Corrosion Science 55 (2012) 385–393

Contents lists available at SciVerse ScienceDirect

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

Crack shape and rust distribution in corrosion-induced cracking concrete

Yuxi Zhao ⇑, Jiang Yu, Bingyan Hu, Weiliang Jin
Institute of Structural Engineering, Zhejiang University, Hangzhou 310058, China

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

Article history: This study investigated a reinforced concrete specimen that had deteriorated in an artificial environment
Received 26 July 2011 for 2 years. The crack width and the rust distribution were observed by digital microscopy. The variation
Accepted 2 November 2011 of the total circumferential crack width along the radial direction is presented using a linear function.
Available online 16 November 2011
Observation reveals that rust does not penetrate into the corrosion-induced cracks before concrete sur-
face cracking. After concrete surface cracking, rust fills the cracks, lining the edges of the cracks due to the
Keywords: circulation of the outer solution. A schematic diagram is proposed to describe crack propagation and rust
A. Steel reinforced concrete
development.
B. Modelling studies
C. Rust
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction corrosion, were used to study the corrosion-induced cracking pro-

cess. It should be noted the analytical and numerical models can-
Corrosion-induced cracking is the major cause of deterioration not accurately reflect real situations without considering the
in reinforced concrete structures [1,2]. As steel corrosion develops, crack patterns and shapes in the concrete cover induced by steel
the corrosion product, which is about 2–6 times [3,4] the volume of corrosion. However, relevant work is scare in the literature.
the original steel, produces an expansive pressure on the surround- Theoretically, before concrete surface cracking, the rust that
ing concrete and eventually leads to concrete cover cracking. Once causes concrete surface cracking can be taken as the sum of three
the cracks penetrate the concrete cover, a path for the rapid ingress processes [16]: the rust migrates to the pores at the steel/concrete
of aggressive agents to the reinforcement is provided; this results interface, the rust creates the stress in the concrete cover, and the
in an acceleration of the steel corrosion process, which will lead rust migrates to the corrosion-induced cracks. The first and second
to progressive deterioration and even spalling of the concrete cover processes have been studied previously [16,19–33]. For the third,
[5–7]. Therefore, investigation of the corrosion-induced cracking however, there are very few published experimental results.
process is crucial for predicting the serviceability and durability Whether or not the corrosion products fill the corrosion-induced
of reinforced concrete structures. cracks is still under discussion. It is, however, one of the key factors
A considerable number of experimental studies have been con- affecting the development of a suitable concrete cracking model for
ducted on corrosion-induced concrete cracking, primarily focusing steel corrosion.
on two aspects of the cracking process: (1) predicting steel corro- Therefore, the aim of this study was to investigate the crack
sion at surface cracking [8–11] and (2) linking the crack width on shape in the concrete surface and rust distribution in corrosion-
the surface of the concrete cover with steel corrosion [8,12–15]. induced cracks, which can help to understand the process that
Empirical models have been established based on experimental causes concrete surface cracking. To achieve this objective, the
data. Fig. 1a shows a comparison between steel corrosion on sur- authors observed the cracks and rust from reinforced concrete
face cracks calculated by empirical models [8–11] and corrosion specimens corroded under simulated natural environmental condi-
observed in accelerated corrosion tests [8–11,13,16–18], whereas tions. The crack widths at different radii were measured, and the
Fig. 1b shows a comparison of surface crack width propagation rust distribution in corrosion-induced cracks was observed. Crack
between model-predicted results [8,12,14,15] and experimental propagation induced by steel corrosion and rust development in
results [13]. It can be clearly observed that none of the empirical cracks are also discussed in this paper.
models provide a good prediction of all of the experimental data,
as empirical models usually cannot take into account all of the 2. Experimental program
relative parameters. Therefore, analytical [19–28] and numerical
[29–32] methods, which are able to consider more relevant param- 2.1. Reinforced concrete specimen
eters such as geometrical dimensions, concrete properties, and
A reinforced concrete specimen with the dimensions shown in
⇑ Corresponding author. Tel.: +86 571 88208726. Fig. 2 was used in this study. This concrete specimen was taken
E-mail address: [email protected] (Y. Zhao). from the same batch of specimens used in a previous study

0010-938X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2011.11.002
386 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393

levels, respectively. The mixture proportions used are reported in

Table 1. The cement used was a Portland cement that conforms
to CEM I 42.5N (EN197-1:2000). The concrete contained a com-
mercial calcium nitrite based corrosion inhibitor. The gravel and
sand were siliceous aggregates at 20 and 5 mm maximum size,
respectively. The 28-day compressive strength of the concrete
was 56 MPa as measured on 150-mm cubes.

