Cement & Concrete Composites 31 (2009) 744–753

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Water permeability of engineered cementitious composites

Michael D. Lepech a,*, Victor C. Li b
a
Department of Civil and Environmental Engineering, Stanford University, United States
b
Department of Civil and Environmental Engineering, University of Michigan, United States

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

Article history: The water permeability of a unique class of high performance fiber reinforced cementitious composites
Received 31 January 2008 (HPFRCC) called engineered cementitious composites (ECC) is investigated. These composites are delib-
Received in revised form 18 July 2009 erately tailored using microcmechanical design principles to exhibit pseudo-strain-hardening character-
Accepted 23 July 2009
istics in uniaxial tension, up to greater than 4% strain. While undergoing tensile deformation, microcracks
Available online 28 July 2009
are designed to saturate the specimen rather than localize into large cracks. This tendency to form micro-
cracks, which are experimentally shown to be approximately 60 lm in width, allows ECC material in the
Keywords:
cracked state to maintain water permeability similar to that of uncracked concrete or mortar, and mag-
Engineered cementitious composites
ECC
nitudes lower than cracked reinforced mortar or concrete. It is also shown that the self-healing properties
Strain-hardening of cracks within ECC material significantly aids in reducing the coefficient of permeability of cracked ECC.
Permeability Ó 2009 Elsevier Ltd. All rights reserved.
Self-healing
Durability

1. Introduction the concrete cover resulting in depassivation of the protective layer

covering the steel reinforcement and leading to the oxidation of
A lack of durability is one of the most serious issues facing the reinforcing steel [12,13]. Worldwide, corrosion of reinforcing
reinforced concrete construction subjected to harsh environmen- steel has been identified as the most prevalent and damaging form
tal conditions. One of the most severe concerns is the drastic of deterioration in reinforced concrete structures [14]. Due to this,
decrease of durability associated with concrete cracking. It is well the reduction of chloride transport is vital to improving concrete
known that the durability of concrete structures in harsh environ- durability.
ments is highly dependent upon the transport properties of the Chloride transport can occur chiefly through three mechanisms;
concrete material itself [1–3], therefore the various transport permeation, absorption, or diffusion. Respectively, these three
mechanisms of fluids within concrete have been the focus of many mechanisms are driven by hydrostatic head, capillary action, or
previous studies. Significant work has been done with regard to ion concentration gradient, all of which rarely occur simulta-
the transport mechanisms of permeability, absorption, and diffu- neously. Therefore, the prevalent transport mechanism for any
sion within hardened concrete. While many of these studies have individual structure is highly dependent on environmental sur-
examined uncracked concrete members [4–6], cracks undoubtedly roundings. For permeation, a constant hydrostatic head is required,
form under both environmental and mechanical loads while in making this the transport mechanism most applicable to deep
service. Once cracked, the transport properties of concrete materi- underwater or water retaining structures.
als change drastically [7–10]. A complete understanding of fluid As mentioned, numerous studies have examined the permeabil-
transport within concrete materials, both cracked and uncracked, ity of both cracked and uncracked concrete. The permeability of
is critical to the development of durable reinforced concrete uncracked concrete has been related principally to the particle
infrastructures. density of the composite, typically a function of water to cement
The durability of reinforced concrete is typically associated with ratio. As particle density increases, there are fewer pores allowing
a minimization of both concrete degradation and reinforcing steel for the flow of fluids through the matrix. Powers et al. [15] found
deterioration. The degradation of concrete can occur through such that as the water to cement ratio decreases, water permeability de-
mechanisms as freeze thaw cycles, alkali–silicate reaction (ASR), creases. This relation of low permeability to high density is one
fatigue stress, or sulfate reaction [11]. Steel deterioration is typi- reason that high strength concretes, with significantly higher par-
cally the result of either chloride penetration or carbonation of ticle packing than normal concretes, are considered highly durable.
However, high strength concretes, in comparison to normal
* Corresponding author. Tel.: +1 650 724 9459; fax: +1 650 723 7514. strength concretes, show much higher levels of thermal shrinkage
E-mail address: mlepech@stanford.edu (M.D. Lepech). [16] and autogenous shrinkage [17]. Together with a high Young’s

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.07.002
 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753 745

