Construction and Building Materials 143 (2017) 92–103

Contents lists available at ScienceDirect

Construction and Building Materials

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

Rheological study of cement paste with metakaolin and/or limestone

filler using Mixture Design of Experiments
Fabiano Nazário Santos a, Sara Raquel Gomes de Sousa b, Antonio José Faria Bombard b,⇑, Sheila Lopes Vieira c
a
Natural Resources Institute (Instituto de Recursos Naturais – IRN), Federal University of Itajubá (Universidade Federal de Itajubá – UNIFEI), Itajubá, MG, Brazil
b
Physics and Chemistry Institute (Instituto de Física e Química – IFQ), Federal University of Itajubá (Universidade Federal de Itajubá – UNIFEI), Itajubá, MG, Brazil
c
Durham, NC, USA

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

Metakaolin acts as a thixotropic

additive when a superplasticizer is
used in cement paste.

Limestone filler does not interfere
with the viscosity or plasticity of the
paste with superplasticizer
admixture.

The ideal metakaolin content
depends on the paste desired
viscosity, thixotropy and workability.

The use of up to 5–10% metakaolin
improve the compressive strength
after 7 days.

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

Article history: Several cement pastes with different amounts of metakaolin (MK) and/or limestone filler (LF) were pre-
Received 27 October 2016 pared. The water/cementitious materials ratio was maintained constant at 0.3, with addition of 0.5% wt/
Received in revised form 27 February 2017 wt of poly-carboxylate ether (PCE) superplasticizer admixture. The following parameters of the fresh
Accepted 1 March 2017
cement pastes were evaluated: the slump and spread, the Marsh funnel time, the plastic viscosity, yield
Available online 22 March 2017
stress, viscoelastic properties and thixotropy. After the curing of 7 day old pastes, compressive strength
tests were performed according to the Brazilian standard using 50  100 mm cylinder specimens. We
Keywords:
conclude that LF alone is not able to avoid segregation or bleeding, and there is no difference between
Rheology
Cement paste
cement pastes mixed with LF and pure OPC pastes, in terms of rheology. On the other hand, if one needs
Metakaolin low slump and low spread, the use of MK is recommended because this material creates a strong, thix-
Limestone filler otropic interconnected net inside of the paste, increasing the yield stress and the thixotropy of the
Mixture design cement paste. By adding 5–10% wt/wt MK, the average increase of compressive strength is approximately
45% at 7 days, compared to the control (only OPC, water and PCE). The maximum recommended amount
of LF or MK substitution in our case was 10% wt.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
The use of superplasticizer (SP) additives (also known as water
reducers) in civil construction has increased in recent years,
⇑ Corresponding author.
E-mail address: bombard@unifei.edu.br (A. José Faria Bombard).
especially in large-sized works or the ones that require the use of
special concretes, such as self-compacting concrete or high perfor-
mance concrete, among others [1,2]. Poly-carboxylate ether (PCE)
is among the additives that has superior performance in terms of
viscosity reduction compared to common plasticizers, such as lig-
nosulfonates and naphthalenesulfonates [3]. On the other hand, if
poly-carboxylate ether is very effective in reducing cement paste

http://dx.doi.org/10.1016/j.conbuildmat.2017.03.001
0950-0618/Ó 2017 Elsevier Ltd. All rights reserved.
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103
viscosity, extra caution with additive overdose is a must. For
instance, a dose of 1 wt.% of water reducing admixture to the
cement leads to cement particle segregation, with cement powder
settling very fast and causing phase separation with a layer con-
taining Portland cement (precipitate) at the bottom of the recipient
and a supernatant containing a lot of water and finer cement par-
ticles. Another negative effect of poly-carboxylate ether overdose is
the loss of thixotropy (and consequently the workability) of the
paste.
Siddique & Khan [4], describe that the use of supplementary
cementitious materials (SCM), such as blast furnace slag, fly ashes,
microsilica, metakaolin, limestone filler, rice husk ashes, among
others, is increasingly growing. The use of such materials may be
advantageous not only by the reduction of economic and environ-
mental costs of Portland cement manufacturing but also because it
can greatly increase the final performance of structures, such as
compressive strength. In addition, some materials may help solv-
ing issues of segregation and workability loss caused by superplas-
ticizers overdose. Martins & Bombard [5], showed that the use of
nanosilica in combination with adequate doses of poly-
carboxylate ether allows the acquisition of a relatively low appar-
ent (plastic) viscosity without workability loss (it maintains the
yield stress and thixotropy of the paste). Pera [6], reports that
the first documented use of metakaolin in a large-scale work was
in the construction of the Jupiá Dam in 1962. Antoni et al. [7],
assessed the replacement of part of a Portland cement segment
with a combination of metakaolin and limestone filler, resulting
in ‘‘45% of substitution by 30% of metakaolin and 15% of limestone
gives better mechanical properties than 100% OPC”. In addition to
that, they argue that ‘‘stoichiometric formation of monocarboalumi-
nate hydrate (MC). . . corresponds to an addition with a weight ratio
of 2:1 metakaolin:limestone.” But the authors did not study the rhe-
ology of mixtures.
A partial literature revision about rheological aspects of cement
pastes with supplementary cementitious materials follows. Cyr
et al. [8], investigated the shear thickening effect of superplasticiz-
ers on the rheological behavior of cement pastes containing or not
mineral additives. They compare the effect of: metakaolin (MK),
quartz (Qtz), fly ash (MFA) or silica fumes (SF). Their superplasti-
cizers (SP) included five different types, but without any detail
about the chemistry of each SP. These authors studied three substi-
tution amounts of Portland cement by the four supplementary
cementitious materials (SCM) above: 0% (only cement), 10% or
25%.They concluded that in terms of shear thickening effect ‘‘can
be amplified (metakaolin), unchanged (quartz, fly ashes) or
reduced (silica fumes)”. Provis et al. [9], studied ‘‘the role of parti-
cle shape” (morphology) of some SCM: ‘‘spherical particles of fly
ash”, ‘‘platy particles of metakaolin”, and the ‘‘angular particles of
blast furnace slag”, ‘‘both in the context of its effect on paste rhe-
ology and on water demand”. The authors focused their report
on particle shape effects in fresh pastes, particle packing and mix
design in geopolymer pastes and geopolymer concretes. However,
they did not mention any water reducer, plastifier or superplasti-
fier. Banfill and Frias studied the rheology of blends cement with
metakaolin or cement with paper sludge wastes, calcined at
700 °C by 2 h [10]. The authors employed a sulfonated naphthalene
formaldehyde condensate as superplasticizer. They concluded that
‘‘the use of low concentrations of calcined paper sludge as a sup-
plementary cementitious material. . . offers a route for utilising this
waste material, as an alternative to the. . . environmental burden
associated with the production of metakaolin from natural kaolin-
ite resources.” Moulin, et al. reported about the effects of ‘‘OPC
blended with 30% (by weight of blend) calcined clay and its rheol-
ogy. However, they also did not use any superplasticizer [11]. Pou-
lesquen et al. studied the rheology of geopolymers prepared with
metakaolin, fumed silice and ‘‘Waterglass activating solutions”,
93
but they did not employ Portland cement, neither any superplasti-
cizer [12]. Janotka et al., investigated in deep the rheology, com-
pressive strength, isothermal calorimetry and setting time of
mixtures of Portland cement with ‘‘metakaolin sands”, a type of
SCM that was not pure metakaolin [13]. Their water/cement ratio
was 0.5, without addition of any water reducer plasticizer. They
concluded: ‘‘. . .the presence of the metakaolin sands reduces the heat
released during the hydration process with respect to non-blended-
cement pastes. The incorporation of metakaolin sand induces a
decrease of the mechanical strength, with the decrease being higher
as the metakaolin sand content increases even though they also pro-
duce a refinement in the pore structure and a decrease of the perme-
ability”. Sonebi et al. made an ‘‘Optimization of rheological
parameters and mechanical properties of superplasticized cement
grouts containing metakaolin and viscosity modifying admixture”,
employing ‘‘Central composite experimental design (CCED)”, a sta-
tistical tool. They employed the same type of superplasticizer we
are studying, PCE. However, they stated: ‘‘The viscosity of cement
grout was determined using a coaxial rotating cylinder viscometer
Fann (smooth cylinders, no serration).” Therefore, slippage could
have occurred during their measurements [14]. Vance et al. pub-
lished a paper with the exact same materials that we are studying.
The title of their paper is: ‘‘The rheological properties of ternary
binders containing Portland cement, limestone, and metakaolin
or fly ash” [15]. However, different from us, they also did not
employ any water reducer admixture. Besides this, in their study,
the water-to-solid ratio (w/s) mass/mass were 0.40 and/or 0.45.
In our study, with the use of PCE SP, we prepared pastes with fixed
water/solids (w/s) ratio = 0.30. Favier et al., compared the rheolog-
ical properties of a geopolymer paste prepared mixing metakaolin
with sodium silicate solution (water glass) versus cement paste.
But not blends of OPC + MK [16]. More recently, Vance et al. com-
pared the rheology of suspensions (pastes) prepared with ‘‘inter-
ground Portland limestone cements”  ‘‘three blended limestone
cements” They described a ratio w/s = 0.45 and again, without
any superplasticizer [17]. Shahriar and Nehdi reported blends of
special cement (oil well API Class G OWC) mixed with four types
of SCMs: MK, SF, (rice husk ashes) RHA, and low calcium FA, with
replacement ranging from 5 to 15%. They also employed a poly-
carboxylate-based high-range water reducing admixture, but with
water-to-binder mass ratio (w/b) of 0.44, which is the usual w/b
recommended for oil well cement formulations. In their study,
they used Design of Experiments too. [18].
For the reader interested in reviewing the significant literature
on the rheology of cement pastes, as well as hundreds of scientific
papers published after 2001, the classical books by Tattersal [19]
and Banfill [20] are advised.
Metakaolin is a material with high pozzolanic activity. In addi-
tion to being advantageous economically and environmentally, it
has the effect of improving mechanical resistance, as compressive
strength, since keeping low amount substitution of OPC by MK
(10%) by such way the hydration heat is similar to 100% OPC [21].
Limestone filler addition to cement accelerates hydration of
Portland clinker grains at early ages, improves the particle packing,
can increase the hydration rate from 1 day to 3 months and pro-
duces the formation of calcium carbo-aluminates (hemicarboalu-
minate or monocarboaluminate), as a result of the reaction
between CaCO3 and C3A of Portland clinker or metakaolin (in case
of ternary blends) [22]. However, if partial substitution of OPC by
LF can be advantageous (same reasons as MK: economic and envi-
ronmental aspects), the formation of carbo-aluminates is a draw-
back, in the case of a sulfate and chloride environment. [23].
Around one hundred papers can be found reporting mixtures of
‘‘limestone AND cement AND metakaolin”. However, very few
[15,17,22,24–27], focus on the rheological properties of ternary
blends of these three cementitious materials. Most of these works,
94 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