2.2. Curing and exposure history

The specimen was covered in damp hessian immediately after

casting and sprayed with water once every day for 2 weeks. There-
after, the panel was moved into the lab and dried at room temper-
ature (approximate 25 °C) for several weeks. The side and bottom
surfaces of each panel were surface treated with epoxy resin and
a polyurethane coating to ensure that chloride predominantly
penetrated through the top cover, with minimal penetration
through the other faces of the panel.
The concrete panel was then subjected to alternate wetting and
drying cycles in a walk-in environmental chamber. Each test cycle
lasted for 3 days and consisted of spraying the blocks with
Fig. 1a. Comparison of steel corrosion at concrete surface cracking between the 3.53 wt.% sodium chloride solution for 4 h and subsequently allow-
model-predicted results and the experimental results. ing them to dry at 40 °C for the remainder of the time to complete
one full cycle.
After 2 years of exposure, longitudinal cracks were observed on
the top face of the specimen, running approximately parallel to
the reinforcing bar. However, the longitudinal cracks did not run
across the entire top surface of the specimen; some sections at the
centre of the top surface did not show cracks, as shown in Fig. 3.
The widths of the cracks in the areas near both sides of the speci-
mens were obviously wider than those in the middle sections.
Brownish-red rust stains were observed on the external side sur-
faces of the specimen. The specimen was then removed from the
environmental chamber for further testing.

2.3. Sample preparation

Fig. 1b. Comparison of concrete surface crack width propagation between the
The selected specimen was cast into a low-viscosity epoxy resin
model-predicted results and the experimental results. to minimise any artificial damage that may have been incurred dur-
ing the sample preparation process for microscopy. Then, it was
carefully cut to extract the corner and middle rebars while keeping
reported in the literature [33,34]. The specimen was cast in situ by their surrounding concrete intact, as shown in Fig. 3. The cut panels
a construction company in accordance with specifications used for were labelled PL, PM and PR, representing the sections containing
reinforced concrete elements of a major sea-crossing bridge in Chi- the left-corner, middle and right-corner rebar, respectively.
na. Each specimen contained three ribbed bars with a nominal Each panel was then sectioned sequentially, starting from the
diameter of 16 mm and a 20-mm cover spacing. The rebars were side face, to produce a series of 10 mm-thick cross-sectional slices.
used as received, and no efforts were made to remove the existing Cutting was carried out using an abrasive cutter and diamond
mill scale. A low cover depth of 20 mm was adopted to reduce the blade suitable for hard brittle materials. An example of a slice is
time for corrosion initiation. shown in Fig. 4a. The locations of the slices are indicated in
The concrete used in the panel was a ternary blended mixture Fig. 3; for example, ‘PR-6’ represents the 6th slice from the side
that contained slag and fly ash at 40 wt.% and 30 wt.% replacement face of this panel containing the right-corner rebar.

Fig. 2. Schematic of the reinforced concrete specimen (dimensions are in mm).

 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393 387

Table 1
Mixture composition of concrete specimens (kg/m3).

Cement Blast-furnace slag Fly ash Sand Aggregate Water (l/m3) Ratio of water/binder Water-reducing admixtures Corrosion inhibitor
126 168 126 735 1068 145 0.345 5.04 8.4

PL PM PR
PM-14
Crack
PR-4
PL-4
PR-1
Rebar

Cutting lines

Concrete panel

Fig. 3. Schematic diagrams of the cut specimens.

Rust
Rebar

Rust 357.3
Crack

Rebar 314.9

349.6
Concrete Concrete

20 mm 1 mm 432.5

(a) field of view: 50×52 mm (b) field of view: 3.74×3.68 mm

Fig. 4. Measurement of the thickness of corrosion layer accumulated at the rebar–concrete interface.

To prevent further corrosion, these slices were kept in a dry The crack width was measured by digital microscopy. For every
environment (relative humidity less than 30%) before observation. crack, measurements were made along the circumferential direc-
Polishing at each stage was performed using different specifica- tion at intervals of 1 mm along the radial direction from the surface
tions of abrasive paper. of the steel bar towards the concrete surface, as shown in Fig. 5.
The total crack width at radius Ri (in mm) can be obtained by sum-
ming all of the crack widths at this radius:
2.4. Measurements
X
j
Wi ¼ wki ð1Þ
2.4.1. Corrosion layer
k¼1
The boundaries of the rebar, corrosion layer and concrete of
each sample were observed by digital microscopy (Pro-micro scan
5866). The rust layer could be clearly distinguished in colour ws1
images as shown in Fig. 4b when the digital microscopy was con-
nected to a computer, which allowed for the accurate measure- w1,20
ments of the corrosion layer thickness, as illustrated in Fig. 4b.
w1i
The area of the corrosion layer was also measured by the digital
microscope. Then, the steel cross-sectional loss area was calculated
by dividing the area of the corrosion layer by the volume expansion
ratio of rust and the original steel n, which was assumed to be 2 in w2i
this paper. ws2

