,
which were found to be the optimum size and shape after a com-
prehensive sensitivity analysis [11]. It is assumed that the steel un-
der the coverage area is depassivated such that it can be
represented by a macrocell consisting of an anode and a cathode
that can be characterized by an anode-to-cathode ratio (A/C)
(Fig. 1b). It is acknowledged by the authors that one can produce
an argument against this assumption such that anodes and cath-
odes on the steel surface can form at many locations at the micro-
scopic level; hence it may not be possible to define an A/C ratio
[12]. Although this claim has its merits, especially for microcell
corrosion, it can still be assumed to be a realistic approximation
when simulating polarization resistance tests, which basically
measure the average corrosion rate along the entire polarized
length of the steel reinforcement within a limited area. Therefore,
for the numerical simulation of the polarization resistance setup, a
macro-level A/C ratio can still be defined, and this approach is
adopted in this paper as well. The validity of this assumption is dis-
cussed in the model verification section of this investigation, which
is presented in ‘‘Part 2: Model verification and parametric study.” It
should also be noted that since A/C ratio is not a known quantity,
each numerical experiment needs to be carried out for a wide
range of A/C values, and the rate that produces the highest corro-
sion current is considered to be the state at which the total electro-
chemical energy of the system is minimized. Further explanation
of this concept will be provided in the results section of this paper.
Using the model that is briefly described above 16,000 polariza-
tion resistance tests were simulated by independently varying
kinetic parameters (by varying temperature), concrete cover
thickness, concrete resistivity and limiting current density for 10
A/C ratios that vary between 0.1 and 1.0. It should be noted that,
in reality, the temperature changes simultaneously affect the
kinetic (i.e. Tafel slopes, exchange current densities, equilibrium
potentials) and concrete/pore solution (i.e. resistivity and limiting
current density) properties. However, varying all parameters
417
Table 1
Experimental variables and their values used in virtual experiments.
Experimental Variable Units Values
Temperaturea (T) K 283, 298, 303, 313, 323
Cover thickness (d) mm 20, 40, 60, 100, 140
Limiting current A/m2 0.002, 0.005, 0.008, 0.02, 0.05, 0.08, 0.2, 0.6
density (iL)
Resistivity (q) Xm 50, 350, 650, 1250, 2500, 5000, 7500, 10,000
A/C – 0.1, 0.18, 0.25, 0.33, 0.43, 0.54, 0.67, 0.76, 0.93, 1
a
Temperature variations directly affect the kinetic parameters only.
simultaneously would not be the most efficient use of the power
of numerical experiments. Although only a fraction of the 16,000
independent cases tested here would have been analyzed if the
changes were considered simultaneously, for the benefit of obtain-
ing more information, a computationally more demanding proce-
dure is selected in this study. In this method, the effect of
temperature is assumed to only affect all the considered kinetic
parameters directly, and the resistivity and limiting current den-
sity are assumed to vary independent from temperature, covering
a wide range of cases that correspond to all possible temperature
levels. The relationship between temperature and resistivity, as
well as temperature and limiting current density, are established
as a post-processing step, as explained in Section 4. With this ap-
proach, the effect of temperature on the corrosion of steel in differ-
ent types of concrete produced with different types of cements and
water/cement ratios (w/c), which are exposed to different levels of
contaminants (chloride ions, etc.) that change the ionic composi-
tion of the solution (and hence the limiting current density and
resistivity), can also be investigated.
In this procedure, as shown in the analysis grid presented in
Table 1, the cover thickness is varied between 20 and 140 mm (5
steps), a range that covers most reinforced concrete structures.
Concrete resistivity can vary in a wide range, as much as five orders
of magnitude for different environmental conditions and materials
(e.g. temperature, the degree of saturation, cement type, etc.) [13];
hence it is considered between 50 and 10,000 X m (8 steps) to
cover all extremes. The lower end of the spectrum corresponds
to saturated poor-quality concrete and the higher end of the
spectrum corresponds to dry good-quality concrete. To investigate
the effect of limiting current density on the corrosion rate (i.e. the
effect of concentration polarization in the cathodic reaction) a wide
range between 0.002 and 0.6 A/m2 (8 steps) is analyzed. The lower
value is chosen such that it is not smaller than passivating current
density of steel [14]. The upper limit represents the condition in
which there is no concentration polarization. The range also repre-
sents the possible variations in limiting current density resulting
from lower oxygen concentrations in the presence of contaminat-
ing ions such as chlorides.
3. The effect of temperature on kinetic parameters
The effect of temperature on the corrosion of steel in concrete
can be due to its influence on the kinetic parameters of corrosion
and on the concrete/pore solution properties. With respect to the
former, the following kinetic parameters are considered in this
study: Tafel slopes, exchange current densities and equilibrium
potentials. The effect of temperature on the concrete/pore solution
properties is included through concrete resistivity and limiting
current density. Most reinforced concrete structures are generally
exposed to a relatively narrow range of temperature variations
(from 40 °C to +50 °C). However, since for most cases corrosion
is negligible at low temperatures, the temperature range
considered in this study is between +10 °C and +50 °C (i.e. 283–
323 K). The following sections describe how the current study
418 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
considers the effects of temperature on the kinetic parameters and
material properties.
