Cement and Concrete Research 30 (2000) 1063 ± 1072

The effect of fly ash composition on the expansion of concrete due to

alkali±silica reaction
Medhat H. Shehata, Michael D.A. Thomas*
Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, Ontario, Canada M5S 1A4
Received 27 July 1999; accepted 13 April 2000

Abstract

This paper presents the results from expansion tests on concrete prisms and mortar bars containing reactive aggregate and different types
and levels of fly ash. Eighteen fly ashes representing those commercially available in North America were tested. The results show that the
bulk chemical composition of the fly ash provides a reasonable indication of its performance in physical expansion tests but cannot be used to
accurately predict the degree of expansion or the minimum safe level of fly ash required to suppress expansion to an acceptable limit.
Generally, for a given fly ash replacement level (RL), the expansion increases as the calcium or alkali content of the ash increases or its silica
content decreases. A corollary to this is that the minimum level of fly ash required to limit the expansion to an acceptable level increases as
the calcium or alkali content of the ash increases or its silica content decreases. Most of the variation in fly ash performance can be explained
on the basis of pore solution composition; those ashes effective in reducing the alkalinity of the pore solution extracted from cement paste
samples were also efficient in controlling expansion. The data from this study provide further support for the use of the accelerated mortar bar
test as a means for evaluating the efficacy of pozzolans in controlling expansion due to alkali ± silica reaction (ASR). D 2000 Elsevier Science
Ltd. All rights reserved.

Keywords: Alkali ± aggregate reaction; Concrete; Expansion; Fly ash; Pore solution

1. Introduction been shown that high - CaO fly ash is not as effective in
reducing the pore solution alkalinity of cement paste sys-
It is now generally accepted that the appropriate use of tems [3,7] and that a greater proportion of the alkalis in
fly ash can prevent expansion due to alkali ±silica reaction high -CaO fly ash may be ``available'' for reaction [8].
(ASR) in concrete. This claim is supported both by exten- The aim of the current study was to determine the effect
sive laboratory research [1] and by field experience [2]. of ash composition on ASR expansion for a wide range of
However, most of the experimental data available relate to commercially available fly ashes and to establish relation-
the use of fly ashes from bituminous coal sources (e.g. ships between the composition of the ash and the mini-
ASTM Class F fly ash), which are characterized by rela- mum level necessary to control deleterious reaction.
tively low calcium contents (i.e. <10% CaO). Relatively few Eighteen different fly ashes were collected from various
detailed studies have been conducted using higher calcium sources within North America and these were combined in
fly ashes from sub - bituminous or lignite coals, especially various proportions with high - alkali cement and reactive
those with calcium contents in excess of 25% CaO. The aggregate for testing in both concrete prism and acceler-
available data invariably indicate that such ashes are less ated mortar bar tests.
efficacious in controlling expansion compared with tradi-
tional Class F fly ashes [3± 6]. The inferior performance
may be largely ascribed to differences in the pore solution 2. Experimental details
chemistry in concretes with low or high -CaO fly ash. It has
Two Portland cements (PC) with alkali contents, of
* Corresponding author. Tel.: + 1 - 416 - 978 - 6238; fax: + 1 - 416 - 978 - 1.02% Na2Oe (HAPC) and 0.60% Na2Oe (LAPC) were
6813. used together with 18 fly ashes with a wide range of
E-mail address: [email protected] (M.D.A. Thomas). chemical composition. The chemical compositions of the

0008-8846/00/$ ± see front matter D 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 8 - 8 8 4 6 ( 0 0 ) 0 0 2 8 3 - 0
1064 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072

