Vous êtes sur la page 1sur 35

Advances in Cement Research

Volume 27 Issue 2
Chloride in cement
Galan and Glasser

Advances in Cement Research, 2015, 27(2), 6397


http://dx.doi.org/10.1680/adcr.13.00067
Paper 1300067
Received 27/09/2013; revised 03/02/2014; accepted 29/04/2014
Published online ahead of print 01/10/2014
ICE Publishing: All rights reserved

Chloride in cement
Isabel Galan

Fredrik P. Glasser

Research Fellow, College of Physical Sciences, University of Aberdeen,


Aberdeen, UK

Professor, College of Physical Sciences, University of Aberdeen, Aberdeen,


UK

Portland cementsteel composites frequently take up chloride from their service environment. The degradation
processes and test methods are described through a critical review of the literature. Plain cement paste is not much
affected by chloride except for increased solubilisation of cement solids, but chloride is detrimental to the passivation
of embedded steel. Emphasis is placed on establishing the underlying physicochemical concepts and mechanisms of
corrosion, and integrating these into a holistic picture. Simplistic calculations and test conditions probably do not
reproduce well the complex reactions with cement components occurring at the steelcement interface and the
concept of a threshold defined in terms of a critical chloride content or chloride:hydroxide ratio is just an
approximation at best: other cement components are involved, as is the oxygen activity at the interface. Chloride
binding can retard the migration of chloride in concrete and specimen calculations are presented showing that the
main binding mechanism arises from the AFm, [Ca2(Al,Fe)(OH)6].X.nH2O, content of the paste. Suggestions are made
to enhance the value of experimental approaches to corrosion.

Notation
C
D
D0
E
E0
Eh
e
F
J
k
Kw
R
r
T
t
V
v
x
z
G0ads

0


0
()
(0)

concentration
diffusion coefficient
pore solution diffusivity
electrical potential
standard redox potential
redox potential
electron charge
Faraday constant
flux
Boltzmanns constant
ion product of water
gas constant
interaction parameter
temperature
time
voltage drop across the membrane
ion velocity
length
electrical charge
free energy of adsorption of chloride
surface coverage, ratio between surface concentration
and maximum surface concentration of adsorbed species
standard chemical potential
defined in Equation 15
sample conductivity
pore solution conductivity
voltage-enhanced time-lag
passive time-lag

Introduction
The role of chloride in cement has had a mixed history. On the
one hand, chloride has long been used to accelerate cement

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

hydration and strength gain. On the other, chloride deprives


embedded steel of the corrosion protection normally developed in
chloride-free cement environments.
Since other means are available to accelerate cement hydration,
chloride has almost universally been banned as a deliberate
additive to cement, and manufacturing processes are conducted to
reduce chloride in the final clinker. This is done by raw materials
selection and kiln design: at high temperatures, alkali chlorides
are volatile and, with good kiln ventilation, chloride can be
distilled off. These developments have left researchers largely
free to concentrate on processes by which chloride enters
hardened concrete in service conditions and of the ensuing
consequences.
Overview of the problem
Although steel is never thermodynamically stable in contact with
hydrated cement paste, the reaction products of corrosion (mainly
solid) develop only slowly and afford good protection to
embedded steel with the result that reaction is slow. However, in
service conditions, chloride may diffuse through cover concrete
and impair the passivation of embedded steel: a regime of
passivity is replaced by one of corrosion. Moreover, steel
corrosion is not the only hazard chloride also affects cement
durability by solubilising calcium, the quality of cover concrete
being impaired. At high concentrations, chloride may also
destabilise portlandite with the formation of calcium chlorohydroxy hydrates, which are damaging. In passing, it is worth
noting that hardened concrete is a mixture of mineral aggregates,
hydrated Portland cement paste and pore water. This paper
generally assumes that the mineral aggregate is inert so that
chemical reactions are confined to cement and its associated pore
water. For these purposes, the aggregate can be neglected.
63

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

However, its physical presence increases the tortuosity of diffusion paths and forms interfacial transition zones, and so on. So,
in the real world, the diffusion of chloride and other chargebalancing species is affected by the presence of aggregate and the
role of concrete must often be considered, not just cement.

speciation is as chloride (Cl). Concentrations range widely, but


are around 0 .5 M in seawater of normal salinity (35 g/l). Table
1 (Eglinton, 2003) gives some representative values. Water in
partially isolated seas, such as the Baltic, is often more dilute
than in the open ocean whereas, in the Arabian Gulf, where
precipitation and runoff are low, evaporation predominates and
salinity can be up to 30% higher than in normal seawater.

The diffusion of chloride in cement is not straightforward for


several reasons.
j

j
j
j

Wind-generated salt spray frequently transports chloride and the


effect is enhanced by evaporation, giving rise to fine particulate
solid, the main component of which is sodium chloride. Dusts
containing sodium chloride may therefore be windborne for
considerable distances.

The cement matrix is somewhat permeable, consisting of a


complex and tortuous network of pores of different sizes and
degrees of interconnectivity.
The moisture state of the water in pores is variable. At its
simplest, diffusion occurs in a saturated medium but many
real-life service regimes are characterised by undersaturation
and diffusion occurs through a network of pores that are only
partially water filled.
The pore size and connectivity are not constant but change
with time as concrete matures.
Chloride may interact strongly with the cement solids by
sorption and binding.
Chloride cannot diffuse on its own: in order to preserve local
electroneutrality in the matrix and associated pore fluid, the
negative charge on chloride has to be balanced locally
within a few atom lengths by positive ions.
In general, the combination and interplay between physical
and chemical factors has often inhibited a fuller
understanding of the problem.

In colder climates, sodium chloride is used as a de-icing salt


where its application affects bridge decks, road surfaces, and so
forth. Many industrial processes use sodium chloride, so it
becomes widely distributed in effluents, in industrial and municipal wastes and so on. Saline environments may also develop as a
result of wicking action, leading to local high concentration:
dilute water, initially low in soluble chloride, migrates into
concrete, typically below grade, and wicks into a zone where
evaporation predominates, allowing local enhancement of the
chloride concentration.

Not surprisingly, it has proven difficult for even experienced


researchers to assess and integrate data from different sources
with a view to giving guidance and developing tests.
This review uses a somewhat unorthodox presentation of the data
but it is hoped that its style will lend clarity to problem definition
and resolution. Most of the data cited have appeared in the
literature but the opportunity has also been taken to introduce
new calculations and as yet unpublished data. These are identified. The approach has also revealed areas where data are lacking.
These are also indicated. Where critical conclusions are reached,
and unless supported by literature reference, they are the authors
opinions.
Chloride in service environments
Chlorine is an abundant geochemical component of the earths
crust and concentrates in ocean and ground water. Its main

Chloride concentration: g/l

The internal environment in Portland cement pastes


As noted, the components of cement paste are slightly soluble
and the nature of soluble species induces a high internal pH.
Strictly, concrete does not have a pH the pH of the associated
pore water is measured instead. However, it is normal practice to
conflate the two. More water is frequently added to concrete
mixes than is necessary to satisfy the chemical demand for water,
with the result that hardened concrete normally contains significant water-filled porosity. Figure 1 (Lothenbach et al., 2008)
shows how the pore water content of the paste changes with time.
To describe hydration, consider first a chemically simplified
cement, containing portlandite, Ca(OH)2, calcium silicate hydrate

Mediterranean

Atlantic

Mean seawater
(35 g/l salinity)

21 .4

17 .8

19 .8

Table 1. Mediterranean, Atlantic and mean seawater chloride


concentration (Eglinton, 2003)

64

A variety of mechanisms thus exist whereby chloride can enter


cement, some of which lead to concentration. We also have to
consider briefly the presence of other water-soluble species that
may be associated and transported with chloride, so that joint
attack by species such as sodium, magnesium and sulfate can
occur with chloride. Carbonation also affects the course and
consequences of reaction with chloride.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

80
Pore solution

Volume: cm3/100 g cement

70
60
50
40
30
20
10

Monosulfate
Hemicarbonate
Hydrotalcite

Monocarbonate Ettringite
Gypsum Brucite
Portlandite
Ferrite
Belite

Aluminate
Calcium silicate hydrate

Alite

0
001

01

1
10
Hydration time: d

100

1000

Figure 1. Hydration evolution of ordinary Portland cement (OPC)


paste without limestone (Lothenbach et al., 2008). The
distribution of hydrates was obtained by coupling thermodynamic modelling, kinetic equations and experimental data

(C-S-H), and the families of phases called AFm and AFt, with
general formula [Ca2(Al,Fe)(OH)6].X.xH2O and [Ca3(Al,Fe)
(OH)6.12H2O]2.X3.xH2O, respectively, phases coexisting with an
aqueous solution. The solubility of the phases is dominated by
portlandite and the resulting pH is 12 .5 at 208C. During the
first few days of cement hydration, the sulfate solubility is
dominated by calcium sulfate (gypsum, anhydrite) but Portland
cement is typically formulated to be undersaturated with respect
to calcium sulfate with the result that free calcium sulfate is
rapidly consumed, leaving AFm and AFt (both of which contain
sulfate) to buffer the aqueous sulfate concentration, typically at
millimolar concentrations. The buffering action of the solid
aqueous trio was explained by Matschei et al. (2007a), as
discussed later.

