Cement & Concrete Composites 31 (2009) 255–262

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Relation between cement composition and compressive strength of pure pastes

Jean-Michel Mechling *, André Lecomte, Cécile Diliberto
Institut Jean Lamour, Dpt CP2S, Equipe Matériaux pour le Génie Civil, Nancy Université, UMR 7198, IUT de Nancy-Brabois, CS 90137, F54601 Villers-lès-Nancy Cedex, France

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

Article history: In Europe, cement paste compressive strength models are frequently described by a power law similar to
Received 11 July 2007 that of Féret. The French LCPC has also adopted this mathematical approach in its recent concrete mixture
Received in revised form 4 February 2009 proportioning model. However the coefficient k and exponent b of the adopted power law were calibrated
Accepted 6 February 2009
with only one Portland cement clinker, starting from pure pastes with various concentrations. These
Available online 15 February 2009
parameters do not take into account the chemical properties of the clinker. The purpose of this work
was to check if the clinker chemical nature had an influence on the mechanical performances of the
Keywords:
pastes (at 28 days). Complementary tests with eight cements coming from six different cement plants
Hydration
Compressive strength
were then carried out. The mathematical treatment of the results made it possible to connect coefficient
Cement paste k to the silica content of the clinkers and more particularly to their C3S rates. It is linked to the strength of
Clinker hydrated cement paste. Thus, the suggested approach significantly improves the accuracy of paste
Cement manufacture strength calculation. These results can be used to calculate concrete strength.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction
Among the most used compressive strength calculation
models – Abram’s law, European methods like Féret’s law and
Bolomey formula – we have used the new concrete mixture
proportioning model designed by the French Laboratoire des Ponts
et Chaussées (LCPC) [1].
In this model, concrete strength is linked to that of cement
paste taking into account parameters describing the effect of
aggregates topology (paste–aggregate model) [2].
The 28-day compressive strength of pure cement pastes (fcp) is
calculated with a model similar to that of Féret [3]. This is a power
law of exponent b whose C0 variable is the cement concentration of
fresh paste: the volume of cement in the volume of paste (Eq. (1)).
Its main parameter is equal to the product of 28-day standard
strength of cement r028 and coefficient k (Eq. (2)):
vc
C0 ¼
vc þ vw þ va
vc, vw and va are the volumes of cement, water and air in fresh paste,
respectively.
b to C3S, C2S, aluminates and ferrite contents has been developed
fcp ¼ r028 kC 0
To be able to predict paste strength starting from its composi-
tion in the concrete (quantities of cement, water and air), coeffi-
cient k and exponent b of the relation (Eq. (2)) were initially
* Corresponding author.
E-mail address: Jean-Michel.Mechling@iutnb.uhp-nancy.fr (J.-M. Mechling).
gauged by the LCPC [1] starting from tests on one pure cement
paste. These tests consisted in measuring compressive strength
at 28 days of a series of pastes made with only one cement CEM
I 52.5 (95% of clinker, EN 197-1 norm) and various quantities of
water (within the range 0.33 < C0 < 0.58). The fitting of the results
provided values of 11.4 and 2.85 for k and b, with an average error
of only 2.3 MPa [2].
In this model, parameters k and b are considered as valid for all
cements whatever their chemical compositions may be. It means
that the 28-day standard strength r028 is sufficient to differentiate
mechanical behaviour in cement pastes. It would be interesting
to consider other cement parameters such as their chemical com-
positions. Tsivilis and Parissakis showed that cement fineness
mainly affects strength at early age (before 7 days) while chemical
and mineralogical parameters influence strength at a later stage
[4].
Previous studies have shown that the chemical properties of ce-
ð1Þ ment can influence many properties of the mixture such as its
compressive strength. Brüggemann and Bentrup [5] have shown
correlations between mineralogical clinkers parameters and ce-
ment strength. A relation linking compressive strength at 28 days
ð2Þ
by Knöfel [6]. Tsivilis and Parissakis [4] have presented three laws
to determine compressive strength at three ages (2, 7 and 28 days)
in relation with the clinker compounds and/or the fineness param-
eter. Jons and Osbaeck [7] have also given relations between ce-
ment composition and cement paste strength suggesting that
clinker phases contribute to increasing strength as well as reducing
porosity. Then the influence of micro structural porosity on cement
paste has been shown in several models such as a law taking into

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.009
256 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262

account the porosity of cement paste either fresh or hardened to 160

calculate compressive strength [8]. We note that the chemical dif-
ferences between several clinkers may change the chemical equi-
librium and thus the composition and structure of hydrates, 120
which is contradictory to the single relation defined by the LCPC.
In this context, the purpose of this paper was to carry out com-
plementary mechanical tests with several cements to check if the 80
relation (Eq. (2)) can be improved by taking into account the clin-
ker chemical composition. The results will then be used to find cor-

fcp (MPa)
40
relations between the parameters of the model and the cement
chemical characteristics. Primarily, this study presents the mate-
rial ingredients used to carry out the tests. Secondly, it presents Co
0
the tests of reference [2], the tests carried out within the frame- 0.0 0.2 0.4 0.6
work of this study as well as their material and instrumental envi-
ronment. Thirdly, the obtained results are discussed. They made it Fig. 1. Experimental values and fitting for the establishment of the initial law [2].

possible to propose a more comprehensive approach of the com-

pressive strength behaviour of pure cement pastes.
2. Material ingredients
The study was carried out with eight Portland cement clinkers
(or type CEM I cements, according to the European standard EN
197-1). These cements were provided by six cement plants located
in France, Luxembourg and Germany. Table 1 gives a letter to each
cement from A to H as well as their commercial designations and
their origins (numbers 1–6). Let us note that cement D comes from
the same cement plant and thus should have the same character-
istics as that used in the initial tests carried out by the LCPC [2].
The samples used in the tests came from recently manufactured
cements. These samples were preserved in hermetic bags or pots
until use (a few weeks after production). Their physical and chem-
ical properties (specific density qc, oxides contents) were given by
the cement plant’s quality control department. The specific densi-
ties vary from 3.13 to 3.18 tons per cubic metre. The compositions
were also provided by the cement plants. The real contents of C3S,
C2S, C3A, and C4AF in cement were obtained by X-ray diffraction
investigations (XRD) according to the Rietveld method [9]. They
generally differ from the theoretical values calculated with the Bo-
gue equations for the clinker because of the presence of sulphate
and possibly of calcium carbonate in cement and of the actual pur-
ity level of the clinker phases. The compressive strength (28-day
standard strength r028 ) of each cement was measured several times
on standardized mortars according to European standards (EN
196-1). All values are given in Table 2.
It is noted that major elements (SiO2, Al2O3, Fe2O3 and CaO) are
obviously constant from one cement to another. However, their
compositions present significant variations. For example, the
contents of C3S and C3A, respectively vary between 51% and 68%
and between 2% and 14%.
One concrete superplasticizer – polycarboxylate water
reducer – was used to cast the mixtures of low w/c ratio. It has a
density of 1.05 and a solid content of 20%. The initial water content
of this product was taken into account in calculations.
The different mixtures were made with distilled water to avoid
possible differences caused by the water’s chemical composition.
3. Experiments and results

Table 1
3.1. The initial tests carried out by LCPC [2]
Reference and designation of cements.

Reference Designation Cement plant no. These initial tests were carried out by the French LCPC with ce-
A CEM I 52.5 R 1 ment whose origin and category are those of cement D. However,
B CEM I 52.5 R 2 this cement had been manufactured twelve years before this study
C CEM I 52.5 R 3
and its chemical composition could not be recovered. It is undoubt-
D CEM I 52.5 N PM 4
E CEM I 52.5 R 5 edly different from that of current cement D. Six pastes with differ-
F CEM I 52.5 N 5 ent volumetric cement concentrations C0 were mixed. They varied
G CEM I 42.5 R 6 between 0.33 and 0.58. The pastes of low volumetric cement con-
H CEM I 42.5 R 2 centration (high w/c ratio) were mixed in two stages in order to
CEM I: category; 52.5: cement strength (MPa); R: quick setting; N: normal setting; avoid a bleeding phenomenon [10]. A diluted paste – of w/c ratio
PM: sea setting. equal to 5 – was first mixed for 24 h. The obtained suspension con-

Table 2
Composition, hydraulic index (I), surface area (S.a.), 28-day standard strength of cement ðr028 Þ of the used cements.

Ref Oxide rate (%) Composition (%) I S.a. r028

2
SiO2 Al2O3 Fe2O3 CaO MgO K2O Na2O SO3 C3S C2S C3A C4AF cm /g MPa
A 20.5 5.2 2.3 64.7 1.0 1.2 0.1 3.7 66.9 8.0 14.1 4.4 0.43 4510 62.7
B 21.0 6.0 2.5 63.5 –b –b 0.44a –b 67.5 7.0 6.0 10.0 0.46 4400 61.7
C 19.9 4.5 3.0 61.9 4.6 1.3 0.3 3.8 58.0 –b 8.0 10.0 0.41 5000 64.0
D 21.3 3.5 4.2 65.0 0.9 0.3 0.1 2.7 63.2 13.3 2.0 12.9 0.44 3550 60.4
E 19.8 5.1 3.0 63.5 1.7 0.9 0.2 3.3 54.5 19.2 8.4 9.8 0.43 –b 66.0
F 19.8 5.1 3.0 63.5 1.7 0.9 0.2 3.3 55.5 18.2 8.4 9.8 0.43 –b 64.2
G 20.0 5.5 3.1 63.0 3.0 1.4 – 3.6 51.0 22.0 9.0 10.0 0.43 4500 60.7
H 21.0 6.0 2.5 63.5 –b –b 0.44a –b 67.5 7.0 6.0 10.0 0.46 3350 55.1
a
Alkaline components in the term of equivalent Na2O.
b
Values not given.
 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262 257

taining hydrated cement particles was then used like mixing water
to make prismatic specimens (dimensions: 3.8  3.8  7.6 cm).
The pastes of strong packing densities (low w/c ratio) were mixed
with a moderate quantity of melamine type superplasticizer to
facilitate casting. The paste specimens were extracted from their
moulds after 2 days then rectified and ground smooth on a lapidary
before being covered with an impervious membrane in order to
avoid desiccation. Compressive strength values obtained at 28 days
are reproduced on Fig. 1. They result from the average of the four
tests. The results made it possible to obtain the initial parameters
of the model (Eq. (3)) – i.e. k = 11.4 and b = 2.85 – with a small
average error (2.3 MPa)
fcp ¼ r028  11:4  C 02:85
3.2. Complementary tests: procedures and results
As for the initial study, the tests were carried out on pure ce-
ment pastes of various cement concentrations C0. These tests are
particularly difficult to carry out taking into account inherent dif-
ficulties of cement pastes: bleeding and segregation of fresh sus-
pension; rise in temperature during the setting and hardening;
drying shrinkage; regularity of the bearing surfaces for the
mechanical tests; etc. To realize this study, original processes had
to be tested and used such as the rotation of the paste during set-
ting, its preservation in a confined environment and a great accu-
racy of rectification of the bearing surfaces before strength tests.
For each cement, eight series of six mixtures (12 for cement D)
ð3Þ were prepared with cement concentrations C0 ranging from 0.31 to

Table 3
Characteristics and compressive strength of the mixtures. Superplasticized mixes are identified with *. Values considered as erratic ones are noted in italic.

Ref qe (t/m3) va (l/m3) C0 fcp (MPa) s (MPa)

A M1 1.674 4 0.318 30.5 0.8
M2 1.747 1 0.351 36.0 2.3
M3 1.822 5 0.388 46.7 4.3
M4 1.932 0 0.438 47.0 3.3
M5* 2.054 4 0.496 85.2 4.7
M6* 2.103 5 0.520 95.1 3.9
B M1 1.679 6 0.314 15.2 2.2
M2 1.747 6 0.345 27.7 0.9
M3 1.827 8 0.383 35.2 1.3
M4 1.941 3 0.433 56.5 1.5
M5* 2.078 0 0.494 103.5 1.6
M6* 2.184 32 0.557 122.0 0.2
C M1 1.676 6 0.316 19.1 1.9
M2 1.746 5 0.347 30.8 0.9
M3 1.831 3 0.386 33.0 3.4
M4 1.936 2 0.435 53.4 4.1
M5* 2.052 10 0.491 85.5 5.7
M6* 2.123 41 0.539 73.1 7.9
D M1 1.678 7 0.314 12.7 1.6
M2 1.745 7 0.345 29.8 0.3
M7 1.790 7 0.366 40.1 1.0
M3 1.825 9 0.383 42.4 2.1
M8 1.880 8 0.408 49.6 5.0
M4 1.936 5 0.432 51.5 4.6
M9 2.016 0 0.466 65.0 7.5
M5* 2.079 0 0.495 69.8 16.0
M10* 2.171 0 0.537 117.1 7.0
M6* 2.240 7 0.572 123.6 6.6
M11* 2.304 32 0.579 126.9 5.0
M12* 2.355 49 0.590 120.5 3.8
E M1 1.675 7 0.315 17.8 1.5
M2 1.733 12 0.345 22.4 1.6
M3 1.829 4 0.386 35.5 3.4
M4 1.935 3 0.434 43.6 3.9
M5* 2.069 2 0.495 71.5 3.0
M6* 2.168 21 0.550 91.6 0.9
F M1 1.686 0 0.318 20.7 0.5
M2 1.741 7 0.346 29.2 0.6
M3 1.837 0 0.388 42.5 0.8
M4 1.941 0 0.436 46.1 2.7
M5* 2.072 0 0.496 82.8 2.6
M6* 2.176 16 0.552 93.6 2.8
G M1 1.681 4 0.316 18.6 0.5
M2 1.750 3 0.347 24.4 1.0
M3 1.837 1 0.386 37.4 4.7
M4 1.941 1 0.434 37.2 2.1
M5* 2.051 12 0.489 54.4 4.9
M6* 2.124 29 0.531 81.3 3.9
H M1 1.681 4 0.316 14.4 1.2
M2 1.745 6 0.346 26.5 0.6
M3 1.825 8 0.384 36.1 2.2
M4 1.941 1 0.434 53.5 5.2
M5* 2.076 0 0.495 92.9 8.6
M6* 2.205 21 0.564 129.2 10.9
258 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262
0.59 (Table 3). It corresponds to w/c ratios ranging from 0.70 to
0.22 (Eq. (7)). To avoid a possible mechanical effect of the hydrated
cement particles caused by the use of highly diluted paste as mix-
ing water (initial tests carried out by the LCPC, 3.1), pure pastes
were mixed with distilled water. Paste mixing was carried out with
a standardized mortar mixer according to EN standard 196-3. Each
1-litre mixture was used to fill two impervious cylindrical moulds
especially built for these tests. They are very regular altuglassTM test
tubes – 18 cm long with 5 cm large interior diameters – provided
at each end with a threading intended to fix PVC seal plugs
(Fig. 2). One of the two plugs was equipped with a sealable vent.
After casting, each mixture was left to set and harden and was pre-
served 28 days in these tubes until testing.
During casting the moulds were vibrated with a mobile plate to
eliminate as much entrapped air as possible. Filled moulds were
sealed with round PVC plugs with vents leaving almost no air in-
side. Sealed test tubes were weighed in order to know experimen-
tal fresh densities qe. They are given in Table 3 (in tons/m3). By
comparison with the theoretical fresh densities qt deduced from
the compositions, it is possible to estimate the entrapped volume
of air in each mixture expressed in unit volume according to the
following relation:
v a ¼ 1000  ð1  qe =qt: Þ
These volumes of air are also given in Table 3 in litres per cubic
metre. They made it possible to calculate the real cement concen-
tration C0 of each mixture (Eq. (5)).
v cinit ð1  v a Þ
C0 ¼ ; important over that limit. The entrapped air volume goes from a
v cinit ð1  v a Þ þ v winit ð1  v a Þ þ v a
v cinit
C0 ¼
v cinit þ v winit þ v a =ð1  v a Þ
Therefore, the initial volumetric proportions of cement (vcinit)
and water (vwinit) in unit volume were adjusted by taking into ac-
count the volume of mixture actually produced
v c ¼ v cinit  ð1  v a Þ and v w ¼ v winit  ð1  v a Þ
The mixtures with high cement concentrations (or low w/c ra-
tio) were cast in the presence of superplasticizer proportioned be-
tween 1% and 1.5% of the cement mass. This proportioning was
voluntarily reduced to mitigate a potential disturbing effect such
as a modification of the usual chemical reactions.
The mixtures of low cement concentrations (or high w/c ratio)
naturally undergo segregation if they are maintained at rest. To
avoid this phenomenon, the altuglassTM test tubes were placed in
Fig. 2. Device intended to prevent the fresh cement paste segregation by slow
rotation.
a particular setting device in slow and offset rotation (seven revo-
lutions per minute) for the first 24 h. The longitudinal axes of the
test tubes turned in a vertical plane around a horizontal axis lo-
cated at ten centimetres from the centre of the mould (Fig. 2).
The test tubes were placed on the machine as soon as filling was
finished. In respect to homogeneity all the mixtures underwent
this treatment even the firmest. From the 24th hour to the release
from the test tubes at 28 days, the sealed moulds were kept at a
temperature of 20 ± 1 °C. A control of their masses over that period
guaranteed that the pastes were not air dried.
The day of the tests, after removal from the tubes, the ends of
the paste specimens were cut off from approximately 1 cm and
then the specimens were sawn in half. The dimensions of the ob-
tained cylindrical half-specimens were 5 cm wide and about
8 cm long. The bearing surfaces were rectified with a precision ma-
chine and carefully ground smooth. Flatness and parallelism of the
supports were checked on a plate surface
For the compressive test, each half-specimen was subjected to a
0.5 MPa/s compressive rate until rupture (EN standard 12390). The
results are given in Table 3. Each value is the average of the four
measurements obtained on each mixture. The standard deviations
of the results s are also given in Table 3.
ð4Þ
3.3. Discussion
Fig. 3 presents the evolution of the estimated entrapped air vol-
ume according to cement concentration C0. This volume remains
low until a concentration of 0.50 (fluid mixtures). It becomes more
few litres per m3 (less than 10 l or 1%) up to 50 l per m3. These tests
ð5Þ show the difficulties in eliminating air in paste with high packing
density and high plastic viscosity.
Fig. 4 presents, for each cement, the experimental strength
compressive values at 28 days according to C0. The dotted curves
describe the theoretical evolution of strength according to the ini-
tial model (Eq. (2)) using the 28-day standard strength of each ce-
ð6Þ
ment. It is noted that experimental fitting does not always satisfy
the initial model. It is close for cements like B and H but for the
others the measured performances deviate from the forecasts.
For example, for cement E, the variation is about 10 MPa for low
cement concentration and it exceeds 45 MPa for high cement con-
centration. In relative value, the theoretical strength of this cement
paste is overestimated by almost 35%. For cement D, the perfor-
mances are a little lower than the ones obtained by the LCPC
although the cement used belongs to the same category (but not
to the same production). Cements resulting from the same clinker
(cement plants 2 and 5) logically present rather similar curves in
spite of setting kinetics (R – quick– for E, N – normal – for F) and
cement strength differences (52.5 for B, 42.5 for H).
60 A B C D
E F G H
Air volume (l/m3 )
40
20
C0 (%)
0
0.3 0.4 0.5 0.6
Fig. 3. Entrapped air volume in the different cement pastes.
 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262 259

150
150
A B

100 100

fcp Mpa, 28-day

fcp Mpa, 28-day

50 50

C0 (%) C0 (%)
0 0
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6

150
C 150
D

100 100

fcp MPa, 28-day

fcp MPa, 28-day

50 50

C0 (%) C0 (%)
0 0
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6

150
E 150
F

100 100
fcp MPa, 28-day
fcp MPa, 28-day

50 50

C0 (%) C0 (%)
0 0
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6

150
G 150
H

100 100
fcp Mpa, 28-day

fcp MPa, 28-day

50 50

C0 (%) C0 (%)
0 0
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6

Fig. 4. Obtained experimental points and model’s curves for the different cements. The initial model (Eq. (2)) is in dotted line, the new suggested model (Eq. (10)) is in fat For
cement D, initial fitting points [2] are represented with white rounds.

Table 4
Fitting of the results gathered by cement strength.

Optimisation on k and b k with b = 2.95 b with k = 10.5

k b s (MPa) k s (MPa) b s (MPa)
CEM I 42.5 13.28 3.20 10.70 11.21 10.94 2.87 11.14
CEM I 52.5 9.47 2.86 8.79 10.07 8.84 3.00 8.91
260 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262

Size effect [11,12] is a first hypothesis to explain the differences
between our experimental test campaign and the initial one on ce-
ment D, which resulted in the theoretical model (see Section 3.1).
Despite the difficult task of determining the actual impact of this
phenomenon, we noticed that the bearing areas and volumes of
the specimens are very similar in both experimental approaches.
Only their forms differ. They were prismatic (3.8  3.8  7.6 cm)
in the initial tests and cylindrical (5 cm wide and 8 cm long) in
ours. Furthermore, in our experimental test campaign differences
in compressive strength have been noticed between the different
cements despite constant experimental conditions. Therefore, size
and shape effects cannot explain the strength differences between
these different cements.
Finally, for all cements, compressive strength measured on pure
paste with cement concentration close to 0.40 are always lower
than standard strength r028 measured on standardized mortar
(comparison in Tables 2 and 3). The concentration of 0.40 is the
one roughly corresponding to a w/c mass ratio of 0.50 (Eq. (7)).
That is to say the cement concentration of the standardized mortar
used to measure the 28-day standard strength cement. By neglect-
ing entrapped air – small quantities are often observed – (Fig. 3)
the relations between w/c and C0 are as follows:
1 1  C0
C0 ¼ and w=c ¼
1 þ qc w=c qc  C 0
Strength differences between pastes and mortars are explained
by the aggregate topological effect on strength (maximum paste
thickness, ceiling effect and bond effect of the grains, etc. [2]).
However, these differences vary from one cement to another. This
is an additional argument to justify that cement 28-day standard
strength r028 is probably not a sufficient criterion to build a general
model to forecast cement paste strength.
4. Influence of chemical composition
4.1. Adjustment of the model parameters
The model parameters (Eq. (2)) (coefficient k, exponent b) were
optimized using the minimum standard deviation s on the ordi-
nates (according to the approach of least squares). Primarily, these
calculations were conducted on all disconcerted points including
those of initial measurements [2] (except some values considered
as erratic and announced in italics in Table 3). The aim was to find
some satisfying constant averages. The found values are k = 10.5
and b = 2.95. These values are rather close to those of the initial
parameters given in [2] but the associated standard deviation
(s = 11.7 MPa) is too high to consider this solution as really satis-
factory. Secondly, fitting curves were calculated by classifying ce-
ments according to two cement strengths (42.5 and 52.5).
Thirdly, fitting curves were independently drawn for each cement.
Table 4 and Table 5 give the results. Three simulations were con-
ducted for each case. For the first one, both parameters k and b
could simultaneously vary. For the two others, exponent b previ-
ously found (b = 2.95) was kept constant whereas coefficient k
was optimized, then the converse was done (k = 10.5). The stan-
dard deviations s between experimental and theoretical compres-
sive strengths were calculated for each adjustment. They were
selected to choose the most relevant fitting curves.
On the one hand, sorting out cement strengths (either 42.5 or
52.5 in Table 4) does not make this model more accurate than
the initial one (higher standard deviations and divergent k and b
parameters). On the other hand, when fitting curves are calculated
for each cement (Table 5), standard deviations are reduced of more
ð7Þ than half whatever the simulation type used. Moreover, cements of
similar chemical nature have relatively similar parameters (B and
H of cement plant 2; E and F of cement plant 5). The improvement
of paste strength calculation model thus probably works by taking
into account the physicochemical characteristics of each cement
used.
Among the three calculation modes used in Table 5, the one
where k and b vary simultaneously offers the best accuracy (aver-
age standard deviation of 3.80 MPa). But this solution comes down
to having to adjust two parameters for each cement. To simplify,
exponent b has been kept constant (b = 2.95) while coefficient k
has been linked to the chemical properties of cements. Further-
more, according to the calculations presented in Table 5, exponent
b gives a better statistical stability (s0 /m = 0.07) and coefficient k a
better accuracy (average standard deviation of 4.85 MPa, that is to
say a value close to that found when k and b are simultaneously
optimized).

Table 5
Fitting of the results for each cement used

Optimisation on k and b k with b = 2.95 b with k = 10.5

(1) (2) (3) (4) (5) (6) (7)
k b s (MPa) k s (MPa) b s (MPa)
A 7.13 2.40 0.74 10.62 4.84 2.92 4.64
B 12.52 3.07 7.64 11.56 7.74 2.82 8.09
C 14.87 3.41 3.06 10.23 4.15 2.99 4.00
D 10.52 2.90 6.73 10.87 6.75 2.90 6.73
E 7.99 2.91 2.17 8.24 2.19 3.29 3.37
F 6.75 2.52 4.97 9.11 5.96 3.14 6.86
G 7.18 2.65 1.53 8.92 2.58 3.17 3.86
H 15.23 3.24 3.52 12.56 4.60 2.71 6.66
Average m 10.27 2.89 3.80 10.26 4.85 2.99 5.53
Standard deviation s’ 3.5 0.4 – 1.4 – 0.2 –
s’/m 0.34 0.12 – 0.14 – 0.07 –

Table 6
Correlation coefficients R enters the coefficient k (b = 2.95) and chemical compounds of cements and specific area (s.a.).

Compound SiO2 Al2O3 Fe2O3 CaO I SiO2/CaO C3S C2S C3A C4AF s.a.
R 0.845 0.221 0.237 0.198 0.713 0.901 0.896 0.904 0.326 0.016 0.632
 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262 261

4.2. Relation with the cement composition
Primarily, several linear adjustments of coefficient k (column 4 of
Table 5) and of the various parameters of cement chemical
composition (Table 2) were tested. Cement chemical composition
has been defined using oxide masses (SiO2, Al2O3, Fe2O3, CaO)
as well as Rietveld composition and the hydraulic index I
(I = (SiO2 + Al2O3 + Fe2O3)/(CaO+MgO)or in SiO2/CaO ratio). The
specific areas (S.a.) of the cements have also been adjusted in
the tests. The obtained correlation coefficients R are given in Ta-
ble 6. Let us recall that they correspond to the co variances of k
and x divided by the product of the standard deviations of k and
x (x being the series of values correlated to the values of k):
cov ðk; xÞ kxkx
R¼ or R ¼
sk  sx s k  sx
Fig. 5 shows fitting curves for C2S, which present the best corre-
lation (R = 0.904). This coefficient R is almost as good for ratio
SiO2/CaO (R = 0.901) and for C3S (R = 0.896). It is still good for
SiO2 (R = 0.845). On the contrary, it is not good for the hydraulic in-
dex I (R = 0.713) or the specific area (R = 0.632) and even worse
for all the other chemical elements. From a physical point of view,
the link between the strength of pure paste and its content in sili-
ceous elements (C3S, C2S, SiO2) can be explained by the fact that
hydrates (HSC – tobermorite) come from these phases, which are
mainly responsible for the material’s compressive strength [13]
[14].
According to these results, three (linear) relations can be de-
fined to determine parameter k with good accuracy, starting from
Rietveld analysis or cement chemical analysis:
k ¼ 0:22  ½C2 S þ 13:28
k ¼ 160  ½SiO2 =CaO  41:2
k ¼ 0:2  ½C3 S  1:65
It will be noticed however that parameter ‘‘C3S” represents the
most abundant phase in the clinker. Its variation range is also the
largest (between 50% and 70% for the cements used in this study).
Therefore, this parameter has been chosen in the model to describe
the strength of a cement paste (at 28 days) as follows (Eq. (12)), by
the combination of Eqs, (2) and (11):
13
12
k= -0.223 [C 2 S] + 13.280
11
10
Coefficient k
 2:95
vc
fcp ¼ r028 ð0:2:½C 3 S  1:65Þ ð12Þ
vc þ vw þ va
4.3. Comparison between the initial model and the new one
The new model (Eq. (12)) was compared with the initial one
(Eq. (3)), [2] by plotting the new theoretical curves on Fig. 4 (full
features) and by calculating the standard deviations with the
experimental points (Table 7).
Most of the time, the new model predicts a much more relevant
experimental compressive strength. Compared to the initial model,
the error is overall reduced of more than half (s goes from 15.75 to
6.87) and it is often similar to that of the individual adjustments.
This new model allows a better estimation of compressive strength
ð8Þ
by taking into account the chemical composition of cement.
This is an improvement of the compressive strength prediction
model of cement pastes and thus of that of concretes defined by
the LCPC in [2]. Several tests are currently carried out to evaluate
the relevance of this new approach regarding high performance
concrete strength calculations.
5. Conclusion
Compressive strength of a pure cement paste at 28 days is mainly
controlled by the 28-day standard strength of this cement. But this
study shows the importance of its chemical composition, in partic-
ular the amount of C3S, which reflects the quantity of silicon present
in the clinker. Let us recall that the content of C3S taken into account
in this paper was determined by treatment of analyses XRD accord-
ing to Rietveld method. Indeed, the many tests carried out and the
ð9Þ equations that have been defined (Eqs. (9) – (12)) showed that for
ð10Þ equivalent 28-day standard strength, cements rich in C3S produced
ð11Þ the strongest pastes. It confirms the fact that hydrates resulting
from this cement constituent are mainly responsible for the
strength of the product. In this hypothesis, coefficient k and some-
times b depend on the chemical composition of the clinker (high
values of the correlation coefficients). Thus, cement probably has
the same value for k (at a defined time: 28 days for example) what-
ever its standard strength, which depends on its granulometry and
its specific area. In other words, two cements from the same clinker
production have the same value of k despite their different stan-
dardized cement strength (depending on grain fineness).
This taking into account of the cement chemistry in a model
similar to that of Féret (power-law) was done by correlating the
content of C3S with the parameter of the law, its constant exponent
b being kept. Other correlations are also possible using C2S or sili-
con content.
The use of this comprehensive model is not particularly harder
because the necessary data are generally available on the cement

chart. It could have a practical application to calculate the strength

9 of high performance concrete (paste with high cement concentra-
tion). Then a significant variation of k leads to a significant change
8 of paste compressive strength and consequently of concrete
C2S content (%) strength.
7 In a future paper, these results will be checked on concretes of
5 10 15 20
identical composition composed of various cements. The evolution
Fig. 5. Relation between parameters k (b = 2.95) and the cement C2S content. of these parameters over time may also be studied.

Table 7
Standard deviations s between experimental points and strength theoretical values calculated according to the initial model and the model suggested.

Model A B C D E F G H Global
Initial 10.53 9.10 12.57 14.95 26.29 20.70 17.84 5.78 15.75
Suggested 7.37 7.76 4.70 8.14 5.98 6.06 3.76 7.30 6.87
262 J.-M. Mechling et al. / Cement & Concrete Composites 31 (2009) 255–262
Acknowledgements
The authors thank the cement-manufacturers who generously
provided them with the samples and chemical analyses used in
this paper. They also thank Mr Yves Angermuller and Mr Bertrand
Ehrhardt (Angermuller company) for their help and their interest
in this study as well as Mr Jean-Paul Glez and Mr Olivier Français
(Nancy Université) for the realization of specific materials neces-
sary for the tests.
References
[1] Sedran T, de Larrard F. BétonlabPro2 – Concrete mix proportioning software,
version 2.0, for Windows [in French], software and information. Paris: Presses
de l’Ecole Nationale des Ponts et Chaussées, 2000.
[2] de Larrard. Spon FE, Spon E, editors. Concrete mixture proportioning, a
scientific approach. London; 1999.
[3] Féret R. About the hydraulic mortars compacity (in French). Annales des Ponts
et Chaussées 1892;7(4):5–164.
[4] Tsivilis S, Parissakis G. A mathematical model for the prediction of cement
strength. Cem Concr Res 1995;25(1):9–14.
[5] Brüggemann H, Bentrup L. Correlations between mineralogical clinker
parameters and cement strength. In: Proceedings of the 11th international
conference on cement microscopy, New Orleans; 1989. p. 226–45.
[6] Knöfel D. Interrelation between proportion of clinker phases and
compressive strength of Portland cements. In: Proceedings of the 11th
international conference on cement microscopy, New Orleans; 1989. p. 246–
62.
[7] Jons ES, Osbaeck B. The effect of cement composition on strength described by
a strength–porosity model. Cem Concr Res 1982;12(2):167–78.
[8] Matusinović T, Šipušić J, Vrbos N. Porosity–strength relation in calcium
aluminate cement pastes. Cem Concr Res 2003;33(11):1801–6.
[9] Rietveld HM. A profile refinement method for nuclear and magnetics
structures. J Appl Crystallogr 1969;2:65–71.
[10] Markestad SA. The Gukild–Carlsen method for making stabilized pastes of
cements. Mater Struct 1976;9(50):115–7.
[11] Tokyay M, Özdemir S. Specimen shape and size effect on the compressive
strength of higher strength concrete. Cem Concr Res 1997;27(8):1281–90.
[12] Semsi Y, Gözde Inan S. The effect of cylindrical specimen size on the
compressive strength of concrete. Build Environ 2007;42:1417–2420.
[13] Baroghel-Bouny V. Cement pastes and concretes characterization, analyses,
interpretations. Paris: Presses de l’Ecole Nationale des Ponts et Chaussées;
1994 [in French].
[14] Bogue RH. Chemistry of Portland cement. New York: Reinhold; 1955.
 Cement & Concrete Composites 31 (2009) 263–267

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Measurement of linear thermal expansion coefficient of alkali-activated

aluminosilicate composites up to 1000 °C
Lucie Zuda, Robert Černý *
Department of Materials Engineering and Chemistry, Faculty of Civil Engineering, Czech Technical University in Prague, Thákurova 7, 166 29 Prague 6, Czech Republic

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

Article history: The effect of high temperatures up to 1000 °C on the length changes of two alkali-activated aluminosil-
Received 18 March 2008 icate composites, one of them with quartz sand aggregates, the second with electrical porcelain, is ana-
Received in revised form 2 February 2009 lyzed in the paper. The thermal strain vs. temperature functions of both materials are found to increase
Accepted 3 February 2009
monotonically in the whole temperature range studied so that the thermal expansion mismatch (the gel
Available online 12 February 2009
undergoes thermal shrinkage, the aggregates expand with increasing temperature) results in positive val-
ues of the apparent linear thermal expansion coefficient. The composite material with electrical porcelain
Keywords:
aggregates exhibits a more desired thermomechanical behavior which is a consequence of the better
Alkali-activated aluminosilicate composites
Electrical porcelain
high-temperature thermal stability of electrical porcelain as compared to quartz. In a comparison with
Quartz sand Portland-cement based composites, the linear thermal expansion coefficient of both studied aluminosil-
Linear thermal expansion coefficient icates is substantially lower in the whole temperature range of 20–1000 °C.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction pipes or chemical process vessels, envelopes or linings of special

Alkali-activated aluminosilicates have been intensively studied
over the past two decades as a promising alternative to Portland-
cement based composites. The unique combination of high
strength, enhanced fire resistance and good chemical resistance,
which are characteristic for this type of materials (for detailed
analysis see, e.g., [1–3]), has attracted many investigators active
in materials science, chemical engineering and civil engineering.
Mechanical properties have been the most frequently investi-
gated parameters of alkali-activated aluminosilicates, similarly as
in the case of Portland-cement composites. In room-temperature
conditions they were analyzed for instance in [4–10], and after
high-temperature exposure in [11–14]. Other parameters have
been studied only rarely: hydric properties were measured in
[13,15], chloride diffusion in [15–17], and thermal properties in
[12,18]. Volumetric changes of alkali-activated aluminosilicates
were studied mostly in the hydration stage [8,9,19] as their drying
and autogenous shrinkage have always been considered a serious
threat. Length changes at elevated temperatures were studied for
several types of geopolymers [20–22] and for alkali activated slag
pastes [23]. In all cases remarkable thermal shrinkage (up to
25%) was observed.
Alkali-activated aluminosilicates can find use in such applica-
tions in building industry where elevated temperatures may be ex-
pected, such as walls and floors adjacent to various heat machines,
vessels such as coal gasification vessels, nuclear safety related
structures in nuclear power plants, tunnel or shaft walls and fire-
protecting linings. In all these cases volumetric changes in the
high-temperature range can play a significant role. Sudden strain
changes, negative strains or thermomechanical incompatibility of
different parts of a structure can lead to serious damage.
In this paper, the linear thermal expansion coefficient of two al-
kali-activated aluminosilicate composites produced from ground
granulated blast furnace slag in the temperature range up to
1000 °C is measured. Two different types of aggregates, quartz
sand and electrical porcelain, are used, and their effect on thermo-
mechanical behavior of the composites is analyzed.
2. Thermal expansion at high temperatures
In the determination of the linear thermal expansion coefficient
a over a narrow temperature interval (T0, T), which is common in
normal temperature measurements, one can usually assume
a = a0 = const. Then, a single experiment is sufficient consisting of
heating a specimen from the initial temperature T0 to the final
temperature T and measuring the thermal strain e.
In wider temperature ranges, this assumption is no longer valid.
The linear thermal expansion coefficient varies with temperature,
a = a(T), and for the thermal strain we have
Z T

* Corresponding author. Tel.: +420 224354429; fax: +420 224354446. eðTÞ ¼ aðTÞdT ð1Þ
T0
E-mail address: cernyr@fsv.cvut.cz (R. Černý).

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.002
264 L. Zuda, R. Černý / Cement & Concrete Composites 31 (2009) 263–267

In a practical experiment we then choose a reference temperature
T0, heat a specimen step by step to the temperatures Ti, i = 1, . . . , n,
covering the desired temperature range, and determine the corre-
sponding values of ei, i = 1, . . . , n,. The point-wise defined function
ei = f(Ti) is then approximated by means of regression analysis and
its continuous representation e(T) is obtained. Finally, the linear
thermal expansion coefficient a(T) is calculated as the first deriva-
tive of e(T) with respect to temperature.
3. Experimental method
Thermal expansion of solid materials in high-temperature range
is measured by commercially produced dilatometers mostly. Vari-
ous treatments are employed, for instance the methods based on
variations of electric resistance, capacitance, inductance, or the
interference methods. However, the applicability of many common
methods which work with small specimen dimensions may be
rather limited for building materials which are mostly nonhomo-
geneous. Therefore, the comparative method proposed by Toman
et al. [24] which was used for various building materials in the past
was employed for the measurements in this paper. As some mod-
ifications of the experimental setup were done to bring the original
method up to date we will give a brief description of the method in
what follows.
The measuring device for determining the linear thermal
expansion of porous materials in the high-temperature range is
based on the application of a comparative technique (see Fig. 1).
A bar sample of the studied material is put into a cylindrical, ver-
tically oriented electric furnace. As it is technically difficult to per-
form length measurements directly in the furnace, a thin ceramic
rod, which passes through the furnace cover, is fixed on the topside
of the measured sample. The length changes can be determined
outside the furnace in this way, for instance by a digital dial indi-
cator, but on the other hand, the temperature field in the ceramic
rod is very badly defined, and it is not possible to determine di-
rectly, which part of the total change of length is due to the mea-
sured sample and due to the ceramic rod.
Therefore, the measurement is performed at the same time on
the sample of a standard material (such as special steels where
a(T) is known) which is put into the furnace together with the
studied material and is provided with an identical ceramic rod
passing through the over. The change of length of the ceramic
rod can be determined in this way, and consequently also the
length change of the measured sample.

1 - furnace shell (thermal insulation), 2 - furnace cover, 3 - heating coil, 4 – standard, 5 -

measured sample, 6 - ceramic rods, 7 – digital dial indicator, T 1-6 – thermocouples
Fig. 1. Scheme of the measuring device for determination of high temperature linear thermal expansion coefficient.
 L. Zuda, R. Černý / Cement & Concrete Composites 31 (2009) 263–267 265

Table 1
Chemical composition of applied slag.

SiO2 (%) Fe2O3 (%) Al2O3 (%) CaO (%) MgO (%) MnO (%) Cl (%) Na2O (%) K2O (%) SO3 (%)
38.6 0.52 7.22 38.77 12.90 0.50 0.06 0.21 0.38 0.36

It should be noted that temperature field is not constant in the DlðT i Þ

whole volume of the furnace due to the differences of heat loss in
the heated walls and in the cover. Therefore, temperature field in
the furnace is measured by thermocouples, and an average value
of temperature is considered in a calculations.
A practical measurement of the linear thermal expansion coef-
ficient of a porous material with the device proposed in the previ-
ous section can be described as follows. The measured sample and
the standard are put into the furnace, provided with contact cera-
mic rods, and the initial reading on the dial indicators is taken.
Then, the electric heating regulation system is adjusted for the de-
sired temperature Ti in the furnace, and the length changes are
monitored. The data acquisition from the digital dial indicators is
done on PC using specially developed software. After the steady
state is achieved, i.e., no temperature changes in the furnace and
no length changes of both measured sample and the standard are
observed, the final readings of length changes are taken. The length
change of the measured sample is calculated from the following
formula:
Z Ti
DlðT i Þ ¼ Dlm ðT i Þ  Dls ðT i Þ þ l0;s as ðTÞdT;
T0
where Dlm, Dls are the final readings of total length changes of the
studied material and of the standard including the length changes
of the ceramic rods, respectively, l0,s is the initial length of the stan-
dard, and as is the known linear thermal expansion coefficient of the
standard. The corresponding value of thermal strain can be ex-
pressed in the form:
Table 2
Slag granulometry.
Sieve residue mm (%)
0.045 0.09
12.4 1.9
eðT i Þ ¼ ; ð3Þ
l0;m
where l0,m is the initial length of the measured sample.
The measurements are then repeated with other chosen values
of furnace temperature Ti, and the a(T) function of the measured
material is determined using the procedure described in the previ-
ous section.
4. Materials and samples
Two composite materials on the basis of alkali activated slag
were investigated. They differed in the type of aggregates. The first
aggregate type was quartz sand (this material will be denoted as
NS in what follows), the second electrical porcelain (EP). Fine-
ground granulated blast furnace slag of Czech origin (Kotouč
Štramberk, Ltd.) was used for sample preparation. Its chemical
composition is shown in Table 1, its granulometry in Table 2. As al-
kali activator, water glass solution (Na2O  xSiO2  yH2O) was used.
It was prepared using Portil-A dried sodium silicate (Na2SiO3) pre-
ð2Þ parative (Cognis Iberia, s.l. Spain). The quartz sand aggregates were
normalized according to ČSN EN 196-1 with the granulometry gi-
ven in Table 3. Electrical porcelain was provided by P-D Refracto-
ries CZ, Velké Opatovice. Its porosity was 0.3%, water absorption
at saturation 0.1%, bulk density 2350 kg/m3. Chemical composition
of the electrical porcelain is given in Table 4, its granulometry in
Table 5.
The composition of mixtures for sample preparation is pre-
sented in Tables 6a, and b. The technology of sample preparation
was as follows: First, the silicate preparative was mixed with
water. The solution was then mixed in the homogenized slag–
aggregate mixture. The final mixture was put into molds and vi-
Specific surface (cm2/g) brated. The specimens were demolded after 24 h and then stored
for additional 27 days in a water bath at laboratory temperature.
3920
The dimension of the specimens was 40  40  100 mm. Three
specimens for each mixture were used for the measurement of lin-
ear thermal expansion coefficient.

Table 3
Sand granulometry. 5. Experimental results and discussion
Sieve mesh (mm) 2 1.6 1.0 0.5 0.16 0.08
Fig. 2 shows the measured thermal strain of both studied mate-
Total sieve residue (%) 0 7±5 33 ± 5 67 ± 5 87 ± 5 99 ± 1
rials as function of temperature up to 1000 °C. The e(T) functions of

Table 4
Chemical composition of electrical porcelain.

SiO2 (%) Fe2O3 (%) Al2O3 (%) CaO (%) MgO (%) Na2O (%) K2O (%) TiO2 (%)
48.6 0.8 45.4 0.3 0.2 1.0 2.9 0.7

Table 5
Electrical porcelain granulometry.

Sieve mesh (mm) 4.00 2.50 1.00 0.50 0.25 0.125 0.090 0.063 0.045
Total sieve residue (%)
0–1 mm fraction – – 0.69 45.24 70.76 89.98 93.4 98.99 99.99
1–3 mm fraction – 4.12 78.33 99.57 99.94 99.94 99.95 98.98 100.00
3–6 mm fraction 69.31 95.52 99.97 99.98 99.99 100.00 – – –
266 L. Zuda, R. Černý / Cement & Concrete Composites 31 (2009) 263–267

Table 6a
Composition of the mixture with quartz sand aggregates for sample preparation.

Sand (g) Slag (g) Alkali-activation silicate admixture (g) Water (ml)
PG1 PG2 PG3
450 450 450 450 90 190

Table 6b
Composition of the mixture with electrical porcelain aggregates for sample preparation.

Electrical porcelain (g) Slag (g) Alkali-activation silicate admixture (g) Water (ml)
0–1 mm fraction 1–3 mm fraction 3–6 mm fraction
450 450 450 450 90 190

both materials were increasing in the whole temperature range so
that no thermal shrinkage was observed up to 1000 °C, contrary to
the results obtained for geopolymers [20–22] and alkali activated
slag pastes [23]. This difference is apparently a consequence of
the presence of aggregates in the alkali-activated aluminosilicate
composites studied in this paper which caused thermal expansion
mismatch in the material (the gel undergoes thermal shrinkage,
the aggregates expand with increasing temperature); the materials
analyzed in [20–23] were pastes. This statement can be supported
by the investigations in [25] where for Portland cement mortar
thermal expansion was observed in the whole studied temperature
range of 20–500 °C while cement paste underwent thermal shrink-
age for temperatures higher than 150 °C.
The linear thermal expansion coefficients calculated using the
e(T) functions from Fig. 2 are presented in Fig. 3 which makes pos-
sible to analyze the effect of temperature increase on length
changes of both studied aluminosilicates in more details. The
a(T) functions were increasing in the lower-temperature range,
for NS up to approximately 400 °C, for EP to 300 °C. Then the linear
1.20E-02
1.00E-02
thermal strain [-]
thermal expansion coefficient of NS began to decrease so that at
800 °C it had similar value as at room-temperature, and finally in
the temperature range of 800–1000 °C it was almost constant. On
the other hand, the a(T) function of EP was only slightly decreasing
in the temperature range of 300–700 °C, and then it decreased fas-
ter so that at 1000 °C it was similar as at room-temperature.
The differences between the a(T) functions in Fig. 3 can be ex-
plained by the different behavior of the two materials used in
the studied aluminosilicate composites as aggregates after their
exposure to elevated temperatures. Electrical porcelain is charac-
terized by low linear thermal expansion coefficient (6–
8  106 K1 in average) and no sudden volumetric changes in
the temperature range up to 1000 °C [26,27]. Quartz, on the other
hand, is subject of displacive phase transition from a (low) to b
(high) modification at 573 °C [28]. The different quartz forms differ
also in their structural response to increasing temperatures. While
the thermal expansion coefficient of a-quartz is positive
(18  106 K1 in average, calculated using the data in [28,29]), it
was found that b-quartz does not exhibit any significant thermal
expansion from 573 °C up to 1000 °C [28,29] or possibly it even
undergoes thermal shrinkage [30]. Thus, the electrical porcelain
aggregates had clearly better prerequisites to compensate for the
thermal shrinkage of aluminosilicate gel than quartz sand, in the

temperature range of 573–1000 °C in particular.