PCs and fly ashes are presented in Table 1. Recently disregarding the alkalis present in the fly ash. Three of the
published changes in the Canadian specifications for sup- mixes (with 20% FM, 30% WM, and 45% OK) were cast
plementary cementing materials (CSA A23.5) included a with an alkali content of 5.25 kg /m3 Na2Oe calculated on
new classification system for fly ash based on the calcium the basis of alkalis from the cement and added NaOH.
oxide content (CaO) of the ash. The new categories of fly Mortar bars were prepared and tested according to the
ash are Type F (  8 ‹ 1% CaO), Type CI (>8% and  20 ‹ Accelerated Mortar Bar Test Method, CSA A23.2 - 25A
2%) and Type CH (>20% CaO). As shown in Table 1, two (equivalent to ASTM C 1260) using the same cementing
of the fly ashes are type F, eight are Type CI and eight are material combinations used for the concrete prism test. The
Type CH. It should be noted that fly ash MN would also reactive coarse aggregate (Spratt) was crushed and graded
meet Type F and ashes PI and C2 would meet Type CI according to the standard test procedure. Mortar samples
because of the tolerances placed on the calcium oxide limits. were cast at an aggregate /cementing materials (CM) ratio of
A reactive siliceous limestone (Spratt) was used as the 2.25 and W/ CM equal to 0.50. Samples were cured at room
coarse aggregate together with a non -deleteriously reactive temperature, in their moulds, for 24 h. Then, they were
natural sand in all the mixes. Concrete prisms were cast and stripped, immersed in water and maintained at 80°C for
tested in accordance with the Canadian Standards Associa- another 24 h. After that, samples were taken out, their initial
tion Concrete Prism Test, CSA A23.2 Test 14A (equivalent lengths were measured and they were soaked in 1 M NaOH
to ASTM C 1293) using 420 kg / m3 of cementitious solution at 80°C throughout the testing period. Length
material and W /CM ranging from 0.42 to 0.45. Five control changes were measured at 3, 7, and 14 days. The 14- day
mixes (without fly ash) were cast at Na2Oe contents of 5.25, expansion results are reported here.
4.20, 3.70, 3.15 and 2.89 kg/m3 of concrete. The first mix,
at 5.25 kg /m3 Na2Oe, represents concrete made with 420
kg /m3 of cement with the alkali content raised to 1.25% 3. Results
Na2Oe (as per CSA) and the other four represent concrete
with the same cement content but with 20%, 30%, 40% and 3.1. Concrete prism test
45% alkali dilution. The alkali contents were adjusted by
mixing the low and high -alkali cement, and by adding The 2 -year expansion results for 42 concrete mixes are
NaOH to the mixing water, if needed. reported in Table 2. Thirty- three of these mixes were cast
A total of 42 mixes were cast using various replacement and tested at the University of Toronto, whereas the results
levels (RLs) of different fly ashes. All these mixes were cast for the other nine mixes were provided by the Corporate
using high - alkali cement with, in the majority of cases, the Technical Services (CTS) Laboratory of Lafarge Canada.
alkali content boosted with NaOH to achieve 1.25% Na2Oe For the purpose of discussion in this paper, it is assumed
on the basis of the mass of the PC in the mixture, that expansions in excess of 0.04% are indicative of sig-

Table 1
Chemical composition of the OPC and fly ashes (mass %)
Material SiO2 Al2O3 Fe2O3 CaO MgO SO3 K2 O Na2O Na2Oea TiO2 P2O5 Na2Oeb LOI
HAPC 20.83 5.11 2.01 62.98 2.43 3.25 1.12 0.28 1.02 0.24 0.05 ± 0.92
LAPC 21.27 4.22 2.98 62.71 2.13 2.82 0.77 0.10 0.60 0.18 0.11 ± 2.40
LG 41.96 19.64 20.07 5.57 1.19 0.95 2.44 0.69 2.30 0.84 0.15 1.10 3.71
FM 47.34 22.34 15.08 6.38 0.82 1.43 1.23 0.60 1.41 1.10 0.32 0.47 2.73
MN 61.5 20.52 4.29 8.68 1.70 0.19 0.60 0.17 0.56 1.38 0.05 0.30 0.08
SD I 50.92 23.64 4.62 13.63 0.86 0.23 0.59 3.38 3.77 0.14 0.73 1.42 0.42
BD 45.66 21.42 5.53 12.34 2.76 0.84 0.96 7.82 8.45 0.65 0.14 2.43 0.35
SD II 51.56 22.90 4.58 15.15 1.16 0.28 0.30 2.60 2.80 0.66 0.12 1.80 0.35
TB 40.68 21.19 4.50 15.87 3.54 2.18 0.49 8.14 8.46 0.96 0.65 3.60 0.53
C1 44.29 20.96 5.23 17.51 4.21 2.13 0.84 1.13 1.68 1.12 0.63 0.77 1.14
WM 39.77 21.46 5.69 18.46 3.77 1.86 0.66 3.71 4.14 1.04 0.54 2.52 1.06
BR 32.71 19.02 5.76 18.85 4.30 4.81 0.68 8.28 8.73 1.24 0.52 4.79 1.18
PI 38.42 20.57 5.64 20.50 4.39 1.76 0.62 2.64 3.05 1.00 0.52 1.84 2.01
C2 39.83 19.56 5.54 21.53 4.62 2.14 0.60 1.55 1.94 1.20 0.71 0.99 1.68
EW 38.22 18.43 5.72 24.61 4.72 1.55 0.44 1.39 1.68 1.42 1.04 0.94 0.18
PP 35.20 18.72 6.06 26.61 5.12 2.49 0.36 1.59 1.83 1.50 1.19 1.33 0.39
IN 36.12 18.64 6.07 26.62 5.41 1.80 0.40 1.34 1.60 1.48 1.12 1.02 0.16
OK I 34.60 16.45 7.13 27.71 5.89 2.71 0.21 1.51 1.65 1.30 0.71 1.23 0.28
OK II 31.65 16.65 7.28 29.10 6.57 3.17 0.20 1.72 1.85 1.33 0.74 1.19 0.36
CC 41.12 11.24 5.93 30.00 4.40 2.13 1.76 1.10 2.26 0.47 0.10 1.05 0.78
a
Acid soluble alkali.
b
Available alkali, expressed as Na2Oe, as per ASTM C311.
 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072 1065

