University of New Haven

Digital Commons @ New Haven

Civil Engineering Faculty Publications Civil Engineering

6-2014

Electrochemical-Mechanistic Model for Concrete

Cover Cracking Due to Corrosion Initiated by
Chloride Diffusion
Goli Nossoni
Manhattan College

Ronald S. Harichandran
University of New Haven, rharichandran@newhaven.edu

Follow this and additional works at: http://digitalcommons.newhaven.edu/civilengineering-facpubs

Part of the Civil Engineering Commons

Publisher Citation
Nossoni, G., Harichandran, R. S. (2014). “Electrochemical-mechanistic model for concrete cover cracking due to corrosion initiated
by chloride diffusion.” Journal of Materials in Civil Engineering, ASCE, 26(6). doi:10.1061/(ASCE)MT.1943-5533.0000470

Comments
This is the authors' accepted version of the article published in Journal of Materials in Civil Engineering. The version of record can be found in the
ASCE library at http://dx.doi.org/10.1061/(ASCE)MT.1943-5533.0000470
 Electrochemical-Mechanistic Model for Concrete Cover Cracking due to

Corrosion Initiated by Chloride Diffusion

G. Nossoni1 and R.S. Harichandran,2 F. ASCE

ABSTRACT

A holistic electrochemical-mechanistic model of the corrosion of steel reinforcing

bars inside concrete and the time to cracking of the concrete cover is presented. The

model accounts for the diffusion of oxygen and moisture into the concrete and rust layers,

the densification of rust due to confinement, the flow of rust into the concrete pores, the

development of internal pressure due to rust build-up, and cracking of the concrete cover.

The relationship between the corrosion current and the pressure build-up due to the

corrosion products for different concrete cover thicknesses was calibrated through

experiments using an accelerated corrosion test with an applied current. Results from

finite element analysis with an inelastic smeared crack concrete model were used to

calibrate a simple analytical model of the critical internal pressure required to cause

cracking of the concrete cover. The various sub-models are linked together to predict the

time for cracking of the concrete cover from the time of corrosion initiation. Results of

parametric studies using the model indicate that the main factors that control the

corrosion current and the time to cracking are the boundary condition, water/cement ratio

and concrete cover.

1
Assistant Professor, Department of Civil and Environmental Engineering, Manhattan College,
Riverdale NY 10471
2
Previously Prof. & Chair, Dept. of Civil and Envir. Engrg., Michigan State Univ., East Lansing
MI 48824-1226. Currently Dean, Tagliatela College of Engineering, Univ. of New Haven, West Haven, CT
06516.

1
Key word: Cathodic reaction, oxygen diffusion, limiting current, pressure loss, rust

thickness, concrete cover cracking

INTRODUCTION

Reinforced concrete is one of the most common construction materials. High quality

concrete is a very durable material and can remain maintenance free for many years when

properly designed. Steel reinforcement is used to improve the tensile strength of concrete

structures, and more generally their mechanical resistance. The high alkalinity of

concrete can prevent the reinforcing steel from corroding within it; in other words steel is

passive inside concrete. Depassivation of steel reinforcement embedded in concrete and

the onset of active corrosion can arise due to two causes: carbonation of the concrete, or

chloride diffusion. For pore solution which has a pH of about 13.5, corrosion will

commence when the [Cl−]/[OH−] ratio reaches a threshold value. For synthetic concrete

pore solution this ratio is about 0.6 (Mindess et al. 2003, Haussmann 1967), but the

threshold values reported in literature vary from 0.3 to above 3 depending on many

factors such as temperature, stress level, pitting potential, experimental conditions, and

experiment methods (Alson et al. 2002, Ann and Song 2007). The depassivation

condition and the time at which the chloride concentration reaches the threshold value

have been the subjects of many investigations (Truc et al. 1999, Gowripalan, et al. 2000,

Stanishn and Thomas 2001, Xi et al. 2000, Glass and Buenfield 1999) and the

depassivation time can be predicted with good accuracy.

The corrosion process is very complex and modeling is often based on observation or

speculation rather than a clear understanding of the physical and chemical processes

involved (Palle 2002, Du et al. 2006). Considerable experimental data and semi-empirical

models are available on the corrosion rate of steel in chloride-contaminated concrete (Liu

2
1996, Newhouse and Weyers 1996, Newhouse 1993). However, to our knowledge no

electrochemical-mechanistic model exists for the corrosion of steel reinforcement in

concrete due to chloride diffusion and the subsequent cracking of the concrete cover

Corrosion of steel in concrete consists of two stages. The first stage is controlled by

the anodic reaction and is very short despite having a high corrosion rate. In the second

stage the current is limited by the cathodic reaction. This is the corrosion current often

measured by researchers (Liu 1996). In this paper an electrochemical-mechanistic model

is developed for the corrosion rate during the second stage based on a fundamental

understanding of the electrochemical corrosion reaction and the effect of oxygen

concentration on the cathodic reaction. The corrosion rate is then used to predict the time

for cracking of the concrete cover due to corrosion. The cathodic control corrosion rate

has been proposed for modeling the corrosion rate in carbonated concrete but has not

been studied for corrosion due to chloride diffusion, although the corrosion reaction in

both cases is similar (Huet et al. 2006, Liang et al. 2004).

The rust thickness is calculated and used to estimate the pressure applied to the

concrete cover, accounting for the amount of rust products that flow into the concrete

pores. The calculation of rust thickness as a function of corrosion current has not been

studied extensively. An elasto-plastic concrete material model was used to calculate the

pressure developed due to thickening of the rust layer. The critical pressure that causes

the concrete cover to crack was calculated and calibrated with finite element analysis and

experimental data.

