Journal of Advanced Concrete Technology Vol.

10, 1-13, January 2012 / Copyright © 2012 Japan Concrete Institute 1

Scientific paper

Enhanced Model and Simulation of Hydration Process of Blast Furnace

Slag in Blended Cement
Yao Luan1, Tetsuya Ishida2, Toyoharu Nawa3 and Takahiro Sagawa4

Received 1 April 2011, accepted 4 October 2011 doi:10.3151/jact.10.1

Abstract
A hydration model for cementitious materials was applied to slag blended cement. The original model was found to
overestimate the hydration degree of slag, and the influence of the slag ratio on the hydration degree not to be well
simulated either. The hydration mechanism of slag was investigated, considering the role of calcium hydroxide as acti-
vator and the Ca/Si ratio of C-S-H. It is assumed that for low Ca/Si ratio, the C-S-H inner product has a stronger resis-
tance against ion diffusion and thus influences the hydration process significantly. Accordingly an enhanced model for
slag hydration is proposed. Finally, the enhanced model is verified by both hydration degree and heat generation tests.

1. Introduction urement of heat generation, such as the adiabatic tem-

perature rise test, is usually used to assess the hydration
In Japan blast furnace slag (BFS) is widely used in the status. However, because it is difficult to separate the
cement and concrete industry. As a by-product during heat released by slag accurately, this method is not ca-
steel manufacture, it is broadly agreed that concrete pable of evaluating the reaction of the slag separately. On
made with slag blended cement has many advantages, the other hand, several test methods have been developed
including better chloride resistance, lower permeability for quantitatively analyzing the hydration degree of slag
and higher strength at later age. From the viewpoint of separately, such as selective dissolution (Kondo and
circular economy and sustainable development, blast Ohsawa 1969; Luke and Glasser 1987; Lumley et al.
furnace slag is a promising mineral admixture in the 1996; Demoulian et al. 1980; Battagin 1992; Escalante et
concrete industry. However, some disadvantages are also al. 2001), XRD (Kondo et al. 1975; Hoshino et al. 2006),
reported, such as larger drying and autogenous shrinkage, Differential Scanning Calorimetry (DSC) (Totani et al.
which tend to induce early age cracks and decrease the 1980; Imoto et al. 2006) and image analysis (Feng et al.
durability of structures. Beside, BFS concrete is more 2004; Kurumisawa et al. 2007). Thanks to advances in
sensitive to the curing conditions. Insufficient water techniques, test methods are becoming more accurate
supply during curing may retard the hydration process and reveal the hydration process and reactivity of slag.
and cause a significant decrease in the long-time prop- For instance, Luke and Glasser (1987) modified the
erties. Therefore, for the purpose of promoting the effi- EDTA dissolution and studied pastes with 30% slag ratio
cient application of blast furnace slag, it is of great ne- until 12 months. They reported that within 1 month
cessity to investigate the properties of BFS blended ce- around 40% of the slag reacted while the hydration de-
ment. gree was up to 60% after 12 months. Lumley et al. (1996)
Hydration is a process that represents the reaction and studied the effects of the water-to-cementitious material
mass conversion of cement, and therefore the percentage (herein refer to both cement and slag, and is simplified
of hydrated cementitious materials is closely tied to mi- as w/b, i.e. water-to-binder) ratio and slag ratio, and they
crostructure development and affects other properties. indicated that 30-55% of the slag hydrated in 28 days and
Slag blended in Portland cement has a more complicated 45-75% in 1-2 years. They also found that as the w/b
hydration mechanism, since slag is a latent hydraulic ratio decreased or the slag ratio increased, the hydration
material and the reactivity is significantly influenced by degree decreased, and the slag type also influenced the
calcium hydroxide (CH), which acts as activator. Meas- reactivity. Escalante et al. (2001) also reported that the
reactivity of slag decreased with the w/b ratio or with
increases of the slag ratio. They pointed out also that the
1 hydration degree increases with the temperature, and the
Researcher, Department of Civil Engineering, values after 6 months were 20-50% in the range of
University of Tokyo, Japan. 10-30℃. In Japan, Iyoda and Dan (2007) and Sakai et al.
E-mail: [email protected] (2004) also confirmed that the hydration degree depends
2
Associate Professor, Department of Civil Engineering, on the slag ratio, and that for a long time most of the
University of Tokyo, Japan. values remain much lower than 100%. Based on the
3
Professor, Division of Sustainable Resources above reviews, the following characteristics of slag hy-
Engineering, Hokkaido University, Japan. dration can be identified: for long time curing, the hy-
4
Researcher, Research Institute, Division of Technology, dration is still far from completed even if sufficient water
Nittetsu Cement Co., Ltd., Japan.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 2