8.00E-03 The dependence of the linear thermal expansion coefficient on
6.00E-03 temperature of both studied materials was in a good qualitative
agreement with the residual strength values measured in [12,13].
4.00E-03 NS The fast decrease of a(T) of the material NS with quartz sand aggre-
2.00E-03 EP gates between 400 °C and 800 °C was well correlated with com-
pressive strength which decreased more than four times in that
0.00E+00
0 100 200 300 400 500 600 700 800 900 1000 temperature range. The 50% reduction in compressive strength of
temperature [C°] the material EP with electrical porcelain aggregates between
600 °C and 800 °C corresponded to the decrease of a(T) for temper-
Fig. 2. Thermal strain of analyzed materials as function of temperature. atures above 700 °C. So, the thermal expansion properties of differ-
ent aggregates played a significant role in the mechanical
performance of the analyzed aluminosilicates. The important role
of aggregates in that respect can also be supported by a compari-
linear thermal expansion

1.60E-05 NS son with the results obtained in [23]. While for the alkali activated
coefficient [K ]

EP slag concrete with basalt coarse aggregates from [23] the reduction
-1

1.20E-05
in compressive strength between room-temperature and 800 °C
was approximately 90%, for NS in [12] it was 80% and for EP in
8.00E-06
[13] only 70%. In the temperature range of 800–1200 °C the role
of the type of aggregates was even more notable. The alkali acti-
4.00E-06
vated slag concrete from [23] lost at 1200 °C its strength com-
pletely but quartz and electrical porcelain aggregates in [12,13]
0.00E+00
0 200 400 600 800 1000 1200 provided the necessary compounds for ceramic bond formation
which caused the compressive strength to increase more than five
temperature [C°]
times for both NS and EP.
Fig. 3. Linear thermal expansion coefficient of analyzed materials as function of The thermal expansion properties of alkali-activated alumino-
temperature. silicate composites obtained in this paper can be compared with
 L. Zuda, R. Černý / Cement & Concrete Composites 31 (2009) 263–267
common cementitious composites in a limited extent only as very
few results were published until now. In [25], the total thermal
strain of cement mortar at 500 °C was approximately 9  103.
This was 20% higher than for the aluminosilicate with quartz sand
in this paper and almost 40% higher than for the material with
electrical porcelain aggregates. The a(T) functions determined up
to 1000 °C for cement mortar in [24] and for high performance con-
crete in [31] were similar in shape to the aluminosilicate material
with quartz sand but 20–50% higher in the whole temperature
range. As lower thermal strain generally results in lower thermal
stress, thus in lower risk of failure of a building element subjected
to one-sided heating, the thermomechanical behavior of the alumi-
nosilicates studied in this paper can be characterized as mostly
better than of Portland-cement based composites.
6. Conclusions
Experimental investigations presented in this paper showed
that the thermal expansion properties of aggregates affected the
thermomechanical behavior of the analyzed alkali-activated alu-
minosilicate composites in a significant way. The main findings
can be summarized as follows:
 Contrary to the results obtained for geopolymers [20–22] and
alkali activated slag pastes [23], no thermal shrinkage was
observed up to 1000 °C; the thermal expansion mismatch
between the contracting aluminosilicate gel and expanding
aggregates resulted, for the aluminosilicate composites with
both quartz sand and electrical porcelain aggregates, in positive
values of the apparent linear thermal expansion coefficient a.
 Electrical porcelain aggregates were identified as a more desired
solution from the point of view of thermomechanical properties
of aluminosilicate composites; lower thermal strain and a
smoother a(T) function were observed as compared with a sim-
ilar composite produced using quartz sand. This was a conse-
quence of better high-temperature volumetric stability of
electrical porcelain, which is characterized by low linear thermal
expansion coefficient and no sudden volumetric changes in the
temperature range up to 1000 °C, while quartz is subject to a
displacive phase transition from a to b modification at 573 °C.
In addition, the different quartz forms differ also in their struc-
tural response to increasing temperatures.
 In a comparison with Portland-cement based composites the
linear thermal expansion coefficient of both studied aluminosil-
icates was in the whole temperature range of 20–1000 °C signif-
icantly lower. As lower thermal strain generally results in lower
thermal stress, and thus in lower risk of failure of a building ele-
ment subjected to one-sided heating, their thermomechanical
behavior can be characterized as better than Portland-cement
based composites.
 The development of thermal strain with temperature was for
both studied materials in a good correlation with the measure-
ments of residual compressive strength presented before
[12,13].
Acknowledgement
This research has been supported by the Ministry of Education,
Youth and Sports of the Czech Republic, under Project No. MSM:
6840770031.
267
References
[1] Davidovits J. Geopolymers and geopolymeric materials. J Therm Anal
1989;35:429–41.
[2] Davidovits J. Geopolymers – inorganic polymeric new materials. J Therm Anal
1991;37:1633–56.
[3] Khale D, Chaudhary R. Mechanism of geopolymerization and factors
influencing its development: a review. J Mater Sci 2007;42:729–46.
[4] Wang SD, Pu XC, Scrivener KL, Pratt PL. Alkali-activated slag cement and
concrete: a review of properties and problems. Adv Cem Res 1995;7:93–102.
[5] Bakharev T, Sanjayan JG, Cheng YB. Effect of admixtures on properties of alkali-
activated slag concrete. Cem Concr Res 2000;30:1367–74.
[6] Collins F, Sanjayan JG. Microcracking and strength development of alkali
activated slag concrete. Cem Concr Compos 2001;23:345–52.
[7] Chang JJ, Yeih W, Hung CC. Effects of gypsum and phosphoric acid on the
properties of sodium silicate-based alkali-activated slag pastes. Cem Concr
Compos 2005;27:85–91.
[8] Palacios M, Puertas F. Effect of superplasticizer and shrinkage-reducing
admixtures on alkali-activated slag pastes and mortars. Cem Concr Res
2005;35:1358–67.
[9] Palacios M, Puertas F. Effect of shrinkage-reducing admixtures on the
properties of alkali-activated slag mortars and pastes. Cem Concr Res
2007;37:691–702.
[10] Jo BW, Park SK, Park MS. Strength and hardening characteristics of activated fly
ash mortars. Mag Concr Res 2007;59:121–9.
[11] Shoaib MM, Ahmed SA, Balaha MM. Effect of fire and cooling mode on the
properties of slag mortars. Cem Concr Res 2001;31:1533–8.
[12] Zuda L, Pavlík Z, Rovnaníková P, Bayer P, Černý R. Properties of alkali activated
aluminosilicate material after thermal load. Int J Thermophys
2006;27:1250–63.
[13] Zuda L, Bayer P, Rovnaník P, Černý R. Mechanical and hydric properties of
alkali-activated aluminosilicate composite with electrical porcelain
aggregates. Cem Concr Compos 2008;30:266–73.
[14] Kong DLY, Sanjayan JG, Sagoe-Crentsil K. Factors affecting the performance of
metakaolin geopolymers exposed to elevated temperature. J Mater Sci
2008;43:824–31.
[15] Shi C. Strength, pore structure and permeability of alkali-activated slag
mortars. Cem Concr Res 1996;26:1789–99.
[16] Douglas E, Bilodeau A, Malhotra VM. Properties and durability of alkali-
activated slag concrete. ACI Mater J 1992;89:509–16.
[17] Roy DM, Jiang W, Silsbee MR. Chloride diffusion in ordinary, blended and
alkali-activated cement pastes and its relation to other properties. Cem Concr
Res 2000;30:1879–84.
[18] Zuda L, Rovnaník P, Bayer P, Černý R. Thermal properties of alkali activated
slag subjected to high temperatures. J Build Phys 2007;30:337–50.
[19] Melo Neto AA, Cincotto MA, Repette W. Drying and autogenous shrinkage of
pastes and mortars with activated slag cement. Cem Concr Res
2008;38:565–74.
[20] Duxson P, Lukey GC, van Deventer JSJ. Thermal evolution of metakaolin
geopolymers: part 1 – physical evolution. J Non-Cryst Solids
2006;352:5541–55.
[21] Duxson P, Lukey GC, van Deventer JSJ. Physical evolution of Na-geopolymer
derived from metakaolin up to 1000 °C. J Mater Sci 2007;42:3044–54.
[22] Subaer, van Riessen A. Thermo-mechanical and microstructural
characterisation of sodium-poly(sialate-siloxo) geopolymers. J Mater Sci
2007;42:3117–23.
[23] Guerrieri M, Sanjayan JG, Collins FG. Residual compressive behavior of alkali
activated concrete exposed to elevated temperatures. Fire Mater
2009;33:51–62.
[24] Toman J, Koudelová P, Černý R. A measuring method for the determination of
linear thermal expansion of porous materials at high temperatures. High
Temp-High Press 1999;31:595–600.
[25] Fu YF, Wong YL, Poon CS, Tang CA, Lin P. Experimental study of micro/macro
crack development and stress–strain relations of cement-based composite
materials at elevated temperatures. Cem Concr Res 2004;34:789–97.
[26] Amigo JM, Clausell JV, Esteve V, Delgado JM, Reventos MM, Ochando LE, et al.
X-ray powder diffraction phase analysis and thermomechanical properties of
silica and alumina porcelains. J Eur Ceram Soc 2004;24:75–81.
[27] Kamseu E, Leonelli C, Boccaccini D. Non-contact dilatometry of hard and soft
porcelain compositions. J Therm Anal Calorim 2007;88:571–6.
[28] Carpenter MA, Salje EKH, Graeme-Barber A, Wruck B, Dove MT, Knight KS.
Calibration of excess thermodynamic properties and elastic constant
variations associated with the a ? b phase transition in quartz. Am Mineral
1998;8:2–22.
[29] Lakshtanov DL, Sinogeikin SV, Bass JD. High-temperature phase transitions and
elasticity of silica polymorphs. Phys Chem Miner 2007;34:11–22.
[30] Huang LP, Kieffer J. Structural origin of negative thermal expansion in high-
temperature silica polymorphs. Phys Rev Lett 2005;95:215901.
[31] Toman J, Černý R. Thermal and hygric expansion of high performance concrete.
Acta Polytech 2001;41:2–4.
 Cement & Concrete Composites 31 (2009) 275

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

News: conferences and symposia

Second International Symposium on Design, Performance and
Use of SELF-CONSOLIDATING CONCRETE (SCC’2009-China), to be
held in Beijing, China, 5–8 June 2009.
Contact: Dr. Bo Peng, College of Civil Engineering, Hunan
University, Changsha, Hunan 410082, China. Tel/Fax: 86-731-882-
3937; E-mail: SCC2009@hnu.cn; Website: http://scc2009.hnu.cn.
1st International Workshop on Concrete Spalling due to Fire
Exposure, to be held in MFPA Institute, Leipzig, Germany, 3–5
September 2009.
Contact: Workshop Secretariat, E-mail: spalling@mfpa-leipzig.de;
Website for more information: www.mfpa-leipzig.de.
Non-conventional Materials and Technologies (IC-NOCMAT
2009): Materials for sustainable and affordable construction, to
be held in Bath, United Kingdom, 6–9 September 2009.
Contact: Prof. Pete Walker, BRE Centre for Innovative Construc-
tion Materials, University of Bath, UK, E-mail: p.walker@bath.ac.uk;
Website: www.bath.ac.uk/ace/nocmat2009.
The Ninth International Symposium on Brittle Matrix Compo-
sites, to be held in Warsaw, Poland, 25–28 October 2009.
Contact: Prof. Dr. A.M. Brandt, Polish Academy of Sciences,
Institute of Fundamental Technological Research (IPPT PAN), Swieto-
krzyska 21, Warsaw 00-049, Poland. E-mail: abrandt@ippt.gov.pl;
Website: http://www.ippt.gov.pl/~abrandt/geninfo1.htm.
International Conference on Computational Design in Engineer-
ing (CODE 2009), to be held in Seoul, Korea, 3–6 November 2009.
Contact: Conference Secretariat, E-mail: code@code2009.org;
Website: http://www.code2009.org.
Advanced Concrete Materials (ACM), International Conference, to
be held in Stellenbosch, South Africa, 17–19 November 2009.
Contact: Dr. Billy Boshoff, University of Stellenbosch, South Africa.
E-mail: admin@concretesociety.co.za; Website: http://www.concre-
tesociety.co.za/acm2009.

doi:10.1016/S0958-9465(09)00051-1
 Cement & Concrete Composites 31 (2009) 268–274

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Cemented paste backfill of sulphide-rich tailings: Importance

of binder type and dosage
_
Bayram Ercikdi *, Ayhan Kesimal, Ferdi Cihangir, Haci Deveci, Ibrahim Alp
Mining Engineering Department, Engineering Faculty, Karadeniz Technical University, 61080 Trabzon, Turkey

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

Article history: In this study, the influence of binder type and dosage on the mechanical properties and microstructure of
Received 21 July 2008 cemented paste backfill (CPB) was investigated using ordinary Portland cement (OPC), Portland compos-
Received in revised form 28 January 2009 ite cement (PCC) and sulphate resistant cement (SRC). The CPB samples of OPC and PCC were observed to
Accepted 30 January 2009
lose their unconfined compressive strengths (UCSs) after 56 days. This could be associated with the sul-
Available online 9 February 2009
phide moiety of the tailings, i.e. the attack on hydration products by sulphate and acid internally gener-
ated via the oxidation of pyrite present. In this respect, those CPB samples of sulphate resistant-based
Keywords:
cements (SRC and a mix of OPC and SRC) maintained good long-term strengths and stability (i.e. no loss
Cemented paste backfill
Binder type
of strength). Increasing binder dosage (5–7 wt.%) improved the UCSs of CPB samples up to 1.9-fold with
Binder dosage no loss of strength at >5 wt.%. Decreasing water-to-cement ratio appeared to produce a beneficial effect
Water-to-cement ratio on the UCSs of CPB samples. SEM studies have provided further insight into the microstucture of CPB and
Unconfined compressive strength confirmed the deleterious formation of gypsum as the expansive phase. These findings have demon-
Sulphide tailings strated the practical importance of binder type/dosage and water-to-cement ratio for the short- and
long-term mechanical performance of CPB.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction unconfined compressive strength of 0.7–2 MPa is requisite for

Mining activities involve the extraction of ores from the earth’s
crust. The underground mine openings created due to the ore
extraction can be filled using waste materials with the process
being known as backfilling [1]. This (i) provides ground support
for the mine structures with a safe working environment, (ii) mit-
igates the risk of surface subsidence associated with the under-
ground stopes, and (iii) provides a disposal site for waste rocks
or mill tailings. In practice, hydraulic fill, rock fill and cemented
paste backfill (CPB) are the backfilling methods; but, the latter
has, in recent years, gained worldwide acceptance for use as an
integral part of underground mining operations [1,2]. The most sig-
nificant benefits of CPB include the alleviation of the environmen-
tal impact of potentially hazardous mill tailings (e.g. sulphide
tailings) by placing them safely into underground openings and
the reduction of the tailings disposal and rehabilitation costs [3–5].
CPB is a mixture of process tailings (70–85wt.% solids), a
hydraulic binder (3–8wt.%) and water [6]. The inherent character-
istics of these constituents essentially control the short- and long-
term mechanical performance of CPB and are, hence, of practical
importance [7,8]. The addition of binder is essential to achieve
the desired strength and stability of CPB for which a 28-day
* Corresponding author. Tel.: +90 462 3773171; fax: +90 462 3257405.
E-mail address: bercikdi@ktu.edu.tr (B. Ercikdi).
most applications [6]. In a typical CPB plant, binder costs (approx-
imately US$1 per ton of CPB for each percentage point of binder)
may comprise up to 80% of the operating costs [9]. Therefore, the
utilization of suitable and cost-effective binders, and the optimiza-
tion of binder dosage assume paramount importance in practice.
Some stability problems (i.e. loss of strength) in long-term have
been reported to occur for CPB of sulphide-rich mill tailings
[10–12]. Although the reasons behind these problems have yet to
be unequivocally elucidated, they are often associated with the
presence of sulphide minerals (pyrite in particular) in the tailings.
In contact with oxygen and moisture, the oxidation of pyrite occurs
and leads to the formation of acid and sulphate, which can then re-
act with hydration products and binder phases such as calcium
hydroxide (CH) and calcium aluminate (C3A). Concomitantly, these
undesirable reactions result in the formation of expansive phases
such as ettringite and gypsum (i.e. referred to as sulphate attack)
[11,12]. These could then culminate in the reduced backfill
strength and potentially collapse of the backfill [8]. In this regard,
the chemical and mineralogical characteristics of binder (i.e. binder
type) and permeability contribute, to a large extent, to the sulphate
resistance of CPB [8,13,14].
Calcium rich binders such as ordinary Portland cement (e.g.
ASTM Type I) are particularly prone to sulphate attack [15]. Port-
land composite cements (PCC) containing active mineral admix-
tures (fly ash, slag, silica fume or natural pozzolan) or sulphate

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.01.008
 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274 269

resistant cements (e.g. ASTM Type V) may be used in sulphate Table 1

environments including CPB of sulphide-rich tailings [11,16–19]. Chemical composition of the sulphide tailings used.

Sulphate resistance of a binder is linked closely with its content Compound % Compound %
of C3A and C4AF [17]. ASTM Type I (Ordinary Portland Cement) typ- MgO 1.19 K2O 0.23
ically contains 8–12% C3A compared with 68% and 65% for PCC and Al2O3 3.27 Na2O 0.17
Type V cements, respectively. In addition to C3A content, the ratio SiO2 11.39 TiO2 <0.01
of C3S/C2S is also claimed to be an indication of the sulphate resis- CaO 0.95 SO3 2.83
FeO 22.92 CI 0.0025
tance of a binder on the ground that the hydration of C3S produces S 26.22 Loss-on-ignition (LOI) 29.02
nearly 2.2 times more CH than that of C2S [17]. CH is often respon-
Total 98.2025
sible for the formation of gypsum and ettringite as aforemen-
tioned. The sulphate performance of Portland composite cements
is associated allegedly with the pozzolanic properties of mineral
admixtures present contributing to pore refinement, dilution of
C3A and consumption of CH by pozzolanic reactions [20].
In this study, the short- and long-term performance of three dif-
ferent binders (ASTM Type I, V and PCC) in CPB of a sulphide-rich
(pyritic) mill tailings were evaluated over a curing period of
360 days. The effects of binder dosage (5–7 wt.%) and water-to-ce-
ment ratio on the strength and durability of CPB were also examined.
SEM studies were also undertaken to correlate microstructure and
mineralogy of CPB samples with the mechanical performance and
the inherent characteristics of binders used.

2. Materials and methods

2.1. Tailings sample

In this study, a mill tailings sample (2 tonnes) produced from

a copper–zinc flotation plant was used as representative of the feed
to the CPB plant currently in operation. Following sampling, the
physical, chemical and mineralogical characterisations of the tail- Fig. 2. XRD profile of the tailings sample (Py (Pyrite), Qtz (Quartz)).
ings were performed. Particle size of the tailings was determined
using a Malvern Mastersizer (S Ver. 2.15 particle size analyzer)
(Fig. 1). The tailings with fines (20 lm) content of 40% can be
(PCC) and sulphate resistant cements ((SRC) (Type V) alone and
classified as a medium size tailings material with a relatively high
as a 50:50 mixture with OPC) were evaluated. It should be noted
potential for water retention to form a paste [21]. The tailings sam-
that PCC contains a natural pozzolan of volcanic origin (14%) and
ple were determined to contain 26.2% S (Table 1), which appeared
limestone (6%) as mineral admixtures. Binder dosage was kept con-
to be predominantly pyritic in nature as XRD analysis already con-
stant at 5 wt.% of solids (tailings plus binder on dry basis) except in
firmed the presence of pyrite as the sole sulphide phase identified
the tests where PCC was used to study the effect of binder dosage
(Fig. 2).
(5–7 wt.%). Prior to their use in the CPB tests, the physical and
chemical properties of these binders were examined in some de-
2.2. Binders used
tail. Fig. 1 illustrates that SRC has a relatively coarser size distribu-
tion than OPC and PCC as also confirmed by its smaller surface area
In this study, the short- and long-term performances of ordin-
(Table 2). Compared with OPC and PCC, SRC was found to have the
ary Portland cement (OPC) (Type I), Portland composite cement
highest content of CaO and Fe2O3 whilst it contained Al2O3 at the
lowest level (Table 2). Derived from the chemical composition of
the binders, Bogue calculations proved that SRC had the lowest
100
OPC proportion of C3A (Table 2), which was consistent with its inherent
PCC characteristic of sulphate resistance.
SRC
Tailings Sample
80
Cumulative Passing (%)

2.3. Preparation and testing of CPB samples

60 A number of CPB samples (192 in total) were prepared by mix-

ing and homogenizing the tailings, binder (OPC, PCC, and SRC) and
tap water in a Univex SRMF20 Stand model blender equipped with
40 a double spiral. The solids contents of paste mixtures were set to
77 wt.%. The effect of binder type and water-to-cement ratio was
examined at a binder dosage of 5 wt.% while binder dosage was
20 tested at 5–7 wt.% (Table 3). The chemical composition of mix
water (the tailings water and tap water) with particular reference
to sulphate was also determined. Having a sulphate content of
0 1283 mg/l, the mix water can be classified as ‘‘aggressive” towards
1 10 100
concrete/backfill durability as suggested by DIN 4030 [22]. It
Particle Size (micron)
should be noted that the residual water of sulphide tailings, which
Fig. 1. Particle size distribution of the tailings sample and binders. is inherently rich in sulphate due to the air and the sulphate con-
270 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274

Table 2
Chemical, physical and mineralogical properties of the binders used in the tests.

Oxide composition OPC (%) PCC (%) SRC (%) Physical properties OPC PCC SRC
SiO2 20.31 24.51 21.31 Specific gravity 3.01 2.94 3.20
Al2O3 5.93 7.15 4.71 Specific surface (cm2/g) 4345 4280 3140
Fe2O3 2.82 3.42 4.31 Retained on 90 lm sieve (%) 2.16 2.09 8.03
CaO 61.02 54.36 62.13 Retained on 45 lm sieve (%) 15.84 15.43 32.96
MgO 1.15 1.24 1.26
SO3 2.95 2.92 2.88 Compound composition
Na2O 0.32 0.53 0.24 C3S 37.10 – 41.43
K2O 1.14 1.38 0.75 C2S 30.30 – 29.90
Free CaO 1.14 – 0.85 C3A 10.95 – 5.20
Loss on ignition 3.78 3.75 1.54 C4AF 8.57 – 13.10

Table 3
A summary of the experimental conditions used in the preparation of CPB samples.

Solids content (SC)a, wt.% Binder dosage (BD)b, wt.% Water to cement ratio (w/c)c Slump (in.)
Binder type 77.0 5 5.97 7.0
Binder dosage 77.0 5 5.97 7.0
6 4.98
7 4.27
Water-to-cement ratio 76.5 5 6.14 7.5
77.0 5.97 7.0
77.5 5.81 6.5
a 100ðMdrytailings þM drybinder Þ
SC : ðM drytailings þM drybinder þM water Þ.
b 100ðM drybinder Þ
BD : ðM drybinder þM drytailings Þ.
c M water
w=c : M drybinder ; ðM : WeightÞ.

taining chemicals used in the flotation process and constitutes 75– press, which had a load capacity of 50 kN and a displacement speed
80% of the mix water, contributes extensively to the sulphate con- of 0.5 mm per minute.
tent of mix water in CPB applications.
The CPB mixtures were thoroughly mixed and poured into plas- 2.4. Studies on microstructure and mineralogy of CPB
tic cylinders (D  H: 10  20 cm) with a perforated bottom. The
open-top cylinders were then allowed to cure in a room main- The microstructure and texture of some 360-day CPB samples
tained at a relative humidity of 80% i.e. to mimic the under- were examined under a LEO scanning electron microscope oper-
ground curing conditions. At the predetermined curing periods ated at an accelerating voltage of 15 kV. Energy dispersive spec-
(up to 360 days), the CPB samples were subjected to unconfined trometry (EDS) of an X-ray probe coupled to the SEM was also
compressive strength (UCS) tests according to ASTM C 39. UCS used to aid the identification of mineral phases present. The frac-
tests were performed using a computer-controlled mechanical tured samples obtained from UCS tests were used in SEM studies.

1
1.6
Compressive strength UCS (MPa)

Compressive strength UCS (MPa)

1.4
0.8

1.2

0.6
1

0.8
0.4
0.6

0.2 0.4
OPC PCC SRC 50%OPC-50%SRC
0.2 5 wt.% 6 wt.% 7 wt.%
0
0 30 60 90 120 150 180 210 240 270 300 330 360 390 0
0 30 60 90 120 150 180 210 240 270 300 330 360 390
Curing time (days)
Curing time (days)
Fig. 3. The effect of binder type on the mechanical performance of CPB samples in
the short and long-term (OPC: ordinary Portland cement, Type I; PCC: Portland Fig. 4. The effect of binder dosage (5–7 wt.% PCC) on the mechanical performance
composite cement; SRC: Sulphate resistant cement, Type V). of CPB samples in the short and long-term.
 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274 271

1 Samples were scanned over a 2-h range of 5–70° with a 0.005° step
size.
Compressive strength UCS (MPa)

0.8
3. Results and discussion

0.6 3.1. Effect of binder type

Fig. 3 illustrates the short- and long-term (over 360 days)

0.4 mechanical performances of CPB samples prepared from OPC,
PCC and sulphate resistant-based cements (SRC and a 50:50 mix-
ture of OPC and SRC) at a fixed binder dosage of 5 wt.%. Over the
0.2 curing period of 56 days, a trend of increase in the strength devel-
w/c: 5.81 5.97 6.14 opment of the CPB samples was observed. However, after 28 days
of curing, only the CPB samples prepared from OPC had the re-
0
0 30 60 90 120 150 180 210 240 270 300 330 360 390 quired UCS of >0.7 MPa. The rate of strength development of CPB
Curing time (days) samples of sulphate resistant cements was consistently slower
than those of OPC and PCC (Fig. 3).
Fig. 5. The effect of water-to-cement (w/c) ratio on the mechanical performance of The strength development of CPB samples prepared from OPC
CPB samples in the short and long-term (at 5 wt.% PCC). peaked (0.833 MPa) at 56 days; thereafter, a trend of loss in
strength (i.e. by 24.6% between 56 and 360 days) was observed.
A similar (though less severe) strength loss for CPB samples of
The samples were dried in an oven controlled at 50 °C and treated PCC also occurred. These findings suggest that OPC and PCC at
with acetone to cease further hydration prior to SEM analysis. 5 wt.% dosage are not particularly suitable for CPB of the sul-
X-ray diffraction (XRD) (Philips X’pert PW 3040 Diffractometer) phide-rich tailings such as that used in this study. The apparent
analyses of the acetone-treated CPB samples were also carried out strength losses of CPB samples observed after 56 days could be as-
to investigate the crystalline phases present with particular refer- cribed to the sulphate attack phenomenon presumably associated
ence to the secondary minerals formed during the curing process. with high sulphide content of the mill tailings, high SO24 content

Fig. 6. SEM images of the CPB samples of OPC (a) and (b), SRC alone (c) and (d) and a blend of SRC/OPC (50–50%) (e) and (f) at a binder dosage of 5 wt.%.
272 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274
of mixing water and relatively high C3A content of these binders.
The development of sulphate attack in CPB can be presumed to oc-
cur via oxidation of pyrite in the presence of air and moisture,
which leads to the formation of acid and sulphate (Eq. (1)). These
then react with hydration products to form expansive phases such
as ettringite (Eq. (2)) and gypsum (Eq. (3)), and to decalcify/de-
struct C–S–H [8,14,15,23,24].
4FeS2 þ 15O2 þ 8H2 O ! 2Fe2 O3 þ 8SO2
4 þ 16H
þ
3CaO  Al2 O3 þ 3CaSO4  2H2 O þ 30H2 O
! 3CaO  Al2 O3  3CaSO4  32H2 O
CaðOHÞ2 þ SO2  beneficial effect of increasing the binder dosage (5–7 wt.% PCC) on
4 þ 2H2 O ! CaSO4  2H2 O þ 2OH
In contrast to OPC and PCC, the sulphate resistant binders (SRC
alone and SRC/OPC mixture) maintained the strength of CPB sam-
ples in the long-term (360 days) and continued to gain strength
even after 56 days (Fig. 3). It is also pertinent to note that both
SRC and SRC/OPC failed to produce the required UCS of >0.7 MPa
over the initial curing period of 28 days. The mixed binder (SRC/
OPC) was observed to enhance the rate of development of strength
of CPB samples compared with SRC alone. The SRC and OPC can be
assumed to contribute to the durability and strength development,
respectively, of CPB samples when used together in the binder
phase. As shown in Table 2, SRC is characterised by its low C3A con-
tent (5.2% c.f. 10.95% OPC), which is an important phase in inti-
mately controlling sulphate resistance [17,25]. The incorporation
Fig. 7. SEM images of the CPB samples prepared from PCC at a binder dosage of 5 wt.% (a) and (b) and 7 wt.% (c) and (d), and at a water-to-cement (w/c) ratio of 5.81 (e) and
(f).
of SRC into the binder phase (OPC/SRC cement) lowers the C3A con-
tent and improves the ability of the binder to endure the sulphate
attack. Conversely, the higher SRC content may also lead to an in-
crease in the C3S/C2S ratio and hence, the availability of CH to react
with SO24 giving increased susceptibility to sulphate attack. It is
also of pertinent to note that SRC is usually more expensive than
OPC and PCC and therefore, the use of SRC blended with OPC will
typically be more cost effective than SRC on its own.
ð1Þ
3.2. Effect of binder dosage
ð2Þ Binder type and dosage exert a profound influence on the short-
and long-term mechanical performance of CPB. Fig. 4 illustrates the
ð3Þ
the strength development of CPB samples over 360 days. At the
binder dosages of 6 and 7 wt.%, CPB samples were able to achieve
the required UCS (P0.7 MPa) over a curing period of 28 days. How-
ever, those CPB samples prepared at 5 wt.% PCC failed to produce a
UCS of above 0.661 MPa even over a 56-day period of curing. In
comparison, the mechanical strength of CPB samples containing
7 wt.% PCC was approximately 1.9 times higher than those contain-
ing 5 wt.% PCC over 56 days.
The increase in the binder dosage from 5 to 6 or 7 wt.% signifi-
cantly improved the long-term performance of PCC (Fig. 4). The
CPB samples at 5 wt.% PCC suffered a strength loss of 15% between
56th and 360th days. On the contrary, those samples prepared at
higher binder dosages gained additional strength (i.e. by 15–17%)
over the same curing period. These findings are consistent with
 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274 273

those of Fall and Benzaazoua [12]. The beneficial effect of increas-
ing binder content is attributed to the production of secondary C–
S–H gels as a result of a pozzolanic reaction of mineral admixtures
in PCC (20.7%) with CH formed during the hydration process. Pro-
viding the additional binding capacity, these C–S–H gels can also
form a protective coating on the alumina-rich phases, which could
otherwise participate in the formation of ettringite [25]. Further-
more, these extra C–S–H gels can improve the microstructure of
CPB with the resultant reduction in permeability and probably in
the oxidation pyrite and hence, the vulnerability of CPB to sulphate
attack [26–29]. Despite these potential benefits, increasing the bin-
der content will augment the operating costs of CPB.
3.3. Effect of water-to-cement ratio
The flowability of CPB mixture is significant for its transport
from the paste preparation plant to the underground openings to
be filled. CPB is therefore required to contain sufficient water to
achieve the desired consistency. However, an increase in the
water-to-cement (w/c) ratio adversely affects strength develop-
ment of CPB [8]. The optimum w/c ratio is essentially a balance be-
tween maximizing the strength required and maintaining the
consistency of paste for the effectiveness of pumping.
Fig. 5 presents the effect of w/c ratio on the mechanical perfor-
mance of the CPB samples prepared from PCC at 5 wt.% binder dos-
age. Only at the lowest w/c ratio of 5.81 tested, the CPB samples
were able to develop the strengths beyond the threshold level of
0.7 MPa only after a curing period of 56 days. Increasing the w/c
ratio resulted in a degree of loss of strength of CPB samples in
the long-term (Fig. 5). The adverse effect of increased w/c levels
can be ascribed to the deterioration of the CPB microstructure
due to the increased porosity and void ratio [13].
detected in the SEM analysis of the samples. Fall and Benzaazoua
[12] also observed the formation of secondary gypsum, which
was claimed to be responsible for the reduction of strength of
CPB samples after 56 days due to its expansive properties. Bing
and Cohen [34] also demonstrated that formation of gypsum as a
result of sulphate attack caused an expansion on concrete and con-
sequently led to the loss of its strength and stability. However, the
role of gypsum formation in the deterioration of cemented materi-
als is controversial by some workers [35,36].
Increasing the binder dosage from 5 wt.% to 6–7 wt.% or
decreasing the w/c ratio from 6.14 to 5.81 (at 5 wt.% PCC) appeared
to improve the microstructure of CPB samples leading to a dense
packing with the visible reduction in porosity and permeability
(Fig. 7a, c, and e). The compactness of CPB matrix can be associated
with the active mineral admixture (e.g. natural pozzolan) present
inherently in PCC binder. These admixtures are presumed to con-
sume Ca(OH)2 liberated during the hydration process through poz-
zolanic reactions and to produce secondary C–S–H gels with
binding properties [18,25,27–29,37–39]. In this regard, a visible
C–S–H densification between pyrite grains were observed to occur
with increasing binder dosage and decreasing w/c ratio (Fig. 7d and
f c.f. Fig. 7b).
Fig. 8 illustrates XRD profiles obtained for CPB samples of OPC
alone and SRC/OPC. The profiles for PCC and SRC were similar to

3.4. Microstructural and mineralogical assessment of CPB samples

After 360 days of curing, CPB samples were examined under

SEM to provide a visual insight into the microstructure of these
samples. EDS unit fitted to SEM allowed the chemical characterisa-
tion and identification of the mineral and cementitious phases
within the CPB matrix. Fig. 6 illustrates the micrographs of the
CPB samples produced from the binders containing only OPC
(Fig. 6a and b), SRC (Fig. 6c and d) and a blend of SRC and OPC
(Fig. 6e and f) at a dosage of 5 wt.%. Fig. 7 shows the influence of
binder dosage (Fig. 7a–d) and w/c ratio (Fig. 7e and f) using PCC
as the binder.
The microstructure of CBP samples (Figs. 6 and 7) was consis-
tent with their respective mechanical performance for stability
over 360 days of curing period (Fig. 3). The importance of micro-
structure for the strength and stability of mortar and CPB has been
highlighted in many studies [16,24,30–32]. These studies have in-
ferred that high strength and stability of mortar/CPB are generally
concomitant with a compact or dense matrix as per the current
findings in this study. To illustrate, accordant with their relatively
poor performance in the long-term, CPB samples of OPC only had a
higher porosity than those of PCC and SRC-based binders (Fig. 6a, c,
and e). The densification of the microstructure appeared to be the
most extensive for those CPB samples obtained from SRC-based
binders (Fig. 6c and e).
The loose microstructure of the CPB samples such as those pre-
pared solely from OPC (Fig. 6a) would be expected to facilitate the
ingress of moisture and air, and hence, the oxidation of pyrite pres-
ent in CPB. In other words, the vulnerability of CPB to sulphate at-
tack increases with an increase in porosity of CPB matrix [31,33].
Fig. 8. XRD patterns for CPB samples prepared from OPC (a), and a blend of OPC and
Gypsum that forms as a potentially deleterious, secondary mineral SRC (b) after a curing period of 360 days (Py (Pyrite), Gp (Gypsum), Qtz (Quartz), Po
phase via sulphate attack was identified in CPB matrix of OPC only (Portlandite), Kln (Kaolinite), Sa (Sanidine), Ill (Illite), Chl (Chlorite), C (Plaster of
as the binder (Fig. 6b). It is relevant to note that no ettringite was Paris), Fsp (Feldspar) and Brt (Barite)).
274 B. Ercikdi et al. / Cement & Concrete Composites 31 (2009) 268–274
that presented in Fig. 8b for SRC/OPC and hence not shown. XRD
analysis of CPB samples (Fig. 8) revealed the presence of gypsum
and portlandite only in the sample prepared from OPC alone as
the binder (Fig. 8a). No portlandite was detected in the CPB sam-
ples of PCC, SRC and SRC/OPC (Fig. 8b). It should be also noted that
no ettringite was identified in XRD studies.
4. Conclusions
This study highlights the potential problems associated with
sulphide-rich tailings and the practical importance of binder type,
dosage and water-to-cement ratio for the strength and stability of
CPB material in the short- and long-term. The mechanical perfor-
mance of a binder in CPB of sulphide-rich tailings is closely associ-
ated with its susceptibility to sulphate attack. The binders,
ordinary Portland cement (OPC), Portland composite cement
(PCC) and sulphate resistant cements (SRC and SRC/OPC) appeared
to be not particularly suitable for CPB of sulphide-rich tailings at
low binder dosages (e.g. 5 wt.%). The long-term mechanical perfor-
mance of CPB was sustained when sulphate resistant cements were
used alone (SRC) or when blended with OPC. The latter would usu-
ally be the most economic solution due to the higher cost of SRC.
However, further tests at higher binder dosages (>5%) are required
to ameliorate the short-term performance of SRC/OPC. The short-
and long-term performances of PCC can be improved by increasing
the dosage beyond 5 wt.% to achieve the desired 28-day strength of
>0.7 MPa and to maintain the stability (i.e. no loss of strength) over
360 days of curing. Reducing the w/c ratio also alleviates, to some
extent, the loss of strength of CPB samples of PCC. SEM observa-
tions for the microstructure of CPB samples have provided visual
evidence for the comparative performances of these binders under
the conditions tested in that the strength and stability of CPB are
linked intimately with the compactness (i.e. permeability and
porosity) of its matrix.
Acknowledgement
The authors would like to express their sincere thanks and
appreciation for the financial support to the Research Foundation
of Karadeniz Technical University (Project No. 2005.112.008.1)
and to TUBITAK (The Scientific and Technological Research Council
of Turkey) (Project No. 107M183).
References
[1] Grice T. Underground mining with backfill. In: The 2nd annual summit-mine
tailings disposal systems. Brisbane, Australia; 1998.
[2] Hassani FP, Archibald JF. Mine backfill. CD-ROM. Canadian Institute of Mine,
Metallurgy and Petroleum; 1998.
[3] Landriault DA, Lidkea W. Paste fill and high density slurry fill. In: Bawden WF,
Archibald JF, editors. Proceedings of the international conference on innovative
mine design or the 21st century. Rotterdam, Kingston (Ontario, Canada): AA
Balkema; 1993. p. 111–8.
[4] Kesimal A, Yilmaz E, Ercikdi B, Alp I, Yumlu M, Ozdemir B. Laboratory testing of
cemented paste backfill. Turkish J Miner Eng 2002;41(4):25–32 [in Turkish].
[5] Regensburg B, Tacey W. Key business issues. In: Jewell RJ, Fourie AB, Lord ER,
editors. Paste and thickened tailings – a guide. Australian Centre for
Geomechanics; 2002. p. 9–22.
[6] Brackebusch FW. Basics of paste backfill systems. Miner Eng 1994;46(10):
1175–8.
[7] Kesimal A, Ercikdi B, Yilmaz E. The effect of desliming by sedimentation on
paste backfill performance. Miner Eng 2003;16:1009–11.
[8] Yilmaz E, Kesimal A, Ercikdi B. The factors affecting strength and stability of
paste backfill. Turkish J An Earth Sci 2003;28:155–69 [in Turkish].
[9] De Souza E, Archibald JF, Dirige APE. Economics and perspectives of
underground backfill practices in Canadian mining. 105th Annual general
meeting of the Canadian Institute of mining. Montreal (Quebec): In:
Metallurgy and Petroleum; 2003.
[10] Hassani FP, Ouellet J, Hossein M. Strength development in underground high
sulphate paste backfill operation. CIM Bull 2001;94(1050):57–62.
[11] Kesimal A, Yilmaz E, Ercikdi B. Evaluation of paste backfill test results obtained
from different size slumps with varying cement contents for sulphur rich mill
tailings. Cem Concr Res 2004;34(10):1817–22.
[12] Fall M, Benzaazoua M. Modelling the effect of sulphate on strength
development of paste backfill and binder mixture optimization. Cem Concr
Res 2004;32(5):301–14.
[13] Benzaazoua M, Fall M, Belem T. A contribution to understanding the hardening
process of cemented pastefill. Miner Eng 2004;17(2):141–52.
[14] Kesimal A, Yilmaz E, Ercikdi B, Deveci H, Alp I. Effect of properties of tailings
and binder on the short- and long-term strength and stability of cemented
paste backfill. Mater Lett 2005;59(28):3703–9.
[15] Neville AM. Properties of concrete. London: Prentice Hall; 2000.
[16] Benzaazoua M, Belem T, Bussiere B. Chemical factors that influence the
performance of mine sulphidic paste backfill. Cem Concr Res 2002;32(7):
1133–44.
[17] Irassar EF, Gonzales M, Rahhal V. Sulphate resistance of type V cements with
limestone filler and natural pozzolana. Cem Concr Res 2000;22:361–8.
[18] Dulaijan SUA, Maslehuddin M, Zahrani MMA, Sharif AM, Shameem M, Ibrahim
M. Sulfate resistance of plain and blended cements exposed to varying
concentrations of sodium sulphate. Cem Concr Comp 2003;25(4):429–37.
[19] Sahmaran M, Kasap O, Duru K, Yaman IO. Effects of mix composition and
water–cement ratio on the sulfate resistance of blended cements. Cem Concr
Comp 2007;29:159–67.
[20] Hooton RD, Emery JJ. Sulfate resistance of a Canadian slag cement. ACI Mater J
1990;87(6):547–55.
[21] Landriault DA. Backfill in underground mining. In: Hustrulid WA, Bullock RL,
editors. Underground mining methods engineering fundamentals and
international case studies. Littleton (Colorado): Society Mining, Metallurgy
and Exploration; 2001. p. 601–14.
[22] DIN 4030, Standard assessment of water, soil, and gases for their
aggressiveness to concrete; principles and limiting values. German Standard,
Construction Materials and Building; 2001.
[23] Bakharev T, Sanjayan JG, Cheng YB. Sulfate attack on alkali-activated slag
concrete. Cem Concr Res 2002;32(2):211–6.
[24] Ouellet S, Bussiere B, Mbonimpa M, Benzaazoua M, Aubertin M. Reactivity and
mineralogical evolution of an underground mine sulphidic cemented paste
backfill. Miner Eng 2006;19(5):407–19.
[25] Hossain KMA, Lachemi A. Performance of volcanic ash and pumice based
blended cement concrete in mixed sulfate environment. Cem Concr Res
2006;36:1123–33.
[26] Tikalsky PJ, Carrasquillo RL. Influence of fly ash on the sulfate resistance of
concrete. ACI Mater J 1992;89(1):69–75.
[27] Ercikdi B, Cihangir F, Kesimal A, Deveci H, Alp I. The effect of natural pozzolan
properties on the strength and stability of paste backfill. Turkish J An Earth Sci
2008;29(1):25–35 [in Turkish].
[28] Klein K, Simon D. Effect of specimen composition on the strength development
in cemented paste backfill. Can Geotech J 2006;43:310–24.
[29] Petrolito J, Anderson RM, Pigdon SP. A review of binder materials used in
stabilized backfills. CIM Bull 2005;98(1085):1–7.
[30] Bakker R. Application and advantages of blended cement concretes. QCL Group
Technical Note; 1999.
[31] Benzaazoua M, Fiset JF, Bussiere B, Villeneuve M, Plente B. Sludge recycling
within cemented paste backfill: study of the mechanical and leachability
properties. Miner Eng 2006;19(5):420–32.
[32] Sobolev K, Türker P, Soboleva S, Iscioglu G. Utilization of waste glass in ECO-
cement: strength properties and microstructural observations. Waste Manage
2007;27(7):971–6.
[33] Benzaazoua M, Ouellet J, Servant S, Newman P, Verburg R. Cementitious
backfill with high sulfur content: physical, chemical and mineralogical
characterization. Cem Concr Res 1999;29(5):719–25.
[34] Bing B, Cohen MD. Does gypsum formation during sulphate attack on concrete
lead to expansion. Cem Concr Res 2000;30(1):117–23.
[35] Mather B. A discussion of the paper ‘‘Theories of expansion in sulfoaluminate-
type expansive cements: school of thought” by Cohen MD. Cem Concr Res
1984;14(4):603–9.
[36] Mehta PK. Sulphate attack on concrete – a critical review. In: Skalny J, editor.
Materials science of concrete III. Ohio (USA): The American Ceramic Society;
1992. p. 105–30.
[37] Fall M, Benzaazoua M, Ouellet S. Experimental characterization of the
influence of tailings fineness and density on the quality of cemented paste
backfill. Miner Eng 2005;18(1):41–4.
[38] Shannag MJ, Shaia AH. Sulfate resistance of high-performance concrete. Cem
Concr Comp 2003;25(3):363–9.
[39] Torres SM, Sharp JH, Swamy RN, Lynsdale CJ, Huntley SA. Long term durability
of Portland-limestone cement mortars exposed to magnesium sulfate attack.
Cem Concr Comp 2003;25:947–54.
 Cement & Concrete Composites 31 (2009) 277–284

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Alkali-aggregate behaviour of alkali-activated slag mortars: Effect of aggregate type

F. Puertas *, M. Palacios, A. Gil-Maroto, T. Vázquez
Eduardo Torroja Institute for Construction Sciences (CSIC), Synthesis Department, 28033, Madrid, Spain

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

Article history: The alkali–silica reaction in waterglass-alkali-activated slag (waterglass-AAS) and ordinary Portland
Received 22 July 2008 cement (OPC) mortars was evaluated using three types of (siliceous and calcareous) aggregates. The tests
Received in revised form 16 January 2009 were conducted to the ASTM C1260-94 standard test method. The mortars were studied by volume sta-
Accepted 3 February 2009
bility, mechanical strength and Hg intrusion porosity. The ASR products were studied with XRD, FTIR and
Available online 14 February 2009
SEM/EDX techniques.
According to the results obtained, under the test conditions applied in this study, waterglass-AAS mor-
Keywords:
tars are stronger and more resistant to alkali-aggregate reactions than OPC mortars. When the mortars
Alkali-activated slag mortars
ASR
were made with a reactive siliceous aggregate, expansion was four times greater in the OPC than in
Porosity the AAS material. When a reactive calcareous (dolomite) aggregate was used, no expansion was detected
Mechanical strength in any of the mortars after 14 days, although the characterization results showed that the dolomite had
Microstructure reacted and calcareous-alkali products (brucite) had in fact formed in both mortars. These reactive pro-
cesses were more intense in OPC than in AAS mortars, probably due to the absence of portlandite in the
latter. When the calcareous aggregate was non-reactive, no expansions were observed in any of the mor-
tars, although a substantial rise was recorded in the mechanical strength of AAS mortars exposed to the
most aggressive conditions (1 M NaOH and 80 °C).
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction
The world-wide need to reduce the energy used and the green-
house gases emitted during cement manufacture has led to the
pursuit of more eco-efficient materials, such as alkali-activated
slag cements (AAS). These cements are obtained by mixing of blast
furnace slags (by-products from the iron industry) with alkaline
solutions. The behaviour and durability of AAS mortars and con-
cretes have been extensively studied [1–15]. These researches have
shown that such materials can develop comparable, and in some
cases greater, mechanical strength than mortars made with ordi-
nary Portland cement (OPC) [3,4]. They are also highly sulphate
and seawater resistant [7,8] and perform well when exposed to
acid media [9,10], chloride ions [11] and freeze-thaw cycles
[12,13]. All these characteristics and properties are essentially
associated with the special nature of the main hydration product
in these systems (a less basic and more highly polymerized CSH
gel than formed in OPC systems) [16–19] and the lower porosity
and permeability of the respective mortars and concretes [20].
Recent studies [21,22] have shown that while AAS mortars do
not generate Ca(OH)2 during hydration, they are more readily
and intensely carbonated than OPC mortars. In AAS systems the
CSH gel itself is carbonated directly, prompting decalcification
* Corresponding author. Fax: +34 913020700.
E-mail address: puertasf@ietcc.csic.es (F. Puertas).
and the formation of silica gels of varying composition and struc-
ture. These same studies have shown, however, that the mechani-
cal strength of AAS mortars is not adversely affected by such more
intense carbonation.
The high concentration of alkalis in alkaline cements (over 3%)
suggests that alkali-activated cements may be more vulnerable
to alkali-aggregate reaction (AAR)-induced expansion than OPC
mortars, although it is likewise true that in these alkaline systems
the Ca2+ ion concentration in the pore solution is lower, a deter-
mining factor in the alkali-aggregate reaction [23]. The studies
conducted to date have yielded varying and at times clearly contra-
dictory results. Metso [24] reported differences in the AAR-med-
iated expansive behaviour of two slags activated with the same
amount of alkali. Gifford and Gillot [25] found behavioural differ-
ences depending on the nature of the aggregate: in the presence
of reactive calcareous aggregate, AAS concretes were more vulner-
able to the reaction and more prone to form expansive products
than OPC concretes, whereas the findings were reversed when
reactive siliceous aggregate was used. The results obtained by
Bakharev et al. [26], in turn, showed that alkali-aggregate reac-
tion-induced expansion was greater in waterglass-AAS than in
OPC concretes, particularly in the first 5 months of the experiment,
when an alkali–siliceous gel was observed to form around the
aggregate. By contrast, Fernandez-Jiménez and Puertas [27] found
that NaOH-AAS mortars made with reactive opaline aggregate
exhibited a lower expansion rate than OPC mortars prepared and

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.008
278 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284

Table 1
Chemical analysis of cement and blast furnace slag (% by weight).
Binder CaO SiO2 Al2O3 MgO Fe2O3 SO
3 S2 Na2O Free CaO L.I.O. I.R. Blaine (m2/Kg)
Cement 65.03 18.61 4.46 2.20 4.39 2.94 – 0.12 0.64 0.43 0.18 527
Slag 41.00 35.34 13.65 4.11 0.39 0.06 1.91 0.01 – 2.72 0.64 325
Mineralogical composition of the cement
C3S C2S C3A C4AF K2SO4 CaSO4  2H2O CaSO4  1/2H2O CaSO4
Cement 70.8(4) 10.3(4) 2.7(1) 13.6(2) 0.6(1) – 1.1(1) 0.9(1)

stored under the same conditions. Similar results have been re-
ported by other authors [28] for waterglass-AAS concretes. Accord-
ing to the results of Krivenko [29], the Al2O3 content in
aluminosilicate constituent is a determining factor in the alkali-
aggregate reaction in alkaline cementitious systems.
These discrepancies may be due to the use of different materials
(nature of the slag, type and concentration of the alkaline activator,
nature and percentage of the reactive aggregate) and the alkali-
aggregate reaction test conditions and standards applied.
In light of the need, for a more rigorous procedure for establish-
ing the behaviour of AAS mortars and concretes made from aggre-
gate with varying degrees of reactivity, the present study was
undertaken to ascertain the behaviour of waterglass-AAS mortars
prepared with calcareous and siliceous aggregate under the condi-
tions laid down in ASTM standard C1260-94.
2. Experimental
2.1. Materials
The chemical composition of CEM I 52.5 N cement and the vit-
reous blast furnace slag used in the study is given in Table 1. The
table also shows the quantitative mineralogical composition, pur-
suant to Rietveld refinement of XRD patterns [30]. The blast fur-
nace slag had a vitreous content of 99% [31].
The slag was alkali-activated with a waterglass solution
(Na2O  nSiO2  mH2O + NaOH) containing 4% Na2O by slag mass
and a SiO2/Na2O ratio of 1.08.
The physical and chemical characteristics of the three types of
aggregate used to prepare the mortars, two calcareous and one sili-
ceous, are summarized in Table 2. Aggregate water absorption and
density were determined as specified in Spanish standard UNE
1097-6. The potential reactivity of the siliceous aggregate was
evaluated to Spanish standard UNE 146507 EX. XRD, FTIR and thin
section microscopic techniques were used to determine the mine-
ralogy of the three types of aggregate.
2.2. Mortar preparation
Table 3 lists all the mortars, which were prepared as specified in
European standard EN-UNE 196-1. The alkali-aggregate reaction
was evaluated according to the method described in ASTM stan-
dard C1260-94. Further to this standard, the aggregate must have
a specific particle size distribution (see Table 4). While all the mor-
tars prepared had the particle size distribution called for in the
standard, the liquid/cement ratio in the AAS mortars had to be
raised to 0.52, for the mortars were unworkable with less. A
water/cement ratio of 0.47 was maintained for the OPC mortars,
however.
Mortar specimens measuring 2.5  2.5  28.7 cm were pre-
pared for each binder/aggregate combination, with a binder/aggre-
gate ratio of 1/2.25.The specimens were cured in a moist cabinet
(99% R.H., 21 ± 2 °C) for the first 24 h and then de-moulded. The
de-moulded samples were submerged in water and kept in an oven
at 80 °C for 24 h, after which the first shrinkage measurement was
taken. They were subsequently submerged in deionized water at
room temperature (control mortars) or in 1 M NaOH at 80 °C and
stored under these conditions for 14 days, according to the ASTM
standard (see storage conditions on Table 3). Previous studies
[27] concluded that for determining the potential expansion due
to alkali-aggregate reaction in AAS mortars, 14 days is not a suit-
able period and a longer period is required. Consequently in the
present study the AAS mortars were additionally stored in the dif-
ferent media during 4 months.

Table 2
Aggregate characteristics and composition.

Aggregate Density (g/cm3) Absorp. coefficient (Aq) Potential reactivity Mineralogical composition
Siliceous (S) 2.64 0.7 Negative Quartz, calcite and albite-type feldspars
Non-reactive (C) calcareous 2.65 0.5 – Calcite and quartz traces
Reactive calcareous (CR) 2.67 0.4 – Calcite, dolomite and quartz traces

Table 3
Mortars prepared.

