Cement and Concrete Research 92 (2017) 1–15

Contents lists available at ScienceDirect

Cement and Concrete Research

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

The mechanism of limited inhibition by fly ash on expansion due to

alkali–silica reaction at the pessimum proportion
Yuichiro Kawabata a,b,⁎, Kazuo Yamada c
a
Port and Airport Research Institute, 3-1-1, Nagase, Yokosuka, Kanagawa 239-0826, Japan
b
Université Paris-Est, IFSTTAR, Marne-la-Vallée, 14-20 bd Newton, Champs sur Marne 77447, France
c
National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan

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

Article history: The inhibitory role of fly ash (FA) on alkali–silica reaction (ASR) expansion is drastically reduced at the pessimum
Received 31 May 2016 proportion relative to its role at full proportion. The main contribution of FA to inhibiting ASR expansion is to re-
Received in revised form 29 September 2016 duce the alkalinity of the pore solution and it is dependent on the thermodynamic equilibrium between the pore
Accepted 8 November 2016
solution and the calcium-silicate-hydrate (C-S-H) gel. C-S-H gel with low Ca/Si ratio can bind a large amount of
Available online xxxx
alkalis. However, when the concentration is higher than the threshold concentration for reaction, C-S-H gel plays
Keywords:
a role in buffering the alkali concentration in the pore solution by supplying alkali. Thus, the inhibitory role of FA
Alkali–silica reaction (C) on ASR expansion decreases at the pessimum proportion because the aggregate can react with alkali even at low
Fly ash (D) alkali concentrations. Therefore, replacing cement with a larger amount of fly ash is essential to reaching the
Pessimum threshold for reaction.
Alkalis (D) © 2016 Elsevier Ltd. All rights reserved.
Modeling (E)

1. Introduction
Extensive studies have revealed that replacement with a suffi-
cient amount of supplementary cementitious materials (SCM) such
as fly ash (FA), blast furnace slag, silica fume, and other natural poz-
zolans can inhibit the alkali–silica reaction (ASR) in mortar or con-
crete. The mechanism of the inhibitory role of FA is mainly
reduction of alkalinity of the pore solution [1–3], although this re-
mains controversial. The inhibitory role of FA on expansion due to
ASR depends not only on its pozzolanic reactivity but also on the re-
activity of the aggregate itself. When the concrete mix contains a
highly reactive aggregate, the level of FA required for replacement
becomes larger [1].
Highly reactive silica minerals such as opal, cristobalite, and
tridymite react with alkalis even at low concentrations. Aggregates
containing these silica minerals often show the pessimum effect [4].
This effect is generally explained in terms of the balance between
the alkali hydroxide concentration of a pore solution and the reac-
tivity of aggregates [5]. Aggregates showing the proportional
pessimum effect have high amounts of dissolved silica (Sc) and a
high tendency to reduce its alkalinity (Rc) in the chemical test spec-
ified by ASTM C 289 [6–8]; they are thus classified as “potentially
⁎ Corresponding author at: Port and Airport Research Institute, 3-1-1, Nagase,
Yokosuka, Kanagawa 239-0826, Japan.
E-mail address: kawabata-y@pari.go.jp (Y. Kawabata).
deleterious”. A high Rc in the chemical test indicates that such ag-
gregates can reduce the alkali hydroxide concentration of the pore
solution. Certain minerals such as clay in the altered aggregate can
also bind these alkalis via cation exchange and render them almost
unavailable for ASR.
A schematic illustration of the proportional pessimum effect is
shown in Fig. 1(a). When the proportion of the reactive aggregate
(Aag in Fig. 1) is larger than the pessimum proportion, the net
amount of alkali hydroxide per unit volume of reactive aggregate
(NAOH in Fig. 1) becomes insufficient for the ASR gel to exert an ex-
pansive pressure that exceeds the tensile strength of the concrete
or aggregate itself [9]. Conversely, when the proportion of the reac-
tive aggregate is smaller than the pessimum proportion, the expan-
sion decreases with the decreasing amount of reactive aggregate.
At the pessimum proportion, the amounts of the available alkali hy-
droxide and reactive aggregate reach an optimal balance such that
the largest expansion occurs.
In this context, the pessimum expansion behavior of mortar and/
or concrete is strongly influenced by the alkali hydroxide concentra-
tion of the pore solution. This influence implies that the pessimum
proportion depends not only on the aggregate characteristics but
also on the alkali hydroxide concentration of concrete. During reduc-
tion of the alkali content of the cement, the extent to which ASR
expansion is reduced varies with the amount of reactive aggregate
(Fig. 1(b)). In fact, cement with lower alkali content exhibits a shift
to a smaller pessimum proportion of the reactive aggregate
(Fig. 1(b)) [10]. When the alkali concentration is sufficiently high

http://dx.doi.org/10.1016/j.cemconres.2016.11.002
0008-8846/© 2016 Elsevier Ltd. All rights reserved.
2 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15

2. Experimental details

(a) 2.1. Materials

Ordinary Portland cement with an alkali content of 0.62 wt% was

(b)
Fig. 1. Schematic illustration of (a) pessimum effect and (b) the influence of reducing
alkalinity of the system on ASR expansion (Aag: reactive aggregate proportion (%), NAOH:
net amount of alkali hydroxide per unit volume of reactive aggregate).
compared to the binding capacity of the aggregate and/or ASR gel, no
pessimum effect is produced.
This discussion can be extended to the inhibitory effect of FA and
other SCMs. Since FA reduces the alkali hydroxide ion concentration
of the pore solution, replacement of cement with FA has the ability to
change the pessimum behavior of the mortar and concrete (Fig.
1(b)). However, little experimental data on this phenomenon has
been obtained [11]. The thermodynamic equilibrium of the alkalis
between the C-S-H gel and the pore solution is of importance when
the cement is replaced with FA; examining such a system may be fa-
cilitated by the use of numerical calculations that take into account
these interactions. Experiments and numerical simulations focusing
on these points are therefore of great importance when designing
the ideal mixture for inhibiting ASR expansion by using FA. The
goal of this study is to evaluate how the inhibitory effect of FA on
ASR expansion decreases at the pessimum proportion and to eluci-
date the mechanism behind it.
This paper describes the inhibitory role of FA in ASR expansion of
mortar, and especially the effect of FA at the pessimum proportion as
determined experimentally. Analysis using scanning electron microsco-
py with energy-dispersive spectrometry (SEM/EDS) of mortar, together
with numerical calculations, was also performed to understand the ex-
perimental results. From the results, the mechanism behind the reduced
effect of FA on the inhibition of ASR expansion at the pessimum propor-
tion is discussed.
used as the reference material. We used two types of FA (labeled R1
and H), both of which are specified as class II under the Japanese Indus-
trial Standard (JIS) A6201 [12]. Table 1 shows the bulk chemical compo-
sition as determined through X-ray fluorescence spectroscopy. Physical
properties such as density and Blaine specific surface area are also sum-
marized in Table 1. According to X-ray Diffraction (XRD)/Rietveld anal-
ysis [3], the glass contents of FA of R1 and H were 74.1% and 65.7%,
respectively.
Four types of reactive aggregates (labeled Ot1, Ot2, Ot3, and Ks)
were collected in Hokkaido, northern Japan. Micrographs of these ag-
gregates obtained through polarized microscopy are shown in Fig. 2.
Ot1, Ot2, and Ot3 are two-pyroxene andesite containing the reactive sil-
ica minerals cristobalite and tridymite that were obtained from different
sites in one mine. The major difference between the Ot aggregates is the
amount of cristobalite (Ot2 N Ot1 N Ot3). Ks is pit sand consisting of a va-
riety of rock types, including tuff particles containing opal, which is
highly reactive.
Results of the chemical test specified by ASTM C 289 are shown in
Table 2. The samples were judged as “potentially deleterious” or “dele-
terious”, meaning that these aggregates have the potential of showing a
proportional pessimum effect [4,6]. Pure limestone from Kyushu was
found to be a non-reactive aggregate. Andesite coarse aggregate from
the Ot mine caused on-site ASR damage in old local structures, but the
estimated total alkali contents of these concretes were significantly
higher than the limit currently regulated in Japan (3.0 kg/m3). Ks caused
ASR damage in several structures with a total alkali content much
b3.0 kg/m3. In the affected structures, Ks was used as a constituent of
sand; the non-reactive coarse aggregate was pure limestone. This mix
of aggregates is believed to have caused the pessimum effect [13].
2.2. Mixture proportions
Two series of experiments using different aggregates were carried
out. In the first series, the andesite aggregate from different locations
in one Ot mine was used to study the inhibitory effect of FA at the
pessimum proportion. Results were compared against those obtained
at full proportion of the reactive aggregate. In the second series, the ef-
fects of the amount of reactive aggregate and the level of FA replace-
ment were investigated using Ks.
2.2.1. Series 1
The objective of Series 1 is to evaluate the impact of the proportion of
reactive aggregate on the inhibitory effect of FA through the mortar-bar
test described in Section 2.3. The water-to-cement ratio and the sand-
to-cement ratio were set at 0.50 and 2.125, respectively. The alkali con-
tent of cement was boosted to 1.20 wt% of cement by addition of NaOH
solution to the mixing water. Non-reactive limestone sand was replaced
with reactive aggregate (Ot1, Ot2, and Ot3) at proportions of 30 and
100 wt% of the total sand content. The weight ratio of reactive aggregate
content to sand is represented as “Rw/Sw”. The final expansions of mor-
tar at 30 and 100 wt% of Rw/Sw are shown in Fig. 3. Pessimum propor-
tions of all aggregates were approximately 30 wt% [11]. At each Rw/Sw

Table 1
Chemical composition and physical properties of cement and FA.