supplied. Most of the data show that the hydration degree ultimate heat in the case of complete hydration of the
is less than 60% and a large proportion of slag remains slag. For per unit weight of slag, the value of Qsg,∞, i.e.
unhydrated. Moreover, the slag ratio influences hydra- the specific heat, is assumed as 525.0 kJ/kg.
tion significantly and a higher ratio induces lower hy- In Eq. 1, except for temperature effect, βsg and λsg are
dration. the other two factors that noticeably affect the hydration
Several models have been proposed to simulate the process. The coefficient considering the effect of avail-
hydration of slag cement. In this paper, a able free water, βsg, is obtained by the following equa-
multi-component hydration heat model developed by tion:
Kishi (1994, 1995, 1996, 1997) and Maekawa et al.
⎧ ⎧⎛ ω s
⎫
(2008) is investigated. In this model, hydration and heat ⎞ ⎪⎫
1
⎪ ⎪ ⎪
generation of components in cement including clinkers β sg = 1 − exp ⎨ −r ⎨⎜⎜ free
⎟⎟ s ⎬
2
⎬ (3)
⋅η sg
sg
⎪⎩ ⎪⎩⎝ 100 ⎠ ⎪⎭ ⎪⎭
such as C3S, C2S, C3A, C4AF, and mineral admixtures
such as fly ash and blast furnace slag are treated respec-
tively by the Arrhenius law of chemical reaction and then where ηsg represents the normalized cluster thickness of
integrated together. This model has been verified by inner product, ωfree is the free water percentage. Physical
adiabatic temperature rise tests for both Portland cement water existing in micro-pores sustains continued hydra-
and slag blended cement concrete. On the other hand, tion and serves as space which hydrates precipitate. This
verification of the hydration degree of slag has not been portion of water is regarded as free water. Lack of free
carried out, because of the scarcity of reliable data when water would induce the stagnation of later hydration, e.g.
the model was established. Based on the above reviews low water-cement ratio concrete. ssg is the normalized
of the hydration degree, recent test methods have be- Blaine value of blast furnace, and r and s are material
come more accurate and some common characteristics constants. The following equation is given for ηsg:
are found. Therefore, in order to further expand the ap- 1
plication of the multi-component hydration heat model, it ⎛ Q ⎞3
is necessary to investigate its applicability on prediction η sg = 1 − ⎜ sg ⎟⎟ (4)
⎜Q
of hydration degree. Before describing this investigation, ⎝ sg ,∞ ⎠
the simulation of slag hydration in the multi-component
For ωfree, under sealed conditions, this equation can be
hydration heat model is introduced briefly first.
simplified as follows:
2. Simulation of slag hydration (Kishi and Wtotal − ∑Wi
Maekawa 1996, 1997) ω free = (5)
C
A reference heat rate that represents the heat generation where Wtotal is the initial water weight, Wi is the water
process under isothermal condition (293 K) is first as- consumed by various components in the cementitious
sumed for slag. Taking into account various factors such binder, and C is the unit cement weight in the paste.
as temperature, available free water, CH content and slag The coefficient expressing hydration stagnation due to
fineness, the hydration heat rate under real condition is CH insufficiency, λsg, is calculated by:
computed following the equation:
⎧⎪ ⎛ FCH ⎞ ⎫⎪
1.5

⎪⎧ E ⎛ 1 1 ⎞ ⎪⎫ λsg = 1 − exp ⎨−2.0 ⎜ ⎟ ⎬ (6)

H sg = min( β sg , λsg ) ⋅ H sg ,T0 (Qsg ) exp ⎨− sg ⎜ − ⎟ ⎬ (1) ⎪⎩ ⎝ RSGCH ⎠ ⎪⎭
⎪⎩ R ⎝ T T0 ⎠ ⎪⎭
where FCH is the CH content in the paste, and RSGCH is the
where Hsg is the hydration heat rate under real condition,
amount of CH required for fully hydration of slag at the
Hsg,T0 is the reference heat rate at 293 K, Qsg is the ac-
next time interval.
cumulated heat during hydration, Esg is the activation
energy of slag, R is the gas constant, T0 is the reference
temperature (293 K), and T is the real temperature. βsg is 3. Investigation on applicability of the
the coefficient considering the effect of available free model to slag
water, and λsg is the coefficient expressing hydration
stagnation due to CH insufficiency. 3.1 Simulation of hydration degree of slag and
The hydration degree is computed by the following CH content
Data of several tests from other researchers were col-
equation:
lected and corresponding simulations based on Eq. 2
Qsg ≡ ∫ H sg were conducted. All the tests were carried out with only
slag blended in Portland cement. Cement pastes were
Q (2)
α sg = sg cured under normal temperature with sufficient water
Qsg ,∞ supply. The experiment details are shown in Table 1.
Comparison between experiment and analysis are illus-
where αsg is the hydration degree of slag and Qsg,∞ is the trated in Fig. 1, which shows that differences exist. For
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 3

Table 1 Details of hydration degree tests.

Curing
No. Researcher Slag ratio w/b Blaine value Test method
temperature
a Iyoda and Dan (2007) 20%, 42%, 67% 0.5 4760 20℃ Selective dissolution

b Sakai et al. (2004) 20%, 50% 0.5 4200 20℃ Selective dissolution
Takamatsu and
c 30%, 45% 0.4 4050 20℃ Selective dissolution
Maruyama (2009)
Sagawa and Nawa
d 50% 0.5, 0.4, 0.3 4000 20℃ XRD-Rietveld
(2006a)
e Sagawa et al. (2010) 45% 0.5 4000 6000 20℃ XRD-Rietveld

