Construction and Building Materials 127 (2016) 918–931

Contents lists available at ScienceDirect

Construction and Building Materials

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

Advanced testing and performance specifications for the cementitious

materials under external sulfate attacks
Georges Massaad a,b, Emmanuel Rozière a, Ahmed Loukili a,⇑, Laurent Izoret b
a
LUNAM Université, Institut de Recherche en Génie Civil et Mécanique (GeM), UMR-CNRS 6183, Ecole Centrale de Nantes, 1 rue de la Noë, 44321 Nantes, France
b
Association Technique de l’Industrie des Liants Hydrauliques (ATILH), 7 place de la Défense, 92974 Paris La défense, France

h i g h l i g h t s

Development of a monitoring strategy.

Decoupling of the sulfate attack phenomena.

Better understanding of micro-degradation mechanism.

Proposition of new performance indicators.

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

Article history: A new monitoring approach was developed to enhance the analysis of performance tests and provide new
Received 7 April 2016 sulfate resistance criteria. Even in the case of low degradation rates or non-swelling materials exposed to
Received in revised form 1 September 2016 external sulfate attacks, the method allows detecting microscopic evolutions via a set of easily accessible
Accepted 28 September 2016
monitored parameters. The monitoring has been performed on a set of sulfate-resisting and non sulfate-
resisting cement-based mortars. The main phenomena induced by external sulfate attack are leaching,
precipitation, aggregates loss and cracking. Based on the monitoring of mass, hydrostatic weighing,
Keywords:
elongation and the amount of leached OH the method provides the samples dimensional variations,
Durability
Cement
the volume and mass changes of minerals and free water, corresponding to the main phenomena of
Sulfate attack the external sulfate attack. On the one hand, the phenomena decoupling allows a better understanding
Leaching of the sulfate attack mechanism. The observed equivalence between the volumes of leached and precip-
Expansion itated products highlights the influence of portlandite, the major leached mineral, on the precipitation
mechanism. The precipitation-deformation correlation and the samples absolute deformation monitoring
showed a common expansion mechanism for the different cement compositions. The cement type was
found to influence the magnitude of deformations. On the other hand, the determination of the volume
variations corresponding to each phenomenon allows the proposition of new performance indicators
from the monitoring approach outputs: the averaged density, the deformation path and the expansion
potential. The three indicators were sensitive to the microstructural and macroscopic evolutions. The
expansion potential, related to crystallization pressure, was found to depend on cement composition.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
External sulfate attacks are caused by the ingress of sulfate ions
from the environment. They result in an increase in porosity, local
expansion and cracking. As a consequence the mechanical proper-
ties and the durability of cement-based materials are affected
[28,23]. Sulfate attacks have been taken into account in the design
of concrete mixtures since the 1920s and the first standards on
cement. The so-called ‘‘sulfate resisting” cements have generally
⇑ Corresponding author.
E-mail address: Ahmed.loukili@ec-nantes.fr (A. Loukili).
shown good behaviour on site, as very few cases of external sulfate
attacks have been reported. However the composition of cements
has been changing for the last decades, in order to design more
sustainable construction products. Changes in clinker content or
chemical composition are likely to affect the sulfate resistance of
cement-based materials [14]. A good understanding of the mecha-
nism of degradation, reliable tests, and relevant indicators are thus
needed to assess the durability of new cements and concrete
exposed to external sulfate attacks.
The mechanism of degradation of cement-based materials is
now relatively well known [12,15]. The ingress of external sulfate
solution modifies the chemical equilibrium between solid phases

http://dx.doi.org/10.1016/j.conbuildmat.2016.09.133
0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 919

and interstitial solution in the cementitious matrix [7,10,24]. The
increase of sulfate concentration in interstitial solution results in
the formation of ettringite (or alumino ferrite trisubstituted AFt)
from monosulfoaluminate 2CaOAl2O3Ca2SO412H2O (alumino fer-
rite monosubstituted AFm). Expansion is normally associated to
ettringite formation [19,18], even gypsum formation could lead
to samples swelling [16]. The theory that is consistent with most
of experiments and observations is the crystal growth pressure.
The crystal growth theory describes the crystallization pressure
produced by supersaturation and confinement [8,9,27]. According
to this theory, the driving force for the crystallization pressure is
the supersaturation of the pore solution with respect to ettringite.
The formation of ettringite does not necessarily result in expansion
or damage. This is supported by numerous field observations in the
past in which no correlation between the amount of formed ettrin-
gite and observed expansion. Significant expansion pressure higher
than tensile strength can only be exerted by ettringite which forms
in small pores within a certain size range [9], approximately 10–
50 nm, depending on the cement used. Microstructural studies
[3] showed that penetrating sulfates first react with monosulfate
in pockets which does not lead to stress or expansion. On depletion
of the pockets content, monosulfate which is finely intermixed in
C-S-H then reacts to form expansive ettringite. The scenario of sul-
fate attack scenario has been confirmed by the zonation observed
on damaged samples [1,21]. These experimental results confirmed
that sulfate attack do not only cause expansion but also softening
and decohesion. The formation of gypsum and ettringite requires a
source of calcium. Calcium can be provided by the leaching of port-
landite and C-S-H. The leaching of the portlandite and the progres-
sive decalcification of C-S-H cause a softening of cement matrix
and a decrease of strength [4].
Performance-based specifications and methodologies of qualifi-
cation of cement-based materials are often associated to testing
procedures. Laboratory conditions should be representative of field
exposure. The typical concentration in sulfate-rich environments is
0.02–3 g/L. Concrete structures are partially or fully saturated, or
exposed to drying and wetting cycles [9,1,13]. Previous studies
have also shown the influence of counter ions, temperature, and
pH control. New experimental procedures have been developed
to provide more representative sulfate resistance tests [5,22].
Most of performance tests and criteria are based on length mea-
surements and expansion limits (ASTM C 1012, 2000). Expansion
data do not provide enough information to assess the chemical
reactions and understand the mechanisms causing damage. More-
over, expansion can be caused by other phenomena [29], and the
phenomena occurring during the first stage of sulfate attack do
not result in significant expansion [21,25].
This paper presents a new monitoring approach aimed at
enhancing performance tests and criteria. Based on some relatively
accessible monitoring parameters such as mass, leaching, and
hydrostatic weighing, we deduced the mass and volume variations
associated to chemical changes and cracking to monitor sulfate
attack. The results of this analysis were directly used to define
new performance indicators to assess the durability of the new
construction materials. This new approach could be the way to cor-
relate the chemical and mechanical evolutions and expansion
potential related to crystallization pressure development.
2. Experimental program
In order to develop a new sulfate attack monitoring strategy, a
double purpose experimental study was designed to understand
the scenario of sulfate attack and then find new monitoring param-
eters to improve existing performance tests and enhance their reli-
ability and relevance.
2.1. Materials and mixtures
The compositions of cements used in the experimental study
are given in Table 1. The NSR is a non sulfate resisting cement with
a relatively high C3A content (6.4 %), the two others cements SR5
and SR3 comply with criteria on sulfate resisting cements with rel-
atively low C3A contents, respectively 0.7 and 0.4%.
A mortar mixture was designed (Table 2) by adapting the com-
position defined in EN 196 standard. The paste volume, i.e. the sum
of water and cement volumes, was kept constant. The water-to-
cement ratio was increased from 0.5 to 0.6. French standard sand
(SNL, 0.08 mm – 1.6 mm size), according to the norm EN 196-1,
was used. It’s natural siliceous sand with generally isometric and
rounded particles shape [2].