Mortar Binder Aggregate Mixing liquid Liquid/solid Storage medium

OPC-S1 OPC S Water 0.47 In water: 99% R.H., T 21 ± °C
OPC-S2 OPC S Water 0.47 1 M NaOH, T 80 °C
OPC-C1 OPC C Water 0.47 In water: 99% R.H., T 21 ± °C
OPC-C2 OPC C Water 0.47 1 M NaOH, T 80 °C
OPC-CR1 OPC CR Water 0.47 In water: 99% R.H., T 21 ± °C
OPC-CR2 OPC CR Water 0.47 1 M NaOH, T 80 °C
AAS-S1 AAS S Wg.4% Na2O 0.52 In water: 99% R.H., T 21 ± °C
AAS-S2 AAS S Wg.4% Na2O 0.52 1 M NaOH, T 80 °C
AAS-C1 AAS C Wg.4% Na2O 0.52 In water: 99% R.H., T 21 ± °C
AAS-C2 AAS C Wg.4% Na2O 0.52 1 M NaOH, T 80 °C
AAS-CR1 AAS CR Wg.4% Na2O 0.52 In water: 99% R.H., T 21 ± °C
AAS-CR2 AAS CR Wg.4% Na2O 0.52 1 M NaOH, T 80 °C
 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284 279

Table 4 are regarded to have undergone an alkali-aggregate reaction.

Grading criteria. Fig. 1 shows the variation in volume recorded for the mortar spec-
Passing Retained on Mass % imens stored in water and in the 1 M NaOH solution at 80 °C up to
4.75 mm 2.36 mm 10 4 months. An analysis of the data in these figures shows that both
2.36 mm 1.18 mm 25 the OPC and AAS mortars made with siliceous aggregate expanded,
1.18 mm 600 lm 25 while no change in volume was observed when calcareous aggre-
600 lm 300 lm 25 gate was used.
600 lm 150 mm 15
In mortars containing siliceous aggregate, the nature of the bin-
der (OPC and AAS) also appeared to affect expansion intensity. In
At the same time, 3  3  3-cm OPC and AAS mortar specimens OPC mortars, volume changes were recorded in the first 72 h of
were prepared and subjected to the alkali-aggregate test described storage in the alkaline medium, with 14 days expansion values
above. Compressive strength was determined on these mortars. on the order of 0.36%. In mortars in which the binder was AAS,
expansion also began after around 72 h of storage, but total 14
2.3. Tests conducted days expansion was just 0.1%. Similar results had been obtained
in previous studies [27], which showed that the alkali-aggregate
The tests conducted were: reaction was greater in OPC than NaOH-AAS siliceous aggregate
mortars when both were subjected to identical test conditions.
2.3.1. Volume stability The primary difference between that research and the present
Volume stability measurements were taken throughout the 14 study was the nature of the alkaline activator, which was water-
days and 4 months-day trial according to UNE 80-112-89 standard. glass in this case.
The mortar properties were studied after 4 months of storage
2.3.2. Compressive strength conditions. Only with siliceous aggregate and accelerated condi-
Specimen compressive strength according to EN 196-1 standard tions the expansion of the mortars was observed, although an
was found after 14 days and 4 months of storage in the media slight increase of the volume were observed in AAS mortars with
described.

2.3.3. Microstructural and mineralogical characterization 140

COMPRESSIVE STRENGTH (MPa)

Total porosity and pore size distribution in tested 14 days spec-

imens were found by Hg intrusion porosimetry (MIP) using a 120
Micromeritics 9320 porosimeter. Also mortar samples were stud-
100
ied under SEM/EDX using a JOEL 5400 microscope attached to an
Oxford-Link ISIS EDX microanalysis unit. The mineralogical study 80
was conducted on binder-enriched samples, tested 14 days sam-
ples. These samples were studied by XRD using a Philips PW- 60
1730 diffractometer and by FTIR using a ATIMATTSON, Genesis
40
FTIR-TM spectrometer.
20
3. Results and discussion
0
3.1. Volume stability OPC-S1 OPC-S2 OPC-C1 OPC-C2 OPC-CR1 OPC-CR2

14 days 4 months
According to ASTM standard C1260, mortars that expand by
more than 0.1% when stored for 14 days in 1 M NaOH at 80 °C Fig. 2. Mechanical strength in OPC mortars.

OPC MORTARS AAS MORTARS

1.2 1.2
1.1 1.1
1.0 1.0 AAS-S1
0.9 0.9 AAS-S2
0.8 AAS-C1
0.8
AAS-C2
% Expansion

0.7
% Expansion

OPC-S1 0.7 AAS-CR1

0.6 OPC-S2 0.6 AAS-CR2
0.5
OPC-C1
0.5
OPC-C2
0.4 OPC-CR1 0.4
0.3 OPC-CR2 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Days Days

Fig. 1. Expansion curves.

280 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284

both calcareous submerged to alkaline medium; but in any case Also at this time (4 months) a slight decrease on mechanical
those expansion values were lower than 0.1%. strengths was observed in the OPC mortar with reactive calcareous
aggregate in aggressive conditions.
3.2. Compressive strengths All the AAS mortars (see Fig. 3) stored during 14 days in the
alkaline medium exhibited much higher strength values than the
The mechanical behaviour of the 14 days and 4 months speci- specimens stored in water, with non-dolomite aggregate AAS mor-
mens stored in water and the alkaline medium are shown in Figs. tars (AAS-C2) reaching values of 124 MPa, i.e., 29% higher than the
2 and 3. water-stored samples. Only the AAS mortar with siliceous aggre-
According to these findings, after 14 days of storage, the gate and stored in alkaline medium during 4 months shows a de-
strength of both OPC and AAS mortars made with siliceous and cal- crease on the compressive strength. In the rest of the mortars an
careous aggregate increased after the accelerated test, except in increase of the strengths were observed.
OPC mortars with siliceous aggregate. The results obtained after
4 months of storage show a decrease (of about 35%) in the com- 3.3. Microstructural and mineralogical characterization
pressive strength in these same mortars with siliceous aggregate.
This mechanical behaviour is explained, in part, by the porosity
and pore size distribution results obtained for the mortars. Fig. 4
140 shows the MIP findings for total porosity and pore size distribution
COMPRESSIVE STRENGTH (MPa)

in OPC mortars after 14 days of accelerated test, while Fig. 5 gives

120 the results for the AAS mortars.
The OPC mortars stored in an alkaline medium had from 10% to
100
24% greater total porosity than the mortars stored in water, with
80 the highest values found for the materials made with siliceous
aggregate. As a rule, the pore size in the mortars stored under
60
the most aggressive conditions was from 0.1 to 0.01 lm greater
40 than in the controls. These mortars also exhibited an increase in
the proportion of larger pores (10–0.1 lm). Such variations in
20
porosity did not appear to suffice to induce a decline in strength
0 values.
AAS-S1 AAS-S2 AAS-C1 AAS-C2 AAS-CR1 AAS-CR2 Total porosity was lower and meso-porosity greater in the AAS
14 days 4 months
than the OPC mortars, a finding that concurs with reports by other
authors [32]. A sharp decline in total porosity was observed in this
Fig. 3. Mechanical strength in AAS mortars. study in all the mortars submerged in the alkaline solution, irre-

16 10

14 8
% TOTAL POROSITY

% POROSITY

12
6
10
4
8
2
6

4 0
> 100 microns 100 - 10 10 - 1,0 1,0 - 0,1 0,1 - 0,01 < 0,01
2 microns microns microns microns microns
0
OPC-S1 OPC-S2 OPC-C1 OPC-C2 OPC-CR1 OPC-CR2 OPC-S1 OPC-S2 OPC-C1 OPC-C2 OPC-CR1 OPC-CR2

Fig. 4. Total porosity and pore size distribution in OPC mortars after 14 days of accelerated test.

16 10

14
8
% TOTAL POROSITY

% POROSITY

12
6
10
4
8

6 2

4 0

2 > 100 100 - 10 10 - 1,0 1,0 - 0,1 0,1 - 0,01 < 0,01
microns microns microns microns microns microns
0
AAS-S1 AAS-S2 AAS-C1 AAS-C2 AAS-CR1 AAS-CR2
AAS-S1 AAS-S2 AAS-C1 AAS-C2 AAS-CR1 AAS-CR2

Fig. 5. Total porosity and pore size distribution in AAS mortars after 14 days of accelerated test.
 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284
Fig. 6. Attack on calcareous aggregate (OPC-C2).
Fig. 7. Attack on calcareous aggregate (AAS-C2).
Fig. 8. SEM/BSE image and analysis of AAS-C2 mortar after 14 days of accelerated test.
281
spective of the type of aggregate, but particularly where non-dolo-
mite or non-reactive calcareous aggregate was used, which would
explain the high strength values obtained for these mortars. The
most prominent effect on pore size distribution was the decline
in the proportion of smaller pores in the mortar stored in an
aggressive medium, although that decline was due, in part, to the
lower total porosity in such mortars.
Porosity was consistently higher in OPC than AAS mortars,
while this parameter declined more steeply in the AAS mortars
stored in an alkaline solution. The explanation may be that the
alkaline medium and the high storage temperature (80 °C) would
continue to stimulate the reaction between any remaining anhy-
drous slag and the NaOH solution, generating new reaction prod-
ucts that would, in turn, reduce porosity and raise strength.
The mortar volume stability, mechanical strength and porosity
results obtained showed that under equally aggressive conditions,
strength and stability raised more in the AAS than in the OPC
mortars.
The microstructural study conducted by SEM/EDX on the non-
reactive calcareous aggregate OPC and AAS mortars showed that
the aggregate surface was attacked in both cases when the mortars
were stored in aggressive environments. The surface attack on the
calcareous aggregate in mortars OPC-C2 and AAS-C2 is depicted in
Figs. 6 and 7, respectively. When the latter mortar was studied
with SEM/BSE techniques, a whitish layer was observed around
the calcareous aggregate that appeared to be spreading into the
matrix. The analysis of this layer showed a higher Ca content than
in other inner-more areas of the matrix (see Fig. 8). These results
are indicative of dissolution and partial attack on the aggregate;
the subsequent diffusion of calcium ions towards the matrix would
favour the formation of a more cohesive interface and explain the
low total porosity and the high strength exhibited by these
mortars.
The SEM/EDX study of the siliceous aggregate OPC and AAS
mortar microstructure confirmed that the aggregate was attacked
under the aggressive test conditions. In the SEM image of mortar
OPC-S2 in Fig. 9, a characteristic morphology of ASR, in a pore,
can be observed. The morphology and composition of an ASR
282 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284

Fig. 9. Mortar OPC-S2 after 14 days of accelerated test. Morphology of ASR in a pore.
products observed in the OPC-S2 mortar are shown in Fig. 10. The
formation of these expansive reaction products induces a decrease
of the total porosity and the compressive strength as it has been
discussed above. On the contrary, the morphology of the ASR
products in the AAS mortars differed slightly from the findings de-
scribed for the OPC mortars. Some of the morphologies and compo-
sitions of these ASR products observed in sample AAS-S2, are given
in Fig. 11. In this case, the formation of the expansive products
does not have a negative effect on the compressive strength (see
Fig. 3). In both mortars (OPC-S2 and AAS-S2) the reactivity of sili-
ceous aggregate is due mainly to the presence of feldspars (see Ta-
ble 2) in this aggregate. These feldspars suffer hydrolysis in the
accelerated conditions due to an exchange between the metal ions
of the feldspars and the ions present in the solution [33]. This
hydrolysis process is increased by the high pH of the solution
and the temperature.
The OPC and AAS mortars with dolomite calcareous aggregate
exhibited no variations in volume or adverse effects on their
strength; nonetheless, the mineralogical characterization study
conducted on the two mortars showed that alkali-calcareous
aggregate reactions did take place after 14 days of test and that
these reactions depended on the nature of the binder. The diffrac-
tograms for binder-enriched samples OPC-CR1 and OPC-CR2 ob-
tained after the accelerated test are reproduced in Fig. 12a. Note
that the diffraction lines assigned to dolomite at d (Å) 2.89 and
2.19 and the ettringite phase at d (Å) 9.66 and 5.60 disappeared
in the sample stored in 1 N NaOH at 80 °C (OPC-CR2). The low

Fig. 10. Morphologies and compositions of ASR products in OPC-S2 mortar after 14 days of accelerated test.

Fig. 11. Morphologies and compositions of ASR products in AAS-S2 mortar after 14 days of accelerated test.
 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284 283

a C b OPC-CR1

3640
3450
C OPC-CR2
OPC-CR1 1115
D PC C C C C 710
P C P A D P A C 3690 988 518
E
3643 3450 873
C
OPC-CR2 C AAS-CR1
710
P C C C C 992
P
C P A C P AC 1432
874 670
B B
3450
710
C
CC AAS-CR2 1432 874
AAS-CR1 D C C C C 995
C D C 670
3690
3450 1432 710

C C 980 874
AAS-CR2 C CC C
C D C
B B D C
1432

10 20 30 40 50 60 4000 3500 3000 1500 1000 500

-1
2-theta Wavenumber (cm )

P = portlandite, C = calcite, D= dolomite, A =alite, E=ettringite, B=Brucite

Fig. 12. (a) XRD and (b) FTIR of OPC and AAS mortars after 14 days of accelerated test.

intensity signals at d (Å) 4.76 and 2.36 on this diffractogram were
attributed to Mg(OH)2. The FTIR study also detected an absorption
peak at 3690 cm1, attributed to the hydroxide O-H stretching
vibrations in brucite (Mg(OH)2) (see Fig. 12b). In addition, these
spectra confirmed the absence of ettringite in the OPC-CR2 mortars,
for no present of S–O vibration bands of SO2 4 groups at around
1115 cm2,that identify that hydrated phase.
Fig. 12a shows the diffractrograms for binder-enriched samples
AAS-CR1 and AAS-CR2. In this case the dolomite signals did not dis-
appear from sample AAS-CR2 (stored in 1 M NaOH at 80 °C); nor
did the XRD study clearly identify the formation of brucite
(Mg(OH)2). The shoulder at around 3690 cm1 on the FTIR spec-
trum (see Fig. 12b), however, may indicate the formation of this
phase, albeit in a lower proportion than observed in the OPC
samples.
The de-dolomitization of dolomite aggregate as a result of the
reaction with alkaline solutions leads to the neo-formation of bru-
cite (Mg(OH)2) and the regeneration of the alkaline hydroxide, as
shown below [34]:
CaMgðCO3 Þ2 þ 2NaOHƒƒ!MgðOHÞ2 þ CaCO3 þ Na2 CO3
Na2 CO3 þ CaðOHÞ2  ƒƒ!CaCO3 þ 2NaOH
The findings of the present study showed that this process was
more intense in OPC than in AAS mortars, probably due to the pres-
ence of Ca(OH)2 in the former and its absence in the latter.
Ettringite decomposed in both types of mortar after storage in
1 M NaOH at 80 °C. The explanation may be found in the instability
of this mineral at temperatures of over 70 °C [35] in the presence of
a high concentration of alkalis. According to Berger [36], ettringite
decomposes in solutions with over 2.7% by weight of Me2O
(Me = Na, K, Li).
4. Conclusions
The OPC and waterglass-AAS mortars containing siliceous and
calcareous aggregate behaved differently under the test conditions
applied in this study (based on ASTM standard C1260-94):
(a) When a non-reactive calcareous aggregate was used, no
variations in volume were observed in any of the mortars,
although a substantial rise was recorded in the mechanical
strength of AAS mortars exposed to the most aggressive con-
ditions (1 M NaOH and 80 °C). Ca2+ ion diffusion from the
aggregate toward the matrix was also observed, with the
formation of highly cohesive aggregate-paste interfaces.
(b) Although both the OPC and the AAS specimens made with
reactive siliceous aggregate expanded under the most
aggressive conditions (submerged in 1 M NaOH at 80 °C),
expansion was nearly four times larger in the OPC than the
AAS mortars.
(c) When a calcareous (dolomite) reactive aggregate was used,
none of the mortars expanded after 14 days; nonetheless,
the results of the characterization study conducted showed
that the dolomite reacted in both mortars and that calcare-
ous-alkali products (brucite) did in fact form. These reactive
processes were more intense in OPC than in AAS mortars,
probably due to the absence of portlandite in the latter.
(d) Unlike the OPC mortars, under the most aggressive condi-
tions for the aggregate the AAS mortars exhibited a decline
in total porosity and a very significant increase in mechani-
cal strength. These effects were probably due to the reaction
between the remaining anhydrous slag and the alkaline
solutions in the medium, with the concomitant formation
of new hydration products.
In summary, under the test conditions applied in this study,
waterglass-AAS mortars proved to be higher in strength and more
resistant to the alkali-aggregate reaction than their OPC counter-
parts. Also, AAS mortars are more vulnerable to alkali-aggregate
reaction and expansion with siliceous aggregate than calcareous
aggregate.
Acknowledgements
This research was funded by the Spanish Ministry of Education
and Science under Project No. BIA2007-61380. The authors thanks
284 F. Puertas et al. / Cement & Concrete Composites 31 (2009) 277–284
to M.M. Alonso and J. Llueca for their support in tests conducted for
this study. M. Palacios worked under a postdoctoral contract
awarded by the CSIC.
References
[1] Glukhovsky VD, Rostovskaja GS, Rumyna GV. High strength slag-alkaline
cements. In: 7th international congress chemistry cement (Paris), vol. 3; 1980.
p. 164–8.
[2] Krivenko PV. Alkaline cements. IX Intern Congr Chem Cem 1992;4:482–8 [New
Delhi].
[3] Byfors K, Klingstedt G, Lehtonen V, Pyy H, Romben L. Durability of concrete
made with alkali-activate slag. In: Proc 3rd int conf on fly ash, silica fume, slag
and natural pozzolans in concrete, Trondheim, vol. 2; 1989. p. 1547–63
[SP114-70].
[4] Fernández-Jiménez A, Puertas F, Palomo JG. Alkali-activated slag mortars:
mechanical strength behaviour. Cem Concr Res 1999;29:593–604.
[5] Wang SD, Pu XC, Scrivener KL, Pratt PL. Alkali-activated slag cement and
concrete: a review of properties and problems. Adv Cem Res
1995;7(27):93–102.
[6] Alcalde JS, Alcocel EG, Puertas F, Lapuente R, Garcés P. Carbon fibre-reinforced
in alkali activated slag mortars. Mater Construcc 2007;57(288):33–48.
[7] Bakharev T, Sanjayan JG, Cheng Y-B. Sulfate attack on alkali-activated slag
concrete. Cem Concr Res 2002;32:211–6.
[8] Puertas F, de Gutierrez R, Fernández-Jiménez A, Delvasto S, Maldonado J.
Alkaline cement mortars. Chemical resistance to sulphate and seawater attack.
Mater Construcc 2002;52(267):55–71.
[9] Pu Xincheng, Yang Changhui, Liu Fan. Studies on resistance of alkali activated
slag concrete to acid attack. In: 2nd international conference alkaline cements
and concretes; 1999. p. 717–22.
[10] Bakharev T, Sanjayan JG, Cheng Y-B. Resistance of alkali-activated slag
concrete to acid attack. Cem Concr Res 2003;33:1607–11.
[11] Shi C. Corrosion resistance of alkali-activated slag cement. Adv Cem Res
2003;15(2):77–81.
[12] Puertas F, Amat T, Fernández-Jiménez A, Vázquez T. Mechanical and durable
behaviour of alkaline cement mortars reinforced with polypropylene fibres.
Cem Concr Res 2003;33:2031–6.
[13] Xu Bin, Pu Xincheng. Study on durability of solid alkaline AAS cement. In:
Second international conference alkaline cements and concretes; 1999. p.
101–8.
[14] Tailing B, Brandstetr J. Present state and future of alkali-activated slag
concretes. In: Proc 3rd int conf on fly ash, silica fume, slag and natural
pozzolans in concrete, Trondheim, vol. 2; 1989. p. 1519–46 [SP114-74].
[15] Mejía de Gutierrez R, Maldonado J, Gutiérrez C. Performance of alkaline
activated slag at high temperatures. Mater Construcc 2004;54(276):87–94.
[16] Wang SD, Scrivener K. Hydration products of alkali activated slag. Cem Concr
Res 1995;25:561–71.
[17] Escalante-García JI, Fuentes AF, Gorokovsky A, Fraire-Luna PE, Mendoza-
Suarez G. Hydration products and reactivity of blast-furnace slag activated by
various alkalis. J Am Ceram Soc 2003;86(12):48–53.
[18] Fernández-Jiménez A, Puertas F, Sobrados I, Sanz J. Structure of calcium silicate
hydrates formed in alkaline activated slag. Influence of the type of alkaline
activator. J Am Ceram Soc 2003;86(8):1389–94.
[19] Puertas F, Fernández-Jiménez A, Blanco-Varela MT. Pore solution in alkali-
activated slag cement pastes. Relation to the composition and structure of
calcium silicate hydrate. Cem Concr Res 2004;34:139–48.
[20] Shi C. Strength, pore structure and permeability of alkali-activated slag
mortars. Cem Concr Res 1996;26:1789–99.
[21] Puertas F, Palacios M, Vázquez T. Carbonation process of alkali-activated slag
mortars. J Mater Sci 2006;41:3071–82.
[22] Palacios M, Puertas F. Effect of carbonation on alkali-activated slag paste. J Am
Ceram Soc 2006;89(10):3211–21.
[23] Tsuneki Ichikawa, Masazumi Miura. Modified model of alkali–silica reaction.
Cem Concr Res 2007;37:1291–7.
[24] Metso J. The alkali reaction of alkali-activated finnish blast furnace slag. Silic
Ind 1982;47(4–5):123–7.
[25] Gifford PM, Gillot JE. Alkali–silica reaction (ASR) and alkali–carbonate reaction
(ACR) in activated blast furnace slag cement (ABFSC) concrete. Cem Concr Res
1996;26:21–6.
[26] Bakharev T, Sanjayan JG, Cheng YB. Resistance of alkali-activated slag concrete
to alkali-aggregate reaction. Cem Concr Res 2001;31:331–4.
[27] Fernandez-Jiménez A, Puertas F. The alkali–silica reaction in alkali-activated
slag mortars with reactive aggregate. Cem Concr Res 2002;32:1019–24.
[28] Sud Al-Otaibi. Durability of concrete incorporating GGBS activated by water-
glass. Constr Build Mater 2008;22:2059–67.
[29] Krivenko P. Alkaline cements, concretes and structures: 50 years of theory and
practice. In: Proceedings of alkali activated materials. Research, production
and utilization; 2007.
[30] De la Torre AG, Aranda MAG. Accuracy in Rietveld quantitative phase analysis
of Portland cements. J Appl Crystallogr 2003;36:1169–76.
[31] Hooton RD, Emery JJ. Glass content determination and strength development
predictions for vitrified blast furnace slag. In: 5th int conf fly ash, silica fume,
slag and other mineral by products in concrete. Montebello, Quebec, Canada;
1983. p. 943–62 [SP 79].
[32] Palacios M. Efecto de aditivos orgánicos en las propiedades de los cementos y
morteros de escoria activada alcalinamente. PhD, Universidad Autónoma de
Madrid; 2006.
[33] Loughnan FC. Chemical weathering of the silicate minerals. New
York: American Elsevier Publishing Company, Inc.; 1969.
[34] Hadley DW. Alkali reactivity of dolomite carbonate rocks. Highway Res Board
1961;40:462–9.
[35] Skalny J, Johansen V, Thaulow N, Palomo A. DEF: as a form of sulphate attack.
Mater Construcc 1996;46:5–29.
[36] Berger AS. On metaestable equilibrium of calcium hydroaluminates in solution
of hydroxides of alkali metals. In: 6th int cong chem cem, vol. III; 1974. p. 1–12
[Moscow].
 Cement & Concrete Composites 31 (2009) 285–289

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Influence of internal curing using lightweight aggregates on interfacial

transition zone percolation and chloride ingress in mortars
Dale P. Bentz *
Building and Fire Research Laboratory, National Institute of Standards and Technology, 100 Bureau Drive Stop 8615, Gaithersburg, MD 20899-8615, USA

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

Article history: The microstructure of the interfacial transition zone (ITZ) between cement paste and aggregate depends
Received 25 August 2008 strongly on the nature of the aggregate, specifically its porosity and water absorption. Lightweight aggre-
Received in revised form 5 March 2009 gates (LWA) with a porous surface layer have been noted to produce a dense ITZ microstructure that is
Accepted 11 March 2009
equivalent to that of the bulk cement paste, as opposed to the more porous ITZ regions that typically sur-
Available online 21 March 2009
round normal weight aggregates. This ITZ microstructure can have a large influence on diffusive transport
into a concrete, especially if the individual ITZ regions are percolated (connected) across the three-
Keywords:
dimensional microstructure. In this paper, the substitution of LWA sand for a portion of the normal
Building technology
Diffusion
weight sand to provide internal curing (IC) for a mortar is examined with respect to its influence on
Interfacial transition zone ITZ percolation and chloride ingress. Experimental measurements of chloride ion penetration depths
Internal curing are combined with computer modeling of the ITZ percolation and random walk diffusion simulations
Lightweight aggregate to determine the magnitude of the reduced diffusivity provided in a mortar with IC vs. one with only nor-
Microstructure mal weight sand. In this study, for a mixture of sands that is 31% LWA and 69% normal weight sand by
Percolation volume, the chloride ion diffusivity is estimated to be reduced by 25% or more, based on the measured
penetration depths.
Published by Elsevier Ltd.

1. Introduction The formation of these ITZ around normal weight aggregates

Lightweight aggregate (LWA) concretes have generally exhib-
ited excellent performance under severe weathering conditions
[1]. One of the reasons cited for this is the high integrity ‘‘contact
zone” formed between the LWA and the neighboring hydrating ce-
ment paste [1]. More recently, the terms interfacial zone and inter-
facial transition zone (ITZ) have been adopted to replace contact
zone. For normal weight aggregates, due to the inherent size differ-
ences between cement and aggregates, a ‘‘wall effect” exists, so
that there is a deficiency of cement particles near the aggregate
surface relative to their concentration in the bulk (non ITZ) cement
paste. Direct microstructural examinations by scanning electron
microscopy (SEM) have revealed that for LWA with a porous outer
layer, this wall effect does not exist and a nearly continuous
uniform microstructure of hydration products abuts and partially
penetrates the LWA [2,3]. As an example, Fig. 1 shows SEM micro-
graphs of blended cement mortars with and without internal cur-
ing [4]. A nearly continuous microstructure is observed near the
porous LWA particle, along with the ability for the cement hydra-
tion products to penetrate into the LWA surface pores and
irregularities.
* Tel.: +1 301 975 5865; fax: +1 301 990 6891.
E-mail address: dale.bentz@nist.gov
will also be influenced by curing conditions. Due to a wall effect
that causes inefficient ‘‘packing” of the cement particles near the
aggregates, the ITZ regions will initially have a higher water-to-ce-
ment ratio (w/c) and a larger interparticle spacing than the bulk ce-
ment paste [5]. If sufficient curing water is not readily available at
early ages, the concrete will undergo self-desiccation, with the
bulk cement paste regions imbibing water from the largest pores
within the ITZ, resulting in less hydration, greater porosity, and lar-
ger empty pores being present in the ITZ (bottom micrograph in
Fig. 1) [6]. If such a concrete later resaturates during its environ-
mental exposure, such porous ITZ regions would likely provide less
resistance to ionic and fluid transport.
When each normal weight aggregate in concrete is surrounded
by such a porous ITZ, their percolation or connectivity across the
three-dimensional microstructure may become an issue for trans-
port and durability [7]. This percolation has been extensively
examined using a hard core-soft shell (HCSS) model developed at
the National Institute of Standards and Technology (NIST) [7,8],
in which the aggregates are considered as hard core (impenetrable)
spherical particles and the ITZ as surrounding concentric soft (pen-
etrable) shells. The HCSS model has been extended to examine
spalling of high-performance concretes containing polymeric fibers
[9], to adapt the protected paste volume concept to internal curing
[10], and most recently to document the influence of water–ce-
ment ratio (w/c) and cement particle size distribution on particle

0958-9465/$ - see front matter Published by Elsevier Ltd.

doi:10.1016/j.cemconcomp.2009.03.001
286 D.P. Bentz / Cement & Concrete Composites 31 (2009) 285–289
Fig. 1. SEM micrographs of blended (20% slag) cement mortars with w/c = 0.30,
cured under sealed conditions for 120 d [4]. Top micrograph is for a mortar with
internal curing (0.08 extra units of water supplied by LWA sand) while bottom
micrograph is for a mortar with normal weight sand only. Dark regions within LWA
sand in top micrograph correspond to pores.
spacing in fresh cement paste [11]. In this paper, the model will be
further adapted to consider a mixture of normal weight sand par-
ticles, with an accompanying ITZ, and LWA where no distinct ITZ is
present, in a fashion similar to that previously employed for
macro-defect-free cements containing both inert and reactive par-
ticles [12]. It should be noted that controversy exists in the litera-
ture regarding the existence of ITZ percolation in concrete [13],
despite microstructural examinations that have validated the exis-
tence of such percolated pathways in concrete [14].
Previously, experimental and computer modeling efforts have
quantified diffusion rates of ions diffusing within these ITZ regions
relative to those diffusing in the bulk paste [15–18]. For example,
Bretton et al. [15] conducted model experiments using a cylindrical
aggregate surrounded by cement paste and concluded that the ITZ
exhibits a chloride ion diffusion coefficient that is 12–15 times that
of bulk paste, for a w/c = 0.5 cement paste cured for 10 d and an as-
sumed ITZ thickness of 100 lm. Bourdette [16] has estimated a
lower ratio of ITZ to bulk paste diffusivity of 3 for a w/c = 0.4 mor-
tar cured for 3 months, assuming an ITZ thickness of 120 lm. Con-
versely, Otsuki et al. [17] have projected that the diffusion
coefficient for the ITZ could be over 100 times that of the bulk
paste for concretes with w/c ranging between 0.4 and 0.7, assum-
ing an ITZ thickness that is a function of aggregate size and varies
between zero and about 80 lm. Based on a multi-scale microstruc-
tural model, Bentz [18] arrived at diffusivity ratios ranging from
0.7 to about 21 for an assumed ITZ thickness of 15 lm and degrees
of hydration ranging from 0.62 to 0.88. In that study, values of less
than one, indicating that the ITZ is more resistant to diffusion than
the bulk paste, were obtained only for low w/c = 0.3 concretes con-
taining silica fume. For these concretes, the very small silica fume
particles can concentrate in the ITZ regions leading to an ITZ micro-
structure that is ultimately denser than that of the bulk paste
[5,18]. The larger values near 20 were observed for w/c = 0.5 con-
cretes and are in reasonable agreement with those of Bretton
et al. discussed above. Based on these results, the general consen-
sus would be that for many conventional concretes, the ITZs could
indeed provide a preferential path for the ingress of deleterious
species such as chloride ions.
As a further concrete example, the lack of distinct and more
porous ITZ regions in lightweight aggregate concretes could con-
tribute to Thomas’ observation of apparent diffusion coefficients
that are reduced by as much as 70% due to the incorporation of
LWA [19]. Furthermore, his results indicated that the greatest
reductions in concrete diffusivity were achieved when both the
coarse and fine aggregates were totally replaced with their light-
weight counterparts, as opposed to a total replacement of the
coarse aggregates combined with a normal weight sand. While it
is only natural to question whether porous LWA might not be ex-
pected to actually increase transport rates, one can observe in Fig. 1
that the pores contained in the LWA appear to be discrete and may
not be percolated across the aggregate particle. In fact, this lack of
three-dimensional percolation has been confirmed by X-ray micro-
tomography studies at NIST for several commercially available
LWA. This discontinuity is also in agreement with the experimental
observation of Zhang and Gjorv that the permeability of high-
strength lightweight concrete is more dependent on the properties
of the cement paste than the porosity of the LWA [20]. Recently,
Pyc et al. have performed mass measurements that suggest that
once the pores in LWA empty while supplying water to the hydrat-
ing cement paste during curing, they are not subsequently resatu-
rated, even upon complete immersion of the specimen [21]. If
these pores in the LWA remain empty, ionic diffusion through
them is a moot point and instead they may potentially function
as part of an effective air void system to provide freeze thaw resis-
tance to the concrete [10]. Of course, it is critical that the pre-wet-
ted LWA be provided an opportunity to empty prior to exposure to
chloride ions. In the current study, this was achieved by employing
sealed curing for 7 d or 28 d for the specimens containing pre-wet-
ted LWA, prior to immersion in Cl solutions.
In typical internal curing (IC) applications, only a fraction of the
normal weight aggregates are replaced by LWA [22]. Such a
replacement can still substantially reduce the total volume of ITZ
paste and could also have a significant influence on its three-
dimensional percolation and chloride ion transport, as illustrated
Fig. 2. Comparison of model mortars with normal weight sand particles only (left)
with their surrounding ITZs and with a 50:50 blend (volume basis) of sand and LWA
(right). Both the volume fraction of ITZ (grey) paste and its percolation are reduced
by the incorporation of the LWA.
 D.P. Bentz / Cement & Concrete Composites 31 (2009) 285–289 287

schematically in two dimensions in Fig. 2. The impetus for the cur-
rent study was an observation made during a recent study on the
effectiveness of viscosity modifiers in reducing the ingress of chlo-
ride ions into mortars (VERDiCT: Viscosity Enhancers Reducing Dif-
fusion in Concrete Technology) [23]. In that study, three methods
for introducing a polyoxyalkylene alkyl ether viscosity modifier
into the mortar were explored, namely direct addition to the
mixing water, application as a topical curing solution, and incorpo-
ration via pre-wetted lightweight fine aggregates. The latter tech-
nology, FLAIR (Fine Lightweight Aggregates as Internal
Reservoirs), had been previously applied to introduce shrinkage-
reducing admixtures (SRA) into a mortar, mitigating the air
detrainment observed when the SRA was added directly to the
mixing water [24]. As shown in Fig. 3, while all three methods of
VERDiCT introduction generally reduced the depth of chloride pen-
etration into a w/c = 0.4 mortar, by far the most successful was
when an aqueous solution of the viscosity enhancer was utilized
to pre-wet the lightweight fine aggregates. Based on these results,
the current study using LWA pre-wetted with water only was ini-
tiated to separate out the beneficial effects of IC from those of
VERDiCT.
2. Experiments and modeling
For this study, mortars with a water-to-cement ratio (w/c) of
0.4 were prepared. The actual batch mixture proportions employed
for mortars with and without IC are provided in Table 1. The
replacement of normal weight sand by LWA in the mortar was per-
formed to provide an extra 0.08 units of curing water per mass of
Control Control-surface
VERDiCT VERDiCT-surface
6 VERDiCT-FLAIR
Cl penetration depth (mm)
cement, while maintaining an overall volume fraction of 55% for
the sand or sand/LWA blend in each mortar. The LWA, an expanded
shale, was obtained from a lightweight aggregate manufacturer. It
has a saturated-surface-dried (SSD) specific gravity of 1.80 and a
total water absorption capacity of 22% by mass; 20.3% water or
about 93% of the total water is released from the LWA during
desorption in a 93% RH environment [22].
After mixing each mortar, cylindrical specimens 50 mm in
diameter by 100 mm in height were prepared. After curing for
1 d in the molds and subsequently curing for either 7 d or 28 d,
the cylinders were exposed in individual sealed plastic bottles to
1 M chloride ion solutions. The control mortar was cured in a solu-
tion of NaOH/KOH/Ca(OH)2, to avoid leaching of alkalis from the
mortar into the curing solution. For the mortar with IC, sealed cur-
ing in double plastic bags (7 d or 28 d) was employed, to promote
water movement from the lightweight fine aggregates to the sur-
rounding cement paste, as discussed previously. Isothermal calo-
rimetry measurements were conducted at 25 °C during the first
7 d of curing on replicate sealed mortar specimens weighing be-
tween 7 g and 7.5 g.
A silver nitrate spraying method [26,27] is being employed for
quantifying the chloride ion ingress into the control and IC mortar
cylinders after exposure times of 28 d, 56 d, 180 d, and 365 d. At
each exposure time, two cylinders are removed from their chloride
solutions and broken down the middle (lengthwise) using a uni-
versal testing machine. For each specimen, one of the two created
faces is sprayed with silver nitrate (AgNO3), the specimen is photo-
graphed, and image processing software is used to visually assess
the penetration depth of the chloride ions (example of results pre-
viously obtained using this technique are provided in Fig. 3).
The sieve size distributions of the different sands and the LWA
had been previously measured. For the mortar with IC, different
portions of the four normal weight sands were replaced by LWA
as indicated in Table 1, in order to maintain a similar overall fine
aggregate size distribution. The specific sieve size distributions

were used as input into the HCSS computer program to determine

5 the volume fraction of percolated ITZ paste as a function of ITZ
thickness for the two mortars. For these simulations, the computer
4
codes [8] were modified so that ITZs were only present surround-
3 ing the normal weight sand particles and not the LWA. While the
simulations were conducted for a range of ITZ thicknesses, in gen-
2 eral the ITZ thickness would be expected to be near the median
particle diameter of the cement [5–7], about 12 lm for the cement
1 employed in this study. A computational volume of 20 mm 
-

20 mm  20 mm was employed for each simulation conducted in

0
28 d 56 d this study, requiring over 750,000 individual sand/LWA particles.
For the ITZ thickness of 12 lm, random walk diffusion studies were
Exposure to Cl-
executed to estimate the relative diffusivity of the mortars with
Fig. 3. Penetration depth as determined by spraying with AgNO3 vs. Cl exposure and without IC. The ‘‘ant” step size was set at 6 lm and the simu-
time (28 d or 56 d) for mortar cylinders first cured for 28 d and then exposed to lations were conducted using 10,000 ants, each taking 100,000 ran-
chloride solutions. Labels of surface and FLAIR indicate 28 d curing in a curing dom steps [8]. Given the early ages at which the experimental
solution composed of a mixture of hydroxides and the VERDiCT admixture, or cured
specimens were exposed to chlorides, for these simulations, the
for 28 d under sealed conditions, respectively. Error bars indicate ± one standard
deviation in the measured average penetration depth [23]. diffusivity of the ITZ regions (surrounding the normal weight
aggregates only) was first set to be 20 times greater than that of
the bulk cement paste, as discussed above. To better understand
the influence of this parameter, a second set of simulations was
Table 1 conducted with a more conservative diffusivity ratio of 5:1. In
Mixture proportions for control and IC mortars. the simulations, no diffusion was allowed within either the normal
Material Control mortar (g) IC mortar (g)
weight sand or the LWA particles.
In a separate experiment, the absorption/release of chloride
Portland cement 1757.3 2300
Water 702.9 920
ions by/from the LWA and coarsest (silica) sand were monitored
F95 fine sand 985.0 954.0 by exposing 100 g of dry aggregate to 500 mL of either distilled
ASTM C778 [25] graded sand 748.6 836.1 water or a one molar solution of chloride ions. Within the experi-
ASTM C778 20–30 sand 748.6 788.2 mental variability of the measurement of chloride ion concentra-
S16 coarse sand 1457.7 982.2
tion, the results indicated no significant absorption or release of
LWA (pre-wetted) – 1100.7
chloride ions during a 28 d evaluation period by either the LWA
288 D.P. Bentz / Cement & Concrete Composites 31 (2009) 285–289

Table 2
Measured penetration depths of chlorides into mortars, via AgNO3 spraying technique.

Curing time (d) Exposure to Cl (d) Penetration depth (mm) for control mortar Penetration depth (mm) for IC mortar lIC/lcont DIC/Dcont
a
7 28 6.7 (0.6) 5.0 (0.4) 0.75 0.56
7 56 6.3 (0.3) 5.3 (0.1) 0.84 0.70
28 28 5.06 (0.33) 3.67 (0.39) 0.73 0.53
28 56 5.11 (0.53) 4.36 (0.30) 0.85 0.73
a
Numbers in parentheses indicate standard deviation in penetration depth between replicate specimens.

or the normal weight sand employed in this study. This suggests penetration depths for the IC and control mortars should be equiv-
that any reduction in chloride penetration depth is not being alent to the square root of the ratio of their diffusivities, or con-
caused by absorption of the chloride ions by the LWA (or sand). versely that the ratio of their diffusivities is given by the ratio of
their penetration depths squared, recognizing that the multi-mode
3. Results and discussion transport/reaction occurring in the real material is being mapped
onto much simpler Fickian diffusion. These ratios for the various
The measured penetration depths for chloride ingress into the curing conditions and chloride exposure periods have been in-
various mortars with and without IC are summarized in Table 2. cluded as the last two columns in Table 2. They suggest that the
In every case, the penetration depth measured by the AgNO3 spray- diffusion coefficient for chloride ions in the IC mortar may be only
ing technique is significantly less for the mortars with IC than for about 55–75% of that for the control mortar, a significant reduc-
their equivalent control mortars with no IC. While the exact ana- tion. In many real world degradation scenarios, concrete service
lytical solution for ingress/binding of chloride ions from a fixed life is proportional to diffusion coefficients, so reducing the diffu-
solution volume into a cylindrical mortar specimen is quite com- sion coefficient by a factor of two, for example, could effectively
plex [28] even when fluid sorption (imbibition) and reaction/bind- double the service life of the concrete [23].
ing are neglected, it contains terms of the form Dt/l2, where D is the Could such a significant reduction in diffusion coefficients be
diffusion coefficient, t is the exposure time, and l is the penetration explained simply by the absence of ITZ regions surrounding the
depth. This implies that for a fixed exposure time, the ratio of the lightweight aggregates in the IC mortar? According to the simula-
tion results provided in Fig. 4, the volume fraction of percolated ITZ
paste (defined as the volume fraction of ITZ paste that is continu-
0.24 ous across the microstructure) for the IC mortar is only 60% of that
Percolated ITZ volume fraction

Control mortar (sand only)

for the control mortar for an assumed ITZ thickness of 12 lm.
0.20 IC mortar (sand/LWA blend) However, the random walk simulations for these two mortar
microstructures indicate that this reduced ITZ volume/percolation
0.16
would produce only about a 22% reduction in diffusivity for the
mortar with IC vs. the control mortar without LWA. By contrast,
0.12
if a 5:1 ratio is assumed for the diffusivity of the ITZ relative to
0.08 the bulk paste, the projected reduction in diffusivity is only 10%.
But these simulations did not account for any possible further
0.04 reductions in bulk paste (and ITZ) diffusivities due to enhanced
hydration provided by the IC water. For example, for a w/c = 0.4 ce-
0.00 ment paste, previously developed equations [29] would indicate
0 5 10 15 20 25 about a 32% reduction in the diffusivity of a cement paste, if its de-
ITZ thickness (µm) gree of hydration were increased from 75% to 80%, due to the avail-
ability of extra IC water for instance. Even an increase from 77% to
Fig. 4. Simulation results for the volume fraction of percolated ITZ paste as a
80% would produce a 13% reduction in diffusivity. Enhanced hydra-
function of ITZ thickness for the control and IC mortar mixtures of Table 1. Dashed
vertical line indicates an ITZ thickness of 12 lm, the median particle diameter of the tion of the mortar with internal curing at later ages is supported by
cement used in this study. the isothermal calorimetry results presented in Fig. 5 that indicate

350
Heat Release (J/g cement)

300

250

200

150 Control-1
Control-2
100
IC
50

0
0 24 48 72 96 120 144 168
Time (h)

Fig. 5. Isothermal calorimetry results (25 °C) for mortars without (Control) and with IC. Results for two replicates for the control mortar are shown to provide an indication of
repeatability.
 D.P. Bentz / Cement & Concrete Composites 31 (2009) 285–289
about a 5% increase in heat released (or degree of hydration) at a
curing age of 7 d for the IC mortar relative to the control mortar,
both cured under sealed conditions. Fig. 1 also qualitatively indi-
cates the enhanced hydration and denser microstructure achieved
at later ages (120 d) in mortars with IC. Thus, the combination of
decreased ITZ volume fraction/percolation and enhanced hydration
could easily produce an IC mortar with a diffusivity that is about
55% to 75% of that of the control mortar (e.g., 0.78  0.68 = 0.53
or 0.9  0.87 = 0.78).
These results have been obtained for mortars with a 55% sand
volume fraction. The absolute reduction in diffusion coefficients
will likely be different for concretes that typically contain 60–
70% aggregates (coarse and fine) by volume fraction. The ITZ re-
gions in a conventional concrete will be highly percolated [7] so
that it may take a larger replacement by lightweight aggregates
to substantially reduce the volume fraction of percolated ITZ paste.
Still, some reduction will surely be achieved along with enhanced
hydration so that one would expect the IC concrete to have a lower
diffusion coefficient than its control counterpart. Furthermore, the
simulation and experimental techniques presented here should be
equally applicable to concrete as to the mortars examined in this
study, as will be pursued in future research.
4. Conclusions
In this study, the mortar prepared with IC and then cured under
sealed conditions exhibited chloride ion penetrations depths that
were significantly reduced with respect to the control mortar.
Modeling results indicate that this reduction in diffusion coeffi-
cient is likely due to both a significant reduction in the volume
fraction of percolated ITZ paste and enhanced long term hydration
in the mortar with IC. Thus, based on the results of this study, it ap-
pears that in addition to providing significant reductions in early-
age autogenous shrinkage and strength enhancement at ages of
28 d and beyond [4,30], IC may also produce mortars and concretes
with significantly enhanced resistance to the penetration of chlo-
ride ions and other deleterious species.
Acknowledgements
The author thanks Mr. Max Peltz of the Building and Fire Re-
search Laboratory for his assistance with the experimental pro-
gram. The provision of materials by the Northeast Solite
Corporation and the Lehigh Portland Cement Co. is gratefully
acknowledged.
References
[1] Holm TA, Bremner TW, Newman JB. Lightweight aggregate concrete subject to
severe weathering. Concr Int 1984;6:49–54.
[2] Zhang MH, Gjorv OE. Microstructure of the interfacial zone between
lightweight aggregate and cement paste. Cem Concr Res 1990;20:610–8.
[3] Lo TY, Cui HZ. Effect of porous lightweight aggregate on strength of concrete.
Mater Lett 2004;58:916–9.
[4] Bentz DP, Stutzman PE. Internal curing and microstructure of high-
performance mortars, ACI SP-256. Internal curing of high performance
289
concretes: laboratory and field experiences. American Concrete Institute,
Farmington Hills, MI; 2008. p. 81–90.
[5] Bentz DP, Garboczi EJ. Simulation studies of the effects of mineral admixtures
on the cement paste-aggregate interfacial zone. ACI Mater J
1991;88(5):518–29.
[6] Bentz, DP. Effects of cement PSD on porosity percolation and chemical
shrinkage. In: Persson B, Fagerlund G, editor. Proceedings of the second
international seminar on self-desiccation and its importance in concrete
technology, Lund University; 1999. p. 127–34.
[7] Winslow DN, Cohen MD, Bentz DP, Snyder KA, Garboczi EJ. Percolation and
pore structure in mortars and concretes. Cem Concr Res 1994;24:25–37.
[8] Bentz DP, Snyder KA, Garboczi EJ. A hard core/soft shell microstructural model
for studying percolation and transport in three-dimensional composite media.
NISTIR 6265. US Department of Commerce; 1999.
[9] Bentz DP. Fibers, percolation, and spalling of high performance concrete. ACI
Mater J 2000;97(3):351–9.
[10] Bentz DP, Snyder KA. Protected paste volume in concrete: extension to internal
curing using saturated lightweight fine aggregates. Cem Concr Res 1999;29:1863–7.
[11] Bentz DP, Aitcin PC. The hidden meaning of water-to-cement ratio. Concr Int
2008;30(5):51–4.
[12] Desai P, Lewis JA, Bentz DP. Unreacted cement content in macro-defect-free
composites: impact on processing-structure–property relations. J Mater Sci
1994;29(24):6445–52.
[13] Diamond S. Percolation due to overlapping ITZs in laboratory mortars? A
microstructural evaluation. Cem Concr Res 2003;33:949–55.
[14] Scrivener KL, Nemati KM. The percolation of pore space in the cement paste/
aggregate interfacial zone of concrete. Cem Concr Res 1996;26:35–40.
[15] Bretton D, Ollivier JP, Ballivy G. Diffusivite des ions chlore dans la zone de
transition entre pate de ciment et roche granitique. In: Maso JC, editor.
Interfaces in cementitious composites. London: E&FN Spon; 1992. p. 269–78.
[16] Bourdette B. Durabilité du mortier: Prise en compte des auréoles de transitions
dans la caractérisation et la modélisation des processus physiques et
chimiques d’altération. Ph.D. thesis, L’Institut National des Sciences
Appliquées de Toulouse; 1994 [in French].
[17] Otsuki N, Nishida T, Wattanachai P, Chin K. Experimental study on influence of
ITZ around aggregate on Cl diffusivity of concrete. J Soc Mater Sci, Jpn
2006;55(10):899–904 [in Japanese].
[18] Bentz DP. Influence of silica fume on diffusivity in cement-based materials II.
Multi-scale modeling of concrete diffusivity. Cem Concr Res 2000;30:1121–9.
[19] Thomas MDA. Chloride diffusion in high-performance lightweight aggregate
concrete. In: Ries JP, Holm TA, editors. Theodore Bremner symposium on high-
performance lightweight concrete; 2003. p. 77–93.
[20] Zhang MH, Gjorv OE. Permeability of high-strength lightweight concrete. ACI
Mater J 1991;88(5):463–9.
[21] Pyc WA, Caldarone MA, Broton D, Reeves D. Internal curing study with
intermediate lightweight aggregates, ACI SP-256. Internal curing of high
performance concretes: laboratory and field experiences. American Concrete
Institute, Farmington Hills, MI; 2008. p. 13–4.
[22] Bentz DP, Lura P, Roberts JW. Mixture proportioning for internal curing. Concr
Int 2005;27(2):35–40.
[23] Bentz DP, Peltz MA, Snyder KA, Davis JM. VERDiCT viscosity enhancers
reducing diffusion in concrete technology. Concr Int 2009;31(1):31–6.
[24] Bentz DP. Capitalizing on self-desiccation for autogenous distribution of
chemical admixtures in concrete. In: Proceedings of the 4th international
seminar on self-desiccation and its importance in concrete technology. Lund
University; 2005. p. 189–96.
[25] ASTM C 778. Standard specification for standard sand; 2005.
[26] Baroghel-Bouny V, Belin P, Maultzsch M, Henry D. AgNO3 spray tests:
advantages, weaknesses, and various applications to quantify chloride
ingress into concrete. Part 1: Non-steady-state diffusion tests and exposure
to natural conditions. Mater Struct 2007;40(8):759–81.
[27] Baroghel-Bouny V, Belin P, Maultzsch M, Henry D. AgNO3 spray tests:
advantages, weaknesses, and various applications to quantify chloride
ingress into concrete. Part 2: Non-steady-state migration tests and chloride
diffusion coefficients. Mater Struct 2007;40(8):783–99.
[28] Crank J. The mathematics of diffusion. Oxford University Press; 1975.
[29] Bentz DP, Jensen OM, Coats AM, Glasser FP. Influence of silica fume on
diffusivity in cement-based materials I. Experimental and computer modeling
studies on cement pastes. Cem Concr Res 2000;30:953–62.
[30] Bentz DP. Internal curing of high performance blended cement mortars. ACI
Mater J 2007;104(4):408–14.
 Cement & Concrete Composites 31 (2009) 290–300

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

FRP-strengthened RC beams under sustained loads and weathering

K.H. Tan a,*, M.K. Saha b, Y.S. Liew c
a
Dept. of Civil Engineering, National University of Singapore, Blk E1A, # 07-03, 1 Engrg. Dr. 2, Singapore 117576, Singapore
b
Structural Engineer, Global Industries Asia Pacific Pte Ltd, 41 Science Park Road, Singapore Science Park II, Singapore 117610, Singapore
c
A/E/C Solutions Specialist, Advance Contech, 3 Raffles Place, #07-01, Singapore 048617, Singapore

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

Article history: Glass FRP-strengthened RC beams were subjected to sustained loads and placed for different periods out-
Received 21 December 2007 doors, indoors, and in chambers that accelerate the effects of outdoor tropical weathering by a factor of
Received in revised form 17 March 2009 six. Beams subjected to outdoor weathering had up to 18% larger crack widths and 16% larger deflections
Accepted 20 March 2009
compared to those kept indoors at the end of 1 and 23=4 years, respectively. The increase in deflections and
Available online 27 March 2009
crack widths was lesser for beams with a higher FRP reinforcement ratio. The residual flexural strength
and ductility of the beams decreased with longer weathering periods. Also, the failure mode of the beams
Keywords:
changed from concrete crushing to FRP rupture, indicating a deterioration in the mechanical properties of
Beams (support)
Cracking
the FRP laminates. Analytical methods which account for material degradation in concrete and FRP lam-
Deflection inates are presented and found to predict the long-term flexural characteristics of the beams well.
Ductility Ó 2009 Elsevier Ltd. All rights reserved.
Flexure
Fiber-reinforced polymer
Long-term serviceability
Residual strength
Sustained loads
Weathering

1. Introduction
Fiber-reinforced polymer (FRP) composite laminates have been
popularly used in structural strengthening of reinforced concrete
(RC) members due to their high strength–weight ratio and ease
of installation [1,2] compared to other materials such as steel
plates. However, the long-term field performance of such FRP-
strengthened members has not been well established.
Limited studies have been reported on the flexural strength of
FRP-strengthened RC beams subjected to weathering [3–5].
A study by Almusallam et al. [3] revealed that both solar radiation
and wet–dry condition did not affect the flexural strength or rigid-
ity of beams strengthened with glass FRP (GFRP) laminates signif-
icantly for an exposure period of 12 months. On the other hand,
Leung et al. [4] reported a reduction in flexural capacity of carbon
FRP-strengthened concrete beams when immersed in water for a
period of 6 months. Liew [5] studied the accelerated weathering
effect of tropical climate on RC beams strengthened with GFRP
laminates, and concluded that a weathering period of 6–9 months
resulted in the change in failure mode with a marginal drop of 2%
in flexural capacity whereas weathering for more than 6 years re-
* Corresponding author. Tel.: +65 65162260; fax: +65 67791635.
E-mail address: cvetankh@nus.edu.sg (K.H. Tan).
duced the flexural strength by 15% due to deterioration of bond be-
tween FRP laminate and concrete.
In reality, structures are subjected to sustained loading and the
combined effect of sustained loading and weathering is likely to
cause further degradation in the mechanical properties of concrete
and FRP composite laminates. This would ultimately reduce the
overall stiffness of the beam, as well as the safety margin against
the ultimate collapse.
This paper reports a study on the long-term flexural character-
istics of glass FRP-strengthened RC beams that were subjected to
sustained loads and placed outdoors, indoors, and in weathering
chambers that accelerate the effects of the outdoor tropical climate
by a factor of six. Analytical methods accounting for the degrada-
tion in material properties are presented to calculate the long-term
deflection and residual flexural strength of the FRP-strengthened
RC beams.
2. Analytical considerations
2.1. Calculation of deflection
Under sustained loading, the concrete in the compression zone
is subjected to creep which leads to a smaller effective modulus of
the material and correspondingly larger beam deflections in the
long-term. Similarly, the FRP laminates are subjected to tensile

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.03.002
 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300 291

Table 1
Test program.
Beam Total sustained load, Ps (kN) Ps/P1a Exposure condition Exposure duration Remarks
(a) Small specimens (100 mm  100 mm  700 mm)
A1/A2-6 m 20.4 0.53 Ambient 6 months Beams under sustained loads in ambient laboratory condition
A1/A2-1y 1 year
E1/E2-6 m Exterior/outdoor 6 months Beams under sustained loads in exterior/outdoor natural condition
E1/E2-1y 1 year
C1/C2-5d 22.7 0.59 Chamber 5 days Beams under sustained loads in chamber
C1/C2-15d 15 days
C1/C2-1 m 1 month
C1/C2-2 m 2 months
C1/C2-3 m 3 months
C1/C2-6 m 6 months

Beam qfrpb (%) Ps (kN) Ps/Puc Exposure condition Exposure duration Remarks
(b) Large specimens (100 mm  125 mm  2000 mm)
A0-3y 0 15.8 0.59 Ambient 23=4 years Beams under sustained loads in ambient laboratory condition
A1-3y 0.64 0.49
A3-3y 1.92 0.40
C1-1y 0.64 17.7 0.55 Chamber 1 year Beams under sustained loads in chamber
C3-1y 1.92 0.45
a
P1: calculated load-carrying capacity of Beam A1-0d.
b
qfrp: FRP reinforcement ratio = FRP laminate area/cross-section area.
c
Pu: calculated load-carrying capacity.