The pH of Portland cement is partially buffered because, upon


adding an acid or an acidic substance, hydroxide thus neutralised by the acid is simply replaced by dissolving more
portlandite, so the pH is not much decreased. In the presence of
stronger bases than portlandite, the pH can be elevated above
the portlandite threshold for example by adding sodium hydroxide (NaOH) (a strong base), but the rise in pH is less than
cumulative because the first additions of sodium hydroxide
reduce the solubility of portlandite with the result that total
hydroxide concentration does not much change: as hydroxide
concentration from sodium hydroxide increases, that from
portlandite decreases, so that the total hydroxide concentration
remains broadly constant until the sodium hydroxide concentration exceeds about 102 M.

Thus, hydroxide is the dominant soluble anion of fresh paste and


its concentration defines pH through the ion product of water Kw

Like all buffer systems


the pH buffers in cement have a limited capacity to buffer:
once this capacity is overcome, the system moves into an
unbuffered regime and
j buffering does not lead to a regime of exact pH, only to
one in which the system is resistant to change: in this
instance, one in which pH changes little upon adding acids
or bases.
j

1a:

H2 O H [OH ] K w  101014 (258C, 100 kPa)

1b:

pH log[OH ]

where [ ] represents concentration. The units are, strictly, molal


but at low concentrations, the molal (mass solute/mass solvent)
scale becomes equal within limits of error to the more widely
used molar (mass solute/volume solvent) scale.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

The buffering concept involving pH is reasonably well known


but what is arguably less well known is that buffering mechanisms also influence the concentration of other species in cement
pore fluid. Taking sulfate as an example, adding or subtracting
calcium sulfate from the bulk composition changes the ratio of
65

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

AFm to AFt in the hydration products but, since pH does not


depend on the amount of substance, only on its presence or
absence, and since the AFm and AFt equilibrate with each other
by way of the aqueous phase, the system of solids and aqueous
phase has the characteristics of a buffer system so, at fixed
temperature, the sulfate concentration of the aqueous phase is
nearly constant.

contact with air so, normally, Eh values do not much change from
the range +100 mV to +300 mV. Exceptions do occur, as for
example when tin salts containing tin(II) are deliberately added to
hydrating cement to lower Eh and thereby chemically reduce
chromium(VI) to the less hazardous chromium(III) state. However, when steel is added to cement it liberates hydrogen, leading
to a decrease in Eh: any dissolved oxygen is scavenged and, once
consumed, the Eh becomes strongly negative, being dominated by
liberation of hydrogen (gas).

Of course, the application of the concept to commercial products


is modified by the presence of other ions. For example, if sodium
is present, effectively as sodium hydroxide, it enhances the
aqueous sulfate concentration, which in turn affects the solubility
of other components and hence the exact numerical value of pH.
But as long as the portlandite buffer is operative for pH and the
AFm/AFt pair is operative for sulfate, the system is buffered and
changes in both pH and sulfate concentration remain small,
requiring relatively large changes in sodium concentration to
change aqueous sulfate concentrations. The same buffering
principles operate to control aqueous concentrations of anions
such as chloride, as will be discussed.
The buffering concept enables regimes to be defined in which
many chemical variables remain fixed, or nearly fixed, giving rise
to a relative constancy of aqueous phase concentrations. Application of the concept of buffering regimes within cement is a great
simplification to achieving a broad-brush approach to its internal
chemistry. A limited number of regimes exist, not as might be
supposed an infinite number. The numerical values for buffering regimes are temperature dependent but, even at constant
temperature, numerical values are slightly fuzzy. For example,
some fine tuning to correct for species activities and include the
impact of minor elements may be required for precise calculations. However, in a broad-brush approach, this does not invalidate the general principle that the regime concept operates and
that only a few regimes are sufficient to define the internal
chemistry.

The Eh is measured on a relative scale defined by a convenient


standard, in this instance the standard voltage for the hydrogen
electrode, taking its numerical value as 0 .00 V at 258C, 1 m
hydrogen ion concentration and 1 bar (100 kPa) pressure of
hydrogen gas.
Analogous to pH, Eh defines a scale of potentials, in this case for
oxidation-reduction reactions. Just as pH buffering capacity can
be measured and its changes used to characterise buffered
systems, so Eh controls can be achieved: this capacity to maintain
a nearly constant Eh is termed poising (Nightingale, 1958). The
poising capacity of Portland cement is weak but can be much
enhanced by adding, for example, iron blast-furnace slag, which
enables redox couples involving sulfur to develop. These couples
tend to generate a much more reducing environment than is found
in plain Portland pastes.
It is usual in redox-related reactions, to find the two half-reactions
occurring in solution such that the overall reaction conserves
mass and charge. However, in corrosion reactions, the two halfreactions of the anode (oxidation) and cathode (reduction) are
often physically separated and the additional mechanisms thus
available for ion and electron transport become an important part
of corrosion science. Thus, the system characteristics with respect
to corrosion are defined by
pH and pH buffers
Eh and poising capacity
j system chemistry
j temperature
j transport properties of the medium and, if anode and cathode
are separated, mechanisms for mass and electron transport.
j

Therefore the concept of regimes is applied here, each of which


has definite boundaries, to increase the generic applicability of
the data. Following geochemical principles, the transition between
regimes will also be associated with crossing a fence; that is,
the boundary between regimes. The regime concept also fits well
with the teachings of corrosion science where the concept is
widely used.
Another characteristic parameter of aqueous solutions is the Eh
function, Eh being shorthand for an oxidation-reduction scale.
Chloride in cement is not normally subject to changes in oxidation
state. But the intrinsic corrosion state of iron and steel in cement
is sensitive to the numerical value of redox potential: the nature,
mechanism and rate of corrosion being dependent on the Eh of
the matrix. So, for that reason, Eh becomes an important parameter with which to characterise cement pastes. Only exceptionally
we do need Eh values to explore cement science because Portland
cements are made under oxidising conditions and are used in
66

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

A widely used representation to depicting corrosion processes


was developed by Pourbaix (1966) and takes the form of the socalled Pourbaix diagram. This form of representation depicts
chemistry as a function of environment, as defined by pH, Eh and
concentration. The chemistry depicted in a Pourbaix diagram may
have to be kept simple to retain a two-dimensional (2D)
representation. To constrain the number of variables, diagrams
are usually constructed for a fixed pressure and temperature, often
(but not exclusively) 1 bar (100 kPa) and 258C. The threshold
concentration above which the species is said to be soluble must
also be defined. Examples of the diagrams and their application
will be developed in the course of this review.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Diffusion processes in hydrated cement

Einstein approach allows a solution to be obtained for the


diffusion coefficient if several assumptions are made, such as
considering convection and diffusion negligible compared to
migration when fields higher than 10 V are applied, assuming
steady-state conditions and neglecting heating/cooling due to
Joule effects (Andrade, 1993). A formation factor relates
diffusivity to electrical resistivity in a porous material and in the
pore solution: its use is, however, associated with practical
difficulties related to the measurement of the conductivity of both
the concrete and its associated pore solution.

Before reaching the steel, chloride has to penetrate from the


service environment through the cover concrete to the reinforcement, normally by migrating through the cement pasteaggregate
mix. The resistance to chloride diffusion through cover concrete
is therefore an important quantity. This resistance is a complex
function of the properties of the cover, including its thickness and
matrix quality. The transport of chloride through uncracked
matrices is also retarded by chemical binding of chloride into
minerals of the cement matrix. Matrix quality is affected by
formulation, including cement content, temperature and curing,
as well as by the incidence of physical cracking. This mixture of
properties, some intrinsic and some extrinsic, also depends on the
geometry of the concrete mass and upon workmanship, and has
proven difficult to quantify and control.
Practical measurements of chloride diffusion are usually interpreted by application of laws of diffusion, such as Ficks laws, the
experimental data being used to derive a diffusion coefficient for
chloride. However, Ficks diffusion laws can strictly only apply to
neutral species that do not interact with the matrix. Furthermore,
we know that chloride is not only charged but also interacts with
the matrix, as for example by forming Friedels salt (itself
basically an AFm type phase). Moreover, diffusion of a charged
species such as chloride is subject to the additional restrictions
noted earlier, including the need to maintain a local electrostatic
balance at every point along the diffusion profile. And, of course,
matrix diffusion is often coupled to physical diffusion through
cracks, the impacts of which may need to be included in realworld applications. Thus, models varying in sophistication have
long been available for treating diffusion. Table 2 summarises the
most common approaches to chloride diffusion.
Data have been modelled most successfully using a Nernst
PlanckEinstein diffusion approach. This approach has the advantage of enabling charge balance conditions to be included in
calculations but has the disadvantage of requiring additional
identification of rate-limiting steps and including more parameters that require numerical evaluation. The NernstPlanck
@C
@x