3.1. Tafel slopes
A wide range of values has been recommended for anodic and
cathodic Tafel slopes of the passive and active steel in concrete
[15–17]. From theoretical point of view, temperature is one of
the factors that influence the Tafel slopes; i.e., Tafel slopes increase
with increasing temperature via:
2:303RT
ba ¼ ð5aÞ
azF
2:303RT
bc ¼ ð5bÞ
ð1  aÞzF
where z is the valance and a is the symmetry factor. It should be
noted that if a in Eq. (5) is considered constant [18], the change
in Tafel slopes for the range of temperatures concerned in this study
(i.e. 10–50 °C) is smaller than 7 per cent. In contrast with this the-
oretical deduction, the experimental investigations suggest some-
what different variation of Tafel slopes with temperature: Elesner
[19] suggests that the anodic and cathodic Tafel slopes slightly de-
crease with increasing temperature. Experimental investigations
carried out by Kriksunov [20] even show that Tafel slopes are inde-
pendent of temperature, and fluctuations can be considered as
experimental noise. Many attempts to explain the observed varia-
tion of Tafel slopes with temperature can be found in [21] and
monograph by Conway et al. [22]. Due to these contradictions, in
the present work, constant values for the anodic (90 mV/dec) and
cathodic (180 mV/dec) Tafel slopes are considered. It should be
noted that these Tafel slopes yield a polarization constant, B, of
ba bc
B¼ ¼ 26 mV=dec ð6Þ
2:303ðba þ bc Þ
which is in agreement with the experimental observations for ac-
tively corroding steel in concrete [5].
3.2. Exchange current density
The exchange current density of metals can be affected by the
pH of the solution, temperature [23] and characteristics of the
substrate [24]. Assuming that the pH of pore solution next to
reinforcing steel after onset of corrosion is relatively constant
due to the formation of saturated hydrous ferrous oxide on the
surface of steel [12] and high buffer capacity of concrete pore
solution, in this study, the exchange current densities of rebar
are considered independent of the pH. The effect of temperature
on the anodic and cathodic exchange current densities, on the
other hand, can not be ignored. Numerical investigations have
proven that with variation of exchange current density, corrosion
rate changes appreciably [25]. The strong dependency of ex-
change current density on temperature in acidic solutions can
be found in earlier works [18]. However, closer examination of
the activation energies obtained by Hurlen [18] suggests that
the activation energy for acidic solutions does not apply to highly
alkaline electrolytes such as concrete pore solution, and other ap-
proaches may be necessary for investigating the effect of temper-
ature on the exchange current density. To the knowledge of
authors, there is a lack of experimental data for exchange current
density of steel in high pH solutions at varying temperatures. To
overcome this problem an alternative approach is used such that
the exchange current density is expressed as suggested by Tanaka
and Tamamushi [26] via:
azFEe

i ¼ zFC o kc e RT ð7Þ
where i is the exchange current density (A/cm2), C o is the molar
concentration of the ion in consideration in the bulk of the solution

at the standard equilibrium conditions, kc is the rate constant at
standard equilibrium condition (cm/s) and Ee (volts) is the equilib-
rium potential (0.780 and 0.160 volts S.C.E. for iron oxidation and
oxygen reduction, respectively). If two values for the exchange cur-
rent density (i;1 and i;2 ) at two different temperatures (T1 and T2,
respectively) are considered, using Eq. (7) the following Arrhe-
nius-type relationship can be obtained:
 
azFEe 1 1
i;1 
¼e R T1 T2 ð8Þ
i;2
Using this equation along with an experimental value for ex-
change current density at a reference temperature, the effect of
temperature on the equilibrium exchange current density can be
analyzed. In the current investigation, the reference values for ano-
dic and cathodic exchange current densities at standard room tem-
perature are considered as 3  104 and 1  105 A/m2,
respectively [1].
3.3. Equilibrium potentials
The variation of the equilibrium potentials of the anodic (/Fe;T )
and cathodic (/O2 ;T ) half-cell reactions with temperature can be ex-
pressed with Eqs. (9) and (10), respectively [27]:
/Fe;T ¼ /Fe;T  þ 0:000052  ðT  T  Þ ð9Þ
/O2 ;T ¼ /O2 ;T   0:001680  ðT  T  Þ ð10Þ
where T  (293 K) is the reference temperature, /Fe;T  (0.780 volts
S.C.E.) is the equilibrium potential of iron reduction at T  , /O2 ;T 
(0.160 volts S.C.E.) is the equilibrium potential of oxygen reduction
at T  . Comparison of two equations suggests that the variation of
the reduction potential of the anodic half-cell reaction is signifi-
cantly smaller than that of the cathodic reaction. In fact, the varia-
tion of the anodic reduction potential with temperature range
considered herein (i.e. 283–323 K) can be considered negligible;
however, in the present work the variations of both equilibrium
potentials are considered.
