Construction and Building Materials 67 (2014) 74–80

Contents lists available at ScienceDirect

Construction and Building Materials

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

Influence of active crack width control on the chloride penetration

resistance and global warming potential of slabs made with fly
ash + silica fume concrete
Philip Van den Heede, Mathias Maes, Nele De Belie ⇑
Magnel Laboratory for Concrete Research, Ghent University, Technologiepark Zwijnaarde 904, B-9052 Ghent, Belgium

h i g h l i g h t s

 Uncracked fly ash + silica fume concrete is very resistant to chloride penetration.

 Uncracked fly ash + silica fume concrete has a long service life (>100 years).
 The seemingly uncracked condition only exists for crack widths below 0.1 mm.
 Limiting the maximum crack width allowed requires more reinforcing steel in concrete slabs.
 More reinforcing steel results in a substantial increase of the slab’s global warming potential.

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

Article history: Service life predictions for concrete exposed to chloride-induced corrosion usually result from durability
Available online 22 November 2013 tests performed on uncracked concrete. Chloride migration coefficients for uncracked concrete should
only be used if the structure can be considered as uncracked. The seemingly uncracked condition requires
Keywords: crack widths below 0.1 mm. The extra reinforcing steel to achieve this in concrete slabs, results in a
Concrete cracking 30–43% increase of the global warming potential. Fly ash + silica fume concrete may be preferred because
Fly ash of its low 28 day migration coefficient (3.4  1012 m2/s), its long service life (>100 years) and its
Silica fume
autogenous healing ability.
Chloride penetration
Service life
Ó 2013 Elsevier Ltd. All rights reserved.
Life cycle assessment (LCA)

1. Introduction for a submerged marine environment according to Eurocode 2

Recent sustainability studies show that concrete’s global warm-
ing potential (GWP) is mainly governed by its binder composition,
strength and service life [1,2]. With respect to the latter, research-
ers are advised to implement data from durability tests into mod-
els that simulate the main deterioration mechanism of the
environment to estimate the concrete’s life span. When looking
at chloride-induced corrosion, the end of service life is often
equaled with steel depassivation. For this failure event the model
of Fib Bulletin 34 [3] based on Fick’s second law, looks straightfor-
ward. Experimental chloride migration coefficients can be used to
estimate when the critical chloride concentration will reach the re-
bars and end service life. However, this approach does not take into
account the unavoidable presence of cracks in concrete due to the
mechanical loads applied. True, a structure should always be de-
signed as such that the maximum allowed crack width (0.3 mm
⇑ Corresponding author. Tel.: +32 9 264 55 22; fax: +32 9 264 58 45.
E-mail address: nele.debelie@ugent.be (N. De Belie).
[6]) is not exceeded. Nevertheless, even 0.3 mm wide cracks in
the tensile zone of a concrete slab – the case study of this paper
– can easily extend beyond the location of the rebars and therefore
offer direct pathways for chlorides. As a consequence, it makes
sense to limit the maximum crack width allowed even more. Of
course, this design approach will have its implications on the
amount of reinforcing steel needed and therefore on the environ-
mental impact of the slab.
In this research, we conducted chloride migration tests on con-
crete representative mortar samples containing a crack of 0.3, 0.2
and 0.1 mm in width. This was done to see whether crack width
reduction could decrease the chloride penetration around the crack
significantly. If not, it may be necessary to aim for very fine crack
widths that can heal autogenously. Jaroenratanapirom and Saha-
mitmongkol reported a fast and complete natural crack closing
for crack widths 60.05 mm in Ordinary Portland cement (OPC)
mortars, OPC + 10% silica fume (SF) mortars and OPC + 30% fly
ash (FA) mortars [7]. Since the binder of the mortar compositions
studied in this paper consisted of a combination of the same

0950-0618/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.conbuildmat.2013.10.032
 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80 75