Ficks first law

J D

Ficks second law

@C @J
@2C

D 2
@t
@x
@x

NernstPlanck equation

@C
@C zF
@E
D

DC
 Cv
@t
@x RT
@x

NernstEinstein equation

RT 
z2 F 2 C

Formation factor

FF


D

 0 D0

Table 2. Theoretical approaches to diffusion, showing the general


equations

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Most investigators would now agree that values of chloride


migration derived from experimental measurement of local
chloride concentrations along a diffusion profile should be termed
apparent diffusion coefficients because numerical values depend
on the specific conditions under which they were determined. To
make the derivation and use of such values more apparent, the
authors prefer the term conditional diffusion coefficient: the
prefix conditional reminds us that the numerical diffusion value
is, strictly, applicable only to the defined conditions used to
derive the experimental value.
These limitations in applicability reduce the generic value of
measurements as it is not generally known how changing an input
variable would affect the numerical values of the diffusion coefficient, for example how cure duration and degree of maturity of the
paste will affect the value. It is however known that in goodquality concrete, steady-state conditions for chloride diffusion
measurements may require several weeks, even months, to establish. Thereafter, chloride penetration to depths comparable with
the thickness of cover concrete in low-permeability matrices may
require years or decades. Very few measurements of natural (i.e.
non-accelerated) diffusion have been reported over long time
scales. Of course real structures can be sampled but, in general,
their exposure history will not have been well documented. These
considerations have led to the development of accelerated tests.
Investigators have sometimes attempted to obviate these difficulties by
using low-quality matrices, perhaps made to high water/solid
(w/s) ratios
j adding chloride to the concrete in the course of mixing, using
a sodium chloride solution instead of pure water (Alonso et
al., 2000; Andrade and Page, 1986; Arya and Xu, 1995;
Manera et al., 2008; Morris et al., 2004; Yonezawa et al.,
1988).
j

The first method is not satisfactory because, in general, we are not


interested in promoting low-quality matrices to protect embedded
steel. Adding chloride from the outset, as in the second method,
creates other problems. For example, although chloride is homogeneously distributed in the mix water, it may not remain in a
homogeneous distribution as the matrix matures. This is because
chloride concentrates selectively in certain constituent mineral
67

Chloride in cement
Galan and Glasser

Reactions between Portland cement paste


and soluble chloride
Solubilisation
The action of sodium chloride on portlandite solubility is
depicted in Figure 2 (Glasser et al., 1999b). Portlandite, an
important constituent of hydrated cement paste, is stable in
contact with sodium chloride over the entire range of conditions
shown. Note that the solubility of portlandite in water and sodium
chloride solutions decreases with rising temperature but, at a
fixed temperature, its solubility increases with rising sodium
chloride concentration. The increase is greatest at low salt
concentrations up to 0 .5 M, remaining nearly constant within
limits of error, over the range 0 .51 .5 M. If the sodium chloride
data are equated to seawater (0 .5 M sodium chloride), substantial increases in calcium solubility relative to plain water can be
expected. However, the pH of saturated portlandite solutions,
relative to portlandite in water, remains essentially unaffected: at
258C, pH values are in excess of 12.
Reaction and binding
Chloride reacts with most of the phases that comprise cement
pastes. However, the strength and nature of the reaction is phase
specific and, moreover, concentration dependent (Table 3).

Increasing solubility

85

128

204

217

Calcium hydroxide stable

55

144

219

25

201

277

236

264

05
10
Sodium chloride: mmol/l

Increasing solubility

solids comprising the matrix and also concentrates in residual pore


water, the volume and distribution of which change as hydration
progresses. The relevant solid phases form over different time
scales, some early in the hydration history, others later, by which
time the pore fluid will have become more concentrated. As the
reaction progresses, chloride also concentrates selectively in the
vicinity of embedded steel, perhaps by selective complexation with
ferrous iron. For all these reasons it is not satisfactory to estimate
chloride diffusion in good-quality matrices using data obtained
from poor-quality matrices or to simulate the impact of chloride
on corrosion behaviour by adding chloride to the mix water. Other
techniques and methods used to measure chloride diffusion in
concrete are discussed later in the paper.

Temperature: C

Advances in Cement Research


Volume 27 Issue 2

266

15

Figure 2. Aqueous analysis showing portlandite solubility in


sodium chloride solutions at 25, 55 and 858C. The data are from
Glasser et al. (1999b), with numerical values in boxes representing
calcium concentration in mmol/l. Note that the initial increase in
solubility is greatest between 0 and 0 .5 M

It is generally agreed that the main contribution to chloride


binding results from the formation of chloride-containing AFm
phase, Friedels salt, and this conclusion is supported by the
calculations presented subsequently. The normal chemistry of
AFm includes sulfate, hydroxy and carbonate ions. But, in contact
with chloride and depending on species concentrations, other
anions in AFm may be replaced by chloride to form Friedels salt.
The structure of the AFm phase is composed of layers, ideally
with the composition Ca2[Al(OH)6]+. The positive layer charge is
balanced by inclusion of singly charged interlayer anions such as
hydroxide, chloride and/or by doubly charged ions such as
carbonate, sulfate, and so on: water molecules, held by van der
Waals forces, complete the interlayer contents. In the thermodynamic sense, and taking temperature as a variable, the least
stable of the AFm phases is the hydroxyl-containing AFm phase
(e.g. hydrocalumite type), which is only stable at ,88C at

Mechanism

Example

Comment

Sorption of chloride

Calcium silicate
hydrate/calcium aluminium
silicate hydrate
AFm ! Friedels salt

Does not affect amorphous nature of the solid. Desorption is probably a


complimentary process. Uptake capacity at ,1 M is low because of
competition from hydroxide concentration for binding sites
Chloride competes for incorporation into the structure in interlayer sites
with other charge-balancing ions (hydroxide, sulfate, carbonate)

Destabilisation of portlandite
and formation of
hydroxychloride phases
(e.g. xCaO.yCaCl2.zH2O)

3:1:15 formation in high-chloride zones: portlandite is consumed.

Ion exchange, perhaps


with minor structural
readjustment
Phase change

Table 3. Reaction between aqueous chloride and cement paste


minerals

68

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

100 kPa and pH  1213 (Matschei et al., 2007a). However,


sulfate, carbonate and chloride AFm phases are generally stable
in chemically complex cement environments in the range 0708C.

phase will project in the triangle ADE in Figure 3). On the


other hand, cements made without added carbonate are likely to
have AFm phase compositions projecting in the triangular region
ABD with the formation of two AFm phases.

The different AFm phases are only partially miscible with each
other. Figure 3 (Matschei et al., 2007b) shows the composition
and coexistence of AFm phases in typical cement environments
(i.e. under conditions such that the pH is buffered by the free
portlandite and chloride is absent). Portlandite is abundant in
Portland cement pastes where its pH buffering action depends on
the presence and accessibility of portlandite, but not on its
amount, so this conclusion about AFm stability has generic
relevance. From Figure 3 it can be anticipated that several AFm
types may coexist in a cement paste. This coexistence seems
complicated until it is realised that the activity of some species
essential to AFm formation are also buffered. For example, many
modern cements contain added calcite. This calcite has a low and
nearly fixed solubility conditioned by the presence of a
portlanditecalcium carbonatecarbon dioxide buffer, with the
result that pH and carbonate activity vary only over narrow limits
and do not depend on how much calcite is added in excess of
saturation, saturation being achieved at low calcite contents
(24% calcium carbonate). So, as a first approximation, all
carbonate-filled Portland cements will either develop hemicarboaluminate or, more likely, monocarboaluminate AFm coexisting
with monosulfate AFm (i.e. the bulk composition of the AFm

Thus, the action of chloride on AFm is not just a matter of


considering the action of chloride on a single AFm phase:
instead, its reaction with one of several possible phase assemblages has to be taken into account. These factors account, in
part, for the range of chloride concentration values reported in
the literature for Friedels salt formation.
Against this background, the impact of chloride can be assessed
as follows. Phase changes will occur when sufficient chloride is
added to attain an aqueous composition that will stabilise
Friedels salt. Thus, a fence (or technically, a phase boundary) has
to be crossed before chloride AFm, Friedels salt, becomes stable.
Because of the complexity of the AFm system Friedels salt can
accommodate a range of hydroxide for chloride substitution
and also because the position of the relevant fence is temperature
dependent, the transition does not occur at a fixed chloride
concentration although, in practice, differences conditioning the
transformation from different AFm precursors to Friedels salt are
small.
Thus, at constant pH and temperature, a range of threshold

A
Monosulfoaluminate (Ms)

Ms-type ss

Limiting Ms-type ss
(C4As05Hx)
Ms-type ss
Hc

Ms Hc

Monosulfoaluminate
monocarboaluminate

Miscibility gap (limiting


Ms-type ss C4AHx)

Ms Hc Mc

Limiting Ms-type ss
C4AHx Hc

C
C4AHx

Hc Mc
D
E
Hemicarboaluminate (Hc)
Monocarboaluminate (Mc)

Hc C4AHx

Figure 3. Phase diagram, taken from (Matschei et al., 2007b)


showing the composition of AFm phases coexisting at 208C in
chloride-free Portland cement. Three end-member compositions
exist, having hydroxide, sulfate and carbonate as principal anions.
Along the join between hydroxide and sulfate (left-hand edge),
sulfate AFm exhibits extensive solid solutions in which sulfate is

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

replaced by hydroxide. These solid solutions extend to point B.