2.4.2. Crack width R

Because some tiny cracks and the cracks that had not reached Ri
the concrete surface could not be penetrated by the epoxy, the
slices with these kinds of cracks were easily damaged during the
cutting process. Only 12 slices cut from panel PR remained undam-
aged. Slices taken from the front side of panel PR were labelled
PR-1 to PR-12 for this crack width measurement. Fig. 5. Measurement of the circumferential crack width.
388 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393

where Wi is the total circumferential crack width at radius Ri, in

mm; wki is the circumferential crack width of a single crack at ra-
dius Ri, in mm; and j is the number of cracks at radius Ri.
According to the measured results, the relationship between the
total circumferential crack width and the radius can be obtained as
W i ¼ f ðRi  RÞ ð2Þ
where R is the radius of steel bar, in mm.
Additionally, the crack width on the surface of the concrete cov-
er was also measured. The sum of all of the crack widths along the
top and side of the slice, the total surface crack width Ws (in mm),
can be expressed as
X
j
Ws ¼ wsk ð3Þ
k¼1

where wsk is the crack width of a single crack at the edge of each
slice, in mm; and j is the number of cracks on the surface of the con-
crete cover.
Fig. 7. Simplified crack model.
2.4.3. Rust distribution
The slices from panels PL, PM and PR were all investigated to
Table 2
study the rust distribution. According to the colours and locations
Parametric regression value in model.
of the rust, the distribution of different types of rust was observed
by digital microscopy. Trends in the rust distribution in the corro- q (%) a b
sion-induced cracking concrete specimen were then analysed. PR-1 6.07 0.035 1.507
PR-2 7.07 0.052 1.495
PR-3 5.89 0.039 0.823
3. Results and discussion PR-4 4.71 0.017 1.163
PR-5 3.50 0.016 0.924
3.1. Crack shape PR-6 1.16 0.011 0.345
PR-7 0.55 0.004 0.378
PR-8 0.75 0.007 0.557
3.1.1. Crack width
PR-9 0.71 0.006 0.595
Using slices PR-1 and PR-5 as examples, Fig. 6 shows the typical PR-10 2.34 0.008 0.451
variation in the measured circumferential crack width along the PR-11 2.78 0.015 0.268
radius direction and the regression line of each slice. In Fig. 6, the PR-12 2.20 0.018 0.209
values of steel corrosion q is the steel cross-sectional loss area
expressed as the percentage of the original rebar cross-sectional width, some regularity can still be found. As shown in Fig. 7, the
area. It can be seen that the total circumferential crack width is total circumferential crack width becomes larger farther away
linearly proportional to the radius as follows: from the surface of the rebar. The slope of the regression line and
W i ¼ aðRi  RÞ þ b ð4Þ the crack width on the surface of the steel bar are influenced by
the steel corrosion. This finding will be discussed in the next
where the parameters a and b can be interpreted physically to section.
describe various characteristics of the cracks. In the simplified crack
model shown in Fig. 7, a is the crack width variation coefficient, and 3.1.2. Parameters discussion
b is the crack width coefficient at the surface of the steel bar. 3.1.2.1. a – Crack width variation coefficient. Parameter a describes
The values of parameters a and b obtained from the regression the variation in the crack width along the radial direction. Fig. 8
analysis for all slices are presented in Table 2. The values of steel shows that a is linearly proportional to the steel corrosion q as
corrosion q for all slices are also listed in Table 2. follows:
Due to the heterogeneity of the concrete, the crack width is
highly variable. However, despite the wide variability in the crack a ¼ 0:67857q  0:00234 ð5Þ

Fig. 6. Measurement and fitting of crack width.

