Construction and Building Materials 169 (2018) 403–413

Contents lists available at ScienceDirect

Construction and Building Materials

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

Effect of curing humidity on the fracture properties of concrete

Zhengxiang Mi, Yu Hu, Qingbin Li ⇑, Zaizhan An
State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, China

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Fracture properties of concrete were

investigated under different
conditions.

At early age, fracture parameters
remained almost constant during
drying.

At later age, fracture parameters
decreased markedly with decreasing
curing humidity.

Fracture mode changed from ductile
to brittle with reduction in curing
humidity.

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

Article history: This study investigates the influence of curing humidity on concrete fracture properties. Concrete fracture
Received 2 November 2017 energy, effective fracture toughness, and characteristic length were tested at five different ages using
Received in revised form 24 February 2018 wedge splitting specimens exposed to four different relative humidity values (30%, 50%, 70% and 98%).
Accepted 1 March 2018
Four environmental chambers capable of automatically controlling temperature and relative humidity
were constructed. A data averaging method for companion specimens was introduced. The test results
show that at the early age, fracture energy, effective fracture toughness, and characteristic length
Keywords:
remained almost constant during drying. As time elapsed, however, the influence of curing humidity
Fracture properties
Wedge splitting test
on the concrete fracture properties became increasingly pronounced; larger values of fracture energy,
Relative humidity effective fracture toughness, and characteristic length were measured by concrete specimens kept at high
Failure mode humidity. The concrete fracture mode also gradually changed from ductile to brittle with curing humidity
decreasing and hydration time elapsing.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction curing conditions owing to climatic variations. This results in the

Concrete is the most common structural material used in the
modern times. It is often poured and operated in different environ-
ments. Therefore, concrete structures face continuously changing
⇑ Corresponding author.
E-mail address: qingbinli@tsinghua.edu.cn (Q. Li).
formation of moisture gradients in concrete structures because of
water evaporation to the surroundings at lower relative humidity
levels. The extent and severity of moisture gradients depend on
many factors, such as ambient air relative humidity, temperature,
local rainfall, wind speed, and age at exposure. The properties of
concrete are also impacted by curing conditions [1,2].