Table 2
Expansion of concrete prisms after 2 years (%)
Fly Ash Replacement Levels (RLs)a (RLs)b
Type Source 0% 15% 20% 25% 30% 40% 45% 50% 60% 20% 30% 45%
Control 0.250 0.241 0.164 0.114 0.044
F LG 0.030
F FM 0.083 0.043 0.099
F MNc 0.039 0.014
CI SD I 0.100 0.067 0.042
CI BD 0.156 0.113 0.087
CI SD IIc 0.062 0.032
CI TB 0.117
CI C1 0.039
CI WM 0.123 0.100 0.068 0.042 0.112
CI BR 0.157
CH PI 0.026
CH C2 0.080
CH EW 0.086 0.051 0.033
CH PPc 0.176 0.082
CH INc 0.191 0.130
CH OK I 0.138 0.054 0.021 0.136
CH OK IIc 0.176
CH CC 0.162
a
Standard concrete prism test, alkali level = 1.25% Na2Oe of mass of PC.
b
Concrete prism test with augmented alkalis [5.25 kg/m3 or 1.25% Na2Oe of the mass of (PC + FA)].
c
Indicates data supplied by CTS.

nificant or damaging expansion, as this is generally con- increases above 4.20 kg /m3. Extrapolation of the data in
sistent with the onset of visible cracking. Fig. 2 indicates that the alkali content needs to be kept
The development of expansion with time for the control below approximately 2.8 kg/m3 Na2Oe in order to keep the
samples (i.e. those without fly ash) at various alkali contents expansion below 0.04% at 2 years. It is interesting to note
is shown in Fig. 1. The expansion of all five concrete mixes that little further expansion is observed when the alkali
exceeded 0.04% after 2 years although the rate of expansion content of the mix is raised above about 4.20 kg /m3 Na2Oe.
and the magnitude of the ultimate expansion is clearly This would indicate that at such high levels of alkali, the
dependent on the alkali content of the concrete. The rela- reaction (or more appropriately the expansion) is limited by
tionship between the 2 -year expansion data and the alkali some other factors such as the quantity of reactive silica or
content of the control samples is shown in Fig. 2. There is a perhaps the availability of calcium.
progressive increase in the prism expansion as the alkali The 2 -year expansion results for the various combina-
content is raised from 2.89 to 4.20 kg/ m3 Na2Oe but little tions of cement and fly ash (presented in Table 2) show that
significant further increase in expansion as the alkali content the effect of fly ash on the expansion varies widely. This fact

Fig. 2. Effect of alkali content of concrete on the 2 - year expansion of

Fig. 1. Effect of alkali content of concrete on expansion due to ASR. concrete prisms.
1066 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072

