Materials and Structures (2008) 41:937–949

DOI 10.1617/s11527-007-9296-z

ORIGINAL ARTICLE

Improving strength, drying shrinkage, and pore structure

of concrete using metakaolin
Erhan Güneyisi Æ Mehmet Gesoğlu Æ
Kasım Mermerdaş

Received: 23 March 2007 / Accepted: 1 August 2007 / Published online: 12 September 2007

RILEM 2007

Abstract This paper presents the results of an Keywords Compressive strength

investigation on the use of metakaolin (MK) as a Concrete
Metakaolin
Pore structure

supplementary cementing material to improve the Portland cement
Shrinkage
Tensile strength
performance of concrete. Two MK replacement
levels were employed in the study: 10% and 20%
by weight of the Portland cement used. Plain and PC- 1 Introduction
MK concretes were designed at two water–cementi-
tious materials (w/cm) ratios of 0.35 and 0.55. The Performance of concrete is determined by its
performance characteristics of the concretes were mechanical and durability properties. There are so
evaluated by measuring compressive and splitting many studies in the literature focusing on the
tensile strengths, water absorption, drying shrinkage, improvement of concrete performance by replace-
and weight loss due to the corresponding drying. The ment of Portland cement to some extents of various
porosity and pore size distribution of the concretes mineral admixtures; such as, fly ash, silica fume,
were also examined by using mercury intrusion blast-furnace slag, etc. Due to pozzolanic and filling
porosimetry (MIP). Tests were conducted at different effects of these certain mineral admixtures, they are
ages up to 120 days. The results revealed that the capable of enhancing the durability through the pore
inclusion of MK remarkably reduced the drying refinement and the reduction in the calcium hydrox-
shrinkage strain, but increased the strengths of the ide of the cement paste matrix [1]. Generally, the
concretes in varying magnitudes, depending mainly effects of mineral admixtures may be assessed as
on the replacement level of MK, w/cm ratio, and age improvement in workability, durability to thermal
of testing. It was also found that the ultrafine MK cracking, durability to chemical attacks, and produc-
enhanced substantially the pore structure of the tion of high performance concrete [2].
concretes and reduced the content of the harmful Recently, there has been a growing interest in the
large pores, hence made concrete more impervious, utilization of high-reactivity metakaolin (MK) as a
especially at a replacement level of 20%. supplementary cementitious material in concrete
industry. MK is an ultrafine pozzolana, produced by
calcining purified kaolinite clay at a temperature
ranging from 700 to 900C to drive off the chemi-
E. Güneyisi
M. Gesoğlu (&)
K. Mermerdaş
cally bound water and destroy the crystalline
Department of Civil Engineering, Gaziantep University,
Gaziantep 27310, Turkey structure [3–5]. Unlike industrial by-products such
e-mail: mgesoglu@gantep.edu.tr as fly ash, silica fume, and blast-furnace slag, MK is
938 Materials and Structures (2008) 41:937–949

refined carefully to lighten its color, remove inert obtained from local sources. Properties of the fine
impurity, and control particle size. The particle size and coarse aggregates are given in Table 2. A
of MK is generally less than 2 lm, which is sulphonated naphthalene formaldehyde-based high
significantly smaller than that of cement particles, range water-reducing admixture (HRWA) was used
though not as fine as silica fume [4, 6]. Moreover, the to give a consistent workability. The properties of the
use of MK in concrete in its present form is relatively HRWA are shown in Table 3.
a new concept [7]. Recent works have shown that the
inclusion of MK greatly influenced the mechanical
and durability properties of concrete [4, 6–14]. It has
also been demonstrated that concrete mixture incor- Table 1 Properties of Portland cement and metakaolin
porating high-reactivity MK gave comparable
Item Portland cement Metakaolin
performance to silica fume mixtures in terms of
strength, permeability, and chemical resistance SiO2 (%) 19.73 51.8
[15–19]. The utilization of this material is also Al2O3 (%) 5.09 45.8
environmentally friendly since it helps in reducing Fe2O3 (%) 3.99 0.35
the CO2 emission to the atmosphere by the minimi- CaO (%) 62.86 0.01
zation of the Portland cement (PC) consumption. MgO (%) 1.61 0.03
The primary objective of this study was to SO3 (%) 2.62 –
investigate systematically the effectiveness of MK Na2O (%) 0.18 0.13
used at various replacement levels on the perfor- K2O (%) 0.80 0.06
mance of the concretes in terms of compressive and Cl (%) 0.01 –
splitting tensile strengths, water absorption, drying Insoluble residue (%) 0.24 –
shrinkage, and pore structures. The normal and high Loss on ignition (%) 1.90 0.91
strength concretes with and without MK were tested Free lime (%) 0.57 –
at different ages up to 120 days for aforementioned Specific gravity (g/cm )3
3.14 2.60
characteristics. Based on the test results, the influence Setting time, Vicat needle 2–46/3–44 –
of MK content, w/cm ratio, and age upon concrete Initial/Final (h-min)
properties were discussed. Expansion, Le Chatelier 1 –
apparatus (mm)
Specific surface area (m2/kg) 327 8600
2 Experimental procedures Color Gray White