modulus, low creep coefficient, and higher brittleness, these high for which maximum limits are set depending on the type of envi-
strength concretes are far more likely than normal strength con- ronmental exposure. The limit imposed within the AASHTO code
cretes to crack at an early age, and produce larger crack widths corresponds to a minimum reinforcement spacing resulting in a
[18]. Once cracked, the permeability of concretes, regardless of ma- crack width limit of approximately 350 lm. Within the American
trix density or compressive strength, increases remarkably. As Concrete Institute (ACI) structural concrete design code [25] a
crack width grows from 100 lm to 500 lm, the permeability of maximum crack width is no longer explicitly calculated, however
concrete material has been shown to increase up to seven orders previous versions of the design code have suggested a maximum
of magnitude [19]. crack width of 300 lm for outdoor exposure. As discussed previ-
The addition of fibers can aid in maintaining a lower permeabil- ously, at these relatively wide crack widths, the permeability of
ity within cracked concrete. Studies performed by Tsukamoto [20] concrete material is nearly five orders of magnitude greater than
found that permeability within cracked concrete composites is that of sound concrete. While these crack width limits are impor-
proportional to the laminar flow rate between two parallel plates, tant in reducing the transport of water and corrosives into rein-
as shown in forced concrete, they are far larger than allowable to effectively
block corrosives from rapidly deteriorating the reinforcing steel,
qR ¼ n  qG ð1Þ
and potentially the entire reinforced concrete structure. Through
where qR is the flow rate through a crack (between two rough the use of ECC material, which exhibits inherently tight crack
walls), n is the flow rate coefficient dependent on the type and tor- widths under large tensile deformations, there exists the possibil-
tuosity of cracking, and qG .ed slot. This laminar flow rate increases ity of designing ECC cover materials with lower permeability and
proportionally with the cube of the distance between the parallel reinforced ECC structures with higher durability.
plates (i.e. crack width). In this work, using micromechanics-based design methods, the
Tsukamoto reported that the addition of fibers to concrete leads theoretical basis for tailoring materials for tight crack width and
to a decrease in the flow rate coefficient, n, due to a reduction in correspondingly low permeability is established. Mechanical tests
cracking tendency and increased microcracking. The incorporation are then carried out to verify the ability of these new materials
of fibers also produced lower crack widths in fiber reinforced con- to maintain tight crack widths under uniaxial tension load. Finally,
crete (FRC) compared to plain concrete. Similarly, Stang and Aarre experimental permeability tests are carried out to compare the
[21] found that the addition of fibers to concrete leads to signifi- performance of new materials with steel reinforced cementitious
cantly lower crack widths than seen in either plain or convention- materials and establish a fundamental crack width versus perme-
ally reinforced concrete. The effect of these two phenomenon, a ability relationship for multiple cracking materials.
decrease in the flow rate coefficient along with a decrease in crack
width, acting simultaneously as a result of the addition of fibers al-
low for the possibility to reduce the permeability of cracked con- 2. Material design methodology
crete materials to a level similar to uncracked concrete.
Engineered cementitious composites (ECC) is a unique type of The primary design focus behind the development of cementi-
high performance fiber reinforced cementitious composite (HPFRCC) tious composites with low permeability in the cracked state lies
which features high tensile ductility with moderate fiber volume behind work performed by Wang et al. [19], Aldea et al. [26], Law-
fraction, typically 2% by volume [22]. Of special interest is the capa- ler et al. [7], and Rapoport et al. [27]. Wang et al. [19] reported that
bility of ECC material to deform to high tensile strains, commonly as crack width increases from 100 lm to 500 lm, the permeability
over 4%, while designed to maintain very tight crack widths. These coefficient increases nearly seven magnitudes from 1.0  1011 m/
cracks have experimentally been shown on the order of 60–80 lm s to 1.0  104 m/s. However, for crack widths under 100 lm, the
[23]. This high performance is possible through the use of steady permeability coefficient remains nearly identical to that of sound
state crack models which provide quantitative links between meso- concrete, suggesting that for crack widths below this threshold
structural properties such as fiber bridging properties and matrix there is no significant increase in permeability after cracking. The
toughness, and composite mechanical behavior such as steady state development of cementitious materials with low permeability
cracking stress and maximum crack width. To accommodate large after cracking can be accomplished by designing for maximum
deformations, rather than forming a small number of cracks which crack widths below 100 lm, even after substantial deformation
widen with increasing load as seen in concrete or tension-softening and crack formation.
FRC, ECC forms numerous microcracks which allow the material to The seemingly diametric objectives of holding maximum crack
undergo pseudo-strain-hardening. Once initial microcracks widen widths below 100 lm while still accommodating large deforma-
to the typical 60 lm width at roughly 1% composite straining, addi- tions can only be accomplished through the formation of large
tional microcracks form to accommodate further deformation while numbers of microcracks, which functionally spread the deforma-
maintaining this tight crack width. This phenomenon is shown in tion out over a large area. As mentioned earlier, ECC material can
Fig. 1. Consequently, crack width in ECC material is more a result be tailored to develop numerous closely spaced microcracks in a
of matrix and fiber interaction rather than steel reinforcement ratio. strain-hardening response when subjected to tensile loading. This
This inherent maximum crack width can be seen as a material prop- tailoring is rooted in micromechanical design principles which fo-
erty, similar to compressive strength or elastic modulus, rather than cus on each of the three phases within the composite; mortar ma-
a structural property, such as reinforcing ratio in reinforced con- trix, fiber, and the interface between them.
crete. This uncoupling of the transport properties of ECC material The first priority when designing ECC material for low perme-
to the required structural performance in any application (i.e. not ability is to ensure the formation of multiple cracks and strain-
requiring heavy reinforcement for crack control when not necessary hardening behavior under load. This allows large deformations to
for load resistance), allows designers significantly greater flexibility be distributed over numerous cracks, each exhibiting a crack width
in meeting design challenges. below the 100 lm threshold discussed previous. The basis of mul-
Reinforced concrete design codes currently do not specify pre- tiple cracking and strain-hardening within ECC is the propagation
cise limits on transport properties, such as a maximum concrete of steady state cracks which were first characterized by Marshall
permeability, or a maximum crack width under load. The American and Cox [28], and extended to fiber reinforced cementitious com-
Association of State and Highway Transportation Officials (AASH- posites by Li and Leung [29] and Lin et al. [30]. By forming steady
TO) [24] design codes relies on the computation of a ‘‘Z” factor state ‘‘flat cracks” which maintain a constant crack width while
746 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753

