Cement & Concrete Composites 32 (2010) 360–368

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Modeling the kinetics of corrosion in concrete patch repairs and identification

of governing parameters
S. Soleimani a, P. Ghods a, O.B. Isgor a,*, J. Zhang b
a
Carleton University, Dept. of Civil & Env. Engineering, Ottawa, Ontario, Canada
b
National Research Council of Canada, Ottawa, Ontario, Canada

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

Article history: Patch repair is a commonly used method for rectifying localized corrosion damage in reinforced concrete
Received 25 September 2008 members. However, the ring-anode effect, which is corrosion at the intersection between the substrate
Received in revised form 28 January 2010 and the repaired concrete, is a commonly observed failure mechanism after patch repairs. In this study,
Accepted 4 February 2010
the kinetics of the corrosion in the ring-anode zones of repaired concrete structures is investigated by
Available online 10 February 2010
numerical modeling. All simulations, in agreement with the existing experimental and in situ observa-
tions, have demonstrated the formation of a ring-anode zone in the 2–5 cm portion of the substrate from
Keywords:
the interface between the substrate and the repaired concrete. Furthermore, the anodic current density in
Reinforcement (D)
Corrosion (C)
the ring-anode zone is found to have a peak near the repaired concrete and to asymptotically approach to
Modeling (E) the corrosion current density observed in the substrate before the patch repair. More importantly, the
Electrical properties (C) simulations demonstrate quantitatively how various parameters affect the ring-anode corrosion. It was
Patch repair found that the resistivities of substrate and repair concretes are the most significant factors that influence
the magnitude of macrocell corrosion, followed by the availability of oxygen in the patch. It was observed
that the ring-anode effect is a localized problem that occurs in the substrate near the repair bond line, and
the size of the patch and the cover thickness are not significant factors affecting the phenomenon.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction Many patch repairs, however, have been found to last for only a
few months to a year before the appearance of new corrosion dam-
Corrosion of steel in concrete is one of the most costly durability age. Recurring corrosion after patch repair can initiate in one of the
problems that causes the degradation of concrete structures in three areas: the substrate (adjacent unrepaired area), patch area, or
terms of structural safety, integrity, and serviceability. Patch repair at the interface between them [2–8]; however, it has often been
is a commonly used method for rectifying localized damage in noted to occur in the substrate very close to the interface or at
reinforced concrete members such as highway bridge decks and the interface. This is known as ring-anode effect, and the cause
columns. The process of patch repair consists of the removal of has been attributed to macrocell corrosion formed between the
loose concrete that has cracked, spalled, or delaminated; (often) steel in the repaired patch (macrocell cathode) and the steel in
the application of surface treatment on the steel; and the replace- the substrate (macrocell anode), as evidenced by experimental
ment of the defective concrete with patching materials that nor- demonstrations [2–6]. It is well understood that the driving force
mally reestablish the original form of the member. Any repair of for macrocell corrosion is the electrochemical incompatibility be-
corrosion-induced damage must aim at stopping one or more of tween the patch and the substrate [9–13], which can be defined
the following processes that are required for the corrosion process as the imbalance in electrochemical potential between different
to proceed: (1) the anodic process; (2) the cathodic process; and/or locations of the reinforcing steel because of their dissimilar envi-
(3) the electrolytic conduction between the anodic and cathodic ronments caused by the patch repair [11]. The dissimilar environ-
sites on the steel reinforcement [1]. The main objective of the ments can be due to the differences in both physical properties
patch repairs of corrosion-induced damage in concrete structures (e.g., porosity) and chemical compositions. For example, the newly
is, at the very least, to prevent the anodic reaction from recurring patched area is usually free of chloride, while the substrate is not.
in the repaired area. Consequently, the corrosion potential of the steel in a patch can be
much higher than that in a substrate; this electrochemical poten-
tial imbalance was shown to be as high as 500 mV [4].
* Corresponding author. Tel.: +1 613 520 2600x2984; fax: +1 613 520 3951. Electrochemical potential imbalance explains the thermody-
E-mail address: burkan_isgor@carleton.ca (O.B. Isgor). namics of macrocell corrosion caused by patch repairs. The kinetics

0958-9465/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2010.02.001
 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368 361

