Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Construction and Building Materials 264 (2020) 120525

Contents lists available at ScienceDirect

Construction and Building Materials


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

The damage mechanism and failure prediction of concrete under


wetting–drying cycles with sodium sulfate solution
Jingge Ren a,b, Yuanming Lai a,b,⇑, Ruiqiang Bai a,b, Yinghong Qin c
a
State Key Laboratory of Frozen Soil Engineering, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, Gansu 730000, China
b
University of Chinese Academy of Sciences, Beijing 100049, China
c
College of Civil Engineering and Architecture, Guangxi University, Guangxi 530000, China

h i g h l i g h t s

 Collision probabilistic factor of the formation of gypsum was calculated.


 Reaction-rate constant calculation formula of gypsum is proposed.
 Formulae of the volume fractions of mirabilite and gypsum in the pore were proposed.
 Damage node of concrete under sulfate wetting–drying cycles was well predicted.
 Concrete was in a dynamic damaging-filling state under wetting–drying cycles.

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

Article history: This study aims to investigate the damage mechanism and predict the failure time of concrete subjected
Received 20 November 2019 to wetting and drying (WD) cycles in a sodium sulfate solution. To explore the growth rules and the
Received in revised form 4 July 2020 growth rates of gypsum, the reaction-rate constant calculation formula of gypsum is proposed based
Accepted 10 August 2020
on theories of chemical reaction kinetics. Subsequently, the formulae of the volume fractions of mirabilite
and gypsum in the pore under each WD cycle are proposed and used to calculate the macroscopical ten-
sile stress to determine the WD cycle number that the concrete can sustain. Finally, the reliability of the
Keywords:
proposed model is verified by an indoor experiment. In addition, a new damage type of concrete, induced
Wetting–drying cycle
Sodium sulfate
by filling-damaging cycles, is also found during the experiment. These findings can be used for evaluating
Gypsum the damage of concrete that is subjected to cyclic WD in sodium sulfate solution.
Crystal growth Ó 2020 Elsevier Ltd. All rights reserved.
Crystallization pressure
Durability

1. Introduction Sodium sulfate physically attacks plain concrete by precipitat-


ing mirabilite in the pore of concrete when the surrounding tem-
Concrete is widely used in many construction fields, such as perature and relative humidity change [5]. At an environment
pavement, bridge and building due to its low cost, simple prepara- with cyclic wetting and drying (WD), the thenardite deposites in
tion process and high performance. However, concrete will be seri- the pore during the drying process, increases the concentration
ously damaged when it is exposed to marine environments, acid of the solution in the pore in the subsequent wetting cycle and
rain and saline soil. In these harsh conditions, the durability of con- even precipitates mirabilite in the pore. The mirabilite cumulates
crete is influenced by inorganic salts, including the chlorine salt in the pore progressively as the increase of the WD cycle index,
and sulfate [1,2]. The chlorine salt is the most harmful salt to rein- crystallization pressure is generated, and the concrete is damaged
forced concrete, and the sulfate is the most harmful salt to the [6]. The physical damage has already been widely accepted [7–9]
plain concrete [3,4]. Sodium sulfate attacks plain concrete physi- and the failure mechanism of the mirabilite has been studied. Car-
cally and chemically. los et al. [10] investigated the transformation rule between differ-
ent crystal phases of sodium sulfate and found that both thenardite
and mirabilite can precipitate directly from a saturated sodium
⇑ Corresponding author at: State Key Laboratory of Frozen Soil Engineering, sulfate solution at room temperature (20 °C). The crystallization
Northwest Institute of Eco-Environment and Resources, Chinese Academy of
pressure of sodium sulfate in the pore of rock was analyzed and
Sciences, Lanzhou, Gansu 730000, China.
E-mail address: ymlai@lzb.ac.cn (Y. Lai). calculated at different times of WD cycles [11].

https://doi.org/10.1016/j.conbuildmat.2020.120525
0950-0618/Ó 2020 Elsevier Ltd. All rights reserved.
2 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

Nomenclature

SM,i Volume fraction of mirabilite after i WD cycles


Symbols
SG,i Volume fraction of gypsum after i WD cycles
ae Ionic activity of e
SM Total volume fraction of mirabilite
b’ Biot coefficient
SG Total volume fraction of gypsum
cNa2 SO4 Concentration of sodium sulfate mol/L
Sm Volume fraction of the crystal after m = n + i + 1 WD cy-
c’ Concentration of the solution mol/L
cles
cCa2þ Concentration of calcium ion mol/L
Sn Volume fraction of the crystal after n WD cycles
cs Saturation concentration of sodium sulfate mol/L
t Any time during the reaction
c’s Concentration of the saturated solution mol/L
t0 The wetting duration for each WD cycle
cSO2 Concentration of sulfate ion
4 T Temperature in Kelvin K
c0SO2 Concentration of sulfate ion in initial state mol/L
sat4
cSO2 Concentration of sulfate ion in saturated sodium sulfate
vc Molar volume of any crystal L/mol
4
solution mol/L
v reac
e Stoichiometric coefficient of ion e
cðCa2þ SO2 Þ Concentration of the particle pair of ðCa2þ SO2
v CaðOHÞ Molar volume of calcium hydroxide L/mol
4 Þ mol=L
vG
2
4 Molar volume of gypsum L/mol
DSO2 Diffusion coefficient of sulfate ion m2/s
e
4
Any ion
vM Molar volume of mirabilite L/mol
f Surface tension N/m
vT Molar volume of thenardite L/mol
Vc,i Total volume of the crystals after i WD cycles L
ISO2 Flux of SO2
4 at the spherical area around the Ca

mol/s
4 Vc,i+1 Total volume of the crystals after i + 1 WD cycles L
J SO2 the flux of SO2
4 in per unit time per unit area mol/(sm )
2
4 VG Volume of gypsum L
kB Boltzmann constant J/K
VG,i Volume of gypsum after i WD cycles L
kG Reaction-rate constant of gypsum
VG,i+1 Volume of gypsum after i + 1 WD cycles L
k1D Diffusion-rate constant of reaction particle
VG,n Volume of gypsum after n WD cycles L
k-1D Separation-rate constant of the particle pair
VM,i Volume of mirabilite after i WD cycles L
k2 Reaction-rate constant of the particle pair
VM,i+1 Volume of mirabilite after i + 1 WD cycles L
KG dissolution equilibrium constant of gypsum
Vi+1
M Volume of the mirabilite produced during i + 1 WD cycle
K MN1 X N2 N3 H2 O Dissolution equilibrium constant ofMN1 X N2  N 3 H2 O
L
Kreac Equilibrium constant of the chemical reaction
Vp Volume of the pore L
KSP,M Thermodynamic solubility product of mirabilite
V 0p Initial volume of the pore L
KSP,T Thermodynamic solubility product of thenardite
V np Volume of the pore after n WD cycles L
MG Molar mass of gypsum g/mol
VQ Volume in the pore that the solution can invade L
NA Avogadro’s constant mol1
VRi Volume in the pore that available for solution invading L
NG Particle number of gypsum mol
Vs Volume of the solid dissolved in the pore solution L
N0G Particle number of gypsum at the initial condition mol
VT,i Volume of thenardite after i WD cycles L
NiG Particle number of gypsum after i WD cycles mol
V DG The volume increment from the formation of gypsum L
NnG Particle number of gypsum after n WD cycles mol
cCa2þ Reaction rate of calcium ion mol/s
NSO2 Particle number of sulfate ion mol
4 cG Formation rate of gypsum mol/(Ls)
p Osmotic pressure Pa
e Poisson’s ratio
Ps Saturated osmotic pressure Pa
n Reaction degree mol
P Probabilistic factor for the molecule collision
h Contacting angel between the crystal liquid interface
Pc Crystallization pressure MPa
and the pore wall
PG Crystallization pressure of gypsum MPa
PM Crystallization pressure of mirabilite MPa
lHCa2þ Standard chemical potential of calcium ion J/mol
Qreac Ionic activity product
lHG Standard chemical potential of gypsum J/mol
le Chemical potential of e
QG Ionic activity product of gypsum
rp Radius of the pore m
lHe Standard chemical potential of e J/mol
r Ca2þ Radius of calcium ion pm
lHSO2 Standard chemical potential of sulfate ion J/mol
r* 4
Macroscopic tensile stress
r SO2 Radius of sulfate ion pm
4
r Ca2þ SO2 Radius of calcium ion plus sulfate ion pm
rc Critical tensile stress
r
4
Distance between Ca2+ and SO2
rG Macroscopic tensile stress of gypsum
4 pm
R Molar gas constant J/(molK)
rM Macroscopic tensile stress of mirabilite
rr Radial compressive stress
RHsat,T Deliquescence relative humidity of thenardite
rT Tensile strength of the concrete
S1 Volume fraction of the crystal after one WD cycles
Sc Volume fraction of the crystal in the pore
p Circular constant
qG Density of gypsum
Si Volume fraction of the crystal after i WD cycles
gw Viscosity of water
Si+1 Volume fraction of the crystal after i + 1 WD cycles

Sodium sulfate chemically attacks plain concrete by reacting sulfate ion. The gypsum should form in situ or on the surface of cal-
with the hydrated calcium silicates (C-S-H) [12]. Sodium sulfate cium hydroxide particle. Supporting the viewpoint of Hansen,
also reacts with the calcium hydroxide in the plain concrete, pro- Mather [16] thought that the formation of gypsum cannot create
duces expansion crystal and causes damages in the concrete [13]. expansion and pointed out that gypsum was primarily formed
Studies on the chemical damage are still controversial. from the reaction of sulfate ions and calcium ions by a through-
One argument is whether the formation of gypsum expands the solution mechanism. On contrary, most researchers thought that
concrete. Hansen [14,15] believed that the precipitation of gypsum the formation of gypsum can cause expansion [17–20]. By studying
is expansion-free. If the gypsum can cause expansion, the volume gypsum in the interfacial zone of a plain concrete, Bonen and
of plain concrete increases during the reaction of calcium ion and Sarkar [21] found that a thick layer of gypsum up to 50 lm width
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 3

precipitates by a through-solution mechanism. This crystallization pore belong to ionic reaction [24]. Both gypsum and ettringite
of gypsum causes tensile stress and expansion in the concrete. The are formed with the reaction of ions by a through-solution mech-
results are effectively contradicted to the opinion of Hansen. Yong anism [39]. The reaction solution belongs to multicomponent
et al. [22] also investigated the damage process of the interfacial dielectric solution, the crystallization pressure correlates with the
zone under the sulfate attack. They got the same viewpoint as ion activity product in the solution. On the basis of the crystalliza-
Bonen and Sarkar, and thought that sulfate ion can react with cal- tion pressure equation of supersaturated solution, Scherer et al.
cium hydroxide and monosulfate (AFm) and lead to expansion and [9,40,41] proposed an ionic-reaction crystallization pressure equa-
cracking. Wang [23] observed the products at different depths in a tion that can be used to calculate the crystallization pressure of
cement paste under sulfate attack by X-ray diffraction. The inves- gypsum and ettringite in the pore solution. Based on this new
tigation result found that the formation of gypsum causes more equation, Flatt and Scherer [42] analyzed the thermodynamics of
damage than the formation of ettringite does. Ping and Beaudion crystallization stress when the ettringite formation is delayed
[24,25] suggested a theory model based on the principles of chem- (DEF). They calculated the lowest content, the storage temperature
ical thermodynamics and pointed out that the crystal induced and the limited space of ettringite crystal that cause damages in
pressure needs to additional conditions. One is that the solid pro- the pore. In addition, on the basis of the ionic-reaction crystalliza-
duct should form and grow in a confined space, and the other is tion pressure equation, Yu et al. [43] calculated the crystallization
that the activity product of the reactants in the pore solution pressure of ettringite in the pore of the concrete under the sulfate
should be larger than the solubility product of the solid products attack and the failure mechanism in the plain concrete.
under atmospheric pressure. Tian and Cohen [26] carried out the Most of the studies centered on the pore crystallization pres-
sulfate attack experiments and demonstrated that the damage in sure which belongs to microscopic force. However, the concrete
the concrete is caused simultaneously by the expansions of ettrin- failure always results from the macroscopic tensile stress, which
gite and gypsum. However, the crystal type varies with sulfate con- can be caused by crystallization pressure. When the macroscopic
centration and this thus results in the different failure mechanism tensile stress exceeds the critical tensile strength of the concrete,
in the concrete. Hence, a new argument was generated on the fail- the concrete is damaged. Hence, a bridge is needed from the micro-
ure mechanism between the laboratory test and the site condition. scopic crystallization pressure to the macroscopic tensile stress for
Gallop and Taylor [27,28] found that the sulfate concentration used forecasting the life of the concrete. Coussy [37] derived the equa-
is usually 30 g/l in the laboratory and the main corrosion product is tion between the macroscopic stress and the crystallization pres-
gypsum. The gypsum forms veins and layers in the first 2 mm sure on the basis of the thermodynamic and physical chemistry.
beneath the surface of the mortar bars. Bellmann et al. [29] demon- Coussy pointed out that the macroscopic stress is a function of
strated the result according to the investigation on the product in crystallization pressure, of Biot coefficient and of the volume frac-
the concrete under different concentrations of the sodium sulfate tion of the pore space filled with crystals. However, Coussy’s equa-
solution. When the concentration of sodium sulfate solution is tion has rarely been used to forecast the damaging time of the
low, the main reaction product is ettringite and the main reaction concrete under sulfate attack, because the volume fraction is hard
product is gypsum under high concentration sodium sulfate to calculate precisely. On one hand, the solid volume is composed
solution. of several kinds of crystals such as mirabilite, gypsum and ettrin-
After the damage of concrete under sulfate attack is understood, gite. On the other hand, the volumes of gypsum and ettringite vary
the research emphasis has been shifted to the crystallization pres- not only with the soak time in the solution but also with the reac-
sure, which usually includes the pressure caused by the formation tion rate of ions in the solution. In addition, calculation of the
of mirabilite, gypsum and ettringite. The crystallization pressure reaction-rate of ionic reaction needs to understand the molecular
caused by mirabilite has already been studied on the basis of the dynamics theory.
principles of thermodynamics and physical chemistry. Tsui et al. The calculation is more complicated when the concrete is sub-
[30] analyzed the phase transformation and crystallization pres- jected to WD cycles, because the volumes of the crystals accumu-
sure of sodium sulfate, and found a threshold of salt content late in the pore with the increase of the cycle time. So far, the
beyond which damage increases substantially. Scherer [31,32] stability and the crystallization pressure of the crystal are studied
found that the crystallization pressure in the pore comes from on the basis of the thermodynamics [44–47]. Limited studies are
the repulsive force between the crystal and the pore wall, and centered on the reaction rate of gypsum and on the volume frac-
the maximum pressure is determined by the separating force on tions of both mirabilite and gypsum in the pore space. However,
the surface of the crystal. The crystallization pressure caused by the research of mirabilite and gypsum in the pore of plain concrete
mirabilite is larger than that caused by other salts, and at most is critical to understand the failure mechanism of concrete and
of the time the concrete and rock materials fail to resist the max- forecast its damage time. Such studies are important for the engi-
imum crystallization pressure [33]. On the basis of the crystalliza- neering application of the concrete and are also meaningful to
tion pressure equation [34], Rijniers et al. [35,36] investigated the understand the destruction process of the concrete subjected to
crystallization pressure of mirabilite and heptehydrated phase by the WD cycles of sodium sulfate solution.
nuclear magnetic resonance. The result indicated that the heptehy- Formulae about the crystallization pressure and the volume
drated phase cannot cause crystallization pressure and damages fraction in the pores are always used to forecast and analyze the
but the mirabilite in the pore can. After investigating the deforma- damage of concrete under WD cycle test. Due to the complexity
tion and crystallization pressure in the pore, Coussy [37] correlated of the in-situ environment, the formulae are usually used in the
the crystallization pressure with the macroscopic tensile stress, a indoor test. For the indoor WD cycle test, the test conditions vary
correlation that can forecast the damage time via the critical stress from person to person [48–50]. However, most of the tests are con-
was proposed. However, most of these studies focused on the ducted according to the relevant test standards [48–49]. Based on
supersaturation of sodium sulfate solution only, but not on the the standard, the concentration of the sodium sulfate solution is
influence of relative humidity on the crystallization pressure. Stei- 5 wt%, the cycle period is 24 h, the total cycle number is 150 times
ger and Asmussen [38] proposed a crystallization-pressure equa- and so on. During the test, the compressive strength, the mass and
tion that considered the influences of both thermodynamic the ultrasonic velocity of plain concrete samples were usually
solubility and the relative humidity. This equation has been widely tested to analyze the durability of concrete subjected to WD cycles.
used by the subsequent researchers. Different from the crystalliza- Among which, the mass change is an important index for the dura-
tion of mirabilite, the formations of gypsum and ettringite in the bility of the concrete and can reflect the damage mode in the
4 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