Fig. 1. Uniaxial stress strain curve and crack development of PVA–ECC M45.

propagating, rather than Griffith-type cracks which widen during

propagation as in typical tension-softening fiber reinforced cemen-
titious materials, ECC material exhibits multiple cracks with small
widths which saturate the specimen while undergoing strain-hard-
ening during extreme tensile deformation. The formation of multi-
ple steady state cracking is governed by the bridging stress versus
crack width opening relation along with the cracking toughness of
the mortar matrix. To achieve this phenomenon the inequality
shown in Eq. (2) must be satisfied.
Z d0
K 2m
J 0b ¼ r0 d0  rðdÞdd P Jtip  ð2Þ
0 Em
where J0b is the complimentary energy shown in Fig. 2, r0 and d0 are
the maximum crack bridging stress and corresponding crack open-
ing, Jtip is the fracture energy of the mortar matrix crack tip, Km is
the fracture toughness of the mortar matrix, and Em is the elastic
modulus of the mortar matrix. In addition to the fracture energy cri-
terion, a strength criterion expressed in Eq. (3) must be satisfied.
r0 > rfc ð3Þ Fig. 2. Crack bridging stress versus crack opening relation.

where r0 is the maximum crack bridging stress and rfc is the first
cracking strength of the mortar matrix. For saturated multiple and strain-hardening performance, can be realized. However, in
cracking, Wang and Li [31] found that Eq. (3) must be satisfied at addition to forming these cracks, the material must also be de-
each potential crack plane, where rfc is understood as the cracking signed to exhibit crack widths below the 100 lm threshold limit.
stress on that crack plane. This can be achieved through tailoring of the crack bridging versus
Once an ECC mixture is selected which sufficiently meets the crack opening relation referenced in Eq. (2). Shown in Fig. 2, the
two above criteria, the formation of multiple steady state cracks, maximum steady state crack width exhibited during ECC multiple
 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753 747

cracking is d0, the crack width corresponding to the maximum ical design tools outlined above to achieve the desired strain-hard-
crack bridging stress, r0. If the crack width were to grow beyond ening response. To minimize the mortar matrix fracture toughness,
d0, the crack bridging stress would begin to fall, in which case no large aggregates are used, and the silica sand has an average and
the crack would localize and multiple crack formation would cease. maximum grain size of 100 lm and 200 lm, respectively. The
By keeping d0 below the 100 lm threshold, the ECC material can poly-vinyl-alcohol (PVA) fibers are purposely manufactured with
exhibit multiple cracking and strain-hardening performance, while a tensile strength, elastic modulus, and maximum elongation
maintaining crack widths small enough to limit permeability matching those needed for strain-hardening performance. Addi-
through the cracked material. tionally, the surface of the PVA fibers is coated 1.2% by weight with
Lin et al. [30] proposed the formulation of the crack bridging a proprietary hydrophobic oiling agent to tailor the interfacial
stress versus opening relationship based on single fiber pullout properties between fiber and matrix for strain-hardening perfor-
performance. This relation in expressed in mance [32]. Specific attributes of the PVA fibers used in this study
Z p=2 Z ðLf =2Þ cos /  are shown in Table 2.
4V f In addition to ECC, the permeability of cracked and uncracked
rðdÞ ¼ PðdÞef / pð/ÞpðzÞdz d/ ð4Þ
pd2f /¼0 z¼0 reinforced (with steel wire mesh) mortar specimens was also
tested as a control test series. The mix proportions of the mortar
where Vf is the fiber volume fraction, df is the fiber diameter, / is the specimens are also shown in Table 1. The components of this mate-
orientation angle of the fiber, Lf is the fiber length, z is the centroidal rial are Type I Portland cement, water, and a standard Ottawa sand.
distance of a fiber from the crack plane, f is a snubbing coefficient, To generate different crack widths and crack spacing among the
and p(/) and p(z) are probability density functions of the fiber ori- mortar samples, the level of reinforcement within the specimens
entation angle and centroidal distance from the crack plane, respec- was varied. The reinforcement ratios of the mortar samples and
tively. P(d) is the pullout load versus displacement relation of a their test designations is shown in Table 3. Mixing of all ECC and
single fiber aligned normal to the crack plane, also described in mortar materials within this study was done using a force-based
Lin et al. [30]. Hobart mixer.
Empirical properties in Eq. (4), such as the fiber snubbing coef-
ficient, have been determined for ECC materials using a variety of
fibers [32]. By varying fiber volume fraction, fiber diameter and 3.1. Test specimen preparation and test execution
length, fiber–matrix interface properties (P(d) behavior), and fiber
orientation (1D, 2D, or 3D), a theoretical composite can be de- For determination of micromechanical characteristics of the
signed to undergo large deformations, up to several percent, by ECC fiber–matrix interface, particularly J 0b and the bridging stress
generating large numbers of microcracks. By limiting the numeri- versus crack opening relation, single strands of PVA fibers were
cal value of d0 (the maximum value of r(d) from Eq. (4)) within pulled from the M45 matrix under displacement control. Single fi-
the composite, maximum crack width is controlled and ductility ber pullout tests allow for the determination of the fiber pullout
is achieved without sacrificing low permeability due to crack load versus pullout displacement curve, P–d, used to compute the
growth and localization. Maximum crack width d0 is not exceeded composite bridging stress versus crack opening relation. This test-
in the composite as a result of the energy and strength multiple ing follows procedures outlined by Redon et al. [33]. Single fiber
cracking constraints (Eqs. (2) and (3)) which ensure that before a pullout specimens were tested after 14 days of wet curing to corre-
crack in ECC reaches a width of d0, an adjacent microcrack will spond with the time at which the permeability specimens would
form and relieve the ambient stress level. The application of these be tested in uniaxial tension.
material design procedures allow materials engineers to carefully Determination of matrix cracking energy on ECC M45 matrices
match material characteristics to specific structural demands, such was performed using notched beams. Prisms of ECC matrix (identical
as strain capacity and low permeability. Using the theoretical con- to ECC material but without fibers) measuring 75 mm  38 mm 
stituent and composite properties determined from the r(d) re- 300 mm were cast and cured in air for 14 days. After curing, a
sponse, actual mix proportions can be developed to fit theoretical 2 mm wide  30 mm deep notch was cut at the midspan of each
requirements of fiber volume fraction, fiber–matrix interface prop- prism. Prisms were tested in bending to determine the load at failure
erties, fiber orientation, etc. (i.e. crack propagation load). The matrix toughness, Km, was calcu-
lated using the appropriate geometric calibration factors.

