Cement and Concrete Research 35 (2005) 125 – 131

Delayed ettringite formation symptoms on mortars induced by high

temperature due to cement heat of hydration or late thermal cycle
R. Barbaruloa,b,c,*, H. Peycelona, S. Prenéb, J. Marchandc
a
DPC/SCCME/LECBA, CEA Saclay, 91191, Gif-sur-Yvette, France
b
Département MMC, Électricité de France, 77818 Moret-sur-Loing, France
c
CRIB, Département de Génie Civil, Université Laval, Québec, Canada G1K 7P4
Received 18 September 2002; accepted 24 May 2004

Abstract

Cases of delayed ettringite formation (DEF) have mainly been detected on mortars or precast concretes steam-cured according to a
predefined temperature cycle during hydration. The present study shows that other situations in which the material is submitted to a
temperature cycle can induce DEF expansions. Mortar bars were made with three different cements (types 10, 20M, and 30). As a first heat
treatment, the mortar bars were steam-cured to reproduce the temperature cycle they would undergo if they were at the center of a large
mortar member. The dimensional variations of these specimens were studied for 1 year. After 1 year, half of the specimens were steam-cured
for 1 month at 85 jC. The expansions were followed for two more years. The early-age steam-cure-induced expansions for mortar types 10
and 30. Late steam-curing induced expansions for the three cements tested. In one case (cement type 20M), the early-age steam cure has
suppressed or delayed the expansion induced by the late steam cure. A scanning electron microscopy (SEM) study showed that typical DEF
symptoms are associated with the expansions.
D 2004 Elsevier Ltd. All rights reserved.

Keywords: Delayed ettringite formation; Heat of hydration; Thermal treatment

1. Introduction The first issue of this work was to determine if a

cementitious material could develop DEF from its own heat
Since the first cases of delayed ettringite formation (DEF) of hydration. A few field observations tend to demonstrate
were reported on heat-cured mortars, DEF has been a subject that DEF could be induced by an excessive rise of temper-
of great concern [1 –5]. As stated by Taylor et al. [5], DEF ature because of the heat of hydration [4,8,9]. However, to
may be defined as ‘‘the formation of ettringite in a cemen- our knowledge, no laboratory experiments have been carried
titious material by a process that begins after hardening is out to validate this hypothesis. The first part of this work
substantially complete and in which none of the sulfate thus consisted in submitting different mortars to steam cures
comes from outside the cement paste.’’ The issue whether directly computed from their measured heats of hydration
ettringite is the cause of expansion is not definitively solved and in following their expansions with time. The length
yet [6,7], but the necessity for the mortar to undergo a high changes of the different mortars have been measured for
temperature (above approximately 70 jC) during hydration 3 years.
is now well established [1 –3]. The purpose of this experi- The second issue of this work was to determine whether
mental work was to determine if other curing conditions, a concrete steam-cured long after its hydration could
during hydration or later, could lead to DEF. present DEF symptoms in the long term. Few data have
been published about the evolution of hydrated mortars at
* Corresponding author. DPC/SCCME/LECBA, CEA Saclay, 91191, high temperature in a saturated environment. The work of
Gif-sur-Yvette, France. Tel.: +33-1-6908-3042; fax: +33-1-6908-8441. Paul and Glasser [10] followed the mineralogical and
E-mail address: [email protected] (R. Barbarulo). structural evolution of cement pastes kept at 85 jC for
0008-8846/$ – see front matter D 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconres.2004.05.041
126 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131