 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393 389

W i ¼ ½0:67857ðRi  RÞ þ 20:93q  0:00234ðRi  RÞ ð7Þ

For the specimens in this study, the total crack width of the con-
crete surface, i.e., Ri  R = 20 mm, can be expressed as
W i ¼ 34:5014q  0:0468 ð70 Þ
For the steel/concrete interface, i.e., Ri  R = 0, the total crack
width can be expressed as follows:
W i ¼ 20:93q ð700 Þ
By substituting a steel corrosion of 0.34%, which is the moment
of concrete surface crack initiation, into Eqs. (70 ) and (700 ), the crack
width on the surface of the concrete cover and the steel/concrete
interface are observed to be very close, both approximately
Fig. 8. Relationship between a and q.
0.07 mm. This shows that the widths of cracks remain constant
during concrete surface cracking.
Molina et al. [29] assumed that a finite number of cracks are
ðR2 ¼ 0:850Þ completely opened in the concrete that surrounds the rebar and
the state is free of stress, so the problem can be described geomet-
The regression line intersects with the x-axis at q = 0.34% when
rically, as shown in Fig. 10. Then, the relationship between the
a = 0, which means that the total crack width is constant along the
crack width at the surface of the steel bar and the steel corrosion
radial direction at this value.
can then be derived:
For the condition q < 0.34%, a is less than 0, which describes the
scenario in which the crack width near the surface of the concrete W i ¼ 2pðn  1Þd ð8Þ
cover is smaller than that near the steel bar. This situation occurs where d (mm) is the radial loss of the steel bar. Taking the steel cor-
before concrete surface cracking and is observed in the slices used rosion qp = ffiffiffiffiffiffiffiffiffiffiffiffi
0.34% at surface cracking obtained for the above example,
in this study. Because most of these cracks were damaged during d ¼ R  1  q  R = 13.6 lm. The calculated result using Molina’s
sample preparation (see Section 2.3), a quantitative study was model is Wi = 0.085 mm, which is slightly larger than the regression
not possible. However, the observation of a few intact cracks that result of the experimental data. The reason for this difference might
did not reach the concrete surface shows that the crack widths ini- be that in this analytical model, the surface crack width is calcu-
tially remained constant along the radius direction near the steel lated geometrically without considering the real conditions of the
bar and gradually became 0 at the ends of these cracks. In this case, concrete specimens, such as the restraining effect of the surround-
a is less than 0, which agrees with the above statement. ing concrete and the friction between the interface of the corroding
Therefore, q = 0.34% can be considered the critical steel corro- steel bar and concrete cover, leading to a larger calculated value.
sion at the surface cracks of the concrete cover for panel PR in this
study. 3.2. Crack width on the surface of the concrete cover

3.1.2.2. b – Crack width coefficient at the surface of the steel bar. Para- 3.2.1. Experimental results
meter b reflects the crack width on the surface of the steel bar. It The surface crack width was measured as discussed in Section
can be seen from Fig. 9 that as the steel corrosion increases, b increases 2.3; the total surface crack width Ws could then be calculated by
linearly: Eq. (3). Fig. 11 plots the tested surface crack width as a function
b ¼ 20:93q ð6Þ of steel corrosion. The figure shows that the surface crack width
is linearly proportional to the steel corrosion according to the fol-
ðR2 ¼ 0:612Þ lowing expression corresponding to the models of Rodriguez et al.
[8], Vidal et al. [12] and Zhang et al. [15]:

3.1.2.3. The total circumferential crack width. Substituting the W s ¼ 39:14q  0:04904 ð9Þ
regression results of a and b into the expression for Wi, i.e., substi-
tuting Eqs. (5) and (6) into Eq. (4), the total circumferential crack
width at any radius with different steel corrosion can be expressed
as follows:

Fig. 9. Relationship between b and q. Fig. 10. Geometrical model of cracks [29].
390 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393

cracks. A cross-section of this slice is shown in Fig. 12a. The top

surface of the slice was exposed to the chloride ingress; the left
side is the polyurethane-treated surface, and the right side and
the bottom are the cutting surfaces. There are five visible cracks
in this slice, cracks 1–5, as shown in Fig. 11a. The surfaces of crack
4 and crack 5 are rather coarse. According to the morphology of
these two cracks, it is assumed that these two cracks formed be-
cause of a cutting disturbance rather than steel corrosion. There-
fore, the rust distribution was investigated only in the other
three cracks, cracks 1, 2 and 3. Typical images showing the local
distribution of rust in these cracks are shown in Fig. 12b–e. The
size of the observed area is 4718 lm  3767 lm.