3. Experimental procedure
Table 2
Micromechanical experimental testing began on ECC material Mechanical properties of PVA fiber.
to verify that it meets both the conditions of strain-hardening per- Nominal Apparent Diameter Length Elastic Elongation
formance, and maximum crack widths below 100 lm. Testing was strength strength (lm) (mm) modulus (%)
done on an ECC mix designs designated Mix 45. The mix propor- (MPa) (MPa) (GPa)
tions for this version of ECC are shown in Table 1. The components 1620 1092 39 12 42.8 6.0
of ECC material are similar to typical fiber reinforced cement com-
posites (FRCC), consisting of Type I Portland cement, sand, class F
fly ash, water, fibers, and a high range water reducing agent
Table 3
(HRWR). However, unlike typical FRCCs, the proportions within
Reinforced mortar and ECC steel reinforcement ratios, observed crack widths, and
ECC are carefully determined through the use of the micromechan- observed crack spacing.

Specimen Reinforcement Crack width Crack spacing

Table 1
series ratio (lm) (mm)
Mix proportions of Ecc M45 and standard mortar (kg/m3).
R/Ma-a 0.019 350–500 20–30
Material Cement Aggregates Fly ash Water HRWRa Fiber R/M-b 0.023 200–350 10–25
ECC M45 578 462 694 319 7.51 26 R/M-c 0.028 125–200 5–15
Mortar 614 1534 – 215 – – ECC 0 40–80 2–5
a
a
High range water reducer. R/M denotes reinforced mortar specimen series.
748 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753

To characterize permeability as a function of crack width for a crack widths and crack spacing among the specimens, various
wide range of crack widths in ECC and reinforced concrete materi- amounts of steel reinforcing mesh (ranging from 1.9% to 2.9%)
als, both ECC and reinforced mortar permeability testing speci- was cast into the plates. All reinforced mortar specimens were de-
mens were cast. These were then loaded in uniaxial tension to formed in uniaxial tension to 1.5% of the specimen gauge length.
produce multiple crack patterns of various widths for a constant Reinforcement levels were varied a number of times until a repre-
specimen strain (elongation). ECC permeability specimens were sentative group of specimens with a variety of crack widths and
cast into tensile testing plates with cross sectional dimensions of spacing was achieved, once again shown in Table 3.
12 mm  75 mm. Tensile specimens measure 300 mm in length
with a gauge length of 180 mm during testing. Tensile specimens 3.2. Permeability testing
were air cured for 14 days prior to uniaxial tension testing. The
testing regime followed that proposed by Li et al. [32]. Following At an age of 28 days, the tensile specimen edges were sealed with
this testing, ECC specimens were kept in water for an additional epoxy to facilitate unidirectional flow through the cross section. Due
14 days to ensure complete water saturation prior to further test- to the length of time associated with this type of testing, crack width
ing. Cracks widths in all specimens were measured load at 30 indi- permeability measurements were performed in the unloaded state.
vidual locations after removing all on the specimen using an To conduct permeability testing, two experimental setups were
optical microscope with an internal crack width gauge. Permeabil- used. A falling head test was used for specimens with a low perme-
ity testing began at 28 days. A deformation limit was chosen a pri- ability, while a constant head test was used for specimens with a per-
ori as an appropriate permeability test level for comparison with meability too high to practically use the falling head test. These two
the control mortar specimens. An initial set of ECC specimens setups are shown schematically in Fig. 4a and 4b, respectively. The
was prepared which was strained to a value of 1.5%. At 1.5% defor- falling head and constant head permeability test setups have been
mation, the reinforced mortar specimens (with reinforcing ratios adapted from Wang et al. [19] and Cernica [34].
ranging from 1.9% to 2.9%) had developed large cracks, and had The permeability of specimens in the falling head test can be
failed in some cases. Details for both the reinforced mortar and determined using Eq. (5), while the permeability of specimens in
ECC specimens are given in Table 3. For the second round of ECC the constant head test can be determined using
permeability testing, separate ECC specimens were strained from  
a  L h0
0% to 3%. Representative stress versus strain curves for the various k¼ ð5Þ
A  t f hf
ECC testing series can be found in Fig. 3.
Reinforced mortar specimens were cast into tensile plates with V L
k¼ ð6Þ
geometry identical to the ECC tensile plates. To generate differing A  h0  t f