materials (50% OPC, 40% FA and 10% SF), it is certainly relevant to Replacing the gravel 2/8 and gravel 8/16 by the much finer sand 0/4 obviously
affects the water demand of the mortar. As a result, its required water amount
consider the 0.05 mm crack width criterion here as well. In a next
needs to be adjusted in accordance with the difference in water absorption between
research phase, concrete slabs made with traditional concrete and the gravels and the sand. This can be done with Eq. (2):
fly ash + silica fume concrete were designed according to the most
suitable crack width criterion. Then, a full probabilistic service life Dfwater ¼ fgravel 2=8  Agravel 2=8  fgravel 8=16  Agravel 8=16 þ Dfsand 0=4  Asand 0=4 ð2Þ
prediction cf. Fib Bulletin 34 [3] was performed in the Comrel soft-
with Agravel 2/8, Agravel 8/16 and Asand 0/4 the water absorption coefficients of the coarse
ware [4], followed by a life cycle assessment (LCA) in the SimaPro
aggregates and the sand. The measured water absorption coefficient and the specific
software [5]. These calculations were done to see the effect of crack surface areas are shown in Table 1. The resulting two MBE mortar compositions can
width limitation on the GWP of the studied concrete slabs. be found there as well. By following this method the workability of the MBE mortars
should be identical to the workability of the corresponding concrete mixtures. The
use of MBE mortar instead of concrete normally reduces the material cost and effort
2. Materials and methods
[9].
2.1. Concrete representative mortar mixes
2.2. Manufacture of MBE mortar with an artificial crack
Two concrete compositions were studied (Table 1). Mix T(0.45) has a cement
content and water-to-cement (W/C) ratio of 340 kg/m3 and 0.45, respectively. It 15 Cylindrical specimens (diameter: 110 mm, height: 53 mm) were made for
is seen as the appropriate OPC reference concrete for exposure class XS2 [8]. The each of the two MBE mortar mixes in PVC tube moulds: 3 samples without crack
exposure class corresponds with an environment where steel reinforced concrete plus 3  4 samples containing an artificial crack as a result of putting thin metal
is permanently submerged in sea water. As a consequence, the concrete is exposed plates with a nominal thickness of 0.1, 0.2 and 0.3 mm at a depth of 15 mm in
to chlorides and this can induce steel corrosion. The other concrete mix is charac- the cylindrical moulds just before casting. Fig. 1a shows a schematic of the mould
terized by the same total binder content (340 kg/m3) as the OPC reference. Only setup with the metal plate fixed at the desired crack depth cf. Mu [10].
now it consisted of three different cementitious materials: 50% Portland cement, Note that artificial cracks created by means of thin metal plates are different
40% FA and 10% SF. The water-to-binder ratio (W/B) equaled 0.35 to ensure a from more naturally induced cracks induced by mechanical loading. Both tech-
strength class at least equal to the strength class of the OPC reference (C50/60). niques are in use and have their advantages and disadvantages. By means of thin
By doing so, composition SF(0.35) was characterized by a strength just one strength metal plates it is indeed not possible to reproduce a concrete crack which is realistic
class higher (C55/67). The environmental consequences associated with difference in all its properies. However, it is seen as a very convenient way to study the effect
in strength between the two mixtures under investigation were taken into account of one crack property in particular, being the crack width (which is the main crack
by choosing a strength related functional unit for LCA (see Section 2.7.1). Because of property considered in crack controlled slab design). The reproducibility of cracks
its high cement replacement level, cement related greenhouse gas emissions could created as such is high. On the other hand, displacement steered mechanical load-
be reduced significantly with the latter concrete mix. Therefore, it is seen as a ing to create more natural cracks does not always guarantee the same predefined
potentially ‘green’ concrete type. crack width. Moreover, with the latter method it is difficult to ensure that the crack
An equivalent mortar mix was designed for the two concrete compositions in does not go all the way through the specimen. This condition is required to be able
accordance with the Concrete Equivalent Mortar (MBE) method [9]. Within a to conduct the chloride migration test. With thin metal plates fixed at a certain
MBE mortar mix, the gravel mass fractions of the corresponding concrete height in the sample mould the crack depth is rather easy to control. For this case
mix – in this case fgravel 2/8 and fgravel 8/16 – are replaced with the amount of sand study we therefore adopted the thin metal plate technique. Though, one should re-
Dfsand 0/4 that has the same specific surface. This sand fraction can be calculated main aware of the differences between these artificial cracks and naturally induced
by means of Eq. (1) in which Sgravel 2/8, Sgravel 8/16 and Ssand 0/4 represent the specific cracks. The walls of the voids created with thin metal plates should be considered as
surface areas of the applied coarse aggregates and sand used in the studied concrete cast surfaces. These surfaces are subject to the so-called wall effect which means
mixes. The ratios 2/8, 8/16 and 0/4 refer to the minimum and maximum aggregate that the more fine (usually cementitious) materials will be present in the vicinity
sizes in mm. The first figure of the ratio represents the lower sieve size, while the of the crack walls. Its unhydrated fraction can still react later on and initiate autog-
second figure is the upper sieve size. enous healing. However, this may be the only effect that favors this mechanism for
fgravel 2=8  Sgravel 2=8þ fgravel 8=16  Sgravel 8=16
the artificial cracks. It is also unknown if the fraction of unhydrated materials near
Dfsand 0=4 ¼ ð1Þ the crack walls is sufficient to induce full closure of the crack. For naturally induced
Ssand 0=4
cracks on the other hand, there can be several beneficial effects. There, the crack
width can seriously vary length of a crack. The crack tortuosity and crack wall
roughness will evidently be higher [11]. Moreover, cracks will contain more con-