Table 1
Cement compositions.

CEM I NSR 52.5 CEM I 52.5 SR5 PM CEM I 52.5 SR3 PM

Clinker (%) 91.9 92.8 94
Mineralogical composition of clinker C3S 67.9 72.9 67.4
C2S 13.1 19 16.1
C3A 6.4 0.7 0.4
C4AF 10.4 6.7 13.7
Lime/Portlandite 0.8 0.3 1.5
Alkali sulfates 1.5 0.3 1
Sulfatted additions Anhydrite 4.4 2.4 1.8
Bassanite 0.5 0.8 3.3
Gypsum 0.2 0.9 0.6
Chemical Composition of the cements PF950 2 1.5 1.4
SiO2 19.4 22.4 20.5
Al2O3 5.1 2.9 3.3
Fe2O3 2.9 2.3 5.2
CaO 63.1 67 64.8
MgO 1.8 0.8 0.7
SO3 3.6 2.3 2.8
K2O 0.9 0.2 0.7
Na2O 0.2 0.1 0.1
SrO 0.1 0.1 0
TiO2 0.3 0.1 0.2
P2O5 0.3 0.1 0.4
MnO 0.1 0 0.1
Cl 0.079 0.027 0.081
920 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931

Table 2 2.2. Laboratory conditions of sulfate attack

Mortar composition.

Water Cement
Content for 1 batch (g) 241 400
Tests were performed on mortar cylindrical samples of
20-mm diameter and 160-mm length. Mortar samples were
kept in mould for 48 h then in saturated limewater for
26 days before testing. Two samples were exposed to sulfate
attack and two control samples from the same batch were
immersed in tap water in order to deduce the influence
of hydration on the parameters monitored during sulfate
attack. This allows decoupling the hydration and sulfate
attack phenomena.
Sand 0/4 mm The laboratory procedure (Fig. 1) consists in immersing the
1350 mortar samples in a 3 g/l sulfate solution (SO24 concentration) at
controlled constant temperature (20 °C) and pH (7.5) [22]. The
pH control is made via nitric acid at 0.5 mol/l added automatically
to equilibrate the OH leached from the cementitious matrix. The
sulfate solution is changed every 20 ml of added acid. The temper-
ature control is provided by a temperature-controlled liquid, which
passes into the double-walled beakers containing samples.
2.3. Monitored parameters
After each renewal of sulfate solution, the same measurements
were made on samples exposed to sulfate solution and control
samples, namely: mass, length, and hydrostatic weighing (Fig. 2).

pH probe

Samples

Double wall
beaker
Nitric acid
0.5 M Magnet
bar

pH regulator Magnec Temperature

agitator regulator

Double wall
Samples
beaker
profile
Sulfate soluon:
pH = 7.5
T = 20 °C
Temperature [NaSO4] = 4.44 g/l
regulator liquid

Fig. 1. Sulfate attack device.

Balance
Water proof
resin layer

Steel Water
string
Sample

Mass Hydrostac weighing Elongaon

measurement measurement measurement
Fig. 2. Devices for macroscopic monitoring.
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 921

Hydrostatic weighing was carried out on samples immersed in
their solution. Then they were wiped until saturated surface dried
(SSD) state to measure the mass. Finally the length was assessed
using LVDT sensor and invar reference bar. The quantity of leached
OH- ions was directly assessed from the cumulated volumes of acid
added to control the pH of sulfate solution.
2.4. Thermogravimetric analysis
Thermogravimetry analyses (TGA) were carried out in a dry
nitrogen atmosphere with a SETARAM analyser. The dry pastes
were crushed to powders, and 90 ± 5 mg samples put in crucibles.
The sample were heated from 20 to 1100 °C at a constant rate of
10 °C/min. To determine the portlandite mass proportions in the
pastes, we applied tangent method [26]. At 425–550 °C the port-
landite decomposition occurs and the TGA slope changes due to
the water loss from the portlandite. The deduced portlandite
mass/volume in the paste is reported to the sample volume,
according the paste proportion in the mortar samples (Table 2).
3. Development of the monitoring strategy
Taking into account the different levels of dependence illus-
trated in Fig. 3, the combinations between the different parameters
established in the following section will allow solving the problem
gradually and eventually quantify changes entrained by each phe-
nomena on each monitoring parameter. Thus decoupling the four
phenomena provides a description of the sulfate attack scenario
and allows the definition performance-based specifications.
3.2. Development of macroscopic evaluation
The specimen volume and the cross section variation of the
specimen can be deduced from three parameters of the monitoring
strategy: the SSD mass, the hydrostatic weighing, and the length
(Fig. 4).
3.2.1. Determination of volume variations
The mass monitoring (1) provides information on the overall
change in the cement matrix in liquid phase (variation of pore
water content due to the change in porosity and cracking) and solid
phase (dissolution and precipitation of minerals). The hydrostatic
weighing monitoring (2) reveals the variation of the solid phase
of the cement matrix.

3.1. Principle of the method M ¼ M sol þ qwater  V liq ð1Þ

The four main phenomena involved in the external sulfate Phyd ¼ M sol  qwater  V sol ð2Þ
attack include precipitation and leaching of cementitious matrix
with M is the SSD mass [g], Phyd the hydrostatic weighing [g], Msol
minerals, aggregates loss and cracking. The parameters monitored
solid phase mass [g], V sol solid phase volume [mL], V liq liquid phase
at macroscopic and intermediate levels give the changes due to the
volume [mL] andqwater water density [g/mL].
phenomena assembled. For a better understanding of these 4 phe-
nomena and their effect on the monitored parameters, a monitor-
ing strategy and a new method of analysis were developed in order
macroscopic
to dissociate the changes caused by each phenomenon.
level
The monitoring strategy illustrated in Fig. 3 is limited to 4 mon-
itoring parameters. Some are directly related to one of the four intermidiate
phenomena mentioned above and others are related to two or level
more of these phenomena. For example, the evolution of the mass microscopic
is influenced by four phenomena. In case of compensation between
level mass
gain and loss of mass caused by leaching and precipitation of min-
erals for example, the global mass variation is not significant. monitoring

c
• leaching r
macroscopic a
level c
• aggregates leached OH-
hydrostatic
intermidiate loss k weighing
level i monitoring monitoring
microscopic • precipitation n
level g
mass
monitoring
leached OH-
c monitoring expansion
• leaching r monitoring
a
• aggregates c hydrostatic
loss k weighing
i monitoring
cross section
• precipitation n variation
g
monitoring

expansion volume
monitoring variation
monitoring

Fig. 3. Diagram of monitoring parameters and main phenomena involved in sulfate

attack. Fig. 4. Monitoring strategy at macroscopic level.
922 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931