https://doi.org/10.1016/j.conbuildmat.2018.03.025
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
404 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
Many studies investigated the effect of moisture gradient on the
behavior of concrete, and the results show that the reduced curing
humidity not only can negatively affect the hydration reaction and
microstructural behavior of concrete, but can also seriously deteri-
orate its mechanical and fracture properties [3–5]. Powers first rec-
ognized that cement hydration slows down at low internal relative
humidity and almost stops below approximately 80% RH [6]. In
other words, due to desaturation and the corresponding reduction
of the internal relative humidity in the pores of a cementitious
material, cement hydration may nearly come to a halt and the
development of concrete properties may be hampered, even
though the system still includes considerable amount of water.
This indicates that the development of various properties of con-
crete structure depend on the available water, and if the structure
is exposed to very low humidity, its surface hydration will cease
and strength and durability will decrease, which may cause surface
cracking and may not afford the desired protection. This will grad-
ually hamper the hydration of the internal parts of concrete struc-
tures, as the pore water is also lost through the drying surface.
Later, other researchers also observed a similar phenomenon. For
instance, Snyder and Bentz [7] found that the hydration of a
cement paste at 90% RH was lower than that under saturated con-
ditions. In particular, Jensen’s experimental results [8] revealed
that the hydration degree of C3S was about 67% after 365 days at
98% RH, but only about 4% when kept at 85% RH for 365 days.
The hydration degree was even smaller at a lower relative humid-
ity, and nearly no hydration took place below 43% RH. On the other
hand, Flatt et al. [9] theoretically proved why C3S cease to hydrate
below 80% RH, and showed that this phenomenon is caused mainly
by the reduced water activity, which changes the solubility of C3S
so that it is in equilibrium with the pore fluid. Du et al. [10] simu-
lated the reduction in the reaction rate caused by decreased
humidity, in which the hydration rate reduction factor indepen-
dent of temperature and hydration degree was 1.0 at 100% RH
and it decreased to smaller values at lower humidity levels.
Regarding the influence of reduced relative humidity on con-
crete microstructure, Cong and Kirkpatrick [11] observed that the
polymerization of C-S-H does not change when the relative humid-
ity was varied from 9% to 100%; however, its local structure
becomes more disordered with decreasing relative humidity. Using
nuclear magnetic resonance spectroscopy, Aparicio et al. [12]
observed that the samples cured at 70% RH presented more tetra-
hedral coordinated aluminum than those cured at 98% RH because
more aluminum was taken up by the C-S-H gel and the transforma-
tion of AFt to AFm was promoted. Furthermore, the mean chemical
reaction chain length of the C-S-H gel formed increased with
decrease in curing humidity, and longer chains correspond to
lower hydration degree. In addition, Sellevold and Bjøntegaard
[13] claimed that the thermal expansion coefficient of concrete
during maturation increase with decreasing humidity, where the
reduction in humidity is owing to self-desiccation and/or drying
of the environment.
In general, the mechanical properties of concrete, both in term
of strength and elastic modulus, decrease monotonically as a func-
tion of its curing humidity. Cebeci [14] showed that the compres-
sive strength of concrete reduced by 30% in 75% RH; and by 45% in
33% RH as compared to the specimens cured in water. Saengsoy
et al. [15] found that the compressive strength of concrete cured
in water and at 95% RH rapidly increased and became almost
steady after 28 days. However, at 80% RH, the concrete strength
develops slowly; especially at 60% RH, the strength hardly
increased after 7 days, leading to a loss of approximately 41% in
the ultimate strength compared to those cured in water. It must
be noted that Un and Baradan [16] concluded that the flexural
strength is more sensitive to curing humidity than compressive
strength, and the smaller values of flexural strength are measured
by the specimens cured at decreased relative humidity. With
regard to the elastic modulus, Beaudoin et al. [17] observed that
the static elastic modulus in compression decreases significantly
with reducing relative humidity, especially below 20% RH. Further-
more, they also reported that the elastic modulus shows a non-
linear response with respect to the relative humidity and that it
depends on the drying history. This is also confirmed by Alizadeh
et al. [18], who also reported that concrete would lose elastic mod-
ulus if exposed to reduced humidity and hypothesized that such
reduction could be attributed to micro-cracking and other micro-
scale phenomena such as high porosity. Moreover, Yalcinkaya
and Yazici [19] studied the effect of relative humidity on the drying
strain and showed that the shrinkage of concrete, compared to that
at 100% RH, was increased by 33%, 46%, and 52% under 70% RH, 60%
RH, and 50% RH, respectively.
However, very limited information is available on the fracture
properties of concrete exposed to low humidity. Bazant et al.
[20] and Lau et al. [21] investigated the effect of moisture on the
fracture properties of hardened concrete. They showed that both
the fracture energy and fracture toughness decrease with the
increasing moisture, and they supposed that the degradation of
fracture properties of concrete under moisture can be attributed
to internal pore water pressure, which would cause a very high dis-
joining pressure between the contacting cement pastes. In particu-
lar, data about the effect of curing humidity on the development of
concrete fracture parameters over time and long-term fracture
properties have been barely reported to date.
Therefore, the objective of this study is to investigate how cur-
ing humidity affects the development of fracture properties of con-
crete with age. Four different curing conditions were considered
with a constant temperature of 20 °C under different relative
humidity levels. The fracture parameters were measured after 3,
7, 14, 28, and 60 days.
2. Experimental program
2.1. Materials
The concrete mix proportions are listed in Table 1. The cement
used was type I ordinary Portland cement with a density of 1450
kg/m3. The fine aggregate utilized was natural river sand with a
fineness modulus of 2.52, conforming to ASTM C33/C33M No.2.
The crushed basalt passing the 20 mm sieve was applied as the
coarse aggregate. Physical properties and gradations of the coarse
aggregate were examined using the test standards of ASTM C33/
C33M No.67. The ratio of fine to total aggregate was 39% for all
concrete mixtures. The substitution rate of fly ash (ASTM Class A)
was 25% for the total binder content, which was in order to reduce
the heat of hydration for applications and save the cement. Potable
water was used for mixing concrete constituents and curing of the
specimens. No further additive was employed. In order to achieve
better uniformity, the concrete was mixed for 6 min. In addition,
an external vibrator was used for concrete compaction.
2.2. Specimen preparation and test set-up
Fracture tests were carried out on notched wedge splitting
specimens. These specimens are characterized by a negligible
influence of their self-weight on test results and by a considerably
large fracture area to the specimen volume ratio, which is approx-
imately 4.6 times greater than that of a commonly used three-
point bending beam of equal volume [22]. However, compared to
a conventional wedge splitting test with one line support, two
hinge roller supports located at the center of the half section of
the specimens were utilized to improve their stability and to avoid
 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
Table 1
Mixture proportions.
w/b Unit Weight (kg/m3)
Cement Fly ash Water Sand
/ / / 0–2.4 mm
0.4 351 117 187 701
unexpected failure of the specimens while preparing for the exper-
iment [23]. The principle of the test is schematically shown in
Fig. 1(a), where two steel loading devices with a roller bearing
on each side were placed on the top of the specimen. A steel profile
with two identical wedges was fixed at the upper plate of the test-
ing machine. The wedges are pressed between the bearings, which
leads to a horizontal splitting force. Neglecting the friction
between the wedge and the roller, the horizontal component act-
ing on the rollers can be evaluated using the following equation:
F
Fs ¼ ð1Þ
2 tan h
where F is the applied vertical load; F s is the horizontal splitting
force; h is wedge angle, which is equal to 15 in this test set up.
The shape and dimensions of wedge splitting specimens
employed in the experiments are shown in Fig. 1(b); the height
of the specimens were larger than that used in the standard test
in order to obtain the fracture parameters independent of the size.
In particular, a single specimen height was adopted here because
the main purpose of our study is to investigate the influence of rel-
ative humidity on the fracture property of concrete rather than the
size effect. The concrete specimens were cast in steel molds, which
were vibrated for at least 60 s in order to increase the density and
to eliminate air bubbles. The molds of the wedge splitting speci-
mens were designed and manufactured as per the test require-
ments. Each mold was thoroughly cleaned, dried, and then
coated with a film of lubricating oil before any concrete was cast.
The film of oil made it simpler to remove specimens from their
molds once the curing process was complete. For all wedge
splitting specimens, the initial notch (pre-crack) was created by
F 50mm
Wedge loading fixture
Roller bearing
Wedge
(a)
Fig. 1. Specimen for the wedge splitting test.
405
Gravel
2.4–5 mm 5–9.5 mm 9.5–16 mm 16–20 mm
11 316 612 155
inserting a steel plate inside the tensile face of a specimen with a
constant thickness of 2 mm during concrete casting. The length
of the initial notch was 120 mm for each specimen, producing a
notch to height ratio of 0.4. The steel plates were carefully removed
after approximately 4.5 h. A groove was made in the upper part of
the specimen to place two loading devices with a roller bearing and
to attach the clip-gauge, where the size of the groove was big
enough to place loading devices and clip-gauge. Three identical
specimens were prepared for each test age and relative humidity.
Although special care was taken while de-molding, moving, and
placing the specimens, some specimens were still severely dam-
aged in the notched section during handling. This is because the
notches are already present in low-strength concrete and also
because a splitting action may occur when the mold for the notch
is removed. For such damaged specimens, the reliable test results
may not be obtained; therefore, some spare specimens were also
prepared.
All wedge splitting tests were carried out on an Instron testing
machine at five different ages (3, 7, 14, 28 and 60 days). The actual
experimental setup used is presented in Fig. 2. During the test, the
primary deformation monitored was the crack mouth opening dis-
placement (CMOD) which was measured with a clip-gauge having
an accuracy and capacity of ±0.0002 mm and 5 mm, respectively.
The clip-gauge was mounted over the mouth of the initial notch.
All signals were continuously recorded by a commercial automatic
date-acquisition system at rate of two readings per second. The
loading actuator was controlled by the crack mouth opening dis-
placement with an initial rate of 0.04 mm/min. In the post-peak
range, the loading rate was gradually increased up to 0.08 mm/
min. This procedure was selected after several trials in order to
30mm
a0=120mm
300mm
tch
No
m
0m
12
300mm
(b)
406 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
Fig. 2. Actual wedge splitting specimen and test set-up.
minimize the time needed to conduct experiment, as a result, the
specimen completely failed in 30–40 min. Because of the use of a
very stiff computer-controlled servo-hydraulic closed-loop testing
machine, the propagation of the main crack was well controlled,
and the complete load–CMOD curves were recorded using the
date-acquisition system.
In addition to wedge splitting specimens, several companion
cubes were also poured with the same batch of concrete to mea-
sure the compressive strength and elastic modulus at 28 days for