control concrete at the same level of dilution. This means

that even the less effective fly ashes can be expected to have
a beneficial impact on the expansion of the concrete beyond
the role of merely diluting the cement alkalis. This phenom-
enon is further illustrated in Fig. 6 which shows the
expansion of the control concrete and three fly ash concretes
all at alkali content of 5.25 kg/ m3 Na2Oe (calculated on the
basis of the alkali from the cement plus the added NaOH
only). In all cases, the expansion of the fly ash concretes is
less than that of the control concrete at the same alkali
content. In addition, samples containing fly ash showed less
evidences of disruption (e.g. cracking and gel exudation)
compared with the control sample cast at Na2Oe content of
5.25 kg /m3 of concrete.
Fig. 7 shows a relation between the CaO content of the
Fig. 3. Effect of low - calcium ash (FM) on expansion due to ASR. ash and the measured or estimated expansion values of the
concrete containing 25% ash. Where data were not avail-
able for a particular fly ash at an RL of 25%, the expansion
is perhaps best illustrated by comparing the performance of value was determined by interpolating or extrapolating the
the fly ashes designated FM and OK I as shown in Figs. 3 data in Table 2. For example, fly ash WM was tested at
and 4, respectively. Both ashes have similar alkali contents; RLs of 20%, 30%, 40%, and 50%. The value for a 25% RL
however, the FM is a Type F fly ash (with 6.38% CaO) and was interpolated to be 0.112%. The graph includes the
the OK is a Type CH fly ash (with 27.7% CaO). Both fly expansion data provided by CTS laboratory. Concrete
ashes meet all the chemical and physical test criteria of both containing fly ash of alkali content > 4.0% Na2Oe are
the Canadian and ASTM specifications for fly ash (CSA represented by the hollow markers while those containing
A23.5 and ASTM C618) and have a history of satisfactory fly ash of alkali content < 4.0% Na2Oe are represented by
use in concrete. However, they clearly perform very dif- solid markers. Also, the Type of the ash (F, CI, or CH) is
ferently with regards to controlling ASR. Fly ash FM is marked on the graph. There appears to be some relationship
very efficient in controlling expansion with 20% fly ash between the CaO content of the fly ash and the expansion
reducing expansion to just 0.043%; an RL of 25% would of concrete prisms for fly ash with a low to moderate alkali
have undoubtedly reduced expansion to less than 0.04%. level (i.e. < 4.0% Na2Oe). There is a slight increase in the
This behaviour is typical of the other low - calcium fly expansion as the CaO content increases ranging from 5% to
ashes used in this study and reported elsewhere [1]. Fly 20% with a sudden and remarkable increase above 20±
ash OK on the other hand is far less effective in this role 22% CaO. In nearly all cases, fly ashes (with low to
and it can be seen that an RL somewhere in the region of moderate alkali) that meet the CSA specification for Type
45% to 60% would be required to limit expansion to less F or Type CI fly ash were effective in limiting the 2- year
than 0.04% with this material. The other high -calcium expansion of concrete prism to  0.04% when used at an
ashes exhibited similar behaviour. RL of 25%. The one exception was fly ash SD I which
Fig. 5 shows the 2 -year expansion values of concrete
samples containing various fly ashes at different RLs. Also
shown in the graph are the 0.04% expansion limit and the
expansion values of the control samples at alkali levels
corresponding to the dilution levels of the ash. The graph
illustrates the variations in the efficiency of different fly
ashes in controlling the expansion, i.e. the level of ash
required to suppress expansion to less than 0.04% varies
considerably between fly ashes. It would appear from Fig. 5
that ashes with either a high - alkali or a high - calcium
content are likely to be less effective at a given level of
replacement. However, all the fly ashes tested yielded
expansion values lower than the control sample with an
alkali content of 5.25 kg /m3 Na2Oe. In other words, the
partial replacement of high -alkali cement for fly ash did not
lead to an increase in expansion regardless of the composi-
tion of the fly ash. Indeed, in most cases, the expansion of
concrete at a particular level of fly ash was less than the Fig. 4. Effect of high - calcium ash (OK) on expansion due to ASR.
 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072 1067

Fig. 5. Effect of ash composition and replacement level on expansion due to ASR.