the authors focused on self-compacted concrete. This motivated us

to study the rheology of cement pastes (Portland cement contain- H
ing blast furnace slag) blended with metakaolin and/or limestone
n H
filler as additive, using experiment designs for mixtures. -
The pastes were prepared with different amounts of metakaolin O O O
+
and/or limestone filler so that the water/cementitious materials Na
ratio was maintained constant and equal to 0.3 and 0.5 wt.% super- O
plasticizer (poly-carboxylate ether type) added to all the pastes.
The statistical approach of using experiment design is a power-
ful tool in the optimization of mixtures when dealing with multi- O
variables. In the present study, we choose to use the design of p H
experiment with mixtures [28,29]. m

Fig. 1. Chemical structure of poly(sodium carboxylate)-graft-poly(oxy-ethylene) –

2. Experiment PCE – hyper-plasticizer or ‘‘high range water reducer”.

2.1. Materials

Fig. 3 shows the particle size distribution (PSD) cumulative

For this study, ordinary Portland cement class CP II E 32, lime-
plots of PC, MK and LF particles. Measurements of each powder dis-
stone filler (LF) of 97.6% calcium carbonate (CaCO3) purity, meta-
persed in Isopropyl alcohol as liquid carriers were made in Malvern
kaolin (MK) and tap water (w) were used. The chemical
Mastersizer 2000 using laser DLS, according Ferraris et al. [31]. The
composition and other properties of all the cementitious compo-
refraction indices nD of the materials are: LF = 1.60; PC = 1.73 and
nents used can be seen in Table 1.
MK = 1.62.
Hyper-plasticizer (SP) (poly-carboxylate ether, provided by
BASF, Fig. 1) is differentiated from conventional superplasticizer
2.2. Mixtures
in that it is based in a unique carboxylate co-polymer with long lat-
eral (graft) poly(ethylene-oxide) chains. This greatly improves
All blends were prepared in the same manner, following the
cement dispersion [2,3].
recommendations of the standard API 10B (American Petroleum
Institute), Appendix A [32]. For the production of all cement paste
2.1.1. Material characterization mixtures, varying the proportions of limestone and metakaolin
The materials were characterized to confirm the composition, (0%–20%), the mixture proportions is shown in Table 2. Ratio
structure and size of the particles. A FTIR spectrophotometer Per- water/cementitious materials was kept fixed = 0.3 and the amount
kin Elmer, model Spectrum 100 was used to confirm the composi- of SP (0.5% wt./wt.) was also the same all blends.
tion and structure of the superplasticizer (poly-carboxylate ether – The motivation for the choice of the mix proportions using
PCE) and results can be seen in Fig. 2. The spectrum (as well as metakaolin was selected based on works by three different groups:
their peak assignment) is in qualitative agreement with the
reported by Andersson et al. [30]. a) A study by Ambroise, Maximilien and Pera. The authors
The average molecular masses were measured by Gel Per- studied mixes prepared with only Portland cement and
meation Chromatography (GPC Shimadzu, model Prominence, metakaolin, without any limestone, with MK contents: 0,
equipped with refractive index differential detector RID-10A, col- 10, 20, 30, 40 and 50% replacement. They did not report
umns Phenogel 5l 10,000–1,000,000). The GPC was calibrated the use of any water reducer admixture to prepare their
with PEO polymer standards of molar mass known. The sample pastes. Additionally, ‘‘the water:solid ratio (W:S) was not
of PCE was diluted in distilled water. We obtained Mn = 18,831 constant; the water added was adjusted to a constant con-
and Mw = 40,052 g/mol. These values are in reasonable sistency.” They concluded: ‘‘At up to 30% replacement, MK acts
qualitative agreement with the literature (Mn = 12,148 and as an accelerating agent, the pore size distribution is displaced
Mw = 22,174 g/mol). [30]. toward small values, the CH (calcium hydroxide) content is con-
siderably reduced, and compressive strengths are not affected.”
Table 1
However, they showed in their paper, that maximum com-
Chemical analysis of Portland cement type II (OPC), metakaolin (MK) and limestone pressive strength was obtained with 10% MK replacement.
filler (LF) There was an improvement in strength, compared to control
Analysis OPC MK LF
(100% OPC), with 10% MK. With 20% MK replacement, the
compressive is practically the same of pre OPC. With 30%
SiO2 (wt.%) 23.52 57 1.6
CaO (wt.%) 56.54 <0.1 53
MK and above, the strength was reduced compared to con-
(CaO + MgO) trol, decreasing the strength with MK replacement. There-
Fe2O3 (wt.%) 3.84 2 <0.1 fore, we chose to test 20% MK (or LF) maximum
Al2O3 (wt.%( 6.52 34 1.4 replacement [33].
Na2O Alkaline equivalent (wt.%) 0.70 <1.5 0.19
b) Helal (2002) studied five cement–limestone blends, using
MgO (wt.%) 1.56 <0.1 0.43
SO3 (wt.%) 1.68 <0.1 <0.1 0%, 5%, 10%, 15%, and 20% of limestone as a partial sub-
TiO2 (wt.%) – 1.5 – stituent of Portland cement, and the cement pastes were
P2O5 (wt.%) 0.29 – – prepared using the standard water of consistency of 0.255,
Mn2O3 (wt.%) 0.80 – – 0.255, 0.258, 0.261, and 0.263, respectively. There was no
Loss on ignition (wt.%) 5.10 3 44
Specific gravity 3.09 2.56 2.60
significant gain in compressive strength compared to 100%
S.S. Blaine 3820 cm2/g – – OPC [34].
BET Area (m2/g) – 23 – c) Antoni et al. (2012) studied ‘‘Cement substitution by a com-
Pozzolanic activity (Chapelle test) – 0.88 g Ca(OH)2 /g – bination of metakaolin and limestone” and they concluded:
Residue # 200 (wt.%) 3.4 1.2 <1
‘‘TGA (Thermogravimetric Analysis) shows that the reactions
