Cement & Concrete Composites 55 (2015) 153161

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Tensile behaviour of early age concrete: New methods of investigation

Emmanuel Roziere, Rachid Cortas, Ahmed Loukili
LUNAM Universit, Ecole Centrale de Nantes, GeM (Research Institute of Civil Engineering and Mechanics), CNRS UMR 6183, 1 rue de la No, BP 92101, 44321 Nantes Cedex 3, France

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

Article history: The assessment of the tensile properties of early-age concrete is essential for reducing the risk of cracking
Received 27 October 2013 due to restrained shrinkage. The tensile strain capacity of concrete, which was dened as a measure of the
Received in revised form 17 June 2014 ability of the material to withstand deformation without cracking, is useful but few data can be found in
Accepted 16 July 2014
available literature and the measure of the displacements of concrete is sometimes questionable. New
Available online 28 September 2014
direct tensile testing apparatus and experimental procedure were designed to provide reliable data on
concrete specimens. The measure of displacements was deduced from digital image correlation. They
Keywords:
enabled determining a stressstrain relationship of concrete before cracking. The results showed a very
Concrete
Tensile strength
rapid increase of strength from the end of setting. The evolution of the tensile strain capacity showed a
Early age minimum corresponding to the period that includes the setting time and early hardening, thus this is a
Tensile strain capacity critical stage for plastic shrinkage cracking. Even if the values are closely linked to the boundary condi-
Shrinkage tions and experimental procedure, the effect of aggregate type could be investigated.
2014 Elsevier Ltd. All rights reserved.

1. Introduction especially aggregate type on tensile properties of early age

Tensile stresses are likely to be generated in concrete even
when no external load is applied. They are due to restrained
shrinkage and often lead to cracking. Cracking affects strength,
durability, and aesthetics of concrete, thus crack opening has to
be minimized. In order to predict the risk of cracking, several prop-
erties of concrete and their evolution must be known, such as
shrinkage, strength, elastic modulus, creep, coefcient of thermal
expansion, and fracture mechanics properties [1,2]. Shrinkage can
be reduced though the optimization of mix design and appropriate
curing. The cracking tendency also depends on the tensile strength
and the tensile strain capacity of concrete. At early age, the risk of
cracking is generally high because of the signicant increase of
shrinkage and the relatively low strength of concrete [3,4].
In standards, concrete specications mainly deals with cement
(or binder) content and water to cement (or water to binder) ratio.
These composition parameters actually have a strong effect on
shrinkage magnitude and kinetics. For a given set of specications,
i.e., for a given evolution of shrinkage, the risk of cracking actually
depends on the tensile behaviour. The specications on the
composition of concrete do generally not include specications
on aggregates, whereas they have a major inuence on the tensile
properties and cracking. This study aims at developing a new
procedure to assess the inuence of mixture proportioning and
Corresponding author.
E-mail address: ahmed.loukili@ec-nantes.fr (A. Loukili).
concrete.
Few experimental data are available to assess the tensile prop-
erties of early age concrete, dened here as the rst 24 h. A uniaxial
direct tensile bench has been developed. Concrete specimens are
cast in two-part steel moulds. The external load is applied to one
part of the mould and the other part is linked to the test frame.
The displacement of concrete is measured through digital image
correlation (DIC). The tensile strain capacity is dened as the strain
when the main crack appears. Experiments were carried out in
endogenous conditions.
The rst part of this paper deals with a review of the existing
literature about the measurement of tensile properties of early-
age concrete. It shows the lack of data and the need for further
study. Then the tensile testing machine and the experimental pro-
cedure are described. The analysis of the experimental data is
detailed and the results are discussed.
2. State of the art
2.1. Direct tensile testing procedures
Drying and hydration cause signicant deformations of early
age concrete, such as plastic shrinkage. The value of 1000 lm/m
is proposed in the literature [5,6] to characterize a high risk to
cracking of the material. Beside the measurement of plastic shrink-
age during the rst hours after casting, the need for assessing the