Table 1 crete particles broken from the surface due to cracking [12]. All these conditions
Concrete compositions, specific surface areas and water absorption coefficients of the contribute to a partial blocking of the crack followed by the autogenous healing
sand and aggregates and MBE mortar mix proportions. phenomenon. These favorable conditions are not present in the artifical cracks. Gi-
ven these differences, a more detailed comparison between artifically and naturally
Concrete composition T(0.45) SF(0.35) cracked specimens in relation to their autogenous healing capacity would certainly
Sand 0/4 (kg/m3) 778 791 be relevant. This investigation is for the moment still ongoing.
Gravel 2/8 (kg/m3) 676 687 After casting, the specimens were kept at a constant temperature and relative
Gravel 8/16 (kg/m3) 447 454 humidity (RH) of 20 °C and 95%, respectively. The metal plates were carefully re-
CEM I 52.5 N (kg/m3) 340 170 moved from the samples after approximately 12 h whereupon the cylinders were
Fly ash (kg/m3) 0 136 demoulded. From then on, they were stored again under the same conditions until
Silica fume (kg/m3) 0 34 the age of 28 days.
Water (kg/m3) 153 119
W/B 0.45 0.35
2.3. Microscopic crack width measurements
FA/B (%) 0 40
SF/B (%) 0 10
After 28 days, the obtained crack widths were measured after mechanical flat-
Sand/aggregate properties Sand 0/4 Gravel 2/8 Gravel 8/16 tening of the cylinders’ troweled surfaces and on the saw cut perpendicular to the
crack of the fourth cylinder of each cracked series. The latter samples were only
Specific surface area (m2/kg) 4.889 0.398 0.194
used for the evaluation of the cross-sectional crack width and not exposed to chlo-
Absorption coefficient 0.008 0.018 0.011
rides. All crack width measurements were done on micrographs taken with a Leica
MBE composition MBE T(0.45) MBE SF(0.35) S8 APO stereo microscope (SM) (magnification: 20) while using the LAS 3.7
software.
Sand 0/4 (kg/m3) 850.8 864.9
After the cross-sectional crack width evaluation with the stereo microscope, the
CEM I 52.5 N (kg/m3) 340 170
cracked area of the non-exposed MBE mortar SF(0.35) with the 0.1 mm wide crack
Fly ash (kg/m3) 0 136
was also subjected to scanning electron microscope (SEM) analysis to study the par-
Silica fume (kg/m3) 0 34
tial closing of the crack more in detail (see Section 3.1). By then, the sample was
Water (kg/m3) 136.5 102.2
196 days old. Three 20  20  10 mm3 prisms containing a cross-section of the
Superplasticizer (ml/kg B) 3.0 14.0
crack, were cut from one cylinder halve and then put in an ultrasonic bath with iso-
W/B 0.40 0.30
propanol to remove all loose particles inside the crack. Afterwards, the samples
FA/B (%) 0 40
were vacuum dried for one week and gold coated by means of a Baltec SCD030
SF/B (%) 0 10
Sputter Coater before being examined in a FEI QUANTA 200F SEM at an accelerating
Strength class C50/60 C55/67
voltage of 20 kV. Secondary electron imaging was used for electron micrography.
76 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80
Fig. 1. Mould setup for creating artificial cracks in the mortar samples (a), expected AgNO3 colour change boundary in a cracked (b) and a seemingly uncracked (c) sample
section.
2.4. Chloride migration tests
A rapid chloride migration test was done on the uncracked MBE mortar samples
in compliance with NT Build 492 [13]. First, the cylindrical specimens were vacuum
saturated in a 4 g/l Ca(OH)2 solution. After 18 ± 2 h of immersion in this solution,
the specimens were fixed inside silicon rubber sleeves with a 0.3 N NaOH (anolyte)
solution on top. The bottom surface of the samples in the sleeves was brought in
contact with a 10% NaCl solution (catholyte). Then, an external electrical potential
was applied axially across each cylinder, which forces the chloride ions to migrate
into the specimens. After 24 h the specimens were removed from the sleeves and
split axially, whereupon a 0.1 M silver nitrate solution was sprayed onto the freshly
split sections. When the white silver chloride precipitation had become clearly vis-
ible, the penetration depth was measured from the center to both edges at intervals
of 10 mm. From the chloride ingress obtained, a non-steady state migration coeffi-
cient can be calculated with Eq. (3):
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
0:0239ð273 þ TÞL ð273 þ TÞLxd
Dnssm ¼ xd  0:0238
ðU  2Þt U2
where Dnssm, U, T, L, xd and t represent the non-steady state migration coefficient
(1012 m2/s), the absolute value of the applied voltage (V), the average value of
the initial and final temperatures in the anolyte solution (°C), the thickness of the
specimen (mm), the average value of the penetration depths (mm) and the test dura-
tion (h), respectively.
The cracked MBE mortar specimens went through the same test procedure ex-
cept for the fact that the external electrical potential was imposed for only 4 h. This
is much less than the normal 24 h test duration. The short test period was chosen to
make sure that the overall chloride ingress would not be more than 13 mm, the
average depth of the artificially induced cracks. Obviously, in case the crack width
is too wide, the chloride penetration close to the crack would be higher than 13 mm
(Fig. 1b). The 10% NaCl solution would almost immediately reach the deepest point
of the crack and chloride migration would start from there on. For each of the spec-
imens containing a 0.1, 0.2 or 0.3 mm crack, it was evaluated whether the chloride
penetration around the crack extended much beyond its deepest point. If not, the
specimen could be considered as uncracked (Fig. 1c) and the Dnssm value measured
for the uncracked specimen would also be valid. As such, service life prediction
could be done using Dnssm values measured on uncracked concrete. However, it also
means that a concrete structure needs to be designed according to the stricter max-
imum crack width criterion.
2.5. Crack width reduction in concrete slabs
When designing a concrete slab for a given mechanical load, an evaluation of
the expected crack width in the service limit state is imperative. Eurocode 2 [6] pro-
vides an Eq. (4) to calculate a characteristic crack width wk (mm) for the structural
element under investigation.
0 1 0 1 from studying a representative mortar of the concrete to the characterization of the
 