Combining the two monitored parameters the water equivalent place) is calculated. Thus we consider that sulfate attack will not
mass of the specimen (3) can be assessed as a function of time, change the hydration, which is a simplifying assumption. The
then the global sample volume variation can be deduced from hydration of the outer layer of samples in contact with sulfate solu-
Eq. (4). tion is almost achieved during the curing phase; however the
hydration effect on monitored data is due to its continuation in
M  Phyd ¼ qwater ðV liq þ V sol Þ ¼ qwater  V t ð3Þ
deeper zones. These zones are affected later by the aggressive sul-
fate solution, when their hydration is achieved. This makes the
DV  qwater ¼ DM  DPhyd ð4Þ
continuation of hydration almost the same for the attacked and
control samples.
3.2.2. Variations of cross section Sulfate attack causes changes in the liquid and solid phase of
The relative variation of cross section radius is deduced from the sample, which makes the mass variation depending on
the relative elongation variation x and the relative volume varia- the variation of these two phases (6) respectively DM water:poro and
tion z monitored during the test (5). DMsol .
rffiffiffiffiffiffiffiffiffiffiffi
1þz DM ¼ DM water:poro þ DMsol ð6Þ
y¼ 1 ð5Þ
1þx
x ¼ DL0L, y ¼ DR0R and z ¼ DVV0 with V 0 initial global volume, L0 initial 3.3.2. Variations of liquid phase
Sulfate attack can be summarized by the four phenomena men-
sample length, and R0 initial radius of cross section.
tioned above. However their occurrence and their effects on the
variation of the liquid phase depend on the zone of the sample
3.3. Development of microscopic evaluation
and its position related to the surrounding aggressive solution
[11]. For a better description of the liquid phase changes 3 zones
3.3.1. Mass variation model
of the tested samples are defined in Fig. 6.
The mass variation during the sulfate attack is due to the leach-
Zone I is in direct contact with the outer sulfate bearing solu-
ing and precipitation of minerals, the aggregates loss, and the infil-
tion. It is the most degraded part of the specimen. In this zone
tration of water in the cracks generated by crystallization
leaching and loss of aggregates are the main phenomena. In Zone
pressures associated to the precipitation of expansive minerals.
II, leaching occurs but the pH is higher than in Zone I, thus the pre-
From experimentally collected data and other physical parameters
cipitation of expansive sulfate minerals takes place causing the
the mass variation is calculated taking into account all mentioned
cracking of the cement matrix under crystallization pressures.
phenomena in order to understand better the coupling of the phe-
Cracks observed in Zone I and II take origin from Zone III where
nomena during the sulfate attack mechanism.
the three others phenomena do not exist.
The following monitoring strategy is based on 3 monitored
Based on the four main phenomena involved in the sulfate
parameters (Fig. 5): mass, hydrostatic weighing, and leached
attack and the different zones previously defined, the expressions
OH. The aim is to assess the microscopic changes corresponding
of mass (7), hydrostatic weighing (8), and volume variation (9)
to the phenomena involved in sulfate attack.
are written.
The hydration of cement is not directly taken into account in
the analysis. In order to decouple hydration from the four other DM ¼ DM lea þ DMag:td þ DMpre þ DM cr ð7Þ
phenomena mentioned before, the difference between the moni-
with lea leaching, prec precipitation, ag.td. aggregates taking-down
toring data from sulfate attack test and control samples immersed
and cr cracking.
in tap water (where hydration is the main phenomenon taking
DMlea ¼ DV lea:I  qlea þ DV lea:II ðqlea  qwater Þ ð7:1Þ
Leached minerals in ZONE II are replaced by water which
macroscopic
level reduces the mass of the solid phase and increases the mass of
water. However in ZONE I the sample is at SSD state when the mass
intermidiate is measured. The volume of water corresponding to leaching is
level dried thus not considered in the mass balance.
microscopic
level DMag:td ¼ DV ag:td:I  qag ð7:2Þ
mass
monitoring The aggregate taking-down occurs especially in ZONE I. The
leached OH-
c water replacing aggregates is dried when measuring SSD mass, this
monitoring
• leaching r is why the mass variation is limited to the mass of taken-down
a aggregates.
• aggregates c hydrostatic
k weighing
DMprec ¼ DV prec:II ðqprec  qwater Þ ð7:3Þ
loss
i monitoring
Precipitated minerals increase the mass of the solid phase and
• precipitation n reduce the volume of free water in the pores.
g
DMcr ¼ DV cr  qwater ð7:4Þ
expansion Water fills the cracks generated during sulfate attack and
monitoring increases the SSD mass of the sample.
In the same way the hydrostatic weighing balance is estab-
lished. It is influenced only by solid phase variations, thus it is pos-
sible to decouple the variations of the solid and liquid phases by
combining the mass and the hydrostatic weighing expressions.
DPhyd ¼ DPhyd:lea þ DPhyd:ag:td: þ DPhyd:pre ð8Þ
Fig. 5. Microscopic monitoring via macroscopic and intermediate monitoring.
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 923

3 4
2
1
1 4 4
4
3 4
2
1 4
1 4
4 3 4
2 4
1
1 4
4 4
2 3 4
1
Initial mortar
1 4 4 matrix
4 3 4
2
1
SULPHATE - BEARING WATER

1 4 4
4
3 4
2 4
1
1 4
4 4
4
ZONE III
ZONE II
ZONE I

(1) Leaching (2) Aggregate

2 3 4 taking-down
1 1 4
4 4 4
2 3 4
1 1 4
4
4
4
2
3
1 4
4 1 4
4
2 4
1 3 4
4
1 4 (3) Precipitation (4) Cracking
4
2 4
3 4
1
4 1
4 4
4 Cement paste
Aggregates
Degradation Water Precipitated
specimen zones minerals

Fig. 6. Definition of degradation zones.

with Phyd the hydrostatic weighing, lea leaching, prec precipitation The volume variation calculation (9) obtained by combining the
and ag.td. aggregates taking-down weight and hydrostatic weighing highlights the volume variations
undergone by the solid part of the zone I (due to leaching and
DPhyd:lea ¼ DV lea ðqlea  qwater Þ
aggregates taking-down) and the variations of the free water vol-
¼ DV lea:1 ðqlea  qwater Þ þ DV lea:2 ðqlea  qwater Þ ð8:1Þ ume in the Zones I, II, III due to the infiltration of water in the
cracks. However the above expression is not sensitive to the local
Due to the minerals leached in ZONES I and II the hydrostatic
volumes changes taking place in Zone II more precisely variations
weighing decreases
due to leaching and precipitation of minerals. The cracks and the
DPhyd:ag:dt ¼ DV ad:dt:1 ðqag  qwater Þ ð8:2Þ pores are actually filled or widen by minerals precipitation or
leaching without a change in the overall volume. In order to take
The aggregates taken-down in ZONE I results in a decrease of
into account previously missed variations, the calculation of the
the hydrostatic weighing.
change of effective free water volume DVFW can be written as:
DPhyd:pre ¼ DV prec:2 ðqprec  qwater Þ ð8:3Þ
DV FW ¼ DV  DV lea:II  DV prec:II ð10Þ
The precipitation of minerals results in an increase in the mass
of the solid phase and the resulting hydrostatic weighing The free water volume variation (10) is in direct relation with
the variation of the porosity of the cementitious matrix during
DM  DPhyd
DV ¼ ¼ DV cr þ DV lea:I þ DV ag:td:I ð9Þ the sulfate attack, which is likely to influence the transport proper-
qwater ties during sulfate attack.
924 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931

The quantification of the volumes of minerals leached and pre-
cipitated in Zone II, and their influence on the volume of free water
is not possible with the developed monitoring strategy. Thus these
changes (Zone II) are combined with the variations due to cracking
determined experimentally (11). In this way the new integrated
ratio f lea:prec expresses the rate of cracks closing or widening due
to mineral modifications. For positive volume variations, when
f lea:prec < 1 we can say that the precipitation takes over leaching
phenomena to reduce the amount of free water in Zone II, however
for f lea:prec > 1 cracks widening takes place and leaching takes over
precipitation.
DMwater:poro ¼ f lea:prec  DV  qw
In the same way the precipitation mass variation is estimated
from experimental data and unknown physical factors measured
experimentally. So the hydrostatic variations corresponding to
the leaching and taken-down aggregates are calculated to get the
precipitation hydrostatic weighing variation DP hydr ¼ DPhydr;prec þ
DPhydr;lea þ DP hydr;ag:td . Two others unknown factors, the densities
of leached and precipitated minerals qlea and qprec, are introduced
to get mass variations from hydrostatic weighing.
DPhydr:prec
DMprec ¼  qprec ð12:3Þ
ðqprec  qwater Þ
With DPhydr:prec ¼ DP hydr þ NleaqMolea ðqlea  qeau Þ þ f gra:dec  NleaqMolea 
ð11Þ lea gra