produced a 2- year expansion of just 0.042% when used at (8.45% Na2Oe) resulted in deleterious expansion even when
a 25% RL. Fly ashes that are classified as CH ashes used at an RL of 40%.
generally failed to control expansion to  0.04% at 25% The relation in Fig. 8 is established in an attempt to
RL although in all cases, expansion was reduced compared correlate the fly ash composition with the expansion of the
to the control concrete without fly ash. The one exception concrete prisms. The samples and the legend used in Fig. 8
was fly ash PI which had a CaO content of 20.5% and a are the same as those used in Fig. 7. Multiple regression
measured expansion of 0.026% when tested in concrete at a analyses were performed between expansion as the depen-
25% RL. However, the calcium content of this fly ash is dent variable and the CaO, Na2Oe, and SiO2 contents of the
only slightly higher than the 20% limit for Type CI ash. fly ash as the independent variables. The best fit (i.e. highest
Indeed, this ash could be classified as CI due to the value of R2 and F ratio) was achieved when the Na2Oe and
tolerance of ‹2% CaO placed on the upper limit for Type CaO were normalized to the SiO2 content. The graph shows
CI fly ash. a general trend of increasing expansion as the Na2Oe and
The expansion results for fly ashes with higher alkali CaO contents of the fly ash increase and as its silica content
contents (i.e. > 4.0% Na2Oe) do not fit the general trend decreases. The graph also indicates that fly ash with
shown in Fig. 7. Indeed, in some cases, significant expan- chemical composition that satisfies the condition
sion and cracking was observed for concrete containing (10Na2Oe + 4.45CaO) / SiO2  2 will likely meet the
high - alkali Type CI ashes at RLs of 25% (see Fig. 7) and 0.04% expansion criterion of the concrete prism test after
higher (see Table 2). Fly ashes WM (4.14% Na2Oe) and BD 2 years when used at an RL of 25%.

3.2. Accelerated mortar bar test

The results of the accelerated mortar bar tests showed the

same general trends as the concrete prism test data, with
higher alkali and higher calcium fly ashes performing less
favorably. Fig. 9 shows the 14- day expansion results from
the accelerated mortar bar test plotted against the 2- year
expansion results from the concrete prism test for the same
combination of materials (i.e. same type and level of fly
ash). It has been suggested by Berube et al. [9] that material
combinations that result in an expansion of less than 0.10%
at 14 days in the mortar bar test are likely to meet the 0.04%
expansion criterion of the concrete prism test after 2 years.
Thomas and Innis [10] found this statement to be applicable
Fig. 6. Effect of ash type and replacement level on expansion of concrete to concrete contain various reactive aggregates, pozzolans
prisms at 5.25 kg / m3 Na2Oe. and slag. The samples tested in the present study show that
1068 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072

Fig. 7. Effect of CaO content of fly ash on expansion of concrete containing 25% fly ash.

it is reasonable to consider that the material combinations solution. The sum of alkali cations was always higher than
that yield a mortar bar expansion value > 0.1% at 14 days the hydroxyl anions, which is attributed mainly to the
will not meet the 0.04% 2 -year expansion criterion from the presence of silica in solution. Sulphate anions may also be
concrete prism test. However, some material combinations present in the pore solution at elevated temperature (80°C)
that show mortar bar expansion  0.1% do not necessarily and are expected to contribute, partly, to the difference
result in a concrete prism expansion  0.04% after 2 years. between the alkali cations and OH anion concentrations.
Although only two combinations (BD at 40% and SD II at Fig. 11 shows that the difference between the sum of
15%) out of a total of 35 tested pass the 0.10% criterion of (Na + K) and OH increases as the sum of alkali cations
the mortar bar test and significantly fail the 0.04% limit of (Na + K) increases, which reflects the presence of higher
the concrete prism test. amounts of other anions (silica and sulphate) in solution at
In an attempt to establish a better understanding of the higher alkalinity. Fig. 12 shows that as the alkali content of
factors that influence the mortar bar test results, the pore the pore solution increases so too does the expansion of the
solution was extracted from the mortar specimens after 14 mortar bars. The alkali in the pore solution of mortar bars at
days immersion in 1 M NaOH solution and analyzed. Fig. 14 days is mainly a function of the availability of alkalis in
10 shows the hydroxyl (OHÿ) ion concentrations plotted the cementitious system and the quantity of alkalis that
against the sum of alkali cations (Na + + K + ) in the pore penetrate from the external solution of 1 M NaOH at 80°C.

Fig. 8. Effect of ash chemical composition on expansion of concrete prisms Fig. 9. Relation between the 14 - day and 2 - year expansions of the
containing 25% fly ash (legend, same as Fig. 7). accelerated mortar bar and concrete prism tests.
 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072 1069

Fig. 12. Effect of (Na + K) ion concentrations in the pore solution on the
Fig. 10. Relation between OH and (Na + K) ion concentrations in the pore 14 - day expansion of mortar bars.
solution extracted from mortar bars at 14 days.