U A @rs fct;eff As A
@
wk ¼ 3:4  c þ 0:425  k1  k2  As   kt  As  1þa
Es  Es Ac;eff
A c;eff A c;eff
with c: concrete cover d + 10 mm, factor k1: coefficient accounting for the bond prop-
erties of the reinforcing steel (=0.8 in case of high bond), factor k2: coefficient
accounting for the strain distribution (=0.5 in bending mode), U: diameter of the re-
bar, As: cross-sectional area of the steel (mm2), Ac,eff: effective cross-sectional area of
the concrete in the tensile zone (mm2), rs: steel stress (N/mm2), Es: design value for
the steel’s modulus of elasticity, kt: factor accounting for the load duration (=0.4 for a
long-term mechanical load), fct,eff: concrete’s effective tensile strength (N/mm2), a:
effective ratio of the moduli of elasticity for the steel (Es) and the concrete (Ecm).
Eurocode 2 [6] imposes a characteristic crack width of 0.3 mm for exposure
class XS2. For a given diameter of the rebars, the characteristic crack width can
be reduced by increasing the overall steel cross-sectional area. Thus, more rebars
would be needed to meet stricter crack width criteria.
2.6. Service life prediction
Fib Bulletin 34 [3] is a design code providing the necessary models for a full
probabilistic service life prediction. The design approach consists of defining a suit-
able limit state Eq. (5) containing the necessary load and resistance variables for the
deterioration mechanism under investigation, in this case chloride-induced
corrosion:
" #
d  Dx
C cr ¼ C 0 þ ðC S;Dx  C 0 Þ  1  erf pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5Þ
2  Dapp;C  t
ð3Þ
with Ccr: the critical chloride content (wt.%/binder), C0: the initial chloride content
(wt.%/binder), CS,Dx: chloride content at depth Dx (wt.%/binder), d: concrete cover,
Dx: depth of the convection zone (mm), t: time (years), erf(.): error function and
Dapp,C: apparent coefficient of chloride diffusion through concrete (mm2/years).
The latter coefficient can be obtained from the experimental non-steady state migra-
tion coefficient using Eq. (6):
    a
1 1 t0
Dapp;C ¼ exp be   DRCM;0  kt  ð6Þ
T ref T real t
with be: a regression variable (K), Tref: the standard test temperature (K), Treal: the
temperature of the structural element or the ambient air (K), DRCM,0: the non-steady
state chloride migration coefficient (mm2/years), kt: a transfer parameter, t0: a refer-
ence point of time (years), t: time (years) and a: the ageing exponent. A combination
of (5) and (6) enables an estimation of the time to steel depassivation. Table 2 gives a
quantification of all the input parameters which are normally used in the model.
Note that the critical chloride content (1.9 wt.%/binder, cf. [14]) differs from the
0.6 wt.%/binder prescribed by Fib Bulletin 34 [3]. This value should be valid for sub-
merged OPC concrete with W/C ratios ranging between 0.5 and 0.4. A parameter
study conducted by Van den Heede et al. [2] showed that 1.9 wt.%/binder is proba-
bly a more realistic than 0.6 wt.%/binder. Since it is still uncertain whether the crit-
ical chloride content increases or decreases in the presence of FA or other
cementitious materials [15], the same Ccr value was adopted for both T(0.45) and
SF(0.35). Nevertheless, further investigation on the actual Ccr value for each con-
crete type is imperative. Therefore, specimens with embedded steel reinforcement
bars would need to be manufactured for corrosion potential monitoring of the rein-
forcing steel as a function of the Cl concentration near the steel surface. However,
within the early stages of optimizing a potentially ‘green’ concrete type such as the
suggested SF(0.35) composition, these long-term experiments were not yet per-
formed. An experimentally determined and mix specific critical chloride threshold
value for service life prediction would certainly be of value once the focus can shift
fully optimized concrete mix design.
ð4Þ
The applied ageing exponents are the values suggested by Fib Bulletin 34 [3] for
OPC (0.3) and FA (0.6) concrete in general. These values are not mix specific. It is not
sure whether the latter value is also valid for FA + SF concrete. In Duracrete [14], a
characteristic ageing exponent of 0.62 is given for concrete that contains SF. For
now, we decided to use the lowest of these two values as the ageing exponent of
FA + SF concrete. To estimate the actual ageing exponents per concrete mix, it
would be necessary to perform chloride diffusion tests on specimens that have been
immersed in realistic sea water solutions. Only if one has obtained the correspond-
ing chloride diffusion coefficients at minimum three different ages, it is possible to
quantify the ageing exponent by means of non-linear regression analysis. The latter
rather time consuming methodology was already applied in Van den Heede et al.
 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80
Table 2
Quantification of the input parameters for the probabilistic limit state function as defined by (5) and (6).
Parameter Distribution Mean
Ccr (wt.%/binder) Beta 1.9
C0 (wt.%/binder) Constant 0 –
CS,Dx (wt.%/binder) Normal 3.0
d (mm) Lognormal 40
Dx (mm) Constant 0 –
be (K) Normal 4800
Tref (K) Constant 293
Treal (K) Normal 283
DRCM,0 (mm2/yrs) Normal 336.2 (MBE T(0.45))
107.5 (MBE SF(0.35))
kt Constant 1 –
t0 (yrs) Constant 0.0767 (28d)
a Beta 0.30 (OPC)
0.60 (FA)
[16] to estimate the ageing exponent of high-volume fly ash concrete (a = 0.4). Its
experimental determination for FA + SF concrete is for the moment still ongoing
at our laboratory.
The probabilities of failure (Pf) and reliability indices (b) that result from the
limit state equation defined by (5) and (6) were calculated using the First Order
Reliability Method (FORM) available in the probabilistic Comrel software [4].
According to Fib Bulletin 34 [3], these parameters need to meet the requirements
for the depassivation limit state (Pf 6 0.10 and b P 1.3) to qualify for use in a XS2
environment.
2.7. Life cycle assessment
In compliance with ISO 14,040 [17], the LCA consisted of four major steps: the
definition of goal and scope, the inventory analysis, the impact analysis and the
interpretation.
2.7.1. Definition of goal and scope
This LCA was conducted to quantify the reduction in greenhouse gas emissions