each investigated humidity level. The dimension of each cube
was 100  100100 mm, which are in accordance with the Chi-
nese standards DL/T 5150-2001 [24]. The mechanical properties
were tested on an oil-pressure controlling testing machine, in
which displacement control with a constant loading rate of 0.2
mm/min of the jack displacement was adopted. The same speci-
mens were used to determine first the elastic modulus and then
the compressive strength. After the elastic modulus was deter-
mined, the load was continuously exerted until the specimen
failed, thus obtaining the compressive strength of each specimen.
The compressive strength and elastic modulus were determined
by averaging the test results of the four identical specimens, and
their values are presented in Table 2.
2.3. Curing conditions
The curing temperature and relative humidity of the specimens
were automatically controlled by using four specialized chambers
capable of producing temperature between 0 °C and 80 °C and rel-
ative humidity between 30% RH and 98% RH. The maximum
humidity deviation and temperature deviation of the chamber
are ±2% RH and ±2 °C, respectively. Four different relative humidity
levels of 30%, 50%, 70% and 98% were used in this experiment to
understand how curing humidity affects the concrete fracture
properties with age. The temperature was controlled at 20 °C for
Table 2
Mechanical properties of concrete cured at different relative humidity.
Relative humidity 30%
Compressive strength (MPa) 32.40
Elastic modulus (GPa) 24.77
each relative humidity level during the period of curing. Since four
variations of relative humidity were used with five scheduled test-
ing ages and three companion specimens as a test group, the total
number of concrete wedge splitting specimens utilized in the
investigation was 60. Note that, in the following, ‘‘T” label repre-
sents the curing temperature, ‘‘RH” denotes the curing relative
humidity, and ‘‘A” indicates the age of curing unless otherwise sta-
ted. For instance, the specimen is labeled as T20RH98A3, whose
curing temperature, curing humidity and age are 20 °C, 98% RH
and 3 days, respectively.
After finishing, the concrete specimens were kept at ambient
conditions. The free surface of the molds was covered with an
impermeable plastic membrane to prevent desiccation in order
to reduce the influence of moisture gradients. When the concrete
specimens were allowed to set for 4.5 h, they become strong
enough to be removed from the steel plates and moved into their
corresponding curing humidity environment. For this, four equal
groups of concrete specimens together with their molds were
stored in four chambers with relative humidity levels that were
previously set to 30% RH, 50% RH, 70% RH and 98% RH and main-
tained there for 24 h. Subsequently, the specimens were removed
from the molds and shifted into the same chambers again and kept
as such until the time of testing. For the low-humidity specimens,
special care was taken while removing them from their molds
because of their slower hydration rate. This curing procedure
was adopted to ensure that all specimens had sufficient strength
to avoid damaging while de-molding, moving and placing. Since
the setting time for the concrete was approximately 4.5 h, this pro-
cedure also guarantee that all specimens had the same properties
when the fracture parameters began to develop for a better com-
parison. In other words, if the concrete specimens were exposed
to the corresponding relative humidity chambers immediately
after mixing, they would achieve different degrees of hydration
and microstructure before the fracture properties started to
develop, which would make the comparison more difficult [25].
Furthermore, the employed curing method also became more con-
sistent with practice, in which the effort is often made to maintain
the curing condition for a certain period before exposure to the
working condition. It must be emphasized here that the curing
method of the cubes was the same as that of the wedge splitting
specimens.
After the desired curing period, the specimens were taken out
from the chamber 3 h before the test and exposed to natural envi-
ronment until testing in order to ensure the relative humidity sta-
bility during the experiment. At the ages of 3, 7, 14, 28, and 60
days, the key fracture parameters of concrete were measured on
three companion specimens according to ASTM E 1820 [26]. The
averaged value of the three identical fracture specimens was used
to analyze the test data. For a given variable, however, if the devi-
ation between one replicate and the average of three companions
was larger than approximately 10%, a spare specimen was tested
to increase the credibility of the test results.
2.4. Data processing method for companion specimens
As mentioned above, three companion wedge splitting speci-
mens were used for each test variable. However, because of the
diversity of individual specimens and the heterogeneity of con-
50% 70% 98%
34.64 37.28 42.69
26.72 29.01 34.36
 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
crete, the experimental results in the same group might be differ-
ent. Moreover, the averaged results are strongly dependent on the
method utilized to average the data. Therefore, in order to more
accurately evaluate the fracture characteristics for each test, a data
process method was shown below.
For each wedge splitting specimen, approximately 3200 original
data points were collected. At first, the points scattered far from
the load-CMOD curve were filtered, and then one point was taken
about every 16 points in the load-CMOD curve in order to facilitate
the optimal fitting curve and make the curve more beautiful, where
the number of data points could sufficiently represent the load-
CMOD curve. Afterwards, in order to eliminate the influence of
fluctuation in the measurement, an average of three adjacent
points (the given point and one point above and one below that
point) was adopted. Subsequently, the filtered and averaged data
were imported into the mathematical software MATLAB and opti-
mally fitted using rational function, with numerator and denomi-
nator degrees of 1 and 3, respectively. The correlation coefficients
of regression curve R2 were more than 0.99 for all cases. The max-
imum load points of each specimen were obtained by averaging
the peak points of the regression curve and measured curve whose
fluctuation has been eliminated, which can further reduce the ran-
domness and discreteness of the peak load. These average values
correspond to the peak load (Fmax) and critical crack mouth open-
ing displacement (CMODc) of each specimen, respectively, and they
are used to calculate the effective fracture toughness, as discussed
later.
When averaging the load-CMOD curve between the compan-
ions, each data point in the load-CMOD curve should denote the
identical mechanical status. According to the equivalent elastic
crack method, the fracture process zone is formed as the applied
load is increased, and the elastically equivalent crack extension
length reaches the maximum at the peak. After Fmax, the fracture
process zone begins to move forward along the crack path, keeping
the maximum equivalent extension length [27]. Therefore, for each
specimen, 100 equally spaced CMOD values were evaluated from
the zero point to CMODc and from the peak to the end point. It
should be noted that the fracture test will never reach the loading
level of zero since complete failure of specimen is approached
asymptotically and the experiment is often stopped before total
energy dissipation [28]. Thus, the end point of CMOD for each spec-
imen was taken such that the increment in the CMOD from the
peak to the end point was identical for each companion specimens
and the fracture energy was calculated at this point. The load val-
ues corresponding to 200 equally spaced CMOD values were calcu-
lated by interpolating the processed test data set using MATLAB.
For each specimen, the 100th data point represents the same status
as that represented by the peak load status, and from 100th to
200th all data points can be regarded as the same status because
each point (or fracture process zone) is equally far from the peak.
Meanwhile, we can assume that the each equally spaced data point
in the ascending region also represents the same fracture status.
Consequently, all of these equally spaced points denote the same
fracture status and the values of CMOD and the corresponding load
can be averaged. The averaged load-CMOD curves are discussed in
the following section.
3. Results
3.1. Load-CMOD curves
Fig. 3 shows the typical load-CMOD curves of concrete exposed
to four different relative humidity levels at the five investigation
ages. These figures illustrate that the concrete was nearly in linear
elastic deformation at the early loading stages. Nonlinear deforma-
407
tion was not observable or not obvious before the load reached the
cracking limit of concrete. The initial slope of the ascending
branches increased with increasing age, because the strength and
stiffness of concrete increases with age. Besides, the peak load also
improved with time for all types of curing methods. For example,
in test specimens cured at 98% RH, the peak load increased from
2.891 kN after 3-day curing to 5.014 kN after 60-day curing. This
increase in the peak load also occured when the specimens were
cured at low relative humidity, but the increased rate was smaller
than that of the concrete cured at high humidity levels. Typically,
for the specimens cured at 30% RH, the peak load increased from
2.714 kN after 3 days to 3.938 kN after 60 days.
As the curing humidity was decreased, the peak load at each age
reduced. Similarly, for a given curing humidity, the peak load gen-
erally increased significantly with the increasing age. For instance,
at 60 days of curing age, the average peak load decreased from
5.014 kN at 98% RH to 4.373 kN at 70% RH, 4.092 kN at 50% RH,
and finally to 3.938 kN at 30% RH with a final drop of 27.3%. Fur-
thermore, the curvature of the post-peak curve which reflected
the property of the cracked specimen until breaking also depended
on the curing humidity, and increased markedly with age. This was
related to the strengthening of the interfacial transition zone (ITZ),
where the stress concentration around the aggregate reduced and
new products in the ITZ pores showed increased crystallizations as
the concrete became older. As a result, more aggregates were
pulled out from the matrix and the fracture path also transformed
gradually from around the aggregate to directly through the
aggregate.
3.2. Fracture energy
Fracture energy is defined as the energy required to create a
crack with unit surface area. For the wedge splitting specimen,
fracture energy was calculated by the area under the measured
load-CMOD curve divided by the ligament area, as shown in Eq.
(2). Since the complete failure of the wedge splitting test was
approached asymptotically, the end point of the test was not easily
determined. Therefore, the test should be stopped at some point
prior to the complete fractures, indicating that the measured work
of fracture does not represent the true fracture energy of the inves-
tigated concrete. For this, Elices et al. [29] suggested that true frac-
ture energy should be estimated by using a far tail constant value.
In this study, the tests were stopped when the descending region of
the load-CMOD curve showed a full tail, and the unmeasured part
of the tail part was inversely extended using the optimally fitted
expression when the fracture energy was calculated. Moreover,
the self-weight of the specimen was neglected in the calculation
of fracture energy because it was compensated by two bottom sup-
ports, and the influence of the self-weight was insignificant for a
not very large wedge splitting specimen.
W
GF ¼ ð2Þ
ðh  a0 ÞB
where GF is fracture energy; W is the total work corresponding to
the area under the horizontal load-CMOD curve; h, a0 , and B are
the effective height, initial notch length, and thickness of the spec-
imen, respectively.
As mentioned above, for each series the mean curve was deter-
mined from the individual curves. The difference between the val-
ues of GF determined from the mean curve and the average value of
GF measured from the individual test results should be lower than
1%. For each curing humidity, the average values of GF measured
from the load-CMOD curve are plotted in Fig. 4 as a function of
the age. Although each point in Fig. 4 represents the average of
three companion specimens tested, the error bars are not pre-
408 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413