4. The effect of temperature on concrete/pore solution
properties
As discussed previously, 16,000 simulated polarization resis-
tance tests were carried out by independently varying kinetic
parameters, concrete cover thickness, concrete resistivity and lim-
iting current density for 10 A/C ratios that vary between 0.1 and
1.0. The effect of temperature is assumed to directly affect the ki-
netic parameters only (see Section 3), and the resistivity and limit-
ing current density are assumed to vary independent from
temperature, covering a wide range of cases that correspond to
all possible temperature levels. The relationship between temper-
ature and resistivity, as well as temperature and limiting current
density, are established as a post-processing step, which is ex-
plained in this section.
4.1. Concrete resistivity
The effect of resistivity is considered by many researchers to be
one of the most important factors affecting the corrosion rate of
reinforcing steel in concrete [1,14,28]. A number of parameters
(e.g. relative humidity, temperature, the presence of external ions
such as chlorides, w/c, and cement type) can significantly change
the resistivity of concrete [13]. Establishing a reasonable relation-
ship between relative humidity and resistivity is not an easy task
 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
due to the coupled nature of the problem. On the other hand, such
a relationship exists between the water saturation and concrete
resistivity [13]. When required, by assuming equilibrium condi-
tions and relating the relative humidity to water saturation using
an adsorption isotherm [29], a relationship between the relative
humidity and concrete resistivity can be established. The appendix
provided in the Part 2 of this investigation summarizes the proce-
dure used in the current study.
The relationship between the concrete resistivity and tempera-
ture has been previously proposed in the form of Arrhenius type
equation [30]:
DU q 1 1
ð  Þ
q ¼ q e R T T ð11Þ
where q (X m) is the concrete resistivity, DU q (kJ/mole) the activa-
tion energy of resistivity. In this equation, the values of activation
energy suggested by Chrisp et al. [30] are a function of water satu-
ration and decrease with increasing water saturation: 39 kJ/mol (for
55% saturation) and 27 kJ/mol (for 100% saturation). These values
are also in agreement with the suggested values by Hope and Ip
[31]. In addition to relative humidity and temperature, other fac-
tors, such as the presence of external ions (e.g. chloride), w/c, ce-
ment type, can also affect concrete resistivity. In the present
work, these effects are considered implicitly as part of the reference
resistivity, q . By investigating cases for a wide range of reference
resistivities (50 and 10,000 X m, 8 steps), it is intended that the ef-
fect of temperature on the resistivity of different concrete types,
which are exposed to different environmental conditions and con-
taminants, can be studied.
4.2. Limiting current density
Temperature affects the limiting current density in two differ-
ent ways: (1) limiting current density increases with temperature
due to the increase of oxygen diffusion coefficient; (2) the increase
of temperature decreases the amount of dissolved oxygen in water
and lowers the limiting current density. The limiting current den-
sity, iL (A/m2), in the current study is related to these parameters
using Eq. (12) as a function of concrete cover thickness, d (m), oxy-
gen diffusion coefficient, DO2 (m2/s) and amount of dissolved oxy-
gen on the surface of concrete, C sO2 (mole/m3) [32]:
DO2 C sO2
iL ¼ zc F ð12Þ
d each case is the area under the anodic section of each current den-
Among these factors, oxygen diffusion coefficient can be repre-
sented as a function of relative humidity and porosity via [33]:
2:2
DO2 ¼ 1:92  106 e1:8
p ð1  hÞ ð13Þ
where ep is the porosity of the cement paste and h is the relative
humidity. Detailed calculation of ep based on concrete mix design
properties can be found in [33]; the appendix provided in the Part
2 of this investigation summarizes the procedure used in the
current investigation. The temperature dependency of the oxygen
diffusion coefficient can be expressed by Arrhenius type equation:
DU D 1 1
ð  Þ
DO2 ;T ¼ DO2 ;T  e R T  T ð14Þ
where DU D (kJ/mole) is the activation energy of the oxygen diffu-
sion coefficient and DO2 ;T  (m2/s) is the oxygen diffusion coefficient
at a reference temperature, T  . The value of activation energy, DU D ,
in Eq. (14) was proposed by Page and Lambert [34] as a function of
w/c (see the appendix in Part 2). It is interesting to note that activa-
tion energy increases by w/c initially (until w/c = 0.5) and decreases
afterwards; the same behavior is observed for the activation energy
of chloride diffusion [35]. Since no explicit relationship between the
value of activation energy and w/c is presented in [34] such a rela-
419
tionship is obtained by conducting regression through the experi-
mental values suggested in [34], details of which can also be
found in [11] and are summarized in the appendix of Part 2.
As it was mentioned before, temperature also affects the
amount of dissolved oxygen in water, C sO2 , and this effect can be
quantified through using [36]:
! !
1:575  105 6:642  107
LnC sO2 ¼ 139:344 þ 
T T2
! !
1:244  1010 8:622  1011
þ 3
 ð15Þ
T T4
where C sO2 (mg/litre) is the amount of dissolved oxygen in water at
standard pressure (i.e. 101.325 kPa). For the purpose of steel corro-
sion in concrete, the Eq. (15) provides good approximations, even in
non-standard conditions. Complete formulation can be found in
[36].