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103 95

100

3410
80

1466
1572
T (%)
60
2881

1342,8

841,9
961
40

FTIR with ATR of PCE admixture

20
(sample dried vacuum oven 80°C) 1101,7

4000 3500 3000 2500 2000 1500 1000 500

wavenumber (cm-1)

Fig. 2. FTIR spectrum of hyper-plasticizer poly-carboxylate ether.

Fig. 3. Cumulative size distribution of Portland cement powder, metakaolin and limestone filler, measured by laser Dynamic Light Scattering, employing isopropyl alcohol as
carrier liquid.

of metakaolin and limestone consume calcium hydroxide,
which may be completely absent in blends with high levels
of substitution at late ages. The metakaolin appears to react
faster in the system with limestone than in the binary meta-
kaolin/Portland cement blend. Also, the limestone reacts fas-
ter in the system with metakaolin than in the binary
limestone/Portland cement blend. These results point to
strong synergistic effects with coupled substitutions of this
type. Of course the consumption of calcium hydroxide could
mean that the high substitution level blends may carbonate
more rapidly. This and other aspects of durability are cur-
rently being studied.” [7].
Therefore, based on these previous studies, we choose, with
help of Mixture Design of Experiments [28,29], to investigate
blends binary or ternary with OPC, MK and LF, with maximum
20% replacement SCM.
2.3. Methods
The flow time of each fresh paste was measured with the Marsh
funnel making it possible to measure the viscosity following the
ASTM D6910 [35,36]. Part of the paste was separated (400 ml)
to perform the rheological measurements using stress-controlled
rheometer (Physica MCR-301, Anton Paar, Germany) equipped
with the measuring system (a stirrer with two hollow vanes, model
ST59-2V-44.3/120). The measuring rotor dimensions are: outer
diameter 59.00 mm, and length 44.3 mm. The cup of the rheometer
have inner diameter 70 mm, and a basket inset cage with serra-
tions inside which prevents wall slippage. Fig. 4 shows the stirrer.
96 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

Table 2 Amplitude sweep ramp of strain c from 0.01% to 100% with

Proportions of Portland cement, limestone and metakaolin. slope 6pt./dec, oscillation frequency 10 rad/s and
Portlant cement (%) Metakaolin (%) Limestone (%) temperature = 25 °C.
100 0 0 Frequency sweep (keeping the strain c = 0.02% constant and
90 10 0 ramp angular frequency from 500 rad/s to 0.05 rad/s with slope
90 0 10 5pt./decade, and temperature 25 °C.
95 5 0
95 0 5
90 5 5
The rheological measurements were performed satisfactorily
93.33 3.33 3.33 during all of the tests and samples did not present any segregation
96.66 1.67 1.67 or slippage of the pastes, due to the use of the appropriate method
91.66 6.67 1.67 and equipment (cup with inner serrated cage and two vane rotor).
91.66 1.67 6.67
The deformation capacity of the cement paste was verified from
93.33 3.33 3.33
93.33 3.33 3.33 the slump flow type test, modified according Roussel and Coussot
93.33 3.33 3.33 [39]. The slump flow test was performed using a square glass plate
93.33 3.33 3.33 with an 80 cm edge. A mold was placed in the center of the glass
80 20 0 plate, then added to the paste inside the cylindrical mold with
80 0 20
dimensions ø 96.8 mm  100.8 mm. Thereafter, the mold was
80 10 10
86.66 6.67 6.67 lifted vertically so that the cement paste flowed until it reached
96.66 1.67 1.67 an equilibrium. After that, the cement paste spreading value was
measured, using the average of two perpendicular diameters. At
the same time, we obtained the slump values (when no spreading
occurs), taking into account the other authors’ essay [39].
All the rheological parameters, and their units followed the rec- For the compression test, sample test specimens were produced
ommendations of standard ASTM C1749-12: ‘‘Standard guide for following the Brazilian standard NBR 7215 [40]. This standard
measurement of the rheological properties of hydraulic cementi- describes the method for the determination of the compressive
tious paste using a rotational rheometer” [37]. With this, the fol- strength of cylindrical specimens measuring 50 mm in diameter
lowing rheological tests were possible: and 100 mm in height. When demolded, the specimens were
placed in a tank containing water saturated with lime for the cure
Flow curve (up-hold-down 3 intervals), with temperature kept until the date of its rupture, at 7 days (or 28 days). At the date of
constant at 25 °C. their rupture, the specimens were removed from the tank and
1st interval: up-ramp of shear rate c_ initial 0.01–100 s1, capped with a thin layer of cement paste in order to regularize
during 205 s. their surfaces. The ruptures were made in a press with capacity
2nd interval: hold shear rate c_ constant of 100 s1 during for 100 metric tons, from the manufacturer TIME Testing Machi-
50 s. nes, model WAW – 1000 C (China). This press is controlled by
3rd interval: down-ramp of shear rate c_ initial 100–0.01 s1, computer and is of the universal type, with control Servo electro-
during 205 s. hydraulic. For the rupture of the ø 50 mm  100 mm specimens, a
Thixotropy (three interval time test) [38] – performed at con- ‘‘RILEM” device (an adapter of the testing compression machine,
stant temperature 25 °C. equipped with two steel bearing blocks with hardened faces, used
First, using shear rate c_ constant of 0.1 s1 during 60 s to better centralize the specimens and distribute equally the force
Then, shear rate c_ constant of 100 s1 during 50 s applied, according note 3 of the ASTM C39/C39-14), was employed
Finally, shear rate c_ constant of 0.1 s1 during 250 s [41].