Thus in the accelerated mortar bar test, one is actually higher RLs to prevent damaging expansion. While 25% of
testing the ability of the fly ash to lower the alkalinity of the low - calcium fly ashes maintained the expansion lower
the pore solution (by binding alkalis in the hydrates) and to than 0.04% after 2 years, the required level for high - alkali
reduce the diffusivity of the mortar bar making it less (e.g. BD or WM) or high -calcium ashes (e.g. EW or OK)
accessible to the external alkali solution. may be in excess of 40% to 50%.
As mentioned previously, much of the reduced efficacy
of high - calcium fly ashes in controlling ASR may be
4. Discussion explained on the basis of pore solution chemistry. A study
of the effects of fly ash composition on the evolution of the
The results of this study show that the efficacy of fly ash pore solution of cement paste was also conducted by the
in controlling expansion due to ASR varies within a wide authors [7] using materials from some of the same sources
range. For instance, concrete containing 25% fly ash shows used for this study. Fig. 13 shows the expansion at 2 years
2 -year expansion values ranging from 0.014% to 0.176% of concrete prisms containing 25% fly ash plotted against
(Table 2). However, all types of ash at all the RLs reduced the hydroxyl ion concentration of the pore solution ex-
expansion compared to the control concrete without fly ash. tracted from cement pastes of the same age and containing
Moreover, all ashes show reduced expansion as the level of 25% of the same fly ash. The data show that the fly ashes
replacement increases. Generally, fly ashes with higher that are capable of lowering the hydroxyl ion concentration
alkali or calcium contents are less effective in controlling of the pore solution of pastes to less than approximately
expansion due to ASR and consequently have to be used at 0.60 mol / l are also generally effective in reducing expan-

Fig. 11. OHÿ(Na + K) vs. (Na + K) ion concentrations in the pore solution Fig. 13. Effect of OH ion concentration on expansion of concrete containing
extracted from mortar bars at 14 days. 25% fly ash.
1070 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072

sion to less than 0.04% at 2 years. A slightly higher value

(0.65 mol /l) was suggested by other authors [11]. If the two
high - alkali ashes are excluded from the analysis, then there
is a broad correlation between the pore solution alkalinity of
the pastes and the expansion of the concrete. High -calcium
fly ashes reduce both the pore solution alkalinity and
expansion compared to the control, but the reductions are
not sufficient to prevent damaging expansion.
The behaviour of the high - alkali fly ashes in Fig. 13 is
interesting. These ashes produce dramatic increases in the
pore solution alkalinity compared with the high - alkali
cement control, but the same ashes do not lead to increased
expansion; indeed, they significantly reduce expansion
compared with the control. This suggests that the effect of
fly ash on expansion is not solely the result of its influence
on the pore solution chemistry. The authors believe that fly
Fig. 14. Relationship between fly ash composition and minimum
ash plays a further role in the ASR by reducing the
replacement level.
availability of calcium in the system. This has been dis-
cussed elsewhere [12,13].
For the purpose of using fly ash to control ASR, the The calculated ``safe levels'' presented in Table 3 are
question that needs to be answered is not whether a grouped into four categories. Group I contains all the Type F
particular fly ash is suitable or not but how much of a and Type CI fly ashes of low to moderate alkali content (i.e.
particular fly ash is required to limit the risk of expansion < 4.0 Na2Oe). For all of these ashes, the safe level falls
to an acceptable level. This minimum or ``safe'' level will somewhere between 20% and 25% replacement. Group II
vary depending on the nature of the reactive aggregate, contains all the CI fly ashes of high -alkali content for which
availability of alkali in the concrete (i.e. from PC), the the safe RL varies between 40% and 55%. Group III
intended exposure conditions of the concrete and, as represents two fly ashes of low alkali content which by
demonstrated in the present study, the composition of the virtue of their calcium contents may be categorized as either
fly ash itself. Type CI or CH ashes (i.e. 20% < CaO < 22%). These fly
To examine the relationship between the composition of ashes are required to be used at levels of 25% to 30% to
the fly ash and the minimum amount required to control control expansion to 0.04% in the conditions used in this
expansion, the concrete prism expansion data in Table 2 study. Group IV includes the fly ashes with calcium contents
were used to estimate the level of fly ash that produced an greater than 25% CaO and with generally low alkali con-
expansion of 0.04%. In many cases, this level of expansion tents (1.60% to 2.26% Na2Oe). The safe levels estimated for
was bracketed by the results for the RLs selected for study these ashes fall in the range, 35% to 60%. Although it is
(e.g. MN, SD I, SD II, EW, OK II) and the ``safe level'' possible to loosely group the behaviour of the fly ashes by
was simply determined by interpolation of the experimen- composition, there is no clear relationship between the
tal data. In other cases, it was necessary to extrapolate the composition of the ash and the safe level of expansion
data to determine the safe level at which an expansion of within a group. For example, in Group II, the improved
0.04% expansion occurs by assuming a linear relationship performance of ash TB compared with BD ash cannot be
between expansion and RL. The authors are aware of the explained on the basis of composition. A similar statement
limitations of this approach but still consider the exercise can be made for the marked improvement in the perfor-
useful for the purposes of discussion. The safe levels, mance of fly ash PP when compared with OK I, which has a
estimated to the nearest 5% RL, for the different fly ashes very similar chemical composition.
are given in Table 3. The chemical composition of the fly ash can be used to
explain some but not all of the performance of the fly ashes
Table 3 in the concrete prism test (and, for that matter, the acceler-
Estimated safe level (%) of fly ash required to limit concrete expansion ated mortar bar test). Fig. 14 shows the ``best fit'' relation-
Group I Group II Group III Group IV ship between the chemical composition of the fly ash (based
Ash Safe level Ash Safe level Ash Safe level Ash Safe level on calcium, silica, and alkali contents) and the safe level
required to limit expansion to 0.04%. This relationship is
LG 25 TB 40 PI 25 EW 45
FM 20 BD 50 C2 30 PP 35 almost identical to the relationship linking the chemical
MN 20 WM 45 IN 45 composition and the 2 -year expansion of prisms with 25%
SD I 25 BR 55 OK I 55 fly ash (Fig. 8). It is clear from Figs. 8 and 14 that the
SD II 20 OK II 60 chemical composition may be used to provide an indication
C1 25 CC 60
of the performance of the fly ash, but cannot be reliably
 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072 1071