http://dx.doi.org/10.1016/j.cemconcomp.2014.07.024
0958-9465/ 2014 Elsevier Ltd. All rights reserved.
154 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161
tensile strain capacity and the strength of concrete has been high-
lighted. Several direct tensile tests have been designed.
The main characteristics of direct tensile tests are given in
Table 1. Referred studies deal with early age concrete (024 h) as
well as hardened concrete. Tests on early age concrete mostly con-
sist in horizontal uniaxial tests [3,4,711], as fresh concrete speci-
mens cannot carry the loads due their own weight. From one day
tests on hardened concrete are done vertically [12,13]. Concrete
specimens generally have a prismatic shape and a central part with
a reduced cross section to promote failure in the middle section
[3,4,711,13]. This reduction of the cross section is steep for tests
rigs inspired by geotechnical testing [3,710,13], whereas tests rigs
inspired by materials science [4,11] show curved transitions which
reduce the stress concentration and allow a better assessment of
strains. The dimensions of the concrete specimens are of the same
order of magnitude, namely: the cross-sectional area is about
10,000 mm2 and the total length is several hundreds of millime-
tres. The gauge length to the width of the central part ratio is gen-
erally between 1 and 2.
As far as early age concrete is concerned, several experimental
problems must be solved. As test rigs are horizontal, the frictions
between the specimens and the frame must be reduced. Materials
with a low coefcient of friction and roller bearings [4,10] are gen-
erally used. The most effective way would be using air-bearing
plates [9,11].
Several methods can be used to grip the specimens, namely:
moulds with variable cross section [3,811], shear keys [9,10],
embedded steel bars [4,1214], gluing and lateral gripping. It is
worth to note that gluing is not recommended because of poor
bonding between steel plates and moist concrete. Whatever the
method, the fabrication and the preparation must be accurate
enough to (i) reduce the stress concentration (ii) avoid load eccen-
tricity and eliminate any bending moment during testing. Concrete
specimens are generally cast in moulds in two parts that grip the
sample when the tensile load is applied. The measure of the dis-
placements on the specimen itself is difcult before or during the
setting of concrete [9,10]. Thus external measures (on the mould)
are sometimes preferred to avoid damaging early-age concrete
[11], but they do not take into account the possible displacements
of the specimen in the mould. Measures without contact are rec-
ommended. Because of the relative displacements between the
concrete specimens and the mould, the actual tensile displacement
rate is likely to differ from the controlled loading rate [10]. In pub-
lished studies, the loading rate varies between 0.033 lm/s [4] and
20 lm/s [7]. These variations are likely to affect the results of ten-
sile testing. Moreover the loading rate should be consistent with
loading due to shrinkage. Plastic shrinkage depends on the envi-
ronmental conditions (temperature, RH, wind speed, etc.) and con-
crete mixture proportions (paste volume, water-to-binder ratio,
etc.). The order of magnitude of plastic shrinkage of concrete with
high paste volume (such as SCC) is 1000 lm/m in 5 h or 200 lm/
(m h) i.e., 20 lm/h for a 0.1 m gauge length. Thermal shrinkage
can be due to the removal of formwork and the temperature gradi-
ent between air and hydrating concrete. A 25 C/h temperature
drop can be observed in the case of massive concrete structures.
Assuming a thermal dilation coefcient of 2  105/C, this results
in a thermal shrinkage of 500 lm/(m h) or 50 lm/h for a 0.1 m
gauge length. Thus the tensile displacement rate of concrete would
be approximately 14  103 lm/s.
Some results of the cited studies are plotted in Fig. 1. Two sets
of data can be distinguished: three studies only deal with very
early age concrete [7,9,11], from 2 to 7 h and two studies were
extended to intermediate ages from 2 to 12 h [4,7]. The variations
of the tensile strain capacity are of several orders of magnitude.
Kasai et al. reported a decrease of the tensile strain capacity during
the rst hours with a minimum at 10 h. The results of Hannant and
Dao also indicated a continuous decrease from 1 to 5 h. Mix design
is a very important parameter inuencing the minimum strain
capacity and the time of minimum strain capacity. After this min-
imum (between 5 and 10 h), the strain at peak stress can be
assumed to increase with time. Setting occurs during this stage
and signicantly affects the properties of concrete. However some
differences between values from different studies show the inu-
ence of the test procedure. The shape and size of the specimen,
curing conditions, loading rate, and boundary conditions actually
affect the results. The other studies deal with concrete aged of
140 days [12,20,21]. The tensile strain capacity is about
100 lm/m and the variations between different sets of data are
lower. The procedures mainly consist in direct vertical tensile tests.
The review shows relatively few data about concrete aged from
7 to 24 h. Only 5 of the reviewed publications have data. A compre-
hensive data set of tensile properties of concrete from 2 to 24 h
could not be found in existing published data. This is not due the
lack of interest for knowledge about these properties, but this
comes from the experimental problems to be overcome in such
studies. Further studies are needed and must concentrate on reli-
able measurements of the displacements of concrete specimens.
2.2. Inuence of aggregates on concrete tensile behaviour
The published data only deal with hardened concrete. They
show a signicant inuence of aggregate properties on concrete
tensile behaviour.
The strength of concrete depends on the bonding between
aggregates and paste and a linear correlation of compressive
strength with the adherence between paste and aggregates was
observed [22]. The quality of bonding is inuenced by the petrogra-
phy, the strength, and the roughness of aggregates [23]. The inter-
facial transition zone (ITZ) between paste and aggregates has
actually been shown to have properties that signicantly differ
from the surrounding cement paste [24]. The properties of ITZ
depend on the microstructure of aggregates and three kinds of
bonding can be distinguished, namely: physical, chemicalphysical,
and mechanical bonding [25]. Physical bonding can be observed
when there is no chemical interaction between paste and aggre-
gates, such as quartz. This leads to poor bonding, thus the ITZ is
the weakest zone of the material. The chemicalphysical bonding
is due to chemical reactions between paste and aggregates. This
mainly occurs with limestone aggregates. Calci-carbo-aluminate
C3ACaCO311H2O is formed, which results in denser interfacial
zone [26,27]. Mechanical bonding can be observed between rough
or porous aggregates and cement paste. The products of cement
hydration are actually formed in the pores of the aggregates.
The properties of aggregates signicantly affect the tensile
strain capacity of concrete [19]. Crushed aggregates lead to higher
tensile strain capacity than gravels. The tensile strain capacity also
depends on the aggregate type. Concrete made of limestone aggre-
gate would show slightly higher tensile strain capacity than con-
crete made of quartzite. The modulus of concrete is actually
affected by the aggregate type. The tensile strain capacity has been
shown to decrease linearly with the modulus of aggregate. Pub-
lished data on the elastic modulus of various aggregate types are
available [28,29]. The modulus of quartzite was 57 and 59 GPa
and the modulus of limestone ranged between 39 and 57 GPa.
3. Experimental program
3.1. Experimental procedures
3.1.1. Direct tensile testing
Fig. 2 shows the tensile testing machine. The load is applied
horizontally by an electric displacement-controlled actuator. The
 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161 155

load cell is placed between the actuator and the mobile part of the
mould. The mould actually comprised two halves with curved

Steel bars embedded

Gauges on concrete
Swaddiwudhipong

transitions to a central part with reduced constant section,

0.017, 0.083, 0.50

1, 3, 7, 14, 28 d
100 mm high, 100 mm long, and 105 mm wide (Fig. 3). The inner

in concrete
100  100 face was covered with PTFE and protected with thin polyethylene
et al. [12]
Vertical

86140
sheets during the test. The two parts of the mould are linked to the

1.63.2

1874
Prism

frame and the load cell by spherical pin connections.