6 5
T20RH98A3 T20RH70A3
5 T20RH98A7 4 T20RH70A7
T20RH98A14 T20RH70A14

Load (kN)
Load (kN)
4
T20RH98A28 3 T20RH70A28

3 T20RH98A60 T20RH98A60
2
2

1
1

0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
CMOD (mm) CMOD (mm)

5 5
T20RH50A3 T20RH30A3

4 T20RH50A7 4 T20RH30A7
T20RH50A14 T20RH30A14
Load (kN)

Load (kN)
3 T20RH50A28 3 T20RH30A28
T20RH50A60 T20RH30A60
2 2

1 1

0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
CMOD (mm) CMOD (mm)

Fig. 3. Load-CMOD curves for wedge splitting specimens.

250 sufficient internal moisture. A similar phenomenon could be

observed for 30% RH, but GF tended to be a stable in a shorter time.
At a lower curing humidity, the shorter stable time can be attribu-
Fracture energy (N/m)

200
ted to the decrease in the internal humidity due to evaporation,
resulting in insufficient free water for the hydration reaction.
150 Fig. 4 also shows that at any age, a lower curing humidity always
led to a smaller value of GF , although the trend is similar. After 3
T20RH98
days of hydration, the GF of specimens kept in 30% RH was smaller
100
T20RH70 by 7.6%, 5.2% and 2.7% when compared to that of the concrete
T20RH50 cured at 98% RH, 70% RH and 50% RH, respectively. Furthermore,
50 the gap gradually increased as the concrete became older. Conse-
T20RH30
quently, at 60 days of curing age, the GF of specimens kept in
0 30% RH was 200.6 N/m, which was 17.4%, 12.3% and 7.1% smaller
0 10 20 30 40 50 60 than that of the concrete exposed to 98% RH, 70% RH and 50%
Age (day) RH, respectively. This indicates that the influence of curing humid-
ity on GF increased gradually with time, and the available free
Fig. 4. Variation of fracture energy with age for concrete cured at different relative
moisture became a critical factor for maintaining the later age
humidity.
hydration reaction.

sented for clarity. In any case, the deviation between one replicate 3.3. Effective fracture toughness
and the average of three companion was less than 10% with respect
to the average. For the nonstandard wedge splitting specimens, the following
As can be observed from Fig. 4, the fracture energy of concrete formula proposed by Xu et al. [30] can be used to calculate the
cured at low relative humidity was similar to that in standard cur- effective fracture toughness
ing condition. At the beginning, the value of GF tended to rapidly
increase in the first 7 days. On prolonging the curing ages, GF kept F smax 3:675½1  0:12ða  0:45Þ ac
K IC ¼ pffiffiffi  a¼ ð3Þ
increasing at a smaller rate, and eventually tended to reach a pla- B h ð1  aÞ3=2 h
teau value. Typically, for the 98% RH, GF was 150.4 N/m at 3 days,
181.3 N/m at 7 days, 199.8 N/m at 14 days, 220.1 N/m at 28 days, where K IC is the effective fracture toughness; F smax is the peak hor-
and 233.3 N/m at 60 days. The fracture energy continuously izontally load; ac is the critical effective crack length corresponding
increased with age and the slope was still relatively large, which to the peak load; B and h are the thickness and of effective height
means that a curing time of 60 days would not be enough to reach the specimen, respectively. Eq. (3) is valid within 1% accuracy for
a stable state for this humidity and the specimens sustained 0:2 6 a 6 0:8.
 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413 409

Once the initial crack propagation occurs under increased exter-

Effective fracture toughness (MPam1/2)