more than 8 years. Expansion was not the main issue of Table 2
their work, but no length change could be observed after Mineralogical compositions of the cements used in this study determined by
the Bogue calculation (wt.%)
the warm moist cure, back to room temperature. However,
Cement Type 10 Type 20M Type 30
pastes are known to present very long induction times
before expanding and very low expansion rates [11]. In C 3S 57.7 39.2 47.4
C2S 14.6 36.9 24.8
the present work, 1-year-old mortars have been submitted
C3A 6.6 4.0 9.1
to a 1-month steam curing at 85 jC, and their expansions C4AF 9.1 12.4 8.3
have been followed for two more years.
Calorimetric measurements during adiabatic hydration of
2. Experimental the mortars described above yielded their total heat of
hydration. The specific heat and the thermal conductivity
2.1. Materials of the different mortars were estimated (see Table 3).
Different simulations, using the software Quadrel [13],
Three different cements were used in this study: a type 10 allowed to find a geometrical configuration for which the
cement (considered as an average Portland cement), a type type 10 mortar, considered as the reference mortar, would
20M cement (low C3A content, low heat of hydration, low reach 85 jC during hydration. This configuration consists of
sulfate content), and a type 30 cement (high C3A content, a 1-m wall of mortar. Of course, such a mortar wall (or
high heat of hydration, high sulfate content). Types 10, concrete with such a high level of cement) would be
20M, and 30 refer to the Canadian Portland cements exceedingly bad engineering and would never be cast since
classification CAN/CSA-A5-98. The chemical and mineral- it would undergo thermal cracking. The experiment was
ogical compositions of the different cements provided by actually designed to compare different cements in an autog-
the manufacturer are given Tables 1 and 2. enous heat curing. The type 10 mortar thus reaches 85 jC in
No reliable relationship based on cement composition about 24 h. Fig. 1(a) shows the different temperature cycles
has been found yet to determine the DEF susceptibility of a at the center of the hypothetic mortar wall during hydration,
cement [1,5]. However, considering a ‘‘classical’’ laboratory up to 7 days. After 7 days, the three curves are identical and
steam curing for precast mortar (for instance, 4 h precure, 12 reach 20 jC after 1 month.
h at 90 jC, and cooling [12], designed to cause DEF in a These temperature cycles were adapted to be easily
laboratory study), type 20M cement would be the least reproducible in the laboratory. The cycles were shortened
susceptible in developing DEF symptoms, whereas type to reach 20 jC in 1 week instead of 1 month. The maximum
30 cement would be the best candidate for DEF. temperature of the type 30 cycle 4 was lowered by 5 jC in
Granitic, alkali– silicate reaction proof sand was used. order not to reach 100 jC. These temperature cycles are
Mortars were made according to the C305-82 ASTM represented Fig. 1(b). Mortar type 10 reaches 85 jC in 24 h,
standard procedure. The water/cement ratio was 0.4, and mortar type 20M reaches 75 jC in 36 h, and mortar type 30
the sand/cement ratio was 2.1. reaches 95 jC in 18 h. Despite these modifications, these
temperature cycles can still be considered as characteristic of
2.2. Definition of the hydration temperature cycle what would undergo in the center of the 1-m wall described
above. As it was designed, the experiment tests as much for
Two distinct temperature cycles have been imposed on heat of hydration as for inherent DEF susceptibility.
the mortars. The first temperature cycle reproduces the
conditions at the center of a large concrete member having 2.3. Specimens curing and conservation
a very high cement content in which temperature rises
because of the heat of hydration and low heat losses. It is Mortars were cast under the form of 35  35  200-mm
called here ‘‘hydration temperature cycle’’. bars in steel molds. For each mortar, half of the samples were
thus readily steam-cured according to their hydration cycle
(see Fig. 1(b)). The remaining specimens were left at room
Table 1 temperature (20 jC), 100% RH (samples left above a surface
Chemical analyses of the cements used in this study (wt.%)
Cement Type 10 Type 20M Type 30
Table 3
SiO2 20.30 23.20 21.12 Calorimetric characteristics of the mortars used in this study
Al2O3 4.42 4.11 5.18
Fe2O3 3.00 4.09 2.73 Cement Type 10 Type 20M Type 30
CaO 62.60 63.13 63.25 Total heat of hydration (kJ/kg) 331 302 380
MgO 2.74 0.98 1.26 Thermal conductivity (kJ/m h jC) 8.1 8.1 8.1
SO3 3.08 2.78 3.80 Specific heat (kJ/kg jC) 1.2 1.2 1.2
K2 O 0.82 0.60 1.08 Total heat of hydration was measured and specific heat and thermal
Na2O 0.29 0.07 0.10 conductivity were estimated.
 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131 127