500

200

X
The concrete specimen was cast in the steel mould just after
mixing then vibrated in one layer. An external frame was placed
LVDT: relative displacements of on the plate to avoid displacements of the two halves of the mould
due to the vibration and pressure of fresh concrete. The specimen
was covered with a polyethylene sheet to avoid drying. The ambi-
ent temperature was 20 C. 1 h before the test, the polyethylene
Split steel mould
split steel mould

sheet was removed and the specimen was covered with white then
Dao et al. [11]

1 h356 h25

0.0010.089

3106722 black painting (Fig. 4). This resulted in a contrast that enabled the

0.520.60
Dog bone

70  100

digital image correlation (DIC).

3245
Horiz.

The loading rate of the actuator was 5 lm per second. A loading

0.83
430

70

rate of 0.5 lm/s was also applied to study the effect of the displace-
ment rate. The displacements of the mould and the concrete spec-
LVDT on cells embedded

Stabilized crushed rocks

Control: 21.4 Measure:

Split steel mould with

imen were actually monitored by two different devices. Linear

Nahlawi et al. [10]

variable displacement transducers (LVDT) gave the relative dis-

placements of the two halves of the mould (Fig. 3). The displace-
ment of two regions of the central part with reduced section was
in concrete

0.0200.75
shear keys
0.04265
1, 7, 28 d
70  100

monitored without any contact by two cameras placed vertically

Horiz.
Cubes

above the frame (Figs. 3 and 4). The failure actually occurred in this
200

60

part of the specimen.

The frequency of image shot was 1 Hz. The size of black and
LVDT on cells embedded

Split steel mould with

white images was 320  240 pixels. The displacements were

deduced from the images through DIC [30]. Digital images without
Hannant et al. [9]

any treatment were used as input data of DIC software VIC2D. Ini-
Control: 12.5

0.0020.160
in concrete

tial calibration was required to convert the displacements (given in

shear keys

5025000
70  100

pixels) in micrometres (lm). One pixel corresponds to 26 lm. The

Horiz.
Cubes

15 h

resulting accuracy of measured displacements is 2 lm, which gives

0.50
208

50

an uncertainty of approximately 20 lm/m for calculated strains.

Split steel mould with
Abel and Hover [8]

3.1.2. Dynamic monitoring of the modulus of elasticity

The FreshCon device which was developed at the University of
Stuttgart was used herein to monitor the early age behaviour
shear keys
100  102

0.010.7

of concrete using ultrasonic waves [31]. This device is composed

0.40.7
Horiz.
Cubes

28 h

of two polymethacrylate (PMMA) walls which form a U-shaped

225

mould (Fig. 5). These two walls are separated by a distance

d = 5.9 cm. The U-shaped mould has high damping properties in
Dog bone

186000
et al. [7]

211 h

order to absorb the waves passing trough the mould and moving
mould
Horiz.
Kasai

520

Steel

around the concrete sample (concrete volume: 450 cm3). The test
X
X

was carried out in an air-conditioned room at 20 C and that the

LVDT on keys embedded

Steel bars embedded in

concrete sample was protected from desiccation by the use of a

plastic foil. A data acquisition system allowed to obtain the ultra-
Hammer et al. [4]

sonic velocity, the energy and the frequency spectrum evolution.

The evolution of both, the dynamic Poisson ratio tdyn and the
in concrete
0.0332.67
100  100

2008500

dynamic elastic modulus Edyn, is obtained using the ultrasonic

Dog bone

concrete
0.010.3
812 h

device. Their values are obtained as a function of the compression

Horiz.

0.35
498

200

and shear wave velocities using Eqs. (1) and (2) [33,34] :
X

1
 v 2p  v 2s
tdyn 2 1
Steel mould

v 2p  v 2s
Ravina and
Shalon [3]

100  100

0.010.15

0.51.05
Horiz.
Cubes

05 h
230

1 tdyn  1  2tdyn
Edyn v 2p  qc 
X

2
1  tdyn
Tensile strength: min

Tensile strain capacity

Effective cross-section

Compressive strength

where qc is the density of fresh concrete, tp is the compression

Loading rate (lm/s)
Gauge length (mm)

displacements

wave velocity and ts the shear wave velocity.

Direct tensile tests.

(L  H, mm2)

max (MPa)
Length (mm)

Measure of

(lm/m)

X: not given.