Material Chemical composition (wt%) Density (g/cm3) Blaine SSA (cm2/g)

LOI SiO2 Al2O3 Fe2O3 CaO MgO Na2O K2O SO3 P2O5 Total

OPC 0.60 21.79 4.98 2.91 65.23 1.21 0.31 0.47 1.72 – 99.22 3.16 3280
FA R1 1.50 60.17 22.24 4.29 5.80 1.66 0.48 0.97 0.47 0.81 98.39 2.30 3910
H 1.93 56.16 26.03 4.82 5.17 1.20 0.55 1.15 0.59 0.43 98.03 2.29 2820
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15 3

(a) Ot1

(b) Ot2

(c) Ot3

(d) Ks

Fig. 2. Thin section micrographs of the aggregates under polarized microscope (left: open nicol, right: cross nicol).

ratio, cement was replaced with 20 vol.% of the FA, R1. The alkali content
of mortar with FA was also boosted, even though the amount of addi-
tional NaOH solution was the same as that for non-replaced mortar.
Table 2
Chemical test results on aggregates (ASTM C 289).
2.2.2. Series 2
Aggregate Sc (mmol/l) Rc (mmol/l) Sc/Rc Result Series 2 was aimed at clarifying through the mortar bar test the ef-
Ot1 545 134 4.1 Potentially deleterious fect of the level of FA replacement on the expansive behavior around
Ot2 801 107 7.5 Deleterious the pessimum proportion. The water-to-cement and sand-to-cement
Ot3 518 135 3.8 Potentially deleterious ratios were set at 0.50 and 2.25, respectively. The alkali content of mor-
Ks 473 200 2.4 Potentially deleterious
tar was boosted so that the alkali content of cement reached 1.20 wt% of
4 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
Fig. 3. Final expansion at 182 days (Rw/Sw: weight percentage of reactive aggregate).
cement by adding NaOH solution to the mixing water. We replaced non-
reactive limestone with the reactive aggregate Ks while keeping the
same volumetric sand content. The volume ratio of reactive aggregate
to sand is represented as “rv/sv”. For the control mix, rv/sv ranged from
5% to 50%. The relationship between rv/sv and the expansion of mortar
at 182 days is shown in Fig. 4. Ks exhibited pessimum at 30 vol.% of rv/
sv, and the expansion at 30 vol.% of rv/sv was twice that observed at
50 vol.%.
At each rv/sv ratio, cement was replaced with 10, 20, and 30 vol.% of
the FA (H), which correspond to 7.5, 15, and 24 wt%, respectively. The
alkali content of mortar with FA replacement was also boosted, even
though the amount of additional NaOH solution was the same as that
for mortar without replacement.
2.3. Mortar bar test
The basic testing protocol was performed in accordance with JIS
A1146. At 24 h after casting, specimens (40 × 40 × 160 mm) were
demolded and the initial length was measured immediately. The mortar
specimens were wrapped with cloth containing water. Thereafter, spec-
imens were stored at above 95% relative humidity (R.H.) at 40 °C in a
humidity chamber. The specimens were kept wet by the wet cloth;
the high humidity led to an internal R.H. close to 100%. The change in
length of the mortar specimen was measured using a dial gauge.
Alkali leaching is an issue of concern [14]. The wrapping procedure
may potentially contribute to the acceleration of alkali leaching at an
early stage [15]. Although a cloth-wrapped concrete specimen can still
be affected by alkali leaching, the amount of alkalis leached is limited
in comparison with that in concrete specimens without wrapping and
thus exposed to the humid atmosphere [16]. The lessening of leaching
is due to the resulting substantial reduction of the extent of evaporation
during cooling and measuring. Even though the amount of alkali
leaching was not measured, the comparison was still performed.
Fig. 4. rv/sv vs. expansion at 182 days (rv/sv: volume percentage of reactive aggregate).
2.4. SEM/EDS analysis
Polished thin sections were prepared from the mortar specimens for
polarized microscopy and SEM/EDS analysis. Each sample was then
coated by carbon to a thickness of 15 nm. Field-emission SEM equipped
with back-scattering electron detection and EDS (JEOL JSM-7001F and
JEOL SM-54060RBEI; INCA Energy, Oxford Instruments) was performed
for observation and elemental analysis. The conditions consisted of an
accelerating voltage of 15 keV, a probe current of 350 pA, a working dis-
tance of 10 mm, a dead time of approximately 30%, and an analysis time
of 100 s/point. For EDS, a magnification of 2000× was used and XPP cor-
rection [17] was carried out.
3. Results
3.1. Series 1
Expansion curves of mortars with or without 20 vol.% of FA at Rw/Sw
ratios of 30 and 100 wt% are shown in Fig. 5. In the case of 100 wt% of
Rw/Sw, Ot2 expanded rapidly, whereas a time lag preceded Ot3 expan-
sion. Ot1 showed a longer latency time prior to expansion, and its ex-
pansion rate was the smallest while the final expansion of Ot1 was
almost the same as that of Ot3. For mortars with Rw/Sw ratios of 30
and 100 wt% without FA, the times for the start of expansion and the
final expansion are different. All aggregates started expansion early in
mortars with an Rw/Sw ratio of 30 wt%; expansion was delayed in mor-
tar with an Rw/Sw ratio of 100 wt%, depending on the aggregate type.
The final expansion for mortar with an Rw/Sw ratio of 30 wt% (0.6–
0.7%) was larger than that for mortar with an Rw/Sw ratio of 100 wt%
(0.5–0.6%).
Replacement of cement with FA delayed the start of expansion and
lowered the rate of expansion at each Rw/Sw ratio. At an Rw/Sw ratio of
100 wt%, FA effectively suppressed the expansion for up to 182 days, ex-
cept in the case of Ot2. Ot2 might thus have a reactivity higher than that
of Ot1 and Ot3 because it contains a higher amount of cristobalite, and
its Sc/Rc ratio is substantially higher, as shown in Table 2. On the con-
trary, all mortars with an Rw/Sw ratio of 30 wt%, in which cement was
replaced with FA, showed the final expansions larger than 0.5%. Thus,
the effect of FA replacement on the final expansion varied with Rw/Sw.
In some specimens, ASR expansion did not reach a plateau over the
test period. Therefore, the experimental data were fitted with a mathe-
matical model using the least-squares method in order to compare the
final expansion [18–20]. The degree of ASR-induced expansion εχ can
be formulated as the product of the swelling potential ε∞ and time of
progress of the chemical reaction ξ(t):
ε χ ¼ ε ∞ ξðt Þ: ð1Þ
This kinetic function represents the degree of reaction, which
evolves from 0 (beginning of the reaction) to 1 (end of the reaction).
It may be described by two parameters, as in the following equation:
ξðt Þ ¼ f1− expð−t=τ C Þg=f1 þ exp½−ðt−τ L Þ=τ C g; ð2Þ
where t is elapsed time, τC is characteristic time, and τL is latency time
(for the schematic, see Fig. 6).
The fitted curves are shown in Fig. 5. In order to compare the effect of
FA replacement on each parameter, the parameters calibrated through
the model were normalized by dividing by the parameter of mortar
without FA replacement. The normalized parameters of FA-containing
mortar are summarized in Fig. 7. At Rw/Sw ratio of 100 wt%, FA reduced
the expansion to different extents (10% to 60% of mortar without FA),
depending on the type of aggregate (Fig. 7(a)). In particular, FA was ef-
fective against expansion due to Ot1 and Ot3, which showed similar
chemical properties as determined through ASTM C 289 (Sc/Rc ratio is
~ 4.0). Ot2 showed high Sc (Table 2), and the inhibitory effect of FA
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
Fig. 5. Expansive behavior of mortar at different Rw/Sw (Legend represents type of reactive aggregate-replacement level of FA).
was relatively small. In contrast, FA reduced the expansion to only 0.8
for every aggregate at an Rw/Sw ratio of 30 wt%, which is the pessimum
proportion. The inhibitory effect of FA was drastically reduced at the
pessimum proportion. The results therefore suggest that replacement
with higher levels of FA is necessary for inhibiting ASR expansion
when the aggregate used is at pessimum proportion.
The effect of FA replacement on the kinetics of ASR expansion was
also considered. Here, the comparison was performed on samples
showing asymptotic final expansion larger than 0.05%; otherwise; the
expansion was insufficient to obtain reasonable fitting due to small
expansion.
Replacement of cement with FA limited its effect on the characteris-
tic time (0.9–1.5 times that of mortar without replacement). No tenden-
cy for characteristic time was observed: most of the cases showed that
FA replacement slightly increased the characteristic time, that is, it de-
creased the reaction rate but it had limited impact on characteristic
time. In contrast, the latency time, that is, the time before expansion,
was increased by FA replacement (1.4–2.5 times that of mortar without
replacement). Replacement with FA reduced the alkalinity of the pore
solution, thus decreasing the dissolution rate of the silica minerals and
extending the latency time. This retarding effect by FA replacement
was enhanced at full proportion more than it was at pessimum propor-
tion (Ot2: 1.8 times; Ot3: 1.5 times).
3.2. Series 2
The expansive behaviors of mortars with and without replacement
with various levels of FA at rv/sv ratios of 5 to 50 vol.% are shown in
Fig. 8. The plot showing the expansive behavior was also fitted with
Eqs. (1) and (2). The expansion of mortar at an rv/sv ratio of 50 vol.%
reached a plateau after 100 days, while other cases exhibited
Fig. 6. Schematic representation of chemical advancement including physical meaning of
latency and characteristic times. After [16].
5
continuous expansion. This expansion suggests that the reaction almost
completely consumed the alkali at an rv/sv ratio of 50 vol.% because of
the abundance of reactive aggregates in the system.
Models describing the expansion characteristics of mortar are
displayed in Fig. 9. In all cases in Fig. 9(a), the replacement of cement
with higher levels of FA reduced further the ASR expansion, although
the replacement level of 10 vol.% was insufficient to reduce the asymp-
totic final expansion in every case. At rv/sv ratios of 20 and 30 vol.%, the
asymptotic final expansion exceeded 0.2% even with cement replace-
ment with 20 vol.% FA. At an rv/sv ratio of 20 wt%, replacement with
30 vol.% could not inhibit the expansion. The larger rv/sv ratio depicted
in Fig. 9(b) led to shorter characteristic time. The latency time was al-
most constant except at an rv/sv ratio of 5%.
Replacement with FA at levels of 20 and 30 vol.% clearly had inhibi-
tory effects that varied with the rv/sv ratio. Interestingly, the pessimum
proportion tended to shift to smaller aggregate proportions (20 vol.%)
from 30 vol.% upon replacement of 30 vol.% of cement with FA.
Each parameter was normalized by dividing it with the correspond-
ing parameter for mortar without replacement (Fig. 10). The level of FA
replacement and the normalized asymptotic final expansion clearly var-
ied with the rv/sv ratio, as described above: the closer the rv/sv ratio is to
the pessimum proportion (30 vol.%), the inhibitory effect becomes
smaller. The effect of FA replacement on the kinetics of ASR expansion
for mortars exceeding an expansion of 0.05% was also studied. The nor-
malized characteristic times ranged from 0.8 to 1.5, which is almost the
same as that for the Series 1 result. The normalized latency time linearly
increased with the increase in the level of FA replacement, irrespective
of the rv/sv ratio. An rv/sv ratio of 20 vol.% resulted in a considerable in-
crease in normalized latency time. At a FA replacement of 20 vol.%, the
normalized latency times ranged from 2.7 to 3.0. Overall, FA replace-
ment strongly affected the latency time but weakly affected the charac-
teristic time at each rv/sv ratio.
After the accelerated mortar bar test, the specimens were wrapped
with plastic film and stored at 20 °C. After 1 year of storage, all mortar
bars containing 20 vol.% of FA showed severe cracking on their surface,
suggestive of expansion after the test, although we did not measure the
expansion owing to defective measurement apparatus. These samples
were used for SEM/EDS, as described below.
3.3. SEM/EDS
SEM/EDS analysis was performed on the Series 2 samples, which had
been stored for 1 year after mortar bar tests. Backscattered-electron im-
ages (Fig. 11) show that gel-filled microcracks extended from the aggre-
gate to the paste, irrespective of the rv/sv ratio and the level of FA
replacement. According to the four grades of petrographic severity pro-
posed by Katayama et al. [21] based on polarized microscopy
6 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15