Fig. 11. Relationship between Ws and q. 3.3.1.1. External cracks. Cracks 1 and 2 both propagate to the sur-
face of the concrete cover because the rust distribution in these
two cracks is similar. Here, crack 1 is taken as an example to illus-
ðR2 ¼ 0:82Þ
trate the characteristics of the rust in the external cracks. Area 1 is
As shown in Fig. 11, the total circumferential crack width at the close to the steel bar. As shown in Fig. 12b, the corrosion products
surface of the concrete cover given by Eq. (70 ) is slightly smaller clearly fill the cracks. Three types of corrosion products can be dis-
than the values calculated using Eq. (9). This disparity can be tinguished: rust r at the steel/concrete interface, rust s filling
explained by the fact that the cracks in the specimens did not prop- into the corrosion-induced cracks and rust t lining the edges of
agate strictly along the radial direction and thus were not perpen- the corrosion-induced cracks and the surrounding microcracks.
dicular to the concrete surface, leading to this inequality of the Rust r appears black with a silver metallic lustre. Rust r ad-
total circumferential crack width at Ri  R = 20 mm and the surface heres to the steel surface and induces expansive pressure on the
crack width. As shown in Fig. 4b, the circumferential crack width of surrounding concrete cover, which is the direct cause of concrete
crack 1 at Ri  R = 20 mm is smaller that its surface crack width. surface cracking. Due to the restriction of the surrounding con-
crete, rust r appears very dense. Because rust r was formed in
the concrete, the oxygen supplement was not sufficient, and its
3.2.2. Crack width at surface cracking
colour is black, we consider it is Fe3O4.
In previous studies, the steel corrosion at surface cracking was
Rust sis a dark-brown appearance. Rust s fills the cracks of
normally defined as the steel corrosion when the surface crack
the rust layer apart from the cracks in the concrete cover, as shown
width equals 0 [8,12–15]. However, according to the discussion
in Fig. 12b. As rust s fills the empty space of the existing cracks, it
in Section 3.1.2, the crack width is not 0 during surface cracking
is not as subject to pressure as rust r; therefore, it is relatively
but maintains the same value along the radial direction in the con-
loose. Rust s was formed after the cracking of concrete cover,
crete cover. Therefore, it is inaccurate to predict the steel corrosion
the solution (H2O, Cl2) and oxygen (O2) came to rust layer through
at concrete surface cracking by letting the surface crack width
the concrete cracks, Fe3O4 was further oxidized. Because it was in a
equal 0.
chloride ingress environment, Rust s is regarded as b-FeO(OH).
In the third part of Section 3.1.2, the surface crack width at sur-
Rust t is reddish brown and is observed along the edges of the
face cracking calculated by Eq. (70 ) is 0.07 mm. By substituting this
corrosion-induced cracks and the concrete microcracks around the
value into Eq. (9), the steel corrosion q = 0.31% is obtained, which is
steel bar and cracks. It occurs because the outer solution enters the
the critical steel corrosion at surface cracking, corresponding to a
cracks after they penetrate the concrete cover; some rust dissolves
critical steel corrosion of 0.34% with respect to coefficient a.
in the solution and contaminates the concrete it contacts. When
Based on the above discussion, it is appropriate to introduce a
the solution dries during the drying cycles, the rust adheres to
non-zero crack width to predict the steel corrosion upon surface
the edge of the cracks. With rich oxygen supplement, the quantiva-
cracking. The value of this critical crack width measured in this
lence of rust compounds became higher, therefore rust t is con-
study is 0.07 mm. In Sections 3.1 and 3.2, the cracks were investi-
sidered as Fe2O3, It is believed that the penetration of rust t
gated to obtain the relationship between the circumferential crack
into the microcracks occurs with water-cooling during the cutting
width and the steel corrosion. However, it should been noted that
process. Although most of the contaminated concrete has been
this experiment was performed only on the unloaded specimens
rubbed off, there are still some residual corrosion products.
with a steel bar diameter of d = 16 mm and concrete cover thick-
In area 2, which is connected to area 1 as shown in Fig. 12c, the
ness of C = 20 mm. The main purpose of this study was to reveal
space of the upper-right crack is empty. The edges of the cracks at
the linear variation regularity of crack width, though the coeffi-
the bottom-left of the area are stained by rust t, which is reddish
cients of this linear model depend on the specimen.
brown in colour. There is no rust observed between areas 2 and 3.
It should be noted that rust r is also observed on the edge of
3.3. Rust distribution the left crack near the steel bar, as shown Fig. 12b, and not only
on the surface of the steel bar. This is probably because there
3.3.1. Rust distribution in different types of cracks was an initial defect (maybe a small crack) in the concrete before
The cracks in this study are classified into two types: external steel corrosion. During the early stages of steel corrosion, the rust
and internal cracks. The cracks that can penetrate the concrete cov- fills the space of this crack. Due to restriction caused by the sur-
er are defined as the external cracks, whereas the cracks that can- rounding concrete, the rust is compacted and ultimately becomes
not propagate to the surface of the concrete cover, including those rust r. As the corrosion products accumulate, the rust layer cracks,
propagating inwards rather than outwards to the concrete surface spreading outwards to the concrete surface through rust r in the
and the ones developing between two steel bars, are defined as the initial crack, which is weaker than the vicinity.
internal cracks. The crack in the rust layer was mentioned above and is shown
Slice PR-6 (located in the middle of the panel PR) is taken as an in Fig. 12b. However, it is not commonly found. More images show
example to introduce the rust distribution in corrosion-induced that the rust layer does not crack when the concrete cracks. For
 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393 391

Epoxy
3
Crack 1
Rust
2
1
Crack 2
Crack 3
Steel bar

Crack

Crack 5

Concrete
20 mm

(a) Slice PR-6

Epoxy

2 mm 2 mm

(b) Area 1 (c) Area 2

Epoxy Epoxy

2 mm 2 mm

(d) Area 3 (e) Area 4

Fig. 12. Photographs of observation for the slice PR-6.