70
LF
Compressive strength 7 days (MPa)

60 MK

50

40

30

20

10

0
0 5 10 15 20 25
SCM substitution (%)

Fig. 5. Compressive strength after 7 days for binary blends of Portland cement with
supplementary cementitious materials: OPC + MK or OPC + LF. Columns heights are
Fig. 4. Stirrer with two hollow vanes, model ST59-2V-44.3/120 (Anton Paar, the average value for 3 test specimens rupture. Error bars are the standard
Germany). deviation.
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103 97

Table 3
Compressive strength after 7-day cure.

CP (%) MK (%) FC (%) R1 (MPa) R2 (MPa) R3 (MPa) Mean (MPa) Std Dev
100.000 0.0000 0.0000 33.14 30.79 50.40 38.11 10.71
96.667 1.6667 1.6667 61.00 49.00 54.00 54.67 6.03
95.000 5.0000 0.0000 56.27 51.00 59.00 55.42 4.07
95.000 0.0000 5.0000 36.53 53.87 44.91 45.10 8.67
93.333 3.3333 3.3333 40.60 40.65 40.45 40.57 0.10
91.667 6.6667 1.6667 40.75 40.75 40.47 40.66 0.16
91.667 1.6667 6.6667 54.92 42.83 27.63 41.79 13.67
90.000 10.0000 0.0000 56.00 50.00 48.00 51.33 4.16
90.000 0.0000 10.0000 51.38 40.90 43.00 45.09 5.54
90.000 5.0000 5.0000 56.00 46.00 61.00 54.33 7.64
93.333 3.3333 3.3333 41.44 42.30 40.00 41.25 1.16
93.333 3.3333 3.3333 43.76 42.43 41.00 42.40 1.38
93.333 3.3333 3.3333 38.00 40.57 41.00 39.86 1.62
93.333 3.3333 3.3333 41.95 39.76 39.02 40.24 1.54
96.667 1.6667 1.6667 38.44 38.66 42.53 39.88 2.30
80.000 20.0000 0.0000 39.28 44.09 35.00 39.46 4.55
80.000 0.0000 20.0000 48.17 43.53 40.35 44.02 3.93
80.000 10.0000 10.0000 47.00 46.89 46.00 46.63 0.55
86.667 6.6667 6.6667 45.00 43.00 33.00 40.33 6.43
90.000 0.0000 10.0000 52.00 50.99 39.00 47.33 7.23
90.000 10.0000 0.0000 58.41 53.44 57.84 56.56 2.72

3. Results and discussion 3.2. Slump, spreading and Marsh funnel time

3.1. Compressive strength after one week
As our focus was on the rheology of the fresh mixtures, com-
pressive strength was only measured with 7 days curing time.
Fig. 5 shows the results for binary blends. The results for all of
the mixtures are in Table 3.
One can see in Fig. 5 that the maximum strength in our formu-
lations seems to be around 5% wt. substitution of Portland cement.
Above 10% wt. MK, the strength start decreases. Besides, the paste
with 20% wt. substitution OPC by MK, was impossible to measure
in the rheometer, because the maximum torque limit of the instru-
ment. In other words: the paste OPC:MK 80:20 was so plastic that
its workability is bad. Even mixing this particular blend was diffi-
cult. Therefore, we focused the study on involving mixtures and sta-
tistical analysis in the range 0–10% wt. substitution OPC by SCM.
The compressive strength of the pastes did not present great
changes compared to the control paste (only cement). However,
in the case that cement was replaced with 5% of metakaolin, the
compressive strength had a 45% average increase.
The spreading and the slump of each paste was measured in the
same test. Next, the time flow of each paste was measured using
the Marsh funnel. Table 4 summarizes the results for all tested
pastes. In some cases, it was not possible to measure all of these
3 responses.
Fig. 6 shows the effect of the increase of metakaolin content as a
function of the spreading (left axis) and the funnel time (right axis).
The dotted line in Fig. 6 at 100 mm represents the spreading
threshold because this value is the internal diameter of modified
frustum. Therefore, spreadings smaller than 100 mm do not make
sense in the test. The curves are only a guideline for the eyes. The
paste with 10% of metakaolin (or more) does not flow inside the
Marsh funnel.
3.3. Yield stress measured by means of oscillatory amplitude sweep
The yield stress can be experimentally measured in different
ways. However, the result strongly depends on the measurement
technique. One of them is the strain amplitude sweep test,

Table 4
Slump, spreading and Marsh funnel time for the tested pastes.

Portland cement (%) Metakaolin (%) Limestone filler (%) Slump (mm) Spreading (mm) Marsh Funnel (seconds)
*
100 0 0 602 31.41
* *
90 10 0 35
*
90 0 10 615 44.78
*
95 5 0 405 112.54
*
95 0 5 604 45.53
*
90 5 5 415 77.35
*
93.33 3.33 3.33 462 58.44
*
96.66 1.66 1.66 530 51.03
*
91.66 6.66 1.66 324 139.56
*
91.66 1.66 6.66 575 45.16
*
93.33 3.33 3.33 491 48.03
*
93.33 3.33 3.33 540 47.38
*
93.33 3.33 3.33 511 49.63
*
93.33 3.33 3.33 526 48.37
* *
80 20 0 11
*
80 0 20 538 71.76
* *
80 10 10 20
* *
86.66 6.66 6.66 50
*
96.66 1.66 1.66 562 42.88
*
Asterisk indicates that it was not possible to measure slump/spreading/time.
98 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

700 160

600 140

Marsh Funnel Time (sec)

500 120
Spread (mm)

400 100

300 80

200 60

100 40
Spread >/= 100 mm

0 20
0 2 4 6 8 10
Metakaolin amount (%)

Fig. 6. Effect of metakaolin content on the spreading of the pastes (left axis) and on
the Marsh funnel time (right axis).

performed in a controlled-stress (or strain) oscillatory rheometer.