f Lumley et al. (1996) 30%, 50%, 60% 0.4 4510 20℃ Selective dissolution

g Lumley et al. (1996) 30% 0.4, 0.6 4510 20℃ Selective dissolution

h Lumley et al. (1996) 60% 0.4, 0.6 4510 20℃ Selective dissolution

most cases, the analysis value appears higher than the factors that may affect hydration in the current model.
test data, which implies that the hydration degree may Only after a clear understanding of the working mecha-
be overestimated to a certain extent. Moreover, the rela- nism of these factors and the causes of the discrepancies
tionship of the hydration and slag ratio seems hardly is gained can modifications be conducted.
obvious in analysis. When the slag ratio is less than 50%,
there appears to be no difference between analytical 3.2 Key parameter tracing in the model
curves. Only when the slag ratio is as high as 67% in As presented in Section 2, in the current model, two
Fig. 1-(a) do the hydration curves reduce after around factors may noticeably affect the hydration process, i.e.
20 days. βsg and λsg. Case (a) in Table 1 is selected, and the values
Figure 2 represents the CH content measured by the of βsg, λsg and min(βsg, λsg) are output and illustrated in
TG-DTA method in test-(a) in Table 1 and the analysis. Fig. 3 according to Eq. 3 and Eq. 6. Their effects on the
For both the analysis and experiment, the amount of CH hydration rate can be clearly grasped. For low ratios such
declines as the slag ratio increases. However, it seems as 20% and 42%, βsg dominates the hydration process,
that when the slag ratio is 67%, the analytical result is while for 67% the alleviative effect at lager age is at-
underestimated. CH is completely consumed and the tributed to λsg. The development of min(βsg, λsg) is con-
value stays zero after 15 days, while the experiment sistent with the analytical results in Fig. 1-(a).
indicates that some CH always remains. Kwak et al. Next, the discussion focuses on the influence of CH.
(2004) also reported that even for a high slag ratio of λsg affects hydration only when the slag ratio is 67%,
70%, CH in the paste is detectable at a later age. There- because the analytical result of CH amount in Fig. 2 is
fore, it can de deduced that analysis may overrate con- almost zero. However, the test result indicates that some
sumption of CH for a high slag ratio, causing remaining amount of CH always exists, while the hydration process
CH to be underestimated. Combined with the analysis in is still reduced. In other words, the fact is that even
Fig. 1, some correlation seems to exist. If the hydration though sufficient CH remains, the hydration rate may
degree is suitably reduced, CH consumption is reduced still be influenced significantly. Thus it appears that the
and a better result in terms of CH content can be ex- current approach does not adequately describe this phe-
pected. nomenon.
Hence, the authors presume that current model has
some limitations. As described in Section 1, the model 3.3 Discussion of hydration mechanism of slag
was verified by the adiabatic temperature rise test, and blended in cement
the heat generation was found to be well simulated and It is well agreed that the principle hydrates of slag are
the temperature history of concrete to be traced properly. C-S-H gels. Compared with Portland cement, research-
On the other hand, it is probable that for slag, the model ers found that as the slag ratio increases, C-S-H gels
has some limitations with regard to the hydration degree. produced by slag show a lower Ca/Si ratio (Harrisson et
As these limitations may hinder further simulation such al. 1987; Shi et al. 1992; Richardson and Groves 1992;
as pore structure development, the authors assume that Gollop and Taylor 1992, 1996), as illustrated in Fig. 4.
appropriate modifications of the model are called for. The physical properties of C-S-H may vary due to the
This must be preceded by an investigation of the major reduced Ca/Si ratio. Richardson et al. (1994, 2004) re
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 4

100 100
Sakai et al. (2004)
Iyoda and Dan (2007) 80
80

Hydration degree (%)

Hydration degree (%)

60 60

40 40
Analysis Test Analysis Test
20% 20% 20% 20%
20 20
42% 42% 50% 50%
67% 67% 0
0
0 20 40 60 80 100 0 20 40 60 80 100
Age (days) Age (days)

(a) (b)
100 100
30% Sagawa and Nawa (2006a)
Analysis
45%
80 80
30%

Hydration degree (%)

Hydration degree (%)

Test
45%
60 60

40 40
Analysis Test
0.5 0.5
20 20
0.4 0.4
Takamatsu and Maruyama (2009) 0.3 0.3
0 0
0 7 14 21 28 0 20 40 60 80 100
Age (days) Age (days)
(c) (d)
100 100
Sagawa et al. (2010) Lumley et al. (1996)
80 80
Hydration degree (%)

Hydration degree (%)

60 60

40 40
Analysis Test
Analysis Test
30% 30%
20 4000 4000 20 50% 50%
6000 6000 60% 60%
0 0
0 10 20 30 40 50 60 0 20 40 60 80 100 120
Age (days) Age (weeks)
(e) (f)
100 100

80 80
Hydration degree (%)
Hydration degree (%)

Lumley et al. (1996)

60 60

40 Lumley et al. (1996) 40

Analysis Test Analysis Test

20 0.4 0.4 20
0.4 0.4
0.6 0.6 0.6 0.6
0 0
0 20 40 60 0 20 40 60 80 100 120
Age (weeks) Age (weeks)
(g) (h)
Fig. 1 Comparison of hydration degree based on original model.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 5

ported that in the case of a high Ca/Si ratio, C-S-H has a 25 Test Analysis
fibrillar-like morphology, while in the case of a low 20% 20%
Ca/Si ratio, the morphology tends to be of the crumpled 20 42% 42%
67% 67%
foil type. This implies a transition of hydrates from es-

CH content (%)
sentially one-dimensional growth to two-dimensional 15
growth, resulting in higher tortuosity and specific sur-
face area in micro-pores. Moreover, the electrical 10
charges on the surface of hydrates are influenced by the
Ca/Si ratio. The surface of hydrates from Portland ce- 5
ment is usually positively charged (Elakneswaran et al. Iyoda and Dan (2007)
2009; Glasser 1992; Nakarai et al. 2006). According to 0
0 20 40 60 80 100
the research from Glasser (1992), as the Ca/Si ratio de-
Age (days)
cline, the electric charges become less positive, and
when the Ca/Si ratio is lower than 1.2-1.3, the surface Fig. 2 Comparison of CH content based on original
charge of C-S-H reverses to negative, and C-S-H tend to model.
bind alkali cations. Studies on the pore solution (Bhatty
and Greening 1978; Chatterji 1979; Hong and Glasser 1.2
1999; Shehata and Thomas 2006) indicated that the
available alkali concentration decreases as the slag ratio 1.0
42%

λsg, factor related to

increases or the Ca/Si ratio decreases, which also sup-

CH insufficiency
0.8 20%
ports the reversal of surface charge and ion binding ca- 20%
pacity of C-S-H. 0.6
Herein the hydration scheme of slag is discussed ac- 42%
cording to the basic information of activation (Tanaka et 0.4
67%
al. 1983; Taylor 1990). At the starting stage, due to the
activation by Ca2+, the protective film on the surface of 0.2
the slag particle is destroyed. The internal structure,
0.0
which is an anionic network composed of O, Si and other 0 20 40 60 80 100
elements, starts to decompose, and silicate ions such as Age (days)
Si(OH)4 are released into the solution. When silicate ions
contact Ca2+, a reaction occurs and hydrates start to pre- 1.0
cipitate. Like in the case of Portland cement, hydrates
from slag can be divided into two categories, i.e. the
0.8
inner product and the outer product. The inner product
βsg, factor related to