A produce physiochemical processes like hydrolysis or plasticization,

thermal decomposition and photo-oxidation of the polymers, caus-
ing a reduction in the modulus of the FRP laminates. However, the
A R6 stirrup @ effect of weathering on concrete is much less significant and may
end U-strip 600 mm 50 mm c/c
therefore be neglected.
700 mm
The compressive creep in concrete due to sustained loads can be
24 mm 100 mm
accounted for by using an effective modulus of elasticity:
2 R6
Ec
R6 Ee ¼ ð1Þ
76 mm 1 þ /t;t0
100 mm
2 T6 where Ec = modulus of elasticity of concrete at age (time) t0, t0 = age
of concrete at the time of application of loading, t = time at which
50 mm for Type 1 and deflection is to be computed, and /t;t0 = creep coefficient of concrete
100 mm for Type 2 FRP at time t as given in ACI 209R [6].
System Similarly, the tensile creep of FRP laminates due to sustained
Section A-A loads can be accounted for using an effective modulus of elasticity,
(a) Small Specimens given by [7]:

A Efrp
Efrp;t ¼ ð2Þ
1 þ /frp

where Efrp = initial modulus of elasticity of FRP composite laminate,

end U-strip A R6 stirrup @ 75 mm c/c and /frp = creep coefficient for FRP composite laminate, given by [7]:
1800 mm  m
2000 mm t
/frp ¼ 1 ð3Þ
t0
24 mm
2 T6 where t = time in hours after application of loading, t0 = 1 h and m is
a coefficient determined experimentally from the relation between
99 mm R6 efrp,t (that is, the FRP strain at time t) and t/t0.
125 mm To account for the effect of tropical weathering on the modulus
of elasticity of FRP laminates, a residual modulus of elasticity func-
2 T10 tion is defined as:

100 mm Efrp;w
uE;w ¼ ð4Þ
Efrp
Section A-A
(b) Large Specimens where Efrp,w = modulus of elasticity of FRP laminates at time t under
the sole effect of weathering. The value of uE,w can be established
Fig. 1. Test beams and section details. from tensile tests on weathered specimens.
The effective modulus of elasticity of FRP laminates as a result
of the combined sustained loading and tropical weathering can
creep, leading to the same result. On the other hand, the action of then be taken as the product of Efrp,t and uE,w, which in view of
weathering factors such as heat, moisture and sunlight, is to Eqs. (2) and (4), can be written as:
292 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300

Table 2
Properties of FRP systems.

Type 1 (unidirectional) GFRP laminate Type 2 (bidirectional) GFRP laminate

Fiber Type E-Glass E-Glass
Sheet form Unidirectional Bidirectional
Tensile strength (MPa) 1700 130
Elastic modulus (GPa) 71 11
Ultimate strain (%) 2.0 1.25
Resin Type Two part, 100% solid, low viscosity amine cured epoxy Same as type 1 GFRP laminate
Tensile strength (MPa) 54
Elastic modulus (GPa) 3
Ultimate strain (%) 2.5
FRP Laminate Volume fraction of fiber (%) 53 –
Volume fraction of resin (%) 47 –
Elastic modulus (GPa) 39 14a
Thickness (mm) 0.8 0.7
a
Obtained from tensile coupon tests conducted by Liew [5].

4 The value of ue,w in Eq. (6) has to be determined experimentally

such as in accordance with JSCE-E-541 [10] test method. Values
have been established for the GFRP systems [5] used in this study
Chamber (0.59 P 1 ) and are reported later herein.
Total Deflection (mm)

3 The flexural failure modes considered were: (1) crushing of con-

crete in compression; (2) rupture of FRP; and (3) flexural crack in-
duced FRP debonding. Based on the method of strain compatibility
Ambient/Outdoor and force equilibrium, the strength corresponding to each mode of
(0.53 P 1 ) failure can be found by equating the corresponding material strain
2 to its limiting strain.
The crushing of concrete in compression would thus occur
when the strain in top compressive fiber of a section reaches the
ultimate strain of concrete, that is:
1
5 30 60 180 360
ec ¼ ecu ð8Þ
0 100 200 300 400 where ecu may be taken as 0.003. Failure by FRP rupture would hap-
Time after Loading (Days) pen when:

Fig. 2. Determination of equivalent sustained loads for weathering in chamber.

efrp ¼ 0:8  efrpu;wt ð9Þ
where efrpu,wt is the ultimate tensile strain of FRP laminates under
combined sustained loading and tropical weathering. The coeffi-
uE;w  Efrp Efrp;w cient of 0.8 in Eq. (9) accounts for the average lower strains of
Efrp;wt ¼ uE;w  Efrp;t ¼ ¼ ð5Þ
1 þ /frp 1 þ /frp FRP rupturing when bonded to beams compared to strains mea-
sured from material tensile test [11].
With the effective modulus of elasticity of concrete and FRP Also, following Teng et al.’s [12] approach, debonding of FRP
laminates determined from Eqs. (1) and (5), respectively, the max- laminate from concrete strata would occur when:
imum deflection of the beam can be calculated using elastic anal- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
ysis methods [7] and incorporating an effective moment of inertia 1:1bw bl Efrp;wt fc0
based on Branson’s formula [8]. efrp ¼ efrp;db ¼ ð10Þ
Efrp;wt tfrp

2.2. Calculation of flexural strength in which, efrp,db = FRP debonding strain, fc0 = concrete cylinder com-
pressive strength, and tfrp = thickness of FRP laminates, bw = bond
In this study, the sustained loads and tropical weathering are width coefficient, and bl = bond length coefficient.
assumed to have insignificant effect on the ultimate strength and The predicted failure mode would be the one with the lowest
strain of concrete. The ultimate strain of FRP composite laminate moment capacity.
after weathering under sustained loads is taken as:
efrpu;wt ¼ ue;t  ue;w  efrp;u ð6Þ 3. Test program

where ue,t = residual ultimate strain function under sustained load- Tests were carried out on two sizes of beam specimens as
ing, ue,w = residual ultimate strain function under weathering, and shown in Table 1 and Fig. 1. Two types of glass fiber-reinforced
efrp,u = ultimate strain of virgin (that is, unweathered, non-load sus- polymer (GFRP) systems, Type 1 and Type 2 as described later,
tained) FRP laminate. were used for the smaller specimens. Type 1 GFRP system was
The value of ue,t is obtained as [9]: used for strengthening of larger specimens.
  The beams were designated as Xm-t, where ‘X’ denotes the
efrp;u  ðefrp;t  efrp Þ efrp
ue;t ¼ ¼ 1  /frp ð7Þ weathering condition (that is, ‘A’ for indoors, ‘E’ for outdoors and
efrp;u efrp;u ‘C’ for in-chamber), ‘m’ is the type of GFRP system for small speci-
where efrp = instantaneous strain in FRP laminate due to sustained mens and the number of ply of GFRP laminate for large specimens
loads, which can be calculated from elastic bending theory using and ‘t’ indicates the actual weathering period in days (d), months
cracked section analysis, and efrp,t = strain in FRP laminate at time t. (m) or years (y). For example, C1-6 m refers to a beam strengthened
 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300 293

200 200 200

2 x P/4 2 x P/4
2 x P/4 2 x P/4
100
Demec pins

100
100

100

50 600 50 2 x P/4 2 x P/4

2 x P/4 Longitudinal elevation 2 x P/4 Transverse
elevation

1 P /4 3 P/4

2 P /4 4 P/4
Plan view
# All measurements are in mm
(a) Small Specimens

100 1800 100

225 450 450 450 225

Test
Beam

Concrete
Blocks/
Steel Plates

Support

Ground
Level

#All measurements are in mm

(b) Large Specimens
Fig. 3. Schematic drawing of test set-ups.

with Type 1 GFRP laminates and placed in the weathering chamber
for 6 months.
3.1. Material properties
The average concrete cube compressive strength and modulus
of elasticity were 42 MPa and 24.8 GPa, respectively, at 28 days.
The average yield strength and modulus of elasticity of the hot-
rolled deformed tensile longitudinal reinforcement averaged
525 MPa and 200 GPa, respectively.
Type 1 FRP system consisted of unidirectional roving E-glass fi-
ber fabric that is commercially available for structural strengthen-
ing works while Type 2 FRP system consisted of bidirectional
woven roving E-glass fiber fabric that is used to produce commer-
cial products. Type 2 system has less superior mechanical proper-
ties than Type 1 system. In both systems, the fabric was
impregnated with a two-part, 100% solid, and low viscosity amine
cured epoxy by hand to form the laminates. The properties of glass
fiber fabrics and resin, as supplied by the manufacturers, are
shown in Table 2.
3.2. Installation of FRP system
For small beams, one ply of Type 1 GFRP laminate (50 mm wide
and 0.8 mm thick) or Type 2 GFRP laminate, (100 mm wide and
0.7 mm thick) was installed on the tensile face of each beam. The
tensile capacity of the laminates was kept the same; hence the
widths of the FRP laminates were different. For large beams, Type
1 GFRP composite laminates, 100 mm in width, were applied in
one ply (0.8 mm thick) for Beams A1 and C1, and in three plies
(2.4 mm thick) for Beams A3 and C3. The GFRP system was installed
21 days after casting following the wet lay-out procedure. At the
ends of the GFRP laminate, a carbon fiber sheet was attached trans-
versely to prevent premature plate-end debonding during testing.
3.3. Application of sustained loads
The service load for the beams was assumed as Pu/1.7, where Pu
is the calculated load-carrying capacity of the beam. The factor of
1.7 is based on the normal loads carried by a structural concrete
member under service condition. For the small beam specimens
294 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300

3.0 0.4

E2

Maximum Crack Widths (mm)

Total Deflections (mm)

2.5 E2 0.3 A2

A2 E1

E1
A1
2.0 0.2
A1

Exposure
Condition FRP Type
1.5 A1 Ambient 1 0.1
A2 Ambient 2
E1 Exterior 1
E2 Exterior 2
1.0 0
0 100 200 300 400 0 100 200 300 400
Outdoor Period (Days)
(a) Small Specimens

Chamber Period (Days)

0 60 120 180 240 300 360 420 0 60 120 180 240 300 360 420
12 12

C1-1y
Total Deflections (mm)

9 9
A1-3y C3-1y

A3-3y
6 6

3 P sus 3 P sus
A1-3y: 0.49P 1 A3-3y: 0.40P 3
ρ frp = 0.64% C1-1y: 0.55P 1 ρ frp = 1.92% C3-1y: 0.45P 3
0 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Outdoor Period (Days)
(b) Large Specimens
Fig. 4. Effect of weathering on long-term serviceability.

placed outdoors or under ambient condition, the sustained load using concrete blocks and steel plates at four-points along the span
was 90% of the service load or 0.53P1, where P1 is the calculated to simulate uniform sustained loads. The beams kept under ambi-
load-carrying capacity of the beam strengthened with one ply of ent laboratory condition (A0-3y, A1-3y, and A3-3y) were subjected
GFRP laminate. to a sustained load equal to the service load or 0.59 times the ulti-
For specimens placed in the weathering chamber, an equivalent mate load-carrying capacity (0.59P0) of the unstrengthened beam.
sustained load had to be applied so as to accelerate the effect by a This is equivalent to 0.49 times the ultimate load-carrying capacity
factor of six, similar to the effect of accelerated weathering (see of the beam strengthened with one ply of GFRP laminates (0.49P1),
Section 3.4). It was determined as follows. Referring to Fig. 2, first, or 0.40 times that of the beam strengthened with three plies of
the deflection history of the beam under a sustained load of 0.53P1 GFRP laminates (0.40P3). For the two beams kept in the weathering
was computed analytically following the method explained earlier chamber, an equivalent sustained load determined in the same
in Section 2.1 but excluding the weathering effect. Then the sus- manner as for small specimens equal to 0.55P1 was applied on
tained loads to cause the same deflections at 30 days, 180 days, C1-1y and 0.45P3 on Beam C3-1y.
and 360 days in one-sixth the time were determined, and the aver-
age value was found to be equal to 0.59P1. 3.4. Weathering
All the beams were clamped in pairs using transverse stain-
less steel bars at the beam ends, as shown in Fig. 3a, with steel The tropical climate is characterized by a hot and humid
rods placed in between the beams at one-third points, to simu- weather with very little seasonal variation in temperature or pre-
late the sustained loads. The stress level was checked by contin- cipitation throughout the year. Just north of the Equator in Singa-
uous monitoring of strain gauges located at top concrete face, pore, the diurnal temperature rarely goes below 23 °C or above
internal tensile steel bars and the FRP laminates at mid-span 33 °C, and the monthly rainfall ranges between 100 to 180 mm.
of the beam. On average, there is not less than four sunshine hours per day.
The larger beams were simply supported over a span of Mean daily relative humidity is as high as 80–90% at night and
1800 mm on a steel frame as shown in Fig 3b. Loads were applied 50–60% during the daytime.
 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300 295

For small beam specimens, a ferrocement weathering chamber
(1500 mm  1000 mm  700 mm) was fabricated to simulate the
outside tropical weathering effect in the laboratory [5]. By consid-
ering the outdoor average diurnal sun hour percentage and
monthly rainfall fraction, a continuous light-wet-dark cycle was
generated in the weathering chamber. In each cycle, the light per-
iod lasted 1.5 h, during which both the UV-A floodlight and cera-
mic heaters worked together to generate UV-A ray and heat that
simulate the daytime weathering condition. It was then followed
by 1.5 h of wet period where the entire specimens were wetted
by water spraying through the atomizers, similar to outdoor wet-
ting of specimens by the rain.
To better simulate the outdoor temperature and humidity fluc-
tuations, 1 h of dark period (that is, idle period) was employed
immediately after the light period, during which all the gadgets
were switched off. Hence one cycle of weathering took 4 h and 6
cycles can be completed within a day. The natural outdoor weath-
ering process was therefore carried out at an accelerated rate of six
in the weathering chamber. This has been verified from tests on
tively. From visual inspection, the interface between the FRP com-
posite laminate and the concrete substrate was found to be
unaffected by combined sustained loading and tropical weather-
ing; the larger deflections and crack widths may be due to the deg-
radation in tensile stiffness of GFRP composite laminate.
Fig. 4b shows the effect of weathering on the deflections of large
beams under sustained loads. After 5½ months weathering in the
chamber (equivalent to 23=4 years outdoors), Beams C1-1y (qfrp =
0.64%) and C3-1y (qfrp = 1.92%) showed about 16% and 13% larger
deflections compared to Beams A1-3y (qfrp = 0.64%) and A3-3y
(qfrp = 1.92%), respectively, which were kept in ambient laboratory
condition. Again, the larger deflections may be due to the degrada-
tion in tensile stiffness of GFRP composite laminates due to the
synergistic effect of sunlight and rainfall.
3.0

Total Deflections (mm)

beam specimens placed outdoors and in the chamber [5]. 2.5 C2-Avg.
For the large beam specimens, a weathering chamber of size
C1-Avg.
2300 mm  1000 mm  625 mm was constructed to enclose the
beams. To reproduce all the weathering factors, the chamber was 2.0
equipped with the required devices, namely UV light, ceramic
heaters, thermostats, water pump and atomizers, similar to the
weathering chamber for small beam specimens. The reproduction 1.5 (a) Small Exposure: Chamber
of the weathering factors at an accelerated rate of six has also been FRP Type
Specimens -
verified [13]. C1 1
Deflections C2 2
1.0
3.5. Instrumentation
0.4
During the weathering period, the mid-span deflection of the
Maximum Crack Widths (mm)

C2-Avg.
small beams was measured using a demec gauge system (Fig. 3)
0.3
with an accuracy of 0.002 mm and the crack width was measured C1-Avg.
using a hand-held microscope which has a graduated scale in divi-
sions of 0.02 mm. Measurements were taken upon application of
sustained loads, and after 1, 3, 5, 7, 10, 14, 21 days, weekly up to 0.2
3 months, every 15 days till 6 months, and thereafter every month
till the end of 1 year.
Deflections of large beams were measured using a transferable 0.1 (b) Small Exposure: Chamber
Specimens – FRP Type
dial beam consisting of a movable frame with three dial gauges lo- C1 1
cated at quarter points and mid-span. The dial gauges had an accu- Max. crack widths
C2 2
racy of 0.01 mm. Measurements were taken at the same time 0
intervals as those for the small beam specimens. 0 50 100 150 200
At the end of the weathering period, the specimens were re- Chamber Period (Days)
lieved of the sustained loads. They were then tested statically un-
der four-point loading using the Instron universal testing machine 0 60 120 180 240 300 360 420
to failure, at a constant cross-head speed of 0.2 mm/min. The mid- 12
span deflection was monitored by a Linear Variable Displacement A0-3y
Transducer (LVDT). Crack widths were measured in the pure mo- C1-1y
ment zone at 5 kN intervals using a hand-held microscope. 9
Total Deflections (mm)

C3-1y

4. Test results and discussion

6
4.1. Long-term deflection and crack width

4.1.1. Effect of weathering P sus ρ frp

3
Fig. 4a compares the total deflections and crack widths of small (c) Large A0-3y: 0.59 P 0 0%
beams under sustained loads and subjected to outdoor weathering Specimens - C1-1y: 0.55 P 1 0.64%
with those kept indoors under ambient laboratory condition. The Deflections C3-1y: 0.45 P 3 1.92%
deflections of beams strengthened with Type 1 GFRP system and 0
kept outdoors were found to be more than those kept indoors by 0 500 1000 1500 2000 2500
about 7% at the end of 1 year. Similarly, for beams strengthened Outdoor Period (Days)
with Type 2 GFRP system, the difference was about 9%. On the
other hand, the crack width was larger by 13% and 18%, respec- Fig. 5. Effect of FRP type and reinforcement ratio on long-term serviceability.
296 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300

3 3 3

2 2 2

1 1 1
Total Deflections (mm)

A1: GFRP Type 1 (Unidirectional) E1: GFRP Type 1 (Unidirectional) C1: GFRP Type 1 (Unidirectional)
Exposure: Ambient Exposure: Outdoor Exposure: Chamber
P sus = 0.53P 1 P sus = 0.53P 1 P sus = 0.59P 1
0 0 00
3 3 3

2 2 2

Analytical
Experimental
1 1 1
A2: GFRP Type 2 (Bidirectional) E2: GFRP Type 2 (Bidirectional) C2: GFRP Type 2 (Bidirectional)
Exposure: Ambient Exposure: Outdoor Exposure: Chamber
P sus = 0.53P 1 P sus = 0.53P 1 P sus = 0.59P 1
0 0 00
0 100 200 300 400 0 100 200 300 400 0 50 100 150 200

Time after Loading (Days)

(a) Small Specimens

16 16

12 12
Total Deflections (mm)

8 8

4 C1-1y 4 C3-1y
ρ frp = 0.64% ρ frp = 1.92% Analytical
P sus = 0.55P 1 P sus = 0.45P 3 Experimental
0 0
0 100 200 300 400 0 100 200 300 400
Time after Loading (Days)
(b) Large Specimens
Fig. 6. Comparison of observed deflections with analytical predictions.

4.1.2. Effect of FRP type and reinforcement ratio
The total deflections and crack widths observed on small beams
kept in the weathering chamber are shown in Fig. 5a and b, respec-
tively. The solid lines give the average of the observed deflections
and crack widths. At the end of 6 months of chamber weathering
(equivalent to 3 years outdoor), beams strengthened with Type 1
FRP system showed 10% and 17% lesser deflection and crack width,
respectively, compared to beams strengthened with Type 2 FRP
system.
On the other hand, after 1 year outdoor weathering (Fig. 4a),
beams strengthened with Type 1 GFRP system showed 9% lesser
deflection compared to beams strengthened with Type 2 GFRP sys-
tem. As for crack widths, it was 21% lesser (Fig. 4a). The lesser
amount of fiber in the direction of stress, initial lack of straightness
of the fiber, and inferior mechanical properties may be responsible
for the larger deflections and crack widths observed in beams
strengthened with Type 2 GFRP system.
Fig. 5c shows the effect of FRP reinforcement ratio on the deflec-
tions of the loaded beams under tropical weathering. The effect of
weathering seems to be more detrimental in Beam C1 with
qfrp = 0.64% than in Beam C3 with qfrp = 1.92%. Beam C1 was found
to deflect 20% more than Beam C3 under the same sustained loads
after 1 year of accelerated weathering. Weathering elements did
not seem to result in deterioration of the interface bond within
the FRP layers or between the FRP laminates and the concrete sub-
strate in Beam C3 (qfrp = 1.92%). However, the deflections of Beams
C3 and C1 were 23% and 5% lesser, respectively, compared to the
control beam, A0-3y (qfrp = 0%) at the end of 5½ months in the
chamber.
4.1.3. Comparison of observed deflections with analytical predictions
In the analysis, the creep coefficient of concrete established for
conventional concrete of a similar strength as in the current inves-
tigation by Tan et al. [14] was used. The creep coefficient of FRP
composite laminate was computed from Eq. (3) with m equal to
0.0095 for Type 1 GFRP and 0.035 for Type 2 GFRP system. These
values were the average values obtained by Holmes and Just [9]
from tensile creep tests on specimens subjected to stress levels
(75–102 MPa for unidirectional and 36–88 MPa for bidirectional
GFRP) similar to the current investigation.
 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300 297

Table 3
Flexural characteristics of beams after weathering.

Specimen type Designation Pu (kN) wsl (mm) Dsla (mm) Dyl (mm) Dul (mm) Dul/Dyl Failure mode
Obs. Pred.
Unstrengthened small beams A0-6 ma 28.44 n.a. 1.53 1.98 5.31 2.68 n.a. n.a.
A0-1ya 27.21 n.a. 1.76 2.21 6.40 2.89 n.a. n.a.
Small beams with Type 1 GFRP A1-0d 45.18 0.28 2.70 3.23 10.58 3.27 CC CC
A1-6 m 32.98b 0.38 2.14 2.90 5.74 1.98 DB CC
A1-1y 42.30 0.20 2.03 2.50 6.51 2.60 CC CC
E1-6m 38.44 0.18 1.49 2.11 5.12 2.43 FR FR
E1-1y 40.14 0.14 1.40 1.82 5.40 2.96 FR FR
C1-5d 42.11 0.16 2.19 3.01 10.13 3.36 CC/DB CC
C1-15d 42.98 0.20 1.71 2.53 8.83 3.49 CC CC
C1-1 m 36.07 0.28 1.79 2.19 6.68 3.05 CC FR
C1-2 m 33.07 0.24 1.50 2.13 5.51 2.59 CC FR
C1-3 m 32.71 0.24 1.64 2.67 6.42 2.40 FR FR
C1-6 m 37.42 0.16 1.75 2.56 5.16 2.02 FR FR
Small beams with Type 2 GFRP A2-0d 39.36 0.22 2.22 2.89 7.18 2.49 CC CC
A2-6 m 37.02 0.14 2.39 2.85 6.13 2.15 FR FR
A2-1y 39.13 0.16 1.99 2.46 5.48 2.23 FR FR
E2-6 m 37.57 0.14 1.19 2.00 4.21 2.11 FR FR
E2-1y 36.41 0.14 1.10 1.93 3.78 1.96 FR FR
C2-5d 36.96 0.20 1.54 2.35 6.76 2.88 CC/FR CC
C2-15d 37.30 0.20 1.54 2.47 5.64 2.28 FR/CC CC
C2-1 m 26.71b 0.12 1.47 4.19 7.32 1.75 CC FR
C2-2 m 29.92b 0.26 1.53 2.42 5.74 2.37 FR FR
C2-3 m 34.75 0.18 2.43 2.99 7.38 2.47 FR FR
C2-6 m 34.79 0.16 1.52 2.25 4.59 2.04 FR FR
Large beams with Type 1 GFRP A1-0d 28.10 0.10 10.11 15.24 39.12 2.57 CC CC
C1-1y 24.59 0.08 6.01 12.58 15.94 1.27 FR FR
A3-0d 38.86 0.12 11.41 16.11 28.74 1.78 DB CC
C3/1y 35.68 0.13 8.79 17.39 26.75 1.54 FR CC

Note: Pu: ultimate load; wsl; max. crack width at service load (i.e., 0.59Pu).
Dsl: deflection at service load; Dyl: deflection at steel yield load; Dul: deflection at ult. load; n.a.: not applicable; CC: concrete crushing; DB: FRP debonding; FR: FRP rupture.
a
Beam without FRP composite laminate.
b
Outlier.

The residual modulus of elasticity functions due to the sole ef-
fect of weathering was established by Liew [5] using tensile cou-
pons as follows:
uE;w ¼ 1  k logðte Þ
where te = exposure period (in days); and k = 0.00103 and 0.00282
for Type 1 GFRP system subjected to outdoor weathering and accel-
erated weathering in the chamber, respectively, with corresponding
values of 0.0502 and 0.0609 for Type 2 GFRP system.
The calculated deflections are compared with the test results in
Fig. 6. The analytical predictions are in general conservative, but
reasonably accurate. The difference between the calculated and ob-
served deflections averaged 13% for small beams at the end of a
3 year equivalent outdoor period while for large beams, the aver-
age difference was 16% at the end of a 6 year equivalent outdoor
period. The analytical results for deflection overestimated those
of the tests mainly because the materials did not degraded uni-
formly and to the extent as the individual component materials
when subjected to the same weathering conditions. That is, the in-
side concrete of the beam and the inner face of the laminates are
not as much affected by the weathering.
4.2. Flexural strength
The flexural characteristics of all specimens tested after various
time periods of weathering, is shown in Table 3. Among the small
beams strengthened with Type 1 GFRP system, Beam A1-6 m
showed unexpectedly low strength due to the premature debond-
ing of GFRP laminate at the mid-span which is irrational consider-
ing the sustained load level and is therefore considered as an
outlier. The other beams, in general, showed a reduction in strength
with longer period of weathering. After 6 months of accelerated
weathering in chamber (equivalent to 3 years of outdoor weather-
ing), the flexural strength of beams strengthened with Type 1 GFRP
laminates was reduced by about 17%, that is, comparing C1-6 m
ð11Þ
with A1-0d. However, the strength is still 32% higher than the beam
without FRP strengthening (A0-6 m). Beam E1-1y that was exposed
to outdoor weathering for 1 year showed only 5% reduction in flex-
ural strength compared to Beam A1-1y which was kept under ambi-
ent condition. Beams strengthened with Type 2 GFRP laminates also
showed a reduction in strength with weathering periods.
The load–deflection curves of small beams with Type 1 GFRP
laminates kept inside the weathering chamber are compared with
the reference Beam A1-0d in Fig. 7a. The initial stiffness of all the
weathered beams was slightly lesser than that of Beam A1-0d. For
short periods of weathering (up to 15 days), the beams showed im-
proved stiffness beyond the sustained load level Ps (that is, 0.59P1 or
22.7 kN). The improved stiffness may be due to the initial curing of
concrete and FRP laminates during short periods of weathering,
which vanished quickly due to deterioration of fiber-resin bond
and fiber strength with longer periods of weathering. Under out-
door weathering (Fig. 7b), however, the stiffness of the beams after
1 year seems to be not much affected compared to the companion
beams kept under ambient condition.
The load–deflection curves of beams with Type 2 GFRP lami-
nates are shown in Fig. 8. The flexural strength of Beam C2-6 m re-
duced by about 12% after 6 months of accelerated weathering, but
the flexural strength was still 22% higher than the unstrengthened
Beam A0-6 m (Fig. 8a). After 1 year, Beam E2-1y showed only 7%
reduction in strength compared to Beam A2-1y (Fig. 8b).
Also, Beams C2-1 m and C2-2 m showed unexpectedly a drop in
strength of 32% and 24%, respectively (see Table 3). Due to the irra-
298 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300

50 50
A1-0d

40 40
C2-15d
C2-5d
30 C1-5d 30
Load (kN)

Load (kN)
C1-6m C2-3m
C1-15d
20 Ps C1-1m 20 Ps A2-0d
C1-3m C2-6m
C2-2m C2-1m
C1-2m
10 10

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Mid-span deflection (mm) Mid-span deflection (mm)
(a) Beams subjected to accelerated weathering in chamber (a) Beams subjected to accelerated weathering in chamber

50 50
A1-0d E2-6m
E2-1y
40 40

A1-1y
30
Load (kN)

Load (kN) 30 A2-1y

E1-1y
A2-6m
A1-6m
20 E1-6m 20
A0-6m
A0-1y A0-6m A0-1y A2-0d

10 10

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Mid-span deflection (mm) Mid-span deflection (mm)
(b) Beams under ambient condition or outdoor weathering (b) Beams under ambient condition or outdoor weathering
Fig. 7. Load–deflection characteristics of small RC beams strengthened with Type 1 Fig. 8. Load–deflection characteristics of small RC beams strengthened with Type 2
GFRP system. GFRP system.

tional low strength of Beams C2-1 m and C2-2 m, especially com-
pared with Beams E2-6 m and E2-1y (with equivalent outdoor
weathering periods) and considering the general trend, these two
beams are regarded as outliers.
The flexural strength of large RC beam specimens, also reported
in Table 3, is compared in Fig. 9a. For beams with qfrp = 0.64%, the
reduction in flexural strength was about 13% (comparing Beam C1-
1y with Beam A1-0d) whereas for beams with qfrp = 1.92%, the
reduction was 8% (comparing Beam C3-1y with Beam A3-0d) at
the end of 1 year of accelerated weathering (equivalent to 6 years
outdoor weathering). The load–deflection curves for these beams
are shown in Fig. 9b. Except for the initial permanent set, the
load–deflection behavior of the weathered beams above the sus-
tained load level remains the same as those of the virgin beams.
From the results, it is evident that effect of weathering is lesser
for beams with higher qfrp. This may be due to better protection
against weathering elements due to thicker FRP laminates.
The ductility of RC beams strengthened with Type 2 GFRP sys-
tem also seemed to decrease with longer weathering periods. For
C2-6 m, the ductility reduced by 18% compared to A2-0d whereas
for E2-1y, the reduction is 12% compared to Beam A2-1y.
For the large beams, the ductility was found to decrease in tan-
dem with the strength behavior, by 50% for Beam C1-1y with
qfrp = 0.64% compared to 15% for Beam C3-1y with qfrp = 1.92% at
the end of 1 year accelerated weathering (Fig. 9a).
The failure mode of the beams changed from concrete crushing
(CC) or flexural crack induced FRP debonding (DB) to FRP rupture
(FR) with longer weathering periods. This indicates that the FRP
system has deteriorated due to weathering. For beams strength-
ened with Type 2 GFRP system, the transition of failure mode
was found to take place at an earlier age than that for beams
strengthened with Type 1 GFRP system.
4.4. Comparison of test results with analytical predictions

4.3. Ductility and failure modes
The ductility of the beams, defined as the ratio of the deflection
at ultimate load (Dul) to that at steel yield load (Dyl), was found in
general to decrease with longer weathering period (see Table 3).
After 6 months of accelerated weathering, the ductility of the
small beams with Type 1 GFRP laminates was found to be reduced
by 38%.
The flexural strength of the GFRP-strengthened RC beams sub-
jected to combined sustained loading and weathering for different
time periods was calculated using the following input values. The
values of efrp,u for Type 1 and Type 2 GFRP laminates were taken
as 0.0214 and 0.0194, respectively, as established by Liew [5]. Also,
the values of ue,w were obtained from experimental studies [5] as:
ue;w ¼ a þ b logðte Þ ð12Þ
 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300 299

50 5
ρ frp = 1.92%

Flexural Strength (kN) 40 4

ρ frp = 0.64%
ρ frp = 0.64%
30 3

Ductility
ρ frp = 1.92%
20 2

10 1

0 0

Beam C1

Beam C3
Beam A1

Beam A3
Beam C1

Beam C3
Beam A1

Beam A3

(a) Flexural strength and ductility after 6-year equivalent outdoor weathering

50
ρ frp = 0.64% ρ frp = 1.92%

40

30 A1-0d
Load (kN)

A3-0d

20 C3-1y
C1-1y

10

0
0 10 20 30 40 50 0 10 20 30 40 50
Mid-span deflection (mm)
(b) Load-deflection characteristics after 6-year equivalent outdoor weathering
Fig. 9. Static behavior of large RC beams strengthened with GFRP laminates.

Chamber Period (Days) Chamber Period (Days)

0 30 60 90 120 150 180 0 30 60 90 120 150 180
1.2 1.2
Normalized Flexural Strength

R2 = 0.45
1.0
Normalized Flexural Strength

1
1.0 1

0.8 0.8
R2 = 0.65

0.6 0.6

0.4 0.4
(a) RC Beams Strengthened Chamber Chamber
with Type 1 (Unidirectional) Exterior/Outdoor (b) RC Beams Strengthened Exterior/Outdoor
0.2 GFRP System 0.2 with Type 2 (Bidirectional)
Ambient Ambient
Analytical GFRP System Analytical
0 0
0 300 600 900 1200 0 300 600 900 1200
Outdoor Period (Days) Outdoor Period (Days)

Fig. 10. Degradation in flexural strength.

where te = period of weathering (in days); and (a, b) = (1.003, The calculated values are compared with the test results in
0.147) and (0.90, 0.151) for Type 1 GFRP laminates under Fig. 10. The analytical approach showed reasonable correlation with
outdoor and accelerated weathering, respectively, with correspond- the test results except for Beams C1-2 m and C1-3 m, which are
ing values of (0.993, 0.0133) and (1.047, 0.011) for Type 2 GFRP considered as outliers as explained before. Also, in general, the
laminates. approach predicted the failure mode accurately as shown in Table 3.
300 K.H. Tan et al. / Cement & Concrete Composites 31 (2009) 290–300
5. Conclusions
From the analytical and experimental investigations carried out
on glass FRP-strengthened RC beams under the combined effect of
sustained loading and tropical weathering, the following conclu-
sions can be drawn:
1. FRP-strengthened RC beams under sustained loads exhibited
larger deflections and crack widths, when subjected to tropical
weathering at the same time. They showed smaller deflections
and crack widths when strengthened with a higher FRP rein-
forcement ratio.
2. Both the strength and ductility of beams under sustained loads
decreased with the longer weathering periods. However, beams
with more FRP laminates showed less degradation in strength
and ductility. The failure mode changed from concrete crushing
or flexural crack induced FRP debonding to FRP rupture, indicat-
ing degradation in the properties of the FRP system.
3. The proposed analytical approach, utilizing the degraded mod-
ulus of concrete and FRP laminates, gives a reasonably accurate
estimate of the long-term deflection of beams subjected to sus-
tained loading under tropical climate. The degradation in flex-
ural strength and change in failure mode are also predicted
reasonably well.
Acknowledgment
The study was supported by a research Grant R-264-000-140-
112 provided by the National University of Singapore.
References
[1] ACI Committee 440. State-of-the-art report on fiber reinforced polymer (FRP)
reinforcement for concrete structures (ACI440R-07). American Concrete
Institute; 2007.
[2] ACI Committee 440. Guide for the design and construction of externally
bonded FRP systems for strengthening concrete structures (ACI440.2R-08).
American Concrete Institute; 2008.
[3] Almusallam TH, Al-Salloum YA, Alsayed SH, Mosallam AS. Durability and long-
term behavior of reinforced concrete beams strengthened with FRP
composites. In: Teng JG, editor. Proceedings International Conference on FRP
Composites in Civil Engineering, vol. II. Hong Kong: Elsevier; 2001. p. 1579–88.
[4] Leung HY, Balendran RV, Lim CW. Flexural capacity of strengthened concrete
beams exposed to different environmental conditions. In: Teng JG, editor.
Proceedings International Conference on FRP Composites in Civil Engineering,
vol. II. Hong Kong: Elsevier; 2001. p. 1597–606.
[5] Liew YS. Durability of FRP composites under tropical climate. MEng thesis.
Singapore: National University of Singapore; 2003.
[6] ACI Committee 209. Prediction of creep, shrinkage, and temperature effects in
concrete structures. Manual of concrete practice (ACI209R-92). American
Concrete Institute; 1992. p. 92.
[7] Tan KH, Saha MK. Long-term deflections of RC beams externally bonded with
FRP system. ASCE J Compos Constr 2006;10(6):474–82.
[8] Branson DE. Deformations of concrete structures. New York: McGraw-Hill;
1977. p. 167–9.
[9] Holmes M, Just DJ. GRP in structural engineering. Essex: Applied Science
Publishers; 1983. p. 20–3 and 213–29.
[10] Japan Society of Civil Engineers. Test method for tensile properties of
continuous fiber sheets (JSCE-E-541-2000). In: Japan society of civil
engineers. Conc eng series, vol. 41; 2000. p. 91–105.
[11] Bonacci JF, Maalej M. Behavioral trends of RC beams strengthened with
externally bonded FRP. J Compos Constr 2001;5(2):102–13.
[12] Teng JG, Chen JF, Smith ST, Lam L. FRP-strengthened RC structures.
England: John Wiley & Sons Ltd.; 2002.
[13] Saha MK. Long-term behaviour of FRP-strengthened RC beams. PhD thesis.
Singapore: National University of Singapore; 2007.
[14] Tan KH, Paramasivam P, Tan KC. Creep and shrinkage deflections of RC beams
with steel fibers. ASCE J Mater Civil Eng 1994;6(4):394–414.
 Cement & Concrete Composites 31 (2009) 301–308

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Hygro-thermo-chemical modeling of high performance concrete. I: Theory

Giovanni Di Luzio a, Gianluca Cusatis b,*
a
Department of Structural Engineering, Politecnico di Milano, Milan, Italy
b
Department of Civil and Environmental Engineering, 4048 Johnsson Engineering Center, Rensselaer Polytechnic Institute, 110 Eighth St., Troy, NY 12180-3590, USA

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

Article history: This study deals with the formulation, calibration, and validation of a new hygro-thermo-chemical model
Received 12 September 2008 for high-performance concrete (HPC) suitable for the analysis of moisture transport and heat transfer at
Received in revised form 19 February 2009 the early age and beyond. In Part I of this study the theoretical formulation is presented and discussed in
Accepted 25 February 2009
detail. Classical macroscopic mass and energy conservation laws are written by using humidity and tem-
Available online 9 March 2009
perature as primary variables and by taking into account explicitly various chemical reactions, such as
cement hydration, silica fume reaction, and silicate polymerization. The effect of cement hydration is
Keywords:
modeled through the classical concept of hydration degree. Silica fume reaction and silicate polymeriza-
Concrete
Early age
tion are modeled by introducing the degree of silica fume reaction and the concentration of silicate
Moisture polymers, along with their evolution laws. The present model can simulate early age phenomena, such
Temperature as self-heating and self-desiccation, with great accuracy. Numerical implementation, calibration and
Relative humidity validation of the model by comparison with experimental test data are postponed to Part II of this study
Hydration [Di Luzio G, Cusatis G. Hygro-thermo-chemical modeling of high performance concrete. II: Numerical
Silica fume reaction implementation, calibration, and validation. Cem Concr Compos, in press].
Polymerization Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction As a consequence of the aforementioned volume changes, uni-

Durability and serviceability are nowadays two of the most
important issues in concrete constructions. Premature deteriora-
tion of reinforced concrete, due to corrosion of the reinforcing
steel, greatly affects the structural performance, shortens the ser-
vice life of structures and increases maintenance costs. To over-
come this problem, since the early 90s, high-performance
concrete (HPC) mixes, characterized by low permeability and high
strength, have been formulated and used in design. Unfortunately,
HPC constructions are more prone to cracking (especially at the
very early age) which produces a direct path for corrosive agents
to reach the reinforcing steel and completely defeats the beneficial
effect of HPC low permeability. The most common cause of crack-
ing at the early age is the restraint of concrete to volume changes.
Typically three different kinds of volume changes occur in concrete
structures: (1) thermal expansion due to heat released by exother-
mic chemical reactions; (2) shrinkage due to release of water
through external concrete surfaces (drying shrinkage); and (3)
shrinkage due to water consumption (autogeneous shrinkage) dur-
ing cement hydration and other chemical reactions, such as the sil-
ica fume (SF) reaction and silicate polymerization. The effect of
autogeneous shrinkage is small (and usually negligible) for stan-
dard concrete whereas it is significant for HPC.
* Corresponding author. Tel.: +1 518 276 3956; fax: +1 518 276 4833.
E-mail addresses: cusatg@rpi.edu, giacusatis@yahoo.it (G. Cusatis).
form and/or non-uniform stress states develop in concrete even
in the absence of external loads. This is the typical situation occur-
ring in composite bridges, in which the deck, subjected to shrink-
age and restrained by the girders underneath, undergoes tensile
stresses, which in some cases reach the tensile strength leading
to transverse cracking. The overall scenario is complicated by the
fact that, at an early age, concrete strength is under development
and the build-up of strength depends on the evolution of the chem-
ical reactions involved in the aging process, which, in turn, de-
pends on temperature, humidity, and admixture materials.
Reliable and predictive analyses of the behavior of concrete struc-
tures at early ages (and beyond) require the formulation of a con-
sistent theoretical framework that includes all of the various
phenomena characterizing early age evolution of concrete.
Many experimental, theoretical, and computational studies on
the behavior of concrete at the early age and, specifically, on self-
heating and self-desiccation can be found in the literature. Among
others, Wittmann [2] calculated the moisture profiles in high
strength concrete by adding a sink term in the diffusion equation
to account for endogenous drying. In the aforementioned study,
however, the effect of hydration heat was neglected. Ulm and
Coussy [3] developed a thermodynamics based theory for the
description of cement hydration. This theory was refined by Cerve-
ra et al. [4], who proposed a coupled thermo-chemo-mechanical
model to simulate the aging effect on strength and damage. The fo-
cus of these studies was mainly the evolution of hydration, and

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.015
302 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
hygral phenomena (which are important for HPC) were neglected.
Also Bažant et al. [5] developed a comprehensive material model
for the simulation of cracking at early age considering only temper-
ature effects. More recently Gawin et al. [6] proposed a refined
model based on the interpretation of concrete as a multi-phase
porous material and on the use of homogenization techniques.
Such model has showed a good capability in fitting experimental
results, but it does not take into account the effect of SF reaction
and silicate polymerization. Similarly, only hydration was modeled
in the study by Oh and Cha [7], in which a model for moisture and
temperature calculation in early age concrete was developed.
Typical HPC mixes are characterized by low (<0.4) water-to-ce-
ment ratio, high cement content, and, most importantly, by the
incorporation of SF or other pozzolanic materials [8]. The pozzola-
nic reaction associated with the presence of SF determines a denser
microstructure and leads to increased performance of concrete in
terms of strength and durability. In addition, SF is known to accel-
erate silicate polymerization [9], which is a gradual bonding of par-
ticles of calcium silicates hydrates (C–S–H). This increases the
average degree of polymerization of the silicate chains and causes
the so-called chemical aging of C–S–H [10]. Both pozzolanic reac-
tion and silicate polymerization interact with cement hydration
[9] influencing concrete mesostructure and, ultimately, concrete
behavior. None of the analytical models currently available in the
literature can simulate the effect of SF reaction and polymerization
on heat transfer and moisture transport in concrete.
The objective of the present study is to overcome the limita-
tions mentioned above and to simulate moisture transport and
heat transfer in HPC by taking into account explicitly the coupled
effect of cement hydration, SF pozzolanic reaction, and silicate
polymerization. In the present Part I, only the theoretical formula-
tion will be presented. The subsequent Part II [1] will present the
numerical implementation and will report the results of extensive
calibration and validation activities performed on the basis of
experimental data available in the literature.
2. Moisture transport governing equation
Water can be found in concrete in various phases [11–13] that
can be classified as follows: (a) capillary water, (b) water vapor, (c)
adsorbed water, and (d) non-evaporable (chemically bound) water.
The capillary water is liquid water filling capillary pores (empty
volume among cement particles) and its state can be described
through the capillary pressure, which is governed by Kelvin’s equa-
tion [14]. If capillary pores are not fully saturated, the capillary
water is in equilibrium with air and water vapor, which can be as-
sumed to behave as incompressible gases [14]. Near pore walls,
van de Waals forces retain water molecules and cause them to
form thin adsorbed water layers. The thickness of the adsorbed lay-
ers is an increasing function of the relative humidity h. In small mi-
cro-pores (gel pores) of few Å the adsorbed water layers cannot
fully develop for high relative humidity conditions and, for this
reason, they are referred to as hindered adsorbed water layers [13].
The mobility of water along adsorbed water layers is rather lim-
ited (especially for hindered adsorbed water) and the transport
mechanism is completely different compared to the transport of
capillary water and water vapor. In addition, all moisture transport
mechanisms are influenced by concrete microstructure (in particu-
lar the microstructure of calcium silicates hydrates), which, in turn,
depends on the extent of the chemical reactions that characterize
concrete at early ages. Moreover, these chemical reactions often in-
volve water, which may be subtracted from the system (e.g. during
cement hydration) or released into the system (e.g. during silicate
polymerization). When water is chemically combined with other
chemical compounds, it looses its mobility completely and it cannot
contribute to moisture transport. In the literature, the chemically
bound water is typically referred to as non-evaporable water be-
cause it is retained even at temperature exceeding 105 °C.
In principle, the different water transport mechanisms should
be modeled independently through the formulation of separate
diffusion equations since each single mechanism has its own driv-
ing force (capillary pressure for the capillary water, vapor pressure
for the water vapor, etc.). However, the complexity of the phenom-
ena hampers this kind of approach and calls for a simplified
approximated analysis. First of all, it is possible to simplify the
problem by postulating the existence of local thermodynamic
equilibrium ([13,15,6], among others). This assumption approxi-
mately holds also in presence of a chemical evolution of the system
(aging), since the characteristic times of the water transport mech-
anisms are much smaller than those associated with concrete
chemical reactions. In Gawin et al. [6] two separate diffusion equa-
tions were used to model separately the transport of water vapor
and the transport of capillary and adsorbed water. Although the
approach is certainly worth mentioning, it seems that the in-
creased complexity (and the increased number of material param-
eters to be calibrated) is not compensated by more accuracy in
modeling experimental evidence. In addition, modeling capillary
water transport and adsorbed water (free and hindered) transport
through only one equation [6] is an approximation of the same or-
der (at least at standard room temperatures) as combining phe-
nomenologically all the transport mechanisms in one single
phenomenological macroscopic equation ([13,16,17], among many
others).
In this study, as widely accepted in the literature, the latter op-
tion is exploited and the overall moisture transport process under
isothermal conditions is described through the Fick’s law that ex-
presses the flux of water mass per unit time J as proportional to
the spatial gradient of the relative humidity h
J ¼ Dh ðh; TÞ$h ð1Þ
The proportionality coefficient Dh ðh; TÞ is called moisture perme-
ability and it is a nonlinear function of the relative humidity h and
temperature T [16]. The choice of the relative humidity h as the
state variable of the problem (in addition to the temperature T)
is done here for convenience but other options can be equivalently
adopted [15].
The moisture mass balance requires that the variation in time of
the water mass per unit volume of concrete (water content w) be
equal to the divergence of the moisture flux J
@w
 ¼$J ð2Þ
@t
The water content w can be expressed as the sum of the evap-
orable water we (capillary water, water vapor, and adsorbed water)
and the non-evaporable (chemically bound) water wn [18,19]. For
reasons discussed later in this paper, it is reasonable to assume
that the evaporable water is a function of relative humidity, h, de-
gree of hydration, ac , and degree of silica fume reaction, as , i.e.
we ¼ we ðh; ac ; as Þ = age-dependent sorption/desorption isotherm
[20]. Under this assumption the water content variation in time
reads
@w @we @wn @we @h @we @w
¼ þ ¼ þ a_ c þ e a_ s þ w_ n ð3Þ
@t @t @t @h @t @ ac @ as
where the total time derivatives of ac ; as , and wn are used instead of
the partial derivatives in order to stress the fact that these variables
will serve as internal variables, as opposed to results from the solu-
tion of the boundary value problem. Note that total and partial
derivatives of these quantities coincide since the mass conservation
equation is written with reference to the initial (undeformed) spa-
tial configuration.
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308 303