In this case, an increasing logarithmic strain ramp is applied to
the sample to be analyzed. Although the strain, which varied from
0.01 to 100%, is in fact the independent variable, a plot of the elas-
tic modulus (or complex viscosity) can be built as a function of the
applied shear stress. It is considered that the sample underwent
plastic deformation (it flowed) where the function G0 = f (r) has
a point of inflection. Thus, the yield stress can be easily obtained
for each paste if the necessary torque is not greater than the max-
imum torque of the rheometer, which is 200 mN-m. As an exam-
ple, Fig. 7 shows G0 (elastic modulus) curves as a function of the
applied shear stress for the control paste (Portland cement, water
and PCE) and the other pastes. More details explaining how the
yield stress and complex viscosity values were measured can be
found in reference [5].
Fig. 7 shows that increasing the metakaolin content will consid-
erably increase the pastes’ yield stress and the G0 value. For exam-
ple, the yield stress is approximately 1 Pa for pastes with 10% filler
Fig. 8. Mixture contour plot (component amounts) of yield stress measured in
oscillatory mode. PC = Portland Cement, MK = Metakaolin, LF = limestone filler.
and for the control. As metakaolin is added to the sample, the yield
stress shows values of 3 Pa for the sample containing 5% metakao-
lin, and values between 20 and 30 Pa for the sample containing 10%
filler. On the other hand, the use of limestone filler (at least up to
10%) does not change the yield stress considerably when compared
to the control.
The yield stress results for mixtures with maximum 10 wt%
substitution were analyzed with help of Minitab software (DOE
Mixtures) considering pseudo-components instead component
amounts. The results were fitted with a Full Cubic Model, excluding
those terms, which inflate the variance. Fig. 8 shows the contour
plot of the full cubic model for the yield stress, which was obtained
using mixture design data analysis. Tables 5 and 6 resumes the
Regression analysis for yield stress of the mixtures: Yield VA versus
PC; MK; LF.
The metakaolin coefficient value of 28.93 in Table 5 confirms
that this additive contributes to the increase of the yield stress
much more than the cement itself or the limestone filler. A full
cubic model (R2 adj. = 99.55%), is much more complex than other,
simpler multilinear regression models (linear, quadratic, special
cubic), that were also tested. But the best fitting, with acceptable
‘‘lack-of-fit” value was reached only with a full cubic model. For
the linear model, standard errors were larger than corresponding
coefficient’s. Quadratic or special cubic models works better than
linear model. However, these models resulted with lack-of-fit,
and thus, we choose the full cubic model for yield stress.

3.4. Complex viscosity of pastes measured using oscillatory frequency

sweep test

In this case, an under fixed and constant strain value of 0.02%

Fig. 7. Elastic modulus (G0 ) as a function of the shear stress (r) for the following
pastes: 10% of filler (blue triangles); control (black line); 5% metakaolin + 5% filler
(green circles) and 10% metakaolin (red squares). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version
of this article.)
(within the linear viscoelastic region), a decreasing logarithmic fre-
quency sweep from 500 to 0.08 rad/s, is applied to the samples and
the following parameters are measured: viscoelastic moduli
(G0 and G0 ) and complex viscosity (g⁄). Fig. 9 shows the complex
viscosity obtained for some pastes and Table 7 summarizes the val-
ues obtained for G0 , G0 and g⁄ (measured using frequency sweep
tests at an angular frequency of x = 5 rad/s) for all pastes.
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103 99

Table 5
Estimated regression coefficients for Yield VA (pseudo-components).

Term Coef SE Coef T P VIF

PC 1.19 0.4516 * * 2.097
MK 28.93 0.4968 * * 2.537
LF 0.89 0.4516 * * 2.097
PC * MK 48.81 2.4044 20.30 0.000 3.481
MK * LF 50.21 2.4044 20.88 0.000 3.481
PC * MK * LF 51.24 10.6317 4.82 0.003 3.586
PC * MK * () 27.72 8.0912 3.43 0.014 1.625
MK * LF * () 25.32 8.0912 3.13 0.020 1.625
S = 0.498610 PRESS = 159.549
R-Sq = 99.79% R-Sq(pred) = 77.63% R-Sq(adj) = 99.55%

Table 6
Analysis of variance for yield VA (pseudo-components).

Source DF Seq SS Adj SS Adj MS F P

Regression 7 711.603 711.603 101.658 408.90 0.000
Linear 2 437.694 583.813 291.907 1174.15 0.000
Quadratic 2 256.392 184.979 92.490 372.02 0.000
PC * MK 1 139.621 102.442 102.442 412.05 0.000
MK * LF 1 116.770 108.403 108.403 436.03 0.000
Special Cubic 1 6.746 5.774 5.774 23.22 0.003
PC * MK * LF 1 6.746 5.774 5.774 23.22 0.003
Full cubic 2 10.772 10.772 5.386 21.66 0.002
PC * MK * () 1 8.337 2.918 2.918 11.74 0014
MK * LF * () 1 2.434 2.434 2.434 9.79 0020
Residual error 6 1.492 1.492 0.249
Lack-of-fit 2 0.720 0.720 0.360 1.86 0.268
Pure error 4 0.772 0.772 0.193
Total 13 713.095

10000 PC-MK-LF (%)

Metakaolin effect 100-00-00
1000 90-3.3-6.7
95-05-00
100 91.6-6.7-1.7
90-10-00
Complex viscosity (Pa.s)

10

10000 PC-MK-LF (%)

Filler effect 100-00-00
1000 93-3.3-3.3
95-00-05
100 90-3.3-6.7
90-00-10
10

0,1 1 10 100 1000

Angular frequency (rad/s)

Fig. 9. Complex viscosity as a function of angular frequency for control pastes containing 3.3, 5, 6.7, and 10% metakaolin or limestone filler. This test was performed in
oscillatory mode at constant strain of 0.02%.

Fig. 9 shows that metakaolin contents above 3.3% lead to an
increase of complex viscosity throughout entire analyzed fre-
quency range and that the pastes exhibited a pseudo plastic behav-
ior. On the other hand, the limestone filler did not cause any
change in viscosity, except for the region above 100 rad/s, where
shear-thickening is observed. The shear-thickening behavior
increased with filler content increase when no metakaolin was
added.
All the values measured of viscoelastic moduli G0 and G00 , as well
as the complex viscosity, are summarized in the Table 7.
3.5. Thixotropy of fresh pastes measured using ‘‘3 ITT” test
A three-interval of time test was used to obtain and evaluate
thixotropy (even if only for comparison) between the different
cement pastes formulations. According to MEZGER (2011), this test
100 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

Table 7
Elastic and viscous moduli (G0 e G00 ) and complex viscosity of pastes measured in frequency sweep tests.