3.1.3. Evolution of the degree of hydration

W/C ratio
Direction

(MPa)
Gripping

The degree of hydration and the free water content of the stud-
Shape
Table 1

Age

ied materials were obtained by using a Thermo-Gravimetric Anal-

ysis (TGA). The experiments were performed on cement pastes
156 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161

10000
Tensile strain capacity, m/m

[4] [7]
[9] [11]
[12] [14]
[15] [16]
1000 [17] [18]
[19]

100

10
0.01 0.10 1.00 10.00 100.00
Age, days
Fig. 3. Measure of displacements on the mould and the concrete specimen.
Fig. 1. Tensile strain capacity from different sets of data. (See above-mentioned
references for further information.)

obtained by sieving fresh concrete just after mixing. Before the

test, the materials were kept in sealed conditions at 20 C. Two sol-
vents (ethanol and ether) were used to stop hydration at different
ages: 3, 6, 9 and 24 h. Then cement pastes were crushed and sieved
to prepare TGA samples. The mass of the samples was approxi-
mately 100 mg.
The TGA was carried out under inert atmosphere (helium) over
a temperature range 251000 C with a heating rate of 20 C/min.
This experiment allows obtaining the amount of free water from
the mass loss between 25 C and 150 C [35]. The amount of chem-
ically bounded water w(t) was obtained from the mass loss
between 150 C and 600 C Dm145  C!600  C and the cement content
mC (Eq. (3)). The cement content of the material was deduced from
the water-to-cement ratio W/C of the concrete mixture (Eq. (4)).
The degree of hydration a (%) was calculated by using Eq. (5):

Dm145  C!600 C
wt 3
mC

msample
mC 4
1 W=C

Fig. 4. Photo of the device and examples of acquired images.

wt
a%  100 5
w1
3.2. Concrete mixtures
where w(t) (%) is the amount of chemically bounded water at the
age t of the material, w(1) (%) is the amount of chemically bounded The concrete mixtures are made of three different aggregate
water for a total hydration of the cement contained in the material. types (Table 2) and Portland cement (Table 3). The compositions
w(1) was calculated from the chemical composition of the cement of the concrete mixtures are given in Table 4. As the study deals
using the Bogue formula [36]. with the inuence of aggregate type, the water, cement, and sand

1: Steel mould
2 3
1 4
2: Concrete specimen

3 5 3: Spherical pin connection

6 4: Electric actuator

5: Load cell

6: LVDT2

Fig. 2. Tensile testing machine.

 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161
contents were kept constant. The superplasticizer content was
adjusted to reach a constant slump.
4. Results and discussions
4.1. Analysis of testing data
The control of the actuator did not allow the displacement-
control of the concrete specimens because displacements between
the specimen and the mould were possible. Thus the data from the
test (Fig. 6) showed the response of a brittle material in tension.
The relative displacements in the central part of the specimen
were assessed as the difference between displacements of regions
image a and image b. The assessment of the displacements of the
specimen from the LVDTs would lead to overestimating the defor-
mations of the specimen. This would explain relatively high values
found in literature [11]. However LVDTs provided useful informa-
tion. When no bending occurred during the direct tensile test,
the LDVT curves were actually identical before cracking.
When the loading rate of the test was 0.5 lm/s, the actual load-
ing rate in the central part of the specimen was 0.04 lm/s
(150 lm/h), which is consistent with plastic shrinkage kinetics
[6,37,38].
The stress was deduced from the measured load and the cross
section (100  105 mm2). It was actually assumed that the stress
distribution was uniform over the central cross section. The
stressstrain can be deduced from the data recorded before the
failure of the specimen (Fig. 6). The loading rate of the test corre-
sponding to Fig. 6 was 5 lm/s. This curve was used to determine
the tensile strength ft and the tensile strain capacity ectu.
Only the rst part of the curves (before the failure) was used.
Complete stressdisplacement curves were given but these studies
were done at earlier ages (before 7 h) [9,11]. At 10 h, the materials
began to show a brittle behaviour and it was not possible to mon-
itor the post-peak stage.
4.2. Repeatability and inuence of displacement rate
The results showed in the following section deal with the effect
of aggregate type. Thus it is necessary to investigate the repeatabil-
ity of the test in order to be able to conclude on signicant effects.
Two series of tensile tests were carried out at 7, 10, 15, 20 and 24 h.
The tensile strength curve is plotted in Fig. 7. The loading rate of
the tests corresponding to Fig. 7 was 5 lm/s. The average differ-
ence between two strength values at the same age was 6%, the
minimum difference is 2% and the maximum is 11%. The maximum
difference will be plotted on following strength curves.
The tensile strain capacity is given in Fig. 8. The average
difference between two strength values at the same age is 21%,
Fig. 5. FreshCon device [32]. (a) Computer with DAQ card, (b) amplier, (c)
piezoelectric sensor, (d) container, and (e) preamplier.
157
the minimum difference is 18% and the maximum is 46%. At 7
and 10 h the concrete actually showed very brittle behaviour.
The displacement at failure is only a few times higher than the
accuracy of the measure of displacements (2 lm). This set of data
is not sufcient to conclude on repeatability. However this part
of the study mainly aims at showing that repeatability tends to
be better for strength than tensile strain capacity.
The duration of the test had to be much lower than the age of
concrete, but the speed had to be consistent with loading due to
plastic or thermal shrinkage. The controlled loading rate of 5 lm/
s was chosen. It resulted in actual speed of 0.032 lm/s, i.e., 56
times higher than the early age shrinkage of concrete. Another con-
trolled rate of 0.5 lm/s was used to investigate the inuence of
loading rate. It resulted in actual speed of 0.021 lm/s. The data
are plotted in Figs. 9 and 10. The tensile strength was not affected
by the loading rate. However the duration of the test signicantly
increased (5900 s at 24 h), which made it impossible to obtain reli-
able values of the tensile strain capacity. As early age concrete was
studied, the autogenous shrinkage and the creep strain were likely
to be of the same order of magnitude [39]. The results plotted in
Fig. 10 actually showed that the tensile strain capacity signicantly
increased with a decrease in the displacement rate.
4.3. Inuence of aggregate type on tensile strength and strain capacity
Fig. 11 shows the evolution of the peak stress of specimens of
concrete made of different aggregate types (Table 4). The tensile
strength increased very rapidly from the age of 7 h, which approx-
imately corresponded to the end of setting (assessed from the Vicat
test). The Vicat test was carried out according to the EN 196-3
standard on mortar samples extracted from fresh concrete using
a 5-mm sieve and a vibrating table. The end of setting was dened
as the time when the Vicat needle rst stopped at height higher
than 38 mm (i.e., less than 2 mm below the top surface of the mor-
tar sample). The gravel type did not signicantly inuence the ten-
sile strength. The compressive strengths of the three concrete
mixtures were actually very close (Table 4). Limestone aggregates
are reported to improve the tensile strength of hardened concrete
by chemicalphysical effects on the ITZ. However the kinetics of
the chemical reactions is not known and the reaction is likely to
remain limited at early age. Finally, the strength of early age con-
crete would be mainly inuenced by the evolution of the cement
paste, whereas the aggregate type or the bonding between paste
and aggregates would not have a signicant effect.
The evolution of the tensile strain capacity is plotted on a log
scale in Fig. 12. Whatever the aggregate type, the tensile strain
capacity decreased of several orders of magnitude between 3 and
7 h. This had been observed in previous studies [4,9,11]. During
setting, the development of hydration products actually links the
cement grains and the aggregates to form a continuous solid skel-
eton. A minimum strain capacity of 2030 lm/m was measured at
7 and 10 h. Then it increased to reach 152, 180 and 185 lm/m at
24 h for porous limestone, siliceous (quartz), and dense limestone
aggregates respectively. These values are consistent with pub-
lished data [12,19]. Also, it was observed that the tensile strain
capacity (measured from 1 day to 28 days) was not signicantly
affected by either the mix proportions or the curing age of concrete
[12]. Published data shows that the tensile strain capacity depends
on the proportion of coarse aggregate and the mean size of aggre-
gate [20]. The three concrete mixtures S, L1, and L2 were designed
keeping the mean size and volume of coarse aggregate constant
(Table 4).
As a consequence the most interesting result is the minimum of
the tensile strain capacity vs. age curve. These ages correspond to a
critical stage for the risk of cracking of concrete at early age. More-
over the aggregate type was expected to inuence the modulus,
158 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161