2.1 Materials
Table 2 Properties of aggregates
Two different binding materials were used in the
study, namely Portland cement and metakaolin (MK). Sieve size (mm) Fine aggregate Coarse aggregate
Their chemical compositions and physical properties River Crushed No I No II
are presented in Table 1. Portland cement (CEM I sand sand
42.5R) was conforming to the Turkish standard TS 16.0 100 100 100 100
EN 197-1 (which is mainly based on the European 8.0 100 100 31.5 1.9
EN 197-1). The MK used in this study is a white 4.0 86.6 95.4 1.0 1.1
powder with a Hunter L whiteness value greater than 2.0 56.7 63.3 0.5 1.0
90 (on a scale from ‘‘0-black’’ to ‘‘100-maximum
1.0 37.7 39.1 0.5 0.9
whiteness’’). It has a specific gravity of about 2.60, a
0.50 25.7 28.4 0.5 0.9
specific surface area of 8600 m2/kg, and an average
0.25 6.7 16.4 0.4 0.8
particle size less than 2 lm. The fine aggregate was a
Fineness modulus 2.87 2.57 5.66 5.93
mix of river sand and crushed sand whereas the
Specific gravity 2.66 2.45 2.72 2.73
coarse aggregate was a river gravel with a maximum
Absorption, % 0.55 0.92 0.45 0.42
particle size of 16 mm. Both aggregates were
Materials and Structures (2008) 41:937–949 939

Table 3 Properties of the high-range water-reducing admixture (HRWA)

SGa State Freezing point Color Chloride content Nitrate content Main component

1.22 Liquid 4C Dark brown None None Sulphonated naphthalene

a
SG: Specific gravity

2.2 Details of mixture proportions and casting the drying shrinkage and weight loss. All specimens
of specimens were poured into the steel moulds in two layers, each
of which being vibrated for a couple of seconds.
Two series of concrete mixtures were designed at low
(0.35) and high (0.55) water–cementitious materials
(w/cm) ratios. The control mixtures included only 2.3 Curing of the specimens
Portland cement with a content of 450 and 350 kg/m3
for the low and high w/cm ratios, respectively. To Drying shrinkage specimens were cured at 20C and
develop the MK-modified concrete mixtures, how- 100% relative humidity and demoulded after 24 h.
ever, the cement was partially replaced with 10% and After that, the specimens were exposed to drying in a
20% MK (by weight) for both series. Thus, totally six humidity cabinet at 23 ± 2C and 50 ± 5% relative
different mixtures were designed in this study. humidity, as per ASTM C157 [20] for about 60 days.
Details of the mixtures are presented in Table 4. All the other specimens were maintained under a
Grading of the aggregate mixture was kept constant plastic sheet for 24 h, then demoulded and water
for all concrete mixtures. The mixtures given in cured until required for testing.
Table 4 were designed to have slump values of
140 ± 20 mm and 180 ± 20 mm for the low and high
w/cm ratios, respectively for the ease of handling, 2.4 Test methods
placing, and consolidation. The high range water-
reducing admixture was added at the time of mixing To evaluate the strength characteristics of the plain
to attain the specified slump at each w/cm ratio. All and MK concretes, the compression test was carried
concretes were mixed in accordance with ASTM out on the cube specimens by means of a 3000 kN
C192 in a power-driven revolving pan mixer. Spec- capacity testing machine according to ASTM C39
imens cast from each mixture consisted of eighteen [21] while the splitting tensile strength was con-
100 · 100 · 100 mm cubes for the compressive ducted on the cylinder specimens according to
strength testing, eighteen 100 · 200 mm cylinders ASTM C496 [22]. The strength measurements of
for the determination of splitting tensile strength, nine concrete were performed at the ages of 1, 3, 7, 28,
100 · 100 · 100 mm cubes for the water absorption 90, and 120 days. Three specimens were used for
test, and three 70 · 70 · 280 mm prisms to monitor each testing age.

Table 4 Mixture proportioning of the concrete

Concrete w/cm MK Cement MK Water Fine aggregate Coarse aggregate HRWAa
series ratio (%) (kg/m3) (kg/m3) (kg/m3) (kg/m3)
River sand Crushed No I No II
(kg/m3) sand (kg/m3) (kg/m3) (kg/m3)

1 0.35 0 450 0 158 724 233 611 241 7.9

0.35 10 405 45 158 720 232 608 240 10.1
0.35 20 360 90 158 717 231 605 239 12.4
2 0.55 0 350 0 193 726 234 612 242 3.5
0.55 10 315 35 193 723 233 610 241 4.4
0.55 20 280 70 193 721 232 608 240 6.1
a
HRWA: High-range water-reducing admixture
940 Materials and Structures (2008) 41:937–949