Portland Cement (%) Metakaolin (%) Limestone filler (%) G’ (Pa) G” (Pa) g* (Pa.s)
100 0 0 36 7.3 7.3
90 10 0 790 83 158
90 0 10 47 8.8 9.5
95 5 0 67 8.3 13.6
95 0 5 33 5.8 6.6
90 5 5 68 8.2 13.7
93.33 3.33 3.33 64 8.0 12.8
96.66 1.66 1.66 42 6.6 8.5
91.66 6.66 1.66 122 13.6 24.6
91.66 1.66 6.66 38 6.4 7.6
93.33 3.33 3.33 35 5.5 7.0
93.33 3.33 3.33 29 5.6 6.0
93.33 3.33 3.33 38 6.2 7.7
93.33 3.33 3.33 37 6.8 7.5
* * *
80 20 0
80 0 20 40 6.0 8.0
80 10 10 1936 292 392
86.66 6.66 6.66 554 64.4 111.5
96.66 1.66 1.66 39 7.4 8.0
*
Asterisk indicates that was not possible to measure this paste in the rheometer, due torque limit.

be used as a reference. In the second time interval, the rotation is

1st interval
1000 suddenly increased to a value 1000 times greater and then it is
maintained constant for a period of time (50 s was the time used
in the present study). The high shear rate of the second time inter-
3rd interval val aims to break the sample’s entire internal structure. Finally, the
100
rotation in the third and last time interval (immediately after the
Viscosity (Pa.s)

second interval) is the same as the one used in the first interval.
In the third interval, the viscosity is monitored during a relatively
10 greater time and its values are recorded every 0.5 s to verify how
much of the initial viscosity is recovered and how much time does
it take for this recovery. In our case, the rotation profile and the
MK 6.6, LF 1.6% interval times used were as follows: 1st interval (0.1 rpm, 60 s,
1 MK 5%, LF 0
2nd interval 12 data points), 2nd interval (100 rpm, 50 s, 100 data points),
MK 5%, LF 5%
MK 3.3, LF 3.3% and 3rd interval (0.1 rpm, 250 s, 500 data points). Fig. 10 shows
MK 0, LF 10% some of the curves obtained in the 3-ITT thixotropy tests.
0,1 Fig. 10 shows that the pastes viscosity increases with metakao-
0 60 120 180 240 300 360 lin content. In addition, thixotropy (which is measured through
Time (s) recovery time (s) for a certain viscosity recovery level and/or vis-
cosity recovery degree (%) after 1 min) also strongly depends on
Fig. 10. Three-Interval thixotropy test. Viscosity as a function of time for 5 pastes
with different contents of metakaolin (MK) and limestone filler (LF).
100 160
MK
Thixotropic Recovery after 1 min (%)

is also called ‘‘3ITT” (three-interval of time test) [38]. In the first 90 LF 140
time interval of this thixotropy test, the sample is sheared under
Recovery time @ 63.2% (sec)

constant low rotation for 1 min to obtain the viscosity, which will 80 120

70 100
Table 8 Control: only
Thixotropy of cement-metakaolin-filler pastes Cement + water + SP
60 80
Paste Thixotropic recovery (%) Time (s) for 63.2% recovery
PC-MK-LF (%) after 1 min (1  1/e) of viscosity 50 60
91.6-6.7-1.7 66 47
95-5.0-0.0 55 105 40 40
90-5.0-5.0 57 82
93.4-3.3-3.3 (b) 52 99 30 MK 20
90-0.0-10 43 110 LF
90-10-0.0 90 14 20 0
96.6-1.7-1.7 45 126 0 5 10
91.6-1.7-6.7 38 158
95-0.0-5.0 48 99 Suppl. Cimentitious Mat. (%)
100-0.0-0.0 39 143
93.4-3.3-3.3 (a) 50 111 Fig. 11. Effect of metakaolin and limestone filler contents on the pastes thixotropy
93.4-3.3-3.3 (c) 49 123 measured on the 3rd interval after 1 min (left axis, black) and time for the viscosity
93.4-3.3-3.3 (d) 53 96 to return to 63.2% (1  1/e) of the reference value (right axis, red). (For interpre-
tation of the references to colour in this figure legend, the reader is referred to the
(a,b,c,d): replicate samples. web version of this article.)
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103 101

Fig. 12. Mixture contour plot of thixotropic recovery (%) after 1 min. PC = Portland
Cement, MK = Metakaolin, LF = limestone filler. Fig. 13. Mixture contour plot (component amounts) for plastic viscosity measured
in rotational mode.

Table 9
Table 11
Estimated Regression coefficients for thixotropic recovery after 1 min (pseudo
Estimated regression coefficients for viscosity plastic (pseudo-components).
components).
Term Coef SE Coef T P VIF
Term Coef SE Coef T P VIF
PC 1.31 0.5312 * * 1.884
PC 39.38 5.316 * * 1.274
MK 11.18 0.5893 * * 2.319
MK 80.38 5.316 * * 1.274
LF 1.31 0.5312 * * 1.884
LF 40.38 5.316 * * 1.274
PC * MK 15.74 2.5925 6.07 0.000 2.628
S = 7.36259 PRESS = 1094.23
MK * LF 18.64 2.5925 7.19 0.000 2.628
R-Sq = 75.17% R-Sq(pred) R-Sq(adj)
S = 0.618696 PRESS = 38.2884
= 49.88% = 70.20%
R-Sq = 96.73% R-Sq(pred) = 63.66% R-Sq(adj) = 95.28%

Table 10
Analysis of variance for thixotropic recovery after 1 min (pseudo components). Table 12
Analysis of variance for viscosity plastic (pseudo-components).
Source DF Seq SS Adj SS Adj MS F P
Source DF Seq SS Adj SS Adj MS F P
Regression 2 1641.00 1641.00 820.500 15.14 0.001
Linear 2 1641.00 1641.00 820.500 15.14 0.001 Regression 4 101.925 101.9249 25.4812 66.57 0.000
Residual error 10 542.08 542.08 54.208 Linear 2 63.523 89.0158 44.5079 116.27 0.000
Lack-of-fit 7 532.08 532.08 76.011 22.80 0.013 Quadratic 2 38.402 38.4020 19.2010 50.16 0.000
Pure error 3 10.00 10.00 3.333 PC * MK 1 18.612 14.1184 14.1184 36.88 0.000
Total 12 2183.08 MK * LF 1 19.790 19.7904 19.7904 51.70 0.000
Residual error 9 3.445 3.4451 0.3828
Lack-of-fit 5 3.374 3.3741 0.6748 38.06 0.002
Pure error 4 0.071 0.0709 0.0177
Total 13 105.370
metakaolin content. Table 8 summarizes thixotropy results for the
tested pastes whenever measurement was possible.
Fig. 11 shows two ways of evaluating thixotropic results: per-
centage recovery of the viscosity after arbitrary time (we choose
1 min), and the time spent to recovery the same level of viscosity In the case of the regression model of thixotropic recovery, the
the 1st interval (commonly this level can be 63.2%). linear model was better than other, more complex models. The
One can see in Fig. 11 that the greater the added kaolin content, thixotropic recovery it is independent of the limestone filler Port-
the greater the thixotropy. On the other hand, this also leads to less land cement ratio and is a function only of the metakaolin content.
spreading and higher Marsh funnel time. Therefore, we concluded Tables 9 and 10 summarize the linear regression and ANOVA for
that metakaolin increases plasticity and thixotropy of cement thixotropic recovery after 1 min.
paste, i.e., it improves paste workability for metakaolin contents Fig. 13 shows the contour plot (in terms of component
of 5 to 8% but makes workability much more difficult for contents amounts) for plastic (or apparent) viscosity, measured in rotational
above 10%. mode in rheometer. The regression model here was quadratic,
Fig. 12 shows the contour plot for the percentage thixotropy because linear was insufficient to explain the viscosity results.
recovery level of the pastes after 1 min. The linear model is a good More complex models are not worth being employed, due to the
fit; and indicates how prevalent the effect of the metakaolin con- ‘‘lack-of-fit” changes. The quadratic model is the simplest with
tent over thixotropy is. acceptable R2 (adjust or predictable).
102 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

Fig. 15. Overlaid contour plot for yield stress and elastic modulus (oscillatory
rheometry); thixotropic recovery after 1 min and plastic viscosity (rotational
rheometry).