1.8
nal load, the nonlinear cohesive fracture process zone appears. In
this case, the practical physical crack is often simplified as an effec-
1.5
tive crack using the equivalent-elastic concept so that the linear
elastic fracture mechanics can be used to calculate K IC . In general,
1.2
the critical effective crack length can be evaluated in different
ways. For example, Karihaloo et al. [31] believed that ac can be
0.9
determined by measuring the peak load and the corresponding T20RH98
load point deflection. The two parameter fracture model proposed T20RH70
0.6
by Jenq and Shah [32] requires an unloading and reloading proce-
T20RH50
dure to estimate the value of critical effective crack length. Based
0.3 T20RH30
on the linear asymptotic superposition assumption, Xu et al. [30]
proposed another method to evaluate ac for nonstandard wedge
0
splitting specimen based on load-CMOD curves. This formula was
0 10 20 30 40 50 60
used in the present study and can be expressed as follows: Age (day)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
13:18 Fig. 5. Variation of effective fracture toughness with age for concrete cured at
ac ¼ ðh þ h0 Þ 1   h0 ð4Þ different relative humidity.
9:16 þ CMODc  E  B=F smax

where E is the elastic modulus obtained from the initial compliance

hydration became low and was negligible below 30% relative
of load-CMOD curve; CMODc is crack opening displacement corre-
humidity [2]. Hence, the hydration reaction was expected to stop
sponding to F smax ; h0 is thickness of clip gauge holder; h and B are
earlier under low humidity condition, and the period that K IC
the effective height and thickness of the specimen respectively.
achieved the steady became slightly shorter with decreasing curing
The empirical expression is valid for0:3 6 a 6 0:7 within 2.9 per-
humidity. Typically, for 98% RH, K IC was 1.061 MPam0.5 at 3 days,
cent accuracy. Thus, the critical effective crack length of each spec-
1.232 MPam0.5 at 7 days, 1.361 MPam0.5 at 14 days, 1.455 MPam0.5
imen can be easily determined by substituting the measured E,
at 28 days, and 1.555 MPam0.5 at 60 days. The value of K IC at 60
CMODc and F smax into Eq. (4). The values of ac at different curing
days was 6.9% larger than that at 28 days. Furthermore, for 30%
humidity levels and ages are listed in Table 3. It is noted that ac
RH, K IC was 0.931 MPam0.5 at 3 days, 1.081 MPam0.5 at 7 days,
was dependent on the curing humidity and age, and its value varied
1.175 MPam0.5 at 14 days, 1.295 MPam0.5 at 28 days, and 1.357
between 165.2 mm and 179.6 mm. Further, ac decreased with rela-
MPam0.5 at 60 days. The value of K IC at 60 days was only 4.8% lar-
tive humidity decreasing and hydration time elapsing. This trend is
ger than that at 28 days. This also indicated that the internal
consistent with previous observations made by Kim [33].
humidity of the specimen did not fall to the critical value before
Fig. 5 illustrates the effective fracture toughness against age for
60 days but was sustained a condition that there was sufficient
different curing humidity, in which each point represents the aver-
internal moisture available for the hydration reaction. It can also
age values of K IC determined by the three replicate specimens. In
be observed from Fig. 5 that K IC of the concrete cured at low rela-
order to avoid presenting a messy plot, the error bars were omitted
tive humidity was smaller than that kept under standard curing
again. Compared with the fracture energy, a smaller scattering of
conditions at the same age; this was especially true at the later
data was observed in K IC , which may be explained by the difference
ages. For instance, at 7 days, for concrete cured at 30% RH, the
in the crack path. Owing to unstable propagation of the crack, its
value of K IC was reduced by about 13.9%, 11.1%, and 4.5% when
path became more arbitrary and redundant cracks also emerged,
compared to the corresponding value of the concrete kept in 98%
but the energy consumed by these random micro-cracks could be
RH, 70% RH and 50% RH, respectively. With the passage of the time,
not easily determined and were therefore neglected when calculat-
the difference in K IC slightly increased. Typically, after 14 days of
ing the fracture energy, resulting in a greater randomness of the
hydration, the K IC of concrete exposed to 30% RH was 1.175
fracture energy. However, K IC was determined before the forma-
MPam0.5, which was approximately 15.9%, 12.7%, and 4.7% smaller
tion of redundant and random micro-cracks.
than that of concrete kept in 98% RH, 70%RH and 50% RH, respec-
As shown in Fig. 5, K IC was a monotonically increasing function
tively. These results demonstrated that the curing humidity exert
of the age, irrespective of the curing humidity investigated, which
a growing effect on K IC with time. Also, we can again confidently
was similar to that noted in the development of GF with age. Dur-
conclude that the available free moisture was a key factor in deter-
ing the first 7 days, K IC increased rapidly with age because the
mining the hydration of concrete, especially at later ages.
cement hydration reaction was very rapidly in this stage. The
growth rate first became gradually larger, and then decreased after
reaching the maximum value, and the corresponding curves 3.4. Characteristic length
became steep. Afterwards, K IC slightly increased with age at a
slower rate and tended to be stable if the curing time was long Fracture energy or effective fracture toughness alone cannot be
enough. Once the internal relative humidity dropped below 80%, used to represent the brittleness of materials, which is commonly

Table 3
Critical effective crack length at different relative humidity levels and ages.

Age (days) Critical effective crack length ac (mm)