3
 CORROSION CURRENT MODELING

The corrosion reaction can be described by two elementary anodic and cathodic

reactions:

Fe ® Fe + + + 2e - (1)

O 2 + 2H 2 O + 4e - ® 4OH - (2)

When there is no limitation of oxygen in the corroding system, the corrosion current

will increase with increasing chloride concentration. However, even if oxygen is

available, the corrosion current will reach the anodic limiting current because the rate of

diffusion of corrosion products away from the anode is limited (Nossoni 2010). This

limiting current is very high and cannot be reached for the corrosion of steel inside

concrete. We assume that the oxygen reduction in the catholic reaction (Equation 2)

controls the corrosion rate. To determine the availability of oxygen for the cathodic

reduction reaction and for corrosion to continue we must determine the path for oxygen

diffusion to the corrosion pit.

Corrosion Scale

Knowledge of the composition and structure of the corrosion products in concrete is

necessary not only to estimate the corrosion expansion and pressure applied on the

concrete cover, but also to estimate the corrosion rate based on the limiting current,

diffusion of oxygen, and reduction of species inside and outside the pit. The structure of

the pit and corrosion products depends on the aqueous phase of the pore solution, steel

type, and pressure adjacent to the bar. In general, the composition of the expansive

corrosion products may be expressed as {a.Fe(OH)2+ b. Fe(OH)3+c.H2O} (Bhagava et al.

4
2005), where a, b and c are variables that depend on the alkalinity of the pore water

solution of the concrete, the oxygen supply, and the moisture content. For example, black

rust (Fe2O3) will form mostly in concrete immersed in deep water where the oxygen

availability is low (Bazant 1997).

The model adopted herein for the corrosion scale is taken from the literature (Sarin et

al. 2004, Jones 1992, Huet et al. 2004) and is shown in Figure 1. The corrosion scale

consists of three layers: the inside part (core layer), the outer part (top layer), and a hard

shell layer that separates these two layers (Sarin et al. 2004).

The core layer is generally a porous mass made up of small particles of different

component phases. Fe(II) is expected to dominate the core part of the scale due to a lack

of hydroxide ions. The wet outside region of the scale on top of the shell layer is the top

layer. Since the top layer is exposed to water and air in the concrete pores, the ferrous

hydroxide is rapidly oxidized to the ferric components Fe(OH)3 and Fe(OH)3+3H2O, and

consequently the concentration of Fe (III) is expected to be higher than in the core layer.

The top layer is saturated with the concrete pore solution. Around the outside of the top

layer particles are loosely held and can be easily transferred to the concrete pore water

under the pressure that is produced (Sarin et al. 2004).

Fe(II) products are more conductive than Fe(III) products and can transfer electrons

outside the core layer. We assume that the cathodic reaction takes place in the top layer

where both oxygen and electrons are available.

Corrosion Current

Corrosion of steel in an alkaline environment starts when the chloride concentration

reaches a certain threshold value. Most existing models treat corrosion as a steady state

5
phenomenon, although it is known that the corrosion rate varies with change in

temperature, chloride content, and corrosion level (Pantaopoulou et al. 2001, Bhargava et

al. 2005, El Maaddawy and Soudki 2007). No comprehensive model is available to take

all of these factors into account and only an electrochemical-mechanistic model that

accounts for most of these factors can predict the long-term behavior of concrete

accurately.

Initially, the rate of corrosion will depend on the chloride concentration and the rate

of the anodic reaction. Shortly thereafter, as the corrosion becomes more severe and the

rust layers build up, the corrosion rate will be controlled by the cathodic reaction and the

availability of oxygen at the cathode.

The limiting current at any time can be calculated through:

ilim = nO2 .F . jo2 ,used (3)

where jO2 ,used is the flux of the oxygen consumed at the cathode, nO2 is 4 for the reduction

of oxygen in the corrosion reaction, and F is Faraday’s constant.

Figure 2 shows a simplified model of the different stages of oxygen diffusion to the

anode. The flux of oxygen to the steel bar can be divided into three stages. In the first

stage oxygen diffuses through the concrete cover and will follow Fick’s Second Law

(Castellote et al. 2001):

¶CO2 ¶ 2 CO2
= D0 (4)
¶t ¶x 2

where CO2 is the oxygen concentration, and DO is the coefficient for diffusion of oxygen

into the concrete cover. Only the dissolved oxygen can be used in the cathodic reaction. It

6
is assumed that the dissolved oxygen is in thermodynamic equilibrium with the oxygen

partial pressure in the gaseous phase, and according to Henry’s Law

RT
CO2 ,l ,in = CO 2, g ,in (5)
KH

where CO2 ,l ,in is the dissolved oxygen concentration, C O2 , g ,in is the concentration of

oxygen gas in the concrete adjacent to the bar, T is the temperature in degrees Kelvin, and

KH is a constant. The degree of concrete saturation is in equilibrium with the outside

relative humidity. The semi-empirical equations proposed by Bazant and Thongutha

(1978, 1979) were used to calculate the water content of the concrete.

CO2 ,l ,in diffuses across the top layer of the rust scale which is assumed to be fully

saturated. The top layer consumes some of the oxygen in the secondary non-

electrochemical reaction, in which Fe(II) is oxidized to Fe(III):

4Fe(OH) 2 + H 2 O + O 2 ¾
¾® 4Fe(OH) 3 (6)

1
¾® 2g - 4FeOOH + H 2 O
4Fe(OH) 2 + O 2 ¾ (7)
2

Assuming that the top layer is a homogeneous porous medium, Fick’s First Law can be

used to predict the flux of the diffused oxygen through the rust layer:

¶CO
2
jO2 ,used = ( Do,r ) (8)
¶x

where DO,r is the coefficient for diffusion of oxygen into the rust layer, and can be

calculated from the general equation for diffusion in porous media (Martys 1994).

7
 The limiting current occurs when CO2 at the bar lever is zero, and all the diffused

oxygen is consumed. The limiting oxygen flux can be estimated from

¶C CO2 ,l ,in - CO2 ,consumed

= (9)
¶x d2

where d2 is the top rust layer thickness.