concrete. The previous studies found that the mass of the concrete heptahydrate has a contact angel of about 90° and cannot cause
increases with the increase of WD times. However, these studies significant stress in a porous medium. Hence, only the crystalliza-
have a limit. Because their test span are usually ten, fifteen, or tion pressures of thenardite and mirabilite at room temperature
thirty cycles [51–53], the test results just can reflect the change are considered.
trend of the mass but fail to represent the specific variation of  
nRT cNa2 SO4
the mass during the experimental process. The specific variation Pc ¼ ln ð1Þ
of the mass can bring a further understanding for the damage in
vc cs
the concrete. Therefore, a high frequency testing during the WD
2f cosh
cycles is needed. In addition, a further study on the damage mode Pc ¼ ð2Þ
of the concrete is needed.
rp
Here the damage of the concrete subjected to WD cycles using where Pc is the crystal pressure in the pore (MPa), n is the number
the sodium sulfate as the solution was studied. The interaction of ions in the salt, R is the molar gas constant (J/(mol∙k)), T is the
between the sodium sulfate solution and the hydration products temperature in Kelvin (K), vc is the molar volume of the salt
in the pore was investigated. The growth rule, growth rate and (L/mol), cNa2 SO4 is the saturation concentration of sodium sulfate in
the volume of the crystals on the basis of the molecular dynamics the pore (mol/L), cs is the saturation concentration of sodium sulfate
and physical chemistry were also studied. Through these studies, (mol/L), f is the surface tension (N/m), h is the contact angel
formulae for the volume fraction of mirabilite and gypsum at dif- between the crystal liquid interface and the pore wall (°) and rp is
ferent WD cycles were proposed. With the volume fraction and the radius of the pore (m) (assuming that the pore is a cylindrical
the crystallization pressure of the crystals, the macroscopical ten- shape).
sile stress developed in a plain concrete can be calculated. By com- The transformation between thenardite and mirabilite is influ-
paring the macroscopical tensile stress during each WD cycle with enced by temperature and relative humidity. The correlation
the critical tensile stress of the concrete, the damage of the con- between the sodium sulfate solution and temperature is shown
crete under WD cycle can be forecasted. Then, the reliability of in Fig. 2 [61]. There are two important bifurcations. One bifurcation
the formulae is verified by our laboratory WD test results. The temperature is 32.38 °C, above which the supersaturation sodium
WD cycles were conducted according to the relevant test stan- sulfate solution precipitates as thenardite and below which, as
dards. During the test, a high frequency testing method was used mirabilite. The crystal phase can also be changed by relative
to explore the specific variation of concrete. The new damage mode humidity (Fig. 3) [60]. Above 32.38 °C, the sodium sulfate directly
of concrete under WD cycles was found. On the basis of these tests precipitates as thenardite at any relative humidity [62,63]. Below
and the theoretical formulae, the life and the damaging process of 32.38 °C, the sodium sulfate precipitates firstly as mirabilite and
the plain concrete, attacked by cyclic wetting and drying, can be then as thenardite with the decrease of the humidity [63,64].
forecasted and understood.
2.1.2. Ettringite and gypsum
The ettringite has a volume that is about 2.27 times of the vol-
2. Growth of crystals under WD cycles
ume of solid reactant. The crystallization of ettringite thus causes
expansion, spalling, and cracking in the concrete [65]. Ettringite
2.1. Properties of the crystals
is formed by ion reactions in the solution, because the solid–solid
conversion is kinetically difficult at room temperature and more
Several kinds of crystals are generated in concrete during WD
likely to occur at high temperature or in very low porosity system
cycles. One is mirabilite formed by the supersaturation of sodium
[66]. Ettringite begins to lose free water when the temperature
sulfate solution; the other two are ettringite and gypsum which
above 40 °C [44,67,68]. On the contrary, gypsum has a good ther-
are produced by the reaction between hydration products and
mostability. Above 104 °C, gypsum begins to lose free water.
sodium sulfate. These crystals change with the temperature, the
Hence, the gypsum is stable during the WD cycles as the common
humidity and the concentration of sodium sulfate solution. Hence,
drying has a temperature up to 75 °C [69], at which the phase of
it is necessary to study the stability and changing rule of the crys-
ettringite has changed.
tals at different conditions in order to analyze the failure mecha-
When the concrete immerses in a sodium sulfate solution, the
nism of concrete under WD cycles.
gydration product calcium hydroxide can react with sodium sul-
fate and produce gypsum (Eqs. (3) and (4)). After that, the gypsum
2.1.1. Sodium sulfate can react with the hydrated calcium aluminate and produce ettrin-
Sodium sulfate belongs to monoclinic and has at least eight gite (Eq. (5)). The main product is determined by the concentration
crystal phases [11]. Phase I and phase II can only exist at the tem- of the sulfate ion [70], with the product as ettringite at low concen-
perature above 270 °C and 225 °C. Phase II have a narrow stability tration but gypsum at high concentration [28,71,72].
zone [54]. The crystalline phase Na2SO48H2O is formed only under
high pressure condition [55–58]. Only phases III, IV, V, sodium sul- 2C 2 S þ 4H2 O ! C 3 S2 H3 þ CaðOHÞ2 ð3Þ
fate heptahydrate, and mirabilite can exist at room temperature
(20 °C), of which phases III, IV and sodium sulfate heptahydrate
are unstable [54,59]. Therefore, only phase V and mirabilite can
stably exist at room temperature.
Na2SO4 usually has three possible crystal phases in the pore
solution: an anhydrous phase (thenardite), a decahydrated phase
(mirabilite), and a metastabel heptahydrated phase. The crystal-
lization pressure in the pore is often calculated by Eq. (1) [21]
based on thermodynamics. The crystallization pressure can also
be calculated by Eq. (2) [60] in which the crystallization pressure
is a function of the contact angel between the crystal liquid
interface and the pore wall. The contact angel is usually larger than
90° (Fig. 1). Rijniers et al. [35,36] found that sodium sulfate Fig. 1. The crystal grows in the pore.
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 5

mirabilite in the pore if the concentration of sodium sulfate solu-


tion in the pore is known. However, the growth rate of gypsum
depends on the reaction rate of the ions, the concentration of the
sodium sulfate solution and the wetting duration. The reaction rate
also correlates with the reaction type. It is necessary to investigate
the reaction type and the reaction rate of the production of
gypsum.

2.2.1. Reaction type and rate


The gypsum forms during the reaction of calcium hydroxide,
sodium sulfate and water in the pore solution. As strong elec-
trolytes, both calcium hydroxide and sodium sulfate can be fully
ionized in the solution as calcium ion, hydroxide ion, sodium ion
and sulfate ion. Hence, the formation of gypsum belongs to liquid
phase ionic reaction [73]. The calcium ion in the pore solution orig-
inates from the dissolution of solid hydration product calcium
Fig. 2. The diagram of the crystal phase changes with the temperature [61]. hydroxide. The solubility of calcium hydroxide is small and
decreases with the increase of temperature [74,75]. Calcium
hydroxide in pore of concrete is sufficient and can furthest dissolve
in the pore solution at the test temperature. Hence, it can be
assumed that the concentration of calcium ion is equal to the sol-
ubility of calcium hydroxide at the experimental temperature.
There are three stages during the generation of gypsum. In the
first stage, calcium hydroxide and sodium sulfate are ionized in the
solution. In the second stage, the ions diffuse in the solution and
form particle pairs when ions are meeting. In the last stage, the
particle pairs react and produce the gypsum. During the second
stage, there are a mass of water molecules around the reaction
ions. The reaction ions need constantly extrude the water mole-
cules and enter into the same cage to meet and react with each
other [76] (Fig. 4).
The crystal of gypsum is composed of calcium ion, sulfate ion
and water molecule. Because the reaction takes place in water, it
can be considered that water has no effect on the reaction rate.
Hence, the reaction rate depends on the concentration and diffu-
Fig. 3. The diagram of sodium sulfate solution changes with the relative humidity sion rate of calcium ion and sulfate ion. The reaction equation
[60]. can be written as:
k1D
  k
Ca2þ þ SO2
4 $ Ca  SO2 ! CaSO4  2H2 O ð6Þ
2þ 2
4
CaðOHÞ2 þ Na2 SO4 þ 2H2 O ! CaSO4  2H2 O þ 2NaOH ð4Þ k1D H2 O

where (Ca2þ  SO2 4 ) is the particle pair, k1D is the diffusion-rate con-
3ðCaSO4  2H2 OÞ þ 3CaO  Al2 O3  12H2 O þ 14H2 O stant of reaction particle, k-1D is the separation-rate constant of the
! 3CaO  Al2 O3  3CaSO4  32H2 O ð5Þ particle pair, k2 is the reaction-rate constant of the particle pair. For
some time later, the concentration of the particle pair reaches the
The concentration of sodium sulfate solution used in the labora- stabilization state:
tory is always much larger than that in site. However, the WD
interface of the concrete in site always suffers cyclic WD. Hence, dcðCa2þ SO2 Þ
the concentration of sodium sulfate at the WD interface is also
4
¼ k1D  cCa2þ  cSO4 2  k1D cðCa2þ SO2 Þ  k2 cðCa2þ SO2 Þ ¼ 0
dt 4 4

much larger than that under site condition, comparable to the con- ð7Þ
centration used in laboratory. Therefore, the WD interface is pref-
erentially damaged. In addition, the damage generated by the where cðCa2þ SO2 Þ is the concentration of the particle pair
4

expansion crystal is mainly induced by gypsum instead of ettrin- (Ca 



SO2
4 )
(mol/L), cCa2þ is the concentration of calcium ion (mol/
gite. Hence, the growth rate of gypsum was considered, while the L), cSO2 is the concentration of sulfate ion (mol/L). From Eq. (7),
4
influence of ettringite was neglected. The plain concrete samples
the concentration of the particle pair at the stabilization state can
are damaged rapidly under WD cycles, during which the influence
be obtained:
of sodium sulfate on the gel materials in the concrete is negligibly.
Hence, only the damage caused by the crystallization of mirabilite k1D  cCa2þ  cSO2
and gypsum was studied. cðCa2þ SO2 Þ ¼ 4
ð8Þ
4 k1D þ k2
The reaction rate can be calculated by:
2.2. Growth rate of the crystal
k2 k1D
cG ¼ k2 cðCa2þ SO2
4 Þ
¼ c 2þ  cSO2 ¼ kG  cCa2þ  cSO2 ð9Þ
The crystallization pressure is determined by the volume of k1D þ k2 Ca 4 4

expansion crystal in the pore. The volume is always determined


by the growth rate of crystal in the pore. The volume of sodium sul- where cG is the formation rate of gypsum (mol/(Ls)), kG is the
k2 k1D
fate in the pore depends on the concentration of the sodium sulfate reaction-rate constant of gypsum andkG ¼ k1D þk2
. When k-1Dk2,
solution and on WD times. It is feasible to calculate the volume of the diffusion rate of the ions is much quicker than the reaction rate
6 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

Fig. 4. The particles react in the solution.

 2
of the particle pair and kG ¼ kk21D
k1D
. The reaction rate of gypsum cCa2þ ¼ 4p rCa2þ þ rSO2 DSO2 cSO2 ð12Þ
4 4 4
mainly depends on the reaction rate of the particle pair. When
k2k-1D, the particle pair reacts in a rate that is faster than the dif- where ISO2 is the flux of sulfate ion at the spherical area around the
4
fusion rate of the ions. Hence, kG = k1D, the reaction rate of gypsum calcium ion (mol/s),r Ca2þ is the radius of Ca2+ (pm), and r SO2 is the
4
mainly depends on the diffusion rate of the ions in the solution [77]. radius of SO2
4 (pm).
As the reactive activation energy of gypsum in the water solution is
Because both Ca2+ and SO2 4 are diffusing in the solution, the dif-
very small, the reaction rate of gypsum depends on the diffusion fusion coefficient DSO2 in Eq. (12) should beDCa2þ þ DSO2 . Also, the
rates of the ions. Hence, the formation of gypsum is a diffusion- 4 4

number of Ca2+ ions is more than one. Hence, the reaction rate
controlled reaction and the reaction rate should be calculated by
should be
the molecular kinetic theory [78,79].
  