of macrocell corrosion – the corrosion current density and its dis- the interface (repair bond line). As demonstrated by his model in
tribution along the reinforcing bar – that directly leads to deterio- Fig. 1b, the macrocell current bypassed a (1–2 cm) portion of rebar
ration, however, is not well understood, and limited knowledge outside the repair bond line, driving the chlorides out of this region
exists on the key factors that affect these key characteristics [14]. with a possible rate of 2–3 mm/month under a macrocell potential
Understanding the distribution of the corrosion current density of 300 mV. As a result, the interface was protected from corrosion
along the rebar is critical to predict the location and the severity and remained passive.
of the corrosion damage. Experimental approaches are not practi- Further study to clarify the location of active corrosion, either
cal or, in most cases, useful to map the current density distribution along the repair bond line or in the substrate, is important to
along the steel reinforcement because of the complex and highly understand the mechanism and to take preventive measures. Cas-
inhomogeneous nature of reinforced concrete and the limitation tro et al.’s model implied that the quality of the interface (i.e., the
of conventionally used electrochemical equipment for isolating repair bond line) between the patch and the substrate was the
the influence from adjacent macrocells when a localized corrosion most critical factor for the induced corrosion, and therefore a
is measured. strong bond between the repair concrete and the substrate should
Besides the distribution of the anodic current density, the pre- be beneficial in preventing the ring-anode effect. In contrast, Bar-
cise location of the anodic site – whether it is in the substrate or key’s model showed the steel at the interface remains in a passive
at the interface – is still not clear. There are two contradictory state bypassed by the macrocell current. Numerical modeling is
experimental findings. Castro et al. [5] monitored the current flow considered as a suitable and useful tool to study the kinetics of
between segmented steel bars to identify the interfacial corrosion reinforcement corrosion after patch repairs in concrete structures
and the corrosion in the substrate. They found that the interface and to specifically address the following questions [14] on the
was mostly anodic, which was directly controlled by the corrosion ring-anode effect: (1) is the anodic site after a patch repair at the
activity in the patch (macrocell cathode). The corrosion in the sub- interface or in the substrate? and (2) what are the anodic corrosion
strate was in turn determined by the corrosion activity at the inter- characteristics of the steel after patch repair with regards to the
face. This is illustrated by the corrosion current distribution in size of active corrosion (e.g., affected length of rebar) and its
Fig. 1a, in which the current flow (only the current flow in the con- current density?
crete is illustrated; the direction of electron flow in the steel would
be in the opposite direction) is from the interface to both the patch 2. Numerical approach
and the substrate. Their experiments showed that the anodic site
extended in both directions (patch and substrate) for 4 cm from 2.1. Initial active microcell corrosion in the substrate
the interface. Barkey [15] reported however, that the interface re-
mained passive and suggested that this passive region might ex- After a patch repair, the steel in the substrate often undergoes
tend out of the repaired region into the substrate. An more active corrosion than in the patch. This is mainly due to
examination of the corrosion products revealed that the inner the fact that the substrate concrete, either by chloride contamina-
boundary of the anode in the substrate was 1–2 cm away from tion or carbonation, is usually more corrosive than the newly
patched concrete. It is also suggested that the substrate can devel-
op more active microcell corrosion after a patch repair, because the
patch repair removes the damaged area, which was an anodic site
for the adjacent substrate, and consequently, its cathodic protec-
tion on the steel in the substrate is lost [16]. In the current inves-
tigation, it is therefore assumed the steel in substrate undergoes
uniform microcell corrosion as an initial condition right after a
patch repair, and the investigations will focus on the interactions
of the passivated steel in the patch area and the steel in the sub-
strate and their final respective corrosion equilibrium state, espe-
cially the corrosion of the steel in the substrate (see Fig. 2).
Chloride contaminated Patch repair The initial microcell corrosion current density of the steel in the
Localized anode
concrete substrate, icor,mic (A/m2), as illustrated in Fig. 2a, can be calculated
from exchange current density following the polarization and the
(a) Castro et al.’s model mix potential theories [17] as follows:

(1) Anodes polarize through activation polarization such that

 
ia
/a ¼ /oa þ ba log ð1Þ
ioa
|fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}
Activ ation

where /a (V) is the anodic corrosion potential, /oa (V) is the

equilibrium potential of the anodic reaction, ba is the anodic
Tafel slope (V/dec), and ioa (A/m2) is the exchange current
density of the anodic reaction.
Patch repair
Chloride contaminated (2) Cathodes polarize through activation and concentration (due
concrete Localized anode to the limiting effect of oxygen availability around the catho-
dic sites) polarization such that
   
(b) Barkey’s model ic 2:303RT iL
/c ¼ /oc þ bc log þ log ð2Þ
ioc zc F iL  ic
Fig. 1. Schematic illustration of two proposed model for ring-anode effect. |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Activ ation Concentration
362 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368

Before the repair After the repair

Microcell corrosion icor,mic Combined microcell and macrocell corrosion
0.12

Current density of steel (A/m2)

0.09
Substrate Patch repair
Substrate(Old concrete) (Old concrete)
0.06

0.03

(a) Steel rebar (b)

Fig. 2. Schematic illustration of the patch repair effect on the mechanism of rebar corrosion (the numerical results were obtained by the authors using substrate resistivity of
500 X m, patch resistivity of 1000 X m, cover thickness of 100 mm, and iL of 0.2 A/m2).

where /c (V) is the cathodic corrosion potential, /oc (V) is the ence very slow microcell corrosion rates in the vicinity of passive
equilibrium potential of the cathodic reaction, bc is the catho- current density of steel in concrete, while the steel sections in
dic Tafel slope (V/dec), ioc (A/m2) is the exchange current den- the substrate undergoes higher microcell activity in the vicinity
sity of the cathodic reaction, iL (A/m2) is the limiting current of microcell corrosion rates that were experienced before the re-
density, R (8.314 J/(mole K)) is the universal gas constant, pair that can be calculated by Eq. (4). In addition, the electrochem-
(F  96; 500 C=mole) is the Faraday’s constant, T (K) is tem- ical potential imbalance between the steel in the repaired zone and
perature, and zc is number of electrons that are involved in the steel in the substrate results in macrocell activity. Therefore,
the cathodic reaction. The limiting current density, iL (A/ after the patch repair, the corrosion mechanism of steel reinforce-
m2), is a measure of oxygen availability around the cathodic ment in concrete changes from uniform microcell corrosion to a
sites on the steel surface, and in the current study, is defined combination of microcell and macrocell corrosion, hence the corro-
by Eq. (3) as a function of concrete cover thickness, d (m), sion current density distribution varies along the rebars and devi-
oxygen diffusion coefficient, DO2 (m2/s), and amount of dis- ates from the microcell corrosion rate calculated from Eq. (4) [20],
solved oxygen on the surface of concrete, C sO2 (mole/m3) [18]: as illustrated in Fig. 2b.
In the repaired zone, the polarization curve of the anodic micro-
DO2 C sO2
iL ¼ zc F ð3Þ cell reaction no longer follows the activation polarization equation
d given in Eq. (1) due to the formation of the protective film but
(3) When the equilibrium is reached, at any point on the steel shows an s-shape behaviour as illustrated in Fig. 3 [21]. This polar-
surface, the rates of anodic and cathodic reactions, ia (A/ ization behaviour can be estimated through the following modified
m2) and ic (A/m2), respectively, will be equal to each other anodic polarization equation [21,22]:
and to the microcell corrosion current density, icor,mic. Since
 