The final limiting current will be

CO2 ,l ,in - CO2 ,consumed

ilim it = nO2 . f .Do,r .( ) (10)
d2

where C O2 ,consumed is the loss in oxygen concentration due to consumption of oxygen for

the secondary reaction presented in Equation 6 in the top layer.

According to Faraday’s Law, the flux of the Fe2+ produced will be

icorr
j Fe(OH ) = (11)
2
n Fe .F

Assuming that Equation 6 dominates the secondary reaction:

j Fe(OH ) = k r J Fe(OH ) 2 (12)

3

kr is the fraction of Fe(OH)2 that is transformed to Fe(OH)3 in the secondary reaction and

needs to be determined experimentally or obtained from the literature. kr can be assumed

to be 1.0 in air since Fe(OH)2 is not a stable product when sufficient oxygen is available,

and for corrosion of steel in concrete it can vary from 1.0 to less than 0.5 depending on

the availability of oxygen. A parametric study was done to evaluate the effect of this

factor on the limiting current. In this paper kr is assumed to be 0.5. This value does not

8
affect the results considerably since the rust thickness was calibrated using experimental

test data.

CO2 ,consumed can be calculated from Equation 12 since the flux of the consumed

oxygen is four times smaller than the flux of the Fe(OH)3 produced (see Equation 6).

Thus,

dC O2 ,consumed j Fe(OH )3
= (13)
dt 4.d 2

Assuming that the density of the porous core and top layer is a function of pressure

and 3.75 and 4.20 times smaller than the density of steel, respectively (Bhagava et al.

2005) the rate of build-up of corrosion products is:

dd1 j Fe(OH )2 . AFe(OH )2 3.75( j Fe(OH )2 - j Fe(OH )3 ).AFe(OH )2

= = (14)
dt r Fe(OH )2 r Fe .K P

dd 2 j Fe(OH )3 . AFe(OH )3 4.20 j Fe(OH )2 . AFe(OH )3

= = (15)
dt r Fe(OH )3 r Fe .K P

where rFe, rFe(OH)2 and rFe(OH)3 are the densities of iron, Fe(OH)2 and Fe(OH)3,

respectively. KP is a correction factor that accounts for pressure dependency of the

density of the corrosion products. KP was calibrated using experimental test result. KP is

an important parameter that affects the pressure applied to the concrete cover, the rust

thickness that affects the oxygen diffusion (Figure 1), and the diffusion coefficient of

oxygen in the rust.

9
 MECHANISTIC MODEL OF CONCRETE COVER CRACKING

Development of Pressure

Concrete surrounding the rebar will be subjected to an internal radial pressure due to

the expansive corrosion products modeled using Equations 14 and 15. It is assumed that

the corrosion occurs uniformly around the steel bar. The radial strain is:

du u
er = +n . (16)
dr r
The rates of radial and circumferential strain due to thickening of the rust layer are:

de r d (d 1 + (d 2 - d 0 ) - d 3 ) de
= +n . q (17)
dt dt .d dt

de q d (d1 + (d 2 - d 0 ) - d 3 )
= 2. (18)
dt d bar .dt

where d3 is the reduction in diameter of the steel bar due to corrosion and can be

calculated using Faraday’s Law, d0 is the loss in rust layer thickness due to diffusion of

rust products into the concrete, and d1 and d2 are the thicknesses of the porous core and

top layer of rust. The rate of reduction in bar diameter is given by

dd 3 j
= Fe (19)
dt r Fe

where jFe is the flux of Fe into the system and rFe is the density of iron.

The rate of pressure growth at the steel surface due to thickening of the rust layer

was calculated using the nonlinear concrete constitutive relation given by (MacGrego

1997):

10
 eq
2 Pcri ( )
e0
Pnet = (20)
e
1 + ( q )2
e0
where Pcri is the critical pressure that causes the concrete to crack. Pnet is net pressure on

the concrete cover from corrosion products, eq is the strain obtained from Equation 18,

and e0 can be calculated using

1.75. ft
ε0 = (21)
Econ
The accuracy of Equation 20 was verified using finite element analysis with a nonlinear

concrete material model (Caré et al. 2008).

The pressure is assumed to be uniform around the bar. The pressure will cause some

amount of corrosion products (mainly red rust in the top layer) to flow into the concrete

pores surrounding the steel rebar before the concrete cracks. Some researchers assumed a

“gap” of 10 to 20 µm around the steel bar to account for the slower build-up of pressure

(Du et al. 2006, El Maaddawy and Soudki 2007, El Maaddawy et al. 2006), and the size

of this “gap” was calibrated based on experimental test results. We explicitly model the

flow of rust products into the concrete pores.

Calculation of d0

The amount of accumulate pressure around the bar will decrease due to the flow of

rust products from the top layer into the concrete pores, thereby increasing the time for

the concrete to crack.

Conservation of mass yields:

11
 ¶w
( j Fe(OH )3 + dj Fe(OH )3 - j Fe(OH )3 ) A = (Vt ) (22)
¶t

where A is the cross sectional area of corrosion pit, jFe(OH)3 is the flux of the top layer

products in concrete pores, w is the mass of Fe(OH)3 per volume of concrete, and Vt is the

total volume of concrete. Expressing the saturation of rust as s = w/n, where n is the

porosity, and since Vt = Adx, Equation 22 can be expressed as

dj Fe(OH )3 ¶s
=n (23)
dx ¶t

Assuming that the flow of rust suspension is similar to the flow of water, Darcy’s

equation can be used to relate the pressure to flow:

d æ dp ö ¶s
ç kc ÷=n (24)
dx è dx ø ¶t

where p is the pore pressure and kc is the conductivity of the colloidal solution of

Fe(OH)3 inside the concrete.

The Van Genuchten equation developed for the flow of liquid and gas in porous

media relates pore pressure and saturation (Monlouis et al. 2007, Van Genuchten 1980):

1
p = M ( s m - 1)1-m (25)

where M and m are empirical parameters for concrete (Coussy et al. 2004). The

conductivity of the colloidal suspension of rust products in water can be related to the

conductivity of water through:

Kk
kc = (26)
hr

12
where Kk is conductivity and hr is the density of water relative to the density of colloidal

suspension of rust products in water.