The water absorption test was conducted on 3 Results and discussions

concrete cubes (100 · 100 · 100 mm). The absorp-
tion test was carried out according to ASTM C642 3.1 Compressive and splitting tensile strengths
[23]. Three specimens from each mixture were tested
at the ages of 28, 90, and 120 days and the average Figure 1 shows the effect of MK on the 1, 3, 7, 28,
values were reported. For the determination of water 90, and 120-day compressive strengths of the
absorption by total immersion, the dry mass (Md) for concretes having different replacement levels of
each sample was recorded and then totally immersed MK and w/cm ratios. As seen from the figure, there
in water at 20C until they achieved a constant mass is an evident increase in the compressive strength
(Ms). Ms was taken as the saturated mass. This took owing to the increase in the MK content. The
up to 48 h. The absorption percentage was then compressive strength values varied between
calculated by the following equation: 12.9–76.8 MPa and 13.8–98.5 MPa for plain and
MK modified concretes, respectively. It was observed
ðM s  M d Þ
WA ð% Þ ¼ 100  ð1Þ that the early (1–7 days) and long (up to 120 days)
Md
term compressive strength of the concretes incorpo-
Drying shrinkage measurements were conducted in rated with MK was about 5–23% and 10–30% greater
accordance with ASTM C157 [20]. The length than that of the plain concretes, respectively, depend-
change was measured by means of a dial gage ing mainly on replacement level of MK and w/cm
extensometer with a 200 mm gage length. Measure- ratio. As it is seen, the incorporation of MK enhanced
ments were carried out every 24 h for the first not only the early-age strength but also provided
3 weeks and then 3 times a week. At the same time, higher strength at later ages. These results are in
weight loss measurements were also performed on aggrement with results reported by Khatib and
the same specimens. Variations in the drying shrink- Hibbert [26] and Bai et al. [27]. In the work of
age strain and the weight loss were monitored during Khatib and Hibbert [26], MK appeared to be very
the 60-day drying period (at 23 ± 2C and 50 ± 5% useful on the compressive strength development and
relative humidity) and the average of three prism the use of MK at 20% replacement level increased the
specimens were used for each property. compressive strength of the concretes by as high as
The tests of pore structure of the concrete samples 38, 54, and 43% at 7, 28, and 90 days, respectively.
were conducted at the age of 120 days using a Furthermore, Bai et al. [27] found that MK contrib-
mercury intrusion porosimetry (MIP). More specifi- uted significantly to early strength development of
cally, the porosity of the specimen as well as its pore concretes. They observed that in the early stages of
size distribution was measured. For this, the samples curing, mixtures with 5% of MK exceeded the control
without coarse aggregate for microstructure testing mixture strength by up to 92%.
were separated from the concrete cube specimens When the level of the compressive strength is
after the compressive strength testing due to the compared at 28 days onwards, the highest strengths
limitation of sample size for the porosimeter. The of 50 and 75 MPa measured at 28 days were
samples were dried at about 105C before mercury achieved at 10% replacement of cement by MK for
intrusion porosimeter test. The samples were sub- the high and low w/cm ratios, respectively. When the
jected to a maximum pressure of up to 227 MPa. replacement level further increased to 20%, the
A cylindrical pore geometry, a constant contact angle increase in the strength values lessened gradually.
of 140, and a constant surface tension of 483 dynes/ However, a marked difference in strength value
cm were assumed [24, 25]. The mercury intruded between plain and MK-modified concretes was
pore diameter d at a pressure of P was calculated by observed at later ages, particularly at 120 days. It
using the following Eq. [25]: was also evident from Fig. 1 that MK had a
remarkable effect on strength development charac-
d ¼ ð1=PÞ4c cos h ð2Þ
teristic of concretes for both w/cm ratios. With
where d is the pore diameter (lm), P is the applied respect to the 28-day compressive strength, the
pressure (MPa), c is the surface tension of mercury strength gain of plain concretes at 120 days was in
(dynes/cm), and h is the contact angle (degree). the range of 12–20% while that of MK modified
Materials and Structures (2008) 41:937–949 941

Fig. 1 Effect of metakaolin 110

(MK) on the compressive w/cm=0.35 w/cm=0.55 1 days
100 3 days
strength development of
7 days
concretes having different 90
28 days
w/cm ratios 90 days

Compressive strength (MPa)

80
120 days
70

60

50

40

30

20

10

0
MK0 MK10 MK20 MK0 MK10 MK20
Replacement level of metakaolin

concretes ranged from 23 to 32%. In MK concretes, of concrete. In other words, as the compressive
metakaolin contributes to the strength of concrete at strength increases, the tensile strength also increases
later ages mainly by the fast pozzolanic reaction [12]. but at a decreasing rate. There are several empirical
The addition of MK into the matrix improves the formulations for evaluating splitting tensile strength
bond between the cement paste and the aggregate fsp and compressive strength fc0 ; and most researchers
particles as well as increasing the density of the achieved the expression of the type:
cement paste, which in turn significantly improves
the compressive strength of the concretes. According fsp ¼ kðfc0 Þn ð3Þ
to the literature, the main factors that affect the
contribution of MK in the strength are (a) the filling where fsp and fc0 are splitting tensile and compressive
effect, (b) the dilution effect, and (c) the pozzolanic strengths measured on 150 · 300-mm cylinders at
reaction of MK with CH [28]. 28 days (MPa), respectively, k and n are coefficient
The strength development pattern for splitting that can be obtained from the regression analysis. The
tensile strength is similar to that of compressive n value is generally within the range of 0.50–0.75.
strength, as can be seen in Fig. 2. The highest tensile The existing expressions for estimating splitting
strength value was obtained for the concrete with 20% tensile strength, as suggested by ACI [30], CEB-
MK for both w/cm ratios. In general, the splitting FIB [31], and TS 500 [32], are given below,
tensile strength increased with the increase in MK respectively.
content at all ages. However, the increase in the
splitting tensile strength was smaller compared to that fsp ¼ 0:59ðfc0 Þ0:5 ð4Þ
obtained in the compressive strength. For example, at
the high w/cm ratio, the compressive strengths of
concretes containing 20% MK were approximately 14, fsp ¼ 0:301ðfc0 Þ0:67 ð5Þ
22, 17, 17, 21, and 30% higher than that of the plain
concretes at 1, 3, 7, 28, 90, and 120 days, respectively. fsp ¼ 0:35ðfc0 Þ0:5 ð6Þ
However, the companion strength increments in the
splitting tensile test were about 6, 28, 7, 8, 18, and 22% Figure 3 shows the observed relationship between
at 1, 3, 7, 28, 90, and 120 days, respectively. the splitting tensile and compressive strengths of the
Neville [29] reported that the splitting tensile concretes tested at 28 days. For comparison, the Eqs.
strength of concrete has a close relationship with 4 through 6 are also included in the diagram. To
compressive strength. However the ratio of the two make use of them, 100-mm cube compressive
strengths depends on the general level of the strength strengths (used in the current study) were converted
942 Materials and Structures (2008) 41:937–949