Table 2 1.8
Properties of aggregates.
1.6 ft = 1.70 MPa L1
Mineralogy Water absorption WA24 Density 1.4

Stress, MPa
(%)
1.2
Sand 0/4 mm 1
Sea sand Diverse 0.6 2.58
Crushed limestone Dense 0.8 2.65 0.8
1 limestone 0.6
Gravels 4/20 mm 0.4
Siliceous Quartzite 0.8 2.59 Tensile strain capacity
Crushed limestone Dense 0.7 2.65
0.2 ctu = 201 m/m
1 limestone 0
Crushed limestone Porous 3.2 2.46 0 100 200 300 400
2 limestone Strain, m/m

Fig. 6. Strain stress curve of concrete L1 at 24 h.

Table 3
Properties of Portland cement CEM I 52.5.
1.8
Chemical analysis (%) 1.6 L1 (1)

Tensile strength (ft), MPa

SiO2 19.9
Al2O3 5.5 1.4 L1 (2)
Fe2O3 2.4 1.2
CaO 64.0
MgO 1.9 1.0
SO3 3.1 0.8
Free lime 0.67
Na2O equivalent 0.41 0.6
Loss on ignition 1.7 0.4
Compound composition of clinker (%) 0.2
C3S 58.1
C2S 10.0 0.0
C3A 11.8 0 4 8 12 16 20 24 28
C4AF 7.5 Age, hours
Physical properties
Blaine neness (cm2/g) 3680 Fig. 7. Tensile strength from two series of tests.
Normal compressive strength (MPa) 65.3
Density (kg/m3) 3.14
content of L2 was actually lower and this could explain earlier
increase of Edyn, as the superplasticizer has generally a retarding
effect on setting. The concrete mixtures made of quartz and dense
Table 4 limestone aggregate showed the highest values at 24 h. The modu-
Composition and properties of concrete mixtures (kg/m3). lus of the concrete with porous aggregate was signicantly lower.
Concrete S Concrete L1 Concrete L2 These results are consistent with the published data about the mod-
ulus of aggregates and their expected effect on the modulus of
Aggregates 4/20 mm
Siliceous 1053
concrete.
Crushed limestone 1 1078 In order to conrm the evolution plotted in Fig. 12, the tensile
Crushed limestone 2 1000 strain capacity was assessed from the tensile strength/elastic mod-
Sea sand 0/4 mm 367 367 367 ulus ratio (Eq. (6)).
Crushed limestone 1 0/4 mm 377 377 377
Cement CEM I 52.5 N 350 350 350
Water (wneta) 175 175 175 ft
Superplasticizer 1.8 1.8 1.3
eten 6
Edyn
W/C 0.50 0.50 0.50
Slump (mm) 160 160 180 The values of the tensile strain capacity eten at 24 h are of the
Compressive strength same order of magnitude as the values obtained from direct assess-
fc,1d (MPa) 19.5 20.5
fc,2d (MPa) 30.5 31.4
ment ectu (Fig. 14). These results also showed that the tensile strain
fc,28d (MPa) 50.9 52.5 50.0 capacity of 1-day concrete was not signicantly affected by the
a
aggregate type. For instance the concrete mixture L2 had actually
wnet = wadded + waggregate + wabsorption.
lower tensile strength ft and lower modulus Edyn. The values of eten
for earlier ages are much lower than the values of ectu from direct
tensile tests. This comes from the dynamic assessment of the elas-
but this effect was not clearly seen on the tensile strain capacity. tic modulus. The ultrasonic method does not involve the time
More data were needed to conrm these results, so the continuous dependent behaviour of concrete at early age, thus higher values
assessment of the elastic modulus through dynamic testing (Edyn) are given for the elastic modulus. The lower values of the tensile
was carried out (Fig. 13). strain capacity eten do not take into account creep, which cannot
The dynamic elastic modulus (Edyn) vs. age curve showed a be avoided during direct tensile testing.
rapid increase between 3 and 10 h. This corresponds to the hydra- The new tensile strain capacity vs. age curves plotted in Fig. 14
tion of cement and the setting period. Then the elastic modulus show a minimum between 3 and 4 h. So the behaviour observed in
increased more slowly. The elastic modulus Edyn of L2 had not the Fig. 12 was not due to an artefact generated by the assessment of
same time-evolution as L1 and S concretes. The superplasticizer the displacement of concrete. The signicant increase of the
 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161 159