20%
lack of free water

consists of hydrates generated inside the radius of the 42%

0.6
original particle and closely bond around an unhydrated 67%
core, while the outer product precipitates outside the
0.4
range of the original particle and fills in capillary pores.
In order to sustain hydration, it is necessary for silicate
0.2
ions released from slag to diffuse outside the inner
product, or alkali cations to diffuse inside. The inner
0.0
product may show resistance against ion diffusion, i.e. as
0 20 40 60 80 100
hydration goes on and the thickness of the inner product
Age (days)
increases, the diffusivity of ions through the inner
product becomes lower. Accordingly, the reaction be-
1.0
comes slower or even stagnates. Therefore, the thickness
of the inner product is a crucial factor that dominates the
0.8
hydration rate of slag.
As discussed above, the Ca/Si ratio of hydrates is re- 20%
Min ( βsg, λsg )

0.6
duced with increases in the slag ratio, and the physical 42%
properties of C-S-H may vary accordingly. Hence, when 67%
0.4
we discuss the resistance effect of the inner product, not
only its thickness but also the various other physical
0.2
properties have to be taken into account. In this study, the
authors assume that the physical properties of C-S-H
0.0
influence the hydration process from two aspects. First,
0 20 40 60 80 100
since C-S-H with a lower Ca/Si ratio has a morphology
Age (days)
that is foil-like rather than fibrillar, the tortuosity of gel
Fig. 3 Parameter tracing.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 6

2.0 move toward the inner product and react with silicate
ions near it. The reverse ion movement implies that
1.8
denser inner product having stronger resistance against
1.6 ion diffusion is formed. Further ion diffusion may be
blocked and the hydration rate afterwards is reduced
Ca/Si ratio

1.4 significantly. The above discussion about the relation-

1.2 Harrisson et al. (1987) ship between the hydration of the slag and inner product
Shi et al. (1992) is illustrated in Fig. 5. Accordingly, the authors conclude
1.0 Richardson and Groves (1992) that for a lower Ca/Si ratio due to both higher tortuosity
Gollop and Taylor (1992, 1996) and reversed electric charge, the inner product layer
0.8
shows higher resistance to ion movement. This may be
0 20 40 60 80 100
Slag ratio (%)
the main reason that the hydration degree decreases as
the slag ratio increases.
Fig. 4 Ca/Si ratio of C-S-H with various slag ratios.
4. Enhanced hydration model for blast
furnace slag
pores becomes higher. This implies more twists and
forks in pores, resulting in a much longer path for ions to 4.1 Modification on hydration rate
diffuse through the inner product. The other factor may From the discussion in Section 3.3, it can be seen that the
be the alternated electric charge on the surface of C-S-H. hydration process is closely associated with the inner
As indicated by Elakneswaran et al. (2009), pores with a product because of its resistance to ion movement. This
diameter of less than 10 nm show the electrical double resistance depends on both the thickness and intrinsic
layer (EDL) effect. The inner product contains mainly properties of C-S-H. In the original model, although the
gel pores and a large proportion of gel pores have a di- thickness of the inner product is included in Eq. 3, it
ameter smaller than 10 nm, and thus the EDL effect of provides constant evaluation for various slag ratios. In
the inner product is obvious. In the case of C-S-H with a other words, the various intrinsic properties of C-S-H
high Ca/Si ratio, the surface is positively charged. The due to the Ca/Si ratio are not taken into account. On the
repelling effect of the positive charge layer restrains the other hand, coefficient λsg, which represents the ineffi-
movement of Ca2+ towards the inner product. The prin- ciency of CH, seems unnecessary because it has been
cipal ion movement is that of silicate ions that are re- proved that a sufficient amount of CH exists in the paste
leased through the inner product. When the electric even for high slag ratios as described in Section 3.2.
charges become less positive or even negative due to the Hence in the modification λsg is not utilized and all
lower Ca/Si ratio, the principal ion movement may factors are integrated into coefficient βsg. Since the hy-
change. Due to the absorbing effect, more Ca2+ ions dration degree is related to the w/b ratio and slag fineness,

Expand and fill in High C/S ratio case

capillary pores

Ca2+
• Silicate ions diffuse outside
Ca2+
inner product due to positive
Outer product
Si(OH)4
Si(OH)4 charges of C-S-H
• Inner product with relatively
Ca2+
Ca2+ coarse structure and low
tortuosity, easy for ion movement
Si(OH)4

Si(OH)4
Ca2+ • High hydration degree
Unhydrated Ca2+

Si(OH) core Ca2+

4

Si(OH)4

Low C/S ratio case

Inner product

Ca2+
• More Ca2+ diffuse inside inner
Ca2+
Si(OH)4
Ca2+ product due to less positive or
negative charges of C-S-H
Si(OH)4
• Much denser inner product
Ion movement during slag hydration Ca2+
with higher tortuosity which is
• Ca2+ diffusion into inner product Ca2+ Si(OH)4
difficult for further ion movement
• Unhydrated core releases silicate ions • Low hydration degree
Ca2+

Fig. 5 Influence of inner product on ion movement and hydration of slag.

Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 7

those factors in Eq. 3, i.e. ωfree and ssg, need to be re- 1.0
served. Moreover, the most important issue is how to