that result from replacing 50% of concrete’s OPC with 40% FA and 10% SF fume while
taking into account the differences in strength and durability with OPC concrete.
Therefore, a concrete slab located in a submerged marine environment carrying a
variable load of 5 kN/m2 was chosen as functional unit. The strength classes of
the two studied MBE mortars (Table 1) were assumed to be similar to the strength
classes of their corresponding concrete compositions. The same strength classes
were considered in the design of the concrete slab (span: 5 m, width: 1 m). Ribbed
steel bars with a diameter of 16 mm and steel quality 500 were used as reinforce-
ments. All design calculations were done in accordance with Eurocode 2 [6]. An-
other reason for choosing a concrete slab as functional unit is because this type
of structural element allows for a crack controlled design. As a result, it can be cal-
culated how much extra reinforcing steel would be needed to obtain a slab where
the concrete in the tensile zone would behave as uncracked cf. the drawing in
Fig. 1c. Only for a slab designed as such, a service life prediction based on the chlo-
ride migration tests performed on uncracked concrete samples would be valid.
Thus, a quantification of the additional environmental impact attributed to the
crack-controlling efforts done to enable the use of chloride migration coefficients
for uncracked concrete as input to the service life prediction models would be very
useful. With our functional unit choice this aspect can be included as well.
2.7.2. Inventory analysis
Per concrete constituent, the necessary inventory data were collected from the
Ecoinvent database [18]. Their proper short descriptions as mentioned in the data-
base together with the amounts used to manufacture 1 m3 of each concrete mix, are
shown in Table 3.
Mean values and standard deviations for the sand and aggregates were calcu-
lated from the amounts of each material needed according to Fuller’s optimal par-
ticle size distribution curve for three deliveries of sand and aggregates to our
laboratory. The probabilistic distribution of these amounts was assumed to be nor-
mal. The required amounts of cement, FA, SF, water and superplasticizer (SP) for
concrete manufacture were assumed to be accurately weighed and therefore con-
sidered as constants. For the allocation of impacts attributed to the industrial by-
products FA and SF, the economic allocation coefficients as proposed by Chen
et al. [19] and Chen [20] were applied (Table 4). For the former by-product, this
is 1.0% of the impact of the coal fired electricity production corresponding with
the production of 1 kg FA. For the latter by-product this is 4.8% of the impact of sil-
icon metal production corresponding with the production of 1 kg SF. SP inventory
data were obtained from an environmental declaration published by the European
Federation of Concrete Admixture Associations [21]. The transport of each
constituent to the concrete plant was not incorporated in the LCA since its environ-
77
Stdv Lower bound Upper bound
0.15 0.2 2.0
– –
0.8 – –
8 – –
– –
700 – –
– – –
5 – –
32.2 – –
20.2 – –
– –
– – –
0.12 0.0 1.0
0.15 0.0 1.0
mental impact is always very case specific. With respect to the steel reinforce-
ments the following Ecoinvent inventory data were used: ‘Reinforcing steel, at
plant/RER U’.
For all Ecoinvent data, unit processes (U) were used in the modelling of each
concrete mix. This was done to enable a full probabilistic uncertainty analysis of
the calculated environmental scores using Monte Carlo simulation.
2.7.3. Impact analysis and interpretation
The IPCC 2007 GWP 100a impact method was used to calculate the Global
Warming Potential (GWP) expressed in CO2 equivalents for a timeframe of
100 years. All calculations were done in the LCA software SimaPro 7.3.3 [5].
3. Results and discussion
3.1. Microscopically measured crack widths
Table 5 shows that the observed crack widths at the surface did
not differ much from the thicknesses of the thin metal plates (0.1,
0.2 and 0.3 mm). The cross-sectional crack widths measured on
one cylinder of each cracked series further confirm this good match
for both mortar compositions. Note that the cross-sectional crack
width observed on the MBE SF(0.35) sample with a 0.1 mm crack
had a rather high standard deviation on the individual values
(stdv.: 0.08 mm). This is mainly due to the fact that the measured
crack widths – recorded over the entire crack depth at regular dis-
tances of 1 mm – quite often equaled zero. At several places along-
side its cross-sectional area, the SM micrographs (age: 28 days)
showed crack closure which could not be attributed to a mere fill-
ing by loose particles originating from the mechanical surface flat-
tening (Fig. 2a). The additional SEM micrographs that were taken
afterwards (age: 196 days) further confirm this observation
(Fig. 2b). The (partial) crack closure practically always consisted
of a bridging with solid material. In the second SEM micrograph,
a very fine crack (<5 lm in width) seems to be going through the
solid crack filling material. Re-cracking was quite often observed
during the SEM analysis. This phenomenon was probably induced
by the sawing operations done to obtain the 20  20  10 mm3
prismatic SEM samples from the MBE SF(0.35)_0.1 mm cylinder
halve. The fact that this very fine crack is going through the whole
local bridging of the crack demonstrates that it was once a solid
crack filling material.
The partial crack closure observed both with SM and SEM may
indicate that for a crack width of ±0.1 mm, partial autogenous
crack healing can occur. Since this phenomenon was not observed
on the MBE T(0.45) sample with a 0.1 mm crack, the partial autog-
enous healing that was detected is most likely induced by further
hydration of the unreacted alternative binders (FA and SF) present
in MBE SF(0.35). Jaroenratanapirom and Sahamitmongkol reported
fast natural healing of cracks P0.05 mm in OPC mortars with 10%
silica fume [7]. Fig. 2 shows that this may also be the case for the
MBE SF(0.35) mortar containing 50% OPC, 40% FA and 10% SF.
78 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80