3.3.3. Variations of solid phase
In the solid part, leaching and precipitation of minerals and
aggregates taking-down occur in the previously defined zones (12).
DMsol ¼ DMsol:lea:I:and:II þ DMsol:ag:td:I þ DMsol:prec:II

OH ions are released during the leaching of hydration prod-
ucts. The mass variation caused by leaching corresponds to the
number of leached OH Nlea multiplied by the unknown molar
mass Molea. The main minerals likely to leach release a number
of OH ions n per leached mole (Table 3). The apparent molar mass
corresponding to the leached minerals will be Moap ¼ Monlea where
the actual molar mass Molea can be estimated according to
(Table 3).
Molea Molea
DMlea ¼ Nlea:1   Nlea:2 
n n
Considering the aggregates taking-down directly related to the
amount of leached minerals and in absence of experimental data
quantifying the taken down aggregates, the unknown ratio fag.td
is defined from the calculated leaching mass variation in order to
estimate the aggregates taken-down mass variation.
Molea
DMag:dt ¼ f ag:td:  N lea 
n
ðqgra  qeau Þ
Combining the different mass variations corresponding to high-
lighted phenomena the mass variation expression (13) is obtained.
DPhydr:prec
DM ¼ Nlea  Moap  f TD :Nlea :Moap þ :q
ðqprec  qwater Þ prec
ð12Þ
þ f lea:prec  DV ð13Þ
3.3.4. Definition of the types of precipitation
During sulfate attack, as the sulfate ions reach the reactive
phases, precipitation products appear. Sulfate bearing minerals
precipitate in the pores and the leached zones of the chemically
affected region of the sample. They exert a crystallization pressure
on the walls of the filled zones, generating the gradual expansion of
the specimen. Three types of precipitated volumes can be distin-
guished: the global precipitation DVglo.prec, the filling precipitation
ð12:1Þ
DV.fil.prec, and the cracks precipitation DVcr.prec (Fig. 7).
The globally measured volume of the sample is the volume
obtained after the aggregates loss (volume loss) and the precipita-
tion of expansive minerals (volume gain) during the sulfate attack
(Fig. 8). From the different volume variations corresponding to dif-
ferent phenomena (Fig. 12), the absolute expansion of the speci-
men can be deduced according to the following equation:
ð12:2Þ DV cr:prec ¼ DV global þ jDV agr:loss j ð14Þ

Table 3
CEM I paste mineralogy (BRGM database [17]).

Mineral Formulation Molar mass (g/mol) Density (g/ml)

CSH(1.6) Ca 1.6 SiO 3.6 : 2.58 H2O + 3.2(H+) M 1.6(Ca2+) + 1(H4SiO4) + 2.18(H2O) 196.3 2.32
Ettringite Ca6Al2(SO4)3(OH)12 : 26 H 2 O + 12(H+) M 6(Ca2+) + 2(Al3+) + 3(SO2 4 ) + 38(H2O) 1255.1 1.76
Gypsum CaSO4 : 2 H2O M 1(Ca2+) + 1(SO2 4 ) + 2(H2O) 172.17 2.3
Hemicarboaluminate Ca8Al4CO16 : 22 H2O + 27(H+) M 8(Ca2+) + 4(Al3+) + 1(HCO 3 ) + 35(H2O) 1128.9 1.98
Hydrotalcite Mg4Al2O7 : 10 H2O + 14(H+) M 4(Mg2+) + 2(Al3+) + 17 (H2O) 443.4 1.95
Monosulfoaluminate Ca4Al2SO10 : 12 H2O + 12(H+) M 4(Ca2+) + 2(Al3+) + 1(SO24 ) + 18(H2O) 622.5 1.99
Portlandite Ca(OH)2 + 2(H+) M 1(Ca2+) + 2(H2O) 74.1 2.24

crack
Vacancy
Expansive
zone
minerals

Fig. 7. Precipitated minerals categories.

 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 925

Calculated (%)
In
Inial secon
-1.0 -0.7 -0.4 -0.1

-0.1

2: Aggregates loss Obtained secon NSR

deformaon
SR5 -0.4

Measured (%)
SR3
1: Precipitaon
deformaon

Fig. 8. Evolution of specimen section during sulfate attack. -0.7

-1.0

Fig. 11. Calculated mass vs measured mass variation.

Deducing the absolute swelling volume variation DVcr.prec from

the global precipitated volume variation DVglo.prec presented at
Fig. 12, we get the filling precipitation volume variation DVfil.prec
(15) or the precipitation volume filling vacant spaces (pores or lea-
ched zones), triggering the expansion. In Eq. (15) we consider that
micro-cracks generated especially at the first stages of the external
sulfate attack and causing the samples swelling (DVcr.prec) are fully
filled by expansive sulfate minerals. At the advanced stages of
degradation, in presence of macro-cracks, the assumption of filled
cracks becomes less relevant and the accuracy of the calculated
DVfil.prec is reduced.

DV fil:prec ¼ DV glo:prec  DV cr:prec ð15Þ

3.4. Parameters calculation

The monitoring strategy is based on the four main chemical and

σexp: compressive stress in expansion zone
σten: generated tensile stress at the sample core
Ften: tensile load induced by precipitation volume
Fig. 9. Expansive pressure calculation concept.
With DVcr.prec the volume variation due only to the precipitation
(including cracks), DVglobal the global resulting volume variation
(including all phenomena) and DVagr.loss the volume variation due
to the aggregates loss mainly occurred at the surface of the sample.
mechanical changes encountered during sulfate attack. To deter-
mine the values of the physical parameters integrated in the
model: Molea, qlea, qprec, flea.prec and fag.td, we proceeded by cali-
brating the calculated mass variation on the experimental mass
variation curve.
By minimizing the sum of squared differences (16) between the
theoretical points and the experimental mass variations, the phys-
ical parameters corresponding to the best fitting curve could be
assessed.
X
R¼ ðx  yÞ2 ð16Þ
x is the measured mass variation and y is the calculated mass
variation.

NSR SR5 SR3

Variaon (%)

Days Days Days

Fig. 10. Influence of cement type on the geometrical variations of samples.

926 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931

0.0
SR3
-2.0 NSR 1.4 NSR

Variaon (%)
SR5
-4.0 0.7

-0.1
SR5
-6.0
SR3
Leaching Free water
-8.0 -0.8
0 40 80 120 160 200 0 40 80 120 160 200

8.0 0.0

SR5
6.0 -2.0 SR3
Variaon (%)

NSR
SR5
4.0 -4.0
SR3 NSR
2.0 -6.0

Precipitaon Agregates loss

0.0 -8.0
0 40 80 120 160 200 0 40 80 120 160 200
Days Days

Fig. 12. Relative volume variations corresponding to four sulfate attack phenomena.