Fig. 3. Uniaxial tensile stress versus strain curves for ECC M45 deformed to: (a) 1.5%, (b) 2%, and (c) 3%.
 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753 749

imens with a single crack can result in very low water

permeability, regardless of the true rate of water permeation.
Therefore, it is more appropriate to use Eqs. (5) or (6) on a cracked
material such as ECC, which exhibits tightly spaced microcracks
after large deformations creating a more homogeneous material,
rather than reinforced or plain concrete or mortar even though
both may be cracked. For comparison with results of other concrete
materials research, the permeability coefficient for concrete or ECC
should be normalized by the number of cracks within the specimen
to produce a coefficient of permeability per crack, or normalized
permeability, and thereby remove the influence of specimen cross
sectional area.

4. Experimental results and discussion

Results from testing of the micromechanical parameters sug-

gest that the proposed ECC M45 mixes should exhibit strain-hard-
ening characteristics, and maintain crack widths below the critical
100 lm threshold. The critical ratio of J 0b =J tip lies above the value of
unity, at approximately 1.08. While this ratio does not fall far
above the strain-hardening cutoff, the current modeling of single
fiber pullout (shown as Eq. (4)) does not take into account two-
directional fiber pullout, or spalling of the matrix from pullout of
angled fibers, both of which would serve to increase the compli-
mentary energy of the composite. Therefore, this value of J0b =J tip
for M45 is assumed to be a lower bound. As seen from Fig. 1,
ECC M45 displays excellent strain-hardening characteristics, and
the observed crack widths of ECC M45 range from 40 lm to
70 lm in the loaded state, well below the 100 lm limit outlined
previously. When unloaded for permeability testing, ECC micro-
cracks are observed to close approximately 10% of their loaded
width (between 5 lm and 10 lm) and are reported as such for
experimental permeability test results.
Overall findings for the water permeability comparison series
between ECC and reinforced mortar samples are shown in Fig. 5
plotting maximum crack width in each material specimen versus
the coefficient of permeability. These results are summarized in
Table 5. As expected, the magnitude of permeability through the
cracked reinforced mortar increases rapidly with an increase in
crack width. For reinforced mortar specimens, water permeability
ranges from 4.58  1011 m/s for uncracked mortar specimens up
to 4.46  104 m/s for crack widths of approximately 500 lm. This
extreme variation, nearly seven orders of magnitude, has a pro-
found impact on the overall performance of reinforced concrete
in service. Historically, ACI [25] and AASHTO [24] design codes
Fig. 4. (a) Falling head permeability test setup and (b) constant head permeability have recommended crack control reinforcement to limit crack
test setup. widths under 300 lm and 350 lm, respectively. Yet even at this
crack width, water permeability has increased six orders of magni-
tude, approximately 1.00  105 m/s.
where k is the coefficient of permeability, a is the cross sectional Many concrete service life models (i.e. Life-365 [35]) assume
area of the standpipe, L is the specimen thickness in the direction that the concrete remains uncracked throughout the majority of
of flow, A is the cross sectional area subject to flow, tf is the test service life, and once cracked it has little resistance to the penetra-
duration, h0 is the initial hydraulic head, hf is the final hydraulic tion of water-borne corrosives resulting in rapid degradation.
head, and V is the volume of liquid passed through the specimen However, it is common for reinforced concrete to crack long before
during the test. Geometric quantities used in these tests are given the end of service life due to free or restrained shrinkage stresses,
in Table 4. thermal stresses, or simple overloading. The occurrence of any one
However, Eq. (5) and Eq. (6) inherently assume homogeneous of these phenomena during service life is practically guaranteed
flow through the material by dividing the rate of flow by the exper- resulting in early age cracking, and premature ingress of corrosives.
imental specimen cross sectional area. The use of very large spec- The durability of reinforced concrete is known to decrease drasti-
cally upon contact between corrosives and reinforcing steel. The
Table 4 wide cracks with large spacing in reinforced concrete have also
Geometric constants of permeability testing setup. been shown to form macro-cell corrosion potentials [36], resulting
Quantity a L A in severe pitting corrosion of the reinforcing steel and spalling of
the concrete cover, undermining the basic assumptions within ser-
Value 2.84  105 m2 0.012 m 8.93  103 m2
vice life models.
750 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753

Fig. 5. Coefficient of permeability versus crack width for ECC and reinforced mortar series deformed to 1.5% in uniaxial tension.

Table 5
Water permeability of cracked and uncracked ECC and reinforced mortar specimens.