Table 3
Overview of the life cycle inventory data used per concrete mix.

Material data (kg) Distribution T(0.45) SF(0.35)

Sand, at mine/CH U Normal 766 ± 44 778 ± 45
Gravel, round, at mine/CH U Normal 1135 ± 109 1154 ± 111
Portland cement, strength class Z 52.5, at plant/CH U Constant 340 170
Fly asha Constant 0 136
Silica fumeb Constant 0 34
Tap water, at user/CH U Constant 153 119
Superplasticizer (EFCA 2006) Constant 1.1 5.2
Processing data (kWh) Distribution T(0.45) SF(0.35)
Electricity, low voltage, production BE, at grid/BE U Constant 3.83 3.83
a
Patially contains Ecoinvent data: ‘Electricity, hard coal, at power plant/BE U’, through allocation.
b
Partially contains Ecoinvent data: ‘MG-silicon, at plant/NO U’, through allocation.

width all broken cylinder surfaces (n = 6) looked similar. The chlo-

ride penetration always extended beyond the deepest point of the
Table 4 crack. This was the case for all three nominal crack widths (0.1, 0.2,
Economic allocation coefficients for FA and SF cf. Chen et al. [19] and Chen [20]. 0.3 mm) that were experimentally assessed in this research, also
Product Mass produced Market price Allocation by economic value when already some partial autogenous healing had occurred for
Electricity 1 kWh* 0.1 €/kWh 99.0%
the 0.1 mm crack of MBE SF(0.35). In other words, none of the
FA 0.052 kg 20 €/t 1.0%a studied cracked samples could be considered as uncracked cf.
Si metal 1 kg 1200 €/t 95.2%
Fig. 1c. This means that within the slab design, the maximum crack
SF 0.15 kg 400 €/t 4.8%b width allowed should be reduced even more, preferably to a value
*
that ensures complete autogenous healing of the crack, i.e. the
Equivalent to 0.367 kg of hard coal used to produce electricity.
a 0.05 mm crack width suggested by Jaroenratanapirom and Saha-
Allocation percentages applied to the following Ecoinvent data: ‘Electricity,
hard coal, at power plant/BE U’. mitmongkol [7]. For the latter crack width, complete crack closure
b
Allocation percentages applied to the following Ecoinvent data: ‘MG-silicon, at was observed within 12 days, which is much earlier than the age of
plant/NO U’. our cracked samples at the time of inspection with SM (at 28 days)
and SEM (at 196 days).
3.2. Measured chloride migration coefficients Obviously, it would be interesting to manufacture another ser-
ies of cracked samples containing a 0.05 mm wide crack and sub-
MBE mortar SF(0.35) is characterized by a 28 day non-steady ject them to chloride migration tests as well. However, it is
state chloride migration coefficient Dnssm of 3.4 ± 0.6  1012 difficult to create these very fine artifical cracks in mortar with
m2/s. This is only around one third of the 28 day Dnssm value the current mould setup (Fig. 1a). Metal plates with a thickness
(10.7 ± 1.0  1012 m2/s) obtained for OPC reference MBE T(0.45). of only 0.05 mm have almost no stiffness. As a result, it is practi-
Thus, when uncracked, the former mortar composition is much cally impossible to maintain them at a fixed position in the fresh
more resistant to chloride penetration than the latter. In terms of mortar. Moreover, their intact removal from the hardened mortar
service life, this finding suggests that less rehabilitation actions is also not evident. More research is needed on how this could be
(repair/replacement) will be needed for an uncracked FA + SF fume achieved in the future.
concrete slab exposed to seawater within a 100 years timeframe.
To evaluate this quantitatively, the Dnssm values were expressed
3.4. Concrete slab dimensioning
in mm2/years and used as DRCM,0 input (Table 2) to the service life
prediction model defined by Eqs. (5) and (6).
When following Eurocode 2 [6], the slab made with T(0.45) con-
crete should have a thickness of 150 mm and contain 6 rebars with
3.3. Determination of the maximum crack width allowed a 16 mm diameter. For the slab made with SF(0.35), the slab thick-
ness can be reduced to 140 mm due to the higher strength class of
Fig. 3 shows one representative photo of the chloride penetra- the concrete. The required number of rebars remained the same.
tion for each mortar-crack combination. Per mortar mix and crack Under these conditions, the estimated crack widths in the tensile

Table 5
Microscopic crack width measurements of MBE T(0.45) and MBE SF(0.35).