5. Results and discussion
This section presents the results of the numerical investigation
of steel corrosion in concrete through simulated polarization resis-
tance tests. The main objective here is to demonstrate the effect of
temperature on the corrosion rate through (1) varying kinetic
parameters, (2) varying concrete/pore solution properties. In this
section only the independent effects of the considered parameters
on the corrosion rate of steel in concrete are discussed. The mutual
dependence of parameters (e.g. dependence of limiting current
density and concrete resistivity to temperature) is discussed in Part
2.
One of the most useful outcomes of the numerical experiments
is the current density distribution at the steel/concrete interface.
Fig. 2a illustrates this current density distribution for a range of
A/C ratios for a given case. The positive values of each curve are
representative of the anodic current density and the negative val-
ues represent the cathodic current density. For every A/C ratio, the
anodic section of each curve reaches to the maximum current den-
sity, icorr,max, at the point of transition from anodic to cathodic sec-
tion. The largest maximum current density in all cases of A/C (e.g.
in Fig. 2a the largest maximum current density corresponds to A/
C = 0.1) is considered the maximum corrosion current density.
The total corrosion current (or simply ‘‘corrosion current”) of
sity curve in Fig. 2a. Fig. 2b illustrates the corrosion currents corre-
sponding to each A/C shown in Fig. 2a. It is clear from this figure
that there is an optimum value of A/C for which the corrosion cur-
rent reaches its maximum value. For example the maximum corro-
sion current in Fig. 2b corresponds to A/C = 0.54, hence, the
optimum A/C is equal to 0.54. Based on the basic thermodynamic
principles, this maximum can be interpreted as corresponding to
the state at which the total energy is minimized by achieving an
optimum (or equilibrium) A/C along the reinforcement. More
information about the concept of optimum A/C ratio and its theo-
retical investigation can be found in previous publications [1,9,11].
When the maximum corrosion current that corresponds to the
optimum A/C ratio is normalized over the anodic section of the
steel, the average corrosion current density, icorr,ave, is obtained.
In this study, the variation of both maximum and average cur-
rent density with respect to the varying independent variables
has been investigated. It was observed that the trends of variation
of both average and maximum corrosion rates with different
parameters are very similar, and the only difference is in the mag-
nitudes. For simplicity, the results obtained from the simulated
experiments are presented here in terms of average corrosion rate,
icorr,ave.
420 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
Fig. 2. (a) Current density distribution at the steel/concrete interface, (b) Variation
of corrosion current with A/C.
5.1. Kinetic parameters
In this section the effect of temperature on the kinetic parame-
ters of corrosion and associated changes in corrosion rate, while all
other properties are kept constant, is presented. Fig. 3 illustrates
the effect of temperature (affecting kinetic parameters only) on
corrosion rate for different values of limiting current density, resis-
tivity and cover thickness. In this figure, it can be observed that the
corrosion rate increases almost linearly with temperature in all
cases. In Fig. 3a (iL = 0.2 A/m2 and q = 5000 X m) corrosion rate in-
creases 16% for 140 mm cover thickness over the range of temper-
ature from 283 to 323 K; the increase for the same range is 11% for
20 mm cover thickness. It can be observed in Fig. 3c (iL = 0.005 A/
m2 and q = 10,000 X m) that the increase of corrosion rate with
temperature for 140 mm and 20 mm cover thickness are 17% and
10%, respectively. In general, changes in kinetic parameters of
corrosion due to temperature variations within the range of inves-
tigation (i.e. 283–323 K) result in a 0–20% change in corrosion
rates. The amount of change depends largely on concrete resistivity
and limiting current density. In order to see the individual effects
of limiting current density and resistivity, while kinetic parameters
are changing due to temperature variations, Figs. 4 and 5 are
plotted.
In Fig. 4, it can be observed that a nearly linear increase of cor-
rosion rate with temperature exists for different values of limiting
current density in concrete with 60 mm cover thickness. Irrespec-
tive of the limiting current density, temperature increases the cor-
rosion rate for all values of concrete resistivity. In addition, for high
values of concrete resistivity, the effect of limiting current density
on the corrosion rate decreases. In Fig. 4c, it is obvious that the lim-
iting current density has almost no effect on the corrosion rate at
different temperatures (i.e. 13% change for all values of limiting
current density). This is due to the fact that for large resistivity val-
ues, resistance control prevails over diffusion control, and the lim-
iting current density loses its significance.
The increase of corrosion rate with kinetic parameters affected
by temperature at different values of resistivity is illustrated in
Fig. 5. As the concrete resistivity increases, the corrosion rate de-
creases significantly and the effect of temperature on the corrosion
rate for all limiting current densities becomes less significant (i.e.
the slope of the lines approach to zero as resistivity increases). It
is also obvious from Fig. 5 that corrosion rate is inversely related
to the concrete resistivity.