98% RH 70% RH 50% RH 30% RH
3 176.9 173.7 173.2 172.3
7 175.9 172.5 172.3 171.1
14 173.8 179.6 171.2 170.5
28 172.2 171.0 170.2 168.7
60 169.8 168.2 167.9 165.2
410
understood as the tendency of a material to fracture abruptly
before any significant irreversible deformation occurs. Therefore,
in order to describe the brittleness of concrete, Hillerborg et al.
[34] introduced characteristic length (Lch ), defined as
EGF
Lch ¼ 2
ft
where GF is the fracture energy; E is the elastic modulus; f t is the
splitting tensile strength obtained using inverse analysis by match-
ing the experimental specimen response data to the theoretical
data. More specifically, the load-CMOD response was modeled
and fit over the experimental data source. A detailed derivation of
this method can be found in the literature [35]. If the elastic mod-
ulus was not measured from the cubes, its value can also be calcu-
lated from the initial compliance of the load-CMOD curve for a
wedge splitting specimen [32]. Kumar et al. [36] concluded that
almost the same value of can be obtained from the load-CMOD
curve and compressive cylinder test. Therefore, in this study, the
value of elastic modulus determined by the initial compliance
was used to calculate the characteristic length. Because the charac-
teristic length includes a combination of energy, stiffness, and
strength parameters, it is regarded as a synthetic brittleness
parameter.
According to the fictitious crack model [34], it is clear that Lch
was proportional to the fracture process zone. The longer fracture
process zone required higher energy, and this energy in micro-
crack propagation was visible in the longer tail stress-strain curve.
Therefore, the smaller value of Lch corresponded with higher brit-
tleness and lower strength against the cracking of concrete and
vice versa.
Variation of Lch with age for concrete cured at different relative
humidity levels is depicted in Fig. 6. It is apparent that Lch was
strongly dependent on the age and curing humidity, and its value
varied between 513.1 mm and 328.5 mm. For the same curing con-
ditions, Lch decreased rapidly with the age until 28 days, and then
kept decreasing at a reduced rate. For instance, in the case of 98%
relative humidity, Lch continuously decreased from 513.1 mm at
3 days to 494.3 mm at 7 days, 462.8 mm at 14 days, 409.4 mm at
28 days, and finally 363.8 mm at 60 days, with a significant loss
of 149.2 mm or 29.1%. The decrease of Lch with age shows that con-
crete is more ductile at early ages. This is because, at early ages, a
large amount of cement and water have not undergone the hydra-
tion reaction yet, and are gathered in the interfacial transition
zone. As a result, the interfacial transition zone was porous and
the cracking path became longer and more tortuous. Fig. 6 also
shows that for the same age, Lch reduced dramatically owing to
550
Characteristic length (mm)
500
450
400
350
300
250
0 10 20 30 40
Age (day)
Fig. 6. Variation of characteristic length with age for concrete cured at different
relative humidity.
Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
decreasing curing humidity. At 3 days, the Lch of concrete cured
at 98% RH was higher by 1.9%, 3.6%, and 5.7% relative to those
cured at 70%RH, 50% RH, and 30% RH, respectively. After 28 days
of hydration, the corresponding gap increased to 5.2% for 70% RH,
8.9% for 50% RH and 12.7% for 30% RH. However, at 60 days of
the curing age, the Lch of the concrete cured at 98% RH was
ð5Þ
363.8 mm, which was only larger by 3.8%, 5.9%, and 9.7% than that
of the concrete cured at 70%RH, 50% RH, and 30% RH, respectively.
This indicates that the decreasing rate of the characteristic length
due to the decreasing curing humidity increased with age until
28 days, and then decreased with age.
4. Discussion
As illustrated in Figs. 4–6, for different curing humidity,
although the overall trend of the development of concrete fracture
properties with age was similar, some deviations were observed.
Typically, at early stages, the cement hydration reaction was very
rapid. Consequently, GF and K IC increased rapidly with age, regard-
less of the curing humidity. The growth rate first became gradually
increased, and then almost decreased after reaching the maximum
value, and the corresponding curves became steep. However, the
period for this stage became slightly longer with the decreasing
relative humidity. In the next stage, the hydration rate became
smaller; therefore, GF and K IC slightly increased with age at a
decreased rate. Normally, there should be a third stage, where GF
and K IC would gradually reached their respective ultimate values
with time. At high humidity, however, the third stage was not
observed or was not very obvious because the age of only up to
60 days was investigated in this study, which also implied that
the internal humidity of the specimen did not reduce to the critical
value. The increase in GF and K IC with age was attributed to the fact
that ITZ between the aggregate and the paste, in which the greatest
number of micro-cracks formed, and it was the most sensitive zone
within the concrete. Therefore, the cement paste became stronger
because of the accumulation of new hydration products in ITZ
pores as the concrete became older. In fact, Elsharief et al. [37]
has demonstrated that the presence of a relatively large amount
of un-hydrated cement grains in the ITZ at early ages made the
ITZ porous. With time elapsing, the pores in the ITZ were filled
with hydration products, and the number and size of these pores
and that of cement paste reduced. The strength of the cement paste
and ITZ also improved. Thus, the bridging effect and pulling of the
aggregate out of the matrix decreased, and the fracture path under
load also transformed from bond zones or around aggregates to
directly through the aggregates, as shown in Fig. 7. Furthermore,
the strength of the aggregate was much larger than that of the
cement paste and ITZ. Consequently, higher energy was required
T20RH98 to overcome the aggregates and the strengthened ITZ and cement
T20RH70 paste. Thus the larger values of GF and K IC were measured. With
further strengthening of cement paste and ITZ, however, the frac-
T20RH50
ture mode did not change significantly, where only more aggre-
T20RH30 gates were pulled out from the matrix. Thus, the values of GF and
K IC kept increasing at a decreasing rate at later ages, and eventually
tended reach to a stable value.
More importantly, as mentioned above, the fracture properties
of the concrete were less dependent on the curing humidity during
early stages, where GF , K IC , and Lch for the concrete did not reduce
significantly as the relative humidity changed after 3 days of
hydration. However, at later ages, when part of the water was used
in the hydration reaction or was lost due to evaporation, the avail-
50 60
able water content became a critical factor for the continued
hydration. Starting at 7 days and continuing till 60 days, the values
of GF , K IC , and Lch became smaller with decreasing curing humidity,
in which GF , K IC , and Lch reduced approximately by 17.4%, 15.9%
 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
(a) At 3 days of curing age
Fig. 7. Typical fracture surfaces of concrete at different ages.
and 14.6% respectively from 98% RH to 30% RH. Such trend agreed
with other limited number of independent investigation reported
by other researchers in a somewhat different way. For instance,
Hordijk et al. [38] observed that the concrete specimens became
more brittle with lower humidity and the characteristic length
decreased with decreasing humidity. Bella et al. [39] reported that
after 7 days of hydration, the elastic modulus steadily decreased
with relative humidity until 33% RH and then kept constant. Inves-