Fig. 7. Comparison of normalized parameters Legend represents weight ratio of the reactive aggregate Rw/Sw. Note that the parameters of FA-replaced mortar were normalized by dividing
by the parameter of mortar without FA replacement.

observation, the severity of ASR may be classified as “moderate (gel-fill-
ing cracks in cement paste)”, although some voids were filled with ASR
gel.
A thin layer around the FA particles can be observed in Fig. 12(a).
The particle may be composed of glass phase only. The FA particle
shown in Fig. 12(b) contains a needle-like crystalline phase, suggestive
of mullite crystals in an amorphous matrix [22] and has a scant reaction
layer around it.
EDS analysis was performed on the ASR gel inside the aggregate and/
or near the aggregate–paste interface, and on the inner calcium-silicate-
hydrate (C-S-H) gel around the unhydrated cement particle. Fig. 13
shows the compositional trends of the ASR gel and inner calcium-sili-
cate-hydrate (C-S-H) gel around the unhydrated cement particle in
terms of the relationship between the Ca/Si and Ca/(Na + K) atomic ra-
tios of the gel [21,23]. The trend for the ASR gel in the Ca/Si–Ca/(Na + K)
diagram is linear, encompassing data points for the high-alkali and low-
Ca ASR gel to those for high-Ca/Si ASR gel. The mean chemical composi-
tion (Table 3) shows that the mean Ca/Si and Ca/(Na + K) atomic ratios
of ASR gel in the control specimens were higher than that of ASR gel for
specimens with FA replacement. The slope of the line in Fig. 13 agrees
well with the analytical results obtained by Katayama et al. [21,23]. It
suggests that the compositional trend of the ASR gel is almost the
same even after replacement of cement with FA.
Even though the unhydrated cement was not analyzed in our study,
the compositional trends of the inner calcium-silicate-hydrate (C-S-H)
gel around the unhydrated cement particle approached that of the

Fig. 8. Expansion behavior of mortar at different rv/sv (legend represents volumetric ratio of the reactive aggregate “rv/sv”-replacement level of FA).
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15 7

Fig. 9. Expansion model characteristics vs. volumetric percentage of reactive aggregate ((a) legend represents replacement level of FA).

final products at Ca/Si atomic ratios of 1.5–2.0. A convergence point
where Ca/Si atomic ratio is within 1.5–1.7, Ca/(Na + K) atomic ratio is
within 100–200, and chemical equilibrium between the ASR gel and
inner calcium-silicate-hydrate (C-S-H) gel around the unhydrated ce-
ment particle is attained can be found [21,23]. Reaction products around
FA were also analyzed. Despite the scattering area of incident electrons
being larger than the thickness of reaction products around FA (b1 μm),
we were able to plot most of the analytical data in the Ca/Si atomic ratio
range of 0.5–1.5 and the Ca/(Na + K) atomic ratio range of 20–100. A
certain amount of alkali was taken up by the reaction products formed
around the FA particle.
Mean (Na + K)/Si atomic ratios of the inner C-S-H gel around the
unhydrated cement particle and FA reaction products are summarized
in Fig. 14. Replacement of FA markedly increased the amount of alkali
taken up by the inner C-S-H gel around the unhydrated cement particle
and the reaction products around FA. This increase may be caused by
the decrease in the Ca/Si atomic ratio of the C-S-H gel by FA [1,3]. It is
of interest that the analytical data for 20-20-FA (rv/sv ratio = 20%,
FA = 20%, reaction product around FA) in Fig. 13 lie between those
for low-Ca ASR gel and C-S-H gel, as shown by the compositional
trend line of the ASR gel. As it contains moderate amounts of Ca, it
shows composition similar to that of ASR gel with moderate Ca content.
Data for 50-20-FA (rv/sv ratio = 50%, FA = 20%, reaction product around
FA), on the other hand, show wide range of Ca/Si atomic ratio (0.57–
1.96) while the Ca/(Na + K) atomic ratio of 50-20-FA is twice that of
20-20-FA. This may be caused by the more amount formation of ASR
gel that adsorbed more alkali and the alkali concentration in C-S-H
and FA products are reduced.