250 2.0
Tensile strain capacity (ctu), m/m

L1 (1) 1.8 S

Tensile strength (ft), MPa

200 L1 (2) 1.6 L1
1.4 L2
150 1.2
1.0
100 0.8
0.6
50 0.4
0.2
0 0.0
0 4 8 12 16 20 24 28 0 4 8 12 16 20 24 28
Age, hours Age, hours

Fig. 8. Tensile strain capacity from two series of tests. Fig. 11. Peak stress of concrete.

4.4. Evolution of the tensile properties as a function of the degree of

2.0 hydration
1.8
Tensile strength (ft), MPa

L1 - 5 m/s
1.6 The evolution of the hydration degree was deduced from TGA
L1 - 0.5 m/s analyses on the cement paste of concrete L2 (Fig. 15). As the three
1.4
concrete mixtures had the same cement content and W/C ratio, the
1.2
evolution of the degree of hydration was assumed to be the same.
1.0
The variations of the Youngs modulus and the tensile strength
0.8
are plotted as a function of the degree of hydration on Fig. 16. The
0.6 modulus increased from degree of hydration of 0.2, whereas the
0.4 tensile strength signicantly increased from degree of hydration
0.2 of 0.3. The tensile strain capacity actually decreased during this
0.0 stage (Fig. 17). The minimum was reached at degrees of hydration
0 4 8 12 16 20 24 28 of 0.26, 0.26, and 0.24 for concrete mixtures S, L1, and L2
Age, hours

Fig. 9. Inuence of the loading rate on the tensile strength.

10000
Tensile strain capacity (ctu ), m/m

1000
Tensile strain capacity (ctu), m/m

500
L1 - 5 m/s
400 L1 - 0.5 m/s 100

300 S
10 L1
200 L2

1
100 0 4 8 12 16 20 24 28
Age, hours
0
Fig. 12. Evolution of the tensile strain capacity.
0 4 8 12 16 20 24 28
Age, hours
50
Fig. 10. Inuence of the loading rate on the tensile strain capacity.
45 S
40 L1
35
L2
Edyn, GPa

modulus (Fig. 13) appeared earlier than the signicant increase of 30

the tensile strength (Fig. 11). At early age the stiffness develop- 25
ment actually evolves faster than the strength development [40].
20
De Schutter and Taerwe [41] gave a degree of hydration-based
15
description for the strength, Youngs modulus, and the peak strain
10
of concrete in compression. For a given degree of hydration, the
5
relative strength was lower than the relative modulus, and the
0
peak strain vs. compressive strength curve showed a minimum. 0 4 8 12 16 20 24 28
This property of the behaviour of early age concrete was also Age, hours
shown by plotting the ultimate strain at compression vs. compres-
sive or tensile strength according to different sources [42]. Fig. 13. Evolution of the modulus of elasticity.
160 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161

100 1000

Tensile strain capacity (ctu), m/m

S
L1
100
ft / Edyn, m/m

L2

10 S
10
L1

L2
1
0.00 0.10 0.20 0.30 0.40 0.50
1
0 4 8 12 16 20 24 28 Degree of hydration
Age, hours
Fig. 17. Variation of tensile strain capacity with degree of hydration.
Fig. 14. Evolution of the tensile strain capacity assessed from the tensile strength/
elastic modulus ratio. modulus is due to the elastic properties of newly formed hydration
products. The degree of hydration of 0.2 corresponds to approxi-
mately 4 h (see Fig. 15). From the initial setting time, a percolation
0.60 threshold has been reached and continuous paths of solid products
can actually be found in the microstructure of cement paste [43].
L2 As the hydration of cement goes on, capillary voids are replaced
0.50
Degree of hydration

by more and more hydration products, especially portlandite and

0.40 y = 0.142Ln(x) + 0.067 CSH, which nally gives signicant strength to the cementitious
material. The degree of hydration of 0.3 approximately corre-
0.30 sponds to the end of setting (given by Vicat test). Moreover during
the rst hours and days of cement hydration, the proportions of
0.20
hydration products vary [44], and the morphology of the precipi-
tating CSH phases also change signicantly [45], thus the prop-
0.10
erties of hydrating cement paste are modied.
0.00
0 4 8 12 16 20 24 28 5. Conclusions
Age, hours
In this paper, a review of the available data and existing exper-
Fig. 15. Evolution of degree of hydration.
imental procedures dealing with the tensile behaviour of early age
concrete has been summarized. The need for a reliable method to
assess the tensile properties of concrete during the rst 24 h has
40 2.0 been highlighted. The main results of the study are given in the fol-
S - Edyn 1.8 lowing conclusions.
35
Tensile strength (ft), MPa