instance, Fig. 13 shows that the cracks at the steel/concrete can be seen from Fig. 12e that although the crack width is about
interface area in slice PM-14 (located in the middle of the panel 60 lm and the rust layer has cracked, very little rust can be observed
PM) are clearly different from those observed in area 1 of slice in the cracks. This is because the outer solution can barely penetrate
PR-6. The same results were observed in the other slices with into the internal cracks to solvate and carry the rust ions. Therefore,
cracks at the steel/concrete interface. This finding illustrates that the ingress of the outer solution is the reason why the rust distribu-
concrete cover cracking occurs earlier than rust layer cracking. tion between external and internal cracks differs.
Rust t in the cracks near the steel/concrete interface may form
with water-cooling during the cutting process as discussed above. 3.3.2. Influence of steel corrosion on rust distribution
Observation of all of the slices shows that the range of rust dis-
3.3.1.2. Internal cracks. In slice PR-6, crack 3 is a crack generated be- tribution becomes wider as steel corrosion increases. Slice PL-4 (lo-
tween two steel bars (i.e., an internal crack). The rust distribution in cated in the middle of the panel PL), for example, shown in Fig. 14,
this type of crack is different from that in external cracks 1 and 2. It features severe cracking and a surface crack width of 2.865 mm for
392 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393

surrounding concrete and is therefore dense. As steel corrosion in-

creases, the concrete at the steel/concrete interface cracks first,
while the rust remains deposited on the steel/concrete interface
due to the protection of the dense rust layer. Only after the rust
layer cracks is the rust able to fill the cracks.
This study shows that for external cracks with small crack
widths, rust is rarely found within the cracks. The test results sug-
gest that rust does not fill cracks before concrete surface cracking.
After concrete surface cracking, the outer solution penetrates
into the concrete through the cracks. Some rust dissolves in the
solution, and the circulation of the solution removes the rust from
2 mm the steel bar. When the specimen dries during the wetting and dry-
ing cycles, the rust remains in the cracks. Therefore, except for rust
filling the cracks close to the rebar, rust can also be observed to
Fig. 13. Crack at steel/concrete interface of the slice PM-14 (6897 lm  6155 lm). absorb onto the edges of the cracks far away from the rebar.
Rust can also move into the internal cracks after the cracking of
the rust layer. However, because the outer solution cannot pene-
Epoxy trate these cracks, there is no rust found lining the edges of the
6
cracks without the circulation of the solution.
Crack 6
5
3.4. Discussion of crack propagation and rust development

Rust As discussed in Section 3.1, the variation of the crack width

along the radial direction can be described using a linear model.
Steel bar
Along with the discussion of rust distribution in Section 3.3, a sche-
matic diagram is used to describe crack propagation and rust
development during the corrosion-induced cracking process, as
shown in Fig. 15.
Concrete
20 mm 3.4.1. Crack propagation
For the external cracks, which can penetrate the concrete cover,
Fig. 14. Slice PL-4 with the severer corroded steel bar. the crack width does not significantly change along the radial
direction near the steel bar before concrete surface cracking. Only
at the end of the crack does the crack width gradually diminish to
crack 6. The rust can be observed lining the surface of the crack in 0, as shown in Fig. 15a. During concrete surface cracking, the crack
area 5, which is far from the steel bar but close to the concrete sur- width can be considered constant along the radial direction, as
face. Rust can even be found on the edges of the crack in area 6, shown in Fig. 15b. After concrete surface cracking, the shape of
which is on the surface of the concrete cover. This is understand- the crack is assumed to be a trapezoid, as shown in Fig. 15c.
able because more severely corroded rebar produces more corro- For the internal cracks between two steel bars, the shape of the
sion product. Moreover, the slices with more severely corroded crack is nearly the same as that of the external crack when the
steel bar normally exhibit wider external cracks, leading to the eas- crack has not broken through between the two bars. After crack
ier ingress and circulation of the outer solution during the wetting transfixion, the crack width grows as steel corrosion increases,
and drying cycles. These findings also suggest that the rust could but the crack width along the crack can be assumed to be constant,
be carried farther away from the rebar. as shown in Fig. 15.

3.3.3. Discussion of rust filling corrosion-induced cracks 3.4.2. Rust development

Before steel/concrete interface cracking, rust deposits on the For external cracks, as shown in Fig. 15, rust develops as
interface gradually. This rust layer is compressed by the follows: (1) the concrete at steel/concrete interface cracks and

Outer solution Outer solution Outer solution

Concrete Concrete Concrete

Rust
Rust

Rust Rebar Rust Rebar Rust Rebar

(a) Before surface cracking (b) Surface cracking (c) After surface cracking

Fig. 15. Schematic diagram on crack propagation and rust development.