MBE T(0.45) MBE SF(0.35)

0.1 mm Flattened surface Cross-section 0.1 mm Flattened surface Cross-section
n 33 12 n 33 12
Mean 0.08 mm 0.09 mm Mean 0.13 mm 0.12 mm
Stdv. 0.01 mm 0.01 mm Stdv. 0.03 mm 0.08 mm
0.2 mm Flattened surface Cross-section 0.2 mm Flattened surface Cross-section
n 33 12 n 33 12
Mean 0.19 mm 0.21 mm Mean 0.21 mm 0.20 mm
Stdv. 0.01 mm 0.02 mm Stdv. 0.02 mm 0.03 mm
0.3 mm Flattened surface Cross-section 0.3 mm Flattened surface Cross-section
n 33 12 n 33 12
Mean 0.27 mm 0.34 mm Mean 0.29 mm 0.33 mm
Stdv. 0.02 mm 0.02 mm Stdv. 0.03 mm 0.05 mm
 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80 79

Fig. 2. Partial crack healing of the 0.1 mm wide crack in MBE SF(0.35) as observed on SM and SEM micrographs.

Fig. 3. Observed AgNO3 colour change boundary on the cracked samples of MBE T(0.45) and MBE SF(0.35). (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)

zone equaled 0.18 and 0.21 mm, respectively. To reduce the theo- Table 6
retical crack width of the two slabs to 0.05 mm – the value pro- Influence of active crack width control on the slab dimensioning, service life and
global warming potential of T(0.45) and SF(0.35) concrete.
posed by Jaroenratanapirom and Sahamitmongkol [7] – the
necessary number of rebars would need to be increased to 14 Without crack width control T(0.45) slab SF(0.35) slab
and 15, respectively (Table 6). Slab thickness 150 mm 140 mm
Number of rebars 6  16 mm 6  16 mm
GWP slab 279 ± 30 kg CO2eq 227 ± 14 kg CO2eq
3.5. Service life prediction Service life 25 years* >100 years*
Number of slab replacements 3* 0*
Comparing the reliability indices b and probabilities of failure Pf GWP slab + replacements 1120 ± 116 kg CO2eq* 227 ± 14 kg CO2eq*
with the prescribed criteria (b P 1.3 and Pf 6 0.1) leads to the fol- With crack width control T(0.45) slab SF(0.35) slab
lowing findings. It shows that the estimated service life of MBE Slab thickness 150 mm 140 mm
T(0.45) is much lower than 100 years (25 years). While for MBE Number of rebars 14  16 mm 15  16 mm
SF(0.35) the predefined service life of 100 years (b = 2.7 and GWP slab 367 ± 29 kg CO2eq 325 ± 18 kg CO2eq
Pf = 0.003) is exceeded by far (Table 6). This means that a concrete Service life 25 years >100 years
Number of slab replacements 3 0
slab made of T(0.45) concrete will need a certain number of reha- GWP slab + replacements 1460 ± 119 kg CO2eq 325 ± 18 kg CO2eq
bilitation actions within the 100 years time span. For this case
*
study it was assumed that rehabilitation comprised complete in case concrete could be considered as uncracked (cf. laboratory specimens),
which is not the case in reality.
replacement of the slab, and not a labor intensive local repair.
The concrete and steel needed for three T(0.45) slab replacements
need to be included in the LCA study to have a correct, durability
related environmental impact calculation. Since a 100 years service crete volume for the initial manufacture of the slab needs to be
life seems easily achievable for composition SF(0.35), only the con- considered in the LCA.
80 P. Van den Heede et al. / Construction and Building Materials 67 (2014) 74–80
3.6. Life cycle assessment
When simply looking at the initial production stage – thus,
without consideration of the service life aspect – it is clear that
the SF(0.35) concrete slab (227 ± 14 kg CO2eq) has a lower carbon
footprint (19%) than the T(0.45) concrete slab (279 ± 30 kg CO2eq)
(Table 6). This can mainly be attributed to the high cement replace-
ment level (50%) applied in the former composition. Replacing 50%
of the cement with supplementary cementitious materials (40%
FA + 10% SF) does not result in 50% reduction of the GWP because
still considerable environmental impacts were assigned to FA and
SF through economic allocation. The environmental benefit for
the SF(0.35) concrete in the production stage adds onto the benefit
that can be achieved with its good service life performance
(replacement-free for more than 100 years). This is in contrast with
for instance ultra-high performance concrete (UHPC) for which the
production related CO2 emissions are usually higher when com-
pared with traditional concrete. Only when considering the long
service life of UHPC, there is a substantial decrease in GWP [22,23].
Without active crack width limitation, the GWPs of the T(0.45)
and SF(0.35) concrete slabs (inclusive the required slab replace-
ments within a 100 years life span) amounts to 1120 ± 116 and
227 ± 14 kg CO2eq, respectively. Increasing the amount of reinforc-
ing steel to 14 and 15 rebars to achieve a characteristic crack width
of only 0.05 mm, will increase the environmental score substan-
tially to 1460 ± 119 kg CO2eq (+30%) and 325 ± 18 kg CO2eq
(+43%), respectively (Table 6).
4. Conclusions
In uncracked condition, a concrete representative FA + SF mor-
tar is characterized by a very low chloride migration coefficient
after 28 days (3.4  1012 m2/s) which highly contributes to the
estimated long service life (>100 years) of the material. As a result,
a slab made with the corresponding FA + SF concrete has a much
lower GWP (80%) when compared with a OPC reference concrete
slab.
This very significant decrease in GWP was obtained in a LCA