used to predict expansion or determine the minimum repla- cretes without fly ash. In all cases, increasing the level of
cement required for a particular fly ash. Various mineralo- replacement of a particular fly ash further reduced expansion.
gical and physical properties of the fly ash, particularly the (2) At a given level of fly ash replacement, the expan-
quantity of glass present and the fineness of the ash, will sion of concrete prisms generally increased as the calcium
also influence the performance. For example, the efficacy of or alkali content of the fly ash increased or as its silica
fly ash in controlling expansion was found to increase as its content decreased.
total silica content increases, however, it is obvious that only (3) The minimum level of replacement required to
the amorphous silica will have a beneficial effect on ASR as control expansion to  0.04% at 2 years generally increased
any crystalline phases (e.g. quartz) can be expected to be as the calcium or alkali content of the fly ash increased or as
inert in concrete. This will be discussed further in a its silica content decreased.
subsequent paper. (4) Much of the variation in performance of different fly
The inability to make accurate predictions of perfor- ashes can be explained on the basis of pore solution
mance based solely on the composition of the fly ash composition. The fly ashes that were found to be most
highlights the need for a performance indicator. Concrete effective in reducing the alkalinity of the pore solution
prism tests, while being reasonably representative of field expressed from paste samples were also found to be the
concrete, suffer from requiring a long testing period. The best for controlling ASR expansion.
accelerated mortar bar test provides a test result just 16 (5) The accelerated mortar bar test provides a reasonable
days after casting the specimen (or after 14 days in 1 M indication of the performance of a particular fly ash in the
NaOH at 80°C). The data presented here and elsewhere concrete prism test.
[9,10] indicate that this test provides a reasonably reliable
indication of the performance of pozzolans and slag in the
concrete prism test. A recent paper by Thomas and Innis Acknowledgments
[14] reports concrete prism and mortar bar test results for
70 different combinations of various pozzolans, slag and The funding for this work was provided by a grant from
reactive aggregates, and concludes that combinations that the Collaborative Research and Development Program of
pass the accelerated test (i.e.  0.10% expansion at 14 the Natural Science and Engineering Research Council
days) can be used in the field with a very low (and (NSERC) of Canada. The industry partner was Lafarge
acceptable) risk of deleterious expansion due to ASR. Canada. In addition to providing financial support, Lafarge
The data presented here generally support that statement. also carried out some of the testing at their Corporate
This test has an advantage over accelerated tests with Technical Services Laboratory in Montreal.
Pyrex glass (ASTM C 441) in that it allows pozzolans
or slags to be evaluated in combination with the particular
reactive aggregate under question. The nature of the References
reactive aggregate will influence the level of prevention
required (e.g. the safe level of fly ash). A further refine- [1] M.D.A. Thomas, Review of the effect of fly ash and slag on alka-
ment of the accelerated mortar bar test is required to allow li ± aggregate reaction in concrete, Building Research Establishment
the test to be used to evaluate combinations of pozzolans Report, BR314, Construction Research Communications, Watford,
UK, 1996.
or slag with cements of varying alkalinity.
[2] M.D.A. Thomas, Field studies of fly ash concrete structures contain-
Although a wide range of fly ashes was selected for the ing reactive aggregates, Mag Concr Res 48 (177) (1996) 265 ± 279.
study, it is apparent that some compositions were not [3] S. Diamond, F. Lopez - Flores, On the distinction in physical and
adequately represented. For instance, all the low and chemical characteristics between lignite and bituminous fly ash, in:
high - calcium fly ashes used were characterized by being S. Diamond (Ed.), Effects of Fly ash Incorporation in Cement and
Concrete, Proceedings of the MRS Symposium N, Materials Research
of low alkali content (i.e. generally below 2.0% Na2Oe).
Society, Boston, 1981, pp. 34 ± 44.
Also, with one exception, the high - alkali Type CI fly [4] E.R. Dunstan, The effect of fly ash on concrete alkali ± aggregate
ashes were characterized by having very high -alkali levels reaction, Cem Concr Aggregates 3 (2) (1981) 101 ± 104.
(i.e. >8.0% Na2Oe) and ashes with alkali contents in the [5] A.D. Buck, K. Mather, Methods for controlling effects of alkali ±
range, 4.5% to 8.0% Na2Oe were not included in this silica reaction in concrete, Technical Report No. SL - 87 - 6, US
Army Corps of Engineers, Waterways Experiment Station, Vicks-
study. Further work is required to fully determine the role
burg, 1987.
of the alkalis in fly ash on its performance in the presence [6] P. Klieger, S. Gebler, Fly ash and concrete durability, Concrete
of reactive aggregates. Durability, Katharine and Bryant Mather International Conference,
ACI SP - 100, vol. 1, American Concrete Institute, Detroit, (1987)
1043 ± 1069.
[7] M.H. Shehata, M.D.A. Thomas, R.F. Bleszynski, The effect of fly ash
5. Conclusions
composition on the chemistry of pore solution, Cem Concr Res 29
(1999) 1915 ± 1920.
(1) All the fly ashes included in this study reduced the [8] C. Lee, Effects of alkalis in Class C fly ash on alkali ± aggregate
expansion of concrete prisms compared with control con- reaction, in: V.M. Malhotra (Ed.), Proceedings of the 3rd International
1072 M.H. Shehata, M.D.A. Thomas / Cement and Concrete Research 30 (2000) 1063±1072