According to reaction kinetics in the diffusion zone, it is cG ¼ 4p DCa2þ þ DSO2 r Ca2þ þ r SO2 cCa2þ  cSO2  NA ð13Þ
assumed that a Ca2+ cannot move, and SO2 4 reacts with Ca2+ if 4 4 4

the SO4 diffuses into the spherical zone around the Ca2+ with a
2 1
where NA is the Avogadro’s constant (mol ). According to Eq. (13),
rCa2þ SO2 radius (Fig. 5). r Ca2þ SO2 is the radius of Ca2+ plus SO2
4 the diffusion rate constant
4 4
  
(pm). There is a concentration gradient of SO2 4 around the Ca2+ iskG ¼ 4p DCa2þ þ DSO2 r Ca2þ þ rSO2 N A . From Stokes-Einstein dif-
[77,80]. The concentration of SO24 decreases with closing to Ca2+. 4 4

fusion coefficient equation, it can be got


The reaction rate is equivalent to the flux at the spherical area
around the Ca2+ with a r Ca2þ SO2 radius. According to Fick’s first
4 kB T
law of diffusion, the flux of SO2 2
4 (J SO2 mol/(sm )) per unit time
DSO2 ¼ ð14Þ
4 4 6pgw rSO2
4
per unit area is proportional to the concentration gradient:
dcSO2 kB T
J SO2 ¼ DSO2 : ð10Þ DCa2þ ¼ ð15Þ
6pgw rCa2þ
4
4 4 dr
where DSO2 is the diffusion coefficient of SO2 2
4 (m /s), r is the dis-
4 where kB is Boltzmann constant, kB = 1.38  1023(J/K); g is the vis-
dcSO2
tance between Ca and 2+
SO2
(pm), and dr4 is the concentration cosity of the solvent (Pas). Substituting Eqs. (14) and (15) into kG
4
gradient of SO2 2+ leads to
4 . Hence, the reactive particle number of Ca
(cCa2þ ) in unit time can be written as:  2
 2 2kB T r Ca2þ þ r SO2
kG ¼ 4
NA ð16Þ
cCa2þ ¼ ISO2 ¼ 4p r Ca2þ þ r SO2 J SO2 ð11Þ 3gw rCa2þ r SO2
4 4 4 4

Substituting Eq. (10) into Eq. (11) gets In fact, not all of the particles effectively collide with each other.
For the effective collision, the particles must have enough energy,
collide in the specific direction, collide for enough time, and exist
in the external of the molecule structure. Hence, the calculative
result of the diffusion rate constant is always much larger than
the practical value. The probabilistic factor P for the molecule col-
lision is added to Eq. (16), and the diffusion rate constant is
 2
2kB T r Ca2þ þ r SO2
kG ¼ 4
NA  P ð17Þ
3gw rCa2þ r SO2
4

The radius of the ion is always dependent on the charge number


and the charge distribution. The radii of calcium ion and the sulfate
ion are 99 pm and 258 pm, respectively. At the temperature of
298.15 K, the value of gw is 0.8937  103 Pas. It can be calculated
that kG = 9.222  109P(Lmol1min1). The reaction rate of calcium
ion and the sulfate ion at the temperature of 298.15 K was tested
and the value of the reaction-rate constant is 0.058 Lmol1min1
(Appendix 1). Using this value, the probabilistic factor is found to
be 1.048  1013. Based on the value of the probabilistic factor,
Eq. (17) can be expressed as
 2
Fig. 5. Schematic diagram of Ca2+ surrounded by SO2
4 ions. - 13 2kB T r Ca2þ þ r SO2
kG ¼ 1:048  10   4
 NA ð18Þ
3gw r Ca2þ r SO2
4
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 7

From Eq. (18), the reaction rate of the calcium ion and the sul- 2.2.3. The volume of thenardite
fate ion at any temperature can be obtained. After saturated, sodium sulfate solution in the pore is supersat-
urated and mirabilite precipitates under the following WD cycles.
2.2.2. The volume of gypsum During the drying process, the mirabilite transforms to thenardite
For the concrete under WD cycles, the calcium ion, the sulfate and the volume of the thenardite after i WD times can be written
ion and the water molecule react in the pore solution and form as
gypsum. As solvent, water has no influence on the reaction rate vT
V T;i ¼ V M;i ð29Þ
of gypsum. For the reaction vM
Ca2þ þ SO2
4 þ 2H2 O ! CaSO4  2H2 O ð19Þ where VM,i is the volume of mirablite after after i WD times (L), VT,i is
its reaction rate can be expressed as the volume of thenardite after i WD times (L), vT is the molar vol-
ume of thenardite (L/mol) and vM is the molar volume of mirabilite
1 dn (L/mol).
cG ¼ ð20Þ
V p dt
2.2.4. The volume fraction of the crystal in the pore.
where n is the reaction degree (mol), t is the reaction time (s), and
The accumulation of crystals in the pore can be divided into two
Vp is the volume of the pore (L). According to law of mass
stages under WD cycles. Before the saturation of sodium sulfate
actioncG ¼ kG  cCa2þ  cSO2 , integrating cG from the original state
4 solution, only gypsum is generated in the pore. After the saturation
(t = 0) to the current state (t0), one gets: of sodium sulfate, mirabilite and gypsum are generated at the same
n ¼ kG  cCa2þ  cSO2  V p  t 0 ð21Þ time. The following investigates these two stages sequentially.
4

The particle number of the reaction product is: (1) The sodium sulfate solution is not saturation
NG ¼ N0G þn ð22Þ
At this stage the pore volume that the solution can invade to
where NG is the particle number of gypsum (mol), N 0G is the particle during each WD cycle was assumed to remain unchanged. If the
number of gypsum at the initial condition (mol). Substituting Eq. sodium sulfate solution is saturated after n WD cycles, the volume
(21) into Eq. (22) gets: of gypsum in the pore can be expressed as
nðn þ 1Þ
NG ¼ N0G þ kG  cCa2þ  cSO2  V p  t0 ð23Þ V G;n ¼  kG  V 0p  t 0  v G  c0SO2  cCa2þ ð30Þ
4
2 4

Due to no gypsum under the initial condition, the particle num- where V0p is the initial volume of the pore (L).
ber of gypsum isN 0G ¼ 0. The particle number of gypsum at the time As vG is a function of the density of gypsum (qG = 2.32  103 g/
of t0 can be written as L), and of the molar mass of gypsum (MG = 172 g/mol), the molar
NG ¼ kG  cCa2þ  cSO2  V p  t 0 ð24Þ volume of gypsum can be calculated by:
4
MG
Assuming that the wetting duration for each WD cycle is t0 and vG ¼ ¼ 74:53  10 - 3
L=mol ð31Þ
qG
the sodium sulfate solution is saturated after n WD cycles. Before
the sodium sulfate solution reaching saturation, it can be assumed Similar to Eq. (31), one can obtain that the molar volume of cal-
that the concentration of the sodium sulfate solution in the pore of cium hydroxide (v CaðOHÞ2 ) is 33.4  103 L/mol. As one molar calcium
concrete is an arithmetic progression whose common difference is hydroxide reacts with one molar sodium sulfate and produces one
the initial concentration of the sodium sulfate solution. Neglecting molar of gypsum, this would result in
the influence of the gypsum on the pore volume, the particle num-
V DG ¼ 0:55V G ð32Þ
ber of gypsum after n cycles of WD (N nG (mol)) can be obtained:
where VDG is the solid volume increment from the formation of
nðn þ 1Þ
NnG ¼  kG  V p  t 0  c0SO2  cCa2þ ð25Þ gypsum (L), VG is the volume of gypsum (L). Hence, the volume frac-
2 4
tion of the pore space filled with crystals after n WD cycles (Sn) can
where c0SO2 is the initial concentration of the sodium sulfate be expressed as
4

(mol/L). The volume of gypsum after n times of WD cycle 0:55V G;n nðn þ 1Þ 
Sn ¼ ¼ 0:55  kG  t 0  v G  c0SO2  cCa2þ ð33Þ
(VG,n (L)) can be expressed by V 0p 2 4

nðn þ 1Þ
V G;n ¼  kG  V p  t 0  v G  c0SO2  cCa2þ ð26Þ The volume of the pore space after n times of WD cycles (Vnp (L))
2 4
is
where vG is the molar volume of gypsum (L/mol).
With the increase of WD cycle, the concentration of sodium sul- V np ¼ V 0p  0:55V G;n ð34Þ
fate reaches saturation and keeps unchanged in the following
cycles. After i WD cycles, the particles number of gypsum
(N iG mol) can be expressed as (2) The sodium sulfate solution is saturation
1) Assuming that the sodium sulfate solution is saturated after
NiG ¼ i  kG  V p  t 0  csat
SO2
 cCa2þ ð27Þ
4 n WD cycles, the solution supersaturates and produces mir-
where csat is the saturation concentration of sodium sulfate abilite in the following cycles. After n + i + 1 WD cycles, the
SO2 4 volume in the pore, which can be used for the solution to
(mol/L). Hence, the volume of gypsum after i WD cycles invade (VQ (L)), is
(VG,i (L)) is
vT
V G;i ¼ NiG  v G ¼ i  kG  V p  t0  v G  csat  cCa2þ ð28Þ V Q ¼ V np  V M;i þV V ð35Þ
SO24 v M S G;i
8 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

where V S is the volume of the solid dissolved in the pore solution As the forming rate of gypsum is much slower than that of mir-
(L). From Eq. (32), it can be written by abilite [11,43] and as the expansion volume is only about half of
the gypsum, it can be assumed that
V S - V G;i ¼ - 0:55V G;i ð36Þ
!
Substituting Eq. (36) into Eq. (35), one gets the volume in the V M;i V M;i V G;i
 þ 0:55 n ¼ Si ð48Þ
pore that can be used for the solution to invade after i + 1 WD V np V np Vp
cycles:
Based on Eq. (48), Eq. (47) can be written as
vT
V Q ¼ V np  V M;i
vM
 0:55V G;i ð37Þ vT
V Q =V np ¼ 1 - Si  ð49Þ
vM
2) From Eq. (37), the volume of the mirabilite produced during In Eq. (45), assuming A = ðv M þ 0:55kG  t0  v G  cCa2þ Þcsat
SO2
and
the i + 1 wetting process (Vi+1
M (L)) can be calculated by:
4

substituting Eq. (49) into Eq. (45), one gets


   
vT vT
V iþ1
M ¼ V np  V M;i  0:55V G;i csat  vM ð38Þ Siþ1 ¼ Si 1  A  þA ð50Þ
vM SO2 4 vM
3) The volume of mirabilite in the pore after i + 1 times of WD It can also be assumed that
cycles can be expressed as vT
A ¼q
  vM
vT
V M;iþ1 ¼ V M;i þ V np  V M;i  0:55V G;i csat  vM ð39Þ then, Eq. (50) can be simplified as
vM SO2 4

Siþ1 ¼ Si ð1  qÞ þ A ð51Þ
Similarly, the volume of gypsum in the pore after i + 1 WD
cycles (VM,i+1 (L)) can be expressed as It can be seen that this is a geometric progression whose com-
 mon ratio is (1-q). The volume fraction of the crystals in the pore
V G;iþ1 ¼ V G;i þ kG  V Q  t0  v G  csat
SO2
 cCa2þ ð40Þ after one WD cycle (S1) can be calculated by
4

The volume increment produced by the formation of gypsum is V np  csat


SO2
 v M þ 0:55V G;1
S1 ¼ 4
ð52Þ
0:55V G;iþ1 ¼ 0:55V G;i þ 0:55  kG  V Q  t 0  v G  csat
SO2
 cCa2þ ð41Þ V np
4

where VG,i+1 is the volume of gypsum after i + 1 WD cycles. With Eqs. (51) and (52), the volume fraction of the crystals in
the pore after i times of WD cycles can be obtained by:
   
4) From Eqs. (39) and (41), the volume fraction of crystals in vM v T ði1Þ v M
Si ¼ A   1A þ ð53Þ
the pore after i + 1 WD cycles (Si+1) can be obtained by
vT vM vT
V C;iþ1 Eq. (53) can be simplified as
Siþ1 ¼ "
V np  i #
  vM vT
vT Si ¼
vT
1 1A
vM
ð54Þ
¼ ðV M;i þ V np  V M;i  0:55V G;i csat  v M Þ=V np
vM SO2 4

þ 0:55V G;iþ1 =V np ð42Þ where A = ðv M þ 0:55kG  t 0  v G  cCa2þ Þcsat


SO2
and is defined as the coef-
4

ficient which influences the growth rate of the crystal in the pore.
where VC,i+1 is the volume of the crystal in the pore after i + 1 WD
cycles (L). Eq. (42) can be simplified as
! (3) The total volume fraction of crystals in the pore
V M;i V G;i VQ
Siþ1 ¼ þ 0:55  n þ n  csat2 According to the volume fraction of crystals in the pore at the
V np Vp V p SO4
first and the second stages, the total volume fraction after m
 ðv M þ 0:55kG  t 0  v G  cCa2þ Þ ð43Þ (m = n + i) WD cycles (Sm) can be obtained by
As the volume fraction of crystals in the pore after i WD cycles 0:55V G;n þ Si  V np
(Si) is Sm ¼ ð55Þ
V 0p
V M;i V G;i
Si ¼ þ 0:55  n ð44Þ where V0p is the initial volume of the pore (L). Eq. (55) can be simpli-
V np Vp
fied as
Substituting Eq. (44) into Eq. (43) gets the volume fraction of
the crystals in the pore after i + 1 WD cycles Sm ¼ Sn þ Si ð1  Sn Þ ð56Þ

Siþ1 ¼ Si þ V Q =V np  csat
SO2
 ðv M þ 0:55kG  t 0  v G  cCa2þ Þ ð45Þ
4 Eq. (56) can be used to calculate the volume fraction of crystals
in the pore.
where
 
vT (4) The volume fraction of mirabilite and gypsum in the pore
V Q =V np ¼ V np  V M;i  0:55V G;i =V np ð46Þ
vM
After the sodium sulfate solution reaching saturation, the vol-
It also can be expressed as
! ume in the pore that is available for the invading solution is VRi
 
V M;i V G;i V M;i vT (L) at the number i WD cycle, where i = 1,2,3i. From Eqs. (39)
V Q =V np ¼1 þ 0:55 n þ 1 ð47Þ
V np Vp Vpn
vM and (41), after i WD cycles, the volume of mirabilite and gypsum
can be obtained by
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 9

  i 
V M;i ¼ csat v V þ csat v V þ    þ csat v V vM
2  v T
so2 M R1 so2 M R2 so2 M Ri Si ¼ 1  1  csat ð68Þ
4 4 4
vT SO
v ðV R1 þ V R2 þ    þ V Ri Þ
4
¼ csat
so2 M
ð57Þ
4
This result is consistent with the result proposed by Flatt et al.
    [11], verifying that the Eq. (56) is reasonable and reliable.
V G;i ¼ 0:55kG t0 csat
so2
cCa2þ v G V R1 þ 0:55kG t 0 csat
so2
cCa2þ v G V R2
 
4 4

þ 0:55kG t0 csatso2
cCa2þ v G V Ri 3. Crystallization pressure and the failure of materials
 
4

¼ 0:55kG t0 csat c
so2 Ca2þ G
v ðV R1 þ V R2 þ    þ V Ri Þ
4
ð58Þ 3.1. Crystallization pressure in the pore

According to Eqs. (57) and (58), the total volume of the crystals In 1939 Correns and Steinborn [81] proposed an equation for
in the pore after i WD cycles (VC,i (L)) can be expressed as crystallization pressure of one molar substance on the basis of
ideal solution and single molecule crystallization:
V C;i ¼ V M;i þ V G;i  0
  RT c
¼ csat v þ 0:55kG t0 csat
so2 M 4
c v ðV R1 þ V R2 þ    þ V Ri Þ
so2 Ca2þ G 4
Pc ¼
vc
ln 0
cs
ð69Þ

¼ AðV R1 þ V R2 þ    þ V Ri Þ where c’ is the concentration of the solution, and cs’ is the concen-
ð59Þ tration of the saturated solution. The equation is only valid for ideal
solutions of molecular solutes. Correns also put forward another
During the WD test with saturated sodium sulfate, the volume
equation for the crystallization pressure. One is the osmotic pres-
ratio of mirabilite and gypsum in the total volume after i WD cycles
sure [34]
can be obtained by:
Pc v c ¼ RTlnðp=ps Þ ð70Þ
V M;i csat
so2
vM csat
so2
vM
¼ sat 4
¼ 4
ð60Þ
V C;i cso2 v M þ 0:55kG t 0 csat c
so2 Ca
2þ v G A where p is the osmotic pressure (Pa) and ps is the saturated osmotic
4 4 pressure (Pa). The other is the saturation pressure