L1 - Edyn 1.6
30 L2 - Edyn  A new experimental procedure has been developed to measure
1.4
S - ft tensile strain capacity, tensile strength and tensile E-modulus of
Edyn, GPa

25 1.2
L1 - ft concrete during the rst 24 h of age. A special feature is the use
20 1.0 of digital image correlation (DIC) to measure deformation.
L2 - ft
15 0.8  The results showed a rapid increase in the tensile strength of
0.6 concrete between 7 h (end of setting period according to Vicat
10
0.4 test) and 24 h. The experiments were carried out in endogenous
5 0.2 conditions. The device would enable experiments in drying con-
ditions, which could be more representative of real exposure
0 0.0
0.00 0.10 0.20 0.30 0.40 0.50 conditions.
Degree of hydration  The evolution of the tensile strain capacity showed a minimum
corresponding to the period that includes the setting time and
Fig. 16. Variation of Youngs modulus and tensile strength with degree of early hardening. This evolution was conrmed by indirect
hydration. assessment of the strain capacity using continuous dynamic
monitoring of the elastic modulus. The signicant increase of
the modulus actually appeared earlier than the signicant
respectively. Then the tensile strain capacity increased and the increase of the tensile strength, thus the strain capacity rstly
aggregate type had no signicant inuence on the evolution. decreased. Then it increased with the tensile strength.
The strength of concrete evolved faster than the stiffness, thus  The aggregate type had a limited inuence on the evolution of
the results are in accordance with previous studies [40,41]. The the tensile strength, and the elastic modulus of early age con-
evolution of the microstructure of hydrating cement paste could crete increased with the modulus of aggregates. Finally the ten-
explain why the properties of concrete become measurable at dif- sile strain capacity was not signicantly affected by the
ferent degrees of hydration. Hydration results from the dissolution aggregate type.
of cement particles and precipitations of solid products, such as  The assessment of the tensile strength was more accurate than
ettringite, portlandite and CSH. The early increase of the elastic the strain capacity and it was less affected by the loading rate.
 E. Roziere et al. / Cement & Concrete Composites 55 (2015) 153161 161