By substituting Eqs. (1) and (3) into Eq. (2) one obtains ~ 1 ) is constant for a given concrete, then
unit of hydrated mass (Q c
one can write [4]
@we @h
$  ðDh $hÞ   F h ðh; ac ; as ; ap Þ ¼ 0 ð4Þ e1
@h @t Q_ c ¼ a_ c c Q c ð8Þ
and where a_ c is the rate of the hydration degree discussed in Section 4
@we @w and c is the cement mass content.
F h ðh; ac ; as ; ap Þ ¼ a_ c þ e a_ s þ w_ n ð5Þ The temperature rise due to cement hydration (commonly re-
@ ac @ as
ferred to as self-heating) can be substantial, especially in massive
where @we =@h is the slope of the sorption/desorption isotherm (also structures; temperature increase up to 60 °C has been reported,
called moisture capacity), ap is the concentration of silicate poly- for example, in [24,26].
mers, and the function F h ðh; ac ; as ; ap Þ expresses the effect of con- Similarly to the hydration heat, also the heat generated by SF
crete aging on the diffusion process. In particular, the first two reaction can be expressed as [25]
terms account for the microstructural changes associated with con-
crete chemical reactions and the last term accounts for the internal e1
Q_ s ¼ a_ s s Q ð9Þ
s
‘‘consumption” of water involved in the chemical reactions. When
where a_ s is the rate of pozzolanic reaction degree, which will be dis-
only cement hydration is accounted for, this last term is called mois- e 1 is the la-
cussed later in Section 5, s is the SF mass content, and Q s
ture sink and it explains the well-known self-desiccation phenome-
tent heat of SF reaction per unit mass of reacted SF.
non observed in high-performance concrete. For old enough normal e 1 , can be measured by conducting
The hydration enthalpy, Q
concrete one has F h ðh; ac ; as ; ap Þ  0, whereas Fðh; ac ; as ; ap Þ–0 for c
calorimetric tests at complete hydration; typical values range from
early age normal concrete and, in general, for HPC.
400 kJ/kg to 550 kJ/kg, depending upon cement composition [24]. If
The governing equation (Eq. (4)) must be completed by appro- e 1 can be
the exact phase composition of the cement is known, Q c
priate boundary and initial conditions. The initial conditions spec-
determined from weighted average of the heat of hydration of each
ify the relative humidity in the whole problem domain at time
single compound
instant t ¼ 0 and the boundary conditions specify either the
relative humidity on the boundary C; hðtÞjC ¼ hðtÞ  (Dirichlet’s type) e 1 ¼ Q 1 W C 3 S þ Q 1 W C2 S þ Q 1 W C 3 A þ Q 1 W C4 AF þ   
Q c C3 S C2 S C3 A C 4 AF ð10Þ
or the moisture flux orthogonal to the boundary ðDh $hÞ  njC ¼
ah ðh  h1 Þ (Cauchy’s type). where Q1
is the individual enthalpy of reaction for each compound,
i
i
and W is the initial percentage weight for each compound.
The enthalpy of SF reaction, Q e 1 , can be assumed to be constant
3. Temperature evolution due to exothermic chemical reactions s
and equal to 780 kJ/kg [26].
By substituting Eqs. (6), (8), and (9), into Eq. (7) one obtains
The coefficients of the diffusion equation presented in Section 2
depend on temperature, therefore the temperature field needs to @T
$  ðk$TÞ  qct þ F T ðac ; as Þ ¼ 0 ð11Þ
be calculated. Note that, since the chemical reactions associated @t
with cement hydration and SF reaction are exothermic, the tem- e 1 þ a_ c c Q
e 1 . For old enough concrete
where F T ðac ; as Þ ¼ a_ s s Q s c
perature field is not uniform for non-adiabatic systems even if
F T ðac ; as Þ  0.
the environmental temperature is constant.
The energy conservation equation (Eq. (11)) must be completed
Heat conduction can be described in concrete, at least for tem-
by appropriate boundary and initial conditions. The initial condi-
perature not exceeding 100 °C [21], by Fourier’s law, which reads
tions specify the temperature in the whole problem domain at
q ¼ k$T ð6Þ time instant t ¼ 0 and the boundary conditions specify either the
temperature at the boundary C; TðtÞjC ¼ TðtÞ  (Dirichlet’s type)
where q is the heat flux, T is the absolute temperature, and k
or the heat flux orthogonal to the boundary ðk$TÞ  njC ¼
is the heat conductivity. For the temperature range considered
aT ðT  T 1 Þ (Cauchy’s type).
in this study ðT < 100  CÞ, the heat conductivity k of concrete
is known to depend mildly on temperature, moisture content,
and aging [22]. However, many authors in the past have ob- 4. Cement hydration
tained good numerical predictions by neglecting this depen-
dence [16]. For the sake of simplicity, the same assumption is Cement hydration is characterized by the reaction of free-water
made herein. An average value of k ¼ 2:3 W=m  C [21] will be with unhydrated cement particles, which are mainly composed by
used for the numerical simulations presented in Part II of this calcium silicates (C3 S and C2 S), calcium aluminates ðC3 AÞ, and cal-
study. cium aluminoferrites ðC4 AFÞ. The products (cement hydrates) of
The enthalpy balance equation can be written as this reaction are mainly calcium silicates hydrates (C–S–H), cal-
cium hydroxide (CH), ettringite (Aft), and monosulfate (Afm) [9].
@T
qct ¼ $  q þ Q_ c þ Q_ s þ Q_ p ð7Þ Although cement hydration has been studied for many years
@t [9,22], the exact stoichiometry of the chemical reactions involved
where q is the mass density of concrete, and ct (ranging from 840 to is not known and their mutual interaction as well as the influence
1170 J/kg °C [22]) is the isobaric heat capacity (specific heat) of con- of external factors (such as temperature, relative humidity, etc.) is
crete. An average value of ct ¼ 1100 J=kg  C can be generally not completely understood. As a consequence, in the literature,
adopted for all concretes. Q_ c ; Q_ s ; and Q_ p represent the rate of heat studies that consider each chemical reaction independently are
generation per unit volume due to cement hydration, SF reaction, very limited [7] whereas many authors [3,25,27–29] have studied
and silicate polymerization, respectively. Since, to the best of the cement hydration with reference to an overall hydration process.
authors knowledge, there is no evidence of temperature variation The present study adopts the latter approach.
associated with the polymerization of silicates, Q_ p ¼ 0 is adopted When a cement grain is immersed in water, a layer of soft gel
hereinafter. (composed by cement hydration products) forms on the surface
If one assumes, as commonly accepted in the literature of the grain. As hydration progresses, this layer grows both inward
[3,25,4,6] and confirmed by various theoretical and experimental and outward around a hard core of unhydrated cement. In order for
investigations [23], that the latent heat of hydration reaction per the hydration to continue, water needs to reach the unhydrated
304 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
grain core and therefore the water diffusion process through the
layer of hydrated compounds is typically considered the dominant
mechanisms governing the hydration kinetics. According to the
thermodynamics based model proposed by Ulm and Coussy
[3,28] and later revisited by Cervera et al. [4], the hydration kinet-
ics can be described by postulating the existence of a Gibb’s free
energy dependent on the external temperature T and the hydration
extent vc . The latter represents, in the case of a closed chemical sys-
tem, the number of chemically combined moles of water per unit
volume of concrete and it expresses the thermodynamic imbalance
between the chemical constituents. The hydration extent can be
also used to define a normalized measure of the hydration process,
the hydration degree, as
vc
ac ¼
v1
c The exact stoichiometry of pozzolanic reaction between SF and
where v1 c is the theoretical asymptotic values of the hydration ex-
tent in ideal hygrometric conditions. The actual asymptotic value of
hydration extent, v1 c , is always smaller than the theoretical value,
v1c , since, in practice, ideal hygrometric conditions are never met.
The ratio between the actual and the theoretical asymptotic values
is the asymptotic degree of hydration a1 c ¼ vc =vc , which has been
1 1
shown to be always smaller than the unity [24].
By assuming that the thermodynamic force conjugate to the
hydration extent, called the chemical affinity, is governed by an
Arrenius-type equation and that the viscosity governing the diffu-
sion of water through the layer of cement hydrates is an exponen-
tial function of the hydration extent [3], Cervera et al. [4] proposed
the following evolution equation for the hydration degree
a_ c ¼ Ac ðac ÞeEac =RT
and
 
Ac2   1
gc ac =ac
Ac ðac Þ ¼ Ac1 þ ac a1
c  ac e
ac
1
where Ac ðac Þ is the so-called normalized chemical affinity, Eac is the
hydration activation energy, R is the universal gas constant, and
gc ; Ac1 ; and Ac2 are material parameters. The ratio Eac =R can be
experimentally determined and it ranges from 3000 to 8000 K for
concrete. In the present study the value Eac =R ¼ 5000 K will be
adopted when specific measurements are not available. The param-
eters gc ; Ac1 ; and Ac2 can be calibrated, for example, by simulating
the temperature evolution during adiabatic tests as it will be dis-
cussed in Part II of this study [1].
Moreover, it is well known from experiments that if the relative
humidity decreases below a certain value ðh  75%Þ, the hydration
process slows down and may even stop [12,30]. This phenomenon
can be taken into account [6] by rewriting Eq. (13) as
a_ c ¼ Ac ðac Þbh ðhÞeEac =RT
where
bh ðhÞ ¼ ½1 þ ða  ahÞb 1
The function bh ðhÞ is an empirical function that was first pro-
posed for the definition of the equivalent hydration period by Ba-
žant and Prasannan [30]. The parameters a and b can be
calibrated through the analysis of experimental data but the con-
stant values a ¼ 5:5 and b ¼ 4 can be generally adopted [6,30].
5. Silica fume reaction
A certain amount of SF is often added in concrete mixture in or-
der to produce HPC [31,32]. SF belongs to the category of highly
pozzolanic materials and it consists of silica in noncrystalline form
with a high specific surface and exhibiting high pozzolanic activity.
From the chemical–physical point of view, three principal
mechanisms have been suggested in the literature to explain SF ef-
fects on concrete: (1) SF particles cause pore blockage in the
hydrating cement producing a denser hydrating gel structure
[33]. (2) Pozzolanic reactions occur between the silica-rich SF par-
ticles and the portlandite (CH) generating calcium silicate hydrates
(C–S–H) [34]. (3) SF particles act as nucleation sites for cement
hydration, accelerating the hydration process [35]. While the CH
is relatively weak and brittle, the C–S–H phase formed by the poz-
zolanic reaction has a microscopic morphology similar to that
developed by cement hydration [36]. Consequently, the additional
C–S–H produced by the pozzolanic reaction leads to higher
strength, lower porosity, and improved chemical resistance [8].
ð12Þ
CH is not known exactly, especially in relation to the water stoichi-
ometry coefficient. On this issue the current literature shows disso-
nant results. According to Sellevold [37] the amount of water
bound in the C–S–H formed by the pozzolanic reaction should be
the same as that contained in the CH. Other experimental investi-
gations [38], however, suggest that the chemically bound water
per gram of cement increases in cement pastes containing SF. In
contrast to these findings, Li et al. [39] found a reduction of chem-
ically bound water in concrete mixes containing SF. In addition,
experimental results by Papadakis [40] suggest that the reaction
takes place without additional water.
A possible reason for the contradictory results reported above is
that the presence of SF accelerates and amplifies the process of
polymerization of C–S–H gel phase [41] which leads to a release
of the water previously combined during cement hydration (reduc-
ð13Þ tion of the chemically bound water) and a consequent increase of
the free water [42]. In the present study, C–S–H polymerization
will be addressed separately as discussed in Section 6 and it is as-
sumed that no water is involved (consumed or released) in the SF
ð14Þ pozzolanic reaction (Section 8).
The effect of SF on moisture transport phenomena is here mod-
eled through the introduction of the degree of SF reaction, as , de-
fined as the ratio between the amount of reacted SF and the total
amount of SF. Since the kinetics of the pozzolanic reaction can be
assumed to be a diffusion controlled process [43–45], the theory
adopted in Section 4 to describe cement hydration can be exploited
to describe the evolution of SF reaction [48]. Accordingly, one has
a_s ¼ As ðas ÞeEas =RT ð17Þ
and
 
As2 gs as =a1
As ðas Þ ¼ As1 þ as ða1
s  as Þe
s ð18Þ
as 1
where As is the SF normalized affinity, Eas is the activation energy of
ð15Þ SF reaction, and a1 s is the asymptotic value of SF reaction degree.
Eas =R ¼ 9700 K can be generally assumed [24]. The material param-
eters As1 ; As2 ; and gs can be calibrated similarly to the parameters
ð16Þ governing cement hydration [1].
Contrarily to cement hydration, the pozzolanic reaction of the
SF seems to be less sensitive to the relative humidity and it is able
to continue to near completion even in presence of very low rela-
tive humidity [46,47]. Consequently, the rate of SF reaction degree
is assumed not to depend on h.
When a sufficient amount of SF is available in concrete, all the
CH produced by cement hydration can be consumed by the pozzo-
lanic reaction. For a water-to-cement ratio, w=c, of about 0.5 (nor-
mal concrete), about s/c = 0.16 is required to consume all the
calcium hydroxide during the pozzolanic reaction [49]. For lower
values of w=c, the required amount of SF, ðs=cÞreq , to consume all
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
the calcium hydroxide is reduced proportionally and the propor-
tionality coefficient of 0.4 can be used as confirmed by experiments
on carbonation shrinkage [49]. One can write
ðs=cÞreq ¼ minð0:16; 0:4w=cÞ
If the amount of SF exceeds the minimum required amount of
SF to consume all the calcium hydroxide, the asymptotic degree
of SF reaction is smaller. Assuming a linear relationship between
a1
s and the inverse of s=c, the asymptotic degree of SF reaction de-
gree can be estimated as
 
eff ðs=cÞreq
a1
s ¼ SF min 1;
s=c
where SF eff is the ‘‘efficiency” of SF, i.e. the mass ratio between the
SiO2 mass content and the total mass of SF. Since only the SiO2 par-
ticipates to the pozzolanic reaction, the remaining part of SF does
not participate in any chemical reaction and has the role of an inert
filler. Typically, SF eff ranges from 0.85 to 0.92 [9].
6. Silicate polymerization
Many experimental investigations [9] revealed that C–S–H par-
ticles bond together over time, increasing the average degree of
polymerization of the silicate chains and causing the C–S–H to be-
come stiffer, stronger, and denser. The key reaction for linking the
C–S–H gel hydrolyzed molecules together is the condensation. In
condensation reactions, covalent bonds are rearranged in such a
way that two monomers are connected to form a polymeric cluster
and water is ‘‘condensed” out.
Portland cement contains, among other things, silicates, which
are minerals that contain silicate anion and various cations (e.g.
calcium, sodium, magnesium, iron, aluminum ions). In presence
of water, silicates become hydrated and some of the oxygens form
a covalent bond with a hydrogen nucleus from the water. Since the
hydrated silicates contain Si–OH bonds, condensation is possible
along with the creation of a polymerized silicate framework. This
process occurs slowly at standard room temperature, but it can
be greatly accelerated by curing at elevated temperature and it
may be induced also by drying [50,51]. This chemical aging process
of C–S–H can thus affect many of the physical properties of cement
paste and concrete, such as permeability, strength capacity, shrink-
age and creep behavior [52].
Silicate polymerization is much more pronounced in the pres-
ence of SF because the SF pozzolanic reaction increases the content
of C–S–H gel. When the water consumed by cement hydration is
less than the water condensed by silicate polymerization, one
can observe a reduction of the non-evaporable water content start-
ing after 2–3 months [49,54]. Zhang and Gjørv [54] observed, for
cement paste containing SF and characterized by various water-
to-binder ratios, an increase of non-evaporable water content for
up to 90 days, and a reduction of the non-evaporable water content
after 550 days. This reduction was higher for higher SF content.
This implies that, even if for standard concrete silicate polymeriza-
tion can be neglected, it needs to be considered in presence of SF.
Silicate products of portland cements will generally be polymer-
ized over time with an increase of the average chain length of
polysilicates in C–S–H gel and with water release. The polymeriza-
tion process can be modeled through the relative concentration of
silicate polymers, ap , defined as the ratio between the mass of
polymeric species and the mass of C–S–H gel. The rate of the rela-
tive concentration of silicate polymers can be formulated as
a_ p ¼ Ap1 Ap ðap ÞBp ðac ÞeEap =RT
and
0
Ap ðap Þ ¼ a1
p  ap ; Bp ðac Þ ¼ Hðac  ac Þ
305
where the term Ap1 eEap =RT describes the kinetics of the chemical
reaction, the coefficient Ap ðap Þ expresses the fact that the rate of
polymerization is reduced as polymerization progresses and it
eventually vanishes when the polymerization is complete
ð19Þ
(ap ¼ a1 p , where ap ¼ 1), the coefficient Bp ðac Þ takes into account
1
the effect of cement hydration (hydration products need to be pres-
ent for the polymerization to occur), and HðÞ is the Heaviside step
function, used here to set a threshold value of hydration degree, a0c ,
below which the polymerization process can be neglected. Accord-
ing to the experimental results of Hirljac et al. [50] on the alite
ðC3 SÞ, the value of a0c ¼ 0:1 can be adopted. Brough et al. [53] found
ð20Þ an Arrhenius energy of 35 kJ/mol for dimeric reactions and of
100 kJ/mol for reactions producing higher polymeric species. In this
study it is assumed an averaged value of 50 kJ/mol, which gives
Eap =R ¼ 6000 K.
The material parameter Ap1 can be calibrated on the basis of
experimental data on the evolution of the polymerized species,
as discussed in Part II of this study [1].
7. Asymptotic hydration degree
As mentioned earlier the asymptotic (ultimate) hydration de-
gree, a1c , is always less than one [24]. This is due to the fact that
the unit value can be reached only in ideal conditions, i.e. with
an adequate water-to-cement ratio to ensure full hydration and
perfect contact between the water and the cement grains. In prac-
tice, these conditions are never fulfilled during curing and com-
plete hydration of concrete is never achieved [12,55].
Pantazopoulo and Mills [19] proposed to calculate a1 c on the ba-
sis of a mass balance between water needed to ensure saturation
conditions and total water content. At infinite time, this balance
equation reads
w0 þ Dw1 1 1
c ¼ we þ wn ð23Þ
where w0 ¼ ðw=cÞc is the initial water content (neglecting the
amount of entrained air content), w1 1
e and wn are the asymptotic
evaporable water and chemically bound water, respectively. Dw1 c
is the amount of water that needs to be provided in order to main-
tain full saturation conditions during hydration and that corre-
sponds to the so-called chemical shrinkage caused by the fact that
hydration products have a lower volume than the sum of anhydrous
cement and water [19]. Experimental studies [18] have shown that
per each gram of hydrated Portland cement: (a) approximately
0.253 g of water is chemically bound ðwn ¼ 0:253cac Þ; (b) the
evaporable water is proportional to the chemically bound water
and the coefficient of proportionality can be estimated as
3:83w=cðwe  3:83ðw=cÞwn Þ; and (c) the amount of water that
needs to be supplied to overcome the chemical shrinkage and main-
tain saturation is Dwc ¼ 0:065cac . Based on these observation and
by using Eq. (23) Pantazopoulo and Mills [19] obtained
1:031w=c
a1
c ¼ ð24Þ
0:194 þ w=c
The presence of SF further reduces the ultimate achievable
hydration degree because, due to the formation of additional
C–S–H gel from the pozzolanic reaction, extra water is hindered
into the gel pores and it becomes unavailable for cement hydration
[24]. In addition, the specific surface of the C–S–H gel increases and
unhydrated cement grain cores become less accessible. Note that
this does not imply that the overall amount of C–S–H created is
less (actually the opposite is true) because the C–S–H coming from
ð21Þ
the SF pozzolanic reaction must be added to the one coming from
cement hydration.
The effect of SF on the water balance equation at saturation (Eq.
ð22Þ (23)) can be taken into account by considering (1) on the right
306 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
hand side, an increased quantity of water held in the gel due to the
pozzolanic reaction plus water released during silicate polymeriza-
tion (roughly 0.50 g/g of reacted SF according to [56]) and (2), on
the left hand side, the amount of water that needs to be provided
in order to maintain saturation conditions and that corresponds
to the chemical shrinkage caused by pozzolanic reactions (roughly
0.22 g/g of reacted SF according to [56]). The SF pozzolanic reaction
is considered not to consume water as discussed in Section 5.
Based on these observations, the balance equation (Eq. (23)) can
be rewritten as
w0 þ 0:065ca1 1 1 1
c þ 0:22sas ¼ 0:253ð3:83w=c þ 1Þcac þ 0:5sas
where s is SF content and a1 s is the asymptotic SF reaction degree
given by Eqs. (19) and (20).
Solving Eq. (25) for the asymptotic degree of cement hydration,
one obtains
1:032w=c  0:279ðs=cÞa1
s
a1
c ¼
0:194 þ w=c
The asymptotic degree of hydration must be positive and so Eq.
(26) holds only if w=c > 0:27ðs=cÞa1
s . This condition is always ful-
filled by typical HPC mixes (e.g. for w=c ¼ 0:3 and s=c ¼ 0:1 one
gets 0:3  0:279  0:1  0:028).
8. Non-evaporable water
As discussed previously, the non-evaporable water is the
amount of water that is chemically bound as consequence of ce-
ment hydration, SF reaction, and silicate polymerization. In gen-
eral, one can write
wn ðac ; as ; ap Þ ¼ jc ac c þ js as s  ap ðjpc c þ jps sÞ
The first term represents the effect of cement hydration and jc
is the mass ratio of non-evaporable water at full hydration. It can
be assumed jc ¼ 0:253 [12,18,20]. The second term represents
the effect of SF reaction and, according to the discussion in Section
5, it can be assumed js ¼ 0. The third term characterizes the effect
of polymerization and the negative sign traduces the fact that dur-
ing silicate polymerization water is released in the system and the
chemically bound water is reduced. Experimental studies showed
that the addition of SF increases the rate of polymerization of C–
S–H gel (for a C3 S paste at an age of 28 days, the mean chain length
was 2.6 in the absence of SF, and 3.6 if SF was present [57]) and
that the C–S–H gel formed by the SF reaction is highly polymerized
[58]. Because of these reasons, for regular concrete mixes without
the addition of SF the effect of the polymerization is rather small
and it can be neglected by assuming jpc  0. On the contrary the
value of jps can not be assumed equal to zero and must be deter-
mined from the analysis of experimental data [1].
9. Sorption/desorption isotherm and moisture capacity
The relation between the amount of evaporable water and rel-
ative humidity is called ‘‘sorption isotherm” if measured with
increasing relativity humidity and ‘‘desorption isotherm” in the
opposite case. Even though sorption and desorption isotherms
are not exactly the same [59] their difference is generally neglected
[60]. For this reason, in the following, ‘‘sorption isotherm” will be
used with reference to both sorption and desorption conditions.
The shape of the sorption isotherm for HPC is influenced by many
parameters, especially those that influence extent and rate of the
chemical reactions and, in turn, determine pore structure and pore
size distribution (water-to-cement ratio, cement chemical compo-
sition, SF content, curing time and method, temperature, mix addi-
tives, etc.).
In the literature various formulations can be found to describe
the sorption isotherm of normal concrete. The most popular are
the Brunauer–Emmett–Teller (BET) model [61], derived from sta-
tistical thermodynamics of adsorption, and modifications of the
BET equation (the BDDT model [62]; the FHH model [63]; Hiller-
borg’s formula [64]; and the BSB model [65]).
However, in the present paper the semi-empirical expression
proposed by Norling Mjornell [20] is adopted because it explicitly
accounts for the evolution of hydration reaction and SF content.
This sorption isotherm reads
 
1
we ðh; ac ; as Þ ¼ G1 ðac ; as Þ 1  10ðg 1 a1
e c ac Þh
ð25Þ
 10ðg 1 a1
þ K 1 ðac ; as Þ e c ac Þh 1 ð28Þ
where the first term (gel isotherm) represents the physically bound
(adsorbed) water and the second term (capillary isotherm) repre-
sents the capillary water. This expression is valid only for low con-
tent of SF. This also justifies the fact that the evaporable water
ð26Þ content does not depend on the relative concentration of silicate
polymers, ap . The material parameter g 1 > 1 governs the shape of
the sorption isotherm.
The coefficient G1 ðac ; as Þ represents the amount of water per
unit volume held in the gel pores at 100% relative humidity, and
it can be expressed [20] as
c s
G1 ðac ; as Þ ¼ kv g ac c þ kv g as s ð29Þ
c s
where kv g and kv g are material parameters.
The maximum amount of water per unit volume that can fill all
pores (both capillary pores and gel pores), we1 ðac ; as Þ
¼ we ðh ¼ 1; ac ; as Þ, can be calculated, for a certain degree of
hydration, by subtracting from the initial water content the differ-
ð27Þ ence between the chemically bound water and the amount of
water that needs to be provided to compensate chemical shrinkage
while maintaining saturation conditions (0.253 g/g–0.065 g/g =
0.188 g/g for cement and 0 g/g–0.22 g/g = 0.22 g/g for SF, see Sec-
tion 7). One can write
we1 ðac ; as Þ ¼ w0  0:188cac þ 0:22sas ð30Þ
By using Eqs. (30) and (28) with h ¼ 1, and solving for K 1 ðac ; as Þ
one obtains

1
w0  0:188ac c þ 0:22as s  G1 1  e10ðg1 ac ac Þ
K 1 ðac ; as Þ ¼ 1 ð31Þ
e10ðg1 ac ac Þ  1
c s
The material parameters kv g ; kv g , and g 1 can be calibrated by fit-
ting experimental data relevant to free (evaporable) water content
in concrete at various ages [1,20].
The moisture capacity @we =@h (derivative of the sorption iso-
therm to respect to h) that is needed in Eq. (1) can be calculated
numerically.
10. Permeability
Moisture diffusion in concrete depends on various diffusion
mechanisms that characterize the mobility of different water
phases present in concrete and that are influenced by the pore
structure of concrete. Three main distinct transport mechanisms
may be identified: molecular diffusion, Knudsen diffusion, and sur-
face diffusion [66,67].
The molecular diffusion process dominates in macro-pores,
which are characterized by a diameter size from 50 nm to 10 lm
[68]. Pores of this size constitute only a small portion of concrete
porosity and, therefore, molecular diffusion occurs only occasion-
ally and cannot be considered the dominant mechanism [67].
The largest portion of concrete pores is composed of
meso-pores (characteristic size from 25 to 500Å) and micro-pores
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
(characteristic size <25 Å). In these pores, frequent collisions
among water molecules as well as between water molecules and
pore walls provide the main diffusion resistance (Knudsen diffu-
sion), for which the diffusion resistance is related to pore size.
The surface diffusion process (movement of water molecules
along layers of adsorbed water) occurs in certain meso-pores and
micro-pores featuring parallel walls and it poses much greater
resistance to transport than Knudsen diffusion for typical concrete
pore sizes. Thus, in general, the surface diffusion can be neglected
in comparison to Knudsen diffusion unless most of the water is ad-
sorbed water, as it occurs at very low relative humidity.
Although each individual mechanism is reasonably understood,
it is not always easy to make an accurate prediction of the total dif-
fusivity because it is difficult to describe and simulate the details of
concrete pore structure formed by randomly oriented pores with
variable pore radii and with various degree of tortuosity and con-
nectivity. In addition, the scenario is further complicated by the
dependence of the pore structure on water-to-cement ratio, ce-
ment content, SF content, age, etcetera.
In this work, as typically done in the literature [16], the above
mechanisms are described phenomenologically through Fick’s
law (Eq. (1)) which is characterized by the moisture permeability.
The proposed formula for the moisture permeability is given by
   1
D1
Dh ðh; TÞ ¼ wðTÞD1 1 þ  1 ð1  hÞn
D0
where
3
e0 w s 2
2:5
e1 w s2
D0 ¼ D 1 ; D1 ¼ D 1
c c c c
and
 
Ead Ead
wðTÞ ¼ exp  ð34Þ
RT 0 RT
In Eq. (32), the coefficient wðTÞ takes into account the effect of
temperature on permeability [69] as earlier proposed by Bažant
and Najjar [16], D0 and D1 represent moisture permeability for a
completely dry situation ðh ¼ 0Þ and completely saturated situa-
tion ðh ¼ 1Þ, respectively. The exponent n governs the rate at which
permeability transitions from D0 to D1 . The exponents of the power
laws in Eq. (33) have been obtained by analyzing drying data gath-
ered from the literature [1]. In Eq. (34), T is the absolute tempera-
ture, T 0 is the reference room temperature ðT 0 ¼ 296KÞ and
Ead =R  2700 K [16].
e 1 , and n should, in
e 0; D
The moisture permeability parameters D
principle, depend on aging but, at this point in time, there is not
sufficient experimental evidence to investigate this dependence.
A good fitting on a large range of experimental data can be
achieved, as presented in Part II of this study [1], by neglecting
the aging dependence. This assumption was used successfully by
Bažant and coworkers in [66].
e 1 , and n can be calibrated by fit-
e 0; D
The material parameters D
ting experimental data relevant to moisture diffusion as discussed
in Part II of this study [1].
11. Closing remarks
In this study a new hygral-thermo-chemical model for the sim-
ulation of moisture transport and heat transfer in HPC has been
formulated. The model is characterized by the following features:
1 Moisture transport is described by one diffusion equation gov-
erning the evolution of pore relative humidity.
2 The diffusion equation is coupled with the heat balance equa-
tion which describes the evolution of temperature and takes
into account the heat released during hydration and silica fume
reaction.
307
3 The hydration process is modeled through the degree of hydra-
tion already used in [3,4,6].
4 The mass balance used in [19] to compute the asymptotic
hydration degree is extended to include the effect of silica fume.
5 The aging effect associated with the silica fume is modeled
through the degree of pozzolanic.
6 The aging effect of silicate polymerization is modeled introduc-
ing the concentration of silicate polymers, formulated for the
first time in this study.
7 The phenomenological model of Norling Mjönell [20] is adopted
to simulate aging and silica fume effects on the sorption
isotherm.
8 The chemically bound (non evaporable) water is modeled
through an expression that takes into account hydration and
silicate polymerization.
9 Moisture permeability is modeled by a novel expression, which
depends on three parameters with a clear physical
interpretation.
The model developed here needs to be implemented in a com-
putational program, calibrated, and validated by comparisons with
experimental data. This task is pursued in Part II of this study
which follows.
ð32Þ
Acknowledgement
The first author gratefully acknowledges the financial support
ð33Þ of CTG (Italcementi Group) through the CIS-E Consortium.
References
[1] DiLuzio G, Cusatis G. Hygro-thermo-chemical modeling of high performance
concrete. II: Numerical implementation, calibration, and validation. Cem Concr
Compos 2009;31(5):309–24.
[2] Wittmann FH. Crack formation and fracture energy of normal and high
strength concrete. Sādhanā 2002;27(4):413–23.
[3] Ulm F-J, Coussy O. Modeling of thermo-chemical–mechanical couplings of
concrete at early age. J Eng Mech, ASCE 1995;121(7):785–94.
[4] Cervera M, Oliver J, Prato T. Thermo-chemical–mechanical model for concrete.
I: Hydration aging. J Eng Mech, ASCE 1999;125(9):1018–27.
[5] Bažant ZP, Kim J-K, Jeon S-E. Cohesive fracturing and stresses caused by
hydration heat in massive concrete wall. J Eng Mech, ASCE 2003;129(1):21–30.
[6] Gawin D, Pesavento F, Schrefler BA. Hygro-thermo-chemo-mechanical
modelling of concrete at early ages and beyond, Part I: hydration and hygro-
thermal phenomena. Int J Numer Methods Eng 2006;67:299–331.
[7] Oh BH, Cha SW. Nonlinear analysis of temperature and moisture distributions
in early-age concrete structures based on degree of hydration. ACI Mater J
2003;100(5):361–70.
[8] Mazloom M, Ramezanianpour AA, Brooks JJ. Effect of silica fume on mechanical
properties of high-strength concrete. Cem Concr Compos 2004;26:347–57.
[9] Taylor HFW. Cement chemistry. London: Thomas Telford; 1997.
[10] Thomas JJ, Jennings HM. A colloidal interpretation of chemical aging of the
C–S–H gel and its effects on the properties of cement paste. Cem Concr Res
2006;36:30–8.
[11] Powers TC, Brownyard TL. Physical properties of hardened cement paste. ACI J
Proc 1946;18(3):250–336.
[12] Powers TC. Physical structure of Portland cement paste. In: Taylor HFW, editor.
Chemistry of cements. London: Academic Press; 1964. p. 391–416.
[13] Bažant ZP. Constitutive equation for concrete creep and shrinkage based on
thermodynamics of multi-phase systems. Mater Struct (RILEM, Paris)
1970;3:3–36.
[14] Powers TC, Brownyard TL. Physical properties of hardened cement paste. ACI J
Proc 1946;18(4):470–504.
[15] Claesson J. Few remarks on moisture flow potentional. Division of building
physics. Sweden: Lund Institute of Thecnology; 1993 [Report TVBH-7163].
[16] Bažant ZP, Najjar LJ. Nonlinear water diffusion in nonsaturated concrete. Mater
Struct 1972;5(25):3–20.
[17] Xi Y, Bažant ZP, Jennings HM. Moisture diffusion in cementitious materials –
adsorption isotherms. Adv Cem Based Mater 1994;1:248–57.
[18] Mills RH. Factors influencing cessation of hydration in water-cured cement
pastes. In: Proceeding of the symposium on the structure of Portland cement
paste and concrete. Washington (DC): Highway research board; 1966. p. 406–
24 [Special report no. 90].
[19] Pantazopoulo SJ, Mills RH. Microstructural aspects of the mechanical response
of plain concrete. ACI Mater J 1995;92(6):605–16.
308 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 301–308
[20] Norling Mjönell K. A model on self-desiccation in high-performance concrete.
In: self-desiccation and its importance in concrete technology, proceedings of
the international research seminar. Sweden: Lund; 1997. p. 141–57.
[21] Bažant ZP, Kaplan MF. Concrete at high temperatures: material properties and
mathematical models. London: Longman Addison-Wesley; 1996.
[22] Neville AM. Properties of concrete. New York: John Wiley and Sons; 1997.
[23] Bentz DP, Stutzman PE. SEM analysis and computer modelling of hydration of
Portland cement particles. In: de Hayes SM, Stark D, editors. Petrography of
cementitious materials ASTM STP 1215. Philadelphia: ASTM; 1994.
[24] Bentz DP, Waller V, De Larrard F. Prediction of adiabatic temperature rise in
conventional and high-performance concretes using a 3-D microstructural
model. Cem Concr Res 1998;28(2):285–97.
[25] de Schutter G, Taerwe L. General hydration model for Portland cement and
blast furnace slag cement. Cem Concr Res 1995;25:593–604.
[26] Waller V, de Larrard F, Roussel P. Modelling the temperature rise in massive
HPC structures. In: Fourth international symposium on utilization of high-
strength/high-performance concrete. RILEM; May 1996. p. 415–21.
[27] van Breugel K. Numerical simulation and microstructural development in
hardening cement-based materials. Heron 1992;37(3):1–61.
[28] Ulm F-J, Coussy O. Strength growth as chemo-plastic hardening in early age
concrete. J Eng Mech, ASCE 1996;122(12):1123–32.
[29] De Schutter G. Influence of hydration reaction on engineering properties of
hardening concrete. Mater Struct 2002;35:447–52.
[30] Bažant ZP, Prasannan S. Solidification theory for concrete creep. I: Formulation.
II: Verification and application. J Eng Mech, ASCE 1989;115:1691–725.
[31] Duchesne J, Bérubé MA. The effectiveness of supplementary cementing
materials in suppressing expansion due to ASR: another look at the reaction
mechanisms Part 1: concrete expansion and portlandite depletion. Cem Concr
Res 1994;24(1):73–82.
[32] Persson B. Seven-year study on the effect of silica fume in concrete. Adv Cem
Based Mater 1998;7:139–55.
[33] Bonen D, Khayat KH. Characterization and pozzolanic properties of silica fume
stored in an open pond. Cem Concr Res 1995;25(2):395–407.
[34] Mitchell DRG, Hinczak I, Day RA. Interaction of silica fume with calcium
hydroxide solutions and hydrated cement pastes. Cem Concr Res
1998;28(11):1571–84.
[35] Kirkbridge TW. Condensed silica fume in concrete: FIP state of art report.
London: FIP commission; 1988.
[36] Cheng Yi H, Feldman RF. Hydration reactions in Portland cement–silica fume
blends. Cem Concr Res 1985;15(4):585–92.
[37] Sellevold E. Condensed silica fume in concrete, FIP/NB symposium on high
strength concrete, Stavanger; 1987.
[38] Diamond S. Effects of microsilica (silica fume) on pore-solution chemistry of
cement pastes. J Am Ceram Soc 1983;66(5):82–4.
[39] Li Y, Langan BW, Ward MA. The strength and microstructure of high-strength
paste containing silica fume. Cem Concr Aggr 1996;18(2):112–7.
[40] Papadakis VG. Experimental investigation and theoretical modeling of silica
fume activity in concrete. Cem Concr Res 1999;29:79–86.
[41] Brough AR, Dobson CM, Richardson IG, Groves GW. Alkali activation of reactive
silicas in cements: in situ 29Si MAS NMR studies of the kinetics of silicate
polymerization. J Mater Sci 1996;31:3365–73.
[42] Persson B. Hydration and strenght of high performance concrete. Adv Cem
Based Mater 1996;3:107–23.
[43] Takemoto K, Uchikawa H. Hydration of pozzolanic cement. In: Proceedings of
the 7th international congress on chemistry of cement, vol. I. Paris; 1980.
p. 1–29.
[44] Wild S, Sabir BB, Khatib JM. Factors influencing strength development of
concrete containing silica fume. Cem Concr Res 1995;25(7):1567–80.
[45] Zelić J, Rušić D, Veža D, Krstulović R. The role of silica fume in the kinetics and
mechanisms during the early stage of cement hydration. Cem Concr Res
2000;30:1655–62.
[46] Justnes H, Sellevold EJ. Lundevall G. high-strength concrete binders. Part A:
reactivity and composition of cement pastes with and without condensed
silica fume. In: Fly ash, silica fume, slag, and natural pozzolans in concrete
proceedings, fourth international conference, May 1992. Istanbul. Detroit:
American Concrete Institute; 1692.
[47] Jensen OM, Hansen PF. Autogenous deformation and change of the relative
humidity in silica fume-modified cement paste. ACI Mater J 1996;93(6):1–5.
[48] Yajun J, Cahyadi JH. Simulation of silica fume blended cement hydration. J
Mater Struct 2004;37:397–404.
[49] Persson B. Eight-year exploration of shrinkage in high-performance concrete.
Cem Concr Res 2002;32:1229–37.
[50] Hirljac J, Wu ZQ, Young JF. Silicate polymerization during the hydration of
alite. Cem Concr Res 1983;13:877–86.
[51] Bentur A, Berger RL, Lawrence Jr FV, Milestone NB, Mindess S, Young JF. Creep
and drying shrinkage of calcium silicate pastes: III. A hypothesis of irreversible
strains. Cem Concr Res 1979;9:83–96.
[52] Jennings HM. A model for the microstructure of calcium silicate hydrate in
cement paste. Cem Concr Res 2000;30:101–16.
[53] Brough AR, Dobson CM, Richardson IG, Groves GW. In situ solid-state NMR
studies of Ca3 SiO5 : hydration at room temperature and at elevated
temperatures using 29Si enrichment. J Mater Sci 1994;29:3926–40.
[54] Zhang M-H, Gjørv OE. Effect of silica fume on cement hydration in low porosity
cement pastes. Cem Concr Res 1991;21:800–8.
[55] Byfors J. Plain concrete at early ages. Stockholm: Swedish Cement and
Concrete Research Institute; 1980 [Technical report no. 3:80].
[56] Jensen OM, Hansen PF. Water-entrained cement-based materials I. Principles
and theoretical background. Cem Concr Res 2001;31:647–54.
[57] Taylor HFW. Nanostructure of C–S–H: current status. Adv Cem Based Mater
1993;1:38–46.
[58] Groves GW, Rodger SA. The hydration of C3S and ordinary Portland cement
with relatively large additions of microsilica. Adv Cem Res 1989;2:
135–40.
[59] Baroghel-Bouny V, Mainguy M, Lassabatere T, Coussy O. Characterization and
identification of equilibrium and transfer moisture properties for ordinary and
high-performance cementitious materials. Cem Concr Res 1999;29:
1225–38.
[60] Xi Y, Bažant ZP, Jennings HM. Moisture diffusion in cementitious materials –
adsorption isotherms. Adv Cem Based Mater 1994;1:258–66.
[61] Brunauer S, Emmett PH, Teller E. Adsorption of gases in multimolecular layers.
J Am Chem Soc 1938;60:309–19.
[62] Brunauer S, Deming LS, Deming WE, Teller E. On a theory of the van der Waals
adsorption of gases. J Am Chem Soc 1940;62:1723–32.
[63] Halsey G. Physical adsorption on non-uniform surfaces. J Chem Phys
1948;16:931–7.
[64] Hillerborg A. A modified absorption theory. Ceme Concr Res 1985;15:809–16.
[65] Brunauer S, Skalny J, Bodor EE. Adsorption on nonporous solids. J Colloid Interf
Sci 1969;30:546–52.
[66] Xi Y, Bažant ZP, Jennings HM. Moisture diffusion in cementitious materials –
moisture capacity and diffusivity. Adv Cem Based Mater 1994;1:258–66.
[67] Karger J, Ruthven DM. Diffusion in zeolites and other microporous solids. New
York: John Wiley and Sons; 1992.
[68] Bažant ZP. Mathematical modeling of creep and shrinkage of
concrete. Chichester: John Wiley and Sons; 1988.
[69] Jooss M, Reinhardt HW. Permeability and diffusivity of concrete as function of
temperature. Cem Concr Res 2002;32:1497–504.
 Cement & Concrete Composites 31 (2009) 309–324

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Hygro-thermo-chemical modeling of high-performance concrete. II: Numerical

implementation, calibration, and validation
Giovanni Di Luzio a, Gianluca Cusatis b,*
a
Department of Structural Engineering, Politecnico di Milano, Milan, Italy
b
Department of Civil and Environmental Engineering, 4048 Johnsson Engineering Center, Rensselaer Polytechnic Institute, 110 Eighth St, Troy, NY 12180-3590, USA

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

Article history: This paper presents the numerical implementation, calibration, and validation of the hygro-thermo-
Received 12 September 2008 chemical model formulated theoretically in the companion paper [Di Luzio G, Cusatis G. Hygro-
Received in revised form 19 February 2009 thermo-chemical modeling of high-performance concrete. I: Theory. Cem Concr Compos, in press,
Accepted 25 February 2009
doi:10.1016/j.cemconcomp.2009.02.015.]. Calibration and validation are based on experimental data
Available online 13 March 2009
gathered from the literature and relevant to hydration, self-desiccation, and drying of standard and
high-performance concrete with and without silica fume.
Keywords:
Ó 2009 Elsevier Ltd. All rights reserved.
Concrete
Early age
Moisture
Temperature
Relative humidity
Hydration
Silica fume
Self-desiccation

1. Introduction by two partial differential equations (governing mass and energy

conservation at the macroscopic level) coupled with three evolu-
In the companion paper (Part I of this study) [1], a new model for tion laws for hydration degree, silica fume reaction degree, and
moisture transport and heat transfer in high-performance concrete concentration of silicate polymers (all treated as internal
(HPC) has been formulated. This model accounts for the effect of ce- variables).
ment hydration, silica fume (SF) reaction, and silicate polymeriza- In an axis-symmetric setting, the mass and energy conservation
tion on early age and long-term aging of HPC, by formulating a laws (Eqs. (4) and (11) in [1]) can be rewritten as
theoretical framework where the kinetics of the chemical reactions    
is described by Arrenius-type evolution laws. This makes it possible @ @h Dh @h @ @h @h
Dh þ þ Dh þ C1 þ C2 ¼ 0 ð1Þ
to simulate self-heating, self-desiccation and drying from the very @r @r r @r @z @z @t
   
early age in a thermodynamically consistent fashion. In the present @ @T k @T @ @T @T
Part II, the numerical implementation of the model is presented and k þ þ k þ C3 þ C4 ¼ 0 ð2Þ
@r @r r @r @z @z @t
the model is calibrated and validated by analyzing experimental
data gathered from the literature. A wide range of typical experi- where C 1 ¼ @we =@h; C 2 ¼ F h ðh; ac ; as ; ap Þ; C 3 ¼ qct , and
mental tests, including desorption tests, isothermal and adiabatic C 4 ¼ F T ðac ; as Þ.
calorimetric tests, self-desiccation tests, and drying tests, are By using the classical Galerkin [2] method based on the use of
numerically simulated in order to check accuracy and predictive the weak formulation of the differential problem and a finite
capability of the proposed formulation. All the definitions and nota- element approximation of the primary variables, hðr; z; tÞ 
PN PN
tions introduced in the companion paper are retained hereinafter. i hi ðtÞN i ðr; zÞ; Tðr; z; tÞ  i T i ðtÞN i ðr; zÞ (Ni(r, z) = shape functions,
N = number of element nodes), one obtains
2. Numerical implementation
K1 h_ þ K2 h þ F1 ¼ 0 ð3Þ
As discussed in the companion paper [1], moisture transport
and heat transfer in high-performance concrete can be modeled K3 T_ þ K4 T þ F2 ¼ 0 ð4Þ

where the matrices K1 ; K2 ; K3 ; K4 , and the vectors F1 , F2 are obtained

* Corresponding author. Tel.: +1 518 276 3956; fax: +1 518 276 4833.
E-mail addresses: cusatg@rpi.edu, giacusatis@yahoo.it (G. Cusatis). by the global assemblage of element level matrices given by

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.016
310 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

Z
Table 1
K ele
1;ij ¼ C 1 Ni Nj dX Material properties with known or independently measurable value.
ele
Z   
@Ni @Nj Nj @Ni @Ni @N j Material property name (unit) Symbol Typical value Reference
K ele
2;ij ¼ D h  þ  dX ð5Þ
ele @r @r r @r @z @z Density (kg/m )3
q 2400 [20]
Z Isobaric heat capacity (J/kg °C) ct 1100 [20]
K ele
3;ij ¼ C 3 Ni Nj dX Heat conductivity (W/m °C) k 2.3 [21]
ele Cement hydration enthalpy (kJ/kg) ~1
Q 500 [22]
Z    ~1
c
@Ni @Nj N j @N i @Ni @Nj Silica fume reaction enthalpy (kJ/kg) Q s 780 [23]
K ele
4;ij ¼ k  þ  dX ð6Þ Hydration activation energy/R (K) Eac =R 5000 [6]
ele @r @r r @r @z @z Silica fume reaction activation energy/R (K) Eas =R 9700 [14]
Z
Diffusivity activation energy/R (K) Ead =R 2700 [3]
F ele
1;i ¼ C 2 N i dX Silica fume efficiency (–) SF eff 0.9 [14,9]
ele
Z Polymerization activation energy/R (K) Eap =R 6000 [24]
F ele C 4 N i dX Hydration threshold for polymerization (–) a0c 0.1 [1]
2;i ¼ ð7Þ
ele n/a (–) kc 0.253 [4]
n/a (–) ks 0.0 [1]
Appendix A reports the calculation of the element matrices n/a (–) kpc 0.0 [1]
above for the case of a three-node triangular element (linear shape
functions).
Time discretization of Eqs. (3) and (4) can be obtained by the formance concrete can be subdivided into two categories. Material
well-known Crank–Nicolson method (a central difference method) properties that have been already established in the past and that
which has been proven to be unconditionally stable in the case of can be measured independently from the current theory (do not
constant matrix coefficients [2]. require a back analysis), belong to the first category. These proper-
Finally, the numerical formulation of the problems requires, at ties are summarized in Table 1 along with typical values that can
each gauss point of the finite element mesh, the numerical integra- be used when specific experimental measurements are not avail-
tion in time of the internal variable evolution laws (governing able. These standard property values will be used throughout the
hydration degree, silica fume reaction degree, concentration of sil- rest of the paper unless otherwise specified. On the contrary, mate-
icate polymers). As presented in the companion paper [1], these rial properties whose calibration requires back analysis of the cur-
equations can be all expressed in the form x_ ¼ f ðx; h; TÞ, where x rent theory, belong to the second category. These properties,
represents an internal variable. A very effective method to inte- referred hereinafter as free parameters of the model, are: hydration
grate such a differential equation is to use the second order Run- degree parameters, gc ; Ac1 , and Ac2 , SF reaction degree parameters,
ge–Kutta formula (midpoint method). By this approach xnþ1 , i.e. gs ; As1 , and As2 , sorption isotherm parameters, kcv g ; ksv g , and g 1 , mois-
the value of the internal variable at time tnþ1 , can be estimated as ture diffusivity parameters, D ~ 1 , and n, polymer concentration
~ 0; D
xnþ1  xn þ Dtn f ðxnþ1=2 ; hnþ1=2 ; T nþ1=2 Þ ð8Þ parameter, Ap1 , and chemically bound water parameter related to
polymerization, kps .
where xn is the value of the internal variable at time t n ; Dt n is the The free parameters of the model can be calibrated on the basis
time increment, xnþ1=2 ¼ ðxnþ1 þ xn Þ=2; hnþ1=2 ¼ ðhnþ1 þ hn Þ=2, and of the following calibration procedure:
T nþ1=2 ¼ ðT nþ1 þ T n Þ=2. At time t 0 ¼ 0 (time of casting) all the inter-
nal variables are set to zero (x0 ¼ 0). 1. Considering a concrete mix without silica fume:
c
Remark 1. The discretized balance equations and the numerical (a) Calibrate kv g and g 1 by fitting experimental sorption/
evolution laws are both nonlinear and need to be solved by means desorption isotherms relevant to two different values of
of an iterative scheme. The numerical simulations performed in hydration degree, which should remain constant during
this study showed that an iterative scheme based on a global tests in order to decouple the calibration of the sorption iso-
measure of convergence (e.g., norm of temperature and/or humid- therm from the calibration of the parameters related to
ity vectors), provides satisfactory accuracy in the iterative solution cement hydration (step 1b). This can be only achieved in
of the internal variable evolution laws. an approximate way because it is not possible to stop the
hydration process while conducting the sorption/desorp-
Remark 2. The system in Eqs. (3) and (4) do not include the tion tests. It is feasible, however, to obtain experimental
boundary conditions. However, Dirichlet’s boundary conditions sorption isotherms characterized by a small variation of
c
can be directly imposed on the system of Eqs. (3) and (4) [3], the hydration degree (6 0:1) [5]. The calibration of kv g
whereas Cauchy’s boundary conditions can be enforced approxi- and g 1 should be performed with reference to the mean
mately by considering an additional layer of material attached to value of the hydration degree during the test.
the actual boundaries and imposing Dirichlet’s boundary condi- (b) Calibrate gc ; Ac1 , and Ac2 by fitting either temperature rise in
tions on these fictitious boundaries. By changing the thickness of an adiabatic calorimetric test [6] or the experimental
the additional layer, different heat and moisture surface emissivity hydration degree calculated as the ratio between the heat
coefficients can be simulated [3]. released in an calorimetric test and the total heat released
at complete cement hydration [7].
(c) Calibrate D ~ 1 , and n by fitting experimental results of
~ 0; D
Remark 3. The discussed axis-symmetric numerical formulation
can be also applied to the solution of two-dimensional problems. drying tests [3,8].
The only modification of the discretized equations consist in 2. Considering the concrete mix in step 1 with the addition of sil-
neglecting the term ðNj @N i =@rÞ=r appearing in the expressions of ica-fume and keeping unaltered the already calibrated
K ele ele parameters:
2;ij and K 4;ij . s
(a) Calibrate kv g by fitting sorption/desorption experimental
data [5]. Same considerations made in step 1a apply here.
3. Calibration and validation procedure (b) Calibrate gs ; As1 , and As2 by fitting the difference in temper-
ature rise (or heat development) for adiabatic calorimetric
The material properties needed to simulate, according to the tests (or isothermal calorimetric tests) with and without
present theory, moisture transport and heat transfer in high-per- (data used in step 1b) SF. Alternatively, gs ; As1 , and As2 can
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 311

be calibrated by fitting direct measurement of the degree of
pozzolanic reaction, which can be obtained by measuring
the variation of CH content due to the presence of SF [9].
(c) Calibrate Ap1 by fitting experimental polymer concentration
data that can be obtained by measuring the evolution of
polymerized species in hydrated concrete mix [9].
(d) Calibrate kps by fitting experimental data on chemically
bound water in presence of polymerization (that is for suf-
ficiently hydrated concrete) [10]. Considering that: (1) the
chemically bound water is proportional to the polymers
concentration through the parameter kps , (2) the polymers
concentration is proportional to the parameter Ap1 , and
(3) polymerization only affects the formula of the chemi-
cally bound water, experimental data on chemically bound
water can be actually used to calibrate directly the poly-
merization parameter C p ¼ kps Ap1 . This allows eliminating
step 2c for which experimental data is extremely rare and
difficult to obtain.
It must be noted here that, in the current literature, a compre-
hensive study providing, for the same concrete, all experimental
data needed for the calibration procedure outlined above, is not
available. In most of the cases, experimental investigations tend
ibration phase. In the validation phase, parameter adjustment is
not permitted. In the following sections examples of calibration
and validation will be presented and discussed.
4. Sorption/desorption isotherms
The calibration of the sorption isotherm parameters is discussed
in this section. As discussed in the companion paper [1], the differ-
ence between sorption and desorption isotherms are neglected in
this study as done in previous studies by other researchers [11].
This also permits to calibrate the sorption isotherm on the basis
of desorption data, which are the most commonly available in
the literature.
Desorption experimental data for two old concretes (2 years of
age), labeled as BO and BH, were published in [12]. Both mixes
were prepared with a type I Portland cement. The first mix (BO)
was a standard concrete, characterized by w/c = 0.48 and
c = 300 Kg/m3. On the contrary, the second mix (BH) was an
high-performance concrete, characterized by w/c = 0.26,
c = 450 Kg/m3, and containing silica fume ðs=c ¼ 0:10Þ. With these
material properties, one can obtain a1 c ¼ 0:73 for mix BO, and
a1c ¼ 0:51; as ¼ 0:87 for mix BH [1]. These values can be assumed
1

to analyze specific aspects in relation to different mixes more that
focusing on a thorough analysis of one single mix. In this cases, the
only option left is an educated guess of the parameters that cannot
be calibrated.
Once the calibration has been performed, the validation has to
be carried out by simulating experimental data, relevant to the
same concrete of the calibration, that has not been used in the cal-
to be the hydration degree and the degree of SF reaction at the time
of the tests. With this assumption, the best fitting of the experi-
mental data without SF (Fig. 1a) can be obtained by setting
c
kv g ¼ 0:2 and g1 = 1.5. The best fitting of the experimental data
with silica fume (Fig. 1b) can be obtained by adopting the previ-
c
ously identified parameters ðkv g ¼ 0:2, g1 = 1.5) and by setting
s
kv g ¼ 0:36.