3.5. New performance indicators

3.5.1. Averaged density indicator

In order to gather the information collected at the macroscopic
Fil.prec (%)

and microscopic levels in a single indicator deduced from the mon-

itoring of performance tests, the ‘‘averaged density” indicator was
defined. The sum of the absolute mass variations generated by the
four phenomena of sulfate attack was divided by the sum of
corresponding absolute volume variations. In this way, the
obtained values at different times of sulfate attack test highlight
the prevailing material and the corresponding phenomena causing
the variations of mass and volume of cementitious materials
(Fig. 15). This allows describing the scenario of sulfate attack using
Leaching (%) a single indicator (17).

Fig. 13. Correlation between filling precipitation Fil.prec and leached volume.

0.09 NSR SR5 SR3

L (%)

0.06

0.03

0
1.5

1
r (%)

0.5

0
0 1 2 3 0 1 2 3 0 1 2 3
Fil.prec (%) Fil.prec (%) Fil.prec (%)

Fig. 14. Absolute longitudinal (DL) and lateral (Dr) deformation as a function of filling precipitation Fil.prec.
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 927

2.4 of the deformations. This approach is useful when testing low-

expansion materials, because lateral deformation was found more
sensitive to sulfate attack than longitudinal expansion.
SR3
Averaged density (g/ml)

2.3

3.5.3. Expansion potential indicator

SR5 In order to estimate the expansion potential of cement-based
2.2
materials in sulfate-rich environments, we introduce in this sec-
tion a procedure to calculate the expansive pressure, combining
the developed tensile stresses in the samples during the sulfate
2.1
NSR attack with the volume of sulfate bearing precipitated minerals.
The ‘‘hinge point” defined on the graphs of expansion variation
(Fig. 17) corresponds to the tensile stress capacity of samples,
2.0
0 50 100 150 200 reached at the values varying between 0.01 and 0.02 % [6,20].
Days Based on compressive strength of the mortars, determined on mor-
tar cubes, we deduced the tensile stress strength ft of mortar tested
Fig. 15. Averaged density variations.
samples according to the Eurocode 2 expressions : rten.NSR = 3.75,
rten.SR5 = 3.85 and rten.SR3 = 3.8 MPa.
The volume of expansive minerals is first represented as an
equivalent volume cube (Fig. 9). In order to assess an equivalent
2.2 expansive pressure, the cube is assumed to be confined like the
sulfate minerals formed in the pores causing expansion. Conse-
NSR quently the volume increase of the precipitated minerals results
1.5
in stresses in the sample until reaching the tensile strength, corre-
Δr (%)

SR5
sponding to hinge point on the longitudinal expansion graphs
0.8 (Fig. 17). In order to calculate the equivalent expansive pressure
SR3
(18), the sides of the equivalent confined cube Shinge are assumed
to be submitted to the resulting tensile load Ften.hinge developed
0.1
0 0.1 0.2 0.3 0.4 0.5 by the core of the sample at thinge (Fig. 9).
ΔL (%) Ften:hinge
EPI ¼ ð18Þ
Fig. 16. Expansion path. Shinge
At thinge the degraded depth is calculated to get the resisting
section of the sample which holds the tensile force generated in
X4
jDmi j the samples core due to the precipitation in outer expansive zones.
AD ¼ Xi¼1 ð17Þ Based on the calculated leached volume variation (Fig. 12), we
j Dv i j
4
i¼1 reach back to the corresponding Portlandite-free degraded depth
(Portlandite is the main leached mineral). Then the resisting sec-
with AD the averaged density indicator, |Dmi| and |Dvi| the elemen-
tion can be deduced and the tensile load corresponding to the ten-
tal absolute mass and volume variations corresponding to the four
sile strength can be determined for each tested material.
sulfate attack phenomena.

3.5.2. Deformation path indicator 4. Results and discussions

The absolute radius variation, i.e. the lateral deformation, can be
deduced from the absolute volume (DVcr.prec in Eq. (15)). The After describing the approach used to identify leached, precipi-
length variations of the cylindrical samples were directly assessed tated minerals, and other parameters representative of the mecha-
from experimental data. The deformation path corresponds to the nisms of sulfate attack, the results are analysed and discussed to
lateral vs. longitudinal deformation graph. It makes it possible to provide more information on the microscopic evolution during sul-
classify cement-based materials as a function of the magnitude fate attack as well as performance indicators.