the distance between the anodic and cathodic sites on the ia;patch
/a;patch ¼ /oa þ ba log þ ia;patch Rf ð5Þ
steel surface are very small, the effect of concrete resistivity ioa
(i.e., IR drop) can be ignored [19], and consequently, the
potentials of anode and cathode can be considered equal in where /a;patch (V) is the anodic corrosion potential of steel in the re-
microcell corrosion. Therefore, using Eqs. (1) and (2), the paired region, Rf (X m2) is the electrical resistance of the protective
microcell corrosion current density, icor,mic, on the steel sur- film, and ia,patch (A/m2) is the anodic current density of steel in the
face in the concrete before the patch repair can be obtained repaired region. The polarization curve of the cathodic microcell
by numerically solving the following non-linear equation: activity in the repaired area still follows the same equation as the
   
icor;mic icor;mic
/oc  /oa þ bc log  ba log
ioc ioa φ(V)
 
2:303RT iL Eq.5
þ log ¼0 ð4Þ
zc F iL  icor;mic
2.2. Corrosion of steel at the intersection of the substrate and the patch
Eq.1
The process of patch repair consists of (1) the removal of the
damaged concrete, (2) the application of surface treatment on (ia,repaired .Rf )
the steel, and (3) the replacement of the defective concrete with
new concrete that is free from chlorides (see Fig. 2). After the appli-
ia,repaired
cation of the patch repair, it is assumed that the reinforcing steel in β a log( )
ioa
the repaired zone is repassivated, while the steel in the unpatched φa ο

zone (the substrate or the old concrete) remains active due to the
presence of chlorides, or later becomes active when chloride con-
centration exceeds a threshold. Reinforcing steel sections in both ioa i (A/m2)
zones continue to experience microcell activity, albeit at different
rates: the steel sections in the repaired chloride-free zone experi- Fig. 3. Schematic illustration of anodic polarization of rebar in the repaired zone.
 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368 363

cathodic polarization equation (Eq. (2)); therefore the following comes smaller, as illustrated in Fig. 4b. However, since the
equation can be written: electrical neutrality must be preserved, the difference between ano-
    dic and cathodic current densities in both zones (i.e., the substrate
ic;patch 2:303RT iL
/c;patch ¼ /oc þ bc log þ log ð6Þ and the repaired zone) should be identical [24] such that:
ioc zc F iL  ic;patch
ic;patch  ia;patch ¼ ia;substrate  ic;substrate ð10Þ
where /c;patch (V) and ic,patch (A/m2) are the cathodic corrosion poten-
tial and the cathodic microcell corrosion density of steel in the re-
paired zone, respectively. 2.3. Numerical solution
In the substrate, where the concrete is still contaminated, the
anodic and cathodic polarization reactions follow the same equa- Although the one-dimensional simplification described by Eq.
tions that have been introduced earlier (Eqs. (1) and (2), respec- (9) is useful for theoretically explaining the concept of macrocell
tively) and can be written respectively as: corrosion in a patch repair, the actual problem given in Fig. 1 needs
  to be solved in the two-dimensional domain (approximating 3D
ia;substrate
/a;substrate ¼ /oa þ ba log ð7Þ phenomenon); therefore, a closed-form solution is not possible.
ioa
In this study, a non-linear finite element solution technique is used
  to solve the governing differential equation for the electrical po-
ic;substrate
/c;substrate ¼ /oc þ bc log tential distribution in the domain of analysis, as illustrated in
ioc
  Fig. 5, and to calculate the corrosion current densities on the steel
2:303RT iL
þ log ð8Þ surface. The governing equation for electric potential distribution
zc F iL  ic;substrate is given by:
where /a;substrate (V) and ia,substrate (A/m2) are the anodic corrosion po- 1
tential and the anodic microcell corrosion density of steel in the r r/ ¼ 0 ð11Þ
q
substrate, respectively; and /c;substrate (V) and ic,substrate (A/m2) are
the cathodic corrosion potential and the cathodic microcell corro- where / (V) is the electrical potential and q (X m) is the concrete
sion density of steel in the substrate. resistivity. Once the nodal potentials (hence the potential gradients
If the two-dimensional problem that is illustrated in Fig. 1 is are determined), the current density distribution on the steel sur-
simplified as a hypothetical one-dimensional setup (as in the case face in both the patch and substrate areas is calculated using Ohm’s
of two parallel rebars, separated by concrete, and experiencing law such that
macrocell corrosion – similar to the ASTM G109 [23] test setup),
the theory behind the macrocell activity in patch repairs can be 1 @/
ipatch ¼  ð12Þ
visualized better. For the simplified one-dimensional macrocell qpatch @n
problem, since the corrosion potentials of the two zones (i.e., the
repaired zone and the substrate) are not the same (see Fig. 4), mac- and
rocell corrosion will proceed such that: 1 @/
isubstrate ¼  ð13Þ
/patch  /substrate ¼ IRcon ð9Þ qsubstarte @n
where /patch (V) and /substrate (V) are the potentials of the repaired where qpatch (X m) and qsubstrate (X m) are the concrete resistivities
zone and the substrate, respectively, Rcon (X) is the concrete resis- of the patch and the substrate area, respectively, and n is the direc-
tance, and I (A) is total macrocell corrosion current. Due to this mac- tion normal to the equipotential lines.
rocell corrosion current, the microcell activities in the repaired zone At each node of the boundary on the steel surface, two condi-
and substrate are also affected. Macrocell current shifts the anodic tions should be satisfied. In the patch zone, the anodic and the
current density along the steel in the substrate towards larger val- cathodic potentials (given by Eqs. (5) and (6), respectively) need
ues while the cathodic current density of the same segment be- to be equal such that
Potential φ