Fig. 2 Effect of metakaolin 5

(MK) on the splitting tensile w/cm=0.35 1 days w/cm=0.55
strength development of 3 days
7 days
concretes having different 4 28 days
w/cm ratios

Splitting tensile strength (MPa)

90 days
120 days
3

0
MK0 MK10 MK20 MK0 MK10 MK20
Replacement level of metakaolin

to 150 · 300-mm cylinder strengths by multiplying the range of TS 500 and ACI and/or CEM-FIB
by a factor of 0.83, assuming that the conversion models.
factors from a 100-mm cube to 100 · 200-mm
cylinder and from a 100 · 200-mm cylinder to
150 · 300-mm cylinder are 0.91, respectively [33]. 3.2 Water absorption
From Fig. 3, it was observed that there was a
considerably high relationship between the splitting Figure 4 shows the variation in water absorption of
tensile and compressive strengths of the concrete so the concrete with different replacement levels of MK,
that a regression analysis provided correlation coef- w/cm ratio, and testing age. It was observed that the
ficient (R2) of 0.98. Within the strength range of this water absorption characteristics of the concrete
study, both ACI and CEB-FIB models appeared to be specimens decreased with increasing MK content,
well close to each other but provided relatively higher irrespective of w/cm ratio and testing age. As can be
predictions. However, TS 500 model provided lower
expected, the water absorption of both the plain and
values compared to the experimental data. Further-
especially MK concretes have a decreasing tendency
more, the proposed equation of this study fell within
with the increase in the curing period. At 28 days of
curing, the beneficial effect of MK in reducing the
water absorption was noticeable due to the filling
7
Theoretical model: effect of ultrafine MK as well as its pozzolanic
y = 0.387x0.5
6 2
reaction. Wild et al. [16] reported that in the presence
Splitting tensile strength (MPa)

R = 0.984
of MK, the filling effect is immediate, the acceler-
5
ation of OPC hydration has its major within the first
4 24 h, and the level of pozzolanic reaction is consid-
erably high within first 7–14 days for all MK levels
3
between 5 and 30%. However, the differences in the
2 Sample with MK
water absorption characteristics of the plain and MK
Sample without MK concretes became more significant at later ages (90
ACI code
1 and 120 days) and were remarkably lower for MK
CEB/FIP code
TS500 code
0
concrete compared to the plain concrete, especially at
20 30 40 50 60 70 80 high w/cm ratio. Furthermore, the reduction in the
Compressive strength (MPa)
water absorption with increasing the test age was
Fig. 3 Relationship between compressive and splitting tensile about 12 and 21% for the plain concretes of the low
strengths and high w/cm ratios, respectively, while it was about
Materials and Structures (2008) 41:937–949 943

Fig. 4 Effect of metakaolin 7

(MK) on the 28, 90, and w/cm=0.35 w/cm=0.55
180 day water absorptions 28 days
6
of concretes having 90 days
120 days
different w/cm ratios
5

Water absorption (%)

4

0
MK0 MK10 MK20 MK0 MK10 MK20
Replacement level of metakaolin

20–29% and 38–45% for the MK concretes, respec- 800

tively. This reduction in the water absorption with 700

age indicates superior performance of the MK
600
Shrinkage (microstrain)

blended cement concretes over the conventional

concrete. This may be explained by the pozzolanic 500

activity of MK during the prolong curing. It is well 400

known that the pozzolanic reactions contribute to the
300
refinement of the binder capillary porosity, with its MK0-w/cm=0.35
direct consequences on the improvement of the 200 MK10-w/cm=0.35
MK20-w/cm=0.35
durability characteristics of the concrete [13]. 100
MK0-w/cm=0.55
MK10-w/cm=0.55
MK20-w/cm=0.55
0
0 10 20 30 40 50 60 70
3.3 Drying shrinkage and weight loss Drying time (days)

Fig. 5 Drying shrinkage vs. drying time graphs of plain and

Drying shrinkage tests alone cannot offer sufficient MK concretes having w/cm ratios of 0.35 and 0.55
information on the behavior of concrete structures
since virtually all concretes are restrained in some
way, either by reinforcement or by the structure. clear and the shrinkage of concrete with MK
However, drying shrinkage tests can provide neces- decreased with increasing replacement levels of
sary information on how the drying shrinkage stresses MK. When the shrinkage strain at 60 days of drying
develop [34]. was considered, the MK concrete had remarkably
The strain developments versus time of the drying lower shrinkage strain in comparison to the plain
shrinkage specimens at two w/cm ratios are presented concrete. It was pointed out that the higher the
in Fig. 5. It was observed from the figure that the replacement of MK, the higher the reduction in the
drying shrinkage of the concretes having 0.35 and shrinkage, irrespective of w/cm ratio. For example, at
0.55 w/cm had decreasing tendency with little fluc- low w/cm ratio, the drying shrinkage of the concrete
tuations with the drying period. At early test ages, the mixtures incorporated with 10 and 20% MK were
shrinkage of all the specimens exhibited a very steep approximately 5 and 15% less than that of the plain
development. The difference between amount of the concrete, respectively. However, at high w/cm ratio,
drying shrinkage of plain and MK modified concretes the effect of MK is more pronounced such that the
was small, especially for the low w/cm ratio. But concretes with 10 and 20% MK had about 13 and
after about 2 weeks the difference became much 18% lower shrinkage than the plain concrete. Similar
944 Materials and Structures (2008) 41:937–949

observation was found by Brooks and Megat-Johari 800

Shrinkage after 60 days (microstrain)

y = -5.2x + 703
[6] in that the total shrinkage measured from 24 h 700 R2 = 0.953

was reduced by the use of MK while drying shrinkage 600

was significantly less for the MK concretes than for 500
the control concrete. They reported that the influence
400
of MK was seen to reduce the drying shrinkage of
y = -4.8x + 536
ordinary portland cement concrete by about 50%. 300
R2 = 0.923