1 NSR SR5 SR3

0.1
L (%)

0.01
thinge thinge thinge
0.001

8 glo.prec
glo.prec
6
glo.prec
4
fil.prec fil.prec
2 fil.prec

0
0 50 100 150 200 0 50 100 150 200 0 50 100 150 200
Days Days Days

Fig. 17. Precipitation volumes determination (DVglo.prec and DVglo.prec) at thinge for the expansion potential indicator calculation.
928 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931
4.1. Macroscopic evolutions
The first output of the monitoring strategy is the sample volume
and the lateral cross section variations (radius). The most widely
monitored geometrical parameter is the longitudinal expansion.
However, in the case of non-swelling material exposed to sulfate,
it does not change significantly. Thus expansion cannot capture
some changes induced by the degradation, whereas the volume
and radius sample monitoring allows assessing leaching activity
and giving the corresponding geometrical variations.
In the Fig. 10, the NSR cement containing the highest C3A per-
centage presents the highest length expansion rate. The SR5 and
SR3 cements showed lower expansion rates since they contain less
C3A, thus less expansive products can be formed. Moving to the
radius and volume variations, we can distinguish two different
behaviours. The first one corresponds to NSR mortar. Even if leach-
ing and aggregates loss take place at the sample surfaces and the-
oretically leads to the radius reduction, the radius actually kept
increasing due to radial expansion and lead to a positive global vol-
ume variation. The second behaviour corresponds to SR5 and SR3
mortars. The material lost at the surface could not be compensated
by the expansion of the mortars, which was lower than NSR expan-
sion. The radius showed a negative variation at the beginning of
the tests for SR5 then it started to increase, which means that first
leaching then expansion prevailed. SR3 mortars showed the lowest
expansion rate and a continuous decrease of radius. The global vol-
ume variations of the three materials followed the radius varia-
tions, which suggest that lateral variations of dimensions had
stronger effects than the longitudinal variations.
4.2. Microscopic evolutions
4.2.1. Discussion on the optimization of parameters
In the previous sections the mass balance was established by
distinguishing the mass variations due to each phenomenon. The
optimisation (Fig. 11) procedure, detailed in the appendix, provides
the values of parameters needed to determine the contributions of
each phenomenon to the mass and volume variations of the
samples (Table 5). Scrolling down the three sets of optimisation
cartographies from the NSR to the SR3 cement containing respec-
tively the highest and the lowest calcium aluminates percentages,
the localisation of red zone varies (Fig. A.1), which reveals varia-
tions in the products affected by sulfate attack. At relatively low
pH, portlandite is the main leached hydration product. Due to
the chemical equilibrium between the cementitious matrix and
Table 4
Variation range of physical parameters.
Parameters Intervals
qlea (g/ml) [1.1, 2.3]
qprec (g/ml) [1.1, 2.3]
Moapp (g/mol) [20, 100]
f lea:prec [0.0, 3.0]
f ag:td [0.0, 3.0]
Table 5
Best fit physical parameters values.
NSR SR5
qlea (g/ml) 2.30 2.27
qprec (g/ml) 2.00 2.00
Moapp (g/mol) 37 30
f lea:prec 0.70 0.40
f ag:td 2.1 1.6
the interstitial water, its dissolution actually maintains the basicity
and the cement paste. Considering that portlandite is the major
leached mineral would lead to the same molar mass as a solution
(red zones positions) for the three different mortars. However
the NSR mortar had higher Molea/n value than the other mortars
(see first column of the Fig. A.1). The dissolution of portlandite
releases two OH ions, thus the corresponding value of Molea/n
would be 37 g/mol. The SR5 and SR3 mortars showed lower values,
which suggests that other hydration products have been leached
(Table 5). The three cements had approximately the same density
of the leached mineral (2.3 g/ml), which corresponds to the
CSH – Portlandite interval (Table 5).
In the third column of Fig. A.1 the NSR and SR5 mortars shows
slightly lower values than the SR3 mortar. The first two cements
compositions actually promoted the ettringite formation in the
paste matrix during sulfate attack. Other sulfate-bearing phases
were formed, such as gypsum. The resulting density is around
2.0 g/ml (Table 5), which is intermediate value between ettringite
density (1.76 g/ml) and gypsum: density (2.30 g/ml). The SR3 mor-
tar showed higher density (2.1 g/ml) of precipitated phase. As a
consequence the gypsum/ettringite ratio of precipitated products
increased with a decrease in C3A content. The formation of gypsum
can be explained by the influence of pH profile within the speci-
men. In the zones where Portlandite is not completely leached,
the pH is high enough to ensure the stability of ettringite. In the
zones closer to the surface, where portlandite has been completely
leached, the pH is lower and it tends to the pH of the external sul-
fate solution (7.5), thus only gypsum is stable.
In the fourth column, the large red zones indicate a much less
clear evolution of the optimal value of the aggregate loss parame-
ter. With the four optimal known parameters, the chosen combina-
tions contain different values for fag.td. Looking for the optimal
solution, the chosen combinations of parameters include the most
convenient fag.td value giving the lowest offset between the exper-
imental and theoretical mass variation curves.
In the fifth column, the optimal solution zones should be related
to the global volume variation DV for a better understanding of the
physical evolution in the samples. According to the Eq. (11) if the
global volume increases (positive DV at Fig. 10), there is an addi-
tional quantity of free water entering into the sample to replace
the leached minerals or to fill the formed cracks, which is the case
of NSR mortar. However if the global volume decreases (negative
DV at Fig. 10), the sample loses a part of its initial free water either
by the degradation of the surface or by precipitation of minerals in
the pores leading to remove water outside the sample which is the
case for SR5 and SR3.
The leached and precipitated minerals and other relevant
parameters could be determined. The obtained values can now
be used to describe the mechanisms of sulfate attack, to define
new types of monitoring, giving more information on the micro-
scopic evolutions during sulfate attack.
4.2.2. Microscopic volume variation
Volume variations corresponding to the four main phenomena
involved in sulfate attack can be deduced from mass variations
(Fig. 12). This monitoring strategy analyses the sulfate attack
mechanism from another angle than usual monitoring parameters,
such as expansion and strength.
The first graph in the Fig. 12 represents the volume variation
SR3
corresponding to leaching. It shows a common behaviour of the
2.30 three series of specimens. The Table 6 giving the portlandite vol-
2.10
ume fractions in the three different mortars seems in good agree-
29
ment with the graph of Fig. 12. The SR5 mortar containing the
0.26
1.4
highest fraction of portlandite showed higher volume lost by
leaching.
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931 929

Table 6 tion has exceeded the tensile strain capacity of the material and
Portlandite volume proportions based on TGA micro-cracks cause damage.
analyses.
The second row of the Fig. 14 shows a different aspect of the
Portlandite volume proportion (%) mechanism. The stairs-like shape highlights the layer by layer
NSR 4.9 degradation at the surface of tested samples.
SR5 5.8 The SR5 mortar showed slightly different filling precipitation/
SR3 5.0 deformation behaviour characterized by a longer initial stage. It
could be explained by the effect of portlandite on the expansion
evolution, mentioned in Section 3.5.3. The SR5 actually has the
The aggregates loss and precipitation graphs show significant
highest portlandite content (Table 6). The precipitated minerals
differences, highlighting different behaviours due to different
could fill larger spaces in capillary pores, in the leached portlandite
cement compositions, for instance at the level of calcium alumi-
region. This could have delayed the supersaturation in the micro-
nates phases. The NSR containing the highest C3A fraction showed
pores thus the effect of crystallization pressure.
the highest precipitation volume due to its higher potential to form
more ettringite than other cements, in addition to other sulfate-
4.3. Performance indicators
bearing minerals such as gypsum. Higher the precipitation volume
is, higher the cracking probability is. This could explain the rela-
4.3.1. Averaged density
tively high volumes of lost aggregates and mortar for the NSR mor-
Three different shapes of averaged density curves can be seen
tar. The SR5 and SR3 mortars, following the same trend of
on Fig. 15. They actually reveal different behaviours during sulfate
aggregates loss variations, presented less precipitation volume
attack. All the average density curves start form 2.2 to 2.25 g/ml.
variation, revealing better durability when exposed to sulfate.
Portlandite (2.24 g/ml density) actually was the most reactive min-
The second graph of the Fig. 12 shows two different evolutions
eral at the beginning of sulfate attack. Then the precipitation took
of free water variation. Taking into account the relatively high rates
place, filling the pores and causing cracking in the mortar matrix
of precipitation and cracking, the NSR mortars showed a positive
with different rates, depending on the mineralogical cement paste
volume variation due to water uptake in the formed cracks. How-
composition and its ability to generate crystallization pressure
ever, for the sulfate resisting cements SR3 and SR5 the mortars pre-
inside the mortars. The decreasing averaged density of the NSR
sented a negative variation due to the loss of pores free water at
mortar curve is due to the formation of less dense minerals, which
the degraded surface (where the aggregates loss is significant),
is consistent with ettringite formation (1.76 g/ml density) and
and the relatively low crystallization pressure of sulfate precipi-
cracking then water uptake (1 g/ml density). Thus a significant
tated minerals causing less cracks and less water uptake than the
decrease of averaged density indicates a poor durability of the
NSR mortar.
cementitious material. For instance, the slopes of SR5 and SR3 (sul-
fate resisting cements) curves are moderate, in comparison with
4.2.3. Filling precipitation and leached volumes equivalence the NSR mortar. For the SR3 mortar, a slight ascending slope can
It is known that during sulfate attack some minerals are leached be observed instead of descending slope, showing the lowest rate
and others precipitate but the relation between leaching and pre- of cracking among the 3 tested mortars. This could correspond to
cipitation is still ambiguous. The Fig. 13 shows a relatively strong the precipitation of other sulfate minerals like gypsum, moving
correlation between the relative filling precipitation volume the averaged density toward its 2.3 g/ml density.
(DVfil.prec/V0) and the relative leached volume. Considering that
portlandite is the main leached phase, the corresponding zones
4.3.2. Deformation path
are favouring the formation of expansive minerals in collaboration
The NSR mortar presented the highest pressure revealed by the
with the reactant zones nearby. Once the portlandite is leached,
longitudinal expansion seen on the Fig. 16, whereas the SR3
the reactive zone is actually rich in calcium (coming from port-
relatively poor in expansive minerals remained at the beginning
landite) and sulfate (coming from the external aggressive solution),
of the deformation path with relatively low longitudinal
which triggers chemical reactions to restore the thermodynamic
deformations.
equilibrium. In this sulfate-saturated environment, the existing
The sulfate-bearing phases produced during sulfate attack
AFm minerals are destabilised to form the ettringite and the over-
depend on the cement composition and the hydration products
flow of sulfate forms gypsum in presence of calcium provided by
in the mortar pastes. Some cement pastes, such as NSR paste, are
leaching. In the chemically affected zone, the precipitated sulfate
more able to produce ettringite then gypsum, whereas other
minerals proportions and the crystallization pressure vary accord-
cements, such as SR5 and SR3, are likely to form more gypsum than
ing to the cement composition and the proportions of hydration
ettringite. Different crystallization pressures are induced but the
products.
expansion mechanism for the different mortars, according to the
Fig. 16, was approximately the same. Considering the mortar
4.2.4. Longitudinal and lateral deformations as a function of filling samples have approximately the same mechanical properties, the
precipitation samples follow the same path of deformation. At the beginning
Crystallization pressure due to the sulfate minerals precipita- of the sulfate attack, lateral deformations mainly take place, and
tions induce longitudinal and lateral deformations of the cylindri- then the precipitated minerals induce higher and higher tensile
cal mortar samples. In this section, for a better understanding of stress level inside the sample, pushing forward the longitudinal
precipitation/swelling mechanism, leaching and aggregates loss expansion.
are pushed aside in order to focus on the lateral and longitudinal
deformations due to precipitation and cracking. Table 7
The Fig. 14 shows a set of linear correlations between filling Expansive pressure estimation according to three methods (MPa).
precipitation and length changes of the mortar samples. At the
Saturation/cube Precipitation/cube Degradation zone
beginning of tests, the first linear trend line represents the first
stage of degradation. The longitudinal deformation increases as NSR 17.9 9.8 18.5
SR5 12.5 8.0 14.8
crystallization pressure develops and tensile loads are generated. SR3 10.7 7.3 10.5
Then the slope of the curves significantly increases as the deforma-
930 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931
4.3.3. Expansion potential
The Table 7 presents three sets of EPI (expansion potential indi-
cators) values. These values cannot be considered as crystallization
pressures exerted on the pore walls at the microscopic level, but
they provide an indicator related to the maximum pressure that
can be generated in a given cement-based material. In the first col-
umn we considered that the filling precipitation volume is the only
source of crystallization pressure inside the samples so the equiv-
alent cube volume takes into account only the filling precipitation
volume (Dfil.prec) at thinge (Fig. 17). In the second column, we con-
sidered that the formed sulfate minerals volumes exceeding the
vacant available space (portlandite-free zones and pores), and pre-
cipitating in the generated cracks, participate to the development
of the tensile stress in the core of the samples. According to this
assumption, the equivalent cube volume correspond to the overall
precipitated volume (Dglo.prec) at thinge, which leads to larger cubes
and surfaces S, so lower calculated crystallization pressure than
the first column. The obtained values agree with the 8 MPa value
given in a previous work aimed at identifying the macroscopic
crystallization pressure on hollow tubes exposed to external
sulfate attack [27]. In the third column, we considered that the
equivalent depth of portlandite-free zones contains the expansible
minerals and all the leached region exerts the tensile stress on
the core of the cylindrical specimen. Then the EPI can be deduced
from the tensile force limit Ften.hinge and the equivalent degraded
section.
The sulfate attack of the Portland cement based mortars lead to
the precipitation of sulfate minerals like gypsum and ettringite but
in different proportions. The resulting crystallization pressure
depends on the proportions of the minerals and the confinement
conditions or the pores where the minerals precipitate. Different
minerals mixes precipitated in the different tested mortars. In
the three sets of the Table 7 the NSR mortar presents the highest
EPI which highlights the ettringite formation. The SR5 and SR3
mortars showed lower EPI, thus lower potential for sulfate attack
induced cracking of the cement paste, which is in good agreement
with the sulfate resisting property of the durable cements.
5. Conclusions
Based on monitoring of sulfate attack tests at the macroscopic
and intermediate levels a new strategy was developed to provide
the estimation of microscopic evolutions, more precisely, the vol-
ume variations entrained by each of the phenomena involved in
the sulfate attack: leaching, precipitation, free water variation
and aggregates loss. It does not require microstructural analysis
to be used in performance tests and criteria. This strategy gives
quantitative information about changes at the microscopic scale
thus a description of the sulfate attack scenario from relatively
simple monitored parameters, which makes it original.