based on a preliminary service prediction in which not all input
parameters to the probabilistic prediction model for chloride-in-
duced corrosion were mix specific. Within the further develop-
ment of the FA + SF mix design, it would certainly be
recommended to determine every relevant model input parameter
(ageing exponent, critical chloride content, etc.) experimentally.
This way, the applicability of the now available default values
mentioned in the Fib Bulletin 34 model code could be evaluated
and further optimized. It would also enable a more precise GWP
quantification for the concrete slabs in the uncracked condition.
To correct for the unavoidable presence of cracks in these slabs,
the additional amount of reinforcing steel needed to reduce the
characteristic crack width to 0.05 mm, a value that should nor-
mally ensure fast and complete autogenous crack healing, needs
to be included in the life cycle assessment as well. This extra
impact should be taken into account when service life prediction
is based on chloride migration coefficients measured on uncracked
concrete/mortar. The extra reinforcing steel can increase the GWP
of the FA + SF concrete slab with another 43%.
Acknowledgements
The authors would like to thank Ghent University for the re-
search funding. The PhD grant of Mathias Maes is funded by the
Agency for Innovation by Science and Technology (IWT).
References
[1] Van den Heede P, De Belie N. Environmental impact and life cycle assessment
(LCA) of traditional and ‘green’ concretes: literature review and theoretical
calculations. Cem Concr Compos 2012;34:431–42.
[2] Van den Heede P, Maes M, Gruyaert E, De Belie N. Full probabilistic service life
prediction and life cycle assessment of concrete with fly ash and blast-furnace
slag in a submerged marine environment: a parameter study. Int J Environ Sust
Dev 2012;11:32–49.
[3] Fib. Fib Bulletin 34. Model code for service life design. Lausanne: Fib; 2006.
[4] RCP Consulting Software. Comrel, TI, Version 8.10 [computer software].
Munich: RCP Consulting Software; 1987–2013.
[5] PRé Consultants bv. SimaPro. Version 7.3.3 [computer software]. Amersfoort:
PRé Consultants bv; 2011.
[6] CEN. NBN EN 1992–1-1, Eurocode 2: design of concrete structures – Part 1–1:
general rules and rules for buildings (+AC: 2008) (+AC: 2010). Brussels: CEN;
2005.
[7] Jaroenratanapirom D, Sahamitmongkol R. Self-crack closing ability of mortar
with different additives. J Met Mater Miner 2011;21:9–17.
[8] BIN. NBN B15–001. Supplement to NBN EN 206–1. Concrete – specification,
performance, production and conformity. Brussels: BIN; 2004.
[9] Schwartzentruber A, Catherine C. Method of the Concrete Equivalent Mortar
(CEM) – a new tool to design concrete containing admixture (in French). Mater
Struct 2000;33:475–82.
[10] Mu S. Chloride penetration and service life prediction of cracked self-
compacting concrete. PhD thesis. Ghent: Ghent University; 2012.
[11] Akhavan A, Shafaatian SMH, Rajabipour F. Quantifying the effects of crack
width, tortuosity, and roughness on water permeability of cracked mortars.
Cem Concr Res 2012;42:313–20.
[12] Edvardsen C. Water permeability and autogenous healing of cracks in
concrete. ACI Mater J 1999;96:448–54.
[13] Nordtest. NT Build 492. Concrete, mortar and cement-based repair materials:
chloride migration coefficient from non-steady state migration experiments.
Espoo: Nordtest; 1999.
[14] Duracrete. Probabilistic performance based durability design of concrete
structures: general guidelines for durability design and redesign. Gouda: CUR;
2000.
[15] Angst U, Elsener B, Larsen CK, Vennesland O. Critical chloride content in
reinforced concrete – a review. Cem Concr Res 2009;39:1122–38.
[16] Van den Heede P, Maes M, Caspeele R, De Belie N. Chloride diffusion tests as
experimental basis for full probabilistic service life prediction and life-cycle
assessment of concrete with fly ash in a submerged marine environment. In:
Strauss A, Frangopol DM, Bergmeister K, editors. IALCCE 2012: proceedings of
the third international symposium on life-cycle civil engineering. Leiden: CRC
Press; 2012. p. 913–20.
[17] ISO. ISO 14,040. Environmental management – life cycle assessment –
principles and framework. Geneva: ISO; 2006.
[18] Frischknecht R, Jungbluth N, editors. Overview and methodology. Final report
ecoinvent v2.0 No. 1. St-Gallen: Swiss Centre for Life Cycle Inventories; 2007.
[19] Chen C, Habert G, Bouzidi Y, Jullien A, Ventura A. LCA allocation procedure
used as an incitative method for waste recycling: an application to mineral
additions in concrete. Resour Conserv Recy 2010;54:1231–40.
[20] Chen C. A study of traditional and alternative structural concretes by means of
the life cycle assessment method (in French). PhD thesis. Troyes: UTT; 2009.
[21] EFCA. EFCA environmental declaration superplasticizing admixures. EFCA;
2006. <http://www.efca.info/publications.html> (26.05.13).
[22] Habert G, Roussel N. Study of two concrete mix-design strategies to reach
carbon mitigation objectives. Cem Concr Compos 2009;31:397–402.
[23] Habert G, Arribe D, Dehove T, Espinasse L, Le Roy R. Reducing environmental
impact by increasing the strength of concrete: quantification of the
improvement to concrete bridges. J Clean Prod 2012;35:250–62.