Beyond this point, at higher hydroxide concentrations, two AFm
phases coexist. ss, solid solution; Ms,monosulfoaluminate; Hc
hemicarboaluminate; Mc, monocarboaluminate; C4AHx,
tetracalcium aluminate hydrate

69

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

chloride concentrations exists, below which Friedels salt is not


stable. When Friedels salt forms in chloride-rich environments,
its composition is close to theoretical: 3CaO.Al2O3.CaCl2.10H2O.
Typically, the threshold for Friedels salt formation occurs at a
few tens of millimolar soluble chloride. Assuming a pH of 12,
this threshold, expressed as a ratio of chloride concentration to
hydroxide concentration, is about 1 for hydroxide AFm but
higher, perhaps in the range 510, for conversion of the more
stable sulfate and carbonate AFm phases to Friedels salt. Note
that while the threshold concept is valid, its numerical value is
fuzzy, partly because of the compositional variability of AFm
precursor phase(s).

Jones et al., 2003; Suryavanshi et al., 1996). A simultaneous


adsorption mechanism has been proposed (Suryavanshi et al.,
1996) but this explanation, in the authors view, is essentially the
same as the exchange process described in the current paper.
Surface sorption is also possible, but the quantitative extent of
surface sorption (2D) is small relative to the 3D interlayer
capacity.

Table 4 shows some values for the amount of chloride gained or


lost per gram of solid at the equilibrium change to/from Friedels
salt. Relative to the mass of chloride sorbed at low concentrations, the quantities of chloride required to form Friedels salt are
large. Partly, this is because the uptake is three-dimensional (3D),
not two, as occurs in surface sorption: the entire bulk of the
crystal is available for uptake of structural chloride.
These arguments, and quantification of the relevant thermodynamics, result in the general statement that chloride AFm is
stable in cement pore fluids at minimum chloride concentrations
estimated by Birnin-Yauri and Glasser (1998) to be greater than
14 mM, and that its formation is responsible for most of the
observed chloride binding.
The formation of Kuzels salt (3CaO.Al2O3.1/2CaCl2.1/2Ca
SO4.10H2O) is estimated to occur above 10 mM chloride at
high sulfate activities (Glasser et al., 1999a). Because sulfate
buffer systems also operate (see previous text) it is difficult to
achieve the high sulfate concentrations necessary to stabilise
Kuzels salt in a normal Portland cement paste. As a result,
Kuzels salt is not normally found except in cements undergoing
external sulfate attack. Solid solutions between hydroxide AFm
and carbonate AFm phases and Friedels salt can also be formed
at Cl/(Cl + OH) and Cl/(Cl + 1/2CO3) ratios in the range of 0 .2
1 .0 and 0 .11 .0 respectively (Balonis et al., 2010) but Portland
cements generally contain insufficient sulfate to form Kuzels salt
and its formation does not much influence the above threshold
values.
Thus the formation of Friedels salt takes place in AFm phases by
an ionic interaction mechanism between SO42, OH and CO32
and Cl (Birnin-Yauri and Glasser, 1998; Hosokawa et al., 2006;

SO4-AFm
(C4ASH12)

OH-AFm
(C4AH13)

CO3-AFm
(C4ACH11)

OH.0 .5CO3-AFm
(C4AC0 .5H12)

11 .4 wt%

12 .7 wt%

12 .5 wt%

12 .6 wt%

Table 4. Percentage of chloride combined in Friedels salt per


gram of reactant

70

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Suryavanshi et al. (1996) suggest that, as chloride is bound,


equivalent sodium ions have to be removed from solution to
maintain charge neutrality, and that the sodium thus removed is
bound in calcium silicate hydrate gel. Jones et al. (2003) used
27Al NMR and pore solution analysis to follow the formation of
Friedels salt from monosulfate AFm. These authors were concerned about charge balances when chloride is supplied from
sodium chloride and proposed that sodium ions are absorbed into
calcium silicate hydrate by reaction with silanol groups. The
protons thus released would balance a deficiency of positive charge
in solution but would be quickly absorbed by free hydroxyl ions to
produce water, which implies that more hydroxide ions are
released by ion exchange in AFm than would be predicted from a
simple ion exchange mechanism (Suryavanshi et al., 1996). However, according to Jones et al. (2003) the ion exchange mechanism
is dominant, at least in the early stages of reaction. These
arguments seem to be theoretical and based on arbitrary assumptions, proof of which is lacking. It is concluded that the complex
mechanism postulated by Suryavanshi et al. (1996), while theoretically possible, is neither convincing nor necessary.
It is reported that hydrated tetracalcium aluminoferrite (C4AF)
phases contribute to the binding of chloride but at a slower rate
than tricalcium aluminate (C3A), forming a chloro-complex
similar to Friedels salt but with iron instead of aluminium
(3CaO.Fe2O3.CaCl2.10H2O) (Csizmadia et al., 2001; Suryavanshi
et al., 1995). Suryavanshi et al. (1995) explored the formation of
the supposed iron analogue to Friedels salt phase in sulfateresistant cement with low tricalcium aluminate content. It was
supposed that ferrite would react with sulfate to form ettringite,
attributing the presence of calcium chloro-aluminate hydrate
(C3F.CaCl2.10H2O) to reaction of chloride with tetracalcium
aluminoferrite. Csizmadia et al. (2001) reported the formation of
calcium-aluminate/ferrite-chloro-hydrate AFm phase (C3 /AF/
.CaCl2.H10) in hydrated tetracalcium aluminoferritegypsum
mixes exposed to wetdry cycles with sodium chloride solutions.
The standard of proof showing that these reactions occur to any
great extent is not high. The iron analogue of Friedels salt
certainly exists, but is unlikely to be stabilised in normal Portland
cement paste environments. Only a small fraction of aluminium in
Friedels salt is normally substituted by iron(III) (Dilnesa et al.,
2011): iron activities are instead controlled by precipitation of
hydrous iron(III) oxide and appear to be too low to sustain
formation of an iron(III) Friedels salt. However, ferrite hydration
is an important source (or potential source) of alumina, which
must be taken into account in determining the overall hydrate
mineralogy.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Besides the contribution to Kuzels and Friedels salt formation,


sulfate/chloride ion exchange occurring among sulfate AFm
phases also promotes the formation of AFt as sulfate is increasingly substituted by chloride. The evidence for this conclusion
has been questioned: since ettringite does not contain chloride,
why should ettringite increase in amount upon addition of
chloride? The answer lies in simple mass balance considerations:
above a threshold concentration, chloride displaces sulfate from
AFm and it is this displaced sulfate that enables formation of
more ettringite (assuming aluminium is available): gypsum
cannot form because the cement normally remains undersaturated
with respect to gypsum.

Regarding the physical interactions of chloride with calcium


silicate hydrate, adsorption of chloride by calcium silicate hydrate
has been reported (Elakneswaran et al., 2009; Henocq et al.,
2006; Hosokawa et al., 2006). Because of the nature of calcium
silicate hydrate, with its high content of nanopores, it is difficult
if not impossible to distinguish between surface and bulk
sorption so the relative importance of the two mechanisms is not
known. Hirao et al. (2005) calculated the binding capacity of
calcium silicate hydrate, measuring (by ion chromatography) the
chloride concentration before and after sorption. The amount of
bound chloride increased with an increase in chloride concentration following a Langmuir-type adsorption isotherm, reaching a
maximum of 0 .4 mmol chloride per gram of calcium silicate
hydrate at chloride concentrations higher than 2 M (Figure 4(a)).
The values obtained were scaled, taking into account that
portlandite was also present in the samples: the correction
increased the maximum binding capacity of calcium silicate
hydrate to 0 .6 mmol/g or 21 .3 mg/g (Figure 4(b)).

Thus the identity of calcium chloroaluminate hydrates is well


established, although the critical chloride content above which
they become stable is not known accurately and the upper limits
of thermal stability of the chlorohydroxy phases are also not well
established, but recent data show they could remain a problem
even in tropical conditions (Galan et al., 2014). The impact of
sodium chloride, as distinct from calcium chloride, in promoting
their formation is not known; their formation may be responsible
for much of the reported deterioration of concrete in splash
zones, but there is not as yet sufficient evidence to quantify the
reaction mechanism and resulting potential for damage. High
chloride contents also enhance the solubility of calcium, with the
result that leaching is accelerated.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

10
y 047001

Bound chloride: mmol/g

According to Brown and Bothe (2004), compounds having the


general composition xCaO.yCaCl2.zH2O are stable but only at
high chloride concentrations: the 3:1:15 ratio compound is said to
be stable above 4 M and the 1:1:2 phase above 9 M chloride.
Brown and Bothe (2004) report the absence of an invariant point
portlanditeFriedels salt3:1:15 in the CaOAl2O3CaCl2H2O
phase diagram, but the invariant existence of portlandite3:1:15
1:1:2 and portlanditeFriedels salt1:1:2. Formation of a high
chloro-complex, 3CaO.3CaCl2.Al2O3.(aqueous), formed by the hydration of tricalcium aluminate in 23% calcium chloride solution
at 108C, has also been claimed (Jones, 1962).

Chloride binding capacity at millimolar concentrations is of most


interest to chloride migration through the solid matrix because of
the dependence of corrosion rates on low chloride concentrations.