Equations 24, 25 and 26 constitute the model for the flow of rust products into the

concrete pores. The internal pressure can be computed through an incremental time step

procedure. After corrosion starts, the pressure begins to accumulate. This pressure pushes

some amount of the corrosion products produced in the top layer (mostly red rust) into

the concrete pores and this reduces the pressure around the bar. When the pressure of the

rust suspension inside the concrete pores exceeds the tensile strength of the concrete, the

concrete will crack.

Critical Pressure

Cracking of the concrete cover is highly dependent on the ratio of the concrete cover

to reinforcing bar diameter defined as

k = D/d (27)

where D is the diameter of the concrete cylinder and d is the diameter of the steel bar.

An empirical model from the literature was adopted in this study and validated and

calibrated using finite element analysis performed with the ABAQUS commercial

software. According to the empirical model, for k < 5 fracture occurs when the mean

value of the circumferential stress reaches the tensile strength of the concrete:

st = sm (28)

Equation 28 implies that cracking occurs when the net pressure on the concrete cover

is

13
 pin - pout = (k - 1).s m (29)

where pin is the pressure that the rust layer applies to the concrete cover and pout is the

pressure lost due to flow of rust from the top layer into concrete pores. However, for k >

5 the first crack opens before the circumferential stress reaches the tensile strength of the

concrete; i.e., when

s t = as m (30)

p in - pout = (ak ).s m (31)

where a is a factor established in the literature through experiments and sm is the

splitting tensile strength of concrete. In this study, finite element analysis was used to

calibrate the a factor and yielded a = 0.72, which compares well with the value of a =

0.8 reported by Shah and Swart (1987).

Finite Element Analysis

Since corrosion causes radial expansion, a 2-D plane strain model can be used to

study the mechanical behavior of concrete. Three different k ratios of 3, 4 and 6 were

used in the model to match the k ratios of experimental specimens described later. Figure

3 shows the geometry of the FE models. Due to symmetry of the model, only half of each

specimen was modeled and appropriate boundary conditions were applied. The concrete

damage plasticity model was used to model concrete (ABAQUS 6.4). When the rebars

corrode, the corroded steel swells to about 2 to 6 times of its initial volume (Mehta 1993).

The expansion of the rust products will apply internal pressure to the concrete cover.

This expansion causes the concrete cover to crack and eventually spall. The objective of

the FE analysis was to find the pressure-displacement relationship and calibrate the a

14
coefficient in Equations 30 and 31. The reinforcing steel was replaced by a uniform

radially outward pressure that was applied in increments over time steps until the

concrete cover cracked. Since the ABAQUS concrete plasticity model cannot account for

the opening of a large crack, the analysis terminates when the crack opens. The amount

of pressure at this time step was considered to be the critical pressure, and the a factor

was calibrated using this pressure. Figure 3 shows the rust thickens and applied pressure

for both model and FEM for a = 0.72.

Calibration of the Model

The factor KP that accounts for densification of rust with pressure needs to be

calibrated experimentally. To calibrate KP, the model was used to predict the rust

thickness and cracking time for accelerated corrosion experiments using the applied

constant current density of 0.0045 A/cm2. The goal was to establish a single

representative value of KP for different k values using experimental data of the time to

cracking. After calibration, the rust thicknesses obtained from the experimental test and

the model were compared to establish the accuracy of the model.

EXPERIMENTAL INVESTIGATION

Experiments were conducted using an accelerated corrosion test with an applied

current to measure the growth of rust thickness. The test was conducted in an

environment in which all the applied current was consumed in corroding the steel bar

(Nossoni et al. 2011). The rust thickness was measured under the microscope using

photographs taken at different time intervals. The time to cracking was obtained as a

function of the measured rust thickness.

15
Specimen Preparation

A normal concrete mix with Portland cement Type I, a maximum coarse aggregate

size of 9.5 mm (3/8²), and a water-cement ratio of 0.41 was used to prepare the

specimens. The maximum aggregate size was chosen to be small due to the small

dimensions of the specimens. No chloride was added to the mix. The concrete had a 28-

day compressive strength of 57.2 MPa (8,300 psi). The specimens were cast in small

cylindrical shapes having a thickness of 25 mm (1-inch) with a #4 carbon steel bar in the

center to serve as an anode. The diameter of the cylinder was chosen to yield k ratios of 3,

4, and 6. Figure 4 shows the schematic of the test specimens.

Three #1 carbon steel wires to be used as cathodes were placed equi-distant from

each other and the anode, close to the outer edge of each specimen. For each specimen

size, tests were conducted on triplicates. The goal of the experiments was to obtain a

single representative KP value for all specimen sizes, and the ratio of the maximum

aggregate size to concrete cover being large was not expected to introduce much error in

the final calibration results.

A certain amount of chloride ions is needed even in the accelerated corrosion test so

that all the applied current is consumed in oxidizing the steel bar (Caré et al. 2008,

Nossoni et al. 2011, Austin et al. 2008). Therefore, before testing the specimens were

soaked in 3.5% NaCl by weight for one week for the smallest specimens having a

concrete cover of 10 mm, and more than one month for the largest specimens. This

allowed the chloride concentration at the bar surface to be sufficient so that all the

applied current oxidized the steel. The specimens were immersed in the salt solution after

28 days of curing.

16
 Before testing, each specimen’s top surface was polished with grade 400 sand paper

and then cleaned with methanol to expose the steel bars (anode and cathodes). The

bottom surface was coated with epoxy to prevent the corrosion products from escaping

out of the concrete. Also, the surfaces of the steel bars that were not embedded inside the

concrete and were needed for the electrical connection were coated with epoxy to prevent

any corrosion in this part and to prevent any variation in current density due to change in

electrode surface area.