Series Crack width (lm) No. of cracks Permeability (m/s) Normalized permeability (m/s per crack)
R/M (uncracked) 0 0 4.58  1011 ± 1.2  1011 4.58  1011 ± 1.2  1011
R/M-a 150 ± 15 17 3.02  108 ± 7.5  109 1.77  109 ± 4.4  1010
R/M-b 240 ± 18 11 7.97  106 ± 2.3  106 7.24  107 ± 2.1  107
R/M-b 300 ± 10 9 2.58  105 ± 6.7  106 2.87  106 ± 7.4  107
R/M-c 500 ± 50 5 4.46  104 ± 4.2  104 8.92  105 ± 8.4  105
ECC (uncracked) 0 0 8.18  1012 ± 2.4  1012 8.18  1012 ± 2.4  1012
ECC (1.5%) 50 ± 10 47 1.95  1010 ± 5.6  1011 4.15  1012 ± 1.2  1012
ECC (2%) 52 ± 10 59 3.00  1010 ± 6.8  1011 5.08  1012 ± 1.2  1012
ECC (3%) 63 ± 10 87 7.74  1010 ± 1.9  1010 8.90  1012 ± 2.2  1012

As designed, the ECC specimens subjected to 1.5% strain prior to

testing exhibited a much lower coefficient of permeability due to
the small crack widths. Cracks within the ECC specimens opened
to a maximum width of approximately 60 lm. This resulted in a
permeability coefficient of 1.95  1010 m/s. This low permeabil-
ity, nearly five orders of magnitude smaller than similarly strained
mortar specimens with ACI or AASHTO recommended maximum
crack widths, should serve to significantly improve the service-life
performance of any structure constructed with ECC material which
is exposed to water pressure.
While the results of ECC permeability measurements may seem
to show an increase in permeability by one order of magnitude
over the uncracked ECC and mortar, this result may be misleading.
Cracked concrete permeability work performed by others
[19,26,7,27], has been conducted primarily on specimens with a
single crack. As mentioned previously Eqs. (5) and (6) inherently
assume homogeneous flow through the material by dividing the
Fig. 6. Coefficient of permeability versus level of tensile strain in ECC.
rate of flow by the experimental specimen cross sectional area.
For comparison with results of other researchers, the permeability
coefficient are normalized by the number of cracks within the 7.74  1010 m/s. Even as the number of cracks along the specimen
specimens to produce a coefficient of permeability per crack or grows, the permeability of the specimens does not change signifi-
normalized permeability. Once this is done, results between ECC cantly, due to the inherently tight 60 lm crack widths maintained
material, with numerous 60 lm cracks, and concrete, with one by the ECC material. This is evidenced by the crack saturation
60 lm crack (1.00  1011 m/s from Wang et al. [19]), match rea- shown in Fig. 1. After normalizing by the number of cracks (from
sonably well. Table 5), the difference among ECC specimens strained to higher
The second series of ECC permeability specimens demonstrates levels becomes even less significant. Once normalized by the num-
that, in addition to exhibiting low permeability at relatively small ber of cracks, the water permeability of ECC specimens strained to
strains (i.e. 1.5%), ECC material can exhibit low water permeability 2% and 3% uniaxial tension falls to 5.08  1012 m/s and
at higher levels of tensile strain. Regardless of the level of tensile 8.90  1012 m/s. This is also shown in Fig. 6.
deformation, ranging from 1.5% to 3.0%, water permeation through Throughout the course of testing, it became apparent that the
the cracked ECC specimens remained very low. This is shown low crack widths exhibited among ECC specimens was not the only
graphically in Fig. 6. The water permeability of the ECC specimens cause for the low water permeability of ECC specimens. As shown
strained to 2% and 3% uniaxial tension is 3.00  1010 m/s and in Fig. 7, the rate of permeation through the ECC specimens
 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753 751

pores on the surface of the ECC specimen. Fig. 8c and d show a

cracked reinforced mortar specimen before and after permeability
testing, respectively.
The effect of self-healing of cracks on permeability has been
investigated by other researchers [37], and may be significant in
the permeability determination of cracked ECC. This can be attrib-
uted primarily to the large binder content and relatively low water
to binder ratio within the ECC mixture. The presence of significant
amounts of unhydrated binders allows for autogeneous healing of
the cracks when exposed to water. This mechanism is particularly
evident in cracked ECC material due to the small crack widths
which facilitate self-healing [37]. However, this phenomenon is
not observed while cracked ECC specimens are simply saturated

Fig. 7. Development of permeability for ECC strain to 1.5%, 2% and 3%.

dropped drastically from the initial values until asymptotically

reaching the recorded value, even though the crack widths during
permeability testing do not change. This phenomenon can be par-
tially attributed to achieving complete saturation and further den-
sification of the matrix throughout the testing period. However,
ECC specimens were saturated in water for 14 days prior to perme-
ability testing at an age of 28 days. By the time of testing, the spec-
imens should have been nearly, if not completely, saturated and
continuing to undergo little matrix hydration. Throughout the
course of permeability testing, a white residue formed within the
cracks and on the surface of the specimens near the cracks. These
formations are shown in Fig. 8. Fig. 8a shows a saturated ECC spec-
imen immediately prior to the beginning of permeability testing,
while Fig. 8b shows the same specimen after permeability testing. Fig. 9. ESEM surface chemical composition analysis of self-healing crack
The white residue forms both within the cracks, and within the formations.