The optimisation maps show the accuracy of the monitoring
strategy to characterize the tested mortars and their different
behaviours during sulfate attack. The best fit physical and
chemical parameters values agree with the existing knowledge
collected from other works on the sulfate attack mechanism,
which confirms the relevance of the physical concepts adopted
in the strategy principle and the reliability of results.

The observed equivalence between the leached minerals
volume and the filling precipitation volume highlights the reac-
tivity of portlandite and its ability to trigger the formation of
expansive products. A potential correlation between the port-
landite quantities, the morphologies of corresponding zones,
and the expansion mechanism including expansive mineral
quantities and reaction kinetics could exist and still to be
established.

The longitudinal expansion curves show different stages of
deformation when correlated with a filling precipitation factor.
First a linear variation can be observed as a function of mineral
precipitation and then the deformation slope increases with the
filling precipitation volume or the time revealing the macro-
cracks generation. The lateral expansion shows a stair shape
as a function of precipitated minerals, highlighting the layer
by layer degradation of the surface exposed to sulfate attack
in contrast with longitudinal deformation and cracking.

Taking into account all the phenomena involved in sulfate
attack and their effect on the durability of tested materials,
the ‘‘averaged density” was defined as a performance criteria.
This indicator proved its capability to describe the performance
of tested materials, pointing out the type of precipitated miner-
als, the rate of leaching and cracking during the sulfate attack.
The averaged density indicator was found sensitive to the dif-
ferent used cements and behaviours, which makes this indica-
tor a potential performance indicator.

Another deformation based performance indicator ‘deformation
path’ was introduced. The longitudinal and lateral variations
combined show up a common length/radius deformation path
for the different mortars. The maximum levels reached on the
deformation path varied, showing up the different levels of
cements durability.