Furthermore, the test results presented herein agree 200

with the findings of the study conducted by Al-Khaja 100 w/cm=0.35

w/cm=0.55
[35] and Jainyong and Yan [36] for the effect of the 0
ultrafine mineral admixtures on the shrinkage and 0 5 10 15 20 25

creep of concretes. In the study of Al-Khaja [35], it Replacement level of MK (%)

was concluded that the shrinkage and creep of plain Fig. 6 Relationship between shrinkage after 60 days and
concrete were considerably or moderately reduced replacement level of MK
with the incorporation of silica fume, showing a
1-month reduction in strain of 34.9 and 18.5% for inverse type of relationship, indicating that the drying
shrinkage and creep, respectively, which led to a shrinkage of the concrete for both w/cm ratios
reduction in the total deformation of 20.8%. Jainyong decreased with increasing MK content with fairly
and Yan [36] have also showed that ultrafine ground good correlation coefficient. It was also noticed that
granulated blast-furnace slag and silica fume can MK had a shrinkage-reducing or compensating
substantially promote hydration of cement and properties. It might be used as an additive in concrete
increase in the amount of AFt crystal hydrates and application where high shrinkage should be avoided
C–S–H gel hydrates in cement paste, which offers a or undesirable.
hardened concrete with a stronger structure and The reason in the reduction of shrinkage owing to
higher resistance to deformation caused by applied the use of MK is surely the decrease in shrinkage rate.
forced. Moreover, these two binders may fill small The shrinkage rates in Fig. 5 have been calculated
pores and voids harmful to the structure of concrete. and presented in Fig. 7. The critical comparison of
That might be the mechanism of reducing effect of the curves given in Fig. 7 for the plain and MK
ultrafine mineral admixtures (i.e. GGBS, SF) on modified concretes indicated that the average shrink-
creep and drying shrinkage of concrete. age rates of MK concretes were lower than the plain
Test results on drying shrinkage in Fig. 5 also concretes during the 60-day drying period, irrespec-
indicated that the concretes with low w/cm ratio had a tive of replacement level and w/cm ratio. It is evident
tendency of higher shrinkage than those with high from the figure that the higher MK content, the
w/cm ratio. At the end of the drying period, indeed,
160
the concrete mixture containing 20% MK had a
MK0-w/cm=0.35
shrinkage strain of 592 and 448 microstrain at low 140 MK10-w/cm=0.35
Average shrinkage rate (10 /day)

MK20-w/cm=0.35
and high w/cm ratios, respectively. However, the 120 MK0-w/cm=0.55
-6

shrinkage data from the earlier investigations indi- MK10-w/cm=0.55

MK20-w/cm=0.55
100
cated that the effect of lower w/cm ratio was to
reduce drying shrinkage of concrete as a result of 80
enhancement in strength and refinement in micro- 60
structure [37]. One possible explanation of such
40
behavior observed in this study may be that self-
desiccation and autogenous shrinkage could have 20

contributed to this higher drying shrinkage of 0

concretes with lower w/cm ratio. 0 10 20 30 40 50 60 70
Drying time (days)
The relationship between the shrinkage after
60 days and the replacement levels of MK can be Fig. 7 Average shrinkage rate vs. drying time graphs of plain
seen in Fig. 6. It was observed that there was an and MK concretes having w/cm ratios of 0.35 and 0.55
Materials and Structures (2008) 41:937–949 945

slower the shrinkage rate of concretes, especially for 800

MK0-w/cm=0.35
the high w/cm ratio. It was also noted that both the 700 MK10-w/cm=0.35
MK20-w/cm=0.35
plain and the MK concretes with low w/cm ratio MK0-w/cm=0.55

Shrinkage (microstrain)
600
showed a somewhat faster development of shrinkage MK10-w/cm=0.55
MK20-w/cm=0.55
500
than those with high w/cm ratio.
The results of weight loss due to drying for 400

different concrete mixtures are also shown in Fig. 8. 300

It was clear from the figure that the concrete mixes
200
with higher unit water contents showed higher weight
loss. Similar to the drying shrinkage test results, 100