The values of the strain capacity are closely linked to the exper- [21] Tasdemir MA, Tasdemir C, Akyz S, Jefferson AD, Lydon FD, Barr BIG.
Evaluation of strains at peak stresses in concrete: a three-phase composite
imental procedure and the parameters of testing, such as the
model approach. Cem Concr Compos 1998;20:30118.
loading rate. From these results, an approach based on strength [22] Alexander KM, Wardlaw JD, Gilbert DJ. Aggregate-cement bond, cement paste
and stresses could be more reliable than a strain based strength and the strength of concrete. In: Brooks AE, Newman K, editors.
approach, as far as early age cracking is concerned. International Conference on the Structure of Concrete. London: Cement and
Concrete Association; 1968. p. 5981.
[23] Tasong WA, Lynsdale CJ, Cripps JC. Aggregate-cement paste interface. Part I.
inuence of aggregate geochemistry. Cem Concr Res 1999;29:101925.
Acknowledgements [24] Scrivener KL, Crumbie AK, Laugesen P. The interfacial transition zone (ITZ)
between cement paste and aggregate in concrete. Interface Sci
2004;12:41121.
The Fdration Nationale des Travaux Publics (from the French [25] Zhang MH, Gjrv OE. Microstructure of the interfacial zone between
building industry) is gratefully acknowledged for their support to lightweight aggregate and cement paste. Cem Concr Res 1990;20:6108.
[26] Ping X, Beaudoin JJ, Brousseau R. Effect of aggregate size on transition zone
the research project. properties at the Portland cement paste interface. Cem Concr Res
1991;21(6):9991005.
References [27] Ollivier JP, Maso JC, Bourdette B. Interfacial transition zone in concrete. Adv
Cem Based Mater 1995;2(1):308.
[28] Davidson RJ. Report on an investigation into the variability of the elastic
[1] Cusson D, Hoogeveen T. An experimental approach for the analysis of early-age
modulus of concrete, United Kingdom Atomic Energy Authority, Contract No.
behaviour of high-performance concrete structures under restrained
E/5A/CON/565/1524, Taylor Woodrow Wind Energy Group, Report No 014H/
shrinkage. Cem Concr Res 2007;37:2009.
87/3241, 1987.
[2] Turcry Ph, Loukili A, Haidar K, Pijaudier-Cabot G, Belarbi A. Cracking tendency
[29] Aulia TB, Deutschmann K. Effect of mechanical properties of aggregate on the
of self-compacting concrete subjected to restrained shrinkage: experimental
ductility of high performance concrete. LACER 1999;4:13348.
study and modeling. J Mater Civil Eng, ASCE 2006:19.
[30] Caduff D, Van Mier JGM. Analysis of compressive fracture of three different
[3] Ravina D, Shalon R. Plastic shrinkage cracking. ACI J 1968;65(22):28294.
concretes by means of 3D-digital image correlation and vacuum impregnation.
[4] Hammer TA, Fossa KT, Bjntegaard . Cracking tendency of HSC: tensile
Cem Concr Compos 2010;32:28190.
strength and self generated stress in the period of setting and early hardening.
[31] Reinhardt HW, Grosse CU. Continuous monitoring of setting and hardening of
Mater Struct 2007;40:31924.
mortar and concrete. Constr Build Mater 2004;18:14554.
[5] Holt E, Leivo M. Cracking risks associated with early age shrinkage. Cement
[32] Robeyst N, Gruyaert E, De Belie N. Ultrasonic monitoring of setting and
Concr Compos 2004;26:52130.
hardening behaviour of concrete and mortar with blast furnace slag cement,
[6] Darquennes A, Khokhar MIA, Rozire E, Loukili A, Grondin F, Staquet S. Early
Montral. In: 12th International congress on the chemistry of cement, 2007.
age deformations of concrete with high content of mineral additions. Constr
[33] Robeyst N, Gruyaert E, Grosse CU, De Belie N. Monitoring the setting of
Build Mater 2011;25:183647.
concrete containing blast-furnace slag by measuring the ultrasonic p-wave
[7] Kasai Y, Yokoyama K, Matsui I. Tensile properties of early age concrete. In:
velocity. Cem Concr Res 2008;38:116976.
Mechanical behavior of materials, society of materials science, Japan, vol. 4;
[34] Krau M, Hariri K. Determination of initial degree of hydration for
1972. pp. 28899.
improvement of early-age properties of concrete using ultrasonic wave
[8] Abel J, Hover K. Effect of water/cement ratio on the early age tensile strength of
propagation. Cem Concr Compos 2006;28:299306.
concrete. Transp Res Rec 1998;98(1510):338.
[35] Loukili A. tude du retrait et du uage de bton ultra-hautes performances,
[9] Hannant DJ, Branch J, Mulheron M. Equipment for tensile testing of fresh
PhD thesis of cole Centrale de Nantes, 1996.
concrete. Mag Concr Res 1999;51(4):2637.
[36] Bogue RH. Calculation of the compounds in Portland cement. Ind Eng Chem
[10] Nahlawi H, Chakrabarti S, Jodikara J. A direct tensile strength testing method
Anal Ed 1929;1:1927.
for unsaturated geomaterials. Geotech Test J 2004;27(4):16.
[37] Turcry Ph, Loukili A. Evaluation of plastic shrinkage cracking of self-
[11] Dao VTN, Dux PF, Morris PH. Tensile properties of early-age concrete. ACI
consolidating concrete. ACI Mater J 2006;103(4):2729.
Mater J 2009;106(6):48392.
[38] Saliba J, Rozire E, Grondin F, Loukili A. Inuence of shrinkage-reducing
[12] Swaddiwudhipong S, Lu HR, Wee TH. Direct tension test and tensile strain
admixtures on plastic and long-term shrinkage. Cem Concr Compos
capacity of concrete at early age. Cem Concr Res 2003;33:207784.
2011;33(2):20917.
[13] Hamanaga Y, Sato Y, Otani T, Ueda K, Kiyohara C. Development of a new
[39] Kovler K, Igarashi S, Bentur A. Tensile creep behavior of high strength
apparatus for concrete shrinkage cracking test. Reports of the Faculty of
concretes at early ages. Mater Struct 1999;32:3837.
Engineering, Oita University. 2006;53:3337.
[40] Weigler H, Karl S, Beton Junger. Teil 1 + II. Betonwerk+ Fertigteil Technik, vol.
[14] Xie N, Liu W. Determining tensile properties of mass concrete by direct tensile
6/1974 and 7/1974, Junger Beton, Tell 1: Beanspruchung Festigkeit
test. ACI Mater J 1989;86:2149.
Verformung, 1974. pp. 392401.
[15] Smeplass S. High strength concrete: mechanical properties normal density
[41] De Schutter G, Taerwe L. Degree of hydration-based description of mechanical
concretes, SP4-Materials Design, Report 4.4, SINTEF, Trondheim, 1992.
properties of early age concrete. Mater Struct 1996;29:33544.
[16] Balendran RV. Properties of high strength lightweight concrete, Ph.D. Thesis,
[42] Bergstrm SG, Byfors J. Properties of concrete at early ages, RILEM reports of
University of Wales, Cardiff, 1980.
technical committees, 42-CEA committee, 1979. pp. 26574.
[17] Phillips DV, Zhang B. Direct tension tests on notched and unnotched plain
[43] Stefan L, Benboudjema F, Torrenti J-M, Bissonnette B. Prediction of elastic
concrete specimens. Mag Concr Res 1993;45:2535.
properties of cement pastes at early ages. Comput Mater Sci 2010;47:77584.
[18] Guo Z, Zhang X. Investigation of complete stress-deformation curves for
[44] Lothenbach B, Winnefeld F. Thermodynamic modelling of the hydration of
concrete in tension. ACI Mater J 1987;84:27885.
Portland cement. Cem Concr Res 2006;36:20926.
[19] Bamforth PB. Early-age thermal crack control in concrete, CIRIA C660, London,
[45] Holzer L, Winnefeld F, Lothenbach B, Zampini D. The early cement hydration: a
2007.
multi-method approach. In: 11th ICCC, 2003. pp. 23648.
[20] Tasdemir MA, Lydon FD, Barr BIG. The tensile strain capacity of concrete. Mag
Concr Res 1996;48(176):2118.