5.2. Concrete/pore solution properties
As described in Section 4, temperature affects concrete resistiv-
ity and limiting current density. This section presents the effect of
temperature on these parameters while all kinetic parameters are
kept constant. Fig. 6 illustrates the effect of temperature on corro-
sion rate through changing concrete resistivity. The plot was ob-
tained after post-processing the raw simulated polarization
resistance data using the procedure provided in Section 4.1. The
reference resistivity used in Eq. (11) was q = 1250 X m at 20 °C
and 65% relative humidity (corresponding to approximately 50%
saturation at equilibrium conditions). The activation energy, DU q ,
corresponding to these conditions was taken as 39 kJ/mole, as sug-
gested by Chrisp et al. [30]. As illustrated in Fig. 6, within the range
of temperatures considered in this investigation (i.e. 283–323 K),
the resistivity changes from 2500 to 350 X m (7 reduction).
The corrosion rate within this range increases from 0.001 to
0.006 A/m2 (for low iL) and to 0.008 A/m2 (for large iL). This
corresponds to a significant increase in corrosion rate in the range
of 6–8 times, which demonstrate that there is a strong relationship
between the corrosion rate and concrete resistivity as affected by
changing temperature.
Fig. 7 illustrates the effect of temperature on corrosion rate
through changing limiting current density. As observed in Fig. 4,
since the effect of limiting current density is not large when the
corrosion is under resistance control, the plot was obtained after
post-processing the raw simulated polarization resistance data
using a resistivity of 50 X m. The limiting current densities corre-
sponding to different temperatures were calculated using the pro-
cedure described in Section 4.2 at the reference values of 20 °C and
65% relative humidity (corresponding to 50% saturation at equilib-
rium conditions). The activation energy, DU D , and the reference
oxygen diffusion coefficient, DO2 ;T  , corresponding to these condi-
tions were taken as 22 kJ/mole and 3  108 m2/s, respectively.
As illustrated in Fig. 7, within the range of temperatures considered
in this investigation (i.e. 283–323 K), the limiting current density
changes from 0.02 to 0.08 A/m2 (4 increase). The corrosion rate
within this range increases from 0.024–0.026 A/m2 to 0.0030–
 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425 421

Fig. 3. Variation of corrosion rate with temperature (affecting only kinetic parameters) for different values of cover thickness: (a) iL = 0.2 A/m2 and q = 5000 X m, (b)
iL = 0.008 A/m2 and q = 7500 X m, (c) iL = 0.005 A/m2 and q = 10,000 X m.

Fig. 4. Variation of corrosion rate with temperature (affecting only kinetic parameters) for different values of iL (60 mm cover thickness): (a) q = 650 X m, (b) q = 1250 X m,
(c) q = 5000 X m.

0.0036 A/m2. This corresponds to an increase in corrosion rate in
the range of 30%.
5.3. Other observations
Concrete cover is the main barrier against external agents
that induce or accelerate steel corrosion. A thick and dense
concrete cover can protect the reinforcing steel from depassiva-
tion for a reasonably long period of time. After depassivation,
the effect of the cover thickness on the kinetics of corrosion
has two opposing effects: (1) thicker concrete cover provides
a better distribution of electric potential in the bulk of the con-
crete, increasing the corrosion rate, a phenomena that was also
observed by other researchers [28]; (2) thicker concrete cover
lowers the rate of oxygen diffusion at high levels of water sat-
uration, resulting in concentration polarization of the cathodes
422 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
Fig. 5. Variation of corrosion rate with temperature (affecting only kinetic parameters) for different values of concrete resistivity (60 mm cover thickness): (a) iL = 0.2 A/m2,
(b) iL = 0.05 A/m2, (c) iL = 0.005 A/m2.
Fig. 6. Variation of corrosion current density with temperature (affecting resistivity
only) for constant values of kinetic parameters (/0Fe = 0.780 volts S.C.E.,
/0O2 = 0.160 volts S.C.E., ba = 90 mV/dec, bc = 180 mV/dec, ioa = 3  104 A/m2 and
ioc = 1  105 A/m2), limiting current density and cover thickness (d = 60 mm). The
concrete used in this plot has the following properties: w/c = 0.5, aggregate/cement
ratio = 3.0, cement type = OPC I, cement content = 400 kg/m3, air content = 3%.
and lowering the corrosion rate. The rate of increase in corro-
sion rate due to the first effect decreases as cover thickness in-
creases, approaching to a plateau for very thick covers. It is
interesting to note in Fig. 8 that, when limiting current density
is assumed to be independent from the cover thickness (i.e.
ignoring the second effect), the corrosion current density in-
crease with increasing cover thickness, and a plateau forms at
very large values of cover thickness. Fig. 8 also suggests that
the effect of cover thickness on the corrosion current density
decreases with increase of resistivity, hence it can be concluded
that the concrete resistivity is a strong controlling factor that
can diminish the effect of potential distribution. The second ef-
fect (i.e. the effect of concrete cover on oxygen availability) is
prominent in highly saturated concretes, in which the consump-
Fig. 7. Variation of corrosion current density with temperature (affecting limiting
current density only) for constant values of kinetic parameters (/0Fe = 0.780 volts
S.C.E., /0O2 = 0.160 volts S.C.E., ba = 90 mV/dec, bc = 180 mV/dec, ioa = 3  104 A/m2
and ioc = 1  105 A/m2), resistivity (50 X m) and cover thickness. The concrete used
in this plot has the same properties given in Fig. 6.
tion rate of oxygen is higher than the rate of diffusion through
the concrete cover. In these cases, by increase of cover thick-
ness the limiting current density decreases, resulting in de-
crease of corrosion rate. For unsaturated concrete with
relatively large concrete cover, cover thickness does not play
an important role in the rate of steel corrosion in concrete.