Fig. 8. Appearance of ECC permeability specimens: (a) before permeability testing and (b) after permeability testing and reinforced mortar specimens; (c) before permeability
testing and (d) after permeability testing.
752 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753

in water. During the 14 days of saturation prior to permeability Within the permeability testing, these carbonates were introduced
testing, cracked ECC specimens showed no signs of autogeneous by the constant addition of water which flows through the speci-
healing of the cracks. After only 3 days in the permeability testing mens. In the case of the partially submerged specimens, the small
apparatus, evidence of self-healing became apparent. A similar amount of carbon dioxide dissolved at the water surface was suffi-
phenomenon was also seen when cracked ECC specimens were cient to cause limited self-healing at that location. However, in the
partially submerged in water. Crack healing was only exhibited absence of this constant carbonate supply, as in the saturation
near the surface of the water, while no healing was observed out- tanks prior to permeability testing, no self-healing of the ECC
side of this near-surface region. microcracks can occur. Ultimately, the formation of these crystals
Surface chemical analysis of the self-healing ECC specimens slows the rate of permeation through the cracked composite and
using an environmental scanning electron microscope (ESEM) further reduces the permeability coefficient.
show that the crystals forming within the cracks, and on the sur-
face adjacent to the cracks, are hydrated cement products, primar- 5. Conclusion
ily calcium carbonate (Fig. 9). These crystal formations within the
self-healed cracks are shown in Fig. 10. To facilitate healing of the Within the study, the design of an ECC composite which al-
cracks, and promote formation of calcium carbonate, a flow of lowed for large tensile deformations while not sacrificing a low
water containing carbonates or bicarbonates must be present. coefficient of permeability was completed. These seemingly dia-
metric goals were achieved by developing a composite which
exhibited multiple microcracks under tension. By saturating the
ECC material with microcracks of approximately 60 lm width,
the unloaded permeability coefficient of the material is not com-
promised, and remains low even after the formation of numerous
microcracks and tensile straining up to 3.0%. When compared to
reinforced mortar specimens cracked under uniaxial tension to
1.5% deformation, ECC material exhibits permeability up to six or-
ders of magnitude lower.
The tight crack widths in ECC are possible by using microme-
chanics as a tool for designing low permeability ECC composites
which meet the two critical criteria of forming multiple cracks un-
der load and ensuring that the maximum of the fiber bridging
stress versus crack opening relationship (r–d) for the composite
occur below a crack width opening of 100 lm. This relationship
can also be used as a guide for tailoring the fiber, matrix, and fi-
ber/matrix interface within the composite to meet the low perme-
ability criteria. Through the use of micromechanical analysis, a
version of ECC material was identified which theoretically meets
the two above criteria, and was experimentally proven to exhibit
low permeability in the cracked state. Additionally, through the
formation of small microcracks and in the presence of water flow,
a significant amount of self-healing was observed within the ECC
cracks which aides in further reducing the permeability coefficient
of the cracked ECC composite.
Through the use of this materials design process, in which
material properties are tailored to meet a specific structural de-
mand, the links between microstructure and structural perfor-
mance are further clarified. Although this approach focused
primarily on the development of a low permeability cementitious
material in the cracked state for durability concerns in this paper,
the application of micromechanics to tailor material properties and
achieve desirable structural performance can be broadly applied
within the civil engineering community.

Acknowledgements

This research was partially funded through an NSF MUSES Bio-

complexity Program Grant (CMS-0223971 and CMS-0329416).
MUSES (Materials Use: Science, Engineering, and Society) supports
projects that study reducing adverse human impact on the total,
interactive system of resource use, the design and synthesis of
new materials with environmentally benign impacts on biocom-
plex systems, as well as maximizing the efficient use of individual
materials throughout their life cycles.

References

Fig. 10. (a) Autogenous self-healing crystalline formations in ECC crack after [1] Ithuralde G. Permeability: the owner’s viewpoint. In: Yves Malier, editor. High
permeability testing and (b) ECC crack before permeability testing. performance concrete: from material to structure; 1992. p. 276–94.
 M.D. Lepech, V.C. Li / Cement & Concrete Composites 31 (2009) 744–753 753