Fig. 14. Mixture contour plot (component amounts) for elastic modulus (G0 )
measured in oscillatory mode, in the rheometer. Fig. 14 clearly suggests that, if one is looking for pastes with an
elastic modulus of G0 > 120 Pa in these particular blends (PC + MK
+ LF, w/b = 0.30 and 0.5% wt. PCE), a minimum amount of 6% wt. of
Table 13 MK must be present in the formulations.
Estimated regression coefficients for elastic modulus (pseudo-components). However, let us see other responses in combination, since the
Term Coef SE Coef T P VIF ideal outcome would be a paste with plastic viscosity as low as
possible; fast thixotropic recovery; yield stress and elastic modulus
PC 30 15.51 * * 2.097
MK 787 17.06 * * 2.537 as high as possible. This is not an easy task. Fig. 13 shows an over-
LF 41 15.51 * * 2.097 laid plot of some of the responses. We arbitrarily chose: a)
PC * MK 1387 82.57 16.80 0.000 3.481 60 < G0 < 120 Pa; b) 2 < yield stress < 10 Pa; c) 2.5 < Plastic viscos-
MK * LF 1405 82.57 17.02 0.000 3.481 ity < 3.5 Pa.s and d) 60% < thixotropic recovery after 100 < 70%. Of
PC * MK * LF 1626 365.12 4.45 0.004 3.586
PC * MK * () 1139 277.87 4.10 0.006 1.625
course, in an ‘‘perfect world”, plastic viscosity should be less than
MK * LF * () 995 277.87 3.58 0.012 1.625 1 Pa.s, thixotropic recovery should approaches 100% as quick as
S = 17.1235 PRESS = 218,282 possible, and yield value or G’ should be as high as possible. The
R-Sq = 99.66% R-Sq(pred) = 57.66% R-Sq(adj) old problem of optimization of functions with maximums and min-
= 99.26%
imums. . . One solution, not ideal, but real and acceptable, is dis-
played in Fig. 15, as an overlaid contour plot.
The Fig. 15 suggests that a paste prepared with PC:MK:LF
90:5:5 must fulfill all the requirements, inside the white region
Table 14 delimited.
Analysis of variance for elastic modulus (pseudo-components).

Source DF Seq SS Adj SS Adj MS F P 4. Conclusions

Regression 7 513,764 513,764 73,395 250.31 0.000
Linear 2 289,295 424,691 212,345 724.20 0.000
Metakaolin acts as a thixotropic additive and a modifier of the
Quadratic 2 199,969 147,094 73,547 250.83 0.000
cement paste viscosity, which did not happen in the case of
PC * MK 1 110,229 82,754 82,754 282.23 0.000
MK * LF 1 89,740 84,916 84,916 289.60 0.000
limestone filler. There is an improvement of paste workability
Special cubic 1 7061 5814 5814 19.83 0.004 when a water reducer admixture is used.
PC * MK * LF 1 7061 5814 5814 19.83 0.004
Excess of superplasticizer (SP), such as poly-carboxylate ether,
Full Cubic 2 17,439 17,439 8719 29.74 0.001 may reduce paste workability especially when contents above
PC * MK * () 1 13,682 4923 4923 16.79 0.006
0.5 wt.% SP are used (even with low water/cement ratio). The par-
MK * LF * () 1 3756 3756 3756 12.81 0.012
Residual error 6 1759 1759 293 tial replacement of Portland cement with metakaolin (up to 10%
Lack-of-fit 2 1026 1026 513 2.80 0.174 wt/wt) maintains workability. On the contrary, binary blends with
Pure error 4 733 733,183 only cement and limestone filler was not able to maintain it.
Total 13 515,523

Because limestone filler does not interfere with the viscosity or
plasticity of the paste with PCE admixture, its use is preferred
when fluidity is desirable. However, with only limestone filler
Tables 11 and 12 summarizes the quadratic regression model and Portland cement, segregation can occurs, especially if
and ANOVA for plastic viscosity. higher contents of this superplasticizer is employed.
Finally, Fig. 14 shows the mixture contour plot for elastic mod-
The ideal metakaolin content depends on the paste desired vis-
ulus, G0 measured in oscillatory rheometer. The elastic modulus, in cosity, thixotropy and workability. For example, the thixotropy
the same trend as yield stress value, are both related to the cohe- recovery of 62.3% only takes 14 s for blend 90% cement with
sion of the pastes, and thus, their ‘‘workability”. Table 13 and 14 10% metakaolin. The slump was of 35 mm; however, this fast
resumes the full cubic regression model and ANOVA, respectively, recovery leads to a loss of fluidity: the paste does not flow
for elastic modulus (pseudo-components). through the Marsh funnel. This also avoids segregation.
 F. Nazário Santos et al. / Construction and Building Materials 143 (2017) 92–103

The use of up to 5–10% metakaolin substitution may result in
30–45% average increase in compressive strength after 7 days.