of water in sealed tanks) for reference samples. After steam

curing, all samples were left at room temperature, 100% RH.
After 1 year, half of the steam-cured samples and half of the
samples cured at room temperature were submitted to a steam
curing at 85 jC, for 1 month, to simulate a long period of
heating due to the presence of exothermic waste. The rate of
temperature rise was about 2 jC/h, slow enough not to
generate thermal gradients in the prisms to avoid micro-
cracking, and the samples were allowed to cool naturally at
100% RH after steam curing. These samples will be referred
to as ‘‘01’’ (samples cured at room temperature and submitted
to a late temperature cycle) and ‘‘11’’ (samples steam-cured
during hydration and submitted to a second temperature
cycle). The samples left at 20 jC will be referred to as
‘‘00’’ (samples left at 20 jC during the whole experiment)
Fig. 2. Different temperature cycles imposed to each mortar. Every mortar
and ‘‘10’’ (samples steam-cured during hydration only). undergoes four types of testing conditions: hydration at room temperature
All samples were left for two more years at room (cycles 00 and 01) or steam curing (cycles 10 and 11) for 1 week; second
temperature, 100% RH. The four different temperature steam curing at 85 jC (cycles 01 and 11; or not: cycles 00 and 10) for 1
cycles combinations applied to the samples are schematized month after 1 year of hydration at 100% RH.
in Fig. 2.
2.4. Measurements and analyses

Length changes of the specimens were recorded for 3

years. For each mortar and curing condition, average expan-
sion was computed from four different samples. Scanning
electron microscope (SEM) observations were performed on
fractured samples.

3. Results

3.1. Effect of a high temperature due to heat of hydration

This part deals with the expansion of the samples

submitted or not to the temperature cycle during hydration
and that have not undergone a second cycle of temperature
(cycles 00 and 10). It thus deals with the possibility of DEF
induced by the heat of hydration of the cement itself.
Fig. 3 shows the length changes of the samples of mortar
types 30, 10, and 20M which have been steam-cured or not
during hydration.
Mortars type 10 and 30 which have been steam-cured
during hydration (cycle 10), present high levels of expansion
(up to 2.2% for mortar type 30; see Fig. 3(a); and 1.7% for
mortar type 10; see Fig. 3(b)) after an induction period of
about 100 days. After 3 years, mortar type 20M shows no sign
of expansion. None of the mortars that have been hydrated at
room temperature (cycle 00) present significant expansion
(Fig. 3(a –c)). SEM studies on these mortars revealed no
alkali– silicate reaction features, such as siliceous gel. DEF
features, such as ettringite deposits around aggregates and in
air voids, are detected on the two mortars that have expanded
Fig. 1. Comparison of the temperature cycles at the center of a 1-m wall of
mortar and the cycles imposed to the mortar prisms. (a) Temperature at the
(see Fig. 4). These artifacts are not believed to be the cause of
center of the 1-m wall depending on the mortars, simulated from calorimetric expansion [3,5] but are systematically observed on mortars
measurements. (b) Temperature cycles imposed to the three different mortars suffering from DEF. Similar secondary ettringite formation in
during hydration. voids (e.g., by Ostwald ripening) is often observed in mature
128 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131

Fig. 3. Expansions of the three different mortars submitted or not to steam curing during hydration (cycles 00 and 10). (a) Expansion of type 30 mortar (cycle
00 and 10). (b) Expansion of type 10 mortar (cycle 00 and 10). (c) Expansion of type 20M mortar (cycles 00 and 10).

concretes, but none of these artifacts could be found in the The absence of expansion of mortar type 20M does not
mortars that did not expand. show that this mortar would not expand if it was subjected to
These results are in accordance with what is expected for a ‘‘standard’’ laboratory steam curing, since the maximum
such mortars if they had been steam-cured according to a temperature reached by this mortar is about 75 jC, which is
laboratory precast mortar steam cycle: a higher temperature close to the threshold temperature 70 jC often considered as
of cure leads to higher levels of expansion [1,3,14] and the minimum temperature to trigger DEF [5].
mortars made with cements with a high level of sulfate The main finding of these first results is that the core of a
(types 10 and 30, as compared to type 20M) are more liable large concrete member that would undergo conditions sim-
to expand because of DEF [1,3,15 –18]. ilar to those applied to the mortars of the present study may