However, as the saturation of concrete increases, this effect be-
comes more significant. The effect of coupling between limiting
current density and cover thickness will be discussed in Part 2
as a parametric study.
Fig. 9 illustrates the effect of iL on the corrosion rate for different
cover thicknesses and temperatures. All the plots are constructed
for resistivity of 50 X m since for higher values of resistivity, the
effect of iL decreases significantly as it was previously discussed.
 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425 423

Fig. 8. Variation of corrosion current density with cover thickness for different values of resistivity at constant values of kinetic parameters calculated at 303 K: (a) iL =
0.2 A/m2, (b) iL = 0.05 A/m2, (c) iL = 0.005 A/m2. In these plots the effect of increasing cover thickness on the limiting current density is ignored.

Fig. 9. Variation of corrosion rate with iL at q = 50 X m for different cover thicknesses and kinetic parameters calculated at (a) T = 323 K, (b) T = 303 K, (c) T = 283 K.

The corrosion rates in Fig. 9 initially increase rapidly and approach
to a plateau for larger values of iL. This observation is in line with
polarization behavior of steel: as iL increases corrosion rate also in-
creases, however, as iL reaches to large values, the anodic and
cathodic polarization curves intersect before reaching to iL and
after this point, the increase of iL does not impose any effect on
the corrosion rate. Also, increase of corrosion rate with cover thick-
ness can also be observed in this figure, which is a result of better
distribution of electric potential.
6. Model development
Although the finite element approach introduced here to carry
out simulated polarization resistance experiments can be consid-
ered to be quite useful, it requires good understanding of the fun-
damental principles of the corrosion theory and the underlying
numerical approach; therefore it is not practical for use by the gen-
eral engineering community. A closed-form equation which uses
easily measurable/predictable parameters as input variables can
424 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425

help eliminate this problem. Obtaining an analytical closed-form Table 2

equation that is based on the solution of the Laplace’s equation The constants of Eq. (18) obtained from the regression analysis.

subjected to polarization boundary conditions in the considered icorr,ave icorr,max

domain is not possible. However, a closed-form solution that is Constant Value Constant Value
based on the non-linear multivariable regression analysis of the
s 1.181102362  103 s 1
test data obtained from the numerical experiments can be con- g 1.414736274  105 g 0.32006292
structed; and such an approach is carried out in this study. Due f 0.00121155206 f 53.1228606
to the large number of cases that have been analyzed through sim- j 0.0847693074 j 0.00550263686
ulated experiments, the considered parameters, which can be trea- k 0.130025167 k 0.120663606
c 0.800505851 c 0.787449933
ted as the main factors affecting corrosion of steel in concrete, are
l 1.23199829  1011 l 3.73825172  107
in a wide range that covers most of the real-world scenarios, the h 0.000102886027 h 47.2478753
outcome of the regression analysis is expected to have a large # 0.475258097 # 0.00712334564
scope of applicability. v 5.03368481  107 v 0.003482058
As it was stated previously, the maximum corrosion current m 90487 m 784679.23
- 0.0721605536 - 0.0102616314
density shows the same trend of variation with the considered
parameters (i.e. temperature, concrete resistivity, limiting current
density and cover thickness) as the average current density; there- cover thickness through d) and their coupling effects. For better
fore the final model representing both average and maximum cor- approximation, a power coefficient is considered for all indepen-
rosion current densities are expected to be in a similar format. dent variables, and also the powers of interaction terms are changed
However, since the rates of the variation of the average and max- to new unknowns; this provides more flexibility to the seed equa-
imum corrosion current density with each parameter are different, tion. A non-linear multivariable regression analysis is conducted
the coefficients of the parameters in the model representing the using a specialized computer program, NLREG [37], on the data ob-
two cases will be different. Therefore, before developing the model tained from the virtual experiments to obtain all the unknown coef-
for the maximum current density, the variation of the average cor- ficients, and insignificant terms are eliminated from the original
rosion current density with considered parameters is examined, seed equation. An insignificant term is a term whose coefficient is
the mutual effects of these parameters are investigated, and the very close to zero such that eliminating this term does not change
model for predicting the average corrosion current density is ob- the values of R2 significantly. Since the value of the coefficient of
tained. Then, a regression analysis on the data associated with each term is dependent on the values of the parameters included
the maximum corrosion current density is conducted considering in that term, after eliminating each insignificant term a new regres-
the final format of the equation obtained for the average corrosion sion analysis was conducted to control the significance of the elim-