inclusion of MK to the concrete mixes decreased 0

0 20 40 60 80 100 120 140
considerably the weight loss for both w/cm ratios.
Weight loss (gr)
During 60 day drying period, the difference of weight
loss between plain and MK blended cement concretes Fig. 9 Relationship between shrinkage and weight loss
were more distinguishable. This is true for all drying
time and the difference increases with increasing the investigated in order to obtain the relationship
drying time. It was observed after 60 day drying that between them. For all mixtures, the shrinkage was
the MK concretes exhibited a weight loss up to 48% approximately proportional to the loss of water.
less compared to plain concretes, depending mainly However, in the case of the mixtures with a w/cm
on replacement level of MK and w/cm ratio. It seems ratio of 0.55, it was observed that the water loss
that the rate of weight loss for MK concretes slows during 60 day drying stage was somewhat high.
down earlier than that of the plain concrete. This may Therefore, the slope of the curves for the high w/cm
be because there is less evaporable water available in mixtures deviated considerably as compared to those
the mixes as hydration and pozzolanic reactions used for the low w/cm mixtures. This may be explained
up significant amount of free water. Overall, the that the water coming from the large capillary pores
foregoing trends lead to confirm that the MK within the high w/cm mixtures results in greater
concretes have a lower porosity and finer pore weight loss [29]. Bissonnette et al. [38] also studied
structure which encourages loss of water by self- the relationship between the shrinkage and weight
desiccation rather than by diffusion to the outside loss for the paste, mortar, and concrete mixtures.
environment [6, 34]. They observed that irrespective of the mixture type,
In Fig. 9, the drying shrinkage is plotted against the shrinkage is almost proportional to the loss of
the corresponding weight loss for all mixtures water. Moreover, it was found that the water–cement
ratio was very effective on this correlation.
140

120 3.4 Effect of MK on the pore structure

of the concrete
100
Weight loss (gr)

80 The results of MIP are given in Table 5 and Fig. 10.

It can be seen from Fig. 10 that, for similar pore
60
sizes, the pore volume of the samples without MK is
MK0-w/cm=0.35
40 MK10-w/cm=0.35 higher than those with MK for both w/cm ratios.
MK20-w/cm=0.35
MK0-w/cm=0.55
The pore size distribution curves indicated the
20
MK10-w/cm=0.55 marked influence that MK additions had in skewing
MK20-w/cm=0.55
0 pore size distribution to the finer sizes compared
0 10 20 30 40 50 60 70
with the samples without MK. It was also observed
Drying time (days)
from Table 5 that the total porosity decreased sub-
Fig. 8 Weight loss vs. drying time graphs of plain and MK stantially with increasing replacement level of MK.
concretes having w/cm ratios of 0.35 and 0.55 The magnitude of this reduction varies between 22
946 Materials and Structures (2008) 41:937–949

Table 5 Results of mercury intrusion porosimetry test for the MK. The effect was particularly beneficial at 20%
low and high w/cm ratio samples tested at 120 days MK content, where the lowest porosity and pore
w/cm MK (%) Porosity Mean pore Median pore diameter were achieved.
ratio (%) diameter (lm) diameter (lm) It is essential to note that MK is a relatively new
mineral admixture used in the cement concrete
0.35 0 10.7 0.286 0.098
products. MK is highly reactive pozzolana and its
0.35 10 7.3 0.139 0.058
reaction mechanism can be divided into physical
0.35 20 5.5 0.088 0.037
and chemical aspects. The physical effect is that
0.55 0 14.6 0.614 0.202
the ultra-fine particles fill the voids in cement,
0.55 10 11.4 0.320 0.129
which makes the microstructure of matrix denser.
0.55 20 10.2 0.283 0.093
The chemical effect is the reaction of MK with the
cement hydrates. The reaction of MK with cement
and 49%, depending mainly on w/cm ratio and hydrates is faster since MK has a loose micro-
replacement level of MK. It was also noted that there structure after heat activated at a temperature of
was a considerable reduction in the mean (or median) 800 C (the chemically bonded water in kaolin is
pore diameter of the samples due to the inclusion of driven out at a higher temperature, and water

Fig. 10 Effect of MK (a) 100

addition on the pore-size
90 MK0
distribution of the samples
MK10
tested at the age of 120 days 80 MK20
(a) w/cm ratio: 0.35; and
Cummulative pore volume (%)

(b) w/cm ratio: 0.55 70

60

50

40

30

20

10

0
100 10 1 0,1 0,01 0,001
Pore diameter (Microns)

(b) 100
90
MK0
80 MK10
MK20
Cummulative pore volume (%)

70

60

50

40

30

20

10

0
100 10 1 0,1 0,01 0,001
Pore diameter (Microns)
Materials and Structures (2008) 41:937–949 947

molecules enter MK more easily when met with porosity from this experimental study and several
water again). During the chemical reaction, the other relations for predicting the compressive
precipitated calcium hydroxide (Ca(OH)2) is trans- strength from the porosity forwarded by the other
formed into secondary C–S–H (C = CaO; S = SiO2; researchers is also given in Fig. 11 for the
and H = H2O) gel, resulting in refinement of the comparison purposes. It was evident from the
pore structure (transformation of coarser pores into figure that the model proposed by Poon et al.
finer pores). Owing to the net reduction in Ca(OH)2 [42] remarkably overpredicted the compressive
content in the hydrated matrix due to pozzolanic strength of the mortar containing MK while the
reaction, the volume of continuous capillary pores Balshin model proposed by Bouguerra et al. [44]
is proportionally decreased; the higher the Ca(OH)2 greatly underestimated the compressive strength.
content in the hydrated matrix, the higher the However, the Ryshkewitch model [44] and the
volume of continuous pores [11, 39]. model proposed by Chan et al. [45] provided
Based on the experimental results, the relation- relatively more reasonable prediction of the exper-
ship between the porosity measured at 120 days imental data.
and the corresponding 120 day compressive In order to analyze the interdependence between
strength of the concrete is also examined. In the the porosity and water absorption measured in the
literatures, it has been shown that the porosity is present investigation, the correlation between the
not only the parameter influencing the strength of porosity and the corresponding 120 day water
concrete, but many other factors such as pore size absorption of the concrete specimens with and
distribution, microcracks, interface, and so on are without MK was also studied. The plot is illustrated
also important factors that determine mechanical in Fig. 12. It was observed that the relationship
properties of cementitious materials [25, 40, 41]. between the two variables was almost exponential
However, porosity, which can be semiemprically with a relatively high correlation coefficient (R2) of
and concisely used to describe the relationship 0.76. As clearly seen from the figure, the water
between strength and microstructure of porous absorption increased exponentially with increasing
materials, is still being studied [42–45]. To observe porosity. On the other hand, the higher porosity
the relationship between the porosity and compres- mixtures absorbed significantly greater amount of
sive strength of the sample containing MK, the plot water. From these results, it could be concluded that
of the porosity versus the corresponding compres- the mixtures with higher absorption are potentially
sive strength, as shown in Fig. 11, is used. less durable than those with lower absorption owing
Additionally, the variation in the strength with the to their greater porosity.