Fig. 4. Ettringite deposits around an aggregate of mortar type 10 steam-cured during hydration. Such deposits are not an evidence for DEF, but are
systematically found on mortars suffering from DEF.
 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131 129

expand if the cement is DEF-prone. However, such a thermal release significant enough to allow DEF or will they delay
profile would hardly be obtained, even in the worst scenario expansion indefinitely?
(large concrete section, excess of cement, high external
temperature). The results of this study could help in under- 3.2. Effect of a long steam-curing cycle on 1-year-old
standing the few field cases reported in the scientific docu- mortars
mentation [4,8,9]. The conditions under which the mortar
may or may not expand are still to be found, since this work This section deals with the effect of a temperature cycle
gives no information on the effect of different parameters, (85 jC steam curing for 1 month) on mature mortars that
which have been studied for more classical steam-curing have or have not been steam-cured during hydration (cycles
cycles. These questions are, considering a temperature cycle 01 and 11). Fig. 5 shows the length changes of the samples
due to the heat of hydration, the following: of mortars type 30, 10, and 20M before and after the second
temperature cycle.
– What would happen if the geometrical configuration Mortar type 30 presents very significant expansion after
would lead to higher/lower temperatures? having been steam-cured at 85 jC for 1 month. Whether the
– What would happen if the mortar would stay a shorter/ mortar had been steam-cured during hydration, the type 30
longer time at high temperature? mortar expands by more than 1% (1.6% for cycle 01, 1.2% for
– What would happen if the rates of temperature rise and cycle 11). The maximum rates of expansion are similar in
fall would differ, due to a different geometry of the con- both cases and similar to the expansion after the first steam
crete member? curing. The main difference between the two behaviors is that
– Water availability and alkali release have been found to be the expansion of the mortar that had been steam-cured during
an important parameter altering the rate of expansion and hydration does not present the 100-day induction time of
the final expansion [5,12]. Since high temperatures due to cycle 01. Whether the origin of this behavior is mechanical
heat of hydration can only be obtained in very large (an already cracked paste is more susceptible of undergoing
concrete members, are the water availability and the alkali expansion) or microstructural (the presence of C3S hydrated

Fig. 5. Expansions of the three different mortars steam-cured 1 year after hydration (cycles 01 and 11, second cycle represented by symbol 1). (a) Expansion of
type 30 mortar (cycles 01 and 11). (b) Expansion of type 10 mortar (cycles 01 and 11). (c) Expansion of type 20M mortar (cycles 01 and 11).
130 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131