Potential

φpatch-φsubstrate=IRcon
φpatch
IRcon
φsubstrate
φ cor,mic

ic,substrate imic ia,substrate ia,patch ic,patch

Corrosion rate, i Corrosion rate, i

(a) Substrate (Old concrete) (b) Patch repair

Fig. 4. Anodic and cathodic polarization curves.
364 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368

domain that is 300 mm long; the cover thickness of concrete was

d= Cover thickness

Cl - Cl - a variable of the investigation. Preliminary simulations have dem-

Cl - onstrated that, mostly due to the high resistivity of concrete, mac-
Substrate concrete Patch repair rocell activity is limited to a narrow area that spans both sides of
- the boundary between the substrate and the repaired zone; hence
Cl
Cl - Cl - Cl - 300 mm length of the domain was found to be sufficient to simu-
Cl - late the patch repair problem accurately. The domain is discretized
by 5  5 mm square finite elements, again, after a preliminary sen-
sitivity analysis that aimed at finding the optimum element size
L = 300 mm that establishes a balance between accuracy and numerical effi-
∂φ ciency. The values of the constant parameters of the simulations
=0 (e.g., exchange current densities, Tafel slopes, etc.) are presented
∂n
in Table 1. These constants were selected to represent typically
measured properties during corrosion of steel in concrete. Further
information on the effect of these constants on the results of the
simulations can be found in another paper [28].

3. Results and discussion

3.1. Ring-anode phenomena

Substrate zone boundary Repaired zone boundary All numerical simulations carried out in this study have clearly
φsubstrate = φa ,substrate = φc ,substrate φ patch = φa , patch = φc , patch demonstrated the formation of a ring-anode zone in the substrate.
As illustrated in Fig. 6, which is a plot of the current density along
Fig. 5. Schematic illustration of the domain, active and passive area and corre- the length of the reinforcing steel for a typical simulation, the for-
sponding boundary conditions. mation of the ring anode is mainly due to the increased corrosion
density along the 2–5 cm portion of the steel reinforcement from
the intersection between the repaired zone and the substrate.
/patch ¼ /a;patch ¼ /c;patch ð14Þ
The size of the ring anode observed in the simulations is also in
2 agreement with the experimental and in situ observations. The
and the current density, ipatch (A/m ), is equal to the difference of
anodic and cathodic current densities as per anodic current density (steel dissolution rate) in this narrow ring
has a peak near the intersection and asymptotically approaches
ipatch ¼ ic;patch  ia;patch ð15Þ to the initial microcell corrosion density before their interactions.
Solving these two equations simultaneously, /patch is obtained The current density in the repaired zone experiences a sharp de-
for each node of the boundary in the repaired area. These values crease at the intersection to negligible values (which can be con-
are used as Dirichlet boundary conditions at each node on the sur- sidered as passive state) towards the centre of the patch. The
face of steel in the patch area. In the substrate, the anodic and the inference from these results is that although the patch repair sys-
cathodic potentials (given by Eqs. (7) and (8), respectively) are also tem decreases the microcell corrosion rate of the repaired area to
identical such that the level of passive current density, the macrocell corrosion rate
created due to the potential gradient between the two zones in-
/substrate ¼ /a;substrate ¼ /c;substrate ð16Þ creases the anodic corrosion rate, particularly for the region close
to the border of repair (see Fig. 2).
and the current density, isubstrate (A/m2), is calculated as:
The simulations show that the ring-anode zone starts exactly at
isubstrate ¼ ia;substrate  ic;substrate ð17Þ the interface between the substrate and the repaired zone, and this
result may initially seem different from Barkey’s observation that
The boundary values at each node on steel in the substrate,
the passive region might extend out of the repaired region into
/substrate , is calculated by simultaneously solving Eqs. (16) and
the substrate such that the inner boundary of the anode in the sub-
(17) and is used as Dirichlet boundary condition for steel in the
strate is 1 cm or 2 cm away from the interface. As Barkey [15] also
substrate.
pointed out, the shift of the ring anode into the substrate can be
During the solution of Eq. (11), the electric neutrality condition,
attributed to the movement of the chlorides from the interface into
which is defined in Eq. (10), is persistently enforced. Due to the
the substrate hence creating a narrow 1–2 cm zone from the inter-
non-linear nature of the boundary conditions, the solution of Eq.
face at which the chloride threshold value is not exceeded. As a re-
(11) requires a non-linear solution algorithm with an appropriate
sult, the interface was protected from corrosion and remained
iteration method. Here, the modified direct iteration method that
passive. Since the numerical approach used in this study did not
is based on a relaxation algorithm is implemented to seek conver-
gence. Nevertheless, attaining the convergence by this method can
also be difficult at times, especially when the oxygen concentration Table 1
around the steel in the substrate becomes low, resulting in signif- Model input parameters and their assumed values.
icant concentration polarization of the cathodic reaction. The de- Parameter Value
tails of the non-linear solution procedure, the finite element
Cathodic exchange current density (ioc) 0.00001 A/m2
algorithm and the numerical difficulties associated with it can be Anodic exchange current density (ioa) 0.0003 A/m2
obtained from [25–27]; therefore they will not be provided here. Cathodic standard potential (/oc ) 0.16 V
Using the described approach, a numerical investigation was Anodic standard potential (/oa ) 0.78 V
carried out to study the kinetics of reinforcement corrosion after Passive film electrical resistance (Rf) 500 X m2
Cathodic slope (bc) 0.180 V/dec
patch repairs and to identify the governing parameters that affect
Anodic slope (ba) 0.090 V/dec
the process. As illustrated in Fig. 5, the analysis is carried out in a
 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368 365