Fig. 10. Normalized expansion model characteristics vs. FA replacement level (legend represents replacement level of FA).
8 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15

(a) rv/sv=20%-FA0% (b) rv/sv=50%-FA0%

(c) rv/sv=20%-FA20% (d) rv/sv=50%-FA20%

Fig. 11. Backscattered Electron Image of cracked aggregates with ASR gel (aggregate: Ks).

4. Numerical modeling
In order to simulate the experimental reduction of the inhibitory effect
of FA at the pessimum proportion and the shift of the pessimum propor-
tion upon replacement of cement with FA, a simplified numerical calcula-
tion was performed. It should be noted that the numerical calculation was
not intended to predict the expansive behavior of mortar quantitatively.
The focus of the numerical model is to simulate how ASR expansion of
mortar changes when cement is replaced with FA, taking into account
the thermodynamic equilibrium of alkalis between the C-S-H gel and
pore solution. The fundamental framework was developed by Furusawa
et al. [5,24,25] on the basis of surface reaction. Some parts of their work
have been applied to other predictive models [26–30].
The previous models have been applied mainly to the Portland ce-
ment system without SCMs. In order to extend the model to the cement
system with SCM replacement, we should take into account the thermo-
dynamic equilibrium between the C-S-H gel and pore solution, because
replacement of cement with SCMs leads to the formation of a C-S-H gel
with a lower Ca/Si ratio and to reduction of the alkalinity [3,31–33].
Therefore, the model by Furusawa et al. was integrated with an alkalin-
ity model developed in a recent work [3].
Many previous models assume that ASR expansion takes place at the
rim, similar to a surface reaction [28,34,35]. There have been arguments
against analysis based on such an assumption, in that it does not realis-
tically reflect the reaction [36–38]. However, the objective of this nu-
merical simulation is to validate the experimental facts shown in the

(a) Glass particle (b) Mullite-deposited particle

Fig. 12. Backscattered Electron Image of FA particle (aggregate: Ks).

 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15 9

Fig. 14. Mean (Na + K)/Si atomic ratio of inner C-S-H gel around unhydrated cement
particle and reaction products around FA.

Fig. 13. Compositional trends of ASR gel and reaction products around FA (legend
represents sv/rv-FA replacement level-analytical point. Analysis was performed on ASR
where CSc is the dissolved silica concentration (ppm), as determined
gel, C-S-H gel and reaction product around FA).
through chemical test.
Eq. (5) may be transformed to the following:
former section, especially focusing on alkali mass balance in the system.
n o
In this context, showing how ASR expansion of mortar changes when x ¼ Ri 1−ð1−α i Þ1=3 : ð7Þ
cement is replaced with FA may be sufficient.

4.1. Reaction model The thickness of the reaction layer can be calculated from the chem-
ical test at different durations and/or concentrations of the alkaline so-
pffiffiffiffiffi
Following the work of Furusawa et al. [5,24,25], we considered the lution. The relationship between Ct and x thus obtained gives k.
model in which the reaction develops one dimensionally from the sur- After calculation of the reaction ratio of each particle of specific size,
face of the aggregate toward its interior. Assuming that the profile of al- the total amount of ASR gel may be obtained as:
kali within the reaction layer is linear, we can express the reaction ratio X
as follows: T pr ¼ Aag α i  βi =60:08; ð8Þ


dx=dt ¼ C cp −C th k=x; ð3Þ where Tpr is the total amount of ASR gel in the mortar (mol/m3-mortar),
Aag is the total reactive aggregate content (g/m3-mortar), and βi is the
where t is duration of the reaction (h), x is thickness of the reaction layer ratio of Ri to the total aggregate.
(cm), k is the reaction rate constant (cm2/h), Ccp is the concentration of Part of the ASR gel fills the porous volume inside or surrounds the
alkali hydroxide in the paste (mol/l), and Cth is the threshold concentra- aggregate without leading to expansion. Derivation of Ter (mol/m3-mor-
tion of alkali hydroxide (mol/l). tar) by Furusawa et al. [24] leads to:
The initial condition in which x is zero when t is zero gives: 8 9
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi X < h0  4πR2 =
x ¼ 2 k∙t∙ðCcp−C th Þ: ð4Þ T er ¼ Aag βi  i  ð9Þ
: 4πR3 =3 ;
i

Assuming that the aggregate particle is spherical and that the parti-
cle is in contact with the pore solution, we can formulate the reaction Replacing 3h′ with h gives:
ratio of the aggregate i with radius Ri, like so: X
T er ¼ Aag βi h=Ri ð10Þ
α i ¼ 1−ð1–x=Ri Þ3 ; ð5Þ
where h (=3h′) is a constant for the reduction of the effective ASR gel
where Ri is the radius of the particle (cm) and αi is the reaction ratio of expansion (mol cm/g).
the particle with radius Ri. According to Eqs. (9) and (10), Ter is a linear function of the surface
Eq. (5) is valid when the thickness of the reaction layer is sufficiently area per unit volume of the reactive aggregate. A larger h implies higher
small compared with Ri. porosity around the reactive aggregate. Thus, a larger h or smaller Ri cor-
The reaction ratio of the particle may also be described by: responds to a larger amount of ASR gel without expansion. Therefore,
the pessimum size effect can also be explained by this model [24]. A
α ¼ C Sc =106 ; ð6Þ similar approach was also applied by Multon et al., although the thick-
ness of porous zone around the aggregate is constant, that is, indepen-
dent of particle size [28].
Table 3 Expansion can occur after filling of pores surrounding the reactive
Mean chemical composition of ASR gel.
aggregate. Assuming that the expansion is linearly related to the reac-
Ca/Si Ca/(Na + K) tion ratio of the aggregate without consideration of the elastic and vis-
rv/sv50-FA0 0.95 87.4
coelastic properties of the ASR gel and cement paste, we can describe
rv/sv50-FA20 0.51 3.2 this process as follows:
rv/sv20-FA0 0.83 20.0
 þ
rv/sv20-FA20 0.50 5.7 ε ¼ E T pr −T er ; ð11Þ
10 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15

where ε is expansion of mortar (%), E is a constant for the conversion of

the amount of ASR gel to the expansion of the mortar (%∙ m3-mortar/
mol), and bXN+ is the positive part of X.

4.2. Alkalinity model

Alkali metal ions are generally balanced with hydroxide ions at 1:1
ratio in the pore solution. The alkalinity of the pore solution therefore
represents the alkali hydroxide concentration in the pore solution
such that modeling the alkalinity of the solution is sufficient in this
study. By considering the closed system without any supply or leaching
of alkali, the total amount of alkali (Naeq) per volume unit of mortar,
Cmortar, can be expressed by:

C mortar ¼ C cp Acp þ C ag Aag þ γNa=Si T pr ; ð12Þ

where Ccp is the total alkali content per unit volume of cement paste
(mol/m3-paste), Acp is the cement paste content in the mortar (m3-
paste/m3-mortar), Cag is the alkali content instantaneously consumed
by the aggregate (mol/g-aggregate), Aag is the total reactive aggregate
content in the mortar (g-aggregate/m3-mortar), and γNa/Si is the
(Na + K)/Si molar ratio of the ASR gel.
By considering the equilibrium between the solid C-S-H gel and pore
solution in the cement paste system, Ccp may be written as follows:

C cp ¼ Rs C CSH þ Rl C fw ; ð13Þ

where CCSH is the amount of C-S-H per unit volume of cement paste
(g/m3-paste), Cfw is the free water per unit volume of cement paste
(ml/m3-paste), Rs is the alkali content in the solid C-S-H (mol/g),
and Rl is the alkali concentration in solution (mol/ml). The distribu-
tion ratio, Rd (ml/g), may be described by:
Rd: ¼ Rs =Rl ð14Þ
and
Rd: ¼ γ ðCa=SiÞδ ; ð15Þ
where Ca/Si is Ca/Si molar ratio, γ and δ are experimental constants,
the calculated values of which are 2.5 and − 3.1, respectively [3,31].
It should be noted that alkali recycling has not been considered [26,
39] because the kinetics for Ca exchange of ASR gel have not yet been
modeled; such a procedure depends on many factors such as the fluidity
of the gel and the path of gel movement through connected pores or
cracks [39–41]. Alkalis resulting from the aggregate have not been
taken into account for the sake of simplicity. The experimental results
suggest that the alkalis originating from FA exist in the glass phase [3].
Assuming that the reaction ratio of FA is 60%, we found the total alkali
content to be nearly the same as it would have been without the contri-
bution from FA. Therefore, alkalis originating from FA can be neglected.
Fig. 15. Schematic flow for calculation.
effective amount of ASR gel and resultant expansion of mortar were cal-
culated. After the amount of alkali consumed by the ASR gel was
subtracted from the total amount of alkali, the amount of alkali between
the C-S-H gel and pore solution at phase equilibrium was re-calculated.
When the concentration of alkali in the pore solution reached the
threshold (bCth), the calculation was terminated.
The models described above were integrated in a calculation system
set up using MATLAB software. The calculation was looped over a period
of up to 200 days. The time step was 0.1 h up to 1 day after the test; this
was then increased to 1 h until 200 days.
4.4. Parameter identification
The reaction rate constant k for the reaction model was calibrated
according to the chemical test on the Ks aggregate. Chemical tests at dif-
ferent concentrations of alkaline solution and duration were performed.
pffiffiffiffiffi
The slope of the relation between Ct and x (Fig. 16), which is equal to
2 k, may be used to derive the reaction rate constant. Dissolution, at
pffiffiffiffiffi
which point Ct = 1.9, was preceded by a time lag. This time lag

4.3. Algorithm for calculation

The Series 2 experiments were used in the calculation. The algorithm

for the calculation is illustrated in Fig. 15. The time was set to zero when
the accelerated mortar bar tests were commenced. At this point, the re-
action ratio of the aggregate was taken to be zero. Thus, the initial alka-
linity was calculated by considering the reduction in alkalinity due to
alkali consumption by the aggregate and to equilibrium between the
C-S-H gel and pore solution. The amount of alkali consumed by the ag-
gregate was kept constant after time equal to zero. Then the alkali hy-
droxide concentration was assumed to be constant during the next
time step. The reaction ratios for each particle size within the time
step were calculated, and then the amounts of ASR gel and consumed al-
kali hydroxide were calculated. The amount of ASR gel that did not con-
tribute to expansion (Ter), was also calculated in parallel. Therefore, the Fig. 16. Chemical test results for thickness calculation.
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
corresponds to b3.8 h and 12.6 h, respectively, for tests on 1.0 mol/l and
0.3 mol/l solutions. The solutions, in turn, correspond to the initial pore
solution of mortar without FA replacement and mortar with FA replace-
ment at 30%, respectively (Table 4, described below). Because the influ-
ence of this time lag observed in this experiment was small, it was
assumed to be negligible in the long-term calculation. As a result, the
calibrated reaction rate constant is 2.6 × 10− 9 (cm2/h). It should be
noted that the reaction rate constant can be affected by temperature;
however, the temperature dependency of the constant was not investi-
gated because of an insufficient amount of aggregate for the test.
The particle size distribution and the total reactive aggregate con-
tent, βi and Aag, were determined to be correct according to the experi-
ments. There are three unknown parameters, namely, γNa/Si, E, and h.
Because the composition of the primary ASR gel is similar to that of
mountainite or shlykovite [39,42], the γNa/Si value in terms of the
(Na + K)/Si molar ratio lies between 3/8 and 1/4. However, the alkali
of the ASR gel can be recycled by exchange with Ca in the cement
paste, as described above. Therefore, these parameters were calibrated
to fit the calculated results to the experimental results for mortar with-
out replacement. The values for γNa/Si, E, and h after calibration are 1/5,
1.0 × 10−3 (%∙m3-mortar/mol), and 3.8 × 10−2 (mol∙cm/g∙m3-mortar),
respectively. The E and h values after calibration, as reported by
Furusawa et al., are 2.35 × 10−3 and 9.9 × 10−2, respectively [24]. Be-
cause these parameters depend on the characteristics of the reactive ag-
gregate, the parameters determined in the present study seem
reasonable.
By assuming that hydration of cement is completed and by summing
the amounts of alkalis from the cement and sodium hydroxide solution,
the value of Cmortar for the alkalinity model after calibration was found to
be 2.2 × 102 mol/m3-mortar. The degree of reduction in alkalinity was
measured according to the chemical test described above. The relation-
ship between the amount of dissolved silica and the degree of reduction
in alkalinity is shown in Fig. 17. From the figure, the amount of alkali
consumed by the aggregate can be determined by calculating the degree
of reduction in alkalinity when the amount of dissolved silica is zero. As
a result, Cag was 6.3 × 10−5 mol/g. It should be noted that Na/Si molar
ratio of reaction product calculated from the chemical test was 2/7
and was slightly higher than the value identified by the fitting (γNa/
Si = 1/5). In the presence of calcium, Na/Si molar ratio is reduced by al-
kali recycling, as shown in Fig. 13. Therefore the value adopted in the
calculation seems reasonable.
Data from the previous study were used as the parameters for the al-
kalinity model (Ca/Si molar ratio, CCSH and Cfw, Table 4; hereafter, the
“Type A” calculation) [3]. Parameters for the alkalinity model were de-
termined by phase-composition analysis over a period of 91 days.
Since the FA used in this study was similar to that in a previous work
[3], the hydration products in our experiment may be similar to those
in the previous study.
To determine the importance of the thermodynamic equilibrium be-
tween the C-S-H gel and pore solution, a complementary calculation
(hereafter, the “Type B” calculation) was performed. In the Type B calcu-
lation, the binding capacity is neglected while the initial total alkali con-
tent of mortar is modified to arrive at the same alkali concentration in
the pore solution (initial Rl in Table 4). Therefore, the alkali-binding ca-
pacity of the C-S-H gel due to a decrease in Ca/Si ratio was not
Table 4
Parameters for alkalinity model (Type A).
Ca/Si CCSH Cfw
(g/m3-mortar) (ml/m3-mortar)
FA 0% 2.21 3.75 × 105 1.30 × 105
FA 10% 1.85 3.76 × 105 1.35 × 105
FA 20% 1.48 3.28 × 105 1.35 × 105
FA 30% 1.18 3.03 × 105 1.28 × 105
a
The alkali fraction consumed by the aggregate (CagAag) is not included in the
calculation.
11
Fig. 17. Dissolved silica vs. reduction in alkalinity.
considered in the calculation. The Cth value obtained with both the
type A and B calculations is 0.25 mol/l, adopted from Diamond [43].
For the sake of simplicity, the phase compositions were assumed to
be constant over the 200-day period (Table 4). The free-water content
was also assumed to be constant over the entire period; no water up-
take by the ASR gel was considered. The calculation errors are larger
in the implementation of the hydration models. The calculation was
thus focused on establishing the expansive behavior when cement is re-
placed with FA at the pessimum proportion. Thus, establishing the trend
through this simplified approach was deemed sufficient.
4.5. Calculated results
The simulated expansive behaviors of mortar are shown in Fig. 18.
The calculated result for mortar without replacement is calibrated.
However, the calibration was insufficient to allow simulation of the ex-
pansive behavior at an rv/sv ratio of 5%, despite the accurate simulation
of the trend. The expansive behavior of the other mortar samples, how-
ever, is consistent with the experimental results. The simulated expan-
sion was sustained, while the experimental results almost reached a
plateau. This difference in trends is attributable to alkali leaching from
the specimens. In this calculation, alkali leaching is not considered;
hence, the simulated expansion over long-term periods was sustained.
On the other hand, the mortar bars in the experiments were subjected
to alkali leaching [14–16].
Upon replacement of cement with FA, the expansion decreased, es-
pecially at rv/sv = 50%, with the increase in the level of FA replacement.
When the level of FA replacement was increased beyond 20%, no expan-
sion could be observed at rv/sv = 50%, whereas other cases showed
expansion.
The expansions of mortar at around 200 days (experimental: 197–
204 days; calculated: 200 days) at different rv/sv ratios and levels of
FA replacement are summarized in Fig. 19. The pessimum proportion
clearly shifted to smaller values upon replacement of the cement with
FA. The simulated results are generally consistent with the experimental
observation. According to the simulation, replacement of 20 vol.% of ce-
ment with FA resulted in a reduction of the expansion to zero at rv/sv =
50%; the expansion reached 52% at rv/sv = 20%. The inhibitory effect of
FA drastically decreased with the decrease in the rv/sv ratio.
The calculated results strongly support experimental evidence of re-
duction of the inhibitory effect of FA at the pessimum proportion and
the shifting of the pessimum proportion upon replacement of cement
Initial Rla
(mol/l) with FA.
Fig. 20 shows the influence of the thermodynamic equilibrium be-
1.06
0.80 tween the C-S-H gel and pore solution on the normalized expansion
0.58 (εFA/ε0), which was obtained by dividing it by the expansion of mortar
0.37 without replacement. Plots from experiments (Fig. 20) clearly show
that the inhibitory effect of replacing cement with FA was
overestimated in the Type B calculation. This point is especially
12 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15