The final performance proposed indicator was ‘expansion
potential indicator’. Different tested cement compositions lead
to different mixes of precipitated sulfate minerals generating
different crystallization pressures inside the tested mortars.
This introduced indicator gave an estimation of the expansive
pressure and a durability performance based classification for
the tested cements.
The developed strategy and the physical parameters deter-
mined in this work can be applied to any cement-based material
whatever the nature of the leached or precipitated minerals since
no assumption on these minerals is required in the model. Micro-
scopic monitoring provided by the strategy makes it applicable in
other attacks such as chloride ingress, monitoring of cracking
under continuous or momentary stress, self-healing.
Appendix
Optimization cartography discussion
Knowing the main hydration products and the corresponding
chemical equilibriums helps defining the variation domains of
physical parameters. Table 3 gives the chemical reactions, the den-
sities and the molar masses of the main hydration products found
in an OPC paste. Based on these chemical data the variation ranges
of the parameters were defined as mentioned in Table 4 in order to
limit the optimization task to a plausible variation domain.
In order to understand the relation between the physical
parameters and their influence on the coefficient R a 6D solution
domain was defined including the variation domain of the 5 phys-
ical parameters and the coefficient R. Then the 6D domain points
were projected on 5x2D plans (R function of one of the physical
parameters) on Fig. A.1.
The domain of precipitated and leached minerals densities was
defined between 1.1 and 2.3 g/ml to include all hydration products
susceptible to react during the sulfate attack. In the same way the
limits of the apparent molar mass Moapp ¼ Molea =n domain were
determined considering that one dissolved mineral molecule gives
only one OH ion. The upper and lower limits of Moapp definition
domain were determined by dividing the molar masses by the
number of OH/ H+ (based on the Table 3 chemical reactions). 20
 G. Massaad et al. / Construction and Building Materials 127 (2016) 918–931
Least square NSR
SR5
SR3
Moap ρlea
Fig. A.1. Least square optimization mapping.
and 100 g/mol were chosen as limits including all the probably
leached minerals, mainly portlandite and CSH. In absence of
experimental data related to taken-down aggregates the fag.td
variation domain was chosen as large as possible to cover the
possible optimal solution intervals. The flea.prec parameter describes
the free water variations during sulfate attack; it varies between
0 and 3 to include the two possible cracks state variations:
widening (flea.prec > 1) or filling (flea.prec < 1) of cracks by minerals
leaching and precipitation.
Once the variation domains of physical parameters were
defined, these domains were divided into intervals to get interme-
diate values between the upper and lower limits of each physical
parameter. Then the least square (14) corresponding to all the pos-
sible combinations between the 5 parameters were calculated. In
the Fig. A.1 each graph presents, via the coloured cartography,
the densities repartition of the calculated R values corresponding
to all the predefined combinations, as a function of one physical
parameter.
Red zones indicate the zones of solution values. Relatively
restrained red zones in the first, second, third and fifth columns
show the strong effect of some parameters of the predefined
domain on the minimization of the least squares. On the fourth
graph the large red zone shows a less decisive role attributed to
this parameter in curve fitting which implies relatively lower rele-
vance of the values obtained via curve fitting procedure.
References
[1] A. Chabrelie, Mechanisms of Degradation of Concrete by External Sulfate Ions
Under Laboratory and Field Conditions PhD thesis, EPFL Lausanne, 2010.
[2] B. Larbi, W. Dridi, P. Dangla, P. Le Bescop, Link between microstructure and
tritiated water duffusivity in mortars: Impact of aggregates, Cem. Concr. Res.
82 (2016) 92–99.
[3] C. Yu, Mechanism of expansion of mortars immersed in sodium sulfate
solution, Cem. Concr. Res. 43 (2013) 105–111.
[4] C. Carde, Effect of the leaching of calcium hydroxide from cement paste on
mechanical and physical properties, Cem. Concr. Res. 27 (1997) 539–550.
[5] D. Planel, J. Sercombe, P. Le Bescop, F. Adenot, J.M. Torrenti, Long-term
performance of cement paste during combined calcium leaching-sulfate
attack: kinetics and size effect, Cem. Concr. Res. 36 (2006) 137–143.
[6] E. Rozière, R. Cortas, A. Loukili, Tensile behaviour of early age concrete: new
methods of investigation cement and concrete composites 55 (2015) 153–161.
[7] F.P. Glasser, The thermodynamics of attack on Portland cement with special
reference to sulfate, in: M.G. Alexander, A. Bertron (Eds.), Proc Concrete in
Aggressive Aqueous Environments, Performance, Testing and Modeling, Vol. 1,
RILEM, Toulouse, France, June 3–5 2009, pp. 3–17.
[8] G.W. Scherer, Crystallization in pores, Cem. Concr. Res. 29 (2000) 1347–1358.
931
ρprec fag.td flea.prec
[9] G.W. Scherer, Stress from crystallization of salt, Cem. Concr. Res. 34 (2004)
1613–1624.
[10] H.F.W. Taylor, Sulfate reactions in concrete – microstructural and chemical
aspects, Cement Technology (Ceramic Transactions), vol. 40, 1989, pp. 61–78.
Westerville, OH.
[11] J. Stroh, M.C. Schlegel, E.F. Irassar, B. Meng, F. Emmerling, Applying high
resolution SyXRD on sulfate attacked concrete field samples, Cem. Concr. Res.
66 (2014) 19–26.
[12] J. Skalny, J. Pierce, Sulfate attack issues, material science of concrete, in: Sulfate
Attack Mechanism, American Ceramic Society, Westerville, OH, 1999, pp. 49–
64.
[13] J. Yuan, Y. Liu, Z. Tan, B. Zhang, Investigating the failure process of concrete
under the coupled actions between sulfate attack and drying-wetting cycles by
using X-ray CT, Constr. Build. Mater. 108 (2016) 129–138.
[14] M.D. Cohen, A. Bentur, Durability of Portland cement–silica fume pastes in
magnesium sulfate and sodium sulfate solutions, ACI Mater. J. 85 (3) (1988)
148–157.
[15] M. Santhanam, M.D. Cohen, J. Olek, Mechanism of sulfate attack: a fresh look,
Part 1: Summary of experimental results, Cem. Concr. Res. (2002) 915–921.
[16] M. Santhanam, M. Cohen, J. Olek, Effects of gypsum formation on the
performance of cement mortars during external sulfate attack, Cem. Concr.
Res. 33 (2003) 325–332.
[17] P. Blanc, A. Lassin, P. Piantone, Thermoddem a Database Devoted to Waste
Minerals, http://thermoddem.brgm.fr (2007).
[18] P. Feng, E.J. Garboczi, C. Miao, J.W. Bullard, Microstructural origins of cement
paste degradation by external sulfate attack, Constr. Build. Mater. 96 (2015)
391–403.
[19] P. Mehta, Mechanism of expansion associated with ettringite formation, Cem.
Concr. Res. 3 (1973) 1–6.
[20] P.W. Brown, H.F.W. Taylor, The role of ettringite in external sulfate attack, in: J.
Marchand, J. Skalny (Eds.), Materials Science of Concrete – Special Volume:
Sulfate Attack Mechanisms, American Ceramic Society, Westerville (Ohio),
1999, pp. 73–97.
[21] R. Hachem, E. Rozière, F. Grondin, A. Loukili, Multi-criteria analysis of the
mechanism of degradation of Portland cement based mortars exposed to
external sulfate attack, Cem. Concr. Res. 42 (2012) 1327–1335.
[22] R. El Hachem, E. Rozière, F. Grondin, A. Loukili, New procedure to investigate
external sulfate attack on cementitious materials, Cement Concr. Compos. 34
(3) (2012) 357–364.
[23] T. Liu, D. Zou, J. Teng, G. Yan, The influence of sulfate attack on the dynamic
properties of concrete column, Constr. Build. Mater. 28 (2012) 201–207.
[24] T. Thorvaldson, Chemical aspects of the durability of cement products, in:
Proceedings of 3rd International Symposium on the Chemistry of Cement,
Cement and Concrete Association, London, 1954.
[25] V. Genovés, L. Soriano, M.V. Borrachero, J. Eiras, J. Paya, Prelimenary study on
short-term sulphate attack evolution by non-linear impact resonance acoustic
spectroscopy technique, Constr. Build. Mater. 78 (2015) 295–302.
[26] V. Kocaba, Development and Evaluation of Methods to Follow Microstructural
Development of Cementitious Systems Including Slags PhD thesis, EPFL
Lausanne, 2009.
[27] W. Muellauer, R.E. Beddoe, D. Heinz, Sulfate attack expansion mechanisms,
Cem. Concr. Res. 52 (2013) 208–215.
[28] W.Y. Ouyang, J.K. Chen, M.Q. Jiang, Evolution of surface hardness of concrete
under sulfate attack, Constr. Build. Mater. 53 (2014) 419–424.
[29] B. Tian, M.D. Cohen, Does gypsum formation during sulfate attack on concrete
lead to expansion? Cem. Concr. Res. 30 (1) (2000) 117–123.