A paste containing 90% Portland cement, blended with 5% Meta-
kaolin and 5% limestone filler, should present good thixotropy
with 60% viscosity recovery after 1 min, elastic modulus above
60 Pa, and plastic viscosity 3.5 Pa.s The compressive strength
(7 days) of this blend was 54 ± 8 MPa.
Acknowledgements
The authors give special thanks to BASF SA for donating the
chemical additives and to Metacaulim do Brasil Indústria e Comér-
cio Ltda (Jundiaí – SP, Brazil) for donating us MK. AJFB thanks the
financial support provided by CAPES and FAPEMIG (Grants: Rede
Mineira de Química CEX-REDE-00010-14, PEP-00680-13, PEP-
00231-15, ETC-00043-15, PEE-00081-16, PEP-00026-17 and APQ-
01824-17). We are in debt to Professor Fábio H. Florenzano (USP
Lorena) by GPC analysis of PCE water reducer. Reading and com-
ments by Prof. Anderson P. Paiva, and Prof. Pedro P. Balestrassi as
well as the reviewer’s suggestions to improve the manuscript are
deeply appreciated.
References
[1] P.K. Mehta, P.J.M. Monteiro, Concrete – Microstructure, Properties And
Materials, third ed., McGraw Hill, New York, 2006.
[2] R. Rixom, N. Mailvaganam, Chemical Admixtures For Concrete, third ed., E & Fn
Spon, London, 1999.
[3] A. Papo, L. Piani, Effect of various superplasticizers on the rheological
properties of portland cement pastes, Cem. Concr. Res. 34 (2004) 2097–2101.
[4] R. Siddique, M.I. Khan, Supplementary Cementing Materials, Springer-Verlag,
Berlin, 2011.
[5] R.M. Martins, A.J.F. Bombard, Rheology of fresh cement paste with
superplasticizer and nanosilica admixtures studied by response surface
methodology, Mater. Struct. 45 (2012) 905–921.
[6] J. Pera, Guest editorial – metakaolin and calcined clays, Cem. Concr. Compos.
23 (6) (2001) iii.
[7] M. Antoni, J. Rossen, F. Martirena, K. Scrivener, Cement substitution by a
combination of metakaolin and limestone, Cem. Concr. Res. 42 (2012) 1579–
1589.
[8] M. Cyr, C. Legrand, M. Mouret, Study of the shear thickening effect of
superplasticizers on the rheological behaviour of cement pastes containing or
not mineral additives, Cem. Concr. Res. 30 (2000) 1477–1483.
[9] J.L. Provis, P. Duxson, J.S.J. van Deventer, The role of particle technology in
developing sustainable construction materials, Adv. Powder Technol. 21 (1)
(2010) 2–7.
[10] P.F.G. Banfill, M. Frias, Rheology and conduction calorimetry of cement
modified with calcined paper sludge, Cem. Concr. Res. 37 (2007) 184–190.
[11] E. Moulin, P. Blanc, D. Sorrentino, Influence of key cement chemical
parameters on the properties of metakaolin blended cements, Cement Concr.
Compos. 23 (2001) 463–469.
[12] A. Poulesquen, F. Frizon, D. Lambertin, Rheological behavior of alkali-activated
metakaolin during geopolymerization, J. Non-Cryst. Solids 357 (2011) 3565–
3571.
[13] I. Janotka, F. Puertas, M. Palacios, M. Kuliffayová, C. Varga, Metakaolin sand-
blended-cement pastes: rheology, hydration process and mechanical
properties, Constr. Build. Mater. 24 (2010) 791–802.
[14] M. Sonebi, M. Lachemi, K.M.A. Hossain, Optimisation of rheological parameters
and mechanical properties of superplasticised cement grouts containing
metakaolin and viscosity modifying admixture, Constr. Build. Mater. 38
(2013) 126–138.
[15] K. Vance, A. Kumar, G. Sant, N. Neithalath, The rheological properties of ternary
binders containing Portland cement, limestone, and metakaolin or fly ash,
Cem. Concr. Res. 52 (2013) 196–207.
103
[16] A. Favier, J. Hot, G. Habert, N. Roussel, JBd’E de Lacaillerie, Flow properties of
MK-based geopolymer pastes. A comparative study with standard Portland
cement pastes, Soft Matter 10 (2014) 1134–1141.
[17] K. Vance, A. Arora, G. Sant, N. Neithalath, Rheological evaluations of
interground and blended cement-limestone suspensions, Constr. Build.
Mater. 79 (2015) 65–72.
[18] A. Shahriar, M.L. Nehdi, Optimization of rheological properties of oil well
cement slurries using experimental design, Mater. Struct. 45 (2012) 1403–
1423.
[19] G.H. Tattersall, P.F.G. Banfill, The Rheology of Fresh Concrete, Pitman Books
Ltd, London, 1983.
[20] P.F.G. Banfill, Rheology of Fresh Cement and Concrete, Taylor & Francis, New
York, 1991.
[21] M. Frias, M.I.S. de Rojas, J. Cabrera, The effect that the pozzolanic reaction of
metakaolin has on the heat evolution in metakaolin-cement mortars, Cem.
Concr. Res. 30 (2000) 209–216.
[22] L.M. Vizcaíno-Andrés, S. Sánchez-Berriel, S. Damas-Carrera, A. Pérez-
Hernández, K.L. Scrivener, J.F. Martirena-Hernández, Industrial trial to
produce a low clinker, low carbon cement, Mater. Constr. 65 (317) (2015) 1–
11.
[23] V.L. Bonavetti, V.F. Rahhal, E.F. Irassar, Studies on the carboaluminate
formation in limestone filler-blended cements, Cem. Concr. Res. 31 (2001)
853–859.
[24] R. Bucher, P. Diederich, M. Mouret, G. Escadeillas, M. Cyr, Self-compacting
concrete using flash-metakaolin: design method, Mater. Struct. 48 (2015)
1717–1737.
[25] I.P. Sfikas, E.G. Badogiannis, K.G. Trezos, Rheology and mechanical
characteristics of self-compacting concrete mixtures containing metakaolin,
Constr. Build. Mater. 64 (2014) 121–129.
[26] C. Perlot, P. Rougeau, S. Dehaudt, Slurry of metakaolin combined with
limestone addition for self-compacted concrete. Application for precast
industry, Cem. Concr. Compos. 44 (2013) 50–57.
[27] A. Mohebbi, M. Shekarchi, M. Mahoutian, S. Mohebbi, Modeling the effects of
additives on rheological properties of fresh self-consolidating cement paste
using artificial neural network, Comput. Concr. 8 (2011) 279–292.
[28] J. Cornell, Experiments with Mixtures, third ed., John Wiley & Sons, New York,
2002.
[29] R.E. Bruns, I.S. Scarminio, B. de Barros Neto, Data Handling in Science and
Technology, Statistical Design – Chemometrics, vol. 25, Elsevier, Amsterdam,
2006. ISBN: 978-0-444-52181-1.
[30] M. Andersson, H. Ervanne, M.A. Glaus, S. Holgersson, P. Karttunen, H. Laine, B.
Lothenbach, I. Puigdomenech, B. Schwyn, M. Snellman, H. Ueda, M. Vuorio, E.
Wieland, T. Yamamoto, Working Report 2008–28: ‘‘Development of
Methodology for Evaluation of Long-term Safety Aspects of Organic Cement
Paste Components”, POSIVA OY Working Report 2008–28, Eurajoki, Finland,
2008, pp. 279.
[31] C.F. Ferraris, V.A. Hackley, A.I. Avilés, Measurement of Particle size distribution
in portland cement powder: analysis of ASTM round robin studies, Cem. Concr.
Aggregates 26 (2) (2004). Paper ID CCA11920.
[32] API Recommended Practice 10B, 22nd ed., Recommended Practice for Testing
Well Cements, Appendix A, 1997, page 123.
[33] J. Ambroise, S. Maximilien, J. Pera, Properties of metakaolin blended cements,
Adv. Cem. Based Mater. 1 (1994) 161–168.
[34] M.A. Helal, Effect of curing time on the physico-mechanical characteristics of
the hardened cement pastes containing limestone, Cem. Concr. Res. 32 (2002)
447–450.
[35] ASTM D6910/D6910M-09: Standard Test Method for Marsh Funnel Viscosity
of Clay Construction Slurries. ASTM International, West Conshohocken, PA
19428–2959. United States.
[36] N. Roussel, R. le Roy, The Marsh cone: a test or a rheological apparatus?, Cem
Concr. Res. 35 (2005) 823–830.
[37] ASTM C1749-12: Standard guide for measurement of the rheological
properties of hydraulic cementitious paste using a rotational rheometer.
ASTM International, West Conshohocken, PA 19428–2959. United States.
[38] T.G. Mezger, The Rheology Handbook, 3rd revised ed., Vincentz Network,
Hannover, 2011, pp. 72–74.
[39] N. Roussel, P. Coussot, Fifty-cent rheometer for yield stress measurements:
from slump to spreading flow, J. Rheol. 49 (2005) 705–718.
[40] NBR 7215 – ‘‘Portland cement - Determination of compressive strength” (in
portuguese), Associação Brasileira de Normas Técnicas, Rio de Janeiro, 1996.
[41] ASTM C39/C39M  16b: Standard Test Method for Compressive Strength of
Cylindrical Concrete Specimens. ASTM International, West Conshohocken, PA
19428–2959. United States.