Fig. 18. Simulated results of Series 2 (plot: experiments, line: simulation, Type A).

important in considering the minimum level of FA replacement for
inhibiting ASR expansion at the pessimum proportion.
5. Discussion
Some previous studies posit that the inhibitory role of FA in ASR ex-
pansion is in increasing the Ca content of the ASR gel, which results in a
reduction of the expansive potential of the ASR gel [44,45]. The analyt-
ical results, however, suggest that the composition of the primary ASR
gel specimens (i.e., gel before Ca exchange) with FA replacement is sim-
ilar to that in mortar specimens without replacement. The chemical
composition of the ASR gel strongly depends on the point of EDS analy-
sis as shown in Fig. 13, but ASR gels containing high or moderate
amounts of Ca are not the cause of expansion but the result of alteration
due to cation exchange with Ca in cement paste [21,23,39,42]. On the
contrary, replacement with FA markedly increased the amount of alkali
uptake in the C-S-H gel around the unhydrated cement particle and FA
products (Fig. 14). It is reasonable, therefore, to conclude that the main
contribution of FA in inhibiting ASR expansion is in reducing the alkalin-
ity of the pore solution via alkali uptake by a C-S-H gel with a low Ca/Si
ratio [1,3]. It has been claimed that aluminium may have a significant
impact on ASR expansion [46,47]. The effect of aluminium on ASR