Accelerated Corrosion Test and Test Set up

The accelerated corrosion test consisted of applying a current so that the reinforcing

bar served as an anode and corroded in a short time. Figure 4 shows the test set up for the

accelerated corrosion test under the microscope. A power supply was used to apply a

constant direct current between the anode and cathodes. The applied current was limited

to 0.05 A yielding a current density of 0.0045 A/cm2 at the anode. The high current

density was not expected to be problematic because the test environment assured that all

the applied current would be consumed in corroding the steel bar (Nossoni et al. 2011).

An optical microscope with a magnification of 25 to 100 was used to capture images

of the concrete and steel interface to measure the rust layer thickness. After immersion in

salt water, the surfaces of specimens were dried and the test was conducted in a dry

condition. The dry condition prevented any escape of rust products outside the concrete

during the test. Initial images were taken before the test began. When the current was

applied, images were captured at 15 to 30 minute intervals depending on the size of

specimens.

17
 Before capturing images, the test was stopped temporarily, and each specimen’s

surface was polished again with sand paper and cleaned with methanol. Methanol was

used to minimize damage to the rust layer during the polishing stage since it has minimal

reaction with rust. The test was continued until cracking initiated in the concrete and the

rust thickness was measured. Although the main purpose of the test was to estimate the

rust thickness and time to concrete cracking as a function of the applied current, the test

was continued even after the concrete cracked.

Measurement of Rust Thickness

Observation of specimens subjected to the accelerated corrosion test under the

optical microscope allowed measurement of the rate of rust thickness build-up as a

function of applied current and time. Figure 6 shows that during the first 15 minutes of

the test no identifiable rust layer was observed. It is likely that during this period the rust

mostly diffused into the micro-pores around the steel bar.

After about 30 minutes the rust layer became visible, mostly around the areas where

some rust had already accumulated in the pores. The rust thickness was not uniform

around the steel bar and the areas surrounding the micro-pores had a thicker rust layer

than other areas.

Figure 7 shows that the average rust thickness had an approximately linear

relationship with time in this experiment since the corrosion current was constant, but this

may not be the case in real corrosion where the corrosion current may vary with time. For

the specimens with k = 3, it took 45 to 60 minutes for the first crack to appear, at which

time the average rust thickness around the steel bar was between 0 0.012 and 0.0142 mm

(0.00046 and 0.00056 in.). For specimens with k = 4 it took between 65 to 80 minutes for

18
the first crack to appear, at which time the average rust thickness around the bar was

between 0.014 and 0.021 mm (0.00057 and 0.00082 in.). For specimens with k = 6 it

took between 95 to 125 minutes for the first crack to appear, at which time the average

rust thickness was between 0.023 and 0.033 mm (0.0009 and 0.0013 in.). As shown in

Figure 7, using Kp = 1.56 in the model yields predictions of the cracking time that match

well with the cracking times observed in the experiments for all three k values.

PARAMETRIC STUDIES

After calibrating the corrosion model using the experimental test results, the model

was used to calculate the corrosion current and the time to cracking of the concrete cover

for typical structural elements. The effect of different boundary conditions, concrete

cover thicknesses, and concrete quality (represented by the water/cement ratio) were

assessed.

The chloride on the concrete surface can originate from various sources. Bridge

structural components can be categorized as superstructure components (boundary

condition type 1), splash zone components (boundary condition type 2), and submerged

components (boundary condition type 3). For submerged components (BC3), there is a

constant supply of water and chloride to the surface of the concrete. However, corrosion

of steel is often limited by the lack of oxygen dissolved in the water. Components in the

splash zone (BC2) have a good supply of both oxygen and chloride, and these

components experience the greatest corrosion. For superstructure components (BC1), the

chloride supply can limit the initiation of corrosion since it is supplied mainly by deicing

salts.

19
 Three different concrete cover thicknesses of 25, 38, and 70 mm, and water cement

ratios of 0.4, 0.5, and 0.6 were used. Table 1 shows typical values of the oxygen and

chloride concentrations for each boundary condition. The oxygen and chloride diffusion

coefficients are highly dependent on concrete quality and can increase by orders of

magnitude from good quality concrete to poor quality concrete (Neville 2000).

The concrete cover and bar diameters used are shown in Table 2. These were chosen

to yield the same k ratios as in the experimental specimens, but with dimensions close to

those of full size structural components.

Effect of Boundary Conditions

Figures 8, 9 and 10 show the calculated corrosion current for the first 200 days

after corrosion starts. The corrosion current decreases as the rust thickness increases. The

decrease in the corrosion current continues over the duration of corrosion until cracking

of the concrete.

The corrosion rates for boundary condition types 1 and 2 (BC1 and BC2) are quite

similar. In the model, the available oxygen controls the corrosion rate and in both cases

the oxygen availability is the same as shown in Table 1. Nevertheless, the corrosion

current is slightly higher for BC2 compared to BC1 due to the greater availability of

water in the concrete pores for BC2. In the model, only dissolved oxygen in water can

contribute to the corrosion process (see Equation 10). The concrete water content

resulting from equilibrium with the outside humidity (Bazant and Thonguthai 1978,

1979) is lower for BC1 than for BC2 and therefore the corrosion current is

correspondingly lower. For comparable concrete cover the corrosion current is almost 5%

higher in BC2 compared to BC1. The time to cracking of the concrete cover for BC1 and

20
BC2 is shown in Figure 11. The 5% increase in corrosion current from BC1 to BC2

decreases the time to cracking by 10% to 25%. For BC3 (substructure submerged in

water), the corrosion current is an order of magnitude smaller than for BC1 and BC2

because the oxygen availability is very limited. This can increase the life of the structure

by more than an order of magnitude. The cracking time for BC3 was not calculated since

it was computationally very expensive and after more than 25 years the concrete cover

did not crack even for the smallest k = 3.