χ, factor affecting hydration rate

BFS20%

due to inner product thickness

simulate the effect of the inner product. When the 0.8 BFS30%
thickness of inner product is zero, i.e. the starting of BFS40%
hydration, no resistance from the inner product exists. As 0.6 BFS50%
the thickness of the inner product increases, the hydra- BFS60%
tion rate should decrease. Furthermore, as explained in 0.4 BFS70%
Section 3.3, for a lower Ca/Si ratio, the inner product
shows stronger resistance towards ion movement, so the 0.2
effect needs to be more significant. Therefore, an avail-
able approach is to establish an influential coefficient 0.0
which includes the thickness of inner product as the 0.0 0.1 0.2 0.3 0.4 0.5
principal variable and meanwhile reflects the effect of ηsg, inner product thickness
Ca/Si ratio. According to Fig. 4, the Ca/Si ratio of C-S-H
shows a clear correlation with the slag ratio, and thus the Fig. 6 function χ versus ηsg.
slag ratio seems a proper index of the influence of the
Ca/Si ratio. For the inner product, as it becomes thicker,
where ωfree, ssg are defined as in Eq. 3, and r and s are
the ion diffusion is strongly resisted so the hydration rate
also the same material constants.
decreases sharply rather than gradually. The exponential
In Eq. 9, χ is integrated into βsg. The effects of the
formula seems appropriate to describe the significant
available free water and slag fineness are reserved by
relationship between inner product thickness and the
ωfree and ssg. Unlike the original model, the thickness of
resistance effect. Therefore, the effect of the inner
the inner product, ηsg, is extracted and included in χ. This
product is proposed by a new coefficient χ, defined as
means that the inner product has an independent effect on
follows:
the hydration rate from the available free water or slag
χ = exp(− a ⋅ η sgb ) (7) fineness. For instance, even if sufficient water is pro-
vided such as a high w/b ratio, later hydration may still
where ηsg is the thickness of the inner product, which has tend to stagnate. Therefore, this separate treatment sub-
the same definition in Eq. 4. a and b are intrinsic pa- stantially corresponds to the mechanism of hydration.
rameters that are the functions of the slag ratio, given by:
4.2 Verification by hydration degree of slag and
a = 19.0 − 20.0 ⋅ psg CH content
(8)
b = 2.2 − 3.0 ⋅ psg The enhanced model is verified by experiments in Table
1, and the results are illustrated in Fig. 7. For most cases,
In Eq. 8, psg is the slag ratio. Therefore, the value the analytical results are much improved, showing the
development of χ depends on the slag ratio. Herein Eq. 8 dependence of the hydration process on the slag ratio.
is not directly derived from experiment. Instead, it is For different w/b ratios, the enhanced model can also
empirically proposed to alter the hydration heat. On the give better results, except the slag ratio 20% in case (a) in
other hand, the basis of Eq. 8 is the intrinsic properties of Table 1, for which neither the analyses before and after
C-S-H which depends on the Ca/Si ratio and its influence modification provide good agreement with the experi-
on ion movement. In other words for Eq. 8 the physical ment, since the test data is even higher than 90%. For
meaning relating to microscopic mechanism of hydration other cases, because hydration degrees are largely de-
exists. Figure 6 plots the curves of χ versus ηsg when the creased in the enhanced model, the values become much
slag ratio changes from 20% to 70%. All the functions closer to the test data especially at early age. For long
decrease as the thickness of the inner product increases, period such as after 91d, the analysis is slightly deviating
which implicitly represents the resistance by the inner from the test data. This is because that the hydration rate
product to ion diffusion. Besides, for a higher slag ratio, in the analysis is not completely zero. On the other hand,
the function decreases more sharply from the beginning. some tests such as case (a) show that hydration com-
This has clear physical meaning that the C-S-H inner pletely stagnates around 91d, while in case (g) the hy-
product has a stronger retardation effect on ion move- dration can gradually proceed until quite long period.
ment the lower the Ca/Si ratio. When χ is integrated into More accurate measurement and simulation until quite
βsg, it is applicable to the simulation of the development long time need to be conducted in the future research. In
of the hydration degree for various slag ratios. any case, comprehensively speaking the simulation of
Accordingly, in the enhanced model, the new expres- slag hydration is greatly improved. Moreover, besides
sion for βsg is given by: the factors that are taken into account in the model, the
hydration degree of slag is also influenced by other fac-
⎧ ⎫⎫
⎪⎧
s
⎪⎧⎛ ω free ⎞ 2 ⎪⎫
1
⎪ ⎪⎪ tors, such as the chemical composition or glass content of
β sg = ⎨1 − exp ⎨ −r ⎨⎜ ⎟ / ssg ⎬ ⎬⎬ ⋅ χ (9) the slag. For those factors which are not included in the
⎪⎩ ⎪⎩ ⎩⎪⎝ 20.0 ⎠ ⎭⎪ ⎪⎪
⎭⎭ model, their influences on the hydration degree should be
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 8

100 100
Sakai et al. (2004)
80 Iyoda and Dan (2007) 80
Hydration degree (%)

Hydration degree (%)

60 60

40 40
Analysis Test Analysis Test
20 20% 20% 20 20% 20%
42% 42%
50% 50%
67% 67%
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Age (days) Age (days)
(a) (b)

100 100
Analysis Test
30% Analysis 0.5 0.5
80 45% 80
0.4 0.4
Hydration degree (%)

30%
45%
Test Hydration degree (%) 0.3 0.3
60 60

40 40

20 20
Takamatsu and Maruyama (2009) Sagawa and Nawa (2006a)
0 0
0 7 14 21 28 0 20 40 60 80 100
Age (days) Age (days)
(c) (d)
100 100
Sagawa et al. (2010) Lumley et al. (1996)
80 80
Hydration degree (%)

Hydration degree (%)

60 60

40 Analysis Test 40
Analysis Test
4000 4000 30% 30%
20 20 50% 50%
6000 6000
60% 60%
0 0
0 10 20 30 40 50 60 0 20 40 60 80 100 120
Age (days) Age (weeks)
(e) (f)
100 100 Analysis Test
Lumley et al. (1996) 0.4 0.4
80 80
0.6 0.6
Hydration degree (%)
Hydration degree (%)