current density to obtain the coefficients for the model represent- inated term (through its effect on R2). The following is the final
ing the maximum corrosion current density. Eq. (16) is the seed format of the equation after the revised regression analysis:
equation used for the non-linear regression analysis, which was i
c15 00
b0124 b124
developed after careful investigation of the data obtained from 1 ðTiL Þb14 T b15 b15
L
Td iL
icorr ¼ a3 þ a14 þ a15 þ a124 þa ð17Þ
the simulated experiments (which is partially presented in this pa- qc1 qc 1 qc1 q c1

per). The development process of the seed equation is not within By conducting a separate regression analysis on the data associ-
the scope of this paper due to space limitations; however the de- ated with the maximum corrosion current density, and considering
tails of this process can be obtained from [11]. the same format of Eq. (17) as the seed equation, a new set of coef-
ficients is obtained. Eq. (18) is the outcome of the regression anal-
b2 b b012
T b1 d 1 iL 4 i
c5
T b12 d ysis for both average and maximum current density:
icorr ¼ a1 þ a2 þ a3 þ a4 þ a5 b5L þ a12
qc 1 qc 2 qc 3 qc 4 qc12  
icorr;ave 1 j k -
T b13 b0
T b14 i 14 ¼ ðgTd iL þ lT miL þ hðTiL Þ# þ vqc þ fÞ ð18Þ
þ a13 þ a14 c L4 icorr;max sqc
q c13 1 q
iL15
c
b23
0
b24 b24 b25 i
c25
b
The coefficients of the equation (i.e. s; c; g; j; k; l; m; -; h; #; v; f)
b15
T b15 d d iL d b25
L
iL 34 are provided in Table 2. The average deviation from any observa-
þ a15 c015
þ a23 þ a24 þ a25 0 þ a34
q qc23 q c24
qc25 qc34 tion for Eq. (18) for average current density, icorr,ave, is
c 5
iL3 b
c
iL45 5  104 A/m2 (in the range of passivating current density); R2
b35 iL 45 b45
þ a35 þ a45 for this model is close to 0.99 (both the adjusted coefficient of mul-
c035 0
q qc45 tiple determination and the proportion of variance explained). The
c125
corresponding R2 for the maximum current density, icorr,max, model
00
b0123 b0124 b124 b0125 00 i
T b123 d T b124 d iL T b125 d b125L
þ a123 þ a124 þ a125 is close to 0.91. Although a better expression with a higher R2 value
qc123 q c124 qc125
c can be obtained for maximum corrosion current density by chang-
0
b134 b134 iL135
T iL T b135 b135 ing the format of the equation, it was decided that it is more prac-
þ a134 þ a135 0
q c134
qc135 tical to have the same format for both equations. Fig. 10 compares
0
b145 b145
c
00 iL145 0
b234 b234 b235 0i
c235
the results of the finite element analysis with those of the model
T iL b145 d iL d b235
L
for average corrosion current density. The verification of the devel-
þ a145 0 þ a234 þ a235 0
qc145 q c234
qc235 oped model and a parametric investigation are provided in Part 2.
c
b iL345
iL 345 b345
þ a345 0 þ a0 7. Conclusions
qc345
ð16Þ
Based on the simulated polarization resistance experiments car-
where the symbols a, b and c represent the coefficients of the equa- ried out to investigate the independent effects of kinetic parame-
tion. It should be noted that Eq. (16) contains all independent obser- ters, limiting current density and concrete resistivity on the rate
vations (i.e. the effects of kinetic parameters through the term T, of steel corrosion in the presence of varying temperatures, the fol-
resistivity through q, the limiting current density through iL and lowing observations are made:
 M. Pour-Ghaz et al. / Corrosion Science 51 (2009) 415–425
Fig. 10. Comparison of the results of the simulated experiments with the developed
model for average current density.
(1) When temperature is assumed to affect only the
kinetic parameters, the corrosion rate of steel in con-
crete increases almost linearly with temperature, and
the rate of increase decreases with increasing con-
crete resistivity.
(2) In general, changes in kinetic parameters of corrosion due
to temperature variations within the range of investigation
10-50 ºC (i.e. 283–323 K) result in change in corrosion rate up to
20%. The amount of change depends largely on concrete
resistivity.
(3) Irrespective of the limiting current density, increasing tem-
perature affects the kinetic parameters such that the corro-
sion rate for all values of concrete resistivity increases. In
addition, for high values of concrete resistivity, the effect
of limiting current density on the corrosion rate decreases.
This is due to the fact that for large resistivity values, resis-
tance control prevails over diffusion control, and the limiting
current density loses its significance.
(4) The effect of temperature on corrosion rate through chang-
ing resistivity (while all other parameters are kept constant)
has been established as the most important effect consid-
ered in this investigation. It has been observed that corro-
sion rates increased 6–8 folds with variations in
temperature from 283 to 323 K.
(5) Corrosion rates have been shown to increase around 30% due
to the changing limiting current density (while all other
parameters are kept constant) caused by temperature varia-
tions from 283 to 323 K.
(6) If the coupling effects of cover thickness with limiting cur-
rent density is ignored, corrosion rate of steel in concrete
increases with cover thickness due to better distribution of
potential in the bulk of concrete, and the rate of this increase
decreases with increasing concrete resistivity.
425
(7) Corrosion rate of steel in concrete shows a rapid initial
increase with increasing limiting current density, and the
rate of the increase decreases significantly, reaching to a pla-
teau, for larger values of limiting current density.