140
Theoretical model: Sample with MK
y = 146.08e-0.0673x Sample without MK 7
120 R2 = 0.9554 Chan et al. model
Sample with MK
Compressive strength (MPa)

Poon et at. model

6 Sample without MK
Ryshkewitch model Theoretical model:
100 0.0629x
Balshin model y = 1.4382e
2
R = 0.7569
Water absorption (%)

5
80
4
60
3
40
2
20
1
0
0 5 10 15 20
0
Porosity (%) 0 5 10 15 20
Porosity (%)
Fig. 11 Relationship between porosity and compressive
strength Fig. 12 Relationship between porosity and water absorption
948 Materials and Structures (2008) 41:937–949

4 Conclusions samples due to the inclusion of MK. The effect

was particularly beneficial at 20% MK content,
The effectiveness of using MK on the performance where the lowest porosity and the pore diameter
properties of concretes were investigated in this were achieved.
study. From the above experimental results, the
following conclusions are drawn: Acknowledgments The authors would like to thank to Mr.
Vakıf Kiriş for his invaluable assistance during the laboratory
• Concretes with high strength and low shrinkage
phase of the study.
can be made by using Portland cement blended
with ultrafine MK.
• The study showed that the MK provided a signif-
icant increase in both the compressive and splitting References
tensile strengths when used as a modifier in
concrete with varying amounts. When MK replaces 1. Chindaprasirt P, Homwuttiwong S, Sirivivatnanon V
(2004) Influence of fly ash fineness on strength, drying
cement, its positive effect on the concrete strength
shrinkage and sulfate resistance of blended cement mortar.
generally starts at early ages and also noticeable Cem Concr Res 34:1087–1092
increase in the strength was observed at later ages. It 2. Mehta PK, Monteiro P (1997) Concrete. Indian Edition,
was observed that the strength of concretes incor- University of California, Berkeley
3. Ambroise J, Murat M, Pera J (1985) Hydration reaction
porated with MK was up to 30% greater than that of
and hardening of calcined clays and related minerals. Cem
the plain concretes, depending mainly on replace- Concr Res 15:261–268
ment level of MK, w/cm ratio, and testing age. 4. Ding J, Li Z (2002) Effects of metakaolin and silica fume
• For all replacement levels, the MK modified on properties of concrete. ACI Mater J 99(4):393–398
5. Khatib JM, Wild S (1996) Pore size distribution of me-
concretes exhibited remarkably lower shrinkage
takaolin paste. Cem Concr Res 26(10):1545–1553
in comparison to the plain concretes, irrespective 6. Brooks JJ, Megat Johari MA (2001) Effect of metakaolin
of w/cm ratio. It is known that the drying on creep and shrinkage of concrete. Cem Concr Comp
shrinkage is influenced by many factors. The 23:495–502
7. Khatib JM, Clay RM (2004) Absorption characteristics of
results demonstrated that the w/cm ratio was the
metakaolin concrete. Cem Concr Res 34(1):19–29
dominating factor because both the plain and 8. Bai J, Sabir BB, Wild S, Kinuthia J (2000) Strength
especially the MK modified concretes with high development in concrete incorporating PFA and metakao-
w/cm exhibited relatively low drying shrinkage. lin. Mag Concr Res 52(3):153–162
9. Sabir BB, Wild S, Bai J (2001) Metakaolin and calcined
• With regard to the rate of drying shrinkage, it is
clays as pozzolans for concrete: a review. Cem Concr
evident that both plain and MK concretes with low Compos 23:441–454
w/cm ratio showed a somewhat faster development 10. Bai J, Wild S, Sabir BB (2002) Sorptivity and strength of
of shrinkage than those with high w/cm ratio. air-cured and water-cured PC-PFA-MK concrete and the
influence of binder composition on carbonation depth. Cem
However, the drying shrinkage rates of the con-
Concr Res 32:1813–1821
cretes had a decreasing tendency with increased 11. Li Z, Ding Z (2003) Property improvement of Portland
drying time, particularly for the MK concretes. cement by incorporating with metakaolin and slag. Cem
• The inclusion of MK as a partial cement- Concr Res 33:579–584
12. Poon CS, Kou SC, Lam L (2006) Compressive strength,
replacement material provided an excellent
chloride diffusivity and pore structure of high performance
improvement in the pore structure of concrete. metakaolin and silica fume concrete. Constr Build Mater
Irrespective of w/cm ratio, the pore size distribu- 20:858–865
tion was shifted to the smaller pore size range due 13. Gleize PJP, Cyr M, Escadeillas G (2007) Effects of me-
takaolin on autogenous shrinkage of cement pastes. Cem
to the incorporation of MK. The total porosity
Concr Comp 29:80–87
decreased substantially with increasing replace- 14. Güneyisi E, Mermerdaş K (2007) Comparative study on
ment level of MK. The magnitude of this strength, sorptivity, and chloride ingress characteristics of air-
reduction ranged from 22 to 49%, depending cured and water-cured concretes modified with metakaolin.
Mater Struct (in press) DOI 10.1617/s11527-007-9258-5
mainly on w/cm ratio and replacement level of
15. Caldarone MA, Gruber KA, Burg RG (1994) High-reac-
MK. Moreover, there was a considerable reduc- tivity metakaolin: a new generation mineral admixture.
tion in the mean (or median) pore diameter of the Concrete Int 16(11):37–40
Materials and Structures (2008) 41:937–949 949