Conference on the Use of Fly Ash, Silica Fume, Slag and Natural at reaction mechanisms: Part 2. Pore solution chemistry, Cem Concr
Pozzolans in Concrete, ACI SP - 114, vol. 1, American Concrete In- Res 24 (2) (1994) 221 ± 230.
stitute, Detroit, 1989 pp. 417 ± 430. [12] R.F. Bleszynski, M.D.A. Thomas, Microstructural studies of alkali ±
[9] M.A. Berube, J. Duchesne, D. Chouinard, Why the accelerated mortar silica reaction in fly ash concrete immersed in alkaline solutions, Adv
bar test method ASTM C1260 is reliable for evaluating the effective- Cem Based Mater 7 (1998) 66 ± 78.
ness of supplementary cement, Cem Concr Aggregates 17 (1) (1995) [13] M.D.A. Thomas, The role of calcium on alkali ± silica reaction, Ma-
26 ± 34. terials Science of Concrete Ð The Sidney Diamond Symposium,
[10] M.D.A. Thomas, F.A. Innis, Effect of slag on expansion due to American Ceramics Society, Westerville, OH, (1998) 325 ± 335.
alkali ± aggregate reaction in concrete, ACI Mater J 95 (6) (1998) [14] M.D.A. Thomas, F.A. Innis, Use of the accelerated mortar bar test
716 ± 724. for evaluating the efficacy of mineral admixtures for controlling
[11] J. Duchesne, M. - A. Berube, The effectiveness of supplementary ce- expansion due to alkali ± silica reaction, Cem Concr Aggregates 21
menting materials in suppressing expansion due to ASR: Another look (2) (Dec. 1999) 157 ± 164.