Fig. 20. Influence of consideration of the interaction between C-S-H gel and pore solution
Fig. 19. Expansion of Series 2 as a function of rv/sv (plot: experiments, line: simulation, (plot: experiments, solid line: simulation with Type A (with considering the interaction),
Type A). dashed line: simulation with Type B (without considering the interaction)).
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
expansion was not clear in this study so that it will be investigated in fu-
ture research.
The main results highlight that the higher level of FA replacement is
required around the pessimum proportion. This important point was
evidenced by the experiments and numerical calculations. The
pessimum effect is controlled by a balance between the alkali hydroxide
concentration of the pore solution and the ratio of reactive aggregates.
The alkalinity, which represents the alkali hydroxide concentration of
the pore solution, may be determined from the equilibrium between
the pore solution and the C-S-H gel [3]. Upon replacement of cement
with FA, the Ca/Si ratio of the C-S-H gel decreases and the alkalinity of
the pore solution decreases. Consequently, replacement with FA chang-
es the pessimum behavior of the mortar/concrete. When the aggregate
proportion is larger than the pessimum proportion, the degree of reduc-
tion in alkalinity due to FA replacement strongly affects the reduction of
expansion of the mortar. Therefore, replacing cement with FA is a very
effective approach to prevent ASR expansion at the full proportion of re-
active aggregate. When the aggregate proportion is near the pessimum
proportion, the net amount of alkali hydroxide per unit volume of reac-
tive aggregate is optimized, thus decreasing the inhibitory effect of FA
replacement on ASR expansion. The test results and numerical calcula-
tions indicate that the pessimum proportion shifts to smaller values
upon replacement of cement with FA. Similarly, cement with a lower al-
kali content causes a shift of the pessimum proportion of the reactive
aggregate to smaller values, a trend that is well known [10]. Because
the balance between the alkalinity and reactive aggregate content con-
trols the pessimum effect, shifting of the pessimum proportion to small-
er amounts of the reactive aggregate is reasonable.
These results are applicable to other SCMs such as blast furnace slag
and silica fume, because the main contribution of these SCMs to ASR in-
hibition is in the reduction of the alkalinity of the pore solution [48–50].
Therefore, a larger amount of cement should be replaced with SCM
when highly reactive aggregate showing a significant pessimum effect
is used. Special precautions should also be taken when the pessimum
proportion is shifted to smaller values in replacing cement with SCM.
The thermodynamic equilibrium between the C-S-H gel and pore so-
lution is also an important consideration in calculations for mortar with
FA replacement. C-S-H gel with a low Ca/Si ratio can be a reservoir of al-
kali. However, when alkali in the pore solution can become available for
reaction above threshold (NCth), reaction can proceed and hence alkali
hydroxide concentration decreases. Because of the consumption of alka-
li due to the reaction, the equilibrium between C-S-H gel and pore solu-
tion changes, and the C-S-H gel supplies alkalis to the system in order to
maintain equilibrium. Consequently, the decrease in Ca/Si ratio of the C-
S-H gel buffers the alkali concentration in the pore solution by supplying
alkali, which is originally bound in the C-S-H gel. This alkali-buffering ef-
fect of the C-S-H gel is inextricably linked to the capability of the C-S-H
gel for alkali sorption, being enhanced up to 1.3, 1.6, and 2.0 times that
of the control specimen at FA replacement levels of 10, 20, and 30 vol.%,
respectively.
Changes in alkali concentrations of the pore solution and the expan-
sion at rv/sv = 30% at different levels of FA replacement under different
assumptions (the Type A and B calculations) are depicted in Fig. 21.
Here, the expansion behaviors of the alkali-boosted mortars with differ-
ent levels of FA replacement were simulated. The Cth value was set to
0.25 mol/l. The temporal trend of reduction in alkalinity is very similar
to the experimental results of Kim et al. [51]. Type B calculations indi-
cate that the reaction led to a significant reduction of alkali concentra-
tion in the pore solution, resulting in termination of ASR expansion.
Type A calculations, on the other hand, suggest a smaller reduction of al-
kali concentration and larger expansion than those obtained through
Type B calculations. This difference is due to the alkali-buffering effect
of the C-S-H gel. This tendency is more prominent when cement is re-
placed with FA, because this buffering effect is strongly influenced by
the amount of alkali bound in the C-S-H gel, depending on its Ca/Si
ratio. The significant reduction of alkali concentration in Type B
13
Fig. 21. Change in alkali concentration and expansion with time (solid line: alkali with
Type A (with considering the interaction), dashed line: alkali with Type B (without
considering the interaction), filled circle: expansion with Type A, unfilled circle:
expansion with Type B).
calculations led to an inhibitory effect that is stronger than that predict-
ed by Type A calculations. For instance, the alkali concentration in mor-
tar without replacement under Type A calculation is, at a maximum, 1.6
times that obtained through Type B calculations. The type-A-to-type-B
ratio of the degree of expansion at 200 days is 2.4. The type-A-to-
type-B ratio of alkali concentration upon replacement of 10% of cement
with FA is 1.6, whereas the ratio of the degree of expansion at 200 days
is 6.4. At an FA replacement of 20%, the expansion under Type A calcu-
lations was 0.2%, while type B indicated no expansion. Type B calcula-
tions indicate that at 20% replacement with FA, the alkali
concentration reached the threshold before expansion. Therefore,
neglecting the thermodynamic equilibrium between alkali in the C-S-
H gel and pore solution results in an overestimate of the inhibitory effect
of FA (Figs. 20 and 21).
The experiments revealed that replacement of cement with FA has a
strong retarding effect on the latency time but has little influence on the
characteristic time. The experimental evidence may be explained by
considering the thermodynamic equilibrium between the C-S-H gel
and pore solution. Latency time is strongly influenced by the alkalinity
of the pore solution when aggregates with the same potential for expan-
sion are used. When cement is replaced with FA, the alkalinity of the
pore solution can be reduced. Since the reaction kinetics are governed
by the alkali concentration of the pore solution and the duration of the
reaction (Eq. (4)), the time prior to expansion is retarded because of
the reduced alkalinity of the pore solution (Figs. 7(c), 10(c)). According
to the numerical simulation, the commencement time of the expansion
corresponds to replacement of cement with FA, as shown in Fig. 21.
Once reaction occurs, the reaction can proceed until the alkali concen-
tration of the pore solution is below Cth. Even in the system with FA re-
placement, a large amount of alkali bound in the C-S-H gel is also
supplied to the pore solution because of weak electrostatic interactions
between the C-S-H gel and pore solution. Consequently, the characteris-
tic time is influenced less by FA replacement. The Type B calculation,
which neglects the thermodynamic equilibrium between the C-S-H
gel and pore solution, shows rapid termination of expansion, owing to
a lack of available alkali (Fig. 21). The Type A calculation, on the other
hand, produced a sustained expansion, owing to the buffering effect of
the alkali. Therefore, the latency time for mortar after replacement of ce-
ment with FA was likely to be similar to that of mortar without replace-
ment (Fig. 21). The kinetics of mortar with FA replacement may be fully
explained by considering the thermodynamic equilibrium between the
C-S-H gel and pore solution.
These calculations thus suggest that implementation of the alkalinity
model is essential when the phase equilibrium is considered, especially
for mortar or concrete with SCM replacement. We emphasize the con-
siderable risk of overestimating the inhibitory effect of SCMs in identify-
ing the minimum level of SCM replacement according to the expansion
14 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
test without considering the aforementioned aspects. Again, a larger
amount of cement should be replaced with SCMs when highly reactive
aggregate at the pessimum proportion is used.
6. Conclusions
This paper presented experimental results documenting the role of
fly ash (FA) as an inhibitor of the expansion of mortar at the pessimum
proportion due to the alkali–silica reaction (ASR). The conclusions can
be summarized as follows:
(1) The inhibitory effect of FA on ASR expansion was drastically re-
duced at the pessimum proportion. The pessimum proportion
shifts to smaller values upon replacement of the cement with
FA. Therefore, a larger amount of cement should be replaced
with FA when a highly reactive aggregate exhibiting a strong
pessimum effect is mixed with non-reactive aggregates.
(2) Replacing cement with FA strongly increases the latency time for
ASR expansion but has little influence on the characteristic time;
these effects, however, depend on the type of reactive aggregate.
The influence of FA replacement on the kinetics depends on the
thermodynamic equilibrium between the C-S-H gel and pore so-
lution.
(3) Elemental analysis using scanning electron microscopy with en-
ergy-dispersive spectrometry revealed that the compositional
trend of the ASR gel could be plotted along one line in the Ca/
Si–Ca/(Na + K) diagram, irrespective of the use of FA. This result
indicates that the composition of the primary ASR gel in mortar
with FA replacement is similar to that of the primary ASR gel in
a control mortar without FA replacement.
(4) Comparison of the mean (Na + K)/Si atomic ratios of the inner C-
S-H gel around the unhydrated cement particle and the FA reac-
tion products showed that replacement with FA increases the
amount of alkali in the inner C-S-H gel around the unhydrated
cement particle and FA reaction products. Therefore, the inhibi-
tory mechanism of FA can be mainly attributed to a reduction
in alkalinity, resulting uptake of alkali by a C-S-H gel with
lower Ca/Si ratio.
(5) A simplified numerical simulation in which the reaction model
and alkalinity model are integrated was proposed. The surface
reaction models based on the chemical test and alkalinity
model, which consider the electrostatic interaction between al-
kali in the C-S-H gel and pore solution, were integrated in the
system. The calculation gave results that were consistent with
the experiments. The results of the numerical simulation strong-
ly supported the experimental evidence for the reduction of the
inhibitory effect of FA at the pessimum proportion and the shift
of the pessimum proportion upon replacement of cement with
FA.
(6) The thermodynamic equilibrium between the pore solution and