60 60

40 40
Analysis Test
0.4 0.4
20 20
0.6 0.6 Lumley et al. (1996)
0 0
0 20 40 60 0 20 40 60 80 100 120
Age (weeks) Age (weeks)
(g) (h)
Fig. 7 Comparison of hydration degree based on enhanced model.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 9

studied in the future to allow more precise prediction. 7, the original model highly overestimates the hydration
Besides the hydration degree, the analysis for CH con- degree of slag, while the analysis based on enhanced
tent for case (a) in Table 1 is shown in Fig. 8. The en- model is consistent with the test data. This means that the
hanced model provides much better simulation. Even for modified model is applicable to various temperature
a high slag ratio of up to almost 70%, CH is still not ranges.
completely consumed, which is simulated well by the
enhanced model. 4.3 Modification of reference heat rate of slag
The tests adopted in Table 1 are all under the reference Based on the modification in Section 4.1, it is probable
temperature (293 K). If the curing temperature increases, that not only the hydration degree but also heat genera-
a higher hydration degree of slag will be obtained (Es- tion are influenced. This is easy to understand, as in the
calante et al. 2001). In the original model, this effect is modified model, hydration degrees are greatly reduced,
included in Eq. 1 and it still remains after modification of but the specific heat of slag Qsg,∞ in Eq. 2 still remains
the model. Therefore in order to expand the applicability 525.0 kJ/kg, which implies that total heat generation is
of the modified model, the hydration degrees of slag reduced. Therefore in order to ensure the predictability
under various curing temperatures are also analyzed. The of heat generation, the authors considered modifying the
results are compared with test data from Sagawa and value of specific heat. Few test data on the specific heat
Nawa (2006b). In the research, 40% of the Portland of slag are available in the literature. Goto et al. (2006,
cement is replaced by slag and the cement pastes are 2009) synthesized slag-simulated glasses in the system
cured in water with temperatures of 5°C, 20°C and 40°C. of CaO-Al2O3-SiO2 and measured heat generation dur-
The w/b ratio is 0.5. The hydration degrees of slag are ing hydration. They proposed that the specific heat of
measured by the XRD-Rietveld method. The analysis slag is 571.2 kJ/kg. Referring to this data, the authors
based on original and enhanced model, and their com- assume that in the enhanced model, the value of Qsg,∞ is
parison with analysis are given in Fig. 9. Similar with 630.0 kJ/kg, which is 1.2 times the value in the original
cases under normal temperature shown in Fig. 1 and Fig. model. This value is not far from Goto’s test considering
the difference between real and synthesized slags.
25 Meanwhile with the increased value, effective compen-
Test Analysis
sation for the heat generation can be obtained. The new
20% 20%
20 42% 42%
function of the reference heat rate is also increased by
67% 67% multiplying both the original reference heat rate with the
CH content (%)

15
constant value of 1.2, as illustrated in Fig. 10. Direct
verification by test curve from slag is unavailable, since
it is difficult to accurately separate the heat of the slag
10
from that of the total blended cement. Herein, the au-
thors decide to compare total heat generation of Port-
5
land cement blended with slag under isothermal condi-
Iyoda and Dan (2007) tion. Gruyaert et al. (2010) measured the total heat gen-
0
0 20 40 60 80 100
eration of BFS cement by isothermal calorimetry with
Age (days)
various slag ratios. The w/b ratio is 0.50 and the iso-
thermal conditions are 10℃, 20℃ and 35℃. In this
Fig. 8 Comparison of CH content based on enhanced study, corresponding analysis by the modified model
model. were carried out and compared with their measurement
as shown in Fig. 11. For slag ratio 0%, i.e. OPC, the

100 100
Sagawa and Nawa (2006b)
80 80
hydration degree (%)

hydration degree (%)

60 60

40 40
Analysis Test Analysis Test
5℃ 5℃ 5℃ 5℃
20 20
20℃ 20℃ 20℃ 20℃
40℃ 40℃ 40℃ 40℃
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Age (days) Age (days)

(a) Original model (b) Enhanced model

Fig. 9 Hydration degree under various temperatures based on the original and enhanced models.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 10

Table 2 Mix proportion of BFS concretes used for 16.0

verification (Maekawa et al. 2008, p.82). 14.0

Referent heat rate (kJ/kg/hr)

12.0 Modified
w/b Unit weight (kg/m3)
Types 10.0
(%)
Water Cement Slag Sand Gravel 8.0
Final heat for
6.0 Original
full hydration
SG400 39.3 157 240 160 645 1,129
4.0

SG300 49.3 148 180 120 757 1,129 2.0

0.0
SG200 78.5 157 120 80 854 1,089 0 200 400 600 800
Accumulated heat (kJ/kg)
Fig. 10 Reference heat rate of slag before and after
modification.
analysis follows Kishi’s original model. Its agreement
with the test data suggests that the disturbance from the 400
Portland cement part in blended cement can be ne- 350 Isothermal condition 10℃
glected. Furthermore, when we turn to the BFS cases,

Total heat production (J/g)

300
for all the isotherm condition and slag ratios, both the
process and final heat can be well traced. Therefore, it 250
can be concluded that the modification of the specific 200
heat and referent heat rate of slag is reasonable. Herein 150 0%BFS
it has to be pointed out that although the enhanced model 30%BFS Analysis
100 50%BFS
is supported by the heat generation test, the specific heat 0%BFS
value 630.0 kJ/kg adopted in the model is not directly 50 30%BFS Test
50%BFS
verified due to the scarcity of test data available. Meas- 0
urement of the specific heat of slag should be carried out 0 50 100 150 200 250
in future research. Age (hours)
Finally, as mentioned in Section 1, the original model
was verified by the adiabatic temperature rise test of BFS 400
concrete. In order to remain consistent with the original 350 Isothermal condition 20℃
Total heat production (J/g)

model, the adiabatic temperature rise adopted in the 300

original verification is analyzed again with the modified
250
model. The mix proportions of BFS concrete are shown
in Table 2, and comparison of results is shown in Fig. 12. 200
0%BFS
Obviously, with increased specific heat and decreased 150 30%BFS Analysis
hydration degree, the modified model provides similar 100
50%BFS
evaluation of adiabatic temperature rise as the original 0%BFS
50 30%BFS Test
model. Therefore, the modified model can be used not 50%BFS
only under isothermal condition but also adiabatic con- 0
dition. 0 50 100 150 200
Age (hours)
5. Summary and conclusions
400
Isothermal condition 35℃
In this study, the hydration process of slag blended in 350
Total heat production (J/g)