at high temperature during hydration could play a predomi- gite. These considerations favor, in the case of DEF induced
nant role in the expansion mechanism [5]) is not determined. by a late thermal cycle, the hypothesis of the conversion
Mortar type 10 presents similar behaviors, but the ex- from hydrated sulfoaluminate (probably calcium monosul-
pansion rates are slower than when induced by a steam cure foaluminate, stable at high temperature) to ettringite.
during hydration and the incremental expansions are smaller
(0.65% cycle 01, 0.35% cycle 11). The induction period 4.2. Possible objections to the relevance of these results
after the second steam curing is a bit shorter for the mortar
that has been steam-cured during hydration. These results seem to show that large concrete members,
The behavior of mortar type 20M is drastically different. reaching high temperatures because of the heat of hydration
It has been seen earlier that this mortar did not expand even of the cement, may suffer DEF with time. However, the
if steam-cured up to about 75 jC during hydration (cycles laboratory experimental conditions applied in this work
00 and 10, Fig. 3(c)). As for mortar types 30 and 10, mortar differ slightly from what would happen in the bulk of a
type 20M subjected to a steam curing at 85 jC for 1 month a large concrete volume. The main differences are the result of
year after hydration (cycle 01) begins to expand significant- the small dimension of the mortar samples used in this
ly (0.5% at 950 days) after a short induction period of about study:
100 days. On the contrary, mortar type 20M, which has been
steam-cured during hydration, presents no sign of expansion – The temperature cycles during hydration were shortened
450 days after the 85 jC steam curing. For unknown as compared to what was given by simulations. Longer
reasons, expansion might be delayed, or inhibited, by the curing times might have suppressed expansion, as
steam cure during hydration. reported [1,3]. However, long curing at 85 jC (1 month
after 1 year of hydration) does produce expansion
in at least one testing condition for the three mortars
4. Discussion tested.
– Alkali leaching, which controls expansion due to DEF
4.1. Concordance of these results with proposed mecha- [12], is very fast on small laboratory samples. This
nisms for DEF leaching of Na + and K + ions would be much slower
on larger concrete members. This might delay, or even
The experimental results of this work seem to broaden inhibit, expansion of large concrete members. Water
the conditions needed to initiating DEF symptoms on ingress would also be slower on large sections.
mortars, since the temperature generated by the heat of – Apart from alkali leaching, mechanical dimensional
cement hydration can trigger DEF. However, it seems that effects might affect the way a large structure reacts as
the conditions that must be met by the mortar are the same compared with a small sample.
as for a classical steam-cured mortar: – CO2 has been suspected of being a triggering agent for
DEF [19]. During steam curing and conservation, the
– None of the samples kept at room temperature expanded. samples are in contact with CO2 of the air. This CO2 can
This result is in accordance with the fact that mortars that diffuse deeply in the samples because of their small
have not been submitted to temperatures above c 70 jC dimensions, and interact with ettringite or simply change
do not develop DEF. the chemical equilibrium of the system. This might
– The cements that are more DEF-prone (in this work, happen much slower on large concrete members.
types 10 and 30 cements) develop DEF more easily than
other cements (cement type 20M for the present work)
and present higher rates of expansion and higher final 5. Conclusions
expansions.
The main conclusions of this 3-year study are the
The results of this study show that DEF can develop after following:
long steam curing on mature mortars. This consideration
excludes a few DEF mechanisms that have been proposed – Some mortars, when exposed to a temperature cycle
before. First, DEF is not only a hydration-related phenom- simulating the temperature induced by the heat of
enon, and the high sulfate concentrations in solution during cement hydration, can expand after an induction period.
hydration are not required to trigger DEF. High temperature Final expansions can reach a few percents after 1 year.
may be sufficient to explain the higher sulfate concentrations From these results, one can imagine that under certain
of sulfate in solution. Secondly, DEF has been explained by conditions, large concrete members may suffer DEF
the positive volume change from unhydrated cement alumi- because of the high temperature reached at the core of
num-bearing phases to hydrated ettringite in the cement the concrete member during hydration. More work is
paste. In 1-year-old mortars, all C3A is hydrated, so the necessary to confirm this conclusion and evaluate the
expansions cannot result from conversion of C3A to ettrin- possible impact of the size of the structure on the
 R. Barbarulo et al. / Cement and Concrete Research 35 (2005) 125–131 131