08

264987x
1 264987x

06
04
02
0
0

10

Bound chloride: mmol/g

The above considerations affect the behaviour of chloride mainly


at low concentrations in the millimolar range. At much higher
concentrations, as may occur in zones where seawater evaporates,
Friedels salt is destabilised and complex compounds containing
calcium chloride, portlandite and/or calcium carbonate have been
reported to be stable below 208C and calcium chloride concentration above 15% (Chatterji, 1978). Damidot et al. (1994)
calculated the phase diagram of the CaOAl2O3CaCl2H2O
system at 258C and defined six isothermal invariant points.
According to these calculations, Friedels salt can coexist with
3CaO.CaCl2.15H2O (3:1:15) and portlandite, corresponding to an
invariant point with a chloride concentration of 3 .3 mol/kg.
Increasing the chloride concentration to 6 .4 mol/kg leads to
another invariant point where Friedels salt coexists with both
(3:1:15) and CaO.CaCl2.2H2O (1:1:2): portlandite is no longer
stable.

2
3
4
Chloride concentration: mol/l
(a)
y 061602

08

264947x
1 264947x

06
04
02
0
0

2
3
4
Chloride concentration: mol/l
(b)

Figure 4. Chloride binding curve of calcium silicate hydrate


without correction (a) and corrected for the amount of portlandite
(b) from Hirao et al. (2005)

71

Chloride in cement
Galan and Glasser

At low chloride concentrations (in the millimolar range), appropriate calcium silicate hydrate sorption values would cluster at
the left-hand edge of Figure 4 and, in this view, it is apparent that
chloride binding in calcium silicate hydrate, while not zero, is
negligible in terms of offering a significant retardation potential.
This is in accordance with the results presented by Elakneswaran
et al. (2009), shown in Figure 5, where experimental data are
interpreted using a surface complexation model in which chloride
is adsorbed on calcium silicate hydrate surface creating additional
negative surface charge due to the formation of SiOHCl from
un-ionised SiOH. The experimental values from Figure 5 were
deduced from the chloride binding isotherms of hydrated cement
paste and those of portlandite by an equilibrium concentration
technique (Tang and Nilsson, 1993) and from the chemical
binding of chloride forming Friedels salt, measured by XRD
Rietveld analysis.
Friedels salt is also said to adsorb a small amount of chloride
ions at its surface (Elakneswaran et al., 2009; Florea and
Brouwers, 2012). However, the authors view is that insufficient
evidence exists to support this: Friedels salt often contains minor
hydroxide (for chloride) substitution, and the data on the final
stages of hydroxide replacement by chloride are, in the authors
view, insufficient to distinguish structural replacement from
surface sorption.

Adsorbed chloride: mg/g of calcium


silicate hydrate

The adsorption of chloride on other phases portlandite and


ettringite is still tentative. Hirao et al. (2005) reported no
chloride binding capacity for either, but Elakneswaran et al.
(2009), based on zeta potential measurements, state that chloride
ions can be adsorbed on the positive surface of dissociated
portlandite ([CaOH]+), forming bonds of the type CaOHCl.
According to Elakneswaran et al. (2009), the adsorption of
chloride on portlandite and Friedels salt follows a Freundlichtype isotherm at free chloride concentrations of less than 1 mol/l
(Figure 6). The significance of this correlation is unclear:
Freundlich is designed to provide an empirical fit to data, so the
14

Experiment

12

Model

10
8
6
4
2
0
1

10
100
Chloride concentration: mmol/l

1000

Figure 5. Comparison of experimental and simulated chloride


adsorption on calcium silicate hydrate surface (Elakneswaran et
al., 2009)

72

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Amount of bound chloride: mg/g of sample

Advances in Cement Research


Volume 27 Issue 2

10

Portlandite

Friedels salt

8
7

y 03105x0458
R 2 099

6
5
4
y 00867x0617
R 2 099

3
2
1
0
1

10
100
Chloride concentration: mmol/l

1000

Figure 6. Adsorbed chloride on portlandite and Friedels salt


(Elakneswaran et al., 2009)

numerical values of the fitting coefficients lack fundamental


physicochemical significance.
Florea and Brouwers (2012), based on previously reported results
(Elakneswaran et al., 2009; Hirao et al., 2005), proposed chloride
binding isotherms for monosulfate AFm, hydroxyl AFm, CSH,
portlandite, Friedels salt and AFt. According to Florea and
Brouwers (2012), including portlandite and Friedels salt as
possible host phases for chloride retention improves the precision
of the proposed model, although the sorptive contributions of
phases other than Friedels salt only amounts to 25% of the
total bound chloride in an ordinary Portland cement (OPC) paste.
The mechanisms and concentration ranges in which chloride
retention mechanisms operate are contrasted and compared in
Figure 7. The data are presented as a series of bar graphs to the
same scale and also show regime boundaries, or fences, for
relevant binding processes. The most crucial benchmark relevant
to corrosion is that which marks the transition of embedded steel
between regimes of passivation and active corrosion; it is shown
in the top bar. This can be used to benchmark the importance of
other processes responsible for chloride bonding. Perusal of other
mechanisms active in this range (0 to 100 mmol) shows that
most chloride binding mechanisms described in the literature are
either relevant but rather have low binding capacities (e.g. the
low mass sorption of chloride by calcium silicate hydrate) or else
do not operate in the low concentration range. The only mechanism that operates in the appropriate concentration range and has
significant chloride binding capacity is that resulting from
conversion of ordinary AFm to Friedels salt. It is also
noteworthy that most of these binding mechanisms are reversible
and that the distinction between free and bound chloride can
only be made if arbitrary definitions are introduced.
Other factors affecting chloride binding
For many years, the chloride binding capacity of cement-based
materials was thought to be directly related to the aluminate content

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Steel

Steel passivation

AFm

AFm with hydroxide, sulfate, carbonate

AFt

Transition
Corrosion
Transition region

Gradual breakdown to form Friedels salt

AFt low sorption

Calcium
silicate
hydrate

Low sorption, gradually increasing

Calcium
hydroxide

Portlandite stable; low sorption

Breakdown

AFm Friedels salt

10

Breakdown
Chch

100
Chloride: mM

1000

Transition
region

10 000

Figure 7. Regimes of steel passivation/corrosion (top bar) relative


to chloride binding mechanisms and binding capacity of cement
solids, assuming aqueous pH  12 .5 and temperature  258C.
Chch: abbreviation for Calcium hydroxy chloride hydrates

of the cement (Hussain et al., 1995). However, importance has been


attached to the role played by alkalis in controlling the distribution
of chloride between the pore solution, AFm phases and calcium
silicate hydrate. Nielsen and co-workers (Geiker et al., 2007;
Nielsen et al., 2005) proposed a model for correlating the binding
of chloride by calcium silicate hydrate and AFm phases with
changes in the composition of the pore solution of Portland cement
pastes. The model considered the following three cases.
Sodium is present in the calcium silicate hydrate and no AFm
phases other than monocarboaluminate are present.
j Sodium is present in the calcium silicate hydrate and AFm
phases other than monocarbonate are present.
j Only alkali-free calcium silicate hydrate is present.
j

The latter case can be taken as a benchmark. According to this


model, the higher the content of alkalis, the lower the content of
bound chloride; that is, soluble alkali favours fractionation of
chloride into the aqueous phase.
The presence of supplementary cementing materials that modify
binder mineralogy may have an important influence on the
chloride binding capacity (Shi et al., 2012). For example, pastes
made with ground granulated blast-furnace slag (GGBS) normally
have higher silica and alumina contents than OPC, and might be
presumed to form higher quantities of precursors capable of
reacting with soluble chloride to form Friedels salt (Dhir et al.,
1996; Luo et al., 2003). Indeed, Dhir et al. (1996) measured
higher amounts of chloride bound in GGBS blended pastes than
in pure OPC, the binding capacity increasing with increasing slag
replacement level. These results were obtained from cement paste
prepared with OPC and additions of GGBS to constitute 0, 33 .3,
50 .0 and 66 .7% of the total binder: 70 mm cubes were water
cured at room temperature for 6 weeks. Sieved samples were
vacuum dried and mixed with 0 .1, 0 .5, 1 .0 and 5 .0 M calcium
chloride solution saturated with calcium hydroxide for 2 weeks
until adsorption equilibrium was reached (Tang and Utgenannt,

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

2009). At a GGBS replacement level of 67%, the chloride


binding capacity was five times that of Portland cement measured
at 5 M chloride. Luo et al. (2003) also reported experimental
evidence of the greater chloride binding capability of GGBS
pastes with the formation of higher amounts of Friedels salt.
These results are not questioned, but the authors have reservations
about their generic applicability. Slags vary widely in chemical
composition and fractional reactivity. The phase distribution of
slag pastes differs from that of Portland cement: the AFm phase
encountered in slag blends is typically a mixture of two AFm
phases: sulfate AFm and stratlingite, an AFm containing an
interlayer aluminosilicate anion. Its reactions with chloride are
not reported but Okoronkwo (unpublished work) has made phasepure stratlingite and found that its uptake of chloride at 208C
from 1 M sodium chloride was low relative to silica-free AFm.
This result would indicate that stratlingite is not responsible for
the higher chloride sorption reported for slag cements.
Other factors controlling the stability of phases are the pH and
temperature (Brown and Bothe, 2004; Wowra and Setzer, 2000).
Suryavanshi and Narayan Swamy (1996) studied the effect of
carbonation on the stability of Friedels salt by exposing concrete
slabs to chloride penetration and atmospheric carbonation with
analysis by XRD and DTA. They note that the solubility of
Friedels salt increases with the degree of carbonation. This is
attributed to the drop in alkalinity arising from carbonation, which
is said to promote the release of chloride from Friedels salt into
the pore solution. The present authors accept the conclusions but
believe that a simpler explanation exists that Friedels salt is
simply destabilised by carbonation and that the products of
reaction with carbonate have poor binding for chloride: later
stages of carbonation destroy Friedels salt and thus tend to release
chloride.
The cation accompanying the chloride salt plays a role both in
the formation of Friedels salt and in sorption by calcium silicate
73