cement is investigated based on the intrinsic properties of 300

the inner product. As the original model appears not to
provide proper prediction of the hydration degree, the 250
hydration mechanism of slag is investigated again, con- 200
0%BFS
sidering the influence of CH as activator and the 150 30%BFS Analysis
chemo-physical properties of C-S-H in slag hydration. 50%BFS
100 0%BFS
Since the Ca/Si ratio of C-S-H decreases as the slag ratio
50 30%BFS Test
increases, resulting in both higher tortuosity and changed 50%BFS
positive electric charge accumulation in C-S-H, it is 0
assumed that the C-S-H inner product has a strong re- 0 50 100 150
sistance to ions movement, and therefore the subsequent Age (hours)
hydration tends to be retarded. Based on the above as- Fig. 11 Comparison of isothermal heat generation based
sumption, an enhanced model is proposed, taking into on modified model.
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 11

70 alkalies with hydrating and hydrated calcium

Initial casting temperature 10℃ silicates.” In: Proceedings of 4th International
60
Conference on Effects of Alkalis in Cement and
Adiabatic temperature (℃)

SG400
50 Concrete, Purdue August 1978. 87-112.
Chatterji, S., (1979). “The role of Ca(OH)2 in the
40 SG300 breakdown of portland cement concrete due to
30 alkali-silica reaction.” Cement and Concrete Research,
SG200 9(2), 185-188.
20
Original model Demoulian, E., Gourdin, P., Hawthorn, F. and Vernet, C.,
10
Modified model (1980). “Influence of slags chemical composition and
0
texture on their hydraulicity.” In: Proceedings of 7th
0 1 2 3 4 5 6 7 International Congress on the Chemistry of Cement,
Age (days) Paris. Vol.2, 89-104.
Elakneswaran, Y., Nawa, T. and Kurumisawa, K., (2009).
70 “Electrokinetic potential of hydrated cement in
Initial casting temperature 20℃ relation to adsorption of chlorides.” Cement and
60 Concrete Research, 39(4), 340-344.
Adiabatic temperature (℃)

SG400
50 Escalante, J. I., Gomez, L. Y., Johal, K. K., Mendoza, G.,
Mancha, H. and Mendez, J., (2001). “Reactivity of
40 SG300 blast-furnace slag in portland cement blends hydrated
30 under different conditions.” Cement and Concrete
SG200 Research, 31(10), 1403-1409.
20
Feng, X., Garboczi, E. J., Bentz, D. P., Stutzman, P. E.
Original model
10 and Mason, T. O., (2004). “Estimation of the degree of
Modified model
hydration of blended cement pastes by scanning
0
electron microscope point-counting procedure.”
0 1 2 3 4 5 6 7
Cement and Concrete Research, 34(10), 1787-1793.
Age (days)
Glasser, F. P., (1992). “Chemistry of alkali-aggregate
70
reaction.” In: R. Swamy, Ed. The Alkali-Silica
Initial casting temperature 30℃ SG400 Reaction in Concrete, Van Nostrand Reinhold, New
60 York. 30-53.
Adiabatic temperature (℃)

50
Gollop, R. S. and Taylor, H. F. W., (1992).
“Microstructural and microanalytical studies of sulfate
40 SG300 attack: I. ordinary Portland cement paste.” Cement and
30
Concrete Research, 22(6), 1027-1038.
SG200 Gollop, R. S. and Taylor, H. F. W., (1996).
20
Original model “Microstructural and microanalytical studies of sulfate
10 Modified model attack: IV. reactions of a slag cement paste with
sodium and magnesium sulfate solutions.” Cement
0 and Concrete Research, 26(7), 1013-1028.
0 1 2 3 4 5 6 7 Goto, S., (2009). “Hydration of hydraulic materials – a
Age (days) discussion on heat liberation and strength
Fig. 12 Adiabatic temperature rises based on original and development.” Advances in Cement Research, 21,
modified model. 113-117.
Goto, S., Fujimori, H., Hidaka, A. and Ioku, K., (2006).
“Effect of components on the rate of heat liberation of
account the resistance effect. Finally, the enhanced hydration in the system of glass/gypsum/lime.”
model is applied to simulation of slag hydration, and the Journal of the European Ceramic Society, 26,
results indicate good agreement with the experiment data. 771-776.
Moreover, the amount of CH in the paste and heat gen- Gruyaert, E., Robeyst, N. and De Belie, N., (2010).
eration of slag cement can also be simulated accurately. “Study of hydration of Portland cement blended with
blast-furnace slag by calorimetry and
References thermogravimetry.” Journal of Thermal Analysis and
Battagin, A. F., (1992). “Influence of degree of hydration Calorimetry, 102, 941-951.
of slag on slag cement.” In: Proceedings of the 9th Harrisson, A. M., Winter, N. B. and Taylor, H. F. W.,
International Congress on the Chemistry of Cement, (1987). “Microstructure and microchemistry of slag
New Delhi, India, Vol.3, 166-172. cement pastes.” In: L. Struble and P. Brown, Eds.
Bhatty, M. S. Y. and Greening, N., (1978). “Interaction of Materials Research Society Symposium Proceedings,
Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 12