tigating three strength levels under humidity of 100% RH to 40%
RH, Kwon et al. [40] demonstrated that the compressive strength
for low, medium, and high strength levels decreased from 40.2 to
25.2 MPa, 44.9 to 29.8 MPa, and 54.0 to 35.8 MPa respectively.
Patel et al. [2] found that curing humidity below 80% RH produced
a coarsened pore structure with a large-diameter porosity three
times greater than that obtained with saturated curing, which
could adversely affect the fracture property and durability of
in situ concrete.
The degradation of the concrete fracture properties at low
humidity raises a question: What phenomena occurred at the
microscopic level under these curing conditions? We believed that
several different mechanisms were responsible for such reduction
in fracture properties at low humidity. Firstly, this reduction was
partly caused by the physical properties of the concrete. With
decreasing environmental humidity, the evaporation of the mois-
ture inside the concrete accelerated, resulting in reduced water
available for the hydration reaction and a partial desaturation of
finer pores. As a result, hydration products could not continue in
the absence of water. Secondly, in the case of low humidity, the
desiccation of concrete caused the shrinkage of the cement paste,
whereas the effect of this desiccation on the aggregates was less
pronounced than on the cement paste [41]. The stress concentra-
tions due to restraint deformation led to the formation of micro-
cracks in the ITZ, originally at larger aggregates even before load-
ing. Furthermore, at a given temperature, the thermal expansion
(a) Cured at 30% RH
Fig. 8. Typical fracture surfaces of concrete specimens cured at different relative humidity.
411
(b) At 60 days of curing age
coefficient of the concrete increased with decreasing relative
humidity [42], which further enlarged the mismatched deforma-
tion between the cement paste and aggregates and increased the
likelihood of micro-cracking. Consequently, the failure mode has
also changed at low humidity levels owing to the weakening of
the bond zone and formation of micro-cracking, where the crack
propagated around the aggregates and lesser aggregates were frac-
tured, as shown in Fig. 8. As it was mentioned before, the strength
of the aggregate was much larger than that of the cement paste and
ITZ, resulting in smaller values of GF , K IC and Lch at low humidity
levels. Further mechanism for the reduction in fracture parameters
was the reduced water activity [43]. Namely, in a concrete hydrat-
ing system, un-hydrated phases continuously dissolved and dif-
fused out into the pore solution forming new solid hydrated
phases. In the case of low humidity, however, the effective water
was lower, and less water in the pores of material was available
for dissolving un-hydrated phases and for reaction with the
cement. What’s worse, evaporation of water from an ionic solution
would be accompanied by the increase in the ionic concentrations,
and if the saturation was reached, ionic precipitation would take
place, which further decreased the water activity and the amount
of the hydrating cement paste. It should be noted that higher con-
centrations would probably give rise to higher reaction rates before
a steady concentration levels were reached again, but in a smaller
volume of water and with a lower reaction rate. Meanwhile, in the
case of insufficient water, the solubility of C3S would shift owing to
the negative capillary pressure in the pore liquid, which was a
main factor in determining the equilibrium between C3S, C-S-H,
and CH, and it can inhibit the hydration reaction of C3S [9]. Finally,
with decreasing curing humidity, the interlayer water reduced and
thus the basal-spacing of C-S-H decreased, yielding a more disor-
dered local structures [11]. As a result, the concrete became a more
porous structure, and the smallest values of GF and K IC were mea-
sured for the concrete kept at 30% RH. In brief, all of these factors
(b) Cured at 98% RH
412 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
were related to phenomenon that fracture properties deteriorated
with decreasing relative humidity.
A question related to our results is still unanswered: Why the
fracture parameters did not change significantly with relative
humidity at the early ages? According to the physical model pro-
posed by Wittman [44], the capillary pressure inside the pores
increased with decreasing moisture content, imposing a volumet-
ric pressure that compressed the material and resulted in the
strengthening of fracture properties. Moreover, this compaction
effect of increasing capillary forces was more pronounced at the
early ages, when the solid skeleton of the hydration products
filled up the overall volume available less efficiently. Hence, the
deterioration of the fracture properties owing to micro-cracking,
reduced water activity, and other such factors was partly com-
pensated by the compaction effect of capillary forces, eventually
resulting in quasi-constant GF , K IC , and Lch . With time, however,
a more developed solid skeleton could better withstand the over-
all compaction. As a result, the compaction effect gradually weak-
ened, resulting in an overall macroscopic reduction in values of
GF , K IC , and Lch . Thus, the variation in the concrete fracture prop-
erties with relative humidity was the result of competition
between the compaction effect and the behaviors of micro-
cracking, lower hydration degree and so on, and the latter could
take over the former at later ages.
5. Conclusions
In this study, the effect of curing humidity on the fracture prop-
erties of concrete was investigated at five different ages (3, 7, 14,
28, and 60 days) using the wedge splitting specimens. The relative
humidity range from 30% RH to 98% RH was considered, which rep-
resented a wide range of dry and wet climatic conditions of under
which concrete structures are usually constructed. The complete
load-CMOD curves of the specimens at each test age and relative
humidity were measured. A data processing method was intro-
duced for averaging the test results of the companion specimens.
The key fracture parameters, such as GF , K IC and Lch , were also eval-
uated from the experimental data of each specimen. The following
conclusions can be drawn:
1) The effect of curing humidity on GF was insignificant at the
early age, but increased dramatically with increasing age.
At 3 days of the curing age, the values of GF was almost con-
stant regardless of the curing humidity. Beginning at 7 days,
a lower curing humidity always resulted in a smaller value
of GF and the gap gradually increased with time elapsing,
where GF reduced approximately by 17.4% due to the reduc-
tion in relative humidity from 98% to 30% at 60 days.
2) For a given age, K IC of concrete monotonically increased as a
function of curing humidity. The value of K IC became smaller
with decreasing relative humidity, and this reduction was
more marked at later ages. Typically, at 14 days, the K IC of
concrete cured at 30% RH was 4.7%, 12.7%, and 15.9% smaller
than that of concrete cured at 50% RH, 70% RH and 98% RH,
respectively.
3) At the early age, the Lch of concrete was less dependent on
the curing humidity, whereas at later ages a steady reduc-
tion was observed owing to decreasing relative humidity.
The reduction rate gradually increased with age until 28
days, and then decreased with age. After 28 days of hydra-
tion, the Lch of concrete cured at 30% RH was 14.6% smaller
than that of the concrete exposed to 98% RH.
4) With decreasing curing humidity, the fracture mode of the
concrete also transformed from ductile behavior to
brittleness.
From the perspective of engineering, the reduction of GF , K IC ,
and Lch due to drying not only increased the likelihood of cracking,