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

hydrate. The cation effect is said to be related to the alkali


content of the pore fluid. According to Andrade and Page (1986),
specimens prepared with calcium chloride in the mix water are
more aggressive towards embedded steel than those made with
sodium chloride. However, this difference in aggression was
attributed to changes occurring in the early stages of hydration
rather than long-term modification of pore solution compositions.
Delagrave et al. (1997) measured more chloride bound in cement
paste specimens immersed in calcium chloride solutions than in
sodium chloride, both solutions having the same initial chloride
concentration. For chloride concentration of 0 .5 M, bound chloride was 7 mg and 12 mg of chloride per gram of paste in
sodium chloride and calcium chloride solutions respectively. At
1 M chloride, the values increased to 10 mg/g and 17 mg/g
respectively. Wowra and Setzer (2000) also reported higher
sorption of chloride in calcium chloride solutions relative to
sodium chloride and interpreted the calcium-induced chloride
adsorption as arising from an increase of sorption sites for
chloride due to the adsorption of calcium from calcium chloride,
resulting in a more positive surface charge; the values reported
for a chloride concentration 42 g/l were about 14 g/kg and 25 g/
kg of sorbed chloride from sodium chloride and calcium chloride
solutions at 258C respectively.

relatively unimportant. Some of the observed results could be due


to the changing stability and solubility of AFm phases, especially
in the presence of sodium.
Calculation of maximum chloride binding capacity
The authors conclude that some of the experimental work being
reported has lost touch with the original goal, namely to assess
the amount of chloride binding under realistic conditions and to
determine the role of chloride binding in corrosion: assuming a
threshold exists, can the retarding action of the cement matrix
significantly affect the time taken for the threshold concentration
to be reached?
To establish the magnitude of the binding it seems sensible to
calculate the maximum amount of chloride that could be bound
in a certain volume of concrete and relate this to
the cement content of concrete
the chloride concentration of the service environment
j numerical values of chloride diffusion and migration,
comparing the local values thus achieved with the corrosion
threshold
j
j

and, finally to
The effect of caesium, lithium and sodium, accompanying the
chloride, on the ability of calcium silicate hydrate to adsorb
chloride has been reported. Henocq et al. (2006) did not find
differences in the chloride-adsorption capability of calcium
silicate hydrate, with calcium/silicon ratio ranging from 0 .6 to
2 .4, and with hydrated tricalcium silicate pastes in the presence
of caesium, lithium and sodium chlorides for chloride concentrations ranging up to 100 mmol/l. The binding was measured by
several methods (change in concentration and zeta potential in
the aqueous phase and microprobe analysis of the solid) and all
showed similar results, namely that sorption of chloride was
independent of the cation. In all cases, at 100 mmol/l of chloride,
bound chloride was below 0 .03 mol of chloride per mole of
silicon. Note this latter value is equivalent to 5 .3 mg of chloride
per gram of calcium silicate hydrate, taking values for the
calculation from Henocq et al. (2006) (the molar weight of
calcium silicate hydrate with calcium/silicon 1 .5 was given as
200 g/mol).
The authors thus conclude that cation effects are either absent or

Cement content in Alumina in


concrete: kg/m3
cement: mass
%
300
500

3 or 7

recast the data in terms of the number of pore fluid


replacements required to saturate the binding capacity.

As an illustrative example, the following two scenarios (with


subsets) are considered.
Cement paste pores are initially filled with sodium chloride
equivalent to seawater concentration 0 .5 M (30 g/l) under
conditions of complete immersion.
j Pores are filled with saturated sodium chloride solution,
360 g/l (6 .2 M), simulating splash zones where salt
concentrations may rise due to evaporation.
j

The calculations were performed for a temperature of 258C and it


was assumed that the alumina content of cement paste is available
to form Friedels salt. It is known that part of the alumina may
initially be in ettringite but, at chloride concentrations of 0 .5 M
and above, ettringite slowly decomposes giving, among other
products, Friedels salt (Glasser et al., 1999a). Therefore, the

Maximum chloride
binding in 1 cm3
concrete: mg

Maximum chloride:
% mass of cement

6 .3 (3%) or 14 .6 (7%)
10 .4 or 24 .4

2 .1 or 4 .9

Table 5. Maximum amounts of chloride bound in concrete,


assuming cement alumina is converted to Friedels salt

74

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

initial presence of ettringite is deemed to be irrelevant because at


sodium chloride concentrations greater than 0 .5 M, aluminium
from all sources, including clinker ferrite, will become available
to form Friedels salt.

Steel in concrete

Table 5 summarises the calculations of the maximum chloride


bound in 1 cm3 of concrete assuming either of two different
cement contents. Table 6 shows data used for the calculation,
namely the dissolved chloride in the pore solution of 1 cm3 of
concrete. Commonly encountered ranges for cement content,
alumina percentage and pore volume are used. No corrections
were made for seawater components other than sodium chloride.
In the case of sorption from a saturated sodium chloride solution,
comparison of the data in Tables 5 and 6 shows that the reduction
in chloride concentration achieved by Friedels salt formation is
negligible in the high-concentration (saturated) scenarios; the
maximum amount of chloride that can be bound is much less
than the available chloride in the pore solution. (In fact this
scenario is unrealistic because, as explained later, portlandite
breaks down with formation of chloroaluminate hydrates: see
subsequent descriptions of the different equations which describe
the relevant reactions.) But in concrete with high cement content,
low porosity and/or high alumina percentages, the two quantities
become more nearly comparable. This is also true for water with
initially low chloride content: in these latter conditions, the
binding contribution from Friedels salt becomes much more
significant when evaluating a retardation scenario.
Thus, in the simulated seawater scenario, the contribution to
chloride removal resulting from Friedels salt formation is
significant to saturate the binding capacity of AFm, seawater
in the pores would have to be replaced several times. For
example, taking a concrete made with 400 kg/m3 cement,
assumed to contain 5% alumina and have 12% pore volume, the
paste could bind essentially all chloride contained in the pore
solution and about six cycles of pore water renewal would be
required to exhaust the binding capacity arising from forming
Cl-AFm. The retardation effect resulting from chloride uptake in
the cover concrete per unit thickness would then become a
relevant factor in assessing concrete performance by delaying
the arrival of chloride at the rebar, and hence the onset of
corrosion.

Concrete porosity:
% volume

7 or 15
7 or 15

Initial sodium
chloride in pore
solution: g/ml

Chloride in pore
solution per cm3 of
concrete: mg

0 .03
0 .36

1 .3 or 2 .7
15 .3 or 32 .8

Table 6. Amounts of chloride available in the pore solution of


1 cm3 of concrete

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

There is general agreement that in normal chloride-free conditions, steel embedded in concrete spontaneously develops a
passivating coating that partially protects the metal surface
against further corrosion. This film contains iron in two oxidation
states, iron(II) and iron(III), as will be further explained. The film
has a spinel-type structure and is well known to have low
permeation by analogy with the same type of film, but composed
of aluminium and oxygen, which passivates aluminium exposed
to moist air and water.
Steel, with its low concentration of (mainly) carbon, receives
broadly the same protection as iron so low-alloy steels are
generally treated using data for iron. This assumption is almost
universally made but it is noted that actual proof seems to be
lacking.
Corrosion rates are often used to define passive regimes, thus the
boundary conditions between regimes may reflect kinetics as well
as equilibrium. Reaction between iron and water in the passive
regime results in the disruption of water with liberation of
hydrogen. The reaction is slow: rates vary, but are typically a few
microns per year (expressed as thickness of iron/steel lost to
corrosion products). Thus, at normal temperatures and cement pH
1213, a passive regime for mild steel is established. However,
iron metal is never thermodynamically stable in contact with
water at high pH: slow diffusion of oxygen to the steel interface
and slow hydrogen release accompanying electrolysis of water
occur. These processes allow oxidative corrosion reactions to
proceed, albeit slowly.
Departures from the regime of spinel stability may occur. For
example, pH may drop as a consequence of carbonation of the
cement and, as a result, other and less protective solid corrosion
products develop. This allows a transition to occur between regimes
of passivation and corrosion. In theory, the transition between
regimes could be sharp, although as will be shown, a number of
perturbing factors, as yet incompletely understood, could lead to
fuzziness in defining the boundary conditions for passivation.
Stresses at a steelconcrete interface are generated because the
volume of corrosion products is frequently much greater than that
of the volume of iron that is oxidised, particularly in regimes of
active corrosion. The corrosion product layer may lose adherence
or crack, thus allowing corrosion to accelerate as aggressive
agents penetrate more readily.
A slow rate of formation of corrosion product is helpful to
preserving passivation: stresses arising from the thickening skin
of corrosion products can, in part, be relieved by slow diffusion
of cement components away from stressed interfaces. Indeed, the
passivating layer may persist and slowly thicken but without loss
of protection, as is evidenced by field examples of old (.100
125 years) reinforced concrete (RC) (Gode, 2010; Jackson, 1996;
Veto, 2010).
75

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

A species abundant in service environments and which is known


to lead to depassivation of steel is chloride. Thus, many of the
problems related to the deterioration of RC in saline environments arise from the presence of chloride ions introduced from
the service environment. Chloride permeates the concrete, eventually reaching the embedded steel, rendering its protective
passivating layer unstable. The consequences of rapid steel
corrosion in RC structures arise in part from damage to the steel
itself, with loss of cross-section by pitting, crevice and general
corrosion, as well as cracking of cover concrete arising from the
expansive nature of the corrosion products.