0:55kG t 0 csat cCa2þ v G 0:55kG t 0 csat cCa2þ v G Pc v c ¼ RTlnðcNa2 SO4 =cs Þ ð71Þ
V G;i so2 so2
¼ sat 4
¼ 4
ð61Þ
V C;i cso2 v M þ 0:55kG t 0 csat c 2þ v G
so2 Ca
A Based on the equations proposed by Correns, Steiger and
4 4
Asmussen [38] put forward the crystallization pressure of the sat-
Hence, the volume fraction of mirabilite (SM,i) and gypsum (SG,i) urated sodium sulfate solution
in the pore at the second stage can be expressed as:   
RT RHsat;T
 " # Pc ¼ lnK SP;T  lnK SP;M þ 10ln ð72Þ
V M;i v M vT i
V M;i vc 100
SM;i ¼ Si ¼ 1 1A ð62Þ
V C;i vT vM V C;i where RHsat,T is the deliquescence relative humidity of thenardite
" and is determined by the water activity in a non-ideal solution cal-
 #
V G;i v M vT i culation using Pitzer coefficients suited for a solution with particu-
V G;i
SG;i ¼ Si ¼ 1 1A ð63Þ
V C;i v T vM V C;i larly high concentrations. KSP,T and KSP,M are the thermodynamic
solubility products of thenardite and mirabilite, respectively.
From Eqs. (62) and (63), the volume fractions of mirabilite (SM) Gypsum is formed by ion reaction in the solution, and the crys-
and gypsum (SG) in the pore after the two stages are tallization pressure of gypsum is different from that of sodium sul-
fate. Based on the crystallization pressure equation proposed by
SM;i  V np
SM ¼ ð64Þ Correns, one can have the crystallization pressure for ion reaction
V 0p [24,39,42,82]:
 
0:55V G;n þ SG;i  V np RT Q
Pc ¼ ln reac ð73Þ
SG ¼
V 0p
ð65Þ vc K reac

According to Eqs. (33) and (34), Eqs. (64) and (65) are simplified where Qreac is the ionic activity product defined by
Y v reac
as Q reac ¼ ae e ð74Þ
SM ¼ SM;i ð1  Sn Þ ð66Þ e

where Kreact is the equilibrium constant of the chemical reaction, ae


SG ¼ Sn þ SG;i ð1  Sn Þ ð67Þ and v e
ðreacÞ
are the activity and the stoichiometric coefficient of ion e
in the solution, respectively.
For any salt M N1 X N2  N 3 H2 O, its electrolytic type can be
(5) The volume fraction of crystals in the pore under other expressed as
conditions M N 1 X N2  N 3 H2 O $ N 1 M þ N 2 X þ N 3 H 2 O ð75Þ

For some other materials, such as rock, only mirabilite is gener- At the dissolution equilibrium state, one has
ated in the pore during WD cycles with sodium sulfate solution.
N1 lMv þ þ N2 lXv  þ N 3 lW  lMN X N N 3 H2 O ¼0 ð76Þ
There is no calcium hydroxide which can react with the sodium 1 2

sulfate and produce gypsum. In Eq. (56), letting Sn ! 0, that is to


  According to the electrolyte balance theory, one has that
say kG  t 0  v G  csat
SO2
 cCa2þ ! 0, it can be obtained that Sm = Si
4
le ¼ lHe þ RTlnae ð77Þ
and A ¼ v M  csat
SO2
. Eq. (54) was simplified as
4
10 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

where le and lH e are the chemical potential and the standard chem-
3.3. Critical tensile stress of concrete
ical potential, respectively, at a given temperature and pressure; ai
is the ionic activity. Substituting Eq. (77) into Eq. (76) yields From Eqs. (66), (67), (72), (73) and (82), the macroscopic tensile
stress caused by the crystallization pressure in the pore can be cal-
lHMN1 XN2 N3 H2 O  N1 lHMv þ  N2 lHXv   N3 lHW culated. Then, it can be used to forecast the damage of the concrete.
K MN1 X N2 N3 H2 O ¼ exp ð78Þ The critical tensile stress of concrete is [86]
RT
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rc ¼ rT = 3ð1  2eÞ ð85Þ
where K MN1 XN2 N3 H2 O is the dissolution equilibrium constant of
MN1 X N2  N 3 H2 O. Harvie and Wear used Pitzer’s formula to calculate where rc is the critical tensile stress (MPa), rT is the tensile strength
the dissolution equilibrium constants of 12 kinds of ions, 55 kinds of the concrete (MPa) and e is the Poisson’s ratio. For concrete, the
of salts and water, at the temperature of 25 °C and the pressure Poisson’s ratio is usually 0.24–0.25 [85–87] and typically e = 0.25 is
of 101.325 KPa [83]. They have also given the standard chemical adopted. If the macroscopic tensile stress exceeds the critical stress,
potential of the ions, salts and water. Hence, the dissolution equilib- the sample will be damaged.
rium constant of gypsum (KG) at the temperature of 25 °C and the
pressure of 101.325 KPa can be calculated by: 4. Experiments
lG  lCa2þ  lSO2
H H H

K G ¼ exp 4
¼ 2:627  10 - 5
ð79Þ To validate the equation used to forecast the damage of con-
RT crete, the compressive strength, the mass and the ultrasonic veloc-
where lHG , lHCa2þ and lHSO2 are the standard chemical potential of ity of plain concrete samples subjected to WD cycles are tested
4
according to Chinese standard GB/T50082-2009. The damage time
gypsum, calcium ion, sulfate ion, respectively. The crystallization
of the samples was worked out.
pressure in the pore of concrete composes of the crystallization of
mirabilite and gypsum and can be expressed as
     4.1. Materials
RT RHsat;T RT Q
PC ¼ lnK SP;T  lnK SP;M þ 10ln þ ln reac ð80Þ
vM 100 v G K reac The cement, used in this study, was 42.5 R Qilianshan cement in
accordance with the relevant Chinese standard. The specific sur-
The crystallization pressure in the pore of rock under WD cycles
face area was 320 m2/kg, the standard consistency was 25%. The
with a sodium sulfate solution mainly comes from the crystalliza-
chemical compositions of the cement were given in Table 1. The
tion of mirabilite and can be calculated by Eq. (72).
sand used was natural river sand which was taken from Liujiaxia
in Gansu province. The fineness modulus of the sand used is
3.2. The macroscopic tensile stress
2.37. The apparent density and water content of the sand are
1.47 kg/cm3 and 1.16%, respectively. The apparent density and
The crystallization pressure in the pore is the microscopic force
water content of the aggregate are 2.96 kg/cm3 and 0.59%, respec-
and only can be felt at the pore scale. A macroscopic tensile stress
tively. The aggregate is basalt with silt content 0.01% and a contin-
must be calculated to assess possible damage to the porous host
uous gradation of 5–20 mm. The water reducer used was
[37,42]. For this, a poromechanic approach is used [37], assuming
polycarboxylate superplasticizer with a water reduction up to
that mirabilite and gypsum are homogeneously distributed through-
25%. The water used to make the samples was tap water and that
out the sample. Averaging the stress over a representative volume
used for sodium sulfate solution is deionized water.
element yields the macroscopic tensile stress (r* (MPa)) [37]:

r  ¼ r r b 0 Sc ð81Þ 4.2. Samples and curing


where b’ is the Biot coefficient; Sc is the volume fraction of the crys-
The concrete used in the test was ordinary plain concrete and
tals in the pore and rr is the radial compressive stress (MPa) which
its composition was given in Table 2. The w/c stands for water
can approximatively equal to the crystallization pressure in the
cement ratio. These materials in Table 2 were mixed and molded
pore (rrDP). Zhang et al. [84] investigated the correlation between
into cubes with a side length of 100 mm. There are three samples
Biot coefficient and hydration degree, and found that the Biot coef-
in one group and there are thirteen groups in total. Eleven groups
ficient of concrete is generally about 0.20–0.40. Ulm et al. [85]
are used for the test and the other two are spare samples. Some
found that the Biot coefficient of concrete is 0.23 ± 0.1. This study
other mixtures were sieved by the sieve No. 4.75 mm and then
then takes the value of the Biot coefficient as 0.23.
pure into the mould (40  40  160 mm3) to make the cement
As the crystallization pressure in the pore of the concrete comes
mortar samples. Three cement mortar samples were prepared.
from mirabilite and gypsum, the macroscopic tensile stress can be
All of the samples were vibrated compactly on the shaking table
expressed as
and then cured in the molds for 24 hours. 24 hours later, the sam-
r ¼ rG þ rM ¼ b0 ðPG SG þ PM SM Þ ð82Þ ples were pulled out and cured in a standard condition for 28 days
with an ambient temperature of 20 ± 2 °C and a relative humidity
where PG (MPa) and rG (MPa) are the crystallization pressure and
over 95%. After that, the samples were dried for 48 hours at the
the macroscopic tensile stress caused by gypsum, respectively; PM
temperature of 75 °C, and then the samples were used for the
(MPa) and rM (MPa) are the crystallization pressure and the macro-
WD cycle test.
scopic tensile stress caused by mirabilite, respectively.
If there is only mirabilite or gypsum crystal in the pore, the
4.3. Test methods
macroscopic tensile stress is

(1) WD cycle test


r ¼ rM ¼ b0 PM SM ð83Þ
The WD cycle test was conducted on samples after 28 day cur-
or ing according to Chinese GB/T50082-2009. The schematic of WD
cycle was illustrated in Fig. 6. During the wetting procedure, the
r ¼ rG ¼ b0 PG SG

ð84Þ solution in the liquid storage tank got into the space where the
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 11

Table 1
The chemical compositions of the cement (wt.%).

SiO2 Al2O3 Fe2O3 CaO MgO SO3


21.78 5.25 3.31 66.32 1.12 0.41

samples stayed in. The concentration of the sodium sulfate solu-


tion used was 5 wt%. The solution was renewed after 15 WD cycles
in order to maintain the concentration of the solution in the
required values. The stainless steel grids were used to make sure
that the samples were in full contact with the solution. During
the drying procedure, the solution was discharged into the storage
tank and the samples were dried by the heating tube. The drying
temperatures were controlled by control panel on the control
box. The test time was 24 hours for one cycle, among which
15 hours were used for wetting the samples and 9 hours drying
the samples. The drying procedure was further divided into four
stages: 2 hours for air drying, 5 h for drying with high temperature,
one and a half hours for air cooling and half an hour for refrigera-
tion. The high drying temperature and the cooling temperature Fig. 6. The schematic of the WD cycle.
were 75 °C and 25 °C, respectively. The airing drying and the air
cooling temperatures were the ambient temperature (The temper-
ature of the room where we conducted the experiment was about 30 WD cycles. To ensure the test effect, cement mortars chosen
20 ± 3 °C). were natural flatness. The cement mortar samples were explored
at different magnifications using a German MERLIN Compact
(2) Compressive strength device. The crystals at the surface of the cement mortar were
observed.
The compressive strength was measured according to Chinese
standard of JTG E30-2005. The compressive strength of one group (6) Energy Dispersive Spectrometer (EDS)
of samples after 28 day curing was tested to detect the initial
strength. During the WD cycles, the compressive strength was After the sample was observed by SEM at different magnifica-
measured on each 3 samples after each three cycles and the load- tion, crystals at the surface of the sample were analyzed by EDS.
ing rate is 0.5 MPa/s. The mean value of the three samples was In the EDS figure, different X coordinates correspond to different
taken as test result, if the differences between the maximum, min- elements. Hence, the element composition of the crystal can be
imum values and the median value were within 15% of the median obtained.
value, respectively. If one of the difference values is not within 15%
of the median value, the median value was used as the test result. If (7) Mercury Intrusion Porosimetry (MIP)
both of the differences were not within 15% of the median value,
the compressive strength should be tested again. Cylindrical samples, with height of 10 mm and diameter of
10 mm, were drilled from the cement mortar. The pore character-
(3) The test of the mass change istics of the samples were tested after suffering 0, 15 and 30 WD
cycles, respectively. The samples were put into absolute ethanol
The three samples, which were used for the compressive
to stop the hydration reaction until they were used for the test.
strength test after 30 cycles, were also used for the mass change
Before the test, the samples were taken out and dried at ambient
test. Their masses were tested after each WD cycle and recorded.
temperature. The MIP (Auto Pore IV 9500 type) with a maximum
pressure of 413.8 MPa was used to evaluate the porosity and
(4) The test of ultrasonic velocity
pore-size distribution of the samples. The density, contact angle
and the surface tensile of mercury were taken as 13.5335 g/ml,
The three samples used for the mass change testing were also
130° and 485 dynes/cm, respectively. The equilibrium time of both
used for the ultrasonic velocity test. The longitudinal wave veloci-
high pressure and low pressure were 10 s.
ties were tested on two fixed opposite surfaces of the samples. The
test was conducted after every three cycles and each sample was
tested for three times. The mean values were chosen as the test
4.4. Results and discussion
results.
To investigate the macro changes of the samples under WD
(5) Scanning Electron Microscope (SEM)
cycle, the compressive strength, the mass change and the ultra-
sonic velocity of the samples were tested during the experiment,
The cement mortar used for SEM was collected from the sample
because the compressive strength directly reflects the deteriora-
which was conducted by compressive strength test after suffering
tion of the sample, the mass change reflects the accumulation of

Table 2
The composition of 1 m3 concrete.