Effect of Concrete Quality

The concrete quality is controlled mainly by the w/c ratio. Concrete with a high w/c

ratio has high porosity and diffusion of oxygen through the concrete cover is highly

dependent on the concrete porosity (Neville 2000). As Figure 8, 9 and 10 show,

decreasing the w/c ratio from 0.6 to 0.5 reduces the corrosion current by about 45%, and

decreasing the w/c ratio from 0.5 to 0.4 reduces it again by about another 45% for all the

cases. Also decreasing the w/c ratio by 0.1 increases the concrete strength by almost

6,895 to 10,342 kPa (1000 to 1500 psi) (Neville 2000). As shown in Figure 11,

decreasing the w/c ratio by 0.1 increases the cracking time by 1.5 to 3 times (depending

on other factors) due to the combination of increased concrete strength and decreased

porosity.

Effect of Concrete Cover

Although the concrete cover has a significant effect on the time for depassivation and

initiation of corrosion, it does not seem to have a significant effect on the corrosion rate.

The corrosion rate is about the same for all three k ratios considered.

21
 After concrete quality, the next most important factor that affects the cracking time

of the concrete cover is the k factor. A larger cover increases the time for chloride

diffusion. However, a larger k value provides more mechanical resistance to cracking and

thereby significantly increases the cracking time as shown in Figure 12. Increasing k from

3 to 4 approximately doubles the cracking time, and increasing it from 4 to 6

approximately doubles it yet again.

Rust Thickness Growth

Several models such as linear, parabolic, and more complicated power laws have

been proposed for the growth of rust with time (Tomoshov 1966) and may be expressed

in the general form:

y n = kt (32)

where y is the rust thickness, t is time, and n is an empirical constant.

Here we present results of the rust thickness growth over time based on the holistic

electrochemical-mechanistic model presented herein and compare them to the empirical

model in Equation 32.

Rust Build-Up

Figure 13 shows the results of the rust growth predicted by the electrochemical

model over time for different w/c ratios. In the model the rate of rust growth is

independent of the concrete cover and is affected only by the boundary condition and w/c

ratio.

22
 Although the electrochemical model is more complex than the simple empirical

equation, the end results are very similar as shown in Figure 13 if the parameters n and k

in Equation 32 are properly calibrated. However, the more complex model is based on

material and environmental characteristics and can be used in a variety of situations

without the need for calibration.

Rust Flow into Pores

One of the factors that affect the time to cracking of the concrete cover is the flow of

rust products into the concrete pores. The process of rust flow has not been effectively

modeled in the past. The rust products can only flow a certain distance from the steel bar

before cracking of the concrete. The calculated distance according to our model can be

between 2 to 15 mm, depending on many factors such as the w/c ratio and boundary

conditions.

The flow of rust products into the concrete pores decreases the amount of

accumulated pressure and increases the time to cracking. Figure 14 shows the reduction

in accumulated pressure for different cases. The loss in pressure is 1% to 2% for a w/c

ratio of 0.4, and 5% to 17% for the w/c ratios of 0.5 and 0.6. Concrete with a high w/c

ratio has higher porosity and therefore more rust can flow into it. The amount of rust that

flows into the surrounding concrete is highest for BC1 due to the lower degree of

saturation, and similar but lower for BC2 and BC3.

CONCLUSIONS

A holistic electrochemical-mechanistic model is developed for the corrosion of

reinforcing steel in concrete in the presence of chloride ions, and the time to cracking of

the concrete cover. The limiting corrosion current is modeled based on the cathodic

23
limiting reaction and oxygen diffusion. The model was calibrated using accelerated

corrosion testing under a microscope to establish the relationship between rust thickness

and the time to cracking of the concrete cover. In the constant current test, the

accumulated rust thickness varied approximately linearly with time until cracking of the

concrete cover.

The developed model was used to predict the corrosion current and the time to

cracking of the cover for three classes of concrete components in bridges. The corrosion

current was calculated for different boundary conditions, concrete properties and concrete

cover. The concrete cover did not have a significant effect on the corrosion current. The

two main factors that influence the corrosion current are the boundary conditions and the

concrete quality. When the w/c ratio is increased, the concrete quality decreases, the

permeability of concrete increases, the diffusion of oxygen and water into the concrete

increase, and consequently the corrosion current increases.

The time to cracking of the concrete cover is influenced by the corrosion rate, the

concrete quality and the cover. Increasing the w/c ratio by 0.1 decreases the time to

cracking by 1.5 to 3 times depending on other factors.

Concrete bridge components in the splash zone have the highest availability of water,

chloride and oxygen, corrode about 5% faster than superstructure components, and the

life of the former can be 10% to 25% less than the latter. Due to the lack of oxygen, the

corrosion rate in submerged components is an order of magnitude lower than components

exposed to air tremendously lengthening their life. Increasing the concrete cover from 38

to 51 mm (1.5 to 2 in.) can double the time to cracking. Increasing the cover further from

51 to 76 mm (2 to 3 in.) further doubles the time to cracking.

24
 Using good quality concrete with a low w/c ratio and the largest affordable cover are

the easiest ways to protect concrete structures from corrosion damage.

REFERENCES

Ann, K. Y., Song, H. W. (2007). “Chloride threshold level for corrosion of steel in

concrete.” Journal of Corrosion Science, 49, 4113–4133.

Alonso, C. M., Castrllote, C., and Andrade (2002). “Chloride threshold dependence of

pitting potential of reinforcements.” Journal of Electrochimica Acta, 47, 3469-3481.

Austin, S. A., Lyons, R., and lng, M. J. (2004). “Electrochemical behavior of steel-

reinforced concrete during accelerated corrosion testing.” Journal of Corrosion

Science, 60, 203–212.

Bhagava, K., Ghosh, A. K., Mori, Y., and Ramanujam (2005). “Modeling of time to

corrosion-induced cover cracking in reinforced concrete structures.” Journal of

Cement and Concrete Research, 35, 2203-2218.