0.12 Substrate Patch repair Node number

0 20 40 60
0.1 A
C
Current density (A/m2)

Substrate Patch repair

0.08 0.12

Current density (A/m2)

A: Anodic current density
B B: Cathodic current density
0.06 C: Corrosion before repair
0.09
D=A-B: Macrocell corrosion

0.04 Substrate resistivity=100 Ω.m

0.06 Cover thickness=50 mm, iL=0.2A/m2
D
0.02 (a) S/P=0.5 Patch resistivity=50 Ωm

0.03 Patch resistivity=100 Ωm

0 Patch resistivity=200 Ωm

-0.02 0
0.12
0 0.2 0.4 0.6 0.8 1

Current density (A/m2)

Normalized Length
0.09
Fig. 6. Anodic and cathodic current density of rebar after the patch repair in the
substrate and in the patch repair.
0.06

consider the movement of chloride in the concrete, the shift of the 0.03
(b) S/P=1
ring anode is not obvious in the results; however, as also stated by
Barkey [15], the macrocell potential gradient at the interface be-
0
tween the substrate and the patch can drive the chloride at rates 0.12
as high as 2–3 mm/month and shift the passive zone by 1–2 cm
Current density (A/m2)
within a year. However, it should be noted that, as demonstrated
0.09
in the current study, the movement of chlorides is not a require-
ment for the formation of the ring-anode zone. The ring-anode
zone form mainly due to the combined macrocell and microcell 0.06
activity after the patch repair. In summary, the numerical simula- (c) S/P=2
tions, which incorporate the kinetics of combined microcell and 0.03
macrocell corrosion in the ring-anode zones after a patch repair,
support Barkey’s observations.
0
0 0.2 0.4 0.6 0.8 1
3.2. Effect of patch size and concrete cover thickness Normalized Length

Fig. 7. Effect of the patch size on the corrosion current density of rebar: (a) S/P = 0.5,
The effect of patch size on the kinetics of the corrosion along the
(b) S/P = 1, (c) S/P = 2.
steel reinforcement is investigated by numerically simulating three
scenarios for the repair of a 100 X m concrete with patches of the
same (100 X m), lower (50 X m) and higher (200 X m) resistivi- 0.128
ties, but with different sizes such that the substrate-to-patch ratios 0.124
(S/P) were 0.5, 1.0, and 2.0. The results of this investigation are
ipeak (A/m2)

0.12
illustrated in (Figs. 7 and 8). It can be observed from this figure that Substrate resistivity=100 Ω.m
the patch size (i.e., S/P ratio) does not have any significant effect on 0.116 Cover thickness=50 mm, i L=0.2A/m 2
the distribution of the current density in the ring-anode zone. The Patch resistivity= 200 Ω.m
0.112
peak current density at the interface and also the size of the ring Patch resistivity= 100 Ω.m
0.108 Patch resistivity= 50 Ω.m
anode were not affected by the S/P ratio or by the selection of
the resistivity of the patch (see Figs. 7 and 8). Therefore it can be 0.104
stated that the ring-anode problem is a much localized problem 0.4 0.8 1.2 1.6 2
that is governed by the kinetics of corrosion around the interface S/P
between the substrate and the repaired zone. The size of the patch Fig. 8. Effect of patch size on the maximum corrosion rate at the ring anode, ipeak.
is not a significant factor that positively or negatively affects the
ring-anode characteristics.
curve of the ring-anode area (i.e., corrosion current, I) increases
The effect of concrete cover thickness on the corrosion current
by the increase of cover thickness, and this increase is larger when
density due to the patch repair was studied numerically using
patch with higher resistivity than that of the substrate is used. In
the developed model. Three concrete resistivities (100, 200, and
other words, when patch with higher resistivity than that of the
500 X m) were simulated with S/P of 0.5. In these cases, the limit-
substrate is used, the dependency of the total corrosion current
ing current density was set to 0.2 A/m2 in order to have enough
to the cover thickness is more pronounced. Therefore, in the fol-
oxygen available around the surface of the reinforcement. The re-
lowing investigations, S/P value and concrete cover thickness are
sults of simulations are presented in Figs. 9 and 10. As illustrated
selected as 0.5 and 50 mm, respectively.
in Figs. 9 and 10a, the cover thickness does not influence the peak
current density and the distribution of corrosion rate in the ring-
anode zone. However, as the cover thickness becomes larger a 3.3. Effect of concrete resistivity
slight rise of corrosion current density along the rebar in the sub-
strate due to the macrocell corrosion can be observed. As illus- In order to investigate the effect of resistivity on the distribu-
trated in Fig. 10b, the area under the corrosion current density tion of corrosion current density along the rebar in concrete after
366 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368