Passivation
The Pourbaix diagrams (Figures 8 and 9) (Pourbaix, 1966) for the
ironoxygenwater system identify a regime of passivation at
high pH and low Eh coincident with the field of Fe3O4. Iron
(III) oxide (Fe2O3) forms at high Eh and pH, giving a solid and
relatively insoluble oxidation product but, unlike spinel, it is not
adherent to metal and is porous with the result that ferric oxide is
much less protective than spinel. On the other hand, at low pH
and all Eh conditions which cannot normally be obtained in
cement pore water iron corrosion products are somewhat
soluble. These regimes, as usefully expressed in a Pourbaix
diagram, one of soluble iron(II) and the other of solid but nonprotective iron(III), are termed regimes of corrosion. From the
shape of the Pourbaix diagram in Figure 8, one of the conclusions
that might be made is that a definite set of pHEh conditions
exists that divides regimes of corrosion from those of passivation.

The above explanation is reasonably well established. But, as is


shown, many of the details are not well known or agreed. And, if
it is possible to control corrosion, many of these details assume
great importance.
22
20
18

16

20

14
b

12

FeO4?

0 2 4 6

Fe

10
08

10

FeOH
3

4
5

11

06

Fe(OH)2

E: V

04

28

Fe2O3

02
0

Fe

7
8

02
04

26
23

06

0
2
4
6

17

Fe3O4

08

6
HFeO2

27

13

10

24

Fe

12

14
16
18
2 1

Figure 8. PotentialpH equilibrium diagram for the system iron


water at 258C (considering as solid substances only iron, iron(II,III)
oxide (Fe3O4) and iron(III) oxide (Fe2O3) (Pourbaix, 1966).
Boundaries between the fields of iron (Fe) and iron(II) (Fe++) and

76

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

7
pH

10

11

12

13

14

15

16

between soluble iron and various solids are shown as a function


of log solubility (only the exponential value is shown) of soluble
iron. The sloping lines b and a mark the stability range of water

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

sulfate) could influence the nature and stability of corrosion


products.

However, it must be recalled that simplifications have been used


to construct the diagrams. One such simplification is that the
transition between regimes is arbitrarily constrained to a fixed
concentration of iron species in the aqueous phase, often set at
1 ppm. This definition is arbitrary and, in practice, contours can
be set at other concentrations: Figure 8 shows how the position of
the boundary shifts with different input assumptions regarding the
limit between soluble and insoluble regimes. Thus, theory
shows that the search for a single numerical value between
regimes is, in part, a matter of what constrains are placed on the
boundary.

For example, Figure 9 shows another and frequently cited


diagram showing a field of ferrous hydroxide largely replacing
that of spinel. Thus, uncertainty arises both from lack of knowledge of the constitution of soluble species as well as thermodynamic stability data for solids that might be stable or, if not
stable, at least persistent, at high pH.
Considering only iron and water, several subsequent reactions
contribute to the formation of the passive layer (shown here as
half-reactions all written as oxidations)

As noted, considerable uncertainty is attached to the construction


of the diagram because
the ironoxygenwater phases are incompletely characterised
the solubility of iron and its speciation are not reliably known
j the action of other cement components (e.g. hydroxide,
j
j

3Fe(OH)2 2OH ! Fe3 O4 4H2 O 2e

2a:

22
20
18

0
9

16

Fe

14

12

10
08

10

FeOH
3

4
5

06
04

E: V

FeO4?

21

20

11

Fe(OH)2

28

6
a

Fe

7
8

02
04

Fe2O3

Fe(OH)3

02

18
23

06

16

0
2
4
6

29

Fe(OH)2
19
12

08

6
4

HFeO2

10

24

12

Fe

14
16
18
2 1

Figure 9. PotentialpH equilibrium diagram for the system iron


water at 258C (considering as solid substances only iron, iron(II)
hydroxide (Fe(OH)2) and iron(III) hydroxide (Fe(OH)3) (Pourbaix,
1966). The limiting equations of interest to corrosion in

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

7
pH

10

11

12

13

14

15

16

cement matrices are (18) and (19); these are defined by (18)
Fe2+ + H2O FeO + 2H+, log(Fe2+) 13 .20  2pH and (19)
FeO + H2O HFeO2 + H+, log(HFeO2) 18 .30 + pH

77

Advances in Cement Research


Volume 27 Issue 2

2b:

Fe 2OH ! Fe(OH)2

2c:

2Fe3 O4 2OH ! 3Fe2 O3 H2 O 2e

2d:

Fe3 O4 OH H2 O ! 3FeOOH e

Although there is not complete agreement on the composition of


the passive layer, most authors conclude it is a physical mix of
iron oxides and hydroxides, the exact composition of which
depends on local pH and Eh conditions, which in turn reflect the
impact of external conditions such as oxygen flow and availability, and the evolution of the internal state with time. The relevant
insoluble solids are underlined in the above equations. Considering only the reaction with oxygen and water, the following
conclusions are possible.
j

Iron metal is not stable in contact with water under any


conditions. However, in a broad range of alkaline conditions,
the driving force/Gibbs free energy for iron to react, while
small, is sufficient to drive corrosion.
The electrochemical potential of steel, as measured relative to
a standard electrode and expressed in volt units, is a good
measure of its passivation/corrosion state while current flow,
in ampere units/surface area, is a good measure of corrosion
rate (Revie, 2011).
Corrosion can occur by hydrogen evolution from water but is
kinetically slow. This condition is characteristic of the
passivated state.
Arguably the most important of the solid corrosion products
is spinel because it forms a dense and adherent coating on
steel and passivates the surface.
The field marked Fe(OH)2 in Figure 9 is probably occupied in
part by spinel (i.e. the authors would prefer to use Figure 8
rather than Figure 9 because it describes what is observed).
In conditions such that oxygen ingresses, the redox potential
of the cement pore fluid in the vicinity of metal is low, so
sacrificial corrosion of the metal must occur, thereby
generating a diffuse halo of low Eh (reducing) conditions in
the vicinity of embedded steel.
Just as pH can be buffered, so can Eh (except that the process
is termed poising (Nightingale, 1958)). Thus the chemically
reduced zone (iron, iron corrosion product(s) and cement)
will have a poising capacity, giving it an ability to resist
change.
pH and Eh gradients may also develop spontaneously in the
vicinity of the steel.

Additional reactions between steel corrosion products and cement


components (e.g. sulfate) cannot be precluded, but conventional
Pourbaix diagrams lack sufficient dimensionality to accommodate
other than simple chemistries.
78

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Chloride in cement
Galan and Glasser

Impact of other ions on passivation


Redox-sensitive species in blended cements
When iron is embedded in Portland cement, its corrosion influences the Eh: iron has to oxidise to reduce the local Eh in the
cement matrix to around 500 mV to 600 mV (the potential set
by hydrogen evolution from water and shown as line a in Figure
9). However, when slag has been added in significant quantities,
its reactions contribute ions such as sulfide species to the pore
fluid, as a consequence of which the oxygen potential decreases
and pore fluid acquires a reducing Eh. The solubility of reduced
sulfur species also adds additional poising capacity not present in
plain Portland matrices. As a consequence of a vigorous sulfur
chemistry, it becomes more difficult to establish a passivated (low
Eh) state: ferric ions, an essential component of spinel, are
reduced to ferrous and as a result, short-term corrosion rates in
slag blends may be higher than in plain cements. The mechanism
is discussed by Macphee and Cao (1993) who conclude that steel
eventually becomes passivated in slag blends but that a longer
time is required to establish the passivated state than in plain
cements.
Interaction with chloride
Once chloride ions have penetrated the concrete cover, the
persistence of the protective passive layer formed on the steel
surface is put at risk. This layer, it will be recalled, was initially
formed in a high pH and low Eh regime from which chloride was
either absent or nearly so. A broad spectrum of observations has
given rise to the belief that a threshold concentration exists that
marks a transition between a stable low-corrosion regime and one
of more rapid corrosion.
The detailed mechanisms by which chloride ions disrupt the
passive layer are still incompletely resolved. Jovancicevic et al.
(1986) classified the existing models for describing the action of
chloride on passive layer breakdown into three stages in terms of
the mechanisms involved
adsorptiondisplacement
chemicalmechanical
j migrationpenetration.
j
j

The first mechanism suggests that the adsorption of chloride and


displacement of oxide and hydroxide (O2 and OH) from a
passivating monolayer leads directly to breakdown. This model
explains the dependence of film breakdown kinetics on chloride
concentration, but does not take into account the film thickness
often much more than a monolayer and its possible role on the
breakdown kinetics (Jovancicevic et al., 1986).
In the second mechanism, chloride adsorbed on the passive layer
surface is postulated to lower the interfacial tension of the oxide/
solution interface resulting from mutually repulsive forces between charged particles. The subsequent cracks and splits
produced in the film can further adsorb anions, which facilitate

Chloride in cement
Galan and Glasser

more reaction so that the process becomes progressive (Hoar,


1967).
45

Lin et al., 1981 also recognise the limitations of the model, for
example that it does not take into account surface structure,
physical inclusions or grain boundaries within the metal. Nevertheless, it includes many relevant factors and it explains why
corrosion often initiates at special sites (pitting) before progressing to general corrosion. It can be concluded that the model is
conceptually correct but needs scaling: it attempts to embrace
processes occurring across a range of scales from atomistic to
macroscopic and it is difficult to see how these mechanisms relate
to each other and how they impact on the observed property
changes.