Bazant, Z. P., Najjar, L. J. (1972). “Nonlinear water diffusion of nonsaturated concrete.”

Materiaux et Constructions, 5 (25).

Bazant, Z. P. (1979). “Physical model for steel corrosion in sea structure-application.”

Journal of Structural Division, ASCE, 105, 1155-1166.

Bazant, Z., and Thonguthai, W. (1979). “Pore pressure in heated concrete walls:

theoretical prediction.” Magazine of Concrete Research, 31(107), 67-76.

Bazant, Z., and Thonguthai, W. (1978). “Pore pressure and drying of concrete at high

temperature.” Journal of the Engineering Mechanics Division, ASCE, 104, 1058-

1080.

25
Castellote, M., Alonso, C., Andrade, C., Chadbourn, G. A., and Page, C. L. (2001).

“Oxygen and chloride diffusion in cement pastes as a validation of chloride diffusion

coefficients obtained by steady-state migration test.” Cement and Concrete Research,

31, 621-625.

Caré, S., Nguyen, Q. T., Hostis, V. L., and, Berthaud, Y. (2008). “Mechanical properties

of the rust layer induced by impressed current method in reinforced mortar.” Cement

& Concrete Research 38, 1079–1091.

Coussy, O., Dangla, P., Lassabatere, T., and Baroghel-Bouny, V. (2004). “The equivalent

pore pressure and the swelling and shrinkage of cement-based materials.” Materials

and Structures / Concrete Science and Engineering, 37, 15-20.

Du, Y. G., Chan, A. H. C., and Clark, L. A (2006). “Finite element analysis of the effects

of radial expansion of corroded reinforcement.” Computers and Structures, 84(13-

14), 917-929.

El Maaddawy, T., Chahrou, A., and Soudki, K. (2006). “Effect of fiber reinforced

polymer wraps on corrosion activity and concrete cracking in chloride-contaminated

concrete cylinders.” Journal of Composites for Construction, ASCE, 10(2), 139-147.

El Maaddawy T., and Soudki, K., (2007). “A model for prediction of time from corrosion

initiation to corrosion cracking.” Cement & Concrete Composites, 29, 168–175.

Glass, G. K., Buenfeld, N. R. (2000). “The influence of chloride binding on the chloride

induced corrosion risk in reinforced concrete.” Journal of Corrosion Science, 42,

329-344.

Hausmann, D. A. (1967). “Steel corrosion in concrete - how does it occur.” Journal of

Materials Protection, 6(11).

26
Huet, B., L’hostis, V. L., Santarini, G., Feron, G. and Idrissi, H. (2007). “Steel corrosion

in concrete: deterministic modelling of cathodic reaction as a function of water

saturation degree”, Corrosion Science, 49, 1918-1932.

Jones, D. A. (1992). Principles and prevention of corrosion. MacMillan, New York.

Liang, M. T., Jin, W. L., Yang, R. J., and Huang, N. (2005). “Predeterminate model of

corrosion rate of steel in concrete.” Cement and Concrete Research, 35, 1827–1833.

Liu, Y. (1996). “Modeling the time to corrosion cracking of the cover concrete in

chloride contaminated reinforced concrete structure.” Ph.D. Dissertation, Virginia

Polytechnic Instituted, Blackburg, VA.

Kassir, K. M. and Ghosn, M. (2002). “Chloride-induced corrosion of reinforced concrete

bridge decks,” Cement and Concrete Research, 32, 139–143.

Martys, S. N. Torquato, S., and Bentz D. P. (1994). “Universal scaling of fluid

permeability for sphere stacking.” Physical Review E, 50(1), 403-408.

MacGregor, J., and Wight, K. J. (1997). Reinforced concrete: mechanics and design.

Prentice-Hall, New Jersey.

Metha, P. K. (1993). Concrete structure, properties and materials. Prentice-Hall, New

Jersey.

Mindess, S. Young, J. F., and Darwin, D. (2003). Concrete, Second Edition, Prentice

Hall, New Jersey.

Monlouis, J. P., Verdier, J., and Perrin, B. (2007). “Predication of the relative

permeability to gas flow of cement-based materials.” Journal of Cement and

Concrete Research, 34, 737-744.

Neville, A. M. (2000). Properties of concrete. Fourth Edition, Prentice-Hall, New Jersey.

27
Nossoni, G. (2010). “Holistic electrochemical and mechanistic modeling of corrosion-

induced cracking in concrete structures.” Ph.D. Dissertation, Michigan State

University, East Lansing, Michigan.

Nossoni, G., Harichandran R. S., and Barton, S. C. (2011). “Chloride-induced corrosion

of reinforcing steel in concrete: electrochemical modeling of initial corrosion rate.”

Corrosion 2011 Conference, NACE, Texas, USA.

Newhouse, C. D., and Weyers, R. E. (1996). “Modeling the measured time to cracking”

Corrosion Activity of Reinforced Concrete Structure, In ASTM STP 1276, 3-21.

Newhouse, C. D. (1993). “Corrosion rates and the time to cracking model for critically

contaminated reinforced concrete structures.” Master of Science Thesis, Virginia

Polytechnic Instituted, Blackburg, VA.

Palle, T. C. (2002). “Deterioration of concrete structures.” Proceedings, First

Pantazopoulou, and K. Papoulia, “Modeling Cover-Cracking Due to Reinforcement

Corrosion in RC Structures.” Journal of Engineering Mechanics, ASCE, Vol. 127,

No. 4, pp.342-451 (2001).

Sarin, P., Snoeyink, V. L., Lytle, D. A., and Kriven, W. M. (2004). “Iron corrosion scale:

model for scale growth, iron release, and colored water formation.” Journal of

Environmental Engineering, ASCE, 130(4), 364-373.

Shah, S. P., and Swart, S. E. (1987). Fracture of concrete and rocks. Springer-Verlag,

New York.