Experimental t0=2 years Numerical c = oo

a 0.4 b 0.4
Concrete BO w/c=0.48 Concrete BH w/c=0.26
0.3
s/c=0.1
0.3

0.2 0.2

0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Relative humidity [-] Relative humidity [-]

c 100 Concrete w/c=0.4

d Concrete w/c=0.25
100
Experimental c=0.2
Experimental c=0.1-0.2
Experimental c=0.4
80 80 Experimental c=0.2-0.3
Experimental c=0.65
Experimental c=0.3-0.5
Numerical c=0.1
Numerical c=0.15
60 Numerical c=0.4 60
Numerical c=0.7 Numerical c=0.3
Numerical c=0.5
40 40

20 20

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Relative humidity [-] Relative humidity [-]
Fig. 1. Fitting of experimental desorption isotherms for concrete mixes with various water-to-cement ratios and various degrees of hydration.
312 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324
In Ref. [5] experimental desorption data, relevant to two differ-
ent concrete mixes characterized by two water-to-cement ratios,
were reported. The first mix (w=c ¼ 0:4) was tested at three differ-
ent ages characterized by ac varying from 0.1, 0.3, and 0.5 to 0.3,
0.5, and 0.6, respectively. The numerical simulations were per-
formed by assuming an average hydration degree, ac = 0.15, 0.4,
and 0.65, respectively, for the three cases. Fig. 1c shows the best
fitting of the experimental data obtained by setting, again,
c
kv g ¼ 0:2 and g1 = 1.5. As shown in Fig. 1d, the same parameters
can be also used to fit the desorption data of the second mix
(w=c ¼ 0:25) tested at four different ages. For this case, the exper-
imental hydration degree ranges and the mean values adopted in
the numerical calculations are reported in the figure legend.
For all the experimental data sets analyzed in this section, the
agreement between the theory and the experiments can be consid-
ered good even though, for some of the tests, the scatter of the
experimental data is relatively high.
5. Isothermal calorimetric tests
For ordinary portland cement, the evolution of the hydration
degree can obtained from the evolution of chemically bound water,
whose measurement can be, in turn, utilized to calibrate the hydra-
tion degree parameters. The experimental data obtained by Pane
and Hansen [13], who measured the evolution in time of the chem-
ically bond water in sealed samples of cement paste cured isother-
mally at three different temperatures (9, 23, and 34 °C), is analyzed
in this section.
The tested cement paste was characterized by c = 438 Kg/m3,
w=c ¼ 0:4, and a1 c ¼ 0:72. In addition, a latent heat of hydration
equal to Q ~ 1 ¼ 439:0 kJ=kg was reported by the authors of the
c c is very good agree-
experimental investigation [13].
Since, in these tests, the temperature and humidity fields do not
feature significant spatial gradients, the numerical simulations
were carried out on a single element mesh subjected to boundary
conditions corresponding to isothermal heat exchange and without
water mass exchange.
The best fitting of the experimental data points relevant to
T = 9 °C was obtained by the parameters reported on the second
column of Table 2. It must be noted that, since desorption data
was not available, the sorption isotherm parameters discussed in
the previous section were adopted. Finally, the simulations rele-
1 13, which includes SF, was used to calibrate the SF reaction degree
0.8
0.6
wn /wn inf
0.4
Chemical bond water for w/c =0.45
Experimental T=9 °C
Numerical T=9 °C
0.2 Experimental T=23 °C
Numerical T=23 °C
Experimental T=34 C
° 7. Self-desiccation tests
Numerical T=34 °C
0 The third numerical example deals with self-desiccation, that is
1 10 100 1000 10000
Time [hours]
Fig. 2. Numerical simulations of the experimental data of Pane and Hansen [13]:
evolution of the chemically bound water under different isothermal conditions.
vant to the other temperatures were performed without changing
the parameters for validation purposes.
Fig. 2 shows the comparison between the experiments and the
numerical simulations with reference to the chemically bound
water normalized by its asymptotic value. The agreement is excel-
lent and the theory correctly predict faster hydration for higher
temperatures. As discussed in [1], this effect is captured automat-
ically through the Arrenius-type dependence of the hydration pro-
cess on temperature. Fig. 2 also shows that the final value of the
chemically bound water (and, similarly, the degree of hydration)
does not depend on temperature.
6. Adiabatic calorimetric tests
The experimental investigation of Bentz et al. [14] on conven-
tional and high-performance concretes aimed at the characteriza-
tion of their adiabatic heat signature is considered herein. Bentz
et al. [14] measured the temperature increase under adiabatic con-
ditions for a large variety of concrete mixes with and without SF. In
Table 3, the first four columns report density, cement content,
water-to-cement ratio, and silica-to-cement ratio, respectively, for
16 of these mixes. The fifth and the seventh columns of the same Ta-
ble report the asymptotic hydration degree and the asymptotic sil-
ica fume reaction degree, calculated on the basis of Eqs. 19, 20, 24
and 26 in [1]. These values are in very good agreement with the val-
ues of the same quantities calculated, with reference to a concrete
age of 3 years, by Bentz et al. [14] through a three-dimensional
microstructural model (sixth and eighth columns of Table 3).
The values Q 1c ¼ 520 kJ=kg, Eac =R ¼ 5490 K; Eas =R ¼ 9620 K can
be assumed for all 16 mixes based on data reported in [14]. It is
worth noting that the assumed value for Q 1
ment with (only 7% higher than) the value calculated from the en-
thalpy of complete hydration of the major phases of cement [14].
Similarly to the previous section, the numerical simulations of
these experiments were performed by a single element mesh.
Boundary conditions not allowing mass and heat exchange
through the external surfaces were imposed to simulate an adia-
batic system.
In the absence of specific measurements, the sorption isotherm
parameters were assumed to be equal to the ones optimized in
Section 4. The hydration degree parameters were calibrated by
the best fitting of the adiabatic temperature rise for Mix 1, which
does not include SF. Then, the adiabatic temperature rise for Mix
parameters. The optimized parameters are reported in the third
column of Table 2. The numerical simulations of the other 14 mixes
were performed for validation.
Figs. 3–6 show comparison between the experimental data and
the numerical simulations. Symbols are the experimental data
points, lines represent the results obtained by the present model
with (solid line) and without (dashed line) the effect of the SF reac-
tion degree. As one can see, without modeling the SF reaction, tem-
perature rise is underestimated by up to 10%. On the contrary, with
the use of the full theory, the agreement is excellent for all 16
mixes. These results clearly show the proposed model can be reli-
ably employed to predict the temperature rise and the heat pro-
duction of a wide variety of concrete mixes with and without SF.
the reduction of relative humidity in early age sealed samples due
to water consumption associated with cement hydration.
The material properties and boundary conditions (sealed condi-
tions for mass exchange at constant temperature) correspond to
those in the experimental study of Lura et al. [15], who measured
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 313

Table 2
Values of the free parameters used in the numerical simulations.

Par. (unit) Section 5 Section 6 Section 7 Section 8 Section 9 Section 10

g 1 (–) 1.5 1.5 1.75 1.2 1.7 1.55
c
kv g (–) 0.2 0.2 0.2 0.255 0.11 0.18
s
kv g (–) – 0.36 0.36 – – –
1
Ac1 (h ) 1:41  107 20  107 10  107 1:5  107 4  107 3:5  107
Ac2 (–) 5  103 1  106 1  106 5  102 5  102 5  103
gc (–) 8 6.5 5.5 8 8.5 8
1
As1 (h ) – 5  107 6  107 – – 6  107
As2 (–) – 1  106 1  106 – – 1106
gs (–) – 9.5 9.5 – – 9
1
C p (h ) – – 0:875  104 – – 0.875104
~ 0 (m2 /h)
D – – – 0.085 0.0024 0.009
~ 1 (m2 /h)
D – – – 6 3.22 8
n (–) – – – 3 3.25 4.5

Table 3
Density, cement content, water-to-cement ratio, silica-to-cement ratio, ultimate degree of hydration, and ultimate degree of silica fume reaction for concrete mixes studied by
Bentz et al. [14].

q (kg/m3) c (kg/m3) w=c s=c a1

c ac (3 years) Ref. [14] a1
s as (3 years) Ref. [14]
1 2443 529.5 0.30 0.0 0.63 0.66 – –
2 2439 455.3 0.35 0.0 0.66 0.74 – –
3 2367 438.1 0.45 0.0 0.72 0.87 – –
4 2366 274.7 0.65 0.0 0.80 0.87 – –
5 2447 474.8 0.30 0.05 0.60 0.62 0.87 0.66
6 2430 483.9 0.30 0.10 0.57 0.57 0.87 0.59
7 2318 453.2 0.30 0.30 0.56 0.48 0.40 0.37
8 2488 453.0 0.35 0.20 0.59 0.59 0.70 0.53
9 2354 455.3 0.45 0.05 0.70 0.80 0.87 0.88
10 2421 344.1 0.45 0.10 0.68 0.77 0.87 0.77
11 2413 323.3 0.45 0.20 0.65 0.72 0.78 0.63
12 2406 333.8 0.45 0.30 0.65 0.68 0.53 0.50
13 2381 307.2 0.55 0.20 0.70 0.78 0.79 0.74
14 2358 281.9 0.65 0.05 0.78 0.88 0.87 0.94
15 2340 300.2 0.65 0.10 0.76 0.85 0.87 0.92
16 2347 295.1 0.65 0.30 0.74 0.81 0.53 0.64

a w/c = 0.30 b w/c = 0.35

100 100
Temperature [C]

Temperature [C]

80 80

60 60

40 40

20 20
0 40 80 120 160 200 0 50 100 150 200
Time [hours] Time [hours]

c w/c = 0.45
d w/c = 0.65
100 100
Temperature [C]

Temperature [C]

80 80

60 60

40 40

20 20
0 40 80 120 160 200 0 40 80 120 160 200
Time [hours] Time [hours]

Fig. 3. Experimental investigation of Bentz et al. [14]: experimental (data points) and numerical (solid lines) adiabatic temperature rise curves for concrete with different
values of water-to-cement ratio and without silica fume.
314 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

a w/c = 0.55 and SF=20% b w/c = 0.65 and SF = 5%

100 100

Temperature [C]
Temperature [C]
80 80

60 60

40 40

20 20
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time [hours] Time [hours]

c d
w/c = 0.65 and SF = 10% w/c = 0.65 and SF = 30%
100 100

Temperature [C]
Temperature [C]

80 80

60 60

40 40

20 20
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time [hours] Time [hours]

Fig. 4. Experimental investigation of Bentz et al. [14]: experimental (data points) and numerical (solid and dashed lines) adiabatic temperature rise curves for concrete with
high values of water-to-cement ratio and silica fume. Dashed lines represent model results obtained neglecting the SF reaction.

a w/c = 0.45 and SF=5% b w/c = 0.45 and SF = 10%

100 100
Temperature [C]
Temperature [C]

80 80

60 60

40 40

20 20
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time [hours] Time [hours]

c w/c = 0.45 and SF = 20% d w/c = 0.45 and SF = 30%

100 100
Temperature [C]
Temperature [C]

80 80

60 60

40 40

20 20

0 30 60 90 120 150 180 0 30 60 90 120 150 180

Time [hours] Time [hours]

Fig. 5. Experimental investigation of Bentz et al. [14]: experimental (data points) and numerical (solid and dashed lines) adiabatic temperature rise curves for concrete with
medium values of water-to-cement ratio and silica fume. Dashed lines represent model results obtained neglecting the SF reaction.
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 315

a w/c = 0.30 and SF=5% b w/c = 0.30 and SF = 10%

100 100

Temperature [C]

Temperature [C]
80 80

60 60

40 40

20 20
0 40 80 120 160 200 0 50 100 150 200
Time [hours] Time [hours]

c w/c = 0.30 and SF = 30% d w/c = 0.35 and SF = 20%

100 100

Temperature [C]
Temperature [C]

80 80

60 60

40 40

20 20
0 40 80 120 160 200 0 40 80 120 160 200
Time [hours] Time [hours]

Fig. 6. Experimental investigation of Bentz et al. [14]: experimental (data points) and numerical (solid and dashed lines) adiabatic temperature rise curves for concrete with
low values of water-to-cement ratio and silica fume. Dashed lines represent model results obtained neglecting the SF reaction.

a 0.8 b 1
Relative humidity [-]
Degree of reaction

0.6
0.95
0.4

0.9
0.2

0 0.85
0 2 4 6 0 2 4 6
Time [days] Time [days]

Fig. 7. Numerical simulations of the experimental data of Lura et al. [15] on high-performance cement paste: evolution in time of (a) degree of reaction and (b) relative
humidity.

the development of the overall degree of reaction and relative a

humidity during the initial 6 days after casting in a cement paste 3 cm 12 cm
10 cm
characterized by w=c ¼ 0:37, and s=c ¼ 0:052. According to these 7 cm
material properties, one has a1 c ¼ 0:653 and as ¼ 0:9.
1

The overall degree of reaction, referred hereinafter as ar , is de-

fined as the ratio between the heat measured in the isothermal 20 cm 10 cm
calorimetric test, Q r , and the heat developed at the completion of
all chemical reactions, Q 1 r . One can write b
~ 1 þ as ðtÞsQ
Q ðtÞ a ðtÞcQ ~1
ar ðtÞ ¼ r 1 ¼ c 1 ~ 1c
~
s
ð9Þ
Qr a cQ þ a sQ
c
1
c
1
s s

Again, since temperature and humidity fields with negligible

gradient characterize these tests, a single element mesh was em-
ployed in the numerical simulations. The best fitting of reaction de- Fig. 8. Numerical simulations of the experimental data of Kim and Lee [17]: (a)
gree and relativity humidity evolution in time was obtained by the geometry and (b) finite element mesh.
316 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324
parameters reported in the fourth column of Table 2. Fig. 7a and b
show the comparison between the experimental data and numer-
ical results and the agreement is very good. In this case, no data
was available for validation.
8. Drying tests
In the previous sections, calibration and validation have focused
on experimental data featuring temperature and humidity fields
without (or negligible) spatial gradients. Consequently, the numer-
ical results were independent of the moisture permeability, whose
effect is analyzed herein with reference to the experimental inves-
tigations of Kim and Lee [17] on moisture drying at early ages.
Drying occurs when concrete is exposed to ambient air with a
relative humidity smaller than the relative humidity of the speci-
men. The imbalance of the relative humidity at the boundaries
causes an outward flow of water, which, in turn, leads to moisture
diffusion from the core towards the external surfaces of the speci-
mens and to nonuniform relative humidity distributions. In addi-
tion, at the early ages, the entire process is influenced by the
self-desiccation phenomenon.
Kim and Lee [17] measured the nonuniform moisture distribu-
tions in concrete samples, relevant to three different concrete
mixes, exposed to ambient air with 50% relative humidity. The
three mixes were characterized by cement content, water-to-ce-
ment ratio, and density equal to 541, 423, 310 Kg/m3, 0.28, 0.4,
0.68, and 2394, 2344, 2257 Kg/m3, respectively. Additional infor-
mation on the chemical characteristics of the mix compounds
and concrete compositions can be found in [17]. The tested speci-
mens were prisms ð10  10  20 cmÞ, which were moist-cured be-
fore drying. Two different ages were considered: t 0 ¼ 3 and 28
days. The internal relative humidity was measured at the distance
of 3, 7, and 12 cm from the exposed surface (Fig. 8a), which was
one of the 10  10 cm square bases. The other five sides were
sealed to generate a uniaxial moisture diffusion. In addition, com-
pletely sealed cubic specimens were also used to measure the var-
iation of relative humidity due to self-desiccation only.
The numerical simulations of these tests were performed by
using the plane finite element mesh showed in Fig. 8b. On the ex-
posed surface, Cauchy’s boundary conditions were imposed by an
equivalent surface thickness of 0.75 mm (see Section 2).
For the three mixes, the asymptotic degree of hydration can be
calculated as 0.61, 0.694, and 0.8, respectively. Since sorption iso-
a 1
Relative humidity [-]
0.95
0.9
0.85
0.8 t0=3days
0.75
0 40 80
Time [days]
Fig. 9. Numerical simulations of the experimental data of Kim and Lee [17]: relative humidity vs. time due to self-desiccation for (a) t0 = 3 day and (b) t0 = 28 day.
therm data and hydration degree data were not available for these
concretes, the sorption isotherm parameters and hydration degree
parameters were calibrated all together by simulating the self-
desiccation experiments relevant to w=c ¼ 0:4 and age of 3 and
28 days (solid curves in Fig. 9a and b). The permeability parameters
were calibrated by simulating drying experimental data relevant to
w=c ¼ 0:4 and age of 3 days (Fig. 10b). The optimized parameters
are reported in the fifth column of Table 2. The numerical
simulations of the other available data served as validation of the
model.
Figs. 9–11 show the comparison between the numerical results
as the experimental data for all the simulations. The agreement is
overall excellent, proving that the present model can be used reli-
ably to estimate moisture diffusion in normal and high-perfor-
mance concretes.
Fig. 9a and b show, for the various concretes, the evolution in
time of relative humidity for the sealed specimens subjected to
self-desiccation after moist-curing for 3 and 28 days, respectively.
The model correctly predicts higher relative humidity decrease for
low water-to-cement ratios at early ages. The initial relative
humidity after moist-curing is slightly less than one for low
water-to-cement ratios suggesting that the water provided
through the moist-curing was not enough to maintain complete
saturation of these concretes.
Fig. 10a, b, and c show, for the three considered water-to-ce-
ment ratios, the evolution in time of relative humidity at the three
different locations mentioned previously, after 3 days of moist-cur-
ing. When drying begins, similarly to the previous case, the initial
internal relative humidity in specimens with low water-to-cement
ratio is less than the one in specimens with high water-to-cement
ratio. In addition, data characterized by high water-to-cement ratio
shows full saturation at the beginning of drying.
Furthermore, the model simulates correctly the spatial gradi-
ent of relative humidity, which decreases with the distance from
the exposed surface. By comparing self-desiccation and drying
data, both the numerical simulations and the experimental data
suggest that, in the inner regions of the specimens, the variation
of relative humidity is mainly due to self-desiccation in the case
of low water-to-cement ratios (slower moisture diffusion and
higher self-desiccation) and to drying in the case of high water-
to-cement ratios (faster moisture diffusion and lower self-desic-
cation). This is in agreement to the fact that, typically, concretes
with low water-to-cement ratio feature a denser microstructure
b 1
Relative humidity [-]
0.95
0.9
0.85
0.8 t0=28days
0.75
120 0 50 100 150 200 250
Time [days]
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 317

a 1 b 1

Relative humidity [-]

Relative humidity [-]

0.9 0.9

0.8 0.8

0.7 0.7
w/c=0.28 w/c=0.40
0.6 0.6
0 40 80 120 0 40 80 120
Time [days] Time [days]

c 1
Relative humidity [-]

0.9 t0=3 days

Exp. 3 cm
Num. 3 cm
0.8 Exp. 7 cm
Num. 7 cm
Exp. 12 cm
0.7 Num. 12 cm
w/c=0.68

0.6
0 40 80 120
Time [days]

Fig. 10. Numerical simulations of the experimental data of Kim and Lee [17]: relative humidity vs. time with t0 = 3 day due to moisture diffusion and self-desiccation for
concrete with different water-to-cement ratios and at different positions in the specimen.

a 1 b 1
Relative humidity [-]
Relative humidity [-]

0.9 0.9

0.8 0.8

0.7 0.7
w/c=0.28 w/c=0.40

0.6 0.6
0 50 100 150 200 250 0 50 100 150 200 250
Time [days] Time [days]

c 1
Relative humidity [-]

t0=28 days
0.9
Exp. 3 cm
Num. 3 cm
Exp. 7 cm
0.8 Num. 7 cm
Exp. 12 cm
Num. 12 cm
0.7
w/c=0.68

0.6
0 50 100 150 200 250
Time [days]

Fig. 11. Numerical simulations of the experimental data of Kim and Lee [17]: relative humidity vs. time with t0 = 28 day due to moisture diffusion and self-desiccation for
concrete with different water-to-cement ratios at different positions in the specimen.
318 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

Fig. 12. Numerical simulations of the experimental data of Ayano and Wittmann [16]: (a) geometry and (b) finite element mesh.

and are characterized by higher consumption of water during
hydration.
Similar observations can be made with reference to Fig. 11a, b,
and c relevant to drying data at the age of 28 days.
9. Size-effect on moisture loss
The proposed model is also able to predict the size-effect on
moisture loss, that is the reduction of the overall moisture content
as function of specimen size. The moisture loss (by weight) can be
calculated by subtracting from the initial water content, w0 , the
sum of non-evaporable water, wn and evaporable water, we . By
integrating this difference over the volume, V, of the specimen
one can obtain the average moisture loss, W loss , as
Z is excellent and the model is able to predict with great accuracy
W loss ¼ w0 V  ðwn þ we ÞdV
V
with relative humidity of 45%. The top and the bottom surfaces of
the cylinders were sealed in order to generate a radial water diffu-
sion. The tested concrete was characterized by c = 400 kg/m3,
q = 2334 Kg/m3, Q~ 1c ¼ 400 KJ=kg, and w=c ¼ 0:5. For this value of
water-to-cement ratio one has a1 c ¼ 0:74.
The three specimens were modeled through axis-symmetric fi-
nite element meshes, which are shown in Fig. 12b. The parameters
of the model were calibrated by simulating the specimen with
/ = 100 cm and their optimized values are reported in the sixth
column of Table 2. The simulation of the other specimens served
as validation. Fig. 13 shows the comparison between the experi-
mental and the numerical results in terms of average water loss
normalized with respect to the initial total water content
W 0 ¼ w0 V. The agreement between experiments and simulations
ð10Þ the size-effect in moisture loss.

In the experimental investigation of Ayano and Wittmann [16], the
average moisture loss was measured for three cylindrical concrete
specimens characterized by different diameters, / = 50, 80, and
150 mm, and lengths, L = 100, 100, and 150 mm, respectively
(Fig. 12a). The moisture loss was due to exposure to ambient air
0.6
=50mm
Moisture loss
10. Effect of curing conditions on moisture content
In this section the numerical simulations of the experimental
results obtained by Persson [10,18,19] are presented. In this exper-
imental investigation, chemically bound (non-evaporable) water
and internal relative humidity were measured in concrete speci-
mens subjected to air, water, and sealed curing, over a period of
450 days. Eight different concrete mixes, characterized by various
values of water-to-cement ratio and cement content, were studied.
Four mixes also included silica fume. The compositions of these
mixes are reported in Table 4 where one can also see the calculated
ultimate degree of hydration and ultimate degree of silica reaction.

0.4

=80mm Table 4
Cement content, water-to-cement ratio, silica-to-cement ratio, ultimate degree of
0.2 hydration, and ultimate degree of silica fume reaction for concrete mixes studied by
Persson [10,18,19].
150mm
q (kg/m3) c (kg/m3) w=c s=c a1
c a1
s

1 2513 456 0.25 0.0 0.582 –

0 2 2469 400 0.33 0.0 0.646 –
1 10 100 3 2441 303 0.46 0.0 0.727 –
4 2424 299 0.58 0.0 0.772 –
Time [days] 5 2533 484 0.22 0.1 0.495 0.792
6 2500 476 0.24 0.1 0.515 0.864
Fig. 13. Numerical simulations of the experimental data of Ayano and Wittmann 7 2456 389 0.36 0.1 0.67 0.9
[16] on moisture loss: symbols are experimental data points and solid lines are 8 2451 298 0.48 0.1 0.734 0.9
numerical results.
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 319

The chemical characteristic of cement, aggregate, silica fume, and column of Table 2, were obtained by the best fitting of self-desic-
additives can be found in [10]. The tested specimens were circular cation data (Figs. 16b and 21b), chemically bound water data (Figs.
slabs of 1 m diameter and 0.1 m thickness. For air and water cur- 16e and 21d), and drying data (Figs. 16d and 21e). The same model
ing, the flat sides of the slabs were covered by thick layers of epoxy parameters were used to simulate the behavior of the other six
resin to impose only radial moisture transport. For sealed curing, mixes for validation purpose. Comparison between the experimen-
also the external circular surface was treated with epoxy resin in tal data and the numerical predictions are shown in Figs. 15–22.
order to obtain perfectly sealed conditions. Temperature and Each figure is relevant to a different concrete Mix and overall the
humidity of the ambient air were recorded during the air curing agreement is good for all eight mixes.
tests and their evolution in time (Fig. 14a and b) provides the Sub-figures a, c, and e in Figs. 19–22 show the chemically
essential boundary conditions for the numerical simulation of bound water ðwn Þ normalized with respect to cement content
these tests. For both air and water curing, relative humidity mea- ðcÞ, for sealed conditions, air curing, and water curing, respec-
surements were obtained through cast-in plastic probes placed at tively. Both the experimental data and the numerical results show
50, 150 and 350 mm from the exposed surface [10]. that the chemically bound water varies slightly under different
The model parameters were calibrated by fitting the experi- curing conditions while it increases for increasing water-to-ce-
mental results relevant to Mix 2 (without silica fume) and Mix 7 ment ratio. At the end of the tests (450 days), wn =c shows more
(with silica fume), both featuring a water-to-cement ratio close than 30% increase (from about 0.15 to about 0.2) as w=c increases
to the average water-to-cement ratio (0.365) of the eight mixes. from 0.25 to 0.58 (Figs. 15a and 18a). The presence of SF reduces
In particular, the optimized parameters, reported in the seventh the amount of chemically bound water but, at the same time,

25 1

Relative humidity [-]

Temperature [C]

20 0.8

15 0.6

10 0.4

5 0.2

0
0 200 400 600 0 200 400 600
Time [days] Time [days]

Fig. 14. Data by Persson [10,18,19]: measured air temperature and relative humidity during air curing.

1 c 0.25
a 0.25 Non-evaporable water b
Sealed condition
Relative Humidity [-]

w/c = 0.25 0.2

0.2 Exp.
Self-Desiccation
wn/c [kg/kg]

Num.
0.9 w/c = 0.25
wn/c [kg/kg]

0.15
Exp. 0.15
Num.

0.1 0.1
0.8 Non-evaporable water
Air curing - w/c = 0.25

0.05 0.05 Exp 50mm

Num 50mm

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 e 0.25 f 1
Air Curing w/c = 0.25
Exp. 50mm
Relative Humidity [-]

Relative Humidity [-]

Num. 50mm 0.2

0.9 Exp. 150mm 0.9
wn/c [kg/kg]

Num. 150mm
Exp. 350mm 0.15
Num. 350mm
0.8 0.8 Water Curing
w/c = 0.25
0.1 Non-evaporable water Exp 50mm
Water curing - w/c = 0.25 Num 50mm
0.7 Exp 50mm 0.7 Exp. 150mm
0.05 Num 50mm Num. 150mm
Exp. 350mm
Num. 350mm
0.6 0 0.6
0 200 400 600 0 200 400 600 0 200 400 600
Time [days] Time [days] Time [days]

Fig. 15. Data by Persson [10,18,19], concrete Mix 1: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).
320 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

a 0.25 b 1 c 0.25

Relative Humidity [-]

0.2 0.2
wn/c [kg/kg]

wn/c [kg/kg]
0.9
0.15 0.15

0.1 Non-evaporable water 0.1 Non-evaporable water

Sealed curing 0.8 Self-Desiccation Air curing - w/c = 0.33
w/c = 0.33 w/c = 0.33 Exp 50mm
0.05 Exp. Exp. 0.05
Num 50mm
Num. Num.
0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 e 0.25 f 1

Relative Humidity [-]

Relative Humidity [-]

0.2
0.9 0.9
wn/c [kg/kg]

0.15
Water Curing
0.8 Air Curing - w/c = 0.33 0.8 w/c = 0.33
Num. 50mm 0.1 Exp. 50mm
Non-evaporable water Num. 50mm
Exp. 50mm
Water curing - w/c = 0.33 Exp. 150mm
Exp. 150mm
0.7 Exp 50mm 0.7
Num. 150mm 0.05 Num. 150mm
Exp. 350mm Num 50mm Exp. 350mm
Num. 350mm Num. 350mm
0.6 0 0.6

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Fig. 16. Data by Persson [10,18,19], concrete Mix 2: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

a 0.25
b 1 c 0.25
Relative Humidity [-]

0.2 0.2
wn/c [kg/kg]
wn/c [kg/kg]

0.9
0.15 0.15
Non-evaporable water
Sealed curing Self-Desiccation Non-evaporable water
0.1 w/c = 0.47
0.1 Air curing - w/c = 0.47
w/c = 0.47
Exp.
0.8 Exp. Exp 50mm
Num. Num. 0.05 Num 50mm
0.05

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 e 0.25 f 1
Relative Humidity [-]
Relative Humidity [-]

0.2
0.9 0.9
wn/c [kg/kg]

Water Curing
0.15 w/c = 0.47
0.8 Num. 50mm
0.8 Air Curing - w/c = 0.47
Num 50mm Non-evaporable water Exp. 50mm
Num. 150mm
0.1 Water - w/c = 0.47 Exp. 150mm
Exp. 150mm Exp 50mm Num. 150mm
0.7 Num. 150mm 0.7 Exp. 350mm
0.05 Num 50mm
Exp. 350mm Num. 350mm
Num. 350mm
0.6 0 0.6

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Fig. 17. Data by Persson [10,18,19], concrete Mix 3: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

emphasizes its increase for increasing water-to-cement ratio. At the chemically bound water does not monotonically increase with
the end of the tests (450 days), mixes with s=c ¼ 0:1 show a time toward an asymptotic value, but rather shows a tendency to
70% increase of wn =c from about 0.1 for w=c ¼ 0:22 (Fig. 19a) to decrease after about 100 days of curing. This is due to silicate
about 0.17 for w=c ¼ 0:48 (Fig. 22a). In addition, for these mixes, polymerization, which produces free evaporable water. For com-
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 321

a 0.25 b 1 c 0.25

Relative Humidity [-]

0.2 0.2

wn/c [kg/kg]
wn/c [kg/kg]

0.9
0.15 0.15
Self-Desiccation
Non-evaporable water Air curing - w/c = 0.58
0.1 w/c = 0.58 0.1
Sealed curing 0.8 Exp. Exp 50mm
w/c = 0.58 Num 50mm
Num.
0.05 Exp. 0.05
Num.

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 e 0.25 f 1
Relative Humidity [-]

Relative Humidity [-]

0.9 0.2
0.9
wn/c [kg/kg]

Water Curing
Air Curing
0.15 w/c = 0.58
0.8 w/c = 0.58 0.8 Exp 50mm
Exp. 50mm Water curing - w/c = 0.58 Num. 50mm
Num. 50mm
0.1 Exp 50mm Exp. 150mm
Exp. 150mm Num 50mm Num. 150mm
0.7 Num. 150mm 0.7
0.05 Exp. 350mm
Exp. 350mm Num. 350mm
Num. 350mm
0.6 0 0.6
0 200 400 600 0 200 400 600 0 200 400 600
Time [days] Time [days] Time [days]

Fig. 18. Data by Persson [10,18,19], concrete Mix 4: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

a 0.25 b 1 c 0.25
Non-evaporable water Non-evaporable water
Sealed curing Self-Desiccation Air curing - w/c = 0.22-S
Relative Humidity [-]

0.2 w/c = 0.22-S w/c = 0.22-S 0.2 Exp 50mm

Exp. Exp.
Num 50mm
wn/c [kg/kg]

wn/c [kg/kg]

Num. 0.9 Num.

0.15 Num. no polymer. Num. no polymer.
0.15

0.1 0.1
0.8
0.05 0.05

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Water Curing - w/c = 0.22-S

d 1 e 0.25 f 1 Exp 50mm
Air Curing - w/c = 0.22-S Non-evaporable water Num 50mm
Water curing - w/c = 0.22-S
Relative Humidity [-]

Exp. 50mm Exp. 150mm

Relative Humidity [-]

Num. 50mm 0.2 Exp 50mm Num. 150mm

0.9 Exp. 150mm
0.9 Exp. 350mm
Num 50mm
wn/c [kg/kg]

Num. 150mm Num. 350mm

Num. no polymer.
Exp. 350mm 0.15
Num. 350mm 0.8
0.8
0.1

0.7 0.7
0.05

0.6 0 0.6

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Fig. 19. Data by Persson [10,18,19], concrete Mix 5: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

pleteness, sub-figures a, c, and e in Figs. 19–22 show both the Sub-figures b in Figs. 19–22 show the evolution of rela-
numerical results with (solid line) and without (dashed line) the tive humidity under sealed conditions (self-desiccation). The
effect of the polymerization. relative humidity decreases with time and its reduction is higher
322 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

a 0.25 b 1 c 0.25
Non-evaporable water Non-evaporable water

Relative Humidity [-]

Sealed curing Self-Desiccation Air curing - w/c = 0.24-S
0.2 w/c = 0.24-S w/c = 0.24-S 0.2
Exp. Exp 50mm
Exp.

wn/c [kg/kg]
0.9
wn/c [kg/kg]

Num. Num 50mm

Num.
0.15 Num. no polymer. 0.15 Num. no polymer.

0.1 0.1
0.8

0.05 0.05

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 e 0.25 f 1
Water Curing - w/c = 0.24-S
Exp 50mm
Air Curing Num 50mm
w/c = 0.24-S Non-evaporable water

Relative Humidity [-]

Relative Humidity [-]

Exp. 150mm
Exp. 50mm
0.2 Water curing - w/c = 0.24-S Num. 150mm
0.9 Num. 50mm Exp 50mm 0.9 Exp. 350mm
wn/c [kg/kg]

Exp. 150mm Num 50mm Num. 350mm

Num. 150mm 0.15 Num. no polymer.
Exp. 350mm
0.8 Num. 350mm 0.8
0.1

0.7 0.7
0.05

0.6 0 0.6

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Fig. 20. Data by Persson [10,18,19], concrete Mix 6: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

a 0.25 b 1 c 0.25
Relative Humidity [-]

0.2 Self-Desiccation 0.2

w/c = 0.36-S
wn/c [kg/kg]

wn/c [kg/kg]

0.9 Exp.
0.15 Num. 0.15

0.1 Non-evaporable water 0.1 Non-evaporable water

Sealed curing 0.8 Air curing - w/c = 0.36-S
w/c = 0.36-S
Exp. Exp 50mm
0.05 0.05 Num 50mm
Num.
Num. no polymer. Num. no polymer.
0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d 1 Air Curing - w/c = 0.36-S e 0.25 f 1

Exp. 50mm
Num. 50mm
Relative Humidity [-]

Relative Humidity [-]

Exp. 150mm 0.2

0.9 Num. 150mm 0.9
wn/c [kg/kg]

Exp. 350mm
Num. 350mm 0.15
0.8 0.8
Non-evaporable water Water Curing - w/c = 0.36-S
0.1 Water curing - w/c = 0.36-S Exp 50mm
Num 50mm
Exp 50mm
0.7 Num 50mm 0.7 Exp. 150mm
0.05 Num. 150mm
Num. no polymer. Exp. 350mm
Num. 350mm
0.6 0 0.6
0 200 400 600 0 200 400 600 0 200 400 600
Time [days] Time [days] Time [days]

Fig. 21. Data by Persson [10,18,19], concrete Mix 7: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

for lower water-to-cement ratio: at 450 days h  0.75 (25%) 0.22 and to less than 0.9 (10%) for w/c = 0.48, after only 100 days
for w/c = 0.25 and h  0.95 (5 %) for w/c = 0.58. The presence of of curing.
silica fume further intensifies the self-desiccation process. With Sub-figures d in Figs. 19–22 show the evolution of relative
s/c = 0.1, the relative humidity drops to 0.75 (25%) for w/c = humidity at three different locations in specimens subjected to
 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324 323

c
a 0.25 b 1 0.25
Self-Desiccation

Relative Humidity [-]

w/c = 0.48-S
0.2 Exp.
0.2

wn/c [kg/kg]
0.9
wn/c [kg/kg]

Num.
0.15 0.15
Non-evaporable water
0.1 Non-evaporable water 0.1 Air curing - w/c = 0.48-S
Sealed curing 0.8
w/c = 0.48-S Exp 50mm
Exp. Num 50mm
0.05 Num.
0.05
Num. no polymer.
Num. no polymer.

0 0.7 0

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

d Air Curing - w/c = 0.48-S

e 0.25
1
Exp. 50mm
Num. 50mm f 1
Exp. 150mm

Relative Humidity [-]

Relative Humidity [-]

Num. 150mm
Exp. 350mm 0.2
0.9 Num. 350mm
0.9
wn/c [kg/kg]

0.15 Water Curing

w/c = 0.48-S
0.8 0.8 Exp 50mm
Non-evaporable water
0.1 Water curing - w/c = 0.48-S
Num 50mm
Exp. 150mm
Exp 50mm Num. 150mm
0.7 0.7
0.05 Num 50mm Exp. 350mm
Num. no polymer. Num. 350mm

0.6 0 0.6

0 200 400 600 0 200 400 600 0 200 400 600

Time [days] Time [days] Time [days]

Fig. 22. Data by Persson [10,18,19], concrete Mix 8: non-evaporable water and relative humidity under sealed conditions (a, b), air curing (c, d), and water curing (e, f).

air curing (drying). During drying the reduction of relative humid- mental data available in the literature. Based on the obtained re-
ity is due to both release of water in the environment and self-des- sults it is possible to draw the following conclusions:
iccation. Both the experimental data and the numerical results
show that the overall relative humidity reduction during drying in- 1. The present formulation can simulate correctly and predict
creases for decreasing water-to-cement ratio. This is a direct con- with great accuracy the evolution of cement hydration and sil-
sequence of the self-desiccation process which is more ica fume (pozzolanic) reaction for concrete mixes characterized
pronounced for low values of water-to-cement ratio. Low values by a wide range of water-to-cement ratio and low silica fume
of water-to-cement ratio also influence the gradient of the humid- content. This is achieved by the formulation of evolution laws
ity profile, which tends to be more uniform along the specimen ra- for the hydration degree and for the silica fume reaction degree.
dius. This is particularly true in presence of silica fume. At 450 2. For high silica fume content mixes, rarely used in practise,
days, the relative humidity gradient between points at 50 mm experimental data are not available in the literature and further
and 350 mm from the exposed surface is less than 0.05 for Mix 5 research is needed to verify the applicability of the present
(low w/c and silica fume) whereas is almost 0.2 for Mix 8 (high model to these cases.
w/c and silica fume). 3. The correct simulation of the effect of hydration and pozzolanic
Finally, sub-figures f in Figs. 19–22 show the evolution of rela- reaction allows an accurate simulation of the phenomena that
tive humidity at three different locations in specimens subjected to characterize early age behavior of standard and high-
water curing (wetting). The relative humidity initially decreases at performance concrete mixes, such as self-desiccation and
all locations as a consequence of self-desiccation. Later, the water self-heating. This capability is an essential prerequisite to simu-
supplied from the environment is not completely consumed by late early age cracking in massive structures (e.g. dams), where
the chemical reaction and the relative humidity increases again. the heat released by the chemical reactions may be significant,
This phenomenon (wetting) is, of course, faster at points located and high-performance structures (e.g. bridge decks), where
close to the external boundaries. It must be noted here, that in this shrinkage induced by self-desiccation cannot be neglected.
case the agreement between the experimental data and the 4. In addition to early age self-desiccation data, the present model
numerical results is not as good as in the other cases, even if it can simulate medium- and long-term increase of free (evapor-
can be still considered satisfactory. The reduced accuracy of the able) water and relative humidity under sealed conditions, as
numerical predications in the case of wetting could depend by shown by some experimental data relevant to concrete mixes
the fact that: (1) the model does not distinguish between the containing silica fume [10]. This is achieved by: (a) assuming
adsorption isotherm and desorption isotherm and (2) sorption/ that this phenomenon is caused by silicate polymerization
desorption isotherms have been calibrated on the basis of desorp- (which is known to release some of the chemically bound
tion data. This point certainly requires further investigation and fu- water) and (b) modeling polymerization through an evolution
ture research. law for the silicate polymer concentration. The present model
is able to reproduce the existing experimental data that features
11. Conclusions the effect of polymerization. Nevertheless, due to the fact that
polymerization data is scarce in the literature, additional cali-
In this paper the model presented in Part I of this study [1] has bration and validation is certainly needed for this component
been calibrated and validated by an extensive analysis of experi- of the model.
324 G. Di Luzio, G. Cusatis / Cement & Concrete Composites 31 (2009) 309–324

5. Finally, the model can simulate and predict with unprecedented References
accuracy drying experimental data for a wide range of standard
and high-performance concrete mixes, with and without silica [1] Di Luzio G, Cusatis G. Hygro-thermo-chemical modeling of high performance
concrete. I: Theory. Cem Concr Compos, in press, doi:10.1016/
fume. j.cemconcomp.2009.02.015.
[2] Belytschko T, Liu WK, Moran B. Nonlinear finite elements for continua and
Since hydration degree and silica fume reaction degree provide structures. New York: John Wiley and Sons; 2000.
[3] Bažant ZP, Najjar LJ. Nonlinear water diffusion in nonsaturated concrete. Mater
the extent of concrete aging, the same theory can be used to sim- Struct 1972;5(25):3–20.
ulate the early age effects on strength (strength build-up), nonlin- [4] Pantazopoulo SJ, Mills RH. Microstructural aspects of the mechanical response
ear mechanical behavior, and fracturing. This subject is part of an of plain concrete. ACI Mater J 1995;92(6):605–16.
[5] Norling Mjönell K. A model on self-desiccation in high-performance concrete.
ongoing research effort by the authors.
In: Self-desiccation and its importance in concrete technology, proceedings of
the international research seminar, Lund, Sweden; 1997. p. 141–57.
Acknowledgement [6] Cervera M, Oliver J, Prato T. Thermo-chemical–mechanical model for concrete.
I: hydration and aging. J Eng Mech, ASCE 1999;125(9):1018–27.
[7] de Schutter G, Taerwe L. General hydration model for portland cement and
The first author gratefully acknowledges the financial support blast furnace slag cement. Cem Concr Res 1995;25:593–604.
of CTG (Italcementi Group) through the CIS-E Consortium. [8] Xi Y, Bažant ZP, Jennings HM. Moisture diffusion in cementitious materials –
moisture capacity and diffusivity. Adv Cem Based Mater 1994;1:258–66.
[9] Taylor HFW. Cement chemistry. Telford (London): Thomas; 1997.
Appendix A [10] Persson B. Hydration and strength of high performance concrete. Adv Cem
Based Mater 1996;3:107–23.
[11] Xi Y, Bažant ZP, Jennings HM. Moisture diffusion in cementitious materials –
For a tree-node triangular element the shape functions can be
adsorption isotherms. Adv Cem Based Mater 1994;1:258–66.
expressed as N i ðr; zÞ ¼ 0:5ðai þ bi r þ ci zÞ=Xe ði ¼ 1; . . . ; 3Þ, where [12] Baroghel-Bouny V, Mainguy M, Lassabatere T, Coussy O. Characterization and
r = radial coordinate, z = longitudinal coordinate, and identification of equilibrium and transfer moisture properties for ordinary and
high-performance cementitious materials. Cem Concr Res 1999;29:1225–38.
Xe ¼ element area [2]. Consequently, one has @Ni =@r ¼ 0:5bi =X,
[13] Pane I, Hansen W. Investigation of blended cement hydration by isothermal
and @N i =@z ¼ 0:5ci =X. By using the previous expressions and by calorimetry and thermal analysis. Cem Concr Res 2005;35:1155–64.
solving the integrals in Eqs. (5)–(7) with a one-point gauss formula, [14] Bentz DP, Waller V, De Larrard F. Prediction of adiabatic temperature rise in
one obtains conventional and high-performance concretes using a 3-D microstructural
model. Cem Concr Res 1998;28(2):285–97.
 1 Xe
C  h ele
D  2 Xe
C [15] Lura P, Jensen OM, van Breugel K. Autogenous shrinkage in high-performance
Kele
1 ¼ A; Kele
2 ¼  B ; Fele
1 ¼ I ð11Þ cement paste: an evaluation of basic mechanisms. Cem Concr Res
12 4Xe 3 2003;33:223–32.