Substrate Patch repair 100 X m, medium resistivity: 200 X m, and high resistivity:
500 X m) were simulated. The rebar cover thickness of 50 mm
and the S/P ratio of 0.5 were used for each case. The limiting cur-
rent density (iL) is set to 0.2 A/m2 in order to eliminate the effect
Current density(A/m2)

0.08 of concentration polarization that may be more obvious in satu-

(a) Patch resistivity=500 Ω.m
rated concrete in which there would be limited oxygen availability.
For each case, three different types of patch repairs with resistivi-
ties lower, equal, and higher than those of the substrate are exam-
0.04
ined. In Table 2, the values of the patch resistivities (qpatch) for each
case are presented. The results of the simulations are presented in
Fig. 11.
In general, as illustrated in Fig. 11, it is obvious that substrate
0
with low resistivity experience the most significant ring-anode ef-
fect, regardless of the resistivity of the patch used in the repair. For
Current density(A/m2)

0.08
(b) Patch resistivity=200 Ω.m example, for cases with patch resistivity of 200 X m, the peak cur-
rent density in the ring-anode zone decreases from 0.1070 A/m2 for
a substrate with resistivity of 100 X m to 0.0981 A/m2 for a sub-
strate with resistivity of 500 X m. In each case, the lowest peak
0.04 corrosion current density in the ring-anode zone was obtained
when a patch with resistivity higher than that of the substrate
was used. As illustrated in Fig. 12, the total corrosion current in
the repaired member decreases as the patch resistivity increases.
0 In other words, if concrete resistivity can be used as an indicator
0.12 of the quality of concrete and the patch (assuming higher quality
(c) Patch resistivity=100 Ω.m may be correlated to higher density and higher electrical resistiv-
Current density(A/m2)

ity), using a higher quality patch (i.e., with higher resistivity that
0.08 of the substrate) corresponds to a better repair strategy. Therefore
Substrate resistivity= 200Ω.m, S/P=0.5
Cover thickness=100 mm
based on the results obtained in this numerical investigation, it can
Cover thickness=50 mm be stated that using a patch with higher resistivity (i.e., quality)
0.04 Cover thickness=25 mm with respect to the substrate concrete diminishes the ring-anode
Corrosion before repair effect. However, if the quality of the substrate and the patch are
significantly different from each other, it is also possible that there
may be mechanical compatibility problems at the interface be-
0
tween the patch and the substrate (e.g., bond failure between
0 0.2 0.4 0.6 0.8 1
two surfaces), which may, in long term, cause cracking and in-
Normalized Length crease the rate of deterioration of the patched zone. It should also
Fig. 9. Effect of cover thickness on the corrosion rate of rebar after the patch repair be pointed out that if the resistivity of the substrate is low, regard-
(iL = 0.2 A/m2). less of the patch quality, the ring-anode effect will be significant;
therefore, for these cases, the best solution is the replacement of
the entire chloride-contaminated substrate concrete with a high
2 quality concrete with high resistance to chloride penetration.
ipeak/icor,mic

1.6
Patch resisitivity= 50 Ω.m
3.4. Effect of oxygen availability
1.2 Patch resisitivity= 100 Ω.m
Patch resisitivity= 200 Ω.m
0.8 The amount of oxygen available on the surface of rebar affects
0.4 the limiting current density through Eq. (3) and affects the concen-
20 40 60 80 100 tration polarization characteristics. Therefore the oxygen concen-
Cover thickness (mm) tration is an important parameter that should be considered in
(a) the study of corrosion kinetics in patch repairs. For this purpose,
a series of patch repair cases were simulated such that the limiting
Total corrosion current (A)

current density was the variable of the investigation. The results of

9.8E-5 this investigation are illustrated in Figs. 13 and 14.
9.6E-5 In Fig. 13, the corrosion current density of reinforcing steel after
9.4E-5 Patch resisitivity= 50 Ω.m
a patch repair is plotted for the different limiting current densities,
9.2E-5 Patch resisitivity= 100 Ω.m
Patch resisitivity= 200 Ω.m
varying from 0.008 to 0.6 A/m2. In the figure, the corresponding
9E-5
8.8E-5 microcell corrosion current densities before the repairs are also
8.6E-5 plotted. The common expectation is that the decrease of the limit-
20 40 60 80 100 ing current density (i.e., lower oxygen concentrations around the
Cover thickness (mm) reinforcement) would reduce both microcell and macrocell corro-
sion rates; hence the ring-anode effect would be diminished. It
(b)
has been observed in the simulations that even at very low values
Fig. 10. Effect of cover thickness on (a) ipeak to icor,mic ratio (b) total corrosion current of limiting current density, the ring anode can still be observed;
(I) (substrate resistivity = 100 X m, S/P = 0.5, iL = 0.2 A/m2). however its damaging effect is less severe when the limiting cur-
rent density is low.
a patch repair, three corroding concrete structures with three dif- The ratio of peak corrosion current density, ipeak, to microcell
ferent resistivities of concrete (qsubstrate) (low resistivity: corrosion density before the repair, icor,mic, versus limiting current
 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368 367

Table 2 densities is plotted in Fig. 14 for the different patch resistivities. As

Three different cases for investigating the effect of patch resistivity. illustrated in this figure, the ratio increases by decreasing the lim-
Cover Limiting current Old concrete Patch iting current density for all S/P ratios. For the limiting current den-
thickness density (A/m2) resistivity resistivity sity less than 0.08 A/m2, the rate of increase is significantly large;
(mm) (X m) (X m) however, the change is not significant for the limiting current den-
Case 1 50 0.2 100 50, 100, 200 sities beyond this value. Additionally, it can be observed that the
Case 2 50 0.2 200 100, 200, patch resistivity also plays an important role; the higher resistivity
500
Case 3 50 0.2 500 200, 500,
of patch results in the lower magnitude of peak corrosion current
1000 density. As concluded before, the S/P ratio does not remarkably af-
fect the corrosion kinetics in the ring anode.