Cement (kg) Sand (kg) Aggregate (kg) Water reducer (kg) w/c
450 673 1097 4 0.4
12 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

is not serious. The mass subsequently has an increasing trend after


the ninth cycle. The reason is that new crystals form and fill the
cracks from the first damage, and thus increase the mass of the
sample. With the continuous increase of cycle index, the cracks
from the first damage will fully be filled again and this causes
the second damage. It can be seen that the mass of the sample
declines significantly after the fifteenth cycle. Because the sample
has been damaged before, the second damage is more serious than
the first time (Fig. 7). After that, the samples are damaged for the
third time and then repeat the cycle again. Besides, the increment
of the mass is larger than the previous one, inferring that the dam-
age is more and more serious. Although the mass curves of the
samples fluctuate locally, the overall trends of the masses decrease
firstly and increase finally. These overall trends are same as the
results of the previous studies [51–53]. Based on the change rule
of the mass, it can be considered that damage is generated again
and again in the sample during the WD cycle. Hence, it can be rede-
Fig. 7. The relationship between compressive strength and WD cycles. fined that the concrete is in a dynamic damaging-filling state dur-
ing the WD cycle.

expansive crystals in the pores and the spalling of mortar around 4.4.3. Ultrasonic velocity
the sample, and the ultrasonic velocity reflects the compactness Fig. 9 shows the ultrasonic velocity of concrete samples, which
of the concrete. After that, the SEM and MIP were conducted to reflects the compactness of the concrete. The ultrasonic velocity
investigate the microcosmic changes in the samples. The crystal increases at the first nine cycles and reaches the maximum at
type can be known by SEM, and the pore characteristics can be the ninth cycle. This trend illustrates that the concrete is not dam-
understood by MIP. aged before the ninth cycle and that the structure of concrete is
being denser. This finding is consistent with the change rule of
4.4.1. Compressive strength compressive strength. After the ninth cycle, the ultrasonic velocity
Fig. 7 shows the results of the compressive strength test. The decreases. This decreasing trend indicates the structure of the con-
compressive strength has a slight decreasing at the first 6 cycles. crete is damaged by the excessively crystallization pressure.
The reason may be that the sodium sulfate solution with the con- Although the crystals can fill the pores and cracks again after the
centration 5 wt% is not saturated until the end of the fifth cycle damage, the ultrasonic velocity still decreases. The reason is that
(Fig. 2). Until the sixth cycle, mirabilite precipitates and starts fill- the aggregate and the mortar in the concrete are closely bonded
ing the pores in the concrete. Hence, the damage in the concrete before damaging. After damage, the bonds are broken and the
before the sixth cycle comes mainly from the corrosion function structure is divided into mortar-crystal-aggregate state. There is
of sodium sulfate (Eqs. (3) and (4)) and the shrinkage in concrete a layer of air between the different phases. The ultrasonic velocity
during the drying stage. However, the corrosion reaction rate is in the air is much lower than that in the aggregate and the mortar.
too slow and the volume of expansion crystal is insufficient to The interface between different phases greatly influences the ultra-
damage the concrete. Therefore, it can be determined that the sonic velocity.
decrease of compressive strength is mainly caused by the shrink-
age of the concrete. The compressive strength reaches the maxi- 4.4.4. Microstructure analysis
mum at the ninth WD cycle. After that, the compressive strength The SEM was conducted on the cement mortar of the concrete
declines in a wave with the increase of cycle number, because mir- after 30 WD cycles. The image of the cement mortar is shown in
abitite forms and then fills the pores and microfracture caused by Fig. 10(a). There are many prismatic crystals in the pores. The ele-
the shrinkage. The compressive strength is enhanced with the mentary composition of the crystals was analyzed by the energy
gradually dense structure and reaches the maximum at the ninth spectrum (Fig. 10(b)). The main elements are calcium, oxygen
cycle. With the increase of the volumes of mirabilite and gypsum and sulfur. This indicates that gypsum is the main product during
in the pore, the crystallization pressure increases. When the crys-
tallization pressure exceeds the critical tensile stress of the con-
crete, the sample is damaged and the compressive strength
decreases with the increase of WD cycles.

4.4.2. Mass change


The masses of the samples after each cycle were tested and
shown in Fig. 8. It can be seen that the change rule has some differ-
ence with the previous studies [51–53]. In the previous studies, the
mass of the sample was tested every 15 or 30 cycles. The mass
always decreases firstly and then increases during the WD cycle.
With high frequency detection, more information can be obtained
from the results (Fig. 8). The mass has a little or no change at the
first eight cycles and has a relatively evident decrease at the ninth
cycle. This trend indicates that the concrete is damaged for the first
time during the ninth WD cycle. Some of the mortars are peeled off
from the sample and decrease the mass of the sample. Before the
first damage, the initial state of the sample is fine, so the damage Fig. 8. The relationship between the mass and the WD cycles.
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 13

the WD cycles and fills the pores in the concrete. Some cracks also increases after WD cycles, and it has a little change during the first
can be seen in the cement mortar and this indicates that concrete is 15 cycles and increases obviously during the latter 15 cycles. These
damaged during the WD cycles duo to the generation of expansion results also indicate that some micro-pores and macro-pores are
crystal. After that the structure of the cement mortar was observed changed into capillary pores by the generation of expansive crys-
at a larger magnification. It can be seen that there are some form- tals. When the volume of the expansive crystals is large enough,
less crystals in the pore which was filled with gypsum (Fig. 10(c)). the pore wall is damaged by the crystallization pressure and more
The main elements of the crystals are sodium, oxygen and sulfur pores are formed in the samples. Therefore, the porosity of the
(Fig. 10(d)). These formless crystals should be sodium sulfate crys- sample increases with the increase of WD times.
tal. This indicates that the pore solution saturates and sodium sul-
fate precipitates in the pore. 5. Model verification and discussion
The MIP was employed to explain the influence of WD cycles on
pore characteristics. The tests were conducted on the cement mor- To verify the volume fraction formula proposed in this paper,
tar samples after 0, 15 and 30 WD cycles. The pore diameter distri- the formula was used to calculate the damage time of the concrete
butions of the samples are shown in Fig. 11. Pores in the samples samples subjected to WD cycles.
include gel micro-pores (<10 nm), capillary pores (10–5000 nm)
and macro-pores (>5000 nm) [88]. Only one peak can be found 5.1. Damage of concrete under WD cycles
for the samples without WD attack. The most probable aperture
is about 77 nm. However, two peaks can be observed for the sam- The conditions in the WD cycles were used to calculate the
ples after 15 and 30 WD cycles. The peaks move to the two sides damage time of the concrete. The experimental conditions include
compared with the sample without WD attack. The peaks of the a wetting temperature of 298.15 K and a concentration of the
sample after 15 WD cycles are 70 nm and 151 nm. The peaks of sodium sulfate solution of 0.352 mol/L (50 g/L). Before the sodium
the sample after 30 WD cycles are 40 nm and 183 nm. The varia- sulfate solution was saturated, it is assumed that the concentration
tion of the most probable aperture indicates that some of the pores of the solution in the pore was an arithmetic progress with the
in the concrete are filled with the expansive crystals and the crys- increase of WD cycles. The solution in the pore was saturated at
tallization pressure of these expansive crystals results in more the fifth cycle, because the maximum concentration of the sodium
large pores in the concrete. With the increasing of WD cycle times, sulfate solution at the temperature of 298.15 K was 1.75 mol/L.
this trend is more and more obvious and aggravates the damage of From the sixth cycle, the concentration of the solution in the pore
concrete. In addition, the volumes of the pores, which are smaller was 1.75 mol/L and not changed at all. The concentrations of the
than 5 nm and larger than 6000 nm, also decrease with the increas- sodium sulfate solution in the pore at the first five cycles were
ing of WD cycle times. This indicates that some gel micro-pores are given in Table 3.
fully filled by the expansive crystals and then burst to capillary From Eq. (18), the reaction rate constant can be calculated by
pores. The macro-pores are absolutely filled and also change to
 2
capillary pores. r 2þ þ r
2kB T Ca SO 2
The critical apertures of the samples are then analyzed in kG ¼ 1:048  10 - 13   4
 NA
Fig. 12. The critical aperture refers to the through pore size in the 3gw r Ca2þ r SO2
4
concrete and is very important for the concrete. It can be seen that 1
1
the critical aperture is increased after WD cycles. The critical aper- ¼ 0:058Lmol min ð86Þ
tures after 15 and 30 cycles are about 1000–2000 nm, and the crit-
where the molar volume of thenardite is vT = 53.3  10 L/mol, the 3
ical aperture is only about 500 nm before the WD cycle. This
molar volume of mirabilite is vM = 217.7  103 L /mol, and the
indicates that the concrete is more easily damaged after WD cycle.
molar volume of gypsum is vG = 74.53  103 L /mol. The concen-
This is consistent with the test result of compressive strength.
tration of calcium hydroxide was assumed equal to its solubility
The proportions of gel micro-pores, capillary pores and macro-
at the temperature of 298.15 K. Hence, the concentration of the cal-
pores in the samples are shown in Fig. 13. It can be seen that the
cium hydroxide in the pore solution should be 2.146  103 mol/L.
volume of capillary pores increases with the increase of WD cycles.
For one WD cycle, the wetting time is 15 hours, i.e., 900 min. For
The volumes of micro-pores and macro-pores decrease with the
A = 0.3885, from Eqs. (60) and (61), the volume fractions can be
increase of WD cycles. Besides, it also can be seen that the porosity
obtained by
vM
sat
V M;i cso2
¼ 4 ¼ 0:0979 ð87Þ
V C;i A

cCa2þ v G
sat
V G;i 0:55kG t 0 cso2
¼ 4
¼ 0:021 ð88Þ
V C;i A
V M;i V
Substituting the values of A, V C;i
; V G;i into Eqs. (62) and (63), the
C;i

volume fractions of mirabilite and gypsum in the pore at the sixth


to tenth WD cycles (as n = 5, i = 1,2,3,4,5) can be calculated and
were given in Table 4. After that, the volume fractions of gypsum
in the first five cycles (n = 1,2,3,4,5) can be calculated by Eq. (33)
and were given in Table 5. Finally, the total volume fractions of
mirabilite and gypsum can be calculated by Eqs. (66) and (67)
(Table 6).
The crystallization pressures of mirabilite and gypsum can be
calculated by Eqs. (72) and (73), respectively. At the wetting stage
of the WD cycles, once the solution arrives at the pore, the thenar-
Fig. 9. The correlation between the ultrasonic velocity and the WD cycles. dite in the pore will dissolve in the solution, even make the
14 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

Fig. 10. The test of SEM: (a) The SEM image of the cement mortar at X1000 magnification; (b) The energy spectrum of gypsum; (c) The SEM image of the cement mortar at
X1780 magnification, (d) The energy spectrum of the sodium sulfate.

Fig. 11. The pore size distributions of the cement mortar samples.

solution supersaturate and precipitate as mirabilite. Assuming that that, with the continuous wetting of concrete, the calcium hydrox-
the generation of mirabilite is completed, the solution is pure. ide dissolves in the solution, reacts with sodium sulfate and gener-
Hence, KSP,T and KSP,M can be calculated by the Pitzer formula for ates gypsum. The solubility of calcium hydroxide is assumed as a
single component electrolyte (Appendix 2) [89]. Substituting the constant. If the sodium sulfate consumes the calcium hydroxide
values of KSP,T and KSP,M into Eq. (72), one obtains the crystallization in the solution, new calcium hydroxide will dissolve and make sure
pressure of mirabilite. The calculated result is 9.3055 MPa. After that the concentration of calcium hydroxide is a constant. Hence,
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 15

the concentration of calcium hydroxide can be considered as a con-


stant during the wetting. During the generation of gypsum, it can
be assumed that there are only sodium sulfate and calcium
hydroxide in the solution. Because the contents of the other ions
are very low, their influence on the activities of the reaction ions
can be neglected. Therefore, the ionic activity product of gypsum
(QG) can be calculated by the Pitzer formula for mixed electrolyte
which proposed by Harvie and Wear (Appendix 3) [90]. As the
sodium sulfate solution in the pore is saturated after five cycles,
the ionic activity products of gypsum in the following cycles are
equal to those in the fifth cycle. Therefore, the ionic activity pro-
duct of gypsum in the first five WD cycles (Table 7) was only cal-
culated. From the results in Table 6, the crystallization pressure
of gypsum in the first WD cycles can be obtained (Table 8).
Because the ionic activity product of gypsum is a constant after
the fifth cycle, the crystallization pressure of gypsum after the fifth
cycle is a constant as well. The total macroscopic tensile stress that
Fig. 12. The pore size distribution of the cement mortar under cumulative
originates from the macroscopic tensile stress of mirabilite and
intrusion.
gypsum can be calculated (Table 9).
The tensile stress of C40 ordinary concrete is usually 3.9 ± 0.1
MPa [91–93]. The critical tensile stress can be obtained by Eq. (85):
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rc ¼ rT = 3ð1  2eÞ ¼ 3:184MPa ð89Þ

Table 6
The total volume fraction of mirabilite and gypsum in the pore.

Cycle number SM SG
1 0.000 1.616  103
2 0.000 4.847  103
3 0.000 9.694  103
4 0.000 0.016
5 0.000 0.024
6 0.371 0.032
7 0.707 0.040
8 1.010 0.045
9 1.284 0.051
10 1.533 0.056
Fig. 13. Porosities of the cement mortar after 0, 15 and 30WD cycles.

Table 3
The concentration of sodium sulfate solution in the pore.

Cycle number 1 2 3 4 5
Concentration of sodium sulfate solution (mol/L) 0.352 0.704 1.056 1.408 1.750

Table 4
The value of SM,i and SG,i from the sixth to tenth WD cycle.

Cycle number 6 7 8 9 10
SM,i 0.380 0.724 1.035 1.316 1.571
SG,i 8.148  103 0.016 0.022 0.028 0.033

Table 5
The volume fraction of gypsum in the pore at the first five WD cycles.

Cycle number 1 2 3 4 5
Sn 1.616  103 4.847  103 9.694  103 0.016 0.024
16 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

Table 7
The ion activity product of gypsum.

Cycle number 1 2 3 4 5
6
Ion activity product (10 ) 7.47855 6.89370 6.30605 5.80789 5.38438

Table 8
The crystallization pressure of gypsum.

Cycle number 1 2 3 4 5
Crystallization pressure (MPa) 24.321113 26.49749 28.088335 29.33512 30.379922

After calculating the crystallization pressure, the macroscopic


Table 9
tensile stress can be calculated using the values of the crystalliza-
The macroscopic tensile stress.
tion pressure and volume fraction of mirabilite in the pore by
Cycle Tensile stress of Tensile stress of Total tensile Eq. (81). The values of macroscopic tensile stress under different
number mirabilite (MPa) gypsum (MPa) stress (MPa)
WD cycles were given in Table 11.
1 0 0.009040 0.009040 The critical tensile stress was calculated by Eq. (85) and the
2 0 0.029540 0.029540
value is
3 0 0.062626 0.062626
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4 0 0.107953 0.107953 rc ¼ rT = 3ð1  2eÞ ¼ 2:25MPa ð91Þ
5 0 0.167697 0.167697
6 0.794038 0.223596 1.017635 where the value of Poisson’s ratio is 0.26 [11] and the tensile
7 1.513167 0.279495 1.792663
stress is 2.7 MPa for the rock. It can be seen that the rock was dam-
8 2.161668 0.314432 2.476100
9 2.748100 0.356356 3.104457 aged after the fourth cycle and this is consistent with the test result
10 3.281026 0.391293 3.672320 of Flatt et al. [11]. Therefore, the volume fraction model, proposed
by this paper, is not only useful for the concrete but also useful for
the noncorrosive materials, such as rock.