C3X e 
k ele 
C4X e [16] Ayano T, Wittmann FH. Drying, moisture distribution, and shrinkage of cement
Kele
3 ¼ A; Kele
4 ¼  eB ; Fele
2 ¼ I ð12Þ based materials. Mater Struct 2002;35:134–40.
2 12 3 4
2 3 X 3 [17] Kim J-K, Lee C-S. Moisture diffusion of concrete considering self-desiccation at
2 1 1 1 early age. Cem Concr Res 1999;29:1921–7.
6 7 6 7 [18] Persson B. Moisture in concrete subjected to different kinds of curing. Mater
A ¼ 41 2 15 I ¼ 415 ð13Þ Struct 1997;30:533–44.
1 1 2 1 [19] Persson B. Seven-year study on the effect of silica fume in concrete. Adv Cem
2 2
3 Based Mater 1998;7:139–55.
b1 þ c21 þ d1 b1 b2 þ c 1 c 2 þ d2 b1 b3 þ c1 c3 þ d3 [20] Neville AM. Properties of concrete. New York: John Wiley and Sons; 1997.
6 7 [21] Bažant ZP, Kaplan MF. Concrete at high temperatures: material properties and
Bele 6
¼ 4 b1 b2 þ c1 c2 þ d1 2
b2 þ c22 þ d2 b2 b3 þ c2 c3 þ d3 7
5 mathematical models. London: Longman (Addison-Wesley); 1996.
2 [22] Mindess S, Young JF. Concrete. Englewood Cliffs, NJ: Prentice-Hall; 1981.
b1 b3 þ c1 c3 þ d1 b2 b3 þ c2 c3 þ d2 b3 þ c23 þ d3 [23] Waller V, de Larrard F, Roussel P. Modelling the temperature rise in massive
ð14Þ HPC structures. In: 4th international symposium on utilization of high-
strength/high-performance concrete, RILEM; May 1996. p. 415–21.
where the overbar means evaluation at the centroid of the element, [24] Brough AR, Dobson CM, Richardson IG, Groves GW. In situ solid-state NMR
studies of Ca3SiO5: hydration at room temperature and at elevated
and di ¼ 2bi Xe =3r for axis-symmetric meshes and di ¼ 0 for two- temperatures using 29Si enrichment. J Mater Sci 1994;29:3926–40.
dimensional planar meshes.
 Cement & Concrete Composites 31 (2009) 325–330

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

The physical and chemical effects of long-term sulphuric acid exposure

on hybrid modified cement mortar
Gengying Li *, Guangjing Xiong, Yunhai lü, Yegao Yin
Dept. of Civil Eng., Shantou Univ., Daxue Road, Shantou, Guangdong 515063, PR China

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

Article history: The properties of a hybrid modified cement mortar (HM) after 5 years storage in a sulphuric acid solution
Received 28 January 2007 were studied. HM was made by mixing ordinary portland cement (OPC), sand and water and subse-
Received in revised form 14 February 2009 quently adding Na2SiO3 (water glass), Na2SiF6, polyvinyl acetate, lignosulphonate, and tributylphosphate.
Accepted 20 February 2009
A control mortar (CM) was prepared by mixing OPC, sand, and water. The compressive strength losses
Available online 1 March 2009
after sulphuric acid attack were measured. The microstructures of HM and CM after sulphuric acid attack
were observed and analyzed by using scanning electron microscopy (SEM), and the porosity and the pore
Keywords:
size distribution were measured by using mercury intrusion porosimetry (MIP). Fourier-transform infra-
Na2SiO3
Polyvinyl acetate
red (FT-IR) spectroscopy was utilized to determine the phases present in HM and CM after sulphuric acid
Hybrid modification attack. Test results show that many cracks appeared in the near-surface region and many needle-like
Acid resistance crystals in the centre region of CM after 5 years sulphuric acid exposure, leading to a higher porosity
Cement mortar of 28.6% and a higher strength loss of 50.6%. By contrast, a basically monolithic structure was kept in
HM after 5 years sulphuric acid exposure, resulting in a lower porosity of 13.3% and a lower strength loss
of 27.4%.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction corroded mortars were observed by scanning electron microscopy

(SEM), and the porosity and pore size distribution were measured
In many parts of the world, most expenditure on sewerage is using mercury intrusion porosimetry (MIP). The phases present
allocated for improving the structural performance and capacity after sulphuric acid attack were analyzed with Fourier-transform
of the existing network. Most collapses and stoppages of sewers infrared (FT-IR) spectroscopy. Compressive strength losses were
are the results of sulphuric acid attack. This is because sulfur con- also measured.
taining compounds in sewage are decomposed by anaerobic bacte-
ria producing H2S, which is either dissolved in water or
2. Methods
metabolized with oxygen by aerobic bacteria to produce sulphuric
acid. The sulphuric acid attack is also supplemented by the poten-
2.1. Materials
tial discharge of sulphuric acid from the chemical industry [1,2].
Good sulphuric acid resistance, therefore, is essential for sew-
A Na2SiO3 with a specific gravity of 1.44 g/cm3 and a viscosity
ers[3,4]. In order to obtain a higher sulphuric acid resistant mortar
index of 1.6 mPs a, was chosen, and a Na2SiF6 was taken as the
at a low cost, a Na2SiO3 and five polymer latexes were chosen to
hardener of Na2SiO3. OPC, fly ash, and sand with a fineness modu-
produce five kinds of modified cement mortars in a previous exper-
lus of 2.35 were also used. The chemical composition and physical
imental study [4], in which the Na2SiO3–polyvinyl acetate latex hy-
properties of the cement and fly ash are presented in Table 1. A
brid modified cement mortar (HM) had been found to be the best.
commercially available lignosulphonate was used as the superp-
The modified effect of HM is similar to that of the other research
lasticizer and tributylphosphate was used as the antifoaming
[5].
agent. A polyvinyl acetate latex with a pH value of 9.0 and a
In order to understand the long-term behavior of HM under sul-
specific gravity of 1.19 g/cm3 was used.
phuric acid attack, in this paper, two kinds of mortars, namely HM
and CM (control mortar), were made and stored in a sulphuric acid
2.2. Experimental program
solution up to 5 years. The morphology and microstructure of the
The mix proportions of HM and CM are presented in Table 2. In
* Corresponding author. the experiment, materials for HM were mixed in the following or-
E-mail address: gyli@stu.edu.cn (G. Li). der. Water, cement, and sand were mixed for 2 min. In a separate

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.014
326 G. Li et al. / Cement & Concrete Composites 31 (2009) 325–330

Table 1
Chemicalcomposition and physical properties of OPC and fly ash.

Chemical analysis SiO2 Al2O3 Fe2O3 CaO MgO SO3 Na2O LOI Total Specific surface, Blaine (m2/ 28 days Compressive strength
(wt.%) percentage kg) (Mpa)
OPC 19.5 4.4 6.22 65.9 1.5 1.09 0.30 1.43 100 462 45.9
Fly ash 52.5 19.1 8.2 14.7 1.98 0.35 0.48 3.6 100 565 –

Table 2 FT-IR spectrometer at a resolution of 4 cm 1 with an accumulation

Mix proportions by weight of HM and CM. of 4000 scans. The unmodified cement mortar cured under a nor-
Material HM CM
mal atmosphere up to 5 years. UCM (air-cured CM) was also pre-
pared for FT-IR spectroscopy.
OPC 100 100
Na2SiO3/ Lignosulphonatea 7/0.05
Na2SiF6 1.05 3. Results and discussion
Polyvinyl acetate latex 3
Lignosulphonate 0.5
Three colored zones could be identified in HM and CM after
Tributylphosphate 6
Fly ash 15 15 5 years in sulphuric acid (Fig. 1). The near-surface zone was white
Sand 200 200 in color and reached a depth of 50 mm in HM and 75 mm in CM
Water 55 55 due to corroding. The transition zone was the part of mortar where
a
Soluble glass and Lignosulphonate were mixed before producing modified color changed from gray to white with a depth of 3–4 mm indicat-
mortar. ing partial corrosion. The centre zone was gray in color, appearing
to be uncorroded.
container polyvinyl acetate latex, Na2SiO3 and fly ash were pre-
3.1. Mechanical properties
mixed for 2 min. Tributylphosphate in the amount of 0.15 vol% of
polymer latex was added to the mix. Then, the two mixtures com-
The compressive strength test results of HM and CM mortars
bined with a further 5 min mixing until a uniform mortar was
with and without sulphuric attack are presented in Table 3. The
obtained.
strength test results indicate HM has a higher resistance to sulphu-
As shown in Table 2, two groups of specimens, namely HM and
ric acid attack, with a strength loss of 29.4% compared to 50.6% for
CM, were made, and each group of specimens was composed of
CM. It should be noted that the strengths are an indication of the
eighteen 31.5  31.5  50 mm prisms. Six prisms from HM and
combined resistance of zones A, B, and C, since the microstructure
CM were used solely to determine the compressive strength after
is affected non-uniformly through the specimen volume with the
28 days. Another six prisms were cured in air up to 5 years. The
near-surface zones more affected. It should also be mentioned that
remaining six prisms were immersed into a 1% sulphuric acid solu-
specimen size effect had not been included in this test. If large size
tion with a volume of 10 L at room temperature up to 5 years. The
specimens were used, the strength losses would be lower.
sulphuric acid solutions were exchanged at intervals of 30 days.

3.2. Morphology and microstructure

2.3. Specimen characterization

Three HM and three CM samples with a size of 1  1  1 cm

2.3.1. Compressive strength
were taken at three regions (the near-surface zone, the transition
The compressive strengths of HM and CM were measured after
zone, and the centre zone) as shown in Fig. 1 for SEM.
28 days curing and again after 5 years exposure to the atmosphere
Portlandite, gypsum, ettringite or other phases could not be
and sulphuric acid solution.
identified in Zone A, which was characterized by numerous inter-
connecting cracks for CM under sulphuric acid attack up to 5 years
2.3.2. SEM
(Fig. 2), probably because the reaction products of H2SO4and port-
Morphology and microstructure of HM and CM samples were
landite had dissolved in water. The observed cracks were generally
observed by using a Philip XL 30 scanning electron microscope
20 lm in length and 2 lm in width, with the largest one being
with an EDX analyzer at an accelerating voltage of 25 kV. The sam-
80 lm long and 2 lm wide. Similarly for HM, portlandite, gypsum,
ples after sulphuric acid attack were dried in a vacuum-dried oven
ettringite or other phases were not observed in the near-surface
at 50 °C for 24 h, and sputter coated with 100 A of Au–Pd prior to
zone A and contained cracks (Fig. 3), although these were generally
imaging.
smaller, measuring generally 2 lm in length and 1 lm in width. A
key difference between Zone A in CM and HM was the develop-
2.3.3. Porosity and pore size
The porosity and pore size distribution of HM and CM were
determined using an AUTOSCA-10 MIP with a measure range of
2–5000 nm radius. The maximum pressure of the porosimeter
was 600 MPa, values of 140° and 480.0 N/cm2 were used for the
contact angle and mercury surface tension, respectively. Six HM A B C
and six CM samples were dried in a vacuum-dried oven for 24 h
before MIP.

Nearsurface Transition zone Center zone

2.3.4. FT-IR Spectroscopy
zone
Samples of HM and CM were prepared for FT-IR spectroscopy by
White White to gray Gray
drying to a constant weight, ground to pass through a 45 lm sieve
and coated on a KBr pellet. FT-IR spectra were subsequently col- Fig. 1. Schematic representation on the three zones (labeled A, B, C) identified in
lected in the mid-region (400–5000 cm 1) using a Nicolet 5DXC HM and CM after 5 years exposure to sulphuric acid.
 G. Li et al. / Cement & Concrete Composites 31 (2009) 325–330 327

Table 3
Compressive strengths of HM and CM cured in air (28 days and 5 years) and sulphuric acid (5 years).

Sample Compressive strength after 28 days Compressive strength after 5 years Compressive strength after 5 years stored in Strength loss after 5 years
cured in air (MPa) cured in air (MPa) sulphuric acid solution (MPa) sulphuric attack (%)
HM 20.7 ± 5.3% 24.8 ± 4.2% 17.5 ± 5.8% 29.4
CM 20.3 ± 4.7% 23.5 ± 2.6% 11.6 ± 4.5% 50.6

Fig. 2. SEM images and EDX analysis of zone A for CM after 5 years exposure to sulphuric acid.

Fig. 3. SEM images and EDX analysis of zone A for HM after 5 years exposure to sulphuric acid.

Fig. 4. SEM images and EDX analysis of zone B for CM after 5 years exposure to sulphuric acid.
328 G. Li et al. / Cement & Concrete Composites 31 (2009) 325–330

ment of a netlike film in the latter. The development of coherent
polymer films has been found elsewhere [6] to be not well devel-
opment, in a modified mortar, therefore the formation of the net-
like film in HM may be explained by the interaction of Na2SiO3
and polyvinyl acetate latex. The chemical analyses performed by
EDX in the near-surface zone A for HM and CM are shown in
Fig. 2b and 3b, indicating the different Ca/Si, carbon, and sulfur
percentage. For CM, a low Ca/Si atomic ratio and a high atomic sul-
fur percentage were found, probably because CaSO4
vH2O, the
reaction product of CASAH and H2SO4, dissolved in water, leading
to the loss of Ca from CASAH.
Fig. 4 shows that the dominance needle-like crystals in zone B
for CM, the crystals were generally 50 lm long, with the largest
one being 100 lm long and 1 lm wide. Some needle-like crystals
(generally 10 lm long) were also observed in zone B for HM
(Fig. 5). A netlike film can be seen in the zone B for HM. The chem-
ical analyses performed by EDX in zone B for HM and CM are
shown in Fig. 4b and 5b indicate that the samples contain sulfur
and carbon containing compounds. Evidence suggests that the
crystals may be a mixture of ettringite and thaumasite which is
in agreement with the research result of Brown and Hooton [7].
In zone C for CM (Fig. 6), needle-like crystals about 5 lm long
can be observed; the number and the length of the crystals, how-
ever are less than those in zone B. As shown in Fig. 7, no needle-like
crystals can be found, and a polymer film can be observed in zone C
for HM, indicating a denser microstructure compared with CM. The
chemical analyses of zone C performed by EDX for HM and CM are
shown in Fig. 6b and 7b; for UMHM, sulfur atomics were found,
indicating that a sulphuric acid attack had already happend; while
in the case of HM, carbon atomics were found, suggesting the pres-
ence of polymer compounds.
3.3. FT-IR Spectroscopy
Three HM and three CM samples were taken at each transition
zone for FT-IR spectroscopy. The FT-IR spectrum of U CM (air-cured
CM)showed bands at 3643, 3436, 1632, 1428, 980, 779, 678, and
457 cm 1 (Curve A, Fig. 8), respectively. According to Geidel et al.
[8], the band observed at 980 cm 1 can be assigned to SiAO
stretching modes of silanol groups in the CASAH phase present
in the cement; the absorption band due to OAH stretching in
Ca(OH)2 is present at 3643 cm 1. According to Ghosh [9], the bands
at 1632 and 3436 cm 1 can be explained by the presence of cal-
cium sulphate in the form of ettringite (mono and trisulphate) in
the cement; the bands at 779 cm 1 and 678 cm 1 are consistent
with b-C2S and a-C2S phases, respectively.
The spectrum of CM (Table 4) showed bands at 3608, 3552,
3410, 1620, 1426, 1153, 1096, 1007, 874, 797, 779, 661, 601, and
461 cm 1, respectively. The majority of these bands can be ex-
plained by the presence of CaSO4
vH2O (the spectrum of CaSO4.2-
H2O shows bands at 1000, 492, 1131, 1142, 602–669, 3555, and
1690 cm 1 or 1006, 413, 1118, 1138, 3500, and 1629 cm 1, respec-

Fig. 5. SEM image and EDX analysis of zone B for HM after 5 years exposure to sulphuric acid.

Fig. 6. SEM image and EDX analysis of zone C for CM after 5 years exposure to sulphuric acid.
 G. Li et al. / Cement & Concrete Composites 31 (2009) 325–330 329

Fig. 7. SEM image and EDX analysis of zone C for HM undergone after 5 years exposure to sulphuric acid.

Fig. 8. FT-IR spectra of UCM (CM air cured for 5 years), CM and HM after 5 years exposure to sulphuric acid.

tively; the spectrum of CaSO4.1/2H2O shows bands at 1012, 465,

Table 4
1158, 1120, 3615, and 1629 cm 1, respectively [10,11]). The ab-
Characteristic transmittance bands for UCM, CM and HM.
sence of the band at 3643 indicates the loss of portlandite by its
dissolution in sulphuric acid. UCM CM HM Assignment
The IR spectrum of HM (Curve C, Fig. 8) showed bands at 3416, 3643 3608 OH bending in CASAH
2926, 1622, and 1417 cm 1, indicating the presence of polymer la- 3438 3552 OH bending in CASAH
3410 3414 OH bending in CASAH
tex. Polymer latex accounts for the broad hydroxyl band centred at
2926 CAH vibration in R1COOR2
about 3400 cm 1, the CAH absorptions at 2926 cm 1, and the car- 2515
boxylate group absorptions at 1622 and 1417 cm 1. 1796 C@O vibration in R1COOR2
1632 1620 1622 OAH bending of water
1428 CO23 vibration
3.4. Porosity and pore size distribution 1417
1153
980 1096 1096
The porosity test results for HM and CM after 5 years sulphuric
1007 SiO stretching
acid attack are shown in Table 5. The total porosity of HM was 779 874 874 CO23 vibration
13.25%. It was much lower than that of CM (28.62%), indicating a 797
denser structure for HM. Moreover, the pores with a size 779 778
d P 50 nm in HM are 3.98%, about 20.32% in that of CM. Mean vol- 678 713
661 662 AlO6 vibration
ume pore radius (MVPR) refers to the minimum radius of pores
601 601 SO4 vibration in CaSO4
vH2O
when half of the total intrusion has taken place. The MVPR value 457 456 459 SiO4 vibration in CASAH
of HM is about 11.26 nm, but the MVPR value of CM is much high-
er, about 31.20 nm. It should be noted that the measured porosities
and pore size distributions were an indications of the combined non-uniform changes in porosity through the specimen thickness
resistance of zones A, B, and C, since sulphuric acid attack caused with the surface regions obviously more affected.
330 G. Li et al. / Cement & Concrete Composites 31 (2009) 325–330

Table 5
Porosity, mean radius and pore size distribution of HM and CM.

Sample Total intruded vol Mean radius Apparent density Porosity Porosity Porosity Porosity
(cm3/g) (nm) (g/cm3) (2 nm < da < 5nm) (%) (d 6 10 nm) (%) (10 nm < d < 50 nm) (%) (d P 50 nm) (%)
CM 0.1655 31.60 1.729 28.62 2.29 6.01 20.32
HM 0.0727 11.26 1.823 13.25 5.30 3.97 3.98
a
d is the mean pore diameter.

4. Conclusions Ministry of Science &Technology of China, and Science and Tech-

nology Plan (2007B020714003).
Based on the results of the experiments, the following conclu-
sions can be drawn: References

[1] De Belie N, Monteny J, Beeldens A, Vincke E, Van Gemert D, Verstraete W.

1) After 5 years sulphuric acid attack, the compressive strength
loss of Na2SiO3–polyvinyl acetate latex modified mortar
(HM) was 29.4%, whereas the control mortar (CM) had a
higher strength loss of 50.6%.
2) After 5 years sulphuric acid attack, three distinct zones
(near-surface zone – Zone A, transition zone – Zone B, and
central zone – Zone C) could be identified for HM and CM.
For CM, Zone A was characterized by SEM as being heavily
cracked, Zone B containing a mass of needle-like crystals,
and Zone C having a few cracks. For HM, a netlike film could
be found in all three Zones, with less cracking and the same
crystals identified in CM, but in smaller quantities.
3) The possible interaction between Na2SiO3 and polyvinyl ace-
tate, interpreted from the FT-IR spectra, leads to the forma-
tion of the netlike film in the HM, making HM more resistant
to sulphuric acid attack than CM.
4) The porosity of HM was 13.25%, while the porosity of CM
was about 28.62%. The netlike film may be the main reason
for decreasing the porosity of HM.
[10] Newman R. Some applications of infrared spectroscopy in the examination of
Acknowledgments
[11] Nanru Yang. Test Method of Inorganic Nonmetal Materials. Wuhan University
The authors gratefully acknowledge the funding provided by
the National Natural Science Foundation of China (59778052),
Experimental research and modelling of the effect of chemical and biogenic
sulfuric acid on different types of commercially produced concrete sewer
pipes. Cem Concr Res 2004;34(12):2223–36.
[2] Neville A. The confused world of sulfate attack on concrete. Cem Concr Res
2004;34:1275–96.
[3] Sahmaran M, Kasap O, Duru K, Yaman IO. Effects of mix composition and
water–cement ratio on the sulfate resistance of blended cements. Cem Concr
Compos 2007;29:159–67.
[4] Xiong GJ, Chen XH, Li GY, Chen LQ. Sulphuric acid resistance of soluble soda
glass – polyvinyl acetate latex – modified cement mortar. Cem Concr Res
2001;31(1):83–6.
[5] Gündüz L, Bekar M, Sapci N. Influence of a new type of additive on the
performance of polymer-lightweight mortar composites. Cem Conc Compos
2007;29(8):594–602.
[6] Afridi MUK, Ohama Y, Demura K, Iqbal MZ. Development of polymer films by
the coalescence of polymer particles in powdered and aqueous polymer-
modified mortars. Cem Concr Res 2003;33(11):1715–21.
[7] Brown P, Hooton RD. Ettringite and thaumasite formation in laboratory
concretes prepared using sulfate-resisting cements. Cem Concr Compos
2002;24(3):361–70.
[8] Geidel E, Lechert H, Döbler J, Jobic H, Calzaferri G, Bauer F. Characterization of
mesoporous materials by vibrational spectroscopic techniques. Micropor
Mesopor Mater 2003;65(1):31–42.
[9] Ghosh SN. Cement and Concrete Science and Technology, vol. I, Part II. New
Delhi: ABI Books; 1992. p. 222–52. (Ghosh SN, editor. Cement and Concrete
Science and Technology, 2 vols. New Delhi, India: ABI Books, Ltd.; 1992)..
painting materials. J Am Inst Conserv 1979;19(1):42–6.
of Technology; 2005. pp. 236–256.
 Cement & Concrete Composites 31 (2009) 331–341

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Effectiveness of the use of steel fibres on the torsional behaviour

of flanged concrete beams
Constantin E. Chalioris *, Chris G. Karayannis
Democritus University of Thrace, Department of Civil Engineering, Xanthi 67100, Greece

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

Article history: The behaviour under torsion of reinforced concrete beams with steel fibres as mass reinforcement is
Received 18 March 2008 experimentally investigated. Short hooked-ended steel fibres with aspect ratio lf/df = 37.5 are used. Test
Received in revised form 30 January 2009 results of 35 beams with rectangular, L-shaped and T-shaped cross-sections tested in pure torsion are
Accepted 5 February 2009
presented and discussed. Various configurations of conventional and fibre steel reinforcement are exam-
Available online 14 February 2009
ined. The experimental program includes (i) plain concrete beams (control specimens), (ii) specimens
with longitudinal reinforcing bars and (iii) specimens with bars and stirrups. All cases are examined with
Keywords:
0%, 1% and 3% steel fibre volume fractions. The use of steel fibres as the only shear torsional reinforcement
Beams
Concrete
is also reported herein, in an attempt to examine the effectiveness of fibres as a potential replacement of
Steel fibre concrete stirrups. Test results indicated that fibrous concrete beams exhibited improved overall torsional perfor-
Tests mance with respect to the corresponding non-fibrous control beams. The addition of steel fibres was
Torsion essential to the tested beams without or with inadequate conventional steel reinforcement. Fibres pre-
vented the sudden brittle failure of both rectangular and non-rectangular beams and proved to be under
some circumstances adequate to provide for enhanced torsional moment capacities, even in the case of
full replacement of stirrups with steel fibres.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction concrete members, improvement of the poor performance of con-

The addition of steel fibres in concrete mix has long been recog-
nised as a non-conventional mass reinforcement that enhances the
mechanical properties of concrete and provides for crack propaga-
tion control [1,2]. This ability is attributed to the tensile stress
transfer capability of the steel fibres across crack surfaces, known
as crack-bridging, and also to the fact that such fibres provide sig-
nificant resistance to shear across developing cracks. Cracking of
steel fibre concrete requires debonding and pull-out of the ran-
domly distributed steel fibres in the concrete. Therefore, steel fibre
concrete demonstrates a pseudo-ductile tensile response and en-
hanced energy dissipations capacities, relative to the brittle behav-
iour of plain concrete [3]. This has inspired investigations of the
possibility of partially replacing conventional steel reinforcement
– mainly stirrups – by steel fibres, especially in cases where design
criteria recommend high steel ratio, which leads to very small stir-
rup spacing [4–6].
Concerning the phenomenon of torsion, early experimental ef-
forts pointed out that the behaviour of an element in pure torsion
is fully characterized by the behaviour of the material in direct
tension [7,8]. Thus, in order to enhance the torsional response of
* Corresponding author. Tel./fax: +30 25410 79632.
E-mail address: chaliori@civil.duth.gr (C.E. Chalioris).
crete in tension by incorporating steel fibres has been proposed
and extensively studied in the last decades [9–23]. Most of the
conducted research on the torsional behaviour of fibrous concrete
elements presents experimental results of rectangular beams
containing steel fibres without conventional reinforcement
[9–12] and rectangular beams with steel fibres, bars and stirrups
[13–18]. There is also an experimental investigation of fibrous con-
crete beams with circular cross-section under torsion [19]. All
these experimental studies have shown that the use of steel fibres
improves the cracking characteristics and the overall torsional
behaviour of concrete beams with rectangular and circular cross-
sections.
Analytical research on the torsional response of fibrous concrete
elements has also been reported in the literature. An efficient the-
oretical model for the prediction of the full torsional behaviour of
steel fibre concrete members without reinforcing bars has been
presented by Karayannis [20,21]. Further, the well-known softened
truss model theory [7] has been properly modified to include the
influence of steel fibres and has successfully been applied in fi-
brous concrete rectangular beams with steel bars and stirrups
[15,17,18]. Moreover, simplified semi-empirical expressions and
formulas for code modifications to calculate the ultimate torsional
capacity of fibrous concrete beams with rectangular cross-section
have also been proposed [9–14,22].

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.007
332 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341
It is obvious that despite the plethora of the carried out research
on this field of study and the fact that in real reinforced concrete
structures most of the beams have L- or T-shaped cross-section
(flanged beams), the published work on non-rectangular fibrous
concrete beams under torsion is quite limited and mainly con-
ducted by the authors [20,21,23]. Furthermore, the use of steel fi-
bres as the only shear reinforcement instead of the use of
common stirrups in torsional beams has not been thoroughly
investigated.
In this paper the influence of the steel fibres on the torsional
behaviour of L- and T-shaped beams is experimentally investi-
gated. Test results of 35 beams in pure torsion with rectangular,
L- and T-shaped cross-sections, with various reinforcement config-
urations are reported. The experimental program includes three
groups of specimens; (i) without conventional steel reinforcement,
(ii) with longitudinal reinforcing bars and (iii) with bars and stir-
rups (conventional steel reinforcement). The examined specimens
also contain different volume fractions of short steel fibres: 0%, 1%
and 3%.
Furthermore, a short review of the proposed analytical models
for the prediction of the torsional strength of steel fibre reinforced
concrete beams is including in this study. The ultimate torsional
moment of the tested fibrous beams without conventional rein-
forcement has been calculated using the analytical expressions of
Craig et al. [9], Mansur & Paramasivam [10] and Wafa et al. [7]
and the numerical model by Karayannis [21]. The entire torsional
behaviour of the tested rectangular fibrous beams with longitudi-
nal bars and stirrups has been predicted using the combined meth-
odology proposed by the authors [17] and the one by Gunneswara
Rao and Rama Seshu [18]. The purpose of this analytical investiga-
tion is to present and to compare the predictions of the proposed
analytical approaches in order to estimate their effectiveness and
accuracy on the experimental results of the tested beams of this
study.
The present study contributes to the limited existing literature
on torsional tests of fibrous concrete beams with non-rectangular
100 150 200
50
200
R Ls L Ts
(a) Beams of group P (without conventional steel reinforcement)
100 200
2 10 70 4 10 70
200
90
2 10 2 10
RL THsL
(b) Beams of group L (with longitudinal bars)
100 200
2 10 70 4 10
200 8/200
8/200
90
2 10 2 10
RR THsR
(c) Beams of group R (with longitudinal bars and stirrups)
Fig. 1. Cross-sectional dimensions and conventional steel reinforcement of the tested beams.
cross-section. Rectangular and comparable flanged beams are
jointly tested in order to acquire comparative results. Furthermore,
the use of steel fibres as the only shear reinforcement in torsion is
also presented herein, in an attempt to examine the effectiveness
of steel fibres as a potential replacement for stirrups.
2. Experimental program
The experimental program includes 35 rectangular and flanged
beams tested under pure torsion and sorted in three groups based
on the conventional steel reinforcement. Tested beams were con-
structed by plain concrete (control specimens) and steel fibre con-
crete with 1% and 3% volume fractions of the used fibres.
2.1. Specimen characteristics
Rectangular, L-shaped and T-shaped sections comprise the
examined cross-sectional schemes of the tested beams. Geometri-
cal characteristics and dimensions of the specimens’ cross-sections
are shown in Fig. 1. All beams have the same height and web
width; h/bw = 200/100 mm. The ratio of the flange width, bf, to
the web width is equal to bf /bw = 1 (rectangular specimens), 1.5
(L-shaped beams), 2 (L- and T-shaped beams) and 3 (T-shaped
beams).
Tested beams are sorted in three groups named as P, L and R,
based on their conventional steel reinforcement. The first group,
group P, includes 17 specimens; 3 with rectangular cross-section
and 14 L-shaped and T-shaped beams without any conventional
steel torsional reinforcement (without bars and stirrups). Second
group, group L, includes 9 specimens; 3 rectangular and 6 T-shaped
beams without stirrups and only longitudinal bars were added as
conventional steel reinforcement. Third group, group R, includes
9 specimens; 3 rectangular and 6 T-shaped beams with bars and
stirrups.
The longitudinal reinforcement of the beams of groups L and R
was the same; four deformed bars of diameter 10 mm (4£10) at
200 300 200 300
50 50 70 70
90 90
T THs TH
300
4 10
90
2 10
THL
300
70 4 10
8/200
90
2 10
THR
 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341 333

the corners of the rectangular beams and six deformed bars of
diameter 10 mm (6£10), 4£10 at the top and 2£10 at the bottom
of the T-shaped beams. The transverse reinforcement of the beams
of group R was 8 mm diameter mild steel closed stirrups at a uni-
form spacing of 200 mm. Steel yield strength was 415 MPa for the
bars and 344 MPa for the stirrups. Cross-sectional and conven-
tional steel reinforcement details of the specimens are shown in
Fig. 1 and summarized in Table 1.
Table 1 also presents the ratios of the steel reinforcements of
the tested beams. It is noted that the ratio values for the bars
and stirrups have been calculated based on the following
expressions:
Asl
q‘ ¼
Ac
Ast  pt
qt ¼
Ac  s
where: Ac: gross area of the concrete cross-section; Asl: total area of
steel longitudinal bars; Ast: area of one single-legged steel stirrup;
pt: perimeter of the stirrup; s: spacing of stirrups.
Beams from plain concrete (without fibres) were used as control
specimens (13 specimens). The steel fibre concrete beams (22
specimens) contained short hooked-ended steel fibres with length
to diameter ratio (aspect ratio) equal to lf/df = 30 mm/
0.80 mm = 37.5 (DRAMIX ZC 30/.80). Two different steel fibre vol-
ume fractions, Vf, were used; a moderate one equal to 1% or 78.5 kg
per 1 m3 concrete and a high one equal to 3% or 235.5 kg per 1 m3
concrete (see also Table 1).
The cement used in this experimental work was a locally man-
ufactured general purpose ordinary Portland type cement (type
35IIa, Greek type pozzolan cement containing 10% fly ash). Sand
with a high fineness modulus and coarse aggregates with a maxi-
mum size of 9.5 mm were used. The concrete mixture was made
using cement, sand and crushed stone aggregates in a proportion
1:2.8:1.2, respectively, and water to cement ratio equal to 0.43.
The maximum size of the used coarse aggregates (9.5 mm), the
length (lf = 30 mm) and the aspect ratio (lf /df = 37.5) of the used
steel fibres were chosen in accordance with the RILEM TC 162-
TDF recommendations [24], based on the minimum dimension of
the tested beams. It is known that the size and the quantity of
aggregates greater than 5 mm, the volume fraction and the aspect
ð1aÞ ratio of the added steel fibres strongly affect the workability of
concrete and the homogenous distribution of the fibres in the con-
ð1bÞ crete mix. Fibrous concrete mixtures with high volume fraction of
steel fibres (Vf = 3%) have workability difficulties. However, the use
of short fibres with low aspect ratio, the use of pozzolan cement
containing 10% fly ash and the following procedure for mixing
and casting the fibrous concrete mixture significantly reduced this
deficiency.
First, cement (with fly ash), sand and crushed stone aggregates
were dry-mixed for about 3 min in a 0.05 m3 laboratory mixer. Sec-
ond, free of clumps steel fibres were dispersed by hand gradually
and very slowly in small amounts to avoid fibre balling, while
mixing was continued for approximately 5–10 min. Third, water
(without plasticizer) was added gradually and mixed for another
7–10 min at least, in order the produced mixture to obtain uniform

Table 1
Reinforcement of the tested beams and concrete strength values.

Beam code name Cross-section Bars (£ in mm) q‘ (%) Stirrups (mm) qt (%) Volume of steel fibres – Vf (%) fc.cube (MPa) fct,split (MPa)
Group P: Beams without bars and stirrups (plain concrete and steel–fibre-reinforced-concrete beams)
RP0 R – – – – – 16.89 1.48
RP1 R – – – – 1 19.80 2.05
RP3 R – – – – 3 19.26 2.26
LsP0 Ls – – – – – 19.24 1.93
LsP1 Ls – – – – 1 21.00 2.25
LP0 L – – – – – 18.38 2.00
LP1 L – – – – 1 30.00 2.85
TsP0 Ts – – – – – 20.70 2.00
TsP1 Ts – – – – 1 25.00 2.36
TP0 T – – – – – 19.52 1.90
TP1 T – – – – 1 29.14 2.83
THsP0 THs – – – – – 30.96 2.17
THsP1 THs – – – – 1 27.26 2.03
THsP3 THs – – – – 3 33.04 3.24
THP0 TH – – – – – 39.56 2.97
THP1 TH – – – – 1 26.00 1.91
THP3 TH – – – – 3 29.51 4.36
Group L: Beams with longitudinal bars
RL0 RL 4£10 1.57 – – – 18.37 1.56
RL1 RL 4£10 1.57 – – 1 17.04 1.79
RL3 RL 4£10 1.57 – – 3 16.44 2.05
THsL0 THsL 6£10 1.62 – – – 30.96 2.52
THsL1 THsL 6£10 1.62 – – 1 24.96 2.15
THsL1 THsL 6£10 1.62 – – 3 26.67 2.88
THL0 THL 6£10 1.24 – – – 32.44 2.41
THL1 THL 6£10 1.24 – – 1 27.63 2.03
THL3 THL 6£10 1.24 – – 3 21.11 2.88
Group R: Beams with bars and stirrups
RR0 RR 4£10 1.57 £8/200 0.63 – 20.12 1.46
RR1 RR 4£10 1.57 £8/200 0.63 1 18.96 1.86
RR3 RR 4£10 1.57 £8/200 0.63 3 16.89 2.57
THsR0 THsR 6£10 1.62 £8/200 0.71 – 36.15 2.26
THsR1 THsR 6£10 1.62 £8/200 0.71 1 32.29 2.05
THsR1 THsR 6£10 1.62 £8/200 0.71 3 26.59 1.86
THR0 THR 6£10 1.24 £8/200 0.66 – 33.19 2.10
THR1 THR 6£10 1.24 £8/200 0.66 1 24.07 1.70
THR3 THR 6£10 1.24 £8/200 0.66 3 25.48 3.14
334 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341
material consistency, adequate workability and homogeneous fibre
distribution. Right afterwards, the prepared fresh fibrous concrete
mixture was carefully placed in the specimens’ moulds using wide
scoops and vibrated using frequency vibrating table. Sufficient
time of vibration was provided to guarantee suitable consolidation
and to prevent fibre protrusion from the finished surface. Extra
time of about 2 min was required for further vibration in fibrous
concrete mixtures with high volume fraction of steel fibres
(Vf = 3%). During mixing and casting of the fresh mixture, no steel
fibre gravitation was observed.
Supplementary compression and splitting tests in order to
determine the cube compression strength of the plain concrete
mixture and the tensile strength of the composite mixture (steel fi-
bre concrete) were also included in the program and results are gi-
ven in Table 1. These strength values are averages from three
standard 150 mm cubes and three standard 150  300 mm cylin-
ders, respectively.
2.2. Test setup
The experimental setup is a commonly used torsional test rig
and it is shown in Fig. 2. The total length of the specimens was
1.60 m. Beams were supported on two roller supports 1.30 m
apart. These supports ensured that the beam was free to twist
and to elongate longitudinally in both end directions during the
test. The load was applied through a diagonally placed steel sprea-
der beam on the ends of two properly configured over-reinforced
concrete arms at the end parts of the specimens, as shown in
Fig. 2. The end parts of the specimens were also heavily reinforced
with high volume of stirrups in order to bear without cracking the
imposed torsional loading at the ends of the 500 mm long concrete
arms. This way, the examined test region was the central part of
the specimens. During the test procedure, torsional helical diago-
nal cracking and, finally, failure were localized within this test re-
gion. The heavily reinforced end parts and the over-reinforced
concrete arms of the beams remained quite intact.
The load was imposed consistently in low rate and measured by
a load cell with accuracy equal to 0.05 kN. The average angle of
twist per unit length of the tested beams was estimated using
the measurements of a set of linear variable displacement trans-
ducers with accuracy 0.02 mm, which was placed at the over-
hanging ends of the specimens to measure the deflections of the
concrete arms as the beam twisted.
Fig. 2. Test setup.
In order to acquire information about the failure modes of the
tested beams, the strains of the steel reinforcement were measured
by electrical resistance strain gauges.
The beams were tested in monotonically increasing torque mo-
ment until the ultimate torsional strength and subsequently in
increasing twist until the total failure of the specimen or the max-
imum twist capacity of the test apparatus.
3. Test results and discussions
3.1. Torsional strength values and behavioural curves
The torsional behaviour of all the tested beams of groups P, L
and R is presented in Figs. 3–5, respectively, in terms of experimen-
tal curves for torsional moment, T, versus angle of twist per unit
length, #. In these figures, the measured torsional response of the
steel fibre concrete beams is compared with the response of the
corresponding control specimens (beams without fibres). It is
noted that all the torsional curves demonstrated in Fig. 3 are com-
plete until failure. Torsional behaviour of beams RL0, THsL0 and
THL0 displayed in Fig. 4 is also complete until failure, while the re-
sponse of the fibrous beams in Fig. 4 and the torsional curves of the
beams of group R in Fig. 5 are truncated.
Furthermore, the measured torsional moment at cracking, Tcr,
the angle of twist per unit length at cracking, #Tcr , the initial pre-
cracking torsional stiffness, K, the post-cracking ultimate torsional
moment, Tu, and the corresponding angle of twist per unit length,
#Tu , of the beams of groups P, L and R are presented in Tables 2–
4, respectively. It is mentioned that specimens without conven-
tional steel reinforcement (beams of group P) and plain concrete
specimens with longitudinal only reinforcement (beams RL0,
THsL0 and THL0 of group L) did not exhibit increasing post-crack-
ing response. Thus, their maximum torsional moment capacities
were equal to the torsional moment at cracking: Tmax = Tcr. For all
the other specimens: Tmax = Tu.
From the values of the elastic torsional stiffness, K (Tables 2–4)
and the slopes of elastic parts of the behavioural curves (Figs. 3–5)
it is deduced that the initial pre-cracking torsional rigidity re-
mained practically unaffected by the amount of steel fibres and
conventional reinforcement used. The values of the cracking tor-
sional moment, Tcr, of the steel fibre concrete beams are higher
than the corresponding values of the control specimens (without
fibres). This increase on the cracking torsional strength is essential
 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341 335

Fig. 3. Experimental curves of the beams of group P (without conventional steel reinforcement).

8 Without fibres
Torsional moment (kN.m)

7
Fibres 1%
6
Fibres 3%
5
4
3
2
1
0
0.00 0.05 0.10 0.00 0.05 0.10 0.00 0.05 0.10 0.15
Angle of twist per length (rad/m)

Fig. 4. Experimental curves of the beams of group L (with longitudinal bars).

8
Without fibres
Torsional moment (kN.m)

7
Fibres 1%
6
Fibres 3%
5
4
3
2
1
0
0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15 0.20
Angle of twist per length (rad/m)

Fig. 5. Experimental curves of the beams of group R (conventionally reinforced beams with longitudinal bars and stirrups).

for the beams without conventional steel reinforcement (speci-
mens of group P, see also Fig. 3 and Table 2). In these cases steel
fibres are the only steel reinforcement and seem capable to provide
enhanced torque capacities, especially in the cases of the beams
with high volume of fibres (Vf = 3%).
After concrete cracking, the favourable effect of steel fibres is
clearly indicated to the post-cracking response of the beams with
longitudinal only reinforcement (specimens of group L). Steel fibre
concrete beams demonstrated further increase on the torsional
strength, whereas the corresponding plain concrete beams (control
specimens) did not exhibit increasing post-cracking response
(Fig. 4 and Table 3). Furthermore, the values of the ultimate tor-
sional moment, Tu, of the conventionally reinforced fibre concrete
beams with bars and stirrups are considerably higher than the val-
ues of the corresponding control specimens (Fig. 5 and Table 4).
These improvements on the torque capacities were observed to
both rectangular and flanged beams. Nevertheless, Fig. 6 demon-
strates the percentage increase on the ultimate torsional moment
due to the use of fibres and indicates that this increase seems to
be reduced in the cases of the flanged beams with respect to the
rectangular beams.
Fig. 6 also shows that the addition of steel fibre volume frac-
tions 1% and 3% to the rectangular beams caused an increase of
the torsional strength that ranges from 18% to 64% and from 35%
to 86%, respectively. Further, the addition of steel fibres with
Vf = 1% and 3% to the T-flanged beams with bf /bw = 2 caused an
336 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341

Table 2
Test results of beams of group P (without longitudinal bars and stirrups).

Beam code name Vf (%) Tmax = Tcr (kN m) #Tmax = #Tcr (rad/m) K (kN m2/rad) #85Tmax (rad/m) lT85max = lT85cr
RP0 – 1.45 0.009 284 – (1.00)
RP1 1 1.84 0.019 324 0.023 1.21
RP3 3 1.95 0.014 341 0.028 2.00
LsP0 – 1.75 0.008 401 – (1.00)
LsP1 1 2.21 0.019 454 0.022 1.19
LP0 – 2.15 0.009 600 – (1.00)
LP1 1 2.65 0.008 610 0.010 1.25
TsP0 – 2.34 0.009 641 – (1.00)
TsP1 1 2.72 0.012 645 0.013 1.08
TP0 – 2.77 0.008 881 – (1.00)
TP1 1 2.96 0.013 891 0.016 1.23
THsP0 – 2.75 0.013 855 – (1.00)
THsP1 1 3.39 0.011 895 0.015 1.36
THsP3 3 3.69 0.021 870 0.037 1.76
THP0 – 3.72 0.018 1525 – (1.00)
THP1 1 4.05 0.015 1571 0.021 1.40
THP3 3 4.42 0.026 1537 0.047 1.81

lT = 1.00.

Table 3
Test results of beams of group L (with longitudinal bars – without stirrups: qt = 0).

Beam q‘ (%) Vf (%) Elastic response until first cracking Post-cracking Tmax #Tmax lT #85Tmax lT85max lT85cr
code (kN m) (rad/m) (rad/m)
Tcr #Tcr K Tu #Tu
name
(kN m) (rad/m) (kN m2/rad) (kN m) (rad/m)
RL0 1.57 – 1.47 0.015 300 – – 1.47 0.015 1.00 0.028 1.87 1.87
RL1 1.57 1 1.80 0.012 301 2.41 0.021 2.41 0.021 1.75 0.041 1.95 3.42
RL3 1.57 3 2.00 0.016 293 2.73 0.055 2.73 0.055 3.44 0.119 2.16 7.44
THsL0 1.62 – 3.18 0.015 850 – – 3.18 0.015 1.00 0.033 2.20 2.20
THsL1 1.62 1 3.57 0.016 855 4.26 0.066 4.26 0.066 4.13 0.162 2.45 10.13
THsL1 1.62 3 3.70 0.018 858 4.92 0.070 4.92 0.070 3.89 0.190 2.71 10.56
THL0 1.24 – 3.80 0.032 1535 – – 3.80 0.032 1.00 0.062 1.94 1.94
THL1 1.24 1 4.30 0.033 1536 4.56 0.068 4.56 0.068 2.06 0.138 2.03 4.18
THL3 1.24 3 4.20 0.041 1541 4.80 0.074 4.80 0.074 1.80 0.166 2.24 4.05

Table 4
Test results of beams of group R (with longitudinal bars and stirrups).

Beam code q‘ (%) qt (%) Vf (%) Elastic response until first cracking Post-cracking lT #85Tmax lT85max lT85cr
name (rad/m)
Tcr #Tcr K Tu = Tmax #Tu = #Tmax
(kN m) (rad/m) (kN m2/rad) (kN m) (rad/m)
RR0 1.57 0.63 – 1.60 0.015 305 2.31 0.073 4.87 0.096 1.32 6.40
RR1 1.57 0.63 1 2.00 0.016 329 2.73 0.088 5.50 0.120 1.36 7.50
RR3 1.57 0.63 3 2.21 0.014 350 3.15 0.101 7.21 0.200 1.98 14.29
THsR0 1.62 0.71 – 3.10 0.017 850 4.41 0.084 4.94 0.151 1.80 8.88
THsR1 1.62 0.71 1 3.28 0.021 836 4.93 0.114 5.43 0.250 2.19 11.90
THsR1 1.62 0.71 3 3.50 0.020 851 5.50 0.250 12.50 1.851 7.40 92.55
THR0 1.24 0.66 – 4.34 0.026 1581 5.60 0.100 3.85 0.153 1.53 5.88
THR1 1.24 0.66 1 4.40 0.030 1585 5.78 0.140 4.67 0.250 1.79 8.33
THR3 1.24 0.66 3 4.30 0.035 1578 7.41 0.156 4.46 1.716 11.00 49.03

increase of the torsional strength that ranges from 12% to 34% and
from 25% to 55%, respectively. Furthermore, for the case of the
T-flanged beams with bf /bw = 3 the addition of 1% and 3% steel
fibre volume fraction caused an increase of the torsional strength
that ranges from 3% to 20% and from 19% to 32%, respectively.
3.2. Cracking behaviour and failure modes
Control specimens (beams without steel fibres) exhibited typi-
cal torsional failures modes (see also photographs of Fig. 7): (i)
Beams without bars and stirrups showed an abrupt collapse by
crushing of concrete and separated in two parts (failure mode
‘‘A”). (ii) Beams with longitudinal bars only also failed in a similar
manner exhibiting brittle torsional failure; they develop an intense
single helical crack, but the bars prevented the splitting of the
specimens in two parts (failure mode ‘‘B”). (iii) Beams with con-
ventional reinforcement including bars and stirrups showed an
increasing post-cracking behaviour along with the formation of a
number of helical and diagonal cracks. In these cases as the torque
increased some of these cracks became excessively wide and
beams failed by transverse steel yielding before concrete crushing
(failure mode ‘‘C”).
The steel fibre beams exhibited different failure mechanisms
since fibres prevented crack growth and crack propagation in con-
crete. The main benefit of steel fibre reinforcement is to establish a
crack-control mechanism and to provide for a pseudo-ductility in
the post-cracking behaviour. The improvement of the first-crack
strength due to steel fibres is usually a secondary consequence
 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341
100
Torsional strength increase (%)
80
60
40 Vf = 3%
20
Vf = 1%
0
1 2
bf /bw
Group P beams
Fig. 6. Increase of the torsional strength of the tested beams due to the use of steel fibres.
which is related to the fibres quantity more than to the fibres effi-
ciency. For low volume fibre content this improvement in the first-
crack strength is of minor importance.
Fibrous beams without bars and stirrups displayed a typical
failure of steel fibre concrete under tension since steel fibre pull-
out was the main strength bearing mechanism. It is emphasized
that fibres prevented the sudden failure of these specimens by con-
trolling crack growth, especially in beams with high volume of fi-
bres (3%) (failure mode ‘‘D”). Fibrous concrete beams with steel
bars without stirrups developed dense diagonal and helical cracks
and failed due to steel fibres debonding (failure mode ‘‘E”). Finally,
conventionally reinforced beams with bars and stirrups containing
steel fibres exhibited a highly ductile post-cracking behaviour,
while the developed cracks were narrower and more closely
spaced with a higher volume fraction of fibres used (failure mode
‘‘F”). Another characteristic of the failure modes ‘‘E” and ‘‘F” is that
increasing the volume of fibres the number of cracks also increases,
yet resulting in a significantly reduced crack width. Cracking pat-
terns of typical tested beams at the end of their loading due to fail-
ure or approximately at the 85% of the maximum torsional
moment (assumed as the end of the reliable response range) are
illustrated in the photographs of Fig. 7.
3.3. Contribution of steel fibres to the post-elastic response
Test results indicated that the increase of the steel fibre fraction
results in a significant improvement of the post-elastic behaviour
of the beams, especially for the response branch after the maxi-
mum torsional moment. For a thorough understanding and evalu-
ation of the fibres’ contribution on the response and for
comparison reasons of test results in terms of ductility, the follow-
ing indices are introduced:
#Tmax
lT ¼
#Tcr
#
lT85max ¼ 85Tmax
#Tmax
#85Tmax
lT85cr ¼
#Tcr
where #Tmax : the rotation at the maximum torsional moment;
#85Tmax : the rotation at the point of 85% of the maximum torsional
moment, assumed as the end of the reliable response range.
Values of the above mentioned indices for the examined beams
are presented in Tables 2–4, for the beams of groups P, L and R,
337
Vf = 3%
Vf = 3%
Vf = 1%
Vf = 1%
3 1 2 3 1 2 3
bf /bw bf /bw
Group L beams Group R beams
respectively. From these values it is obvious that the primary
advantage of the use of steel fibres is the significant improvement
of the post-cracking torsional response, increasing the torque rota-
tion capabilities of the conventionally reinforced beams without
considerable loss of their strength.
3.4. Steel fibres as the only shear torsional reinforcement
The use of non-conventional fibrous reinforcement in torsional
beams as the only shear reinforcement instead of common stirrups
is an interesting alternative that has been proposed for special
cases such as in strengthening works [25]. Recent experimental
works also indicated that the addition of steel fibres in reinforced
concrete beams under shear allows the reduction, and in special
cases the total replacement, of the required stirrups [4–6]. This
reduction of stirrups is suggested in cases where design criteria
recommend high steel ratio of shear reinforcement, which leads
to small stirrup spacing and, therefore, the use of steel fibres is
more convenient.
The possibility of the use of steel fibres as the only shear rein-
forcement in torsional beams is also discussed in this experimental
research. For this reason, comparisons of the torsional behavioural
curves of (i) the non-fibrous reinforced concrete beams with bars
and stirrups with (ii) the fibrous beams with longitudinal rein-
forcement and steel fibres with volume fraction 1% and 3% are pre-
sented in Fig. 8a, b and c.
Despite the rather limited experimental data curves shown in
Fig. 8 (nine tested beams) these comparisons indicate a potential
for using steel fibres as a replacement for stirrups in torsional
beams. The experimental curves of Fig. 8 show that this alternative
could be achieved under specific circumstances with adequate
quantities of steel fibres and this replacement seems to be more
efficient in rectangular beams than in flanged ones.
ð2Þ It is also mentioned that in order to completely substitute the
conventional shear reinforcement with steel fibres, it is necessary
ð3Þ the fibrous concrete beams to obtain at least equal performances
in terms of strength and ductility with the corresponding conven-
ð4Þ tionally reinforced beams. Regarding the strength, it is possible to
provide the same value if adequate percentage of fibres is added.
However, the ductility resources differ in the case of use of fibres
comparing to the ones of conventional steel reinforcement, be-
cause in the first case it is related to the gradual pull-out of fibres,
whereas in the second it is related to the plastic deformation of the
steel reinforcement. For this reason, the complete substitution is
338 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341

Fig. 7. Failure modes of typical tested specimens.

not efficacious when high amplitude of the shear crack width
occurs.
4. Analytical results
4.1. Fibrous beams without conventional reinforcement
The classical elastic Saint Venant’s theory to the torsion prob-
lem, although it properly describes the elastic behaviour, fails to
predict the ultimate torsional strength of concrete elements. Ana-
lytical solutions of the governing equation deduced by this ap-
proach are based on the assumption that brittle failure occurs
when the maximum developing tensile stress reaches the material
maximum tensile strength. By ignoring the post-cracking tension
softening phenomenon of the material, they consistently underesti-
mate the ultimate torsional strength of the element. For plain con-
crete members the ultimate torque has been experimentally found
to be roughly up to 50% greater than predicted ones [7,26]. Further-
 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341 339

Fig. 8. Investigation of the use of steel fibres as the only shear torsional reinforcement instead of stirrups.