Node number
0 20 40 60 0.12

0.12 Substrate Patch repair Combined microcell and macrocell

Current density (A/m2)

corrosion after repair

iL = 0.008 A/m 2
0.09 iL = 0.08 A/m 2

Current density (A/m2)

0.08 iL = 0.2 A/m 2
Substrate resistivity=100 Ω.m iL = 0.6 A/m 2
0.06 Patch resistivity=50 Ωm
Patch resistivity=100 Ωm
(a) Case 1 Patch resistivity=200 Ωm Peak=0.1070 A/m 2
0.03 Microcell corrosion before repair
iL = 0.008 A/m2
0.04 iL = 0.08 A/m2
0
0.12 iL = 0.2 A/m2
iL = 0.6 A/m2
Current density (A/m2)

0.09

Substrate resistivity=200 Ω.m 0

0.06 Patch resistivity=100 Ωm
(b) Case 2 Patch resistivity=200 Ωm Peak=0.1040 A/m 2 0 0.2 0.4 0.6 0.8 1
Patch resistivity=500 Ωm
0.03 Normalized Length

Fig. 13. Effect of iL on the corrosion rate of rebar due to the patch repair (substrate
0 resistivity = 100 X m, patch resistivity = 200 X m, cover thickness = 50 mm, S/
0.12 P = 0.5).
Current density (A/m2)

0.09

Substrate resistivity=500 Ω.m

0.06 (c) Case 3 Patch resistivity=200 Ωm Peak=0.0981 A/m 2
6 6
Patch resistivity=500 Ωm
5 5
Patch resistivity=1000 Ωm
ipeak/ icor,mic

(a) S/P= 0.5 (b) S/P= 1.0

ipeak/ icor,mic

0.03
4 4

0 3 3
0 0.2 0.4 0.6 0.8 1
2 2
Normalized Length

Fig. 11. Effect of patch resistivity on the corrosion of rebar in concrete for three
1 1
different hypothetical cases (cover thickness = 50 mm, S/P = 0.5, iL = 0.2 A/m2). 0 0.2 0.4 0.6 0 0.2 0.4 0.6
iL(A/m2) iL(A/m2)

2.0x10-5
Total macrocell current, Imac (A)

6
-5 Substrate resisitivity= 500 Ω.m
1.6x10 5 Substrate resistivity=100 Ω m
Substrate resisitivity= 200 Ω.m (c) S/P= 2.0
ipeak/ icor,mic

Cover thickness= 50 mm
1.2x10 -5
Substrate resisitivity= 100 Ω.m 4 Patch Resistivity=50 Ω m
Patch Resistivity=100 Ω m
8.0x10
-6 3 Patch Resistivity=200 Ω m

-6 2
4.0x10
1
0.0x100
0 200 400 600 800 1000 0 0.2 0.4 0.6
iL(A/m2)
Patch resistivity (Ω.m)
Fig. 14. Effect of iL on the ratio of ipeak to icor,mic for the different patch resistivities:
Fig. 12. Effect of patch resistivity on the total macrocell corrosion current. (a) S/P = 0.5, (b) S/P = 1, (c) S/P = 2.
368 S. Soleimani et al. / Cement & Concrete Composites 32 (2010) 360–368