 Y. Zhao et al. / Corrosion Science 55 (2012) 385–393 393

the rust (rust r) still accumulates at the steel/concrete interface, References

leading to the propagation of the cracks in the concrete cover as
shown in Fig. 15a. (2) The dense rust layer cracks, and the cracks [1] Z.P. Bazant, Physical model for steel corrosion in sea structures–applications, J.
Struct. Div. 105 (1979) 1137–1153.
propagate to the concrete surface in a relatively short amount of [2] P.B. Bamforth, Enhancing reinforced concrete durability: guidance on selecting
time; the rust filling the cracks can be ignored before surface crack- measures for minimizing the risk of corrosion of reinforcement in concrete,
ing. (3) After concrete surface cracking, the outer solution ingresses Concrete Society Technical Report No. 61, 2004.
[3] K. Suda, S. Misra, K. Motohashi, Corrosion products of reinforced bars
into the cracks, and steel corrosion is accelerated, as shown in embedded in concrete, Corros. Sci. 35 (1993) 1543–1549.
Fig. 15b. If there is no solution on the surface of the sample, the [4] Y.X. Zhao, H.Y. Ren, H. Dai, W.L. Jin, Composition and expansion coefficient of
moisture (H2O) and oxygen (O2) will ingress through the cracks rust based on X-ray diffraction and thermal analysis, Corros. Sci. 53 (2011)
1646–1658.
and accelerate the steel corrosion as well. (4) As the steel corrosion
[5] A. Bentur, S. Diamond, N.S. Berke, Steel Corrosion in Concrete, E&FN Spon,
grows, rust s fills the cracks, and the rust is carried away from the London, 1997.
rebar by the solution, lining the edges of the cracks, i.e., rust t, as [6] M. Ohtsu, S. Yosimura, Analysis of crack propagation, crack initiation due to
shown in Fig. 15c. corrosion of reinforcement, Constr. Build. Mater. 11 (1997) 437–442.
[7] S.J. Williamson, L.A. Clark, Pressure required to cause cover cracking of
Rust 2 fills the internal cracks after the cracking of the rust layer concrete due to reinforcement corrosion, Mag. Concr. Res. 52 (2000) 455–467.
though at a very slow speed, and no rust lines the edges of the cracks. [8] J. Rodriguez, L.M. Ortega, J. Casal, J.M. Diez, Corrosion of reinforcement and
According to these discussions, it can be assumed that the filling service life of concrete structures, in: Proceedings of the Seventh International
Conference on Durability of Building Materials and Components, London,
corrosion-induced cracks with rust need not to be considered in 1996, pp. 117–126.
studying the concrete cracking process before surface cracking. [9] C. Alonso, C. Andrade, J. Rodriguez, J.M. Diez, Factors controlling cracking of
Thus, the three-stage model [16] should be modified. The rust concrete affected by reinforcement corrosion, Mater. Struct. 31 (1998) 435–441.
[10] M.P. Webster, L.A. Clark, The structural effect of corrosion – an overview of the
causing concrete cracking develops via only two processes: the mechanism, in: Proceedings of the Concrete Communication, Birmingham, UK,
rust migrates to the pores at the steel/concrete interface, and the 2000, pp. 409–421.
rust creates stress in the concrete cover. Further experimental [11] B.H. Oh, K.H. Kim, B.S. Jang, Critical corrosion amount to cause cracking of
reinforced concrete structures, ACI Mater. J. 106 (2009) 333–339.
work is currently ongoing to find more evidence to support this [12] T. Vidal, A. Castel, R. Francois, Analyzing crack width to predict corrosion in
preliminary conclusion. reinforced concrete, Cem. Concr. Res. 34 (2004) 165–174.
[13] K. Vu, M.G. Stewart, J. Mullard, Corrosion-induced cracking: experimental data
and predictive models, ACI Struct. J. 102 (2005) 719–726.
4. Conclusions [14] J.A. Mullard, M.G. Stewart, Corrosion-induced cover cracking of RC structures:
new experimental data and predictive models, Research Report No. 275.05,
A reinforced concrete specimen deteriorated in an artificial 2009.
[15] R. Zhang, A. Castel, R. Francois, Concrete cover cracking with reinforcement
environment for 2 years was investigated in this study. The crack corrosion of RC beam during chloride-induced corrosion process, Cem. Concr.
width and the rust distribution were observed by digital micros- Res. 40 (2010) 415–425.
copy, and a schematic diagram was proposed to describe crack [16] Y.P. Liu, R.E. Weyers, Modeling the time to corrosion cracking in chloride
contaminated reinforced concrete structures, ACI Mater. J. 95 (1998) 675–681.