2
more, in the case of steel fibre concrete elements the post-cracking b h
strength of the composite material represents an even more impor- — Mansur & Paramasivam ½10 : T u;sb ¼ ð0:71f rf Þ ð6Þ
3
tant part of the element strength and energy absorption capacity. 2
b h
Hsu has proposed the skew bending theory for the prediction of — Wafa et al: ½11 : T u;sb ¼ ð0:71f ctf ;split Þ ð7Þ
3
the torsional strength of plain concrete beams [7]. Other subse-
quent researchers have proposed modified versions of this theory where fcf0 , frf, and fctf,split are the compression, the rupture and the
in order to predict the torsional strength of steel fibre concrete splitting strength of the composite mixture (steel fibre concrete),
members [9–11]. Relations yielded by these efforts are presented respectively.
and used for the prediction of the torsional strength of beams Recently, a new efficient numerical algorithm that employs a
tested in the present work: smeared crack approach for the analysis of plain concrete elements
! in torsion has been proposed by Karayannis [27]. This analytical
b
2 qffiffiffiffiffi
3 model is initially based on Saint Venant’s theory and utilizes a spe-
— Craig et al: ½9 : T u;sb ¼ 140:23 þ 10 h fcf0 ð5aÞ
645:16 cial numerical technique that can be regarded as a combination of
2
b h the numerical techniques of finite elements and finite differences
T u;sb ¼ ð0:85f rf Þ ð5bÞ
3 [8]. This scheme can easily be applied to elements with practically
340 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341

Table 5
Comparisons between experimental and analytical results for beams of group P.
ð5aÞ T exp ð5bÞ T exp ð6Þ T exp ð7Þ T exp T exp
Beam code name Texp (kN m) T cal (kN m) T cal (kN m) T cal (kN m) T cal (kN m) T mod
cal (kN m)
ð5aÞ
T cal
ð5bÞ
T cal
ð6Þ
T cal
ð7Þ
T cal T mod
cal

RP1 1.84 1.79 1.03 2.32 0.79 1.94 0.95 1.37 1.34 1.65 1.12
RP3 1.95 1.85 1.05 2.56 0.76 2.14 0.91 1.51 1.29 1.94 1.01
LsP1 2.21 2.24 0.99 2.71 0.82 2.26 0.98 1.59 1.39 2.25 0.98
LP1 2.65 2.81 0.94 3.63 0.73 3.04 0.87 2.14 1.24 2.74 0.97
TsP1 2.72 2.65 1.03 3.01 0.90 2.51 1.08 1.77 1.54 2.91 0.93
TP1 2.96 3.37 0.88 4.01 0.74 3.35 0.88 2.36 1.25 3.28 0.90
THsP1 3.39 2.85 1.19 3.23 1.05 2.70 1.26 1.90 1.78 3.19 1.06
THsP3 3.69 3.11 1.19 5.16 0.72 4.31 0.86 3.04 1.21 3.77 0.98
THP1 4.05 3.66 1.11 3.92 1.03 3.27 1.24 2.31 1.75 4.00 1.01
THP3 4.42 3.93 1.12 8.94 0.49 7.47 0.59 5.26 0.84 4.53 0.98
1.05 0.80 0.96 1.36 0.99
0.102 0.163 0.195 0.277 0.061

any cross-section because it includes numerical mapping and it has
no need of extended computing power since it does not make use
of any stiffness matrix calculations. The smeared crack model uses
the behaviour of a crack process zone which is described by consti-
tutive relations expressed in terms of normal stress and crack
width. The total strain is considered as the sum of elastic and frac-
ture strain components. For the determination of the fracture
strain the crack process zone width is used, which is attributed
to the material, the type of loading and the size of the beam
[26,27].
This method has been extended and modified [20,21] in order
to incorporate a stress–strain model with a non-linear post-crack-
ing branch for the fibrous material behaviour in tension to make
this method applicable to the torsion problem of steel fibre con-
crete beams. This approach takes into account the volume fraction
of fibres, the tensile strength of concrete and parameters
concerning the geometry, the orientation and the slip mechanism
of the fibres. Detailed derivation of the equations and the solution
technique of this analytical model can be found in Refs. [20], [21]
and [27]. Extensive comparisons between analytical results yielded
by these analyses and experimental data derived from a broad
range of parametrical studies established the validity of these ana-
lytical models for plain and steel fibre concrete beams subjected to
pure torsion or combined torsion with shear force, bending mo-
ment and axial force [21,26,28].
Based on the above mentioned analytical relationships and the
analytical model of the author, predictions of the torsional moment
capacities for the tested beams of group P are presented and com-
pared with the experimental ones in Table 5. In this Table, the cal-
ð5aÞ ð5bÞ ð6Þ ð7Þ
culated values of T cal ; T cal ; T cal ; T cal and Tmod represent the
analytically predicted ultimate torsional moments of the steel fibre
beams derived from Eqs. (5a), (5b), (6), (7), and the analytical mod-
el, respectively. The mean values and the standard deviation values
of the ratios of the experimental to the calculated torsional
strengths are also summarized and compared in Table 5.
4.2. Rectangular fibrous beams with conventional reinforcement
of the torque moment at a lower rate, depending on the volume of
the transverse reinforcement. The consistently decreasing tor-
sional stiffness reveals the different nature of the mechanism of
resistance. It is noted that in conventionally reinforced concrete
beams with bars and stirrups (without fibres), the transition from
the one region to the other is usually abrupt and it could be char-
acterized by the lack of a bearing mechanism soon after cracking
and before the activation of the reinforcements and the concrete
struts [7,15,17,29]. This is more or less justified by the formation
of an almost horizontal plateau in the experimental curves of T ver-
sus # between these two regions (see for example the experimental
curves of the beams RR0, THsR0 and THR0 in Figs. 5 and 8). How-
ever, in fibrous concrete beams with bars and stirrups, the transi-
tion from the pre-cracking to the post-cracking region seems to be
gradual because of the steel fibres which provide crack-bridging
and significant resistance to shear across developing cracks (see
also the T versus # curves of the beams with steel fibres in Fig. 5).
Based on the this observation and aiming to the description of
the entire torsional behaviour of reinforced concrete elements,
the combination of two different theoretical models has already
been adopted for reinforced concrete beams by Chalioris [29],
[30] and for fibrous reinforced concrete beams by the authors
[17] and by a recent study of Gunneswara Rao and Rama Seshu
[18]. In these combined methodologies, the elastic response until
the first cracking is described by specially modified approaches
of the classical elastic theory, whereas the post-cracking part is de-
scribed by specially extended versions of the softened truss model
developed by Hsu & Mo [31]. In these modified truss models the
contribution of steel fibres has been incorporated as an additional
reinforcement component that contributes to the torsional resis-
tance along with the conventional steel reinforcement. It is also
noted that these combined methods can provide full analytical tor-
sional curves for the behaviour of rectangular fibrous beams with
conventional reinforcement.
The above described combined methods were used in this study
in order to analytically predict the values of the torsional moment
at cracking, the angle of twist per unit length at cracking, the initial

The typical torque moment versus angle of twist experimental Table 6

curve of a steel fibre concrete beam with longitudinal bars and stir- Comparisons between experimental and analytical results for the rectangular beams
rups (see for example the experimental curve of beam RR3 in of group R.
Fig. 5) comprises two distinct regions: an elastic region up until Beam code Elastic response until Post-cracking response
first cracking and a post-cracking branch. The different character name first cracking
of the response in these regions reveals the different nature of T exp
cr
#exp
Tcr K exp T exp
u
#exp
Tu
T cr #cal K cal T cal #cal
the load resisting mechanism in each case. The pre-cracking part (kN m)
Tcr
(rad/m)
2
(kN m /rad) (kN m)
u Tu
(rad/m)
is characterized by the high value of the torsional stiffness RR0 1:60
¼ 1:13 0:015
¼ 1:15 305
¼ 1:02 2:31
¼ 1:02 0:073
¼ 0:82
1:41 0:013 298 2:27 0:089
and the element behaves as a homogeneous concrete element 2:00 0:016 329 2:73 0:088
RR1 1:65 ¼ 1:21 0:014 ¼ 1:14 332 ¼ 0:99 2:74 ¼ 1:00 0:085 ¼ 1:04
since the influence of the steel reinforcement is of minor impor- 2:21 0:014 350 3:15 0:101
RR3 1:93 ¼ 1:15 0:014 ¼ 1:00 340 ¼ 1:03 3:14 ¼ 1:00 0:087 ¼ 1:16
tance. The post-cracking part is characterized by a further increase
 C.E. Chalioris, C.G. Karayannis / Cement & Concrete Composites 31 (2009) 331–341
pre-cracking torsional stiffness, the post-cracking ultimate tor-
sional moment and the corresponding angle of twist per unit
length for the rectangular beams of group R (RR0, RR1 and RR3).
These analytical values are presented and compared with the
experimental ones in Table 6.
5. Concluding remarks
The experimental program of this study includes torsional tests
of 35 rectangular and flanged beams with various steel fibre and
conventional steel reinforcement arrangements. The main objec-
tive of this study is to investigate the effectiveness of the use of
short steel fibres on the torsional response of concrete beams with
rectangular, L- and T-shaped cross-section. Based on the test re-
sults presented herein, the following concluding remarks are
drawn:
Fibrous concrete beams exhibited improved overall torsional
performance with respect to the corresponding non-fibrous control
beams. The main contribution of steel fibres on the torsional
behaviour is mainly observed after concrete cracking.
The addition of steel fibres was essential to the beams without
conventional steel reinforcement since fibres were the only rein-
forcement and proved capable to provide enhanced torsional mo-
ment capacities, especially in the beams with high volume of
fibres (Vf = 3%). The presence of fibre reinforcement even at low
percentage was very effective in preventing the sudden brittle fail-
ure of both rectangular and non-rectangular beams by implying
crack-control mechanism.
After concrete cracking, the favourable effect of steel fibres was
clearly demonstrated to the post-cracking response of the beams
with longitudinal only reinforcement. Fibre concrete beams
displayed further increase on the torsional strength, whereas the
corresponding control specimens did not exhibit increasing post-
cracking response.
Increasing the steel fibre volume fraction a significant improve-
ment of the post-elastic response of the beams is achieved, espe-
cially after the maximum torque moment. However, these
improvements on the torque capacities seemed to be reduced in
the cases of the flanged beams with respect to the rectangular
ones.
The use of steel fibres as the only shear torsional reinforcement
in beams without stirrups has also been investigated in this study.
Experimental torque curves indicated that short steel fibres is a
reliable way to increase the post-elastic twist capabilities of the
beams. Furthermore, comparisons between the experimental re-
sults showed that the use of high fractions of steel fibres (Vf = 3%)
in torsional beams with longitudinal bars (without stirrups) pro-
vided increased strength and ductility capabilities. This improve-
ment due to the addition of steel fibres seemed to be more
effective in the beams with rectangular cross-section than in the
corresponding flanged beams.
The analytical results of existing numerical approaches for pre-
dicting the torsional strength of steel fibre reinforced concrete
beams have been presented and compared to the observed data
to estimate the effectiveness of these models. It is emphasized that
the prediction of the entire torsional behaviour of the rectangular
fibrous concrete beams with longitudinal bars and stirrups has
been achieved using analytical methods that combine (a) a modi-
fied approach of the classical elastic theory for the elastic response
341
until the first cracking and (b) an extended version of the softened
truss model for the post-cracking part. Analytical results yielded
from this combined methodology proved to be in good compliance
with the experimental ones.
References
[1] ACI Committee 544. State-of-the-art on fiber reinforced concrete. ACI manual
of concrete practice. Farmington Hills (MI): American Concrete Institute; 1998.
[2] ACI Committee 544. Design considerations for steel fiber reinforced concrete.
ACI manual of concrete practice. Farmington Hills (MI): American Concrete
Institute; 1999.
[3] Nanni A. Fatigue behavior of steel fiber reinforced concrete. Cement Concrete
Comp 1991;13:239–45.
[4] Lim DH, Oh BH. Experimental and theoretical investigation on the shear of
steel fibre reinforced concrete beams. Eng Struct 1999;21:937–44.
[5] Cucchiara C, Mendola L, Papia M. Effectiveness of stirrups and steel fibres as
shear reinforcement. Cement Concrete Comp 2004;26:777–86.
[6] Juarez C, Valdez P, Duran A, Sobolev K. The diagonal tension behavior of fiber
reinforced concrete beams. Cement Concrete Comp 2007;29:402–8.
[7] Hsu TTC. Torsion of reinforced concrete. New York: Van Nostrand Reinhold;
1984.
[8] Karayannis CG. Torsional analysis of flanged concrete elements with tension
softening. Comput Struct 1995;54(1):97–110.
[9] Craig RJ, Parr JA, Germain E, Mosquera V, Kamilares S. Fiber-reinforced beams
in torsion. J ACI 1986;83(6):934–42.
[10] Mansur MA, Paramasivam P. Fiber-reinforced concrete beams in torsion,
bending and shear. J ACI 1985;82(2):33–9.
[11] Wafa FF, Hasnat A, Tarabolsi OF. Prestressed fiber-reinforced concrete beams
subjected to torsion. ACI Struct J 1992;89(3):272–83.
[12] Gunneswara Rao TD, Rama Seshu D. Torsion of steel fiber reinforced concrete
members. Cem Concr Res 2003;33:1783–8.
[13] Mansur MA, Lim TY. Torsional behaviour of reinforced fibre concrete beams.
Cem Compos Ligh Concr 1985;7(4):261–7.
[14] Narayanan R, Kareem-Palanjian AS. Torsion in beams reinforced with bars and
fibers. J Struct Eng ASCE 1986;112(1):53–66.
[15] Mansur MA, Nagataki S, Lee SH, Oosumimoto Y. Torsional response of
reinforced fibrous concrete beams. ACI Struct J 1989;86(1):36–44.
[16] El-Niema EI. Fiber reinforced concrete beams under torsion. ACI Struct J
1993;90(5):489–95.
[17] Karayannis CG, Chalioris CE. Torsional behaviour of reinforced fibrous concrete
elements. Technika Chronika I, Sci J Tech Chamb Greece 1996;16(1–2):53–67
[in Greek].
[18] Gunneswara Rao TD, Rama Seshu D. Analytical model for the torsional
response of steel fiber reinforced concrete members under pure torsion.
Cement Concrete Comp 2005;27:493–501.
[19] Tegos IA. Fiber reinforced concrete beams with circular section in torsion. ACI
Struct J 1989;86(4):473–82.
[20] Karayannis CG. A numerical approach to steel–fibre reinforced concrete under
torsion. Struct Eng Rev 1995;7(2):83–91.
[21] Karayannis CG. Nonlinear analysis and tests of steel–fiber concrete beams in
torsion. Struct Eng Mech 2000;9(4):323–38.
[22] Nanni A. Design for torsion using steel fiber reinforced concrete. ACI Mater J
1990;87(6):556–64.
[23] Karayannis CG, Chalioris CE. Influence of steel–fibers on the capacity of flanged
beams in torsion. Technika Chronika I, Sci J Tech Chamb Greece
2000;20(2):111–22 [in Greek].
[24] RILEM TC-162-TDF: Test and design methods for steel fibre reinforced
concrete. Mater Struct 2001;34(1):3–6.
[25] Chalioris CE. Torsional strengthening of rectangular and flanged beams using
carbon fibre-reinforced-polymers – experimental study. Constr Build Mater
2008;22(1):21–9.
[26] Karayannis CG, Chalioris CE. Experimental validation of smeared analysis for
plain concrete in torsion. J Struct Eng ASCE 2000;126(6):646–53.
[27] Karayannis CG. Smeared crack analysis for plain concrete in torsion. J Struct
Eng ASCE 2000;126(6):638–45.
[28] Karayannis CG, Chalioris CE. Strength of prestressed concrete beams in torsion.
Struct Eng Mech 2000;10(2):165–80.
[29] Chalioris CE. Experimental study of the torsion of reinforced concrete
members. Struct Eng Mech 2006;23(6):713–37.
[30] Chalioris CE. Analytical model for the torsional behaviour of reinforced
concrete beams retrofitted with FRP materials. Eng Struct
2007;29(12):3263–76.
[31] Hsu TTC, Mo YL. Softening of concrete in torsional members – theory and tests.
J ACI 1985;82(3):290–303.
 Cement & Concrete Composites 31 (2009) 342–346

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

Characterization of historical mortars from Santa Catarina (Brazil)

P.J.P. Gleize *, E.V. Motta, D.A. Silva, H.R. Roman
Universidade Federal de Santa Catarina, Depto de Engenharia Civil, Cx. P 476, 88040-900 Florianópolis-SC, Brazil

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

Article history: The aim of this research is to characterize the mortars (binder and aggregate) of nine historical buildings,
Received 13 June 2008 built between 1750 and 1922, in the State of Santa Catarina, Brazil, using X-ray diffraction (XRD), ther-
Received in revised form 13 February 2009 mogravimetry (TG), Fourier transform infra-red spectrometry (FT-IR) and atomic absorption spectrome-
Accepted 20 February 2009
try (AAS). The compositions and granulometric distributions of the aggregates were determined by
Available online 28 February 2009
sedimentology, and the aggregate content was estimated by hot hydrochloric acid attack.
The results showed that the principal binder is hydrated lime, obtained from the burning of seashells,
Keywords:
in some cases mixed with hydraulic materials (clay, ground ceramic tile or brick, and hydraulic lime). The
Historical mortar
Mineralogical characterization
results of the hot HCl attack revealed three average binder/aggregate ratios.
Lime Ó 2009 Elsevier Ltd. All rights reserved.
Binder
Binder/aggregate ratio

1. Introduction
The rehabilitation of historical buildings is very important in
terms of the history and culture of cities. Inadequate intervention
in historical buildings and the use of Portland cement-based mor-
tars to restore historic renderings can cause more damage than
benefit through negative interactions between the cement and
the original materials (bricks, stones or binders) [1]. The damage
caused is due to the fact that, if compared to historical mortars,
Portland cement-based mortars (i) are less permeable, retained
an excess water which can generate alteration phenomena; (ii)
have a high content of soluble salts triggering intense corrosion
of the original materials through salt decay; (iii) are less able to
accommodate the masonry structure movements (thermal and
humidity effects, creep) due to its higher hardness and rigidity
[2–5]. Therefore, firstly, a historic-scientific study of the original
material needs to be carried out before any intervention in histor-
ical buildings, and, secondly, the data obtained must be used either
to produce a similar material or to formulate a restoration mortar
from modern materials without significant negative interactions
with the pre-existing materials.
Historical mortars are complex systems, containing aerial or
hydraulic binders or a blend of them, aggregates (not always crys-
talline) and admixtures that interact with the binder [6]. The use of
techniques for microstructural characterization of materials, e.g.
optical and electron microscopy, X-ray diffractometry, thermal
analysis, infra-red spectroscopy, chemical analysis, etc., allows
for the determination of the composition and of some characteris-
* Corresponding author. Tel.: +55 48 3315176; fax: +55 48 3315191.
E-mail address: ecv1phg@ecv.ufsc.br (P.J.P. Gleize).
tics of the mortars. However, each technique has its own limits and
in many cases several characterization techniques must be used to
obtain coherent and reliable results [7]. Such modern techniques of
investigation have been used to characterize Roman and medieval
mortars [3,4,8–13].
The characterization of the binder allows the classification of
the mortars as typical lime mortars, crushed brick-lime mortars,
rubble masonry mortars, cementitious mortars and mortars with
gypsum [3,11]. However, these studies did not investigate the ori-
gin of the lime, which could be mussel shells, natural limestone or
other sources.
The first binder used in Brazil was calcareous, obtained from the
seashell deposits that coated the sea-bed in Baía de Todos os San-
tos (All Saints Bay) in the 16th century, in the city of Salvador [14].
Locals have reported that mixtures of ‘seashell lime’, sand and
whale or fish oil were used for the production of mortars used in
the construction of many historical buildings in Brazil. In Flori-
anópolis, Santa Catarina State, Caieira beach was historically
known as the place where quicklime was produced from seashells,
extracted from extensive millenarian deposits, that were burnt in
holes in the ground. However, this artisanal process discontinued
with the beginning of industrial limestone mining (around 1940).
Thus, it seems that the main historical mortars in Brazil were com-
posed of sand, whale or fish oil, and lime obtained from the burn-
ing of oyster shells and/or coral blocks in artisanal kilns [15].
The main purpose of this research is to identify and characterize
the binders and aggregates of mortars used in nine historical build-
ings built in Brazil between 1750 and 1922 in the cities of Flori-
anópolis, Laguna and São Francisco do Sul, all of them in the
State of Santa Catarina, southern Brazil. The final goal is the formu-
lation of restoration mortars.

0958-9465/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconcomp.2009.02.013
 P.J.P. Gleize et al. / Cement & Concrete Composites 31 (2009) 342–346 343

2. Materials and methods Table 2

XRD binder characterization.

Nine rendering mortars extracted from historical constructions Calcite Quartz Calcium Kaolinite– Kaolinite Gehlenite Anortite
were studied. The function, age and provenance of the mortar sam- silicium smectite
hydrate
ples are shown in Table 1.
The samples were removed from the rendering mortar of ma- LAG1 ++++ ++ ++ ++ +++ + t
LAG2 ++++ + + + ++ – t
sonry walls at an approximate height of 130 cm from regions
LAG3 ++++ ++ ++ + ++ t +
apparently not damaged. Three 150 g samples were extracted from SFS1 ++++ + ++ t + t –
each of the nine buildings with a chisel. Their external part was SFS2 ++++ + ++ – – – t
scraped off and discarded in order to ensure the collection of SFS3 ++++ + t – + t –
non-altered material [16]. FLN1 ++++ + + – – – –
FLN2 ++++ + t – – – –
For the characterization of the binder and of the aggregate, the
FLN3 ++++ + – – – – –
sample was gently ground in a porcelain mortar. The fraction pass-
ing through a sieve number 200 (US Standard Sieve Openings – Peak intensity: ++++: very strong, +++: strong, ++: medium, +: weak, t: traces, –: not
found.
0.074 mm) was considered as the binder [17].
Approximately 150 mg of the fine binder powder was used for
the thermal analysis; for the other tests (500 mg for XRD, 10 mg
for FT-IR, 5 g for AAS and 5 g for the gravimetric analysis) the pow-
der was dried at a temperature not higher than 70 °C to avoid the samples the presence of clay minerals, such as kaolinite (basal
elimination of organic compounds. interplanar distance dhkl = 7.17, 1.49, 3.58 Å) and kaolinite-smec-
X-ray diffractometry (XRD) was performed with a Philips pow- tite – Al–Si–O–OH–H2O (dhkl = 7.24, 4.31, 3.55 Å), could be de-
der diffractometer to determine the nature of the crystalline com- tected, as well as traces of anortite (dhkl = 3.20, 3.18, 4.04 Å) and
ponents. The type and content of lime and the presence of gehlenite (dhkl = 2.86, 1.93, 1.82 Å), and a small amount of calcium
hydraulic binders were determined by thermogravimetric analysis silicate hydrate (dhkl = 3.27, 2.42, 4.22 Å) which indicates the pres-
(TGA) (Netzsch STA 409EP). TGA has been used to classify mortars ence of hydraulic compounds.
as typical lime mortars, crushed brick-lime mortars, rubble ma- Table 3 shows the results of the TG analysis. According to the
sonry mortars, cementitious mortars and mortars with gypsum classification proposed by Moropoulou et al. [6] and Bakolas et
[3,11]. The temperature in a nitrogen atmosphere was raised from al. [11], the weight loss up to 120 °C is due to adsorbed water
ambient temperature to 1000 °C at a constant heating rate of 10 °C/ which is characteristic of the presence of hydraulic binders that
min. Fourier transform infra-red (FT-IR) spectrometry was used to are hygroscopic. The weight loss in the temperature range 120–
detect the presence of salts and organic compounds and to confirm 200 °C is attributed to the crystallization water of hydrated salts
the results obtained from the thermal analysis and the XRD. Atom- that may originate from the sulfatation of carbonates, accumula-
ic absorption spectrometry (AAS) was used to identify and quantify tion of salts in the masonry, or the presence of gypsum in the ori-
oxides, and gravimetric analysis was used to quantify the silica. ginal binder. The weight loss between 200 and 600 °C is attributed
The aggregate obtained after the gentle grinding and sieving of to the loss of the chemically bound water of hydraulic compounds
the mortars was washed with water to eliminate possible salts, and (C–S–A–H and/or C–S–H) when techniques such as XRD, FT-IR do
dried at 70 °C. The sedimentological analysis provided information not detect other compounds that decompose in the same temper-
on the maximum, minimum and characteristic grain diameters, the ature range, such as calcium hydroxide, magnesium hydroxide,
grain size distribution, the probable origin of the aggregate (dune, hydromagnesite, organic compounds, etc. The weight loss above
river or sea sand), and the mineral composition, from 30 g samples. 600 °C is due to the decomposition of carbonates. Furthermore,
The binder/aggregate ratio was determined by the dissolution the CO2/H2O ratio, i.e., the ratio of the weight loss due to the
of the mortar samples through hot hydrochloric acid attack, as de- decomposition of carbonates (>600 °C, in%) to that attributed to
scribed in [16]. the chemically bound water of hydraulic compounds (200–
600 °C, in%), can provide important information regarding the
3. Results and discussion hydraulic nature of the binder [4,6,11,18]. It is well known that
as the hydraulic character of the binder increases, the weight loss
3.1. Binder identification and characterization between 200 and 600 °C increases, the weight loss above 600 °C
decreases, and the CO2/H2O ratio decreases [3,4,11,18–20]. In
The XRD results (Table 2) indicated that calcite is the main com- Fig. 1, three distinctive groups of binder were identified, whose
ponent of the binder, and that quartz is present in the smaller characteristics are showed in Table 4 [18], leading to the following
aggregate grains which passed through the 74 lm sieve. In some classification of the mortars:

Table 1
Mortar description.

Mortar Function Date of construction Provenancea

FLN1 External rendering mortar Nearly 1750 Vicar’s house of the Lagoa da Conceição Church – Florianópolis
FLN2 Internal rendering mortar Nearly 1750 Nossa Senhora das Necessidades Church of Santo Antônio de Lisboa – Florianópolis
FLN3 Internal rendering mortar 1806 Nossa Senhora da Lapa do Ribeirão Church – Florianópolis
LAG1 Internal rendering mortar 1850–1900 Mansion – Laguna
LAG2 Internal rendering mortar 1800–1850 Mansion – Laguna
LAG3 External rendering mortar Around 1800 Historical hotel – Laguna
SFS1 Internal rendering mortar Around 1900 Mansion – São Francisco do Sul
SFS2 External rendering mortar 1922 Mansion – São Francisco do Sul
SFS3 External rendering mortar Around 1900 Terminal Marítimo – São Francisco do Sul
a
The environmental conditions are the same for all mortars (coastal and sub-tropical) with an average RH and temperature around 80% and 21 °C, respectively, over the
year.
344 P.J.P. Gleize et al. / Cement & Concrete Composites 31 (2009) 342–346

Table 3
TGA results for the binder according to the classification proposed by Bakolas,
Moropoulou, and co-workers [6,11].

T < 120 °Ca 120–200 °Ca 200–600 °Ca T > 600°Ca CO2/H2O
LAG1 1.53 0.23 7.04 19.34 2.75
LAG2 1.39 0.74 8.26 18.9 2.29
LAG3 1.12 0.00 8.30 12.03 1.45
SFS1 1.14 0.31 5.19 25.88 4.98
SFS2 1.42 1.42 5.35 20.35 3.80
SFS3 1.15 0.83 4.47 30.34 6.78
FLN1 1.26 0.68 4.29 31.86 7.43
FLN2 1.09 0.61 4.44 34.93 7.87
FLN3 0.62 0.60 3.85 39.39 10.23
a
% Weight loss.

Fig. 2. Typical TGA curves of the three groups of mortar binders: lime, hydraulic
lime and crushed brick-lime.

Fig. 2 shows typical TGA curves of the three groups of binders.

The binder in lime mortars is a typical hydrated or air-lime mortar
which has become totally carbonated over time. The slightly
hydraulic character of the binding material in hydraulic lime mor-
tars is attributed to the presence of impurities (silica) in the raw
carbonate material or to the accidental introduction of ‘clayed
materials’ during the lime production, since it is known that histor-
ically ovens were covered with clay before burning. The highly
hydraulic character of the binder of crushed brick-lime mortars is
due to the use of marly lime, as mentioned above, and also to lime/
ceramic dust interactions [4,11,18]. The mixture of burnt clayed
material (e.g. crushed brick) to lime was used in ancient times by
the Phoenicians (10th century BC) as a way to improve the bearing
capacity and durability in a marine atmosphere [1,20,21]. The
hydraulicity results from interactions between the calcium
hydroxide and silicates, silico-aluminates and aluminate present
in the crushed ceramic material which is not well burnt, leading
to the formation of cementitious compounds such as C–S–H, C–
S–A–H and C–A–H [1,3,4,11,22–24].
The infra-red analysis (FT-IR) confirmed the TGA and XRD re-
sults, as follows:

(i) All binding materials show the absorption bands of carbon-

Fig. 1. (a) Percent CO2/H2O ratio vs. CO2. (b) Percent hydraulic water vs. CO2/H2O
ratio.
(a) Typical lime mortars (FLN3) with hydraulic water <4% and
CO2/H2O > 10.
(b) Hydraulic lime mortars (FLN1, FLN2, SFS1, SFS2 and SFS3)
with intermediate hydraulic character with hydraulic water
between 4% and 6% and CO2/H2O between 4 and 9.
(c) Crushed brick-lime mortars (LAG1, LAG2 and LAG3) with
higher hydraulic character than the hydraulic lime mortars
with hydraulic water between 6% and 10% and CO2/H2O < 4.
ates (2870, 2515, 1798, 1427, 876, 709 cm1) [25–27]: the
stronger the hydraulic character of the binder the less
intense the bands.
(ii) Mortars with hydraulic binders show absorption bands due
to silicates and alumino-silicates (1110, 1033, 790, 756,
and 539 cm1) [25] which are more intense for more
hydraulic binders.
(iii) Organic compounds and hydrate salts were not detected.
Atomic absorption spectrometry – AAS allows the quantifica-
tion of the oxides in the binder, as well as the estimation of the bin-
der hydraulicity through the cementation index – CI [Eq. (1)]. The
higher the CI, the more hydraulic the binder [2]. As expected, the
AAS results (Table 5) showed that as the hydraulic character of
the binder increases the CI index increases.

Table 4
Binder classification (adapted from [18]).

Mortar type T < 120 °C 200–600 °C – H2O T > 600 °C – CO2 CO2/H2O Sample
(hygroscopic water) (hydraulic water) (decomposition of carbonates)
Lime mortars <1 <4 >30 >10 FLN3
Hydraulic lime mortars >1 4–6 20–34 4–9 FLN1, FLN2, SFS1, SFS2*, SFS3
Crushed brick-lime mortars >1 6–10 12–20 <4 LAG1, LAG2, LAG3
*
SFS2 shows characteristics of both hydraulic lime and crushed brick-lime mortars.
 P.J.P. Gleize et al. / Cement & Concrete Composites 31 (2009) 342–346 345

Table 5 3.3. General considerations

Chemical analysis of the binder (%wt) and cementation index – CI.

CaO (%) SiO2 (%)a Al2O3 (%) Fe2O3 (%) MgO (%) CIb Via visual inspection, it was found that the majority of the ren-
LAG1 12.46 29.63 5.81 2.60 0.18 7.17 dering mortars analyzed in this research are in an excellent state of
LAG2 18.23 30.56 5.87 1.19 0.11 5.05 conservation, despite their advanced age (more than 100 years-
LAG3 15.01 29.37 5.78 2.62 0.21 5.90 old), considering that some of them are in a state of total abandon-
SFS1 29.81 19.42 2.76 1.36 0.21 1.93 ment. This durability is essentially due to the fact that lime-based
SFS2 23.00 25.76 2.13 1.07 0.64 3.15
SFS3 34.59 19.00 1.41 0.44 0.17 1.58
mortars do not release soluble salts and are more chemically,
FLN1 40.64 7.84 0.76 0.61 0.55 0.56 structurally and mechanically compatible with the ancient ma-
FLN2 47.07 3.85 1.84 0.29 0.11 0.28 sonry [2,4,6,11,27,28]. So, our recommendation is the use of similar
FLN3 48.69 1.92 0.41 0.21 0.41 0.12 original materials for the mortars restoration purposes of the stud-
a
Quantified by gravimetric analysis. ied buildings, for example seashell lime and clay brick dust as bin-
b
CI < 0.3 – aerial; CI = 0.3–0.5 – low hydraulicity; CI = 0.5–0.7 – moderate der. However, the main problem is that these original ingredients
hydraulicity; CI = 0.7 to 1.1 – hydraulic. are no longer available:

2; 8x%SiO2 þ 1; 1x%Al2 O3 þ 0; 7x%Fe2 O3
CI ¼
ð%CaO þ 1; 4%MgOÞ
The AAS results also show that the lime used in all mortars is
calcitic, with magnesium only present in trace amounts (0.11–
0.64%). However, the lime produced in the coastal region of Santa
Catarina is dolomitic (CaO/MgO between 1.2 and 1.4), leading to
the following hypotheses about the origin of the lime used in the
mortars:
1. The lime, produced with calcitic limestone as the raw material,
was ‘‘imported” from other Brazilian states such as Minas
Gerais.
2. The lime was manufactured from the burning of mussel shells.
The first hypothesis is reasonable, but its probability is very low
due to the lack of transport facilities in historical times which led
to high transportation cost and to a preference for local materials.
The use of ‘seashell lime’ is thus the most realistic hypothesis,
although no information on its use in historical mortars could be
found in the literature, probably because of the local aspect of this
type of production.
3.2. Aggregate identification and characterization
The sedimentological analysis showed that the aggregates were
composed mainly of quartz, feldspar, some traces of mica, rock res-
idues and seashell fragments. The analysis allowed the classifica-
tion of the aggregates into two groups: beach sand for the São
Francisco do Sul mortars (SFS1, SFS2, SFS3), and river sand for
the others (LAG1, LAG2, LAG3, FLN1, FLN2, FLN3).
The grain size distributions of the sands are continuous. The
maximum and minimum dimensions of the sand grains are
2.8284–0.0884 mm for three of the buildings (FLN2, SFS1 and
LAG2) and 2.8284–0.0625 mm for the others.
Silts and clays were not detected. The percentage of seashells in
the aggregates varies from 7.73% in the LAG2 mortar to 46.03% for
FLN1 mortar. The gravel percentages in the sands are highest for
FLN2 19.52%, LAG2 – 16.73%, LAG2 – 14.95%, SFS3 – 10.33% and
SFS2 – 7.67%; and for the other mortars these percentages are
low. The method of hot HCl attack [16] allowed the classification
of the mortars according to their binder/aggregate ratio into three
groups: 1:1.2 (FLN1, FLN2); 1:2 (FLN3, LAG1, LAG2, LAG3, SFS2,
SFS3); and 1:3.44 (SFS1).
It is worth noting that the proportion of aggregates decreases
with the age of the building. The higher aggregate amount may
have been used in an attempt to prevent cracking or due to eco-
nomic aspects.
(a) ‘‘Modern” clay brick dust does not present the same physical
and chemical characteristics as the traditional ones because
production processes have changed a lot through the last
ð1Þ century. For example, nowadays, the temperatures for clay
brick production are higher than 900 °C and, to develop
some pozzolanic activity, clay brick dust must be fired
between 600 and 900 °C [21,29]).
(b) In the State of Santa Catarina, as mentioned above, commer-
cial lime is originated from limestone mining which turns
seashell lime production economically unviable.
However, two factors can lead to a regain of interest of the use
of seashell lime for restoration purpose. The first one is that com-
mercial lime originated from limestone mining produced in the
coastal region of Santa Catarina State is dolomitic, and it is well
known that, when compared to high calcium limes which is the
case of seashell lime, the late hydration of MgO from dolomitic
limes can be result in pitting [5] and the attack by SO2 from pol-
luted environments lead to the formation of deleterious magne-
sium sulfate salts [30]. The second one is due to a recent
specificity of our region. Currently the coastal region of the State
of Santa Catarina is leading the national production of oysters.
However, such industry has a potential of serious problem about
disposal of oyster shell waste which is, for the most part, illegally
dumped into the sea, leading to serious environmental problems
as, for example, waterfront silting.
Therefore, for restoration purposes and to ensure a compatibil-
ity with the old masonry, it can be considered the study of a repair
mortar with a binder composed by high calcium lime from the
burning of the oyster shell waste with small amounts of carefully
controlled calcined clay dust. A binder:aggregate ratio between
1:2 and 1:3 was determined as a benchmark. However, further
investigations on these materials are needed.
4. Conclusions
The main purpose of this study was to identify and characterize
the binders and aggregates of mortars used in nine historical build-
ings in order to collect data that will be considered for the formu-
lation of restoration mortars compatible with ancient masonry.
Seashell fragments were found in all of the mortars studied.
From the XRD, TGA, AAS and FT-IR results it was possible to con-
clude that the main binder of the mortars is a calcitic lime manu-
factured from seashells, which was totally carbonated.
Regarding the binder composition, it was possible to classify the
mortars into three groups:
1. Typical hydrated lime mortars in which the predominant binder
is hydrated lime.
346 P.J.P. Gleize et al. / Cement & Concrete Composites 31 (2009) 342–346
2. Hydraulic lime mortars with low hydraulic binder content, or
marly lime mortars; the trace amounts of hydraulic compounds
is probably due to impurities present in the raw material or
they could have been introduced during the lime production.
3. Crushed brick-lime mortars with medium hydraulic binder con-
tent: lime mixed with small fragments or dust of ceramic
materials.
The aggregates are of beach origin in the case of the San Fran-
cisco do Sul mortars (SFS1, SFS2 and SFS3), and the others are from
river origin. They are mainly comprised of quartz and seashell frag-
ments. Three average binder/aggregate ratios were measured:
1:1.2; 1:2; 1:3.44. It was also observed that the binder proportion
is lower for the more recent buildings.
Thus, for restoration purposes, it was suggested the study of re-
pair mortars with a binder composed by a mixture of seashell lime
and calcined clay dust and, a binder:aggregate ratio between 1:2
and 1:3.
Acknowledgements
The authors acknowledge the Brazilian research agencies Conse-
lho Nacional de Desenvolvimento Científico e Tecnológico – CNPq for
the financial support, and Coordenação de Aperfeiçoamento de Pes-
soal do Nível Superior – CAPES for providing the scholarship to the
second author. The authors also acknowledge the Instituto do
Patrimônio Histórico e Artístico Nacional – IPHAN of Santa Catarina
State, and the Instituto de Planejamento Urbano de Florianópolis –
IPUF, for the access to the historical buildings.
References
[1] Collepardi M. Degradation and restoration of masonry walls of historical
buildings. Mater Struct 1990;23(2):81–102.
[2] Callebaut K, Elsen J, Van Balen K, Viaene W. Nineteenth century hydraulic
restoration mortars in the Saint Michael0 s Church (Leuven, Belgium). Natural
hydraulic lime or cement? Cem Concr Res 2001;31(3):397–403.
[3] Moropoulou A, Bakolas A, Bisbikou K. Characterization of ancient, Byzantine
and later historic mortars by thermal and X-ray diffraction techniques.
Thermochim Acta 1995;269–270:779–95.
[4] Maravelaki-Kalaitzaki P, Bakolas A, Moropoulou A. Physico-chemical study of
Cretan ancient mortars. Cem Concr Res 2003;33(5):651–61.
[5] Rodriguez-Navarro C, Hansen E, Ginell WS. Calcium hydroxide crystal
evolution upon aging of lime putty. J Am Ceram Soc 1998;81(11):3032–4.
[6] Moropoulou A, Bakolas A, Bisbikou K. Investigation of the technology of
historic mortar. J Cult Herit 2000;1(1):45–58.
[7] Paama L, Pitkanen I, Ronkkomaki H, Peramaki P. Thermal and infrared
spectroscopic characterization of historical mortars. Thermochim Acta
1998;320(1–2):127–33.
[8] Riccardi MP, Duminuco C, Tomasi P, Ferloni P. Thermal, microscopic and X-ray
diffraction studies on some ancient mortars. Thermochim Acta 1998;321(1–2):
207–14.
[9] Alvarez JI, Navarro I, Martin A, Garcia Casado PJ. A study of the ancient mortars
in the north tower of Pamplona0 s San Cernin Church. Cem Concr Res
2000;30(9):1413–9.
[10] Bruni S, Cariati F, Ferno P, Pozzi A, Toniolo L. Characterization of ancient
mortars coming from northern Italy. Thermochim Acta 1998;321(1–2):161–5.
[11] Bakolas A, Biscontin G, Moropoulou A, Zendri E. Characterization of structural
byzantine mortars by thermogravimetric analysis. Thermochim Acta
1998;321(1–2):151–60.
[12] Franzini M, Leoni L, Lezzerini M. A procedure for determining the chemical
composition of binder and aggregate in ancient mortars: its application to
mortars from some medieval buildings in Pisa. J Cult Herit 2000;1(4):365–73.
[13] Gleize PJP, Silva DA, Nappi S. Ancient rendering from a Brazilian palace. Its
characterization and microstructure. Cem Concr Res 2000;30(10):1609–14.
[14] Guimarães JEP. A cal: Fundamentos e aplicações na engenharia Civil. São Paulo,
Brazil: Pini; 1998.
[15] Mascárelo SNPR. Arquitetura no Brasil. São Leopoldo, Brazil: Unisinos; 1982.
[16] Alvarez JI, Martin A, Garcia Casado PJ, Navarro I, Zornoza A. Methodology and
validation of a hot hydrochloric acid attack for the characterization of ancient
mortars. Cem Concr Res 1999;29(7):1061–5.
[17] Veiga MR. Methodologies for characterization and repair of mortars of ancient
buildings, Historical construction. In: Proceedings of the 3rd internal seminar
on historical construction. Guimarães: Portugal; 2001. p. 353–61 [in
Portuguese].
[18] Moropoulou A, Bakolas A, Anagnostopoulou S. Composites materials in ancient
structures. Cem Concr Compos 2005;27(2):295–300.
[19] Bakolas A, Biscontin G, Contardi V, Franceschi E, Moropoulou A, Palazzi D, et al.
Thermoanalytical research on traditional mortars in Venice. Thermochim Acta
1995;269–270:817–28.
[20] Ingo GM, Fragalà I, Bultrini G, De Caro T, Riccucci C, Chiozzini G. Thermal and
microchemical investigation of Phoenician–Punic mortars used for lining
cisterns at Tharros (western Sardinia, Italy). Thermochim Acta 2004;418(1–2):
53–60.
[21] Baronio G, Binda L. Study of the pozzolanicity of some brick and clays. Constr
Build Mater 1997;11(1):41–6.
[22] Moropoulou A, Bakolas A, Bisbikou K. Physico-chemical adhesion and cohesion
bonds in join mortars imparting durability to historic structures. Constr Build
Mater 2000;14(1):35–46.
[23] Moropoulou A, Tsiourva Th, Bisbikou K, Biscontin G, Bakolas A, Zendri E. Hot
lime technology imparting high strength to historic mortars. Construc Build
Mater 1996;10(2):151–9.
[24] Alvarez JI, Navarro I, Casado PJG. Thermal, mineralogical and chemical studies
of the mortars used in the cathedral Pamplona (Spain). Thermochim Acta
2000;365(1–2):177–87.
[25] Bakolas A, Biscontin G, Moropoulou A, Zendri E. Characterization of the lumps
in mortars of historic masonry. Thermochim Acta 1995;269–270:809–16.
[26] Hughes TL, Methven CM, Jones TGJ, Pelham SE, Fletcher P, Hall C. Determining
cement composition by Fourier transform infrared spectroscopy. Adv Cem Bas
Mater 1995;2(3):91–104.
[27] Genestar C, Pons C. Ancient covering plaster mortars from several convents
and Islamic and Gothic palaces in Palma de Mallorca (Spain). Analytical
characterization. J Cult Herit 2003;4(4):291–8.
[28] Montoya C, Lanas J, Arandigoyen M, Navarro I, Garcia Casado PJ, Alvarez JI.
Study of ancient dolomitic mortars of the church of Santa Maria de Zamarce in
Navarra (Spain): comparison with simulated standards. Thermochim Acta
2003;398:107–22.
[29] He C, Osbaeck B, Makavicky E. Pozzolanic reactions of six principal clay
minerals: activation, reactivity assessments and technological effects. Cem
Concr Res 1995;25(8):1691–702.
[30] Cultrone G, Arizzi A, Sebastián E, Rodriguez-Navarro C. Sulfation of calcitic and
dolomitic lime mortars in the presence of diesel particulate matter. Environ
Geol 2008;56:741–52.
 Cement & Concrete Composites 31 (2009) 347

Contents lists available at ScienceDirect

Cement & Concrete Composites

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

News: conferences and symposia

Second International Symposium on Design, Performance and
Use of SELF-CONSOLIDATING CONCRETE (SCC’2009-China), to be
held in Beijing, China, 5–8 June 2009.
Contact: Dr. Bo Peng, College of Civil Engineering, Hunan
University, Changsha, Hunan 410082, China. Tel/Fax: 86-731-882-
3937; E-mail: SCC2009@hnu.cn; Website: http://scc2009.hnu.cn.
1st International Workshop on Concrete Spalling due to Fire
Exposure, to be held in MFPA Institute, Leipzig, Germany, 3–5
September 2009.
Contact: Workshop Secretariat, E-mail: spalling@mfpa-leipzig.de;
Website for more information: www.mfpa-leipzig.de.
Non-conventional Materials and Technologies (IC-NOCMAT
2009): Materials for sustainable and affordable construction, to
be held in Bath, United Kingdom, 6–9 September 2009.
Contact: Prof. Pete Walker, BRE Centre for Innovative Construction
Materials, University of Bath, UK, E-mail: p.walker@bath.ac.uk;
Website: www.bath.ac.uk/ace/nocmat2009.
The Ninth International Symposium on Brittle Matrix Compo-
sites, to be held in Warsaw, Poland, 25–28 October 2009.
Contact: Prof. Dr. A.M. Brandt, Polish Academy of Sciences,
Institute of Fundamental Technological Research (IPPT PAN), Swieto-
krzyska 21, Warsaw 00-049, Poland. E-mail: abrandt@ippt.gov.pl;
Website: http://www.ippt.gov.pl/~abrandt/geninfo1.htm.
International Conference on Computational Design in Engineer-
ing (CODE 2009), to be held in Seoul, Korea, 3–6 November 2009.
Contact: Conference Secretariat, E-mail: code@code2009.org;
Website: http://www.code2009.org.
1st International Conference on Concrete Technology, to be held
in Tabriz, Iran, 6–7 November 2009.
Website: http://www.icct.ir/
Advanced Concrete Materials (ACM), International Conference, to
be held in Stellenbosch, South Africa, 17–19 November 2009.
Contact: Dr. Billy Boshoff, University of Stellenbosch, South Africa.
E-mail: admin@concretesociety.co.za; Website: http://www.concre-
tesociety.co.za/acm2009.
Iberian Conference on Fracture and Structural Integrity 2010, to
be held in Porto, Portugal, March 17–19, 2010.
Contact: Secretariat: Ms. Fernanda Fonseca, Faculdade de Engen-
haria da Universidade do Porto, Portugal. E-mail: ffonseca@
fe.up.pthttp://paginas.fe.up.pt/~cifie/.

doi:10.1016/S0958-9465(09)00074-2