Thoft-Christensen, P. (2002) “Deterioration of concrete structures.” Proceedings, First

International Conference on Bridge Maintenance, Safety and Management, IABMAS,

Barcelona, July 14-17, 16 pp.

28
Tomoshov, N. D. (1966). Theory of corrosion and protection of metals. McMillan Co.,

New York.

Van Genuchten, M. Th. (1980). “A close form equation for predicting the hydraulic

conductivity of unsaturated soils.” Journal of the Soil Sci. Soc. Am., 44(5), 892-898.

Xi, Y., Willam, K., and Frangopol, D. M. (2000). “Multiscale modeling of interactive

diffusion processes in concrete.” Journal of Engineering Mechanics, ASCE, 126(3),

258-265.

29
List of Tables

Table 1: Different Concentration Boundary Conditions (gr/cm3)

Table 2: Geometry

List of Figures

Figure 1: Corrosion scale model (adopted from Jones 1992, Sarin et al. 2004)
Figure 2: Oxygen diffusion path
Figure 3: Finite element analysis for calibrating the model
Figure 4: Test Specimens
Figure 5: Test set up under microscope
Figure 6: Images of rust thickness under the microscope for specimens with k = 6
Figure 7: Rust thickness over time for specimens with k = 3, 4, and 6
Figure 8: Corrosion current for BC1
Figure 9: Corrosion current for BC2
Figure 10: Corrosion current for BC3
Figure 11: Effect of water cement ratio on concrete cracking time: (a) BC1 and (b) BC2
Figure 12: Effect of concrete cover on cracking time (a) BC1 and (b) BC2
Figure 13: Comparison between model and Equation 32
Figure 14: Reduction in pressure as a percentage of the final pressure

30
Table 1: Different Concentration Boundary Conditions (g/cm3)

Boundary Oxygen Chloride

Condition Concentration Concentration
Type 1 3.04 x10-4 1.495 x10-3
Type 2 3.04 x10-4 1.301 x10-2
Type 3 7.0x10-7 1.91 x10-2

Table 2: Geometry

D (Concrete Cover) d (Bar Diameter)

Ratio (k)
mm (in.) mm (in.)
3 25 (1.0) 25 (1.0)
4 38 (1.5) 25 (1.0)
6 63 (2.5) 25 (1.0)

31
 Pore Solution
Of Concrete
Fe (III)
Top Fe(OH)3
Surface
e-
e- OH- e-
e-
Fe (II)- Fe (III)
Shell layer Fe2O3; Fe3O4
OH- Cl-
Porous Cl-
OH-
Core Cl-
Fe (II)
Cl-
Fe(OH)2
Pit electrolyte

Corroded Fe
Floor

Figure 1: Corrosion scale model (adopted from Jones 1992, Sarin et al. 2004)
Concrete

1 Co2,g,out
Cover
Porous

layer
Steel

Top
core

2 Co2,g,in
3 Co2,l,in
4 Cinter,sink
5 C0=zero

d1= thickness of core layer

d2= thickness of top layer
2
4 dc= thickness of concrete cover
3
5

d1 d2 dc

Figure 2: Oxygen diffusion path

32
 2500

Internal Pressure (psi)

2000

1500

k=3- Model
k=3- FEM 1000
k=4- Model
k=4- FEM
k=6- Model 500
k=6- FEM

0
-4.0E-04 -3.4E-04 -2.8E-04 -2.2E-04 -1.6E-04 -1.0E-04 -4.0E-05
Rust Thickness (in.)
Figure 3: Finite element analysis for calibrating the model

33
 Bottom surface coated with
epoxy (concrete and bars)

k=3

Anode: # 4 Bar
Cathode: 3 # 1 Bars

D = 1.5”, 2”, and 3” Top surface polished

Q = 120° d = 0.5”

k=4

h= 1”

k=5
Figure 4: Test Specimens

Computer Connected to Camera

Camera

Microscope
Power Supply

Lights
034 05
034 050

+ -
Specimen

Figure 5: Test set up under microscope

34
 Concrete Cover
Steel bar
Anode Concrete Cover Rust layer

Steel bar
Anode

0.1 mm
0.1 mm

(a) t=0min (b) t=35 min

Steel bar
Rust layer Anode Steel bar
Anode
Rust layer

Concrete Cover
0.1 mm Concrete Cover 0.1 mm

(c) t= 70 min (d) t=90 min-Cracking

Steel bar
Steel bar Anode
Anode

Rust layer
0.1 mm 0.1 mm
Concrete Cover
Rust layer Concrete Cover

(e) t=115 min-Cracking (f) t=130 min-Cracking

Figure 6: Images of rust thickness under the microscope for specimens with k = 6

35
Figure 7: Rust thickness over time for specimens with k = 3, 4,and 6
 
 



3.0e-6

2.5e-6 

2.0e-6 

1.5e-6  


1.0e-6


5.0e-7


0.0
0.0 50.0 100.0 150.0 200.0

 

Figure 8: Corrosion current for BC1

36
Corrosion Current (A cm-2)
3.0e-6
K=3
2.5e-6 K=4
K=6
2.0e-6

1.5e-6 w/c=0.6

1.0e-6 0.5

5.0e-7
0.4
0.0
0.0 50.0 100.0 150.0 200.0
Time (Day)

Figure 9: Corrosion current for BC2

2.0e-7
Corrosion Current (A cm-2)

K=3
1.5e-7 w/c=0.6 K=4
K=6

1.0e-7
0.5

5.0e-8

0.4
0.0
0.0 50.0 100.0 150.0 200.0
Time (Day)

Figure 10: Corrosion current for BC3

37
 (a)

(b)
Figure 11: Effect of water cement ratio on concrete cracking time: (a) BC1
and (b) BC2

38
 (a)

(b)
Figure 12: Effect of concrete cover on cracking time (a) BC1 and (b) BC2

39
 Figure 13: Comparison between model and equation 32

Figure 14: Reduction in pressure as a percentage of the final pressure

40