phenomena—for instance, the effective temperature in microstructural and microanalytical study, Adv. Cem. Res. 5 (20)
(1993) 139 – 146.
the structure during hydration, the water ingress in the
[7] C. Famy, K.-L. Scrivener, A.-R. Brough, A. Atkinson, E. Lachowski,
structure, or the rate of alkali leaching. Characterization of calcium-silicate-hydrate products in expansive and
– One-year-old mortars, when subjected to a 1-month non-expansive heat-cured mortars—Electron microscopy study, ACI
steam curing at 85 jC, can expand with similar features SP-192, vol. 1, American Concrete Institute, Farmington Hills, MI,
as DEF-affected mortars. Mortars that have previously 2000, pp. 385 – 401.
expanded because of DEF expand again. Mortars that [8] L. Divet, F. Guerrier, G. Le Mestre, Existe-t-il un risque d’attaque
sulfatique endogère dans les pièces de béton de grande masse? (Is
have not suffered from DEF or that have been hydrated at there a risk of internal sulfate attack on massive concrete members?)
room temperature can expand as well. No such result has (in French), Bull. Lab. Ponts Chaussées 213 (1998) 59 – 72.
been reported before and allows to conclude that DEF is [9] D.-W. Hobbs, Expansion and cracking in concrete associated with
not a phenomenon only related to a high temperature delayed ettringite formation, in: B. Erlin (Ed.), ACI SP-177, Ettrin-
during hydration. It also appears that an early mild heat gite—The Sometimes Host of Destruction, American Concrete Insti-
tute, Farmington Hills, MI, 1999, pp. 159 – 181.
treatment can suppress (or delay significantly) expansion [10] M. Paul, F.-P. Glasser, Impact of prolonged warm (85 jC) moist cure
due to heating at later age. The conditions of success of on Portland cement paste, Cem. Concr. Res. 30 (2000) 1869 – 1877.
this process are to be determined. More work is needed to [11] R. Yang, C.-D. Lawrence, J.-H. Sharp, Delayed ettringite formation
determine the exact mechanisms of DEF. The results of in 4-year old cement pastes, Cem. Concr. Res. 26 (11) (1996)
the present study should help develop new tests and 1649 – 1659.
[12] C. Famy, K.-L. Scrivener, A. Atkinson, A.-R. Brough, Influence of the
models for the study of the occurrence and mechanisms storage conditions on the dimensional changes of heat-cured mortars,
of DEF. Cem. Concr. Res. 31 (2001) 795 – 803.
[13] Quadrel Software Technical Booklet, Digital Site Systems, Suite 315,
4516 Henry Street, Pittsburgh, PA 15213.
References [14] Y. Fu, J. Ding, J.-J. Beaudoin, Expansion of Portland cement mor-
tar due to internal sulfate attack, Cem. Concr. Res. 27 (9) (1997)
[1] C.-D. Lawrence, Delayed ettringite formation: an issue? in: J. Skalny, 1299 – 1306.
S. Mindess (Eds.), Materials Science of Concrete IV, American Ce- [15] H.-F.-W. Taylor, Delayed ettringite formation, Adv. Cen. Concr. 5 (20)
ramic Society, Westerville, OH, 1995, pp. 113 – 154. (1994) 122 – 131.
[2] Y. Fu, Delayed ettringite formation in Portland cement products, PhD [16] I. Odler, Y. Chen, Effect of cement composition on the expansion
thesis, CNRC, Ottawa, Canada, 1996. of heat-cured cement pastes, Cem. Concr. Res. 25 (4) (1995)
[3] C. Famy, Expansion of heat-cured mortars, PhD thesis, Imperial Col- 853 – 862.
lege, London, UK, 1999. [17] I. Odler, Y. Chen, On the delayed expansion of heat-cured Portland
[4] L. Divet, Les réactions sulfatiques internes au béton: contribution à cement pastes and concretes, Cem. Compos. 18 (1996) 181 – 185.
l’étude des mécanismes de la formation différée d’ettingite (Internal [18] S. Kelham, The influence of cement composition on the volume sta-
sulfate reactions in concrete: study of the mechanisms of delayed bility of mortar, in: B. Erlin (Ed.), ACI SP-177, Ettringite—The
ettringite formation) (in French), PhD thesis, Conservatoire National Sometimes Host of Destruction, American Concrete Institute, Farm-
des Arts et Métiers, 2001. ington Hills, MI, 1999, pp. 27 – 45.
[5] H.-F.-W. Taylor, C. Famy, K.-L. Scrivener, Delayed ettringite forma- [19] H.-J. Kuzel, Initial hydration reactions and mechanisms of delayed
tion, Cem. Concr. Res. 31 (2001) 683 – 693. ettringite formation in Portland cements, Cem. Concr. Compos. 18
[6] K.-L. Scrivener, H.-F.-W. Taylor, Delayed ettringite formation: a (1996) 195 – 203.