6. Conclusions
Comparing the calculated result with the macroscopic tensile
stress in Table 8, it is found that the macroscopic tensile stress To investigate the damage process and mechanism of concrete
was close to the critical tensile stress in the ninth cycle and under WD cycles, the crystal growth and the crystallization pres-
exceeded the critical tensile stress in the tenth cycle. The concrete sure were calculated and the WD cycles that the plain concrete
was damaged after the tenth cycle. This variation regularity is con- can sustain was forecasted. The following conclusions can be
sistent with that of experimental result in which the compressive made:
strength and the ultrasonic velocity of the concrete decreased after
the ninth cycle. Although the mass of the concrete decreased after (1) The damage in the concrete composes of physical damage
the eighth cycle, the deviation is within the range of error. The pro- caused by the crystallization of mirabilite, and the physico-
posed models, in this paper, were thus reliable to forecast the dam- chemical damage which refers to the losing function of gel
age of the concrete. material and the generation of expansion crystal in the cor-
rosion reaction. The type and content of the expansive crys-
5.2. The damage in corrosion-free material under WD cycles tal change with the concentration of sodium sulfate solution.
A high concentration leads to gypsum precipitation, while a
If there is no reaction in the material under WD cycles, only low one, ettringite formation. However, there always gener-
mirabilite generates in the pore of the material. The rock studied ate gypsum at the wet-dry interface of concrete in field and
by Flatt et al. [11] was calculated as an instance to verify the pro- in the pore of the concrete in the laboratory.
posed model in this paper. The volume fraction of mirabilite in the (2) The reaction of the generation of gypsum is diffusion-
pore was calculated by Eq. (68) at the temperature of 298.15 K. The controlled liquid ionic reaction. The reaction-rate constant
results were given in Table 10 (i = 1,2,3,4,5). formula was derived based on the diffusion controlled reac-
From Eq. (72), the crystallization pressure of mirabilite can be tion theory. The formula for collision probabilistic factor was
calculated by worked out and then the reaction rate constant at any tem-
   perature can be calculated by the formula.
RT RHsat;T (3) Based on the chemical reaction dynamics and phase transi-
Pc ¼ K SP;T  K SP;M þ 10ln ¼ - 9:3055MPa ð90Þ
vM 100 tion laws, the volume formulae of gypsum and mirabilite
in the pore at any WD cycles were proposed. The total vol-

Table 10
Volume fraction of the crystals in the pore.

Cycle number 1 2 3 4 5
Volume fraction 0.0932 0.1844 0.2730 0.3603 0.4453
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 17

ume fraction formula and the volume fraction formulae of SO2


4 þ Ca

þ 2H2 O ! CaSO4  2H2 O: ðA1:1Þ
mirabilite and gypsum in the pore were derived for the con-
crete samples under WD cycles, in the conditions that the As water is the solvent, the reaction rate mainly depends on the
sodium sulfate solution saturates or not saturate. concentrations of calcium and sulfate ions and on the temperature
(4) By testing the compressive strength, mass change and the of the solution. The reaction rate constant is mainly dependent on
ultrasonic velocity of the concrete in laboratory, the number the temperature. Under the condition that the wetting tempera-
of WD cycle that the concrete can sustain under the experi- ture in the WD cycle testing was 298.15 K, the reaction rate con-
mental condition was determined. Under the experimental stant was measured at the temperature of 298.15 K. Here the
condition, the macroscopic tensile tress was calculated using electric conductivity rate instrument was used to test the reaction
the volume fraction in the pore and the crystallization pres- rate.
sure in the pore. By comparing the macroscopic tensile tress During the WD cycles, the reaction of calcium hydroxide and
with the critical tensile stress of the concrete, the cycle num- the sodium sulfate forms gypsum. Hence, reactants used for testing
ber that the concrete can sustain was obtained. The theoret- the reaction rate constant should be calcium hydroxide and
ical result is consistent with the experimental result. The sodium sulfate. However, the calcium hydroxide solution also
proposed formulae were also used to calculate the damage reacts with the carbon dioxide and generates calcium carbonate
time of the rock which does not react with sodium sulfate. (Eq. (A1.2)). Moreover, the solubility of calcium carbonate is smal-
The result is also consistent with Flatt’s experimental result, ler than that of calcium sulfate. Hence, if the calcium hydroxide is
verifying the reliability of the formulas. chosen as the reactant, the reaction product is more likely to be
(5) Most of the current studies thought that the mass simply calcium carbonate rather than the calcium sulfate. Therefore, the
increases with the increase of WD times. By measuring the calcium chloride was chosen to replace the calcium hydroxide.
mass of the test concrete after each cycle, a different change Because the calcium chloride is also strong electrolytes and can
rule was found. The mass has few changes at the initial provide equal numbers of calcium ion in the solution. This replace-
cycles and then presents an increasing in wave. Hence, the ment will not influence the reaction rate and it is reasonable.
concrete is not straightly damaged but in a dynamic
Ca2þ þ 2OH1 þ CO2 ! CaCO3 þ H2 O ðA1:2Þ
damaging-filling state under the WD cycles.

CRediT authorship contribution statement A1.2.2. Experiment


0.1 mol/L sodium sulfate solution and 0.1 mol/L calcium chlo-
Jingge Ren: Conceptualization, Data curation, Validation, Inves- ride solution were prepared and poured into the beakers. Before
tigation, Writing - original draft, Writing - review & editing. Yuan- the experiment, the two solutions in the beakers were placed in
ming Lai: Conceptualization, Writing - review & editing, the water bath with the temperature of 298.15 K. After 10 min,
Resources, Supervision, Funding acquisition. Ruiqiang Bai: Writing the solution were taken out if their temperatures reached to
- review & editing. Yinghong Qin: Writing - review & editing. 298.15 K. Equal volumes of these two solutions were taken out
and mixed in a beaker which was also put in the water bath with
Declaration of Competing Interest the temperature of 298.15 K. At the same time, the computing time
was started. The conductivity of the mixed solution was detected
The authors declare that they have no known competing finan- and recorded each 10 min for four times, then the conductivity
cial interests or personal relationships that could have appeared of the mixed solution was detected and recorded each 20 min for
to influence the work reported in this paper. another four times. The test was completed after the eight values
were got. After that, the conductivities of 0.05 mol/L sodium sulfate
Acknowledgements solution and calcium chloride solution at the temperature of
298.15 K were tested and recorded.
This research was supported by National Key Research and
Development Program of China [grant number A1.3. Results and calculation
2018YFC0809605], Key Research Program of Frontier Sciences of
Chinese Academy of Sciences [grant number QYZDY-SSW- As the reaction between calcium and sulfate ions is the second
DQC015], National key Basic Research Program of China [973 pro- order reaction, when the reaction lasts for a time of t, the reaction
gram number 2012CB026102], National Natural Science Founda- rate can be expressed as
tion of China [grant numbers 41230630].
dx  
¼ kG  c0SO2  x  c0Ca2þ  x : ðA1:3Þ
Appendix 1. Measure of the reaction rate constant of calcium dt 4

ion and the sulfate ion


where x (mol) is the variation of the reactant at the time of t (min),
c0SO2 (mol/L) and c0Ca2þ (mol/L) are the initial concentration of cal-
A1.1. Material 4

cium ion and sulfate ion and kG is the reaction rate constant of
The testing reagents include sodium sulfate, calcium chloride, the formation of gypsum. If the initial concentration of sulfate ion
and distilled water. The instrument includes cylinders, beakers, equals to that of calcium ion, Eq. (A1.3) can be simplified as
weighing papers, droppers, a balance with the precision 0.001 g,  2
a conductivity meter (DDS-307), and constant temperature baths. dx
¼ kG  c0SO2  x : ðA1:4Þ
dt 4

A1.2. Test method and experiment Integrating Eq. (A1.4) yields

A1.2.1. Test method 1 x


kG ¼  : ðA1:5Þ
The reaction of calcium and sulfate ions in the solution can be tc0SO2 c0SO2  x
written as 4 4
18 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

Table 11
Macroscopic tensile stress.

Cycle number 1 2 3 4 5
Tensile stress (MPa) 0.6071 1.2698 1.8799 2.4811 3.0644

However, the concentration of the ions in the solution cannot be


measured directly. Hence, the reaction rate was reflected by the
conductivity of the solution. The conductivity of the solution
always depends on the concentration of metal ions and the tem-
perature of the solution. It can be known that the concentration
of the mixed solution is 0.05 mol/L as the concentrations of both
sulfate and calcium ions are 0.1 mol/L. If the conductivities of
0.05 mol/L calcium chloride solution and 0.05 mol/L sodium sulfate
solution are LCa2þ and LSO2 , respectively, the conductivity of the
4

mixed solution (L0), which is composed of the same volume of


0.1 mol/L calcium chloride solution and 0.1 mol/L sodium sulfate
solution, is equal to LCa2þ þ LSO2 . Hence, at the time of t, the con-
4

ductivity of the mixed solution is Lt. Eq. (A1.5) can be written as


1 L0  Lt
kG ¼  : ðA1:6Þ
tc0SO2 Lt
4
L0 Lt
Fig. A1. The correlation between t
and Lt.
Eq. (A1.6) can also be expressed as
1 L0  Lt
Lt ¼  : ðA1:7Þ
kG c0SO2 t Table A2
4 The fitting equation and coefficients.

There is a linear correlation between L0 L


t
t
and Lt as the reaction Equation R2 Slope (p) Intercept (q)
1
is a second order reaction. The slope of the correlation line is k 0 . y = px + q 0.9183 343.00 11,384
Gc
SO2
4

Hence, the reaction rate constant can be calculated by the slope.


The test results were given in Table A1. The conductivities of
0.05 mol/L sodium sulfate solution and calcium chloride solution This equation neglects the other force except for Coulombic
were tested and they were 15.02 ms and 16.76 ms, respectively. force between the ions and it is only useful for the solution whose
According to the results in Table A1, the correlation between L0 L
t
t
concentration is smaller than 0.1 mol/L. Based on Debye-Hückel’s
and Lt can be analyzed (Fig A1). equation, Pitzer (1973) proposed formulae to calculate the average
L0 Lt
t
and Lt follows a linear correlation, with a R2 is 0.9183 activity coefficient of the electrolyte and the permeability
(Table A2). From Eq. (A1.7) we can obtain: coefficient of the solution. It is available for the high concentration
solution with a concentration up to 6 mol/L. Hence, the formulae of
1
kG c0SO2 ¼ 343 min : ðA1:8Þ Pitzer were used to calculate the average activity coefficient of the
4
electrolyte and the permeability coefficient of the solution.
kG = 0.058Lmol1min1 can be obtained from Eq. (A1.8). For the electrolyte MX whose molar concentration is m, the for-
mulae to calculate the average activity coefficient (c±MX) and the
Appendix 2. The calculation of the equilibrium constants of permeability coefficient (u) are expressed as [88]
thenardite and mirabilite 2dM dX c 2ðdM dX Þ
3=2
c
lnc MX ¼ jZ M Z X jf þ m BMX þ m2 C cMX ðA2:2Þ
d d
A2.1. The formulae of Pitzer
3=2
2dM dX u 2ðdM dX Þ
In the electrolyte solution, the calculation of equilibrium con- u  1 ¼ jZ M Z X jf u þ m BMX þ m2 Cu
MX : ðA2:3Þ
d d
stant of the electrolyte is crucial to calculate the ionic activity.
Debye-Hückel firstly proposed the model to calculate the ionic where M, X are the cation and the anion, dM, dX are the number of
activity of the electrolyte: the cation and the anion, and Z M ; Z X are the charge number of the
cation and the anion, respectively. d = dM + dX is the total ion num-
pffiffi c u
Ajzþ z j I ber within the electrolyte. f ; f ; BcMX and Bu
MX are all the functions of
lgc ¼ pffiffi : ðA2:1Þ the ionic strength (I) and they are defined as:
1 þ aB I

Table A1
The conductivity of the mixed solution.

t(min) 10 20 30 40 60 80 100 120


Lt(ms/cm) 15,760 15,640 15,530 15,430 15,240 15,050 14,860 14,680
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 19

" #
c I1=2 2   The dissolution equilibrium constant of thenardite is
f ¼ Au þ ln 1 þ bI1=2 ðA2:4Þ
1 þ bI1=2 b
w ¼ mNaþ cNaþ mSO2 cSO2 aw :
K Na2 SO4 ¼ ðaNaþ Þ2 aSO2 a10 ðA2:15Þ
2 2 10
4 4 4
!
u I1=2 The water activity (aw) equation is
f ¼ Au 1=2
ðA2:5Þ X
1 þ bI lnaw ¼ u ðme =55:508Þ: ðA2:16Þ
   
where me (mol/kg) is the molar concentration of the ion (e). The
BcMX ¼ 2bMX þ bMX g a1 I1=2 þ bMX g a2 I1=2
ð0 Þ ð1 Þ ð2Þ
ðA2:6Þ
value of water activity at the temperature of 298.15 K is 0.9428.
From Eqs. (A2.14) and (A2.15), the ratio of K Na2 SO4 to K Na2 SO4 10H2 O
ð0 Þ ð1 Þ ð2 Þ
B/MX ¼ bMX þ bMX ea1 I
1=2 1=2
þ bMX ea2 I ðA2:7Þ can be calculated by

3 u K Na2 SO4
C cMX ¼ C : ðA2:8Þ ¼ a10
w : ðA2:17Þ
2 MX K Na2 SO4 10H2 O

where b(0), b(1), b(2), and C u are the Pitzer parameters of the elec- After that, the ratio of K Na2 SO4 to K Na2 SO4 10H2 O can be calculated by
trolyte; b = 1.2 kg1/2/mol1/2 is an empirical constant; Au is the per- the value of water activity.
meability coefficient of Debye-Hückel which is determined by the
property of the solution and the temperature. It is defined as: Appendix 3. The calculation of the ionic activity and
 1=2  2 3=2 permeability coefficient in the mixed solution of sodium sulfate
1 2p N 0 q W ee
Au ¼ : ðA2:9Þ and calcium hydroxide
3 1000 DkB T
where N0 is the Avogadro constant; kB is the Boltzmann constant; ec A3.1. Calculation formulae
is the charge of the electron; qw is the density of the solvent at the
temperature of T; D is the dielectric constant of the solvent. The During the wetting stage of the WD cycles, the hydration pro-
value of Au is 0.3915 kg1/2/mol1/2 at the temperature of 298.15 K. duct calcium hydroxide dissolves in the sodium sulfate solution
ð2Þ on the pore. The solution in the pore is a mixed electrolyte solu-
For non-2-2 electrolyte, the value 0f bMX is zero in Eqs. (A2.6) and
tion. Pitzer formulae were arranged by Harvie and Wear; the for-
(A2.7). The value of a1 is 2 kg1/2/mol1/2. For 2-2 electrolyte, the val-
mulae used for calculating the ionic activity in the mixed
ues of a1 and a2 are 1.4 kg1/2/mol1/2 and 12 kg1/2/mol1/2, respec-
solution were obtained. The calculation formulae are [92]
tively. The I in Eqs (A2.6) and (A2.7) is the ionic strength and
  N Na

defined as:
Ri mi ðu  1Þ ¼ 2½Au I3=2 = 1 þ 1:2I1=2 þ
c
R R mc ma Buca þ ZC ca
1X  
c¼1 a¼1
I¼ mi Z 2i : ðA2:10Þ N c 1 Nc Na N a 1 Na
2 þ R R mc mc0 Uucc0 þ R ma wcc0 a þ R R ma ma0
c¼1 c0 ¼cþ1 a¼1 a¼1 a0 ¼aþ1
The expression of g ðxÞ is  
Nc Na Na Nn Nc

   Uuaa0 þ R mc waa0 c þ R R mn ma knc þ R R mn mc knc
1 c¼1 n¼1 a¼1 n¼1 a¼1
g ðxÞ ¼ 2 1  1 þ x  x2 ex =x2 ðA2:11Þ
2 ðA3:1Þ

Na
A2.2. The calculation of the equilibrium constants of thenardite and lncM ¼ z2M F þ R ma ð2BMa þ ZC Ma Þ
a¼1
mirabilite   N 1
Nc Na a
þ R mc 2UMc þ R ma wMca þ R
Sodium sulfate is 2–1 electrolyte. The average ionic activity of c¼1 a¼1 a¼1

sodium sulfate and the permeability coefficient of the solution Na Nc Na

can be obtained by:  R ma ma0 waa0 M þ jzM j R R mc ma C ca


a0 ¼aþ1 c¼1 a¼1

4 c 25=2 2 c Nn
lnc MX ¼ 2f þ mBcMX þ m C MX ðA2:12Þ þ R mn ð2knM Þ ðA3:2Þ
3 3 n¼1