16. Wild S, Khatib JM, Jones A (1996) Relative strength, 32. TS 500 (2000) Requirements for design and construction of
pozzolanic activity and cement hydration in superplasti- reinforced concrete structures. Turkish Standard, TSE,
cized metakaolin concrete. Cem Concr Res 26:1537–1544 Ankara, Turkey
17. Zhang MH, Malhotra VM (1995) Characteristics of a 33. Zhang MH, Gjorv OE (1991) Mechanical properties of
thermally activated alumino-silicate pozzolanic material high-strength lightweight concrete. ACI Mater J
ant its use in concrete. Cem Concr Res 25:1713–1725 83(3):240–247
18. Khatri RP, Sirivivatnanon V, Yu LK (1997) Effect of 34. Wiegrink K, Marikunte S, Shah SP (1996) Shrinkage
curing on water permeability of concretes prepared with cracking of high-strength concrete. ACI Mater J 93(5):1–8
normal Portland cement and with slag and silica fume. 35. Al-Khaja WA (1994) Strength and time-depended defor-
Mag Concr Res 49(180):162–172 mations of silica fume concrete for use in Bahrain. Constr
19. Khatib JM, Wild S (1998) Sulphate resistance of metaka- Build Mater 8(3):169–172
olin mortar. Cem Concr Res 28(1):83–92 36. Jianyong L, Yan Y (2001) A study on creep and drying
20. ASTM C157 (1997) Standard test method for length shrinkage of high performance concrete. Cem Concr Res
change of hardened hydraulic-cement mortar and concrete. 31:1203–1206
Annual Book of ASTM Standards, 04.02 Concrete and 37. Regourd M (1985) Microstructure of high strength cement
Concrete Aggregates based systems. In: Materials research society proceedings
21. ASTM C39 (1999) Standard test method for compressive series, vol 2. Materials Research Society, Warrendale, PA,
strength of cylindrical concrete specimens. Annual Book of pp 3–17
ASTM Standards, 04.02 Concrete and Concrete Aggregates 38. Bissonnette B, Pierre P, Pigeon M (1999) Influence of key
22. ASTM C496 (1996) Standard test method for splitting parameters on drying shrinkage of cementitious materials.
tensile strength of cylindrical concrete specimens. Annual Cem Concr Res 29:1655–1662
Book of ASTM Standards, 04.02 Concrete and Concrete 39. Feldman RF (1983) Significance of porosity measurements
Aggregates on blended cement performance. In: Malhotra VM (ed)
23. ASTM C642 (1993) Standard test method for specific Proceedings, SP-79, American Concrete Institute, Farm-
gravity, absorption, and voids in hardened concrete. ington Hills, MI, pp 415–433
Annual Book of ASTM Standards, 04.02 Concrete and 40. Feldman RF, Beaudoin JJ (1976) Microstructure and
Concrete Aggregates strength of hydrated paste. Cem Concr Res 6:398–400
24. Day RL, Marsh BK (1988) Measurement of porosity in 41. Mehta PK (1993) Concrete: structure, properties, and
blended cement pastes. Cem Concr Res 18:63–73 materials. Prentice Hall, New York
25. Taylor HFW (1990) Cement chemistry. Academic Press 42. Poon CS, Wong YL, Lam L (1997) The influence of dif-
Inc., London ferent curing conditions on the pore structure and related
26. Khatib JM, Hibbert JJ (2005) Selected engineering prop- properties of fly ash cement pastes and mortars. Const
erties of concrete incorporating slag and metakaolin. Build Mater 11(7–8):383–393
Constr Build Mater 19:460–472 43. Zhang B (1998) Relationship between pore structure and
27. Bai J, Wild S, Gailius A (2004) Accelerating early strength mechanical properties of ordinary concrete under bending
development of concrete using metakaolin as an admix- fatigue. Cem Concr Res 28:699–711
ture. Mater Sci 10(4):338–344 44. Bouguerra A, Ledhem A, De-Barquin F, Dheilly RM,
28. Wild S, Khatib JM, Jones A (1996) Relative strength, Queneudec M (1998) Effect of microstructure on the
pozzolanic activity and cement hydration in superplasti- mechanical and thermal properties of lightweight concrete
cized metakaolin concrete. Cem Concr Res 26:1537–1544 prepared from clay, cement, and wood aggregates. Cem
29. Neville AM (2006) Properties of concrete, 4th edn. Pre- Concr Res 28(8):1179–1190
ntice Hall, Essex, England 45. Chan YN, Luo X, Sun W (2000) Compressive strength and
30. ACI 363R-92 (1997) State of art report on high strength pore structure of high performance concrete after exposed
concrete (ACI Committee 363) (Reapproved 1997). to high temperature up to 800C. Cem Concr Res 30:247–
American Concrete Institute, Farmington Hills, MI 251
31. CEB-FIB (1993) Committee Euro-International du Beton
(CEB-FIB). CEB-FIB Model Code 1990. Thomas Telford,
London