[2] Shah SP, Wang K, Weiss WJ. Is high strength concrete durable? In: Gjørv OE, [19] Wang K, Jansen D, Shah S, Karr A. Permeability study of cracked concrete.
Sakai K, editors. Concrete technology for a sustainable development in the 21st Cement Concr Res 1997;27(3):381–93.
century; 2000. p. 102–14. [20] Tsukamoto M. Tightness of fiber concrete. Darmstadt Concr 1990;5:215–25.
[3] Buenfeld NR. Measuring and modeling transport phenomena in concrete for [21] Stang H, Aarre T. Evaluation of crack width in FRC with conventional
life prediction of structures. In: Glanville J, Neville A, editors. Prediction of reinforcement. Cement Concr Compos 1992;14:143–54.
concrete durability: proceedings of STATS 21st anniversary conference; 1995. [22] Li VC. On engineered cementitious composites (ECC) – a review of the material
p. 77–90. and its applications. J Adv Concr Technol 2003;1(3):215–30.
[4] Perraton D, Aitcin PC, Vezina D. Permeabilities of silica fume concrete. In: [23] Weimann MB, Li VC. Drying shrinkage and crack width of ECC. Brittle matrix
Permeability of concrete ACI SP-108. Detroit (MI): American Concrete composites 7, Warsaw, Poland; October 2003. p 37–46.
Institute; 1988. p. 63–84. [24] American Association of State Highway and Transportation Officials. AASHTO
[5] Powers TC. Structure and physical properties of hardened Portland cement LFRD bridge design specifications: second edition. Washington (DC, USA):
paste. J Am Ceram Soc 1991;41(1):1–6. AASHTO; 1998. p. 5–41.
[6] Reinhardt HW, Hearn N, Sosoro M. Transport properties of concrete. In: [25] American Concrete Institute. Building code requirements for structural
Reinhardt HW, editor. Penetration and permeability of concrete: RILEM report concrete and commentary. Detroit (MI): ACI; 2002.
16; 1997. p. 213–64. [26] Aldea CM, Shah SP, Karr A. Effect of cracking on water and chloride
[7] Lawler JS, Zampini D, Shah SP. Permeability of Cracked Hybrid Fiber-Reinforced permeability of concrete. J Mater Civil Eng 1999(August):181–7.
Mortar under Load. ACI Mater J 2002;99(4):379–85. [27] Rapoport J, Aldea CM, Shah SP, Ankenman B, Karr A. Permeability of cracked
[8] Hearn N. Effect of Shrinkage and Load-Induced Cracking on Water Permeability steel fiber-reinforced concrete. J Mater Civil Eng 2002(July/August):355–8.
of Concrete. ACI Mater J 1999;96(2):234–41. [28] Marshall DB, Cox BN. A J-integral method for calculating steady-state matrix
[9] Gerard B, Reinhardt HW, Breysse D. Measured transport in cracked concrete. cracking stresses in composites. Mech Mater 1988(8):127–33.
In: Reinhardt HW, editor. Penetration and permeability of concrete: RILEM [29] Li VC, Leung CKY. Theory of steady state and multiple cracking of random
report 16; 1997. p. 265–331. discontinuous fiber reinforced brittle matrix composites. ASCE J Eng Mech
[10] Gerard B, Jacobsen S, Marchand J. Concrete cracks II: observation and 1992;118(11):2246–64.
permeability – a review. In: Gjorv, Sakai K, Banthia N, editors. Concrete [30] Lin Z, Kanda T, Li VC. On interface property characterization and performance
under severe conditions 2, environment and loading, vol. 1, E&FN Spon; 1998. of fiber reinforced cementitious composites. J Concr Sci Eng, RILEM
p. 183–97. 1999;1:173–84.
[11] Nowak AS, Szerszen MM. Life-cycle deterioration models for concrete deck [31] Wang S, Li VC. Tailoring of pre-existing flaws in ECC matrix for saturated strain
slabs. In: Life-cycle performance of deteriorating structures, ASCE Publication; hardening. In: Proceedings of FRAMCOS-5, Vail, Colorado, USA; April 2004. p.
August 2003. p. 133–40. 1005–12.
[12] Tuuti K. Corrosion of steel in concrete. Stockholm: Swedish Cement and [32] Li VC, Wu C, Wang S, Ogawa A, Saito T. Interface tailoring for strain-hardening
Concrete Research Institute; 1982. PVA–ECC. ACI Mater J 2002;99(5):463–72.
[13] Bakker RFM. Initiation period of corrosion. In: Schiessl P, editor. RILEM report: [33] Redon C, Li VC, Wu C, Hoshiro H, Saito T, Ogawa A. Measuring and modifying
corrosion of steel in concrete; 1988. p. 22–55. interface properties of PVA fibers in ECC matrix. ASCE J Mater Civil Eng
[14] Ciampoli M (chair). Aging effects, maintenance, and structural health 2001;13(6):399–406.
monitoring. In: Corotis et al., editors. Structural safety and reliability, Swets [34] Cernica JN. Geotechnical engineering. New York: Holt, Reinhart & Winston;
& Zeitlinger; 2001. 1982 [p. 97–9].
[15] Powers TC, Copland LE, Hayes JC, Mann HM. Permeability of Portland cement [35] ‘‘Life-365 Service Life Prediction Model: Computer Program for Predicting the
paste. ACI J Proc 1954;51(3):285–98. Service Life and Life-Cycle Costs of Reinforced Concrete Exposed to Chlorides”
[16] Mehta PK. Durability–critical issues for the future. Concr Int Silica Fume Association. 2001.
1997;19(6):27–33. [36] Raupach M, Dauberschmidt C. Macrocell corrosion of reinforcement in the
[17] Wittmann FH. Crack formation and life cycle performance. In: 5th areas of cracks of high performance concrete. In: CONSEC ’01: third
international conference on fracture mechanics of concrete structures, Vail, international conference on concrete under severe conditions. Vancouver,
Colorado, USA; 2004. p. 3–10. BC, Canada; June 2001. p. 1506–13.
[18] Li VC, Stang H. Elevating FRC material ductility to infrastructure durability. In: [37] Evardsen C. Water permeability and autogenous healing of cracks in concrete.
Proceedings of 6th RILEM symposium on FRC, Varenna, Italy; 2004. p. 171–86. ACI Mater J 1999;96(4):448–54.