C-S-H gel must be considered in the calculation of inhibitory
role of FA, as C-S-H gel with a lower Ca/Si ratio plays a role in
buffering the alkali concentration of the pore solution when its
concentration is higher than the threshold for ASR (e.g.,
0.25 mol/l). In order to inhibit ASR expansion, therefore, replac-
ing cement with a larger amount of FA (or other supplementary
cementitious materials) is essential to reaching the threshold
level for reaction.
(7) The effect of FA replacement on ASR kinetics was validated
through numerical simulation. Since the reaction kinetics are
strongly governed by the alkali concentration in the pore solu-
tion and by the duration, the latency time is retarded as a result
of the reduced alkalinity of the pore solution when cement is re-
placed with FA. Once reaction occurs, however, the reaction can
proceed until the alkali concentration of the pore solution is
below threshold concentration of alkali. The reaction can also
be sustained by the alkali-buffering effect of the C-S-H gel. Con-
sequently, the characteristic time is less influenced by FA re-
placement. The effect of FA replacement on ASR kinetics is in
accordance with the results of numerical simulations that con-
sider the thermodynamic equilibrium between alkali in the C-
S-H gel and pore solution.
Acknowledgments
Part of this work was financially supported by the Japan Society for
the Promotion of Science (JSPS, No. 16H04393) and the JSPS Postdoctor-
al Fellowship for Research Abroad (H26-719). We also would like to
thank Dr. Shin-ichi Hirono (Taiheiyo Consultant Co., Ltd.) for performing
polarized microscopy observations and Dr. François Toutlemonde
(Université Paris-Est, IFSTTAR) for insightful discussions.
References
[1] M. Thomas, The effect of supplementary cementing materials on alkali-silica reac-
tion: a review, Cem. Concr. Res. 41 (2011) 1224–1231.
[2] F. Rajabipour, E. Giannini, C. Dunant, J.H. Ideker, M. Thomas, Alkali–silica reaction:
current understanding of the reaction mechanisms and the knowledge gaps, Cem.
Concr. Res. 76 (2015) 130–146.
[3] Y. Kawabata, K. Yamada, Evaluation of alkalinity of pore solution based on the phase
composition of cement hydrates with supplementary cementitious materials and its
relation to suppressing ASR expansion, J. Adv. Con. Tech. 13 (2015) 538–553.
[4] T. Katayama, Petrography of alkali-aggregate reactions in concrete – reactive min-
erals and reaction products, Proceedings of East Asia Alkali-Aggregate Reaction
Seminar 1997, pp. A45–A59 Supplementary papers.
[5] Y. Furusawa, H. Ohga, T. Uomoto, An analytical study concerning prediction of con-
crete expansion due to Alkali-Silica Reaction, Proceedings of 3rd CANMET/ACI Inter-
national Conference on Durability of Concrete 1994, pp. 757–779.
[6] ASTM C 289-07, Standard test method for potential alkali-silica reactivity of aggre-
gates (Chemical method), American Society for Testing Materials, Annual Book of
ASTM Standards, 04.02 2007, pp. 179–185.
[7] R.C. Mielenz, K.T. Greene, E.J. Benton, Chemical test for the reactivity of aggregates
with cement alkalis: chemical processes in cement-aggregate reaction, ACI Journal
Proceedings 44 (1948) 193–221.
[8] T. Katayama, T.S. Helgason, H. Olfasson, Petrography of alkali-reactivity of some vol-
canic aggregates from Iceland, Proceedings of 10th International Conference on Al-
kali-Aggregate Reaction in Concrete 1996, pp. 377–384.
[9] T. Ichikawa, Alkali-silica reaction, pessimum effects and pozzolanic effect, Cem.
Concr. Res. 39 (2009) 716–726.
[10] D.W. Hobbs, Alkali-Silica Reaction in Concrete, Thomas Telford, London, 1988.
[11] Y. Kawabata, T. Ikeda, K. Yamada, Y. Sagawa, Suppression effect of fly ash on ASR ex-
pansion of mortar/concrete at the pessimum proportion, Proceedings of 14th Inter-
national Conference on Alkali-Aggregate Reaction in Concrete, 2012 031711-
KAWA-01.
[12] JIS A 6201, Fly ash for use in concrete, 2015
[13] K. Yamada, S. Hirono, T. Miyagawa, New findings of ASR degradation in Japan, Pro-
ceedings of 14th International Congress on the Chemistry of Cement, 2011 No. 589,
CD-R 7 pages.
[14] P. Rivard, M.A. Bérubé, J.P. Ollivier, G. Ballivy, Alkali mass balance during the accel-
erated concrete prism test for alkali-aggregate reactivity, Cem. Concr. Res. 33
(2003) 1147–1153.
[15] J. Lindgård, M.D.A. Thomas, E.J. Sellevold, B. Pedersen, Ö. Andiç-Çakır, H. Justnes, T.F.
Rønning, Alkali–silica reaction (ASR) - performance testing: influence of specimen
pre-treatment, exposure conditions and prism size on alkali leaching and prism ex-
pansion, Cem. Concr. Res. 53 (2013) 68–90.
[16] K. Yamada, S. Karasuda, S. Ogawa, Y. Sagawa, M. Osako, H. Hamada, M. Isneini, CPT
as an evaluation method of concrete mixture for ASR expansion, Constr. Build.
Mater. 64 (2014) 184–191.
[17] J.-L. Pouchou, F. Pichoir, in: D. Newbury (Ed.), Microbeam Analysis 1988, Dan
Francisco Press 1988, p. 315.
[18] C. Larive, Apports Combinés de l'Expérimentation et de la Modélisation à la Compré-
hension de l'Alcali-Réaction et de ses Effets Mécaniques, Laboratoire Central des
Ponts et Chaussées, 1998 (in French).
[19] S. Multon, Evaluation expérimentale et théorique des effets mécaniques de l'alcali-
réaction sur des structures modèlesPh.D. Thesis Université de Marne la Vallée,
2003 (in French).
[20] S. Poyet, Etude de la dégradation des ouvrages en béton atteints par la réaction
alcali-silice: approche expérimentale et modélisation numérique multi-échelles
des dégradations dans un environnement hydro-chemo-mécanique variablePhD.
Thesis Université de Marne la Vallée, 2003 (in French).
[21] T. Katayama, T. Oshiro, Y. Sarai, K. Zaha, T. Yamato, Late-expansive ASR due to
imported sand and local aggregates in Okinawa island, southwestern Japan, Pro-
ceedings of 13th International Conference on Alkali-Aggregate Reaction in Concrete
2008, pp. 862–873.
[22] H. Takahashi, K. Yamada, Critical characters of fly ash suppressing ASR, Proceedings
of 14th International Conference on Alkali-Aggregate Reaction in Concrete, 2012
030211-TAKA.
 Y. Kawabata, K. Yamada / Cement and Concrete Research 92 (2017) 1–15
[23] T. Katayama, The so-called alkali-carbonate reaction (ACR) – its mineralogical and
geological details, with special reference to ASR, Cem. Concr. Res. 40 (2010)
643–675.
[24] T. Uomoto, Y. Furusawa, A simple model for predicting expansion in mortar bars due
to alkali-silica reaction, Concrete Research and Technology 3 (1992) 109–119 (in
Japanese).
[25] Y. Furusawa, T. Uomoto, A new evaluation system to quantitatively predict the prog-
ress of alkali-silica reaction, Concrete Research and Technology 3 (1992) 15–25 (in
Japanese).
[26] S. Poyet, A. Sellier, B. Capra, G. Foray, J.M. Torrenti, H. Cognon, E. Bourdarot, Chemical
modeling of alkali silica reaction: influence of the reactive aggregate size distribu-
tion, Mat. Struc. 40 (2007) 229–239.
[27] E. Grimal, A. Sellier, Y. Le Pape, E. Bourdarot, Creep, shrinkage, and anisotropic dam-
age in alkali aggregate reaction swelling mechanism –part I: a constitutive model,
ACI Mater. J. 105 (2008) 227–235.
[28] S. Multon, A. Sellier, M. Cyr, Chemo-mechanical modeling for prediction of alkali sil-
ica reaction (ASR) expansion, Cem. Concr. Res. 39 (2009) 490–500.
[29] S. Multon, M. Cyr, A. Sellier, P. Diederich, L. Petit, Effect of aggregate size and alkali
content on ASR expansion, Cem. Concr. Res. 40 (2010) 508–516.
[30] X.X. Gao, S. Multon, M. Cyr, A. Sellier, Alkali-silica reaction (ASR) expansion:
pessimum effect versus scale effect, Cem. Concr. Res. 44 (2013) 25–33.
[31] S.-Y. Hong, F.P. Glasser, Alkali binding in cement pastes Part I. The C-S-H phase, Cem.
Concr. Res. 29 (1999) 1893–1903.
[32] M.H. Shehata, M.D.A. Thomas, R.F. Bleszynsli, The effect of fly ash composition on
the chemistry of pore solution in hydrated cement pastes, Cem. Concr. Res. 29
(1999) 1915–1920.
[33] M.H. Shehata, M.D.A. Thomas, The effect of fly ash composition on the expansion of
concrete due to alkali-silica reaction, Cem. Concr. Res. 30 (2000) 1063–1072.
[34] M. Alnaggar, G. Cusatis, G.D. Luzio, Lattice discrete particle modeling (LDPM) of al-
kali silica reaction (ASR) deterioration of concrete structures, Cem. Con. Comp. 41
(2013) 45–59.
[35] L. Charpin, A. Ehrlacher, Microporomechanics study of anisotropy of ASR under
loading, Cem. Concr. Res. 63 (2014) 143–157.
[36] C.F. Dunant, K.L. Scrivener, Micro-mechanical modelling of alkali-silica-reaction-in-
duced degradation using the AMIE framework, Cem. Concr. Res. 40 (2010) 517–525.
[37] A.B. Giorla, K.L. Scrivener, C.F. Dunant, Influence of visco-elasticity on the stress de-
velopment induced by alkali-silica reaction, Cem. Concr. Res. 70 (2015) 1–8.
15
[38] C.F. Dunant, K. Scrivener, Physically based models to study the alkali-silica reaction,
Proceedings of the Institution of Civil Engineers - Construction Materials 2016,
pp. 1–9.
[39] T. Katayama, ASR gels and their crystalline phases in concrete –Universal products
in alkali-silica, alkali-silicate and alkali-carbonate reactions, Proceedings of 14th In-
ternational Conference on Alkali-Aggregate Reaction in Concrete, 2012 030411-
KATA-03.
[40] M. Kawamura, K. Iwahori, ASR gel composition and expansive pressure in mortars
under restraint, Cem. Con. Comp. 26 (2004) 47–56.
[41] A.G. Vayghan, F. Rajabipour, J.L. Rosenberger, Composition-rheology relationships in
alkali-silica reaction gels and the impact on the gel's deleterious behavior, Cem.
Concr. Res. 83 (2016) 45–56.
[42] T. Katayama, Late-expansive ASR in a 30-year old PC structures in eastern Japan,
Proceedings of 14th International Conference on Alkali-Aggregate Reaction in Con-
crete2012 030411-KATA-05.
[43] S. Diamond, Alkali reactions in concrete-pore solution effects, Proceedings. of 6th In-
ternational Conference on Alkali-Aggregate Reaction in Concrete 1983, pp. 155–166.
[44] M. Prezzi, P.J.M. Monteiro, G. Sposito, The alkali-silica reaction part I: use of the dou-
ble-layer theory to explain the behavior of reaction-product gels, ACI Mater. J. 94
(1997) 10–16.
[45] F. Bektas, L. Turanlim, T. Topal, M.C. Goncuoglu, Alkali reactivity of mortars contain-
ing chert and incorporating moderate-calcium fly ash, Cem. Concr. Res. 34 (2004)
2209–2214.
[46] T. Chappex, K. Scrivener, The influence of aluminium on the dissolution of amor-
phous silica and its relation to alkali silica reaction, Cem. Concr. Res. 42 (2012)
1645–1649.
[47] T. Chappex, K. Scrivener, Alkali fixation of C-S-H in blended cement pastes and its
relation to alkali silica reaction, Cem. Concr. Res. 42 (2012) 1049–1054.
[48] M.D.A. Thomas, F.A. Innis, Effect of slag on expansion due to alkali aggregate reaction
in concrete, ACI Mater. J. 95 (1998) 716–724.
[49] M.H. Shehata, M.D.A. Thomas, Use of ternary blends containing silica fume and fly
ash to suppress expansion due to alkali-silica reaction, Cem. Concr. Res. 32 (2002)
341–349.
[50] T. Ramlochan, M. Thomas, K.A. Gruber, The effect of metakaolin on alkali-silica reac-
tion in concrete, Cem. Concr. Res. 30 (2000) 339–344.
[51] T. Kim, J. Olek, H. Jeong, Alkali-silica reaction: kinetics of chemistry of pore solution and
calcium hydroxide content in cementitious system, Cem. Concr. Res. 71 (2015) 36–45.