4 25=2 2 u Nc
u  1 ¼ 2f u þ mBuMX þ m C MX : ðA2:13Þ lncX ¼ z2X F þ R mc ð2BcX þ ZC cX Þ
3 3 c¼1
  N 1
The Pitzer parameters of sodium sulfate are: b(0) = 0.0196, Na Nc c Nc
þ R ma 2UXa þ R mc wXac þ R R mc mc0 wcc0 X
b = 1.113 and C u ¼ 0:00497. The concentration of the saturated
(1) a¼1 c¼1 c¼1 c0 ¼cþ1

sodium sulfate solution is 1.75 mol/kg. Hence, the values of Nc Na Nn


c u
f ; f ; BcMX and Bu þ jzX j R R mc ma C ca þ R mn ð2knX Þ ðA3:3Þ
MX can be calculated by Eqs (A2.3-A2.6). Substitut- c¼1 a¼1 n¼1
ing the results into Eqs (A2.12) and (A2.13), one can obtain that the
value of the average ionic activity of sodium sulfate is 0.1638 and where M, c and c0 are the cations; X, a and a0 are the anions; Nc, Na
the value of the permeability coefficient of the solution is 0.6225 at and Nn are the number of the cation, anion and neutral molecule,
the temperature of 298.15 K. According to the principle of chemical respectively; cM, ZM and mc are the activity coefficient, ionic
equilibrium, when a salt reaches the dissolution equilibrium in valence and the molality of the cation; cX, ZX and ma are the activity
solution at constant temperature and pressure, the dissolution coefficient, ionic valence and the molality of the anion; and cn, mn,
equilibrium constant of the salt is a constant. Therefore, the disso- knc and kna are the activity coefficient of the neutral molecule, the
lution equilibrium constant of thenardite is molality of the neutral molecule, the interaction coefficient
between the cation and the neutral molecule and the interaction
 1
K Na2 SO4 ¼ ðaNaþ Þ2 aSO2 ¼ m2Naþ c2Naþ mSO2 cSO2 : ðA2:14Þ coefficient between the anion and the neutral molecule,
4 4 4
respectively.
20 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

F, C, Z, Bu and B are defined as: Table A3


h    i Nc Pitzer parameters.
F ¼ Au I1=2 = 1 þ 1:2I1=2 þ 2=1:2ln 1 þ 1:2I1=2 þ R Electrolyte bð0Þ bð1Þ bð2Þ Cu
c¼1
Na N c1 Nc N a1 Na2 SO4 0.04604 0.93350 – 0.00483
 R mc ma B0ca þ R R mc mc0 U0cc0 þ R CaðOHÞ2 0.1747 0.2303 5.72 –
a¼1 c¼1 c0 ¼cþ1 a¼1
NaOH 0.17067 0.08411 – 0.00342
Na
CaSO4 0.2 3.7762 58.388 –
 R ma ma0 U0aa0 ðA3:4Þ
a0 ¼aþ1
 
C MX ¼ C u
MX = 2jzM zX j
1=2
ðA3:5Þ
Table A4
Z ¼ Ri¼1 jzi jmi ðA3:6Þ The interaction parameters of the ions.

    hNa,Ca wNa,Ca,SO4 wNa,Ca,OH hSO4,OH wSO4,OH,Na wSO4,OH,Ca


Bu
ð0Þ ð1Þ 1=2 ð2Þ
CA ¼ bCA þ bCA exp a1 I þ bCA exp a2 I1=2 ðA3:7Þ 0.07 0.055 0 0.013 0.009 0

   
ð0Þ ð1Þ ð2Þ
BCA ¼ bCA þ bCA g a1 I1=2 þ bCA g a2 I1=2 ðA3:8Þ

h    i Hence, the values of Ehij and Eh0ij can be calculated by Eqs.


ð1Þ ð2Þ
B0CA ¼ bCA g 0 a1 I1=2 þ bCA g 0 a2 I1=2 =I ðA3:9Þ (A3.15-A3.19). Substituting the results to Eqs. (A3.12-A3.14), one
gets the values of Uu 0
ij ; Uij and Uij . Finally, the ionic activity of the
where Au is the permeability coefficient of Debye-Hückel and the
ions and the permeability coefficient of the mixed solution can
value of Au is 0.3915 kg1/2/mol1/2 at the temperature of 298.15 K;
be obtained.
w is the interaction parameter of three different ions (two cations
and one anion or two anions and one cation); Bu and B are the sec-
A3.2. The calculation of ionic activity and permeability coefficient
ond virial coefficients; B0 is the differentiation of B about ionic
ð0Þ ð1Þ ð2Þ
strength. bCA ;
and bCA bCA
in Eqs. ((A3-7)-(A3-9)) are the Pitzer
Assuming that there are only sodium sulfate and calcium
parameters of the electrolyte. For non-2-2 electrolyte, the values hydroxide in the pore solution, the formulae arranged by Harvie
ð2Þ
of bCA and a2 equal to zero, and a1 is 2 kg1/2/mol1/2. For 2-2 elec- and Wear were used to calculate the ionic activity and the perme-
trolyte, the values of a1 and a2 are 1.4 kg1/2/mol1/2 and 12 kg1/2/- ability coefficient. Pitzer parameters of sodium sulfate, calcium
mol1/2, respectively. hydroxide, calcium sulfate and sodium hydroxide were given in
g and g0 are defined as: Table A3. The interaction parameters of the ions were given in
Table A3.
g ðxÞ ¼ 2½1  ð1 þ xÞexpðxÞ =x2 ðA3:10Þ
Based on the values in Tables A3 and A4, the activity coefficients
of the ions in the mixed solution at different concentration can be
g 0 ðxÞ ¼ 2 1  1 þ x þ x2 =2 expðxÞ =x2 ðA3:11Þ
calculated.
Uuij ; Uij and U0ij are also the second virial coefficients; they corre-
late with the ionic strength and can be expressed as: References
u
Uij ¼ hij þ Ehij þ IEh0ij ðA3:12Þ [1] A.S. Goudie, H.A. Viles, Salt Weathering Hazard, Wiley, Chichester, 1997.
[2] G.A. Novak, A.A. Colville, Efflorescent minerals assemblages associated with
cracked and degraded residential concrete foundations in Southern California,
Uij ¼ hij þ Ehij ðA3:13Þ Cem. Concr. Res. 19 (1989) 1–6, https://doi.org/10.1016/0008-8846(89)90059-8.
[3] M.A. Halliwell, N.J. Crammond, in: Durability of Building Materials and
U0ij ¼ Eh0ij : ðA3:14Þ Components, E&FN Spon, London, 1996, pp. 235–244.
[4] D.A. St, John, An unusual case of ground water sulphate attack on concrete,
Cem. Concr. Res. 12 (1982) 633–639, https://doi.org/10.1016/0008-8846(82)
where hij is the Pitzer force parameter between two different
90025-4.
cations or anions. When the charge of i is equal to that of j, the val- [5] S. Chatterji, A.D. Jensen, Efflorescence and breakdown of building materials,
ues of Ehij and Eh0ij are zero. Nord. Concr. Res. 8 (1989) 56–61. http://xueshu.baidu.com/
9eb9d3c12ab3670613eca29b51215255&site=.
Ehij ¼ Z i Z j =ð4IÞ J xij  J ðxii Þ=2  J xjj =2 ðA3:15Þ [6] Y.L. Wang, X.Y. Guo, S.Y.H. Shu, Y.C. Guo, X.M. Qin, Effect of salt solution wet-
dry cycling on the bond behavior of FRP-concrete interface, Constr. Build.
Mater. 254 (2020) 119317, https://doi.org/10.1016/
Eh0ij ¼  Ehij =I j.conbuildmat.2020.119317.
h  i [7] G.W. Scherer, Stress from crystallization of salt, Cem. Concr. Res. 34 (2004)
þ Z i Z j = 8I2 xij J 0 xij  xii J 0 ðxii Þ=2  xjj J 0 xjj =2 ðA3:16Þ 1613–1624, https://doi.org/10.1016/j.cemconres.2003.12.034.
[8] M. Sabthabam, M.D. Cohen, J. Olek, Mechanism of sulfate attack: a fresh look:
Part 2. Proposed mechanisms, Cem. Concr. Res. 33 (2003) 341–346.
xij ¼ 6Z i Z j Au I1=2 : ðA3:17Þ [9] I.S. Evans, Salt crystallization and rock weathering: a review, Rev. Geomorphol.
Dyn. 19 (1970) 153–177.
[10] R.N. Carlos, D. ERIC, S. Eduardo, How does sodium sulfate crystallize?
The expression of J(x) is defined as:
Implications for the decay and testing of building materials, Cem. Concr. Res.
1 30 (2000) 1527–1534, https://doi.org/10.1016/S0008-8846(00)00381-1.
J ðxÞ ¼ x 4 þ C 1 xC 2 exp C 3 xC4 ðA3:18Þ [11] R.J. Flatt, C. Francesco, M.A.S. Asel, W.S. George, Chemomechanics of salt
damage in stone, Nat. Commun. 5 (2014) 1–5, https://doi.org/10.1038/
1 ncomms5823.
J 0 ðxÞ ¼ 4 þ C 1 xC 2 exp C 3 xC 4 [12] J.K. Chen, M.Q. Jiang, Long-term evolution of delayed ettringite and Gypsum in
2 Portland cement mortars under sulfate erosion, Constr. Build. Mater. 23 (2009)
þ 4 þ C 1 xC 2 exp C 3 xC 4 C 1 xexp C 3 xC 4 C 2 xC 2 1 þ C 3 C 4 xC 4 1 xC 2
812–816, https://doi.org/10.1016/j.conbuildmat.2008.03.002.
ðA3:19Þ [13] B. Bary, N. Leterrier, E. Deville, P.L. Bescop, Coupled chemo-transport-
mechanical modelling and numerical simulation of external sulfate attack in
where the values of C1, C2, C3 and C4 are 4.581, 0.7237, 0.0120 and mortar, Cem. Concr. Comp. 49 (2014) 70–83, https://doi.org/10.1016/j.
0.528, respectively. cemconcomp.2013.12.010.
J. Ren et al. / Construction and Building Materials 264 (2020) 120525 21