These observations are in agreement with available experimen-
tal observations [6,12]. The data obtained from the simulated
experiments have been used to carry out a statistical analysis to
develop a closed-form mathematical model to predict maximum
and average corrosion rate of actively corroding steel in concrete.
The mathematical model has been shown to represent the simu-
lated polarization resistance tests very accurately.
Acknowledgements
This research has been funded by Natural Sciences and
Engineering Research Council of Canada (NSERC).
References
[1] P. Ghods, O.B. Isgor, M. Pour-Ghaz, Mater. Corros. 58 (2007) 265–272.
[2] S. Morinaga, in: Proceedings of the 5th International Conference on Durability
of Building Materials and Components, Brighton, UK, 1990.
[3] S. Morinaga, in: Proceedings of the 7th International Conference on Durability
of Building Materials and Components, Stockholm, Sweden, 1996.
[4] T. Liu, R.W. Weyers, Cem. Concr. Res. 28 (1998) 365–379.
[5] C. Andrade, J.A. Gonzalez, Werkst. Korros. 29 (1978) 515–519.
[6] D.A. Jones, Principles and Prevention of Corrosion, second ed., Prentice-Hall
Inc., Upper Saddle River, NJ, 1996.
[7] T. Luping, SP REPORT 2002:25, SP Swedish National Testing & Research Inst.,
2002.
[8] R.S. Munn, O.F. Devereux, Corrosion 47 (1991) 612–618.
[9] M. Stern, A.L. Geary, J. Electrochem. Soc. 104 (1957) 56–63.
[10] O.B. Isgor, A.G. Razaqpur, Mater. Struct. 39 (2006) 259.
[11] M. Pour-Ghaz, A Novel Approach for Practical Modelling of Steel Corrosion in
Concrete, M.Sc Thesis, Carleton University, Ottawa, Canada, 2007.
[12] H.H. Uhlig, R.W. Revie, Corrosion and Corrosion Control, third ed., John Wiley &
Sons, New York, 1985.
[13] O.E. GjØrv, Ø. Vennesland, A.H.S. El-Busiady, in: Proceeding of the 9th Annual
Offshore Technology Conference, Houston, Texas, 1977, pp. 581–588.
[14] C. Alonso, C. Andrade, J.A. Gonzalez, Cem. Concr. Res. 18 (1988) 687–698.
[15] B. Elsener, Cem. Concr. Compos. 24 (2002) 65–72.
[16] P. Garces, M.C. Andrade, A. Saez, M.C. Alonso, Corros. Sci. 47 (2005) 289–306.
[17] A.A. Sagues, S.C. Kranc, Corrosion 48 (1992) 624–633.
[18] T. Hurlen, Acta Chem. Scand. 14 (1960) 1564–1570.
[19] B. Elsener, Corros. Sci. 47 (2005) 3019–3033.
[20] L. Kriksunov, Electrochim. Acta 40 (1995) 2553–2555.
[21] E. Gileadi, J. Electrochem. Soc. 134 (1987) 117–120.
[22] B.E. Conwoy, D.F. Tessier, D.P. Wilkinson, J. Electrochem. Soc. 136 (1989)
2486–2493.
[23] T. Hurlen, Acta Chem. Scand. 14 (1960) 1555–1563.
[24] J. O’M. Bockris, R.J. Mannan, A. Damjanovic, J. Chem. Phys. 48 (1968) 1898–
1904.
[25] J. Ge, O.B. Isgor, Mater. Corros. 58 (2007) 573–582.
[26] N. Tanaka, R. Tamamushi, Electrochim. Acta 9 (1964) 963–989.
[27] A.J. Brad, R. Parsons, J. Jordan, Standard Potentials in Aqueous Solution, CRC
Press, New York, 1985.
[28] J. Gulikers, Mater. Corros. 56 (2005) 393–403.
[29] Y. Xi, Z.P. Bazant, H.M. Jennings, Adv. Cem. Based. Mat. 1 (1994) 248–257.
[30] T.M. Chrisp, G. Starrs, W.J. McCarter, E. Rouchotas, J. Blewett, J. Mater. Sci. Lett.
20 (2001) 1085–1087.
[31] B.B. Hope, K.A.C. Ip, Report ME-87-02, Ontario Ministry of Transportation and
Communication, Ontario, Canada, 1987.
[32] H. Bohni, Corrosion in Reinforced Concrete Structures, CRC Press, New York,
2005.
[33] V.G. Papadakis, C.G. Vayenas, M.N. Fardis, ACI Mater. J. 8 (1991) 186–196.
[34] C.L. Page, P. Lambert, J. Mater. Sci. 22 (1987) 942–946.
[35] C.L. Page, N.R. Short, A. El Tarras, Cem. Concr. Res. 11 (1981) 395–406.
[36] A.D.F. Eaton, A.H. Mary, Standard Methods for the Examination of Water and
Wastewater, twenty first ed., American Public Health Association,
Washington, DC, 2005.
[37] NLREG, Nonlinear Regression Analysis Program Version 6.3, Copyright Ó
1992–2005, Phillip H. Sherrod, <http://www.nlreg.com>, 2007.