propagation and rust development. The following conclusions have [17] C. Andrade, C. Alonso, F.J. Molina, Cover cracking as a function of bar corrosion:
been drawn from this study: Part 1 – experimental test, Mater. Struct. 26 (1993) 453–464.
[18] Rasheeduzzafar, S.S. Al-Saadoun, A.S. Al-Gahtani, Corrosion cracking in
relation to bar diameter, cover, and concrete quality, J. Mater. Civ. Eng. 4
(1) A linear model is adopted to describe the variation of the (1992) 327–343.
total circumferential crack width along the radial direc- [19] K. Bhargava, A.K. Ghosh, Y. Mori, et al., Modeling of time to corrosion-induced
tion. The parameters in the linear model (i.e., the crack cover cracking in reinforced concrete structures, Cem. Concr. Res. 34 (2004)
2203–2218.
width variation coefficient a and the crack width coeffi- [20] Y.X. Zhao, W.L. Jin, Modeling the amount of steel corrosion at the cracking of
cient b on the surface of the steel bar) are discussed, concrete cover, Adv. Struct. Eng. 9 (2006) 687–696.
and their relationship with the steel corrosion is also [21] K. Bhargava, A.K. Ghosh, Y. Mori, S. Ramanujam, Model for cover cracking due
to rebar corrosion in RC structures, Eng. Struct. 28 (2006) 1093–1109.
proposed.
[22] C.Q. Li, R.E. Melchers, J.J. Zheng, Analytical model for corrosion-induced crack
(2) The surface crack width is linearly proportional the steel width in reinforced concrete structures, ACI Struct. J. 103 (2006) 479–487.
corrosion. A critical crack width of a certain value [23] T.E. Maaddawy, K. Soudki, A model for prediction of time from corrosion
(0.07 mm obtained in this study), rather than 0, is sug- initiation to corrosion cracking, Cem. Concr. Compos. 29 (2007) 168–175.
[24] L. Chernin, D.V. Val, K.Y. Volokh, Analytical modeling of concrete cover
gested to predict steel corrosion at concrete surface cracking caused by corrosion of reinforcement, Mater. Struct. 43 (2010) 543–
cracking. 556.
(3) Rust does not penetrate into the corrosion-induced cracks [25] K.H. Kim, S.Y. Jang, B.S. Jang, B.H. Oh, Modeling mechanical behavior of
reinforced concrete due to corrosion of steel bar, ACI Struct. J. 107 (2010) 106–
before the cracks reach the concrete surface. Thus, the 113.
three-stage model should be modified. After surface [26] G. Malumbela, M. Alexander, P. Moyo, Model for cover cracking of RC beams
cracking, rust fills the cracks and develops in the external due to partial surface steel corrosion, Constr. Build. Mater. 25 (2011) 987–991.
[27] C.H. Lu, W.L. Jin, R.G. Liu, Reinforcement corrosion-induced cover cracking and
cracks much more quickly than that in the internal cracks. its time prediction for reinforced concrete structures, Corros. Sci. 53 (2011)
Moreover, rust lines the edges of the external cracks due 1337–1347.
to the circulation of the outer solution. [28] Y.X. Zhao, J. Yu, W.L. Jin, Damage analysis, cracking model of reinforced
concrete structures with rebar corrosion, Corros. Sci. 53 (2011) 3388–3397.
(4) A schematic diagram is proposed to describe crack prop- [29] F.J. Molina, C. Alonso, C. Andrade, Cover cracking as a function of rebar
agation and rust development during steel corrosion corrosion: Part 2 – numerical model, Mater. Struct. 26 (1993) 532–548.
induced concrete cracking process. [30] E.J. Hansen, V.E. Saouma, Numerical simulation of reinforced concrete
deterioration: Part II – steel corrosion and concrete cracking, ACI Mater. J. 96
(1999) 331–338.
[31] Y.G. Du, A.H.C. Chan, L.A. Clark, Finite element analysis of the effects of radial
Acknowledgements expansion of corroded reinforcement, Comp. Struct. 84 (2006) 917–929.
[32] D.V. Val, L. Chernin, M.G. Stewart, Experimental and numerical investigation of
corrosion-induced cover cracking in reinforced concrete structures, J. Struct.
Financial support from the National Natural Science Foundation Eng. ASCE 135 (2009) 376–385.
of China with Grant Nos. 50808157 and 50920105806, is gratefully [33] H.S. Wong, Y.X. Zhao, A.R. Karimi, N.R. Buenfeld, W.L. Jin, On the penetration of
corrosion products from reinforcing steel into concrete due to chloride-
acknowledged. The authors also appreciate the finical support from
induced corrosion, Corros. Sci. 52 (2010) 2469–2480.
Science and Technology Department of Zhejiang Province with [34] Y.X. Zhao, B.Y. Hu, J. Yu, W.L. Jin, Non-uniform distribution of rust layer around
Grant No. 2010R10099. steel bar in concrete, Corros. Sci. 53 (2011) 300–4308.