Materials Research Society, Pittshurgh. 85, 213-222. W., (1996). “Degrees of reaction of the slag in some
Hong, S. Y. and Glasser, F. P., (1999). “Alkali binding in blends with portland cements.” Cement and Concrete
cement pastes: Part I. the C-S-H phase.” Cement and Research, 26(1), 139-151.
Concrete Research, 29(12), 1893-1903. Maekawa, K., Ishida, T. and Kishi, T., (2008).
Hoshino, S., Yamada, K. and Hirao, H., (2006). “Multi-scale modeling of structural concrete.” London
“XRD/Rietveld analysis of the hydration and strength and New York: Taylor & Francis.
development of slag and limestone blended cement.” Nakarai, K., Ishida, T. and Maekawa, K., (2006).
Journal of Advanced Concrete Technology, 4(3), “Modeling of calcium leaching from cement hydrates
357-367 coupled with micro-pore formation.” Journal of
Imoto, H., Sakai, E., Hukui, R. And Daimon, M., (2006). Advanced Concrete Technology, 4(3), 395-407.
“Measurement of blast furnace slag content ratio in Richardson, I. G., (2004). “Tobermorite/jennite and
cement by DSC.” In: Proceedings of the Japan tobermorite/calcium hydroxide-based models for the
Concrete Institute. 28(1), 29-34. (In Japanese) structure of C-S-H: applicability of hardened pastes of
Iyoda, T. and Dan, Y., (2007). “The Effect of temperature tricalcium silicate, β-dicalcium silicate, portland
on hydration rate of slag in blended cement.” In: cement, and blends of portland cement with
Proceedings of JSCE 333 Committee Symposium. blast-furnace slag, metakaolin, or silica fume.”
II59-62. (In Japanese) Cement and Concrete Research, 34(9), 1733-1777.
Kishi, T. and Maekawa, K., (1994). “Thermal and Richardson, I. G., Brough, A. R., Groves, G. W. and
mechanical modeling of young concrete based on Dobson, C. M., (1994). “The characterization of
hydration process of multi-component cement hardened alkali-activated blast-furnace slag pastes and
minerals.” In: R. Springeschmid, Ed. Proceedings of the nature of the calcium silicate hydrate (C-S-H)
the International RILEM Symposium on Thermal phase.” Cement and Concrete Research, 24(5),
Cracking in Concrete at Early Ages, Munich. 11-18. 813-829.
Kishi, T. and Maekawa, K., (1995). “Multi-component Richardson, I. G. and Groves, G. W., (1992).
model for hydration heat of portland cement.” Journal “Microstructure and microanalysis of hardened
of Materials, Concrete Structure and Pavements cement pastes involving ground granulated
(JSCE), 29, 97-109. (In Japanese) blast-furnace slag.” Journal of Materials Science, 27,
Kishi, T. and Maekawa, K., (1996). “Multi-component 6204-6212.
model for hydration heat of blended cement with blast Sagawa, T., Ishida, T., Luan, Y. and Nawa, T., (2010).
slag and fly ash, Journal of Materials.” Concrete “Hydrate composition analysis and micro structure
Structures and Pavements (JSCE), 33, 131-143. (In characteristics of portland cement-blast furnace slag
Japanese) system.” Journal of JSCE E, 66(3), 311-324. (In
Kishi, T. and Maekawa, K., (1997). “Multi-component Japanese)
model for hydration heating of blended cement with Sagawa, T. and Nawa, T., (2006a). “Hydration of blast
blast furnace slag and fly ash.” Concrete library of furnace slag cement and effect of water to cement ratio
JSCE, 30, 125-139. on hydration of blast furnace slag cement. ” Cement
Kondo, R., Abo-el-enein, S. A. and Daimon, M., (1975). Science and Concrete Technology, 60, 82-87. (In
“Kinetics and mechanisms of hydrothermal reaction of Japanese)
granulated blast furnace slag.” Bulletin of the Sagawa, T. and Nawa, T., (2006b). “Quantitative
Chemical Society of Japan, 48(1), 222-226. hydration analysis of blast furnace slag cement by
Kondo, R. and Ohsawa, S., (1969). “Studies on a method Rietveld method.” Cement Research and Technology,
to determine the amount of granulated blast furnace 17, 1-11. (In Japanese)
slag in cement.” In: Proceedings of the 5th Sakai, E., Imoto, H. and Daimon, M., (2004). “Hydration
International Symposium on the Chemistry of Cement, and strength development of blastfurnace slag
Tokyo. Ⅳ, 255-262. cement.” In: Proceedings of JCI, 26(1), 135-140. (In
Kurumisawa, K., Nawa, T. and Elakneswaran, Y., (2007). Japanese)
“Study on chloride permeability and microstructure of Shehata, M. H. and Thomas, M. D. A., (2006). “Alkali
blast furnace slag blended paste.” In: Proceedings of release characteristics of blended cements.” Cement
JSCE 333 Committee Symposium, II37-40. (In and Concrete Research, 36(6), 1166-1175.
Japanese) Shi, C., Day, R. L., Wu, X. and Tang, M., (1992).
Kwak, D., Kokubu, K., Lee,C. and Lee, K., (2004). “Microstructure and performances of alkali–slag
“Influence of replacement ratio on pore structure of cement pastes.” In: Proceedings of 9th International
concrete with ground granulated blast furnace.” In: Congress on the Chemistry of Cements, New Delhi,
Proceedings of JCI, 26(1), 783-788. (In Japanese) India. 3, 298-304.
Luke, K. and Glasser, F. P., (1987). “Selective dissolution Takamatsu, N. and Maruyama, I., (2009). “Study on
of hydrated blast furnace slag cements.” Cement and volume deformation and hydration degree of blast
Concrete Research 17(2), 273-282. furnace slag replaced matrix.” In: Proceedings of JCI,
Lumley, J. S., Gollop, R. S., Moir, G. K. and Taylor, H. F. 31, 187-192. (In Japanese)
 Y. Luan, T. Ishida, T. Nawa and T. Sagawa / Journal of Advanced Concrete Technology Vol. 10, 1-13, 2012 13

Taylor, H. F. W., (1997). “Cement Chemistry.” 2nd ed. Totani, Y., Saito, Y., Kageyama, M. and Tanaka, H.,
Thomas Telford Ltd. (1980). “The hydration of blast furnace slag cement.”
Tanaka, H., Totani, Y. and Saito, Y., (1983). “Structure of In: Proceedings of the 7th International Congress on
hydrated glasy blast furnace slag contents.” ACI the Chemistry of Cement, Paris, 1980. 2 III, 95-98.
Special Publication, 79, 963-977.

View publication stats