but also exacerbated the durability of concrete structures. Thus,
the influence of ambient humidity on the fracture parameters of
concrete should not be ignored while designing concrete structures
to avoid undesirable results. We believe that our work will provide
a better understanding of the fracture properties of concrete sub-
jected to practical curing regimes and will also help designers in
checking the safety factor and the adequacy of the structure to
safely withstand all applied loads.
Acknowledgements
This research work was supported by the National Natural Science
Foundation of China (51339003, 51579134), and the Research
Fund of Tsinghua University (Grant No. 20161080079).
References
[1] P. Altcin, The durability characteristics of high performance concrete: a view,
Cem. Concr. Compos. 25 (2003) 409–420.
[2] R.G. Patel, D.C. Killoh, L.J. Parrott, W.A. Gutteridge, Influence of curing at
different relative humidities upon compound reactions and porosity in
Portland cement paste, Mater. Struct. 21 (1988) 192–197.
[3] G. Michael, W. Lars, A method based on isothermal calorimetry to quantify the
influence of moisture on the hydration rate of young cement pastes, Cem.
Concr. Res. 40 (2010) 867–874.
[4] S.N. Shoukry, G.W. William, B. Downie, M. Riad, Effect of moisture and
temperature on the mechanical properties of concrete, Constr. Build. Mater. 25
(2011) 688–696.
[5] W. Zhou, H. Li, H. Nasser, Study on variability of model parameters of concrete
structure: humidity and moisture effect, Proceedings of the SPIE 6934, San
Diego: California, USA, 2008.
[6] D.C. Powers, A discussion of cement hydration in relation to the curing of
concrete, Highway Res. Board Proc. 27 (1947) 178–188.
[7] K.A. Snyder, D.P. Bentz, Suspended hydration and loss of freezable water in
cement pastes exposed to 90% relative humidity, Cem. Concr. Res. 34 (11)
(2004) 2045–2056.
[8] O.M. Jensen, P.F. Hansen, E.E. Lachowski, F.P. Glasser, Clinker mineral hydration
at reduced relative humidities, Cem. Concr. Res. 29 (1999) 1505–1512.
[9] R.J. Flatt, G.W. Scherer, J.W. Bullard, Why alite stops hydrating below 80%
relative humidity, Cem. Concr. Res. (2011) 987–992.
[10] M.Y. Du, X.Y. Jin, H.L. Ye, N.G. Jin, Y. Tian, A coupled hygro-thermal model of
early-age concrete based on micro-pore structure evolution, Constr. Build.
Mater. 111 (2016) 689–698.
[11] X.D. Cong, R.J. Kirkpatrick, Effects of the temperature and relative humidity on
the structure of C-S-H gel, Cem. Concr. Res. 25 (6) (1995) 1237–1245.
[12] S. Aparicio, M.R. Sagrario, M.A. Miguel, V.F. Jose, The effect of curing relative
humidity on the microstructure of self-compacting concrete, Constr. Build.
Mater. 104 (2016) 154–159.
[13] E.J. Sellevold, Ø. Bjøntegaard, Coefficient of thermal expansion of cement paste
and concrete: mechanisms of moisture interaction, Mater. Struct. 39 (2006)
809–815.
[14] O.Z. Cebeci, Strength of concrete in warm and dry environment, Mater. Struct.
20 (1987) 270–272.
[15] W. Saengsoy, T. Nawa, P. Termkhajornkit, Influence of relative humidity on
compressive strength of fly ash cement paste, J. Struct. Constr. Eng. 73 (631)
(2008) 1433–1441.
[16] H. Un, B. Baradan, The effect of curing temperature and relative humidity on
the strength development on the Portland cement mortar, Sci. Res. Essays 6
(12) (2011) 2504–2511.
[17] J.J. Beaudoin, L. Raki, R. Alizadeh, L. Mitchell, Dimensional change and elastic
behavior of layered silicates and Portland cement paste, Cem. Concr. Compos.
32 (2010) 25–33.
[18] R. Alizadeh, J.J. Beaudoin, L. Raki, Mechanical properties of calcium silicate
hydrates, Mater. Struct. 44 (2011) 13–28.
[19] C. Yalcinkaya, H. Yazici, Effect of ambient temperature and relative humidity
on early-age shrinkage of UHPC with high-volume mineral admixtures, Constr.
Build. Mater. 144 (2017) 252–259.
[20] Z.P. Bazant, P.C. Prat, Effect of temperature and humidity on the fracture
energy of concrete, ACI Mater. J. 85 (4) (1988) 262–271.
[21] D. Lau, O. Buyukozturk, Fracture characterization of concrete/epoxy interface
affected by moisture, Mech. Mater. 42 (2010) 1031–1042.
[22] E. Bruhwiler, F.H. Wittman, The wedge splitting test, a new method of
performing stable fracture mechanics tests, Eng. Fract. Mech. 35 (1990) 117–
125.
[23] N. Bretschneider, V. Slowik, B. Villmann, V. Mechtcherine, Boundary effect on
the softening curve of concrete, Eng. Fract. Mech. 78 (2011) 2896–2906.
[24] DL/T 5150-2001, Test Code for Hydraulic Concrete, China Electric Power Press,
Beijing, 2006.
 Z. Mi et al. / Construction and Building Materials 169 (2018) 403–413
[25] J.Y. Zhang, D. Cusson, P. Monterio, J. Harvey, New perspectives on maturity
method and approach for high performace concrete applications, Cem. Concr.
Res. 38 (2008) 1438–1446.
[26] ASTM E 1820, Standard Test Method for Measurement of Fracture Toughness,
Annual Book of ASTM Standards, 2003.
[27] Z.F. Zhao, S.H. Kwon, S.P. Shah, Effect of specimen size on fracture energy and
softening curve of concrete: Part I. Experiments and fracture energy, Cem.
Concr. Res. 38 (2008) 1049–1060.
[28] J. Lee, M.M. Lopez, An experimental study on fracture energy of plain concrete,
Int. J. Concr. Struct. Mater. 8 (2) (2014) 129–139.
[29] M. Elices, G.V. Guinea, J. Planas, Measurement of the fracture energy using
three-point bend tests: Part 3 – Influence of cutting the P-d tail, Mater. Struct.
25 (1992) 327–334.
[30] S.L. Xu, H.W. Reinhardt, Determination of Double-K criterion for crack
propagation in quasi-brittle fracture, Part III: compact tension
specimens and wedge splitting specimens, Int. J. Fract. 98 (2) (1999)
179–193.
[31] B.L. Karihaloo, P. Nallathambi, Effective crack model for the determination of
fracture toughness (K eIC ) of concrete, Eng. Fract. Mech. 35 (4/5) (1990) 637–
645.
[32] Y.S. Jenq, S.P. Shah, Two parameters fracture model for concrete, J. Eng. Mech.
111 (10) (1985) 1227–1241.
[33] J.K. Kim, Y. Lee, S.T. Yi, Fracture characteristics of concrete at early ages, Cem.
Concr. Res. 34 (2004) 507–519.
[34] A. Hillerborg, M. Modeer, P.E. Peterson, Analysis of crack formation and crack
growth in concrete by means of fracture mechanics and finite elements, Cem.
Concr. Res. 6 (6) (1976) 773–782.
413
[35] J. Zhang, X.C. Ju, Investigation on stress-crack opening relationship of
engineered cementitious composites using inverse approach, Cem. Concr.
Res. 41 (2011) 903–912.
[36] S. Kumar, S.R. Pandey, A.K.L. Srivastava, Determination of double-K fracture
parameters of concrete using peak load method, Eng. Fract. Mech. 131 (2014)
471–484.
[37] A. Elsharief, M.D. Cohen, J. Olek, Influence of aggregate size, water cement ratio
and age on the microstructure of the interfacial transition zone, Cem. Concr.
Res. 33 (2003) 1837–1849.
[38] D.A. Hordijk, H.W. Reinhardt, Fracture of concrete in uniaxial tensile
experiments as influenced by curing conditions, Eng. Fract. Mech. 35 (4)
(1990) 819–826.
[39] C.D. Bella, M. Griffa, T.J. Ulrich, P. Lura, Early-age elastic properties of cement
based materials as a function of decreasing moisture content, Cem. Concr. Res.
89 (2016) 87–96.
[40] S.H. Kwon, K.P. Jang, J.W. Bang, J.H. Lee, Y.Y. Kim, Prediction of concrete
compressive strength considering humidity and temperature in the
construction of nuclear power plants, Nucl. Eng. Des. 275 (2014) 23–29.
[41] V. Mechtcherine, Fracture mechanical behavior of concrete and the condition
of its fracture surface, Cem. Concr. Res. 39 (2009) 620–628.
[42] W. Mateusz, L. Pietro, Moisture dependence of thermal expansion in cement-
based materials at early ages, Cem. Concr. Res. 53 (2013) 25–35.
[43] M. Wyrzykowski, P. Lura, Effect of relative humidity decrease due to self-
desiccation on the hydration kinetics of cement, Cem. Concr. Res. 85 (2016)
75–81.
[44] F. Wittmann, Interaction of hardened cement paste and water, J. Am. Ceram.
Soc. 56 (8) (1973) 409–415.