4. Conclusions References

A comprehensive numerical investigation on the kinetics of cor- [1] RILEM Draft Recommendation. Repair strategies for concrete structures
damaged by reinforcement corrosion. Mater Struct 1994;27:415–36.
rosion after patch repairs of reinforced concrete structures have [2] Wheat HG, Harding KS. Galvanic corrosion in repaired reinforced concrete
been carried out. The outcome of the study revealed the following slabs – an update. Mater Select Des 1993;5:58–62.
conclusions: [3] Schiebl P, Breit W. Local repair measures at concrete structures damaged by
reinforcement corrosion – aspect of durability. In: International symposium on
corrosion of reinforcement in concrete construction. Royal Society of
 Numerical simulations carried out in this study have clearly Chemistry, SP, vol. 183; 1996. p. 327–36.
demonstrated the formation of a ring-anode zone in the [4] Pruckner F, Gj/rv OE. Patch repair and macrocell activity in concrete
structures. ACI Mater J 2002;99:143–8.
substrate, supporting the experimental observations made by [5] Castro P, Pazini E, Andrade C, Alonso C. Macrocell activity in slightly chloride-
Barkey [15]. contaminated concrete induced by reinforcement primers. Corrosion
 The size of the ring anode observed in the simulations, which 2003;59:535–46.
[6] Li G, Yuan YS. Electrochemical incompatibility for patch-repaired corroded
spans a 2–5 cm portion of the steel reinforcement from the
reinforced concrete. J China Univ Min Technol 2003;32:44–7.
intersection between the repaired zone and the substrate, is in [7] Cleland DJ, Yeoh KM, Long AE. Corrosion of reinforcement in concrete repair.
agreement with the experimental and in situ observations. Constr Build Mater 1997;11:233–8.
[8] Kasselouri V, Kouloumbi N, Thomopoulos TH. Performance of silica fume–
 The anodic current density in the ring-anode zone has a peak
calcium hydroxide mixture as a repair material. Cem Concr Compos
near the intersection of the substrate and the repaired concrete 2001;23:103–10.
and asymptotically approaches to the microcell corrosion den- [9] Emberson NK, Mays GC. Significance of properties mismatch in the patch
sity observed before the patch repair. repair of structural concrete, part 1: properties of repair systems. Mag Concr
Res 1990;42:147–60.
 Barkey’s observation of the shift of the ring anode into the sub- [10] Raupach M, Schiebl P. Macrocell sensor systems for monitoring of the
strate can be attributed to the movement of the chlorides from corrosion risk of the reinforcement in concrete structures. NDTE Int
the interface into the substrate, hence creating a narrow 1– 2001;34(6):435–42.
[11] Gu P, Beaudoin JJ, Tumidajski PJ, Mailvaganam NP. Electrochemical
2 cm zone from the interface at which the chloride threshold incompatibility of patches in reinforced concrete. Concr Int 1997;19:68–72.
is not exceeded. However, as demonstrated in the current study, [12] Emmons PH, Vaysburd AM. Corrosion protection in concrete repair: myth and
the movement of chlorides is not a requirement for the forma- reality. Concr Int 1997;19:47–56.
[13] Mailvaganam NP. Concrete repair and rehabilitation: issues and trends. The
tion of the ring-anode zone. The ring-anode zone form mainly Indian Concr J 2001;75:759–64.
due to the combined macrocell and microcell activity after the [14] Zhang J, Mailvaganam N. Corrosion characteristics and key electrochemical
patch repair. factors in patch repair. Can J Civ Eng 2006;33(6):785–93.
[15] Barkey DP. Corrosion of steel reinforcement in concrete adjacent to surface
 The ring-anode problem is a much localized problem that is gov-
repairs. ACI Mater J 2004;101:266–72.
erned by the kinetics of corrosion around the interface between [16] Raupach M. Chloride-induced macrocell corrosion of steel in concrete –
the substrate and the repaired zone. The size of the patch or the theoretical background and practical consequences. Constr Build Mater
1996;10(5):329–38.
thickness of the concrete cover is not a significant factor that
[17] Jones DA. Principles and prevention of corrosion. Prentice Hall; 1995.
positively or negatively affects the ring-anode characteristics. [18] Bohni H. Corrosion in reinforced concrete structures. New York: CRC Press;
 The substrate with low resistivity experience the worst ring- 2005.
anode effect, regardless of the resistivity of the patch used in [19] Gulikers J. Theoretical considerations on the supposed linear relationship
between concrete resistivity and corrosion rate of steel reinforcement. Mater
the repair. Corros 2005;56:393–403.
 If concrete resistivity can be used as an indicator of the quality of [20] Elsener B. Macrocell corrosion of steel in concrete – implications for corrosion
concrete and the patch (assuming higher quality may be corre- monitoring. Cem Concr Compos 2002;24(1):65–72.
[21] Revie RW, Uhlig HH. Corrosion and corrosion control. 4th ed. John Wiley &
lated to higher density and higher resistivity), using a higher Sons; 2008.
quality patch (i.e., with higher resistivity that of the substrate) [22] Ghods P, Isgor OB. Technical report. Carleton University, Ottawa, Canada; 2008.
corresponds to a better repair strategy. However, considering [23] ASTM G109-07. Standard test method for determining effects of chemical
admixtures on corrosion of embedded steel reinforcement in concrete exposed
the fact that the high quality patch may also cause mechanical to chloride environments. ASTM International, West Conshohocken, PA; 2007.
compatibility problems at the interface between the patch and [24] Qian S, Zhang J, Qu D. Theoretical and experimental study of microcell and
the substrate, choosing patches with similar properties as the macrocell corrosion in patch repairs of concrete structures. Cem Concr Compos
2006;28(8):685–95.
substrate is recommended.
[25] Ghods P, Isgor OB, Pour-Ghaz M. A practical method for calculating the
 If the resistivity of the substrate is low, regardless of the patch corrosion rate of uniformly depassivated reinforcing bars in concrete. Mater
quality, the ring-anode effect will be significant; therefore, for Corros 2007;58:265–72.
[26] Ge J. On the numerical solution of laplace’s equation with nonlinear boundary
these cases, the best solution is the replacement of the entire
condition for corrosion of steel in concrete. MASc Thesis, Carleton University,
chloride-contaminated substrate concrete with a high quality Ottawa, Canada; 2006.
concrete with high resistance to chloride penetration. [27] Pour-Ghaz M, Isgor OB, Ghods P. Virtual experiments to investigate steel
 At very low values of limiting current density (i.e., during peri- corrosion in concrete. In: Lucio Soibelman, Burcu Akinci, editors. ASCE
international workshop on computing in civil engineering, Pittsburgh (PA);
ods of oxygen deprivation), the ring anode can still be observed; 2007. p. 542–9.
however its damaging effect is less severe when the limiting [28] Ge J, Isgor OB. Effects of Tafel slope, exchange current density and electrode
current density is low. potential on the corrosion of steel in concrete. Mater Corros 2007;58(8):573–82.