[14] W.C. Hansen, Attack on Portland cement concrete by alkali soil and water-a [43] C. Yu, W. Sun, K. Scrivener, Mechanism of expansion of mortars immersed in
critical review, Highway Res. Rec. 113 (1966) 1–32. sodium sulfate solution, Cem. Concr. Res. 43 (2013) 105–111, https://doi.org/
[15] W.C. Hansen, Crystal growth as a source of expansion in Portland cement 10.1016/j.cemconres.2012.10.001.
concrete, Proc. Am. Soc. Test. Mater. (1963) 932–945. [44] Q. Zhou, F.P. Glasser, Thermal stability and decomposition mechanisms of
[16] B. Mather, Discussion of ‘‘the process of sulfate attack on cement mortars”, ettringite at <120°C, Cem. Concr. Res. 31 (2001) 1333–1339, https://doi.org/
Adv. Cem. Based Mater. 5 (1996) 109–110, https://doi.org/10.1016/S1065- 10.1016/S0008-8846(01)00558-0.
7355(96)90057-7. [45] B. Albert, B. Guy, D. Damidot, Water chemical potential: a key parameter to
[17] P.K. Mehta, Sulfate attack on concrete-a critical review, Materials Science of determine the thermodynamic stability of some hydrated cement phase in
Concrete III, Am. Ceram. Soc, Westerville, OH, 1992. concrete?, Cem. Concr. Res. 36 (2006) 783–790, https://doi.org/10.1016/j.
[18] J. Nielsen, Investigation of resistance of cement paste to sulfate attack, cemconres.2005.12.016.
Highway Res. Rec. 113 (1966) 114–117. [46] L.G. Baquerizo, T. Matschei, K.L. Scrivener, Impact of water activity on the
[19] D. Bonen, M.D. Cohen, Magnesium sulfate attack on Portland cement paste-I. stability of ettringite, Cem. Concr. Res. 79 (2016) 31–44, https://doi.org/
Microstructural analysis, Cem. Concr. Res. 22 (1992) 169–180, https://doi.org/ 10.1016/j.cemconres.2015.07.008.
10.1016/0008-8846(92)90147-N. [47] R. EI-Hachem, E. Roziere, F. Grondin, A. Loukili, Multi-criteria analysis of the
[20] D. Bonen, M.D. Cohen, Magnesium sulfate attack on Portland cement paste-II. mechanism of degradation of portland cement based mortars exposed to
Chemical and mineralogical analysis, Cem. Concr. Res. 22 (1992) 707–718, external sulphate attack, Cem. Concr. Res. 42 (2012) 1327–1335, https://doi.
https://doi.org/10.1016/0008-8846(92)90023-O. org/10.1016/j.cemconres.2012.06.005.
[21] D. Bonen, S.L. Sarkar, Replacement of Portlandite by gypsum in the [48] J.M. Gao, Z.X. Yu, L.G. Song, T.X. Wang, S. Wei, Durability of concrete exposed to
interfacial zone and cracking related to crystallization pressure, in: sulfate attack under flexural loading and drying-wetting cycles, Constr. Build.
Ceramics Transtractions, 37, Cement-Based Materials: Present, Future, Mater. 39 (2013) 33–38, https://doi.org/10.1016/j.conbuildmat.2012.05.033.
and Environmental Aspects, Am. Ceram. Soc, Westerville, OH, 1993, pp. [49] J. Yuan, Y. Liu, Z.C. Tan, B.K. Zhang, Investigating the failure process of concrete
49–59. under the coupled actions between sulfate attack and drying-wetting cycles by
[22] S. Yang, Z. Zhongzi, T. Mingsu, The process of sulfate attack on cement mortars, using X-ray CT, Constr. Build. Mater. 108 (2016) 129–138, https://doi.org/
Adv. Cem. Based Mater. 4 (1996) 1–5, https://doi.org/10.1016/S1065-7355(96) 10.1016/j.conbuildmat.2016.01.040.
90057-7. [50] J. Gong, J. Gao, Y.F. Wang, Effects of sulfate attack and dry-wet circulation on
[23] J.G. Wang, Sulfate attack on hardened cement paste, Cem. Concr. Res. 24 creep of fly-ash slay concrete, Constr. Build. Mater. 125 (2016) 12–20, https://
(1994) 735–742, https://doi.org/10.1016/0008-8846(94)90199-6. doi.org/10.1016/j.conbuildmat.2016.08.023.
[24] X. Ping, J.J. Beaudoin, Mechanism of sulfate expansion I. Thermodynamic [51] Y.J. Chen, J.M. Gao, L.P. Tang, X.H. Li, Resistance of concrete against combined
principles of crystallization pressure, Cem. Concr. Res. 22 (1992) 631–640, attack of chloride and sulfate under drying-wetting cycles, Constr. Build.
https://doi.org/10.1016/0008-8846(92)90015-N. Mater. 106 (2016) 650–658, https://doi.org/10.1016/
[25] X. Ping, J.J. Beaudoin, Mechanism of sulfate expansion II. Validation of j.conbuildmat.2015.12.151.
thermodynamic theory, Cem. Concr. Res. 22 (1992) 845–854, https://doi.org/ [52] L. Jiang, D.T. Niu, Study of deterioration of concrete exposed to different types
10.1016/0008-8846(92)90109-9. of sulfate solutions under drying-wetting cycles, Constr. Build. Mater. 117
[26] B. Tian, M.D. Cohen, Does gypsum formation during sulfate attack on concrete (2016) 86–98, https://doi.org/10.1016/j.conbuildmat.2016.04.094.
lead to expansion?, Cem. Concr. Res. 30 (2000) 117–123, https://doi.org/ [53] J.M. Abualgasem, J.C. Cripps, C.J. Lynsdale, Effects of wetting and drying cycles
10.1016/S0008-8846(99)00211-2. on thaumasite formation in cement mortars, J. Mater. Civil Eng. 27 (2015),
[27] R.S. Gallop, H.F.W. Taylor, Microstructural and microanalytical studies of https://doi.org/10.1061/(ASCE)MT.1943-5533.0001083.
sulfate attack: I. Ordinary portland cement paste, Cem. Concr. Res. 22 (1992) [54] W. Eysel, Crystal chemistry of the system Na2SO4-K2SO4-K2CrO4-Na2CrO4 and
1027–1038, https://doi.org/10.1016/0008-8846(92)90033-R. the glaserite phase, Am. Mineral. 58 (1973) 736–747.
[28] R.S. Gallop, H.F.W. Taylor, Microstructural and microanalytical studies of [55] C. Rodriguez-Navarro, E. Doehnea, E. Sebastian, How dose sodium sulfate
sulfate attack: III. Sulfate-resisting portland cement: reactions with sodium crystallization? Implications for the decay and testing of building materials,
and magnesium sulfate solutions, Cem. Concr. Res. 25 (1995) 1581–1590, Cem. Concr. Res. 30 (2000) 1527–1534, https://doi.org/10.1016/S0008-8846
https://doi.org/10.1016/0008-8846(95)00151-2. (00)00381-1.
[29] F. Bellmann, B. Möser, J. Stark, Influence of sulfate solution concentration on [56] K. Linnow, A. Zeunert, M. Steiger, Investigation of sodium sulfate phase
the formation of gypsum in sulfate resistance test specimen, Cem. Concr. Res. transitions in a porous material using humidity and temperature controlled X-
36 (2006) 358–363, https://doi.org/10.1016/j.cemconres.2005.04.006. ray diffraction, Anal. Chem. 78 (2006) 4638–4689, https://doi.org/10.1021/
[30] N. Tsui, R.J. Flatt, G.W. Scherer, Crystallization damage by sodium sulfate, J. Cult. ac0603936.
Herit. 4 (2003) 109–115, https://doi.org/10.1016/S1296-2074(03)00022-0. [57] L.D.H. Oswald, A. Hamilton, C. Hall, W.G. Marshall, T.J. Prior, C.R. Pulham, In
[31] G.W. Scherer, Crystallization in pores, Cem. Concr. Res. 29 (1999) 1347–1358, situ characterization of elusive salt hydrates-the crystal structures of the
https://doi.org/10.1016/S0008-8846(99)00002-2. heptahydrate and octahydrate of sodium sulfate, J. Am. Chem. Soc. 130 (2008)
[32] G.W. Scherer, Stress from crystallization of salt in pores, in: V. Fassina (Ed.), 17795–17800, https://doi.org/10.1021/ja805429m.
Proceedings of the Ninth International Congress on Deterioration and [58] H. Derluyn, T.A. Saidov, R.M. Espinosa-Marzai, L. Pei, G.W. Scherer, Soldium
Conservation of Stone, Venice, 19–25 June, Elsevier, Amsterdam, 2000, pp. sulfate heptahydrate Ⅰ: the growth of single crystals, J. Cryst. Growth 329
187–194. (2011) 44–51, https://doi.org/10.1016/j.jcrysgro.2011.06.024.
[33] G.W. Scherer, Reply to the discussion by S. Chatterji of the paper, [59] H. Naruse, K. Tanaka, H. Moriikawa, F. Marumo, Structure of Na2SO4 (1) at
‘‘Crystallization in pores”, Cem. Concr. Res. 20 (2000) 673–675, https://doi. 693K, Acta Crystallogr. B 43 (1987) 143–146, https://doi.org/10.1107/
org/10.1016/S0008-8846(00)00232-5. S010876818709815X.
[34] C.W. Correns, Growth and dissolution of crystals under linear pressure, Disc. [60] R.J. Flatt, Salt damage in porous materials: how high supersaturations are
Far. Soc. 5 (1949) 267–271, https://doi.org/10.1039/DF9490500267. generated, J. Cryst. Growth 242 (2002) 435–454, https://doi.org/10.1016/
[35] L.A. Rijniers, H.P. Huinink, L. Pel, K. Kopinga, Experimental evidence of S0022-0248(02)01429-X.
crystallization pressure inside porous media, Phys. Rev. Lett. 94 (2005) 1–4, [61] P.W. Atkins, Physical Chemistry, Oxford University Press, Oxford, 1990.
https://doi.org/10.1103/PhysRevLett.94.075503. [62] W.F. Linke, Solubilities, Inorganic and Metal Organic Compounds: A
[36] L.A. Rijniers, L. Pel, H.P. Huinink, K. Kopinga, Salt crystallization as damage Complication of Solubility Data from the Periodical Literature, Van
mechanism in porous building materials-a nuclear magnetic resonance study, Nostrand-Reinhold, Princeton, 1958.
Magn. Reson. Imag. 23 (2005) 273–276, https://doi.org/10.1016/j. [63] A. Arnold, Behavior of some soluble salts in stone monuments, in: Skoulikdis
mri.2004.11.023. (Ed.), 2nd International Symposium on the Deterioration of building Stones,
[37] O. Coussy, Deformation and stress from in-pore drying-induced crystallization National Technical University, Athens, 1976, pp. 27–36.
of salt, J. Mech. Phys. Sol. 54 (2006) 1517–1547, https://doi.org/10.1016/j. [64] Handbuch Gmelin, der Anorganischen Chemie, Verlag Chemie, Weinheim,
jmps.2006.03.002. 1927.
[38] M. Steiger, S. Asmussen, Crystallization of sodium sulfate phases in porous [65] M.R. Hartman, R. Berliner, Investigation of the structure of ettringite by time-
materials: the phase diagram Na2SO4-H2O and the generation of stress, of-flight neutron powder diffraction techniques, Cem. Concr. Res. 9 (2004)
Geochim. Cosmochim. Ac. 72 (2008) 4291–4306, https://doi.org/10.1016/j. 364–370, https://doi.org/10.1016/j.cemconres.2005.08.004.
gca.2008.05.053. [66] V.S. Ramachandran, J.J. Beaudoin, Significance of water/solid ratio and
[39] S. Chatterji, N. Thaulow, Unambiguous demonstration of destructive crystal temperature on the physico-mechanical characteristics of hydrating 4CaO.
growth pressure, Cem. Concr. Res. 27 (1997) 811–816, https://doi.org/ Al2O3.Fe2O3, J. Mater. Sci. 11 (1976) 1893–1910, https://doi.org/10.1007/
10.1016/S0008-8846(97)00078-1. BF00708268.
[40] G.W. Scherer, Factors affecting crystallization pressure, in: K. Scrivener, J. [67] Y. Shimada, J.F. Young, Thermal stability of ettringite in alkaline solutions at 80
Skalny (Eds.), Proceedings of international RILEM 186-isa workshop and °C, Cem. Concr. Res. 34 (2004) 2261–2268, https://doi.org/10.1016/j.
internal sulfate attack and delayed ettringite formation, 35, RILEM cemconres.2004.04.008.
publications, Paris, 2004, pp. 139–154. [68] P.K. Mehta, Stability of ettringite on heating, J. Am. Ceram. Soc. 55 (1972) 55–
[41] M. Steiger, Crystal growth in porous materials-II: influence of crystal size on 56, https://doi.org/10.1111/j.1151-2916.1972.tb13403.x.
the crystallization pressure, J. Cryst. Growth 282 (2005) 455–469, https://doi. [69] K. Serafeimidis, G. Anagnostou, The solubilities and thermodynamic
org/10.1016/j.jcrysgro.2005.05.008. equilibrium of anhydrite and gypsum, Rock Mach. Rock Eng. 48 (2015) 15–
[42] R.J. Flatt, G.W. Scherer, Thermodynamics of crystallization stress in DEF, Cem. 31, https://doi.org/10.1007/s00603-014-0557-1.
Concr. Res. 38 (2008) 325–336, https://doi.org/10.1016/j.cemconres.2007. [70] I. Biczok, Concrete Corrosion Concrete Protection, Chemical Publishing, New
10.002. York, 1967.
22 J. Ren et al. / Construction and Building Materials 264 (2020) 120525

[71] T. Schmidt, B. Lothenbach, M. Romer, J. Neuenschwander, K. Scrivener, Physical high ionic strength salt at 25℃, Geochim. Cosmochim. Ac. 48 (1984) 723–751,
and microstructural aspects of sulfate attack on ordinary and limestone https://doi.org/10.1016/0016-7037(84)90098-X.
blended portland cements, Cem. Concr. Res. 39 (2009) 1111–1121, https://doi. [84] Y.M. Zhang, C. Pichler, Y. Yuan, M. Zeiml, R. Lackner, Micromechanics-based
org/10.1016/j.cemconres.2009.08.005. multifield framework for early-age concrete, Eng. Struct. 47 (2003) 16–24,
[72] J. Marchand, E. Samson, Y. Maltais, J. Beaudoin, Theoretical analysis of the https://doi.org/10.1016/j.engstruct.2012.08.015.
effect of weak sodium sulfate solutions on the durability of concrete, Cem. [85] F.J. Ulm, G. Constantinides, F.H. Heukamp, Is concrete a poromechanics
Concr. Comp. 24 (2002) 317–329, https://doi.org/10.1016/S0958-9465(01) material?-a multiscale investigation of poroelastic properties, Mater. Struct.
00083-X. 37 (2004) 43–58, https://doi.org/10.1007/BF02481626.
[73] K. Serafeimidis, G. Anagnostou, The solubilities and thermodynamic [86] R.M. Espinosa-Marzal, A. Hamilton, M. Mcnall, K. Whiteaker, G.W. Scherer, The
equilibrium of anhydrite and gypsum, Rock Mech. Rock Eng. 48 (48) (2015) chemomechanics of crystallization during rewetting of limestone impregnated
15–31, https://doi.org/10.1007/s00603-014-0557-1. with sodium sulfate, J. Mater. Res. 26 (2011) 1472–1481, https://doi.org/
[74] L.B. Miller, J.C. Witt, Solubility of calcium hydroxide, J. Phys. Chem. 33 (1929) 10.1557/jmr.2011.137.
285–289, https://doi.org/10.1021/j150296a010. [87] N. Venkovic, L. Sorelli, B. Sudret, T. Yalamas, R. Gagné, Uncertainty propagation
[75] K. Johannsen, S. Rademacher, Modelling the kinetics of calcium hydroxide of a multiscale poromechanics-hydration model for poroelastic properties of
dissolution in water, Act. Hydr. Hydr. 27 (1999) 72–78, https://doi.org/ cement paste at early-age, Prob. Eng. Mech. 32 (2013) 5–20, https://doi.org/
10.1002/(SICI)1521-401X. 10.1016/j.probengmech.2012.12.003.
[76] S.H. Northrup, J.T. Hynes, Short range caging effects for reactions in solution. Ⅰ. [88] W.N. Meng, K.H. Khayat, Effect of graphite nanoplatelets and carbon
Reaction rate constants and short range caging picture, J. Chem. Phys. 71 nanofibers on rheology, hydration, shrinkage, mechanical properties, and
(1979) 871–883, https://doi.org/10.1063/1.438378. microstructure of UHPC, Cem. Concr. Res. 105 (2018) 64–71, https://doi.org/
[77] A.J. Melean, M.J. Muldoon, C.M. Gordon, L.R. Dunkin, Bimolecular rate 10.1016/j.cemconres.2018.01.001.
constants for diffusion in ionic liquids, Chem. Comm. 17 (2002) 1880–1881, [89] K.S. Pitzer, Thermodynamics of electrolytes. Ⅰ. Theoretical basis and general
https://doi.org/10.1039/b202944h. equations, J. Phys. Chem. 77 (1973) 268–277, https://doi.org/10.1021/
[78] O.G. Berg, P.H. Hippel, Diffusion-controlled macromolecular interactions, j100621a026.
Annu. Rev. Biophys. Chem. 14 (1985) 131–160, https://doi.org/10.1146/ [90] C.E. Harvie, J.H. Were, The prediction of mineral solubilities in natural waters:
annurev.bb.14.060185.001023. the Na-K-Mg-Ca-Cl-SO4-H2O system from zero to high concentration at 25℃,
[79] S.D. Traytak, A.V. Barzykin, M. Tachiya, Competition effects in diffusion- Geochim. Cosmochim. Ac. 44 (1980) 981–997, https://doi.org/10.1016/0016-
controlled bulk reactions between ions, J. Chem. Phys. 126 (2007) 1–7, https:// 7037(80)90287-2.
doi.org/10.1063/1.2717181. [91] Y. Yao, W.H. Zhong, Effect of polypropylene fibers on the longterm tensile
[80] D. Shoup, G. Lipari, A. Szabo, Diffusion-controlled bimolecular reaction rates. strength of concrete, J. Wuhan Univ. Technol. 22 (2007) 52–55, https://doi.org/
The effect of rotational diffusion and orientation constants, Biophys. J. 36 10.1007/s11595-005-1052-z.
(1981) 697–714, https://doi.org/10.1016/S0006-3495(81)84759-5. [92] S.B. Zhao, X.X. Ding, M.S. Zhao, C.Y. Li, S.W. Pei, Experimental study on tensile
[81] R.J. Flatt, M. Steiger, G.W. Scherer, A commented translation of the paper by C. strength development of concrete with manufactured sand, Constr. Build.
W. Correns and W. Steinborn on crystallization pressure, Environ. Geol. 52 Mater. 138 (2017) 247–253, https://doi.org/10.1016/j.conbuildmat.2017.
(2007) 187–203, https://doi.org/10.1007/s00254-006-0509-5. 01.093.
[82] M. Steiger, Crystal growth in porous materials-Ⅰ: the crystallization pressure of [93] S. Bhanja, B. Sengypta, Influence of silica fume on the tensile strength of
large crystals, J. Cryst. Growth 282 (2005) 455–469, https://doi.org/10.1016/j. concrete, Cem. Concr. Res. 35 (2005) 743–747, https://doi.org/10.1016/j.
jcrysgro.2005.05.007. cemconres.2004.05.024.
[83] C.E. Harvie, N. Moller, J.H. Weare, The prediction of mineral solubilities in
natural water: The Na-K-Mg-Ca-H-Cl-SO4-OH-HCO3-CO3-CO2-H2O system to

You might also like