G 0ads

G 0ads: kcal/mol

35
40
30

35

30

25

20
08
06
04
02
07
05
03
V/NHE

Figure 10. Free energy of adsorption of chloride G0ads and


interaction parameter r as a function of potential (Jovancicevic et
al., 1986). NHE refers to a voltage, V, measured against a normal
hydrogen electrode (the reference standard)

04 V/NHE
Borate buffer (pH 84)

Breakdown

4
01 V

: 10-9 mol/cm2

The migration models of the third group explain the breakdown


by an exchange process by way of cation vacancies and migration
of ions, oxide and hydroxide (Kruger, 1976). A point defect
model (Lin et al., 1981) is included in this description; it
considers that the cation diffusion rate in the passivating film is
enhanced by complexing iron(II) with chloride, the latter having
reached the film initially by occupying oxide vacancies. This
occupation results in the formation of voids at the metal/film
interface arising from accelerated iron dissolution in these zones
and, subsequently, to pit growth. According to Lin et al., 1981,
the proposed model is in quantitative agreement with experimental data for the pitting of iron in chloride-containing aqueous
solutions, explaining both the dependence of the critical pitting
potential on chloride concentration and of the incubation time on
breakdown overpotentials (defined as the difference between
applied potential and the minimum potential critical to initiate
pitting).

Transition
region
Bare iron
Passive layer

r: kcal/mol

Advances in Cement Research


Volume 27 Issue 2

10

02 V

05
1

Jovancicevic et al. (1986) conclude that desorption and absorption of chloride on iron are two distinct processes. Adsorption,
being potential and concentration dependent, is stronger in the
passive layer than on bare iron, which is explained by a charge
transfer between chloride (Cl) and iron (Fe2+ and/or Fe3+). The
adsorbed chloride displaces water molecules from the passive
layer, giving rise to further charge transfer. The adsorption on
bare iron and passive layer surfaces is described by a Temkintype isotherm

3:

G0ads RT
ln C

r
r

103
102
Chloride concentration: mol/l

101

Figure 11. Concentration dependence of absorption of chloride


(mol/cm2) for different potentials (Jovancicevic et al., 1986). is
the surface concentration of adsorbed and/or absorbed species.
The parameter on the right-hand vertical scale is the surface
coverage, defined as the ratio between and the maximum
surface concentration of adsorbed species (max). Note that max
only accounts for adsorbed chloride, which can lead to values of
greater than one

where G0ads is the free energy of adsorption of chloride and r an


interaction parameter. The results are shown in Figure 10: the
free energy of adsorption of chloride is higher on the passive
layer than on bare iron.

repulsion of the anions within the film, with an increase of the


saturation value as the potential grows more positive, corresponding to the increasing film thickness (Jovancicevic et al., 1986).

Absorption is assumed to be controlled by the diffusion of


chloride into the film. Figure 11 shows the linear dependence of
absorption on concentration. This relation is attributed to mutual

In summary, Figures 10 and 11 give proof of the two possible


interaction processes occurring between chloride and passive
layer on iron: adsorption and absorption.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

79

In addition to the already mentioned factors affecting chloride


binding by cement, factors related to the steel are reported to
affect the onset of corrosion. These include the steel surface
characteristics (Alonso et al., 2000; Mammoliti et al., 1996),
electrical potential of the rebar (Alonso et al., 2002; Angst et al.,
2009) and steel bar diameter (Liang et al., 2005; Lu et al., 2011;
Rasheeduzzafar et al., 1992; Sakashita et al., 1999). The nature
of the steelconcrete interface has also been assigned an important effect on the onset of corrosion but this is still subject to
debate: different interface microstructures with and without local
concentration of portlandite have been reported (Monteiro et al.,
1985; Page, 2009), which may be attributed, at least partially, to
formulation. Stress can also be a factor in corrosion but is not
included in the scope of this review.
Involvement of cement components in the corrosion
process
The breakdown of the passivating film may involve cement
components other than hydroxide, water and chloride. For
example, an association between the formation of green rust
(GR) and disruption of the passive layer has been reported
(Sagoe-Crentsil and Glasser, 1993). GR refers to a family of
crystalline compounds whose layer structures are based on
iron(II)iron(III) double-layer hydroxysalts (hydroxyl-chloride,
-sulfate, -carbonate, etc.). These phases have long been known in
the laboratory and some compositions have natural mineral
equivalents. The GR phases are typically not stable in air so,
upon exposure to oxygen, iron(II) in GR rapidly oxidises and its
crystalline structure is destroyed. Nevertheless, the GR phases are
believed to have physicochemical ranges of stability at high pH,
similar to conditions that develop spontaneously at steelcement
interfaces.

Chloride in cement
Galan and Glasser

04

Fe(OH)2
FeOOH

16
4

0
2
15

Eh: V

Advances in Cement Research


Volume 27 Issue 2

FeOH

Fe
04

HFeO2

19

0
2

Fe4(OH)8Cl
17

14

10

12

13

11

pH
(a)
04

Fe(OH)2

Fe2O3

26

0
23

Figures 12 and 13 show the corresponding Pourbaix diagrams,


modified to include chloride and showing calculated stability for
chloride-containing GR phases. The standard chemical potential,
0, of GR1 corresponding to 3Fe(OH)2.Fe(OH)2Cl was determined
from the work of Refait and Genin (1993) using experimental
values of Eh and pH at invariant points involving this compound
with other phases: ferrous hydroxide (Equations 4a4d), -iron
oxide-hydroxide, lepidocrocite (-FeOOH) (Equations 5a5d) and
hydrated magnetite (Equations 6a6d).
4a:

80

4Fe(OH)2 Cl 3Fe(OH)2 :Fe(OH)2 Cl e

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Eh: V

Measurement of the Eh potential during oxidation processes


leading to the formation of GR compounds has allowed calculation of their standard Gibbs free energy of formation, enabling
them to be included in a modified Pourbaix diagram for the iron
water system. The stoichiometry and stability of GR compounds,
Fe(II)3Fe(III)(OH)8Cl.nH2O (Refait and Genin, 1993; Refait et
al., 1998), [Fe(II)4Fe(III)2(OH)12][CO3.2H2O] (Drissi et al., 1995)
and [Fe(III)4Fe(III)2(OH)12][SO4.2H2O] (Genin et al., 1996) has
been reported.

13

6
Fe(OH)2

Fe
08

11

18

27

15

HFeO2

FeOH
24

Fe
04

Fe

0
2

4 (O

14

H)

8 Cl

17

21

4
6

Fe(
OH
)2 ,
6

11

Fe(

Fe

08

2Fe

OO

7
18

25

9
pH
(b)

11

22

20

OH

)2

12

13

Figure 12. Calculated EhpH equilibrium diagram at 258C and


activity [Cl] 0 .35 mol/l for the systems FeGR1FeOOHH2O
Cl (a) and FeGR1Fe2O3H2OCl (b) with GR1 Fe4(OH)8Cl
(Refait and Genin, 1993)

Advances in Cement Research


Volume 27 Issue 2

4b:

Chloride in cement
Galan and Glasser

0 (3Fe(OH)2 :Fe(OH)2 Cl) 509 720  500 cal=mol

Eh E0  0.0591 log [Cl ]


5d:

(or 2 134 198  2094 J=mol)

E0 [0 (GR1)  40 (Fe(OH)2 )  0 (Cl )]=23 060

4c:

3[3Fe(OH)2 :Fe(OH)2 Cl] 4[Fe(OH)2 :2FeOOH]


where E0 0:554  0.01V

3Cl 8H 5e

6a:

0 (3Fe(OH)2 :Fe(OH)2 Cl) 509 325  230 cal=mol


6b:

4d:

(or 2 132 544  961 J=mol)

5a:

3Fe(OH)2 :Fe(OH)2 Cl 4FeOOH 3e Cl 4H

E0 [40 (Fe(OH)2 :2FeOOH) 30 (Cl )


6c:

5b:

Eh E0 0.0197log[Cl ]  0.0788 pH

Eh E0  0.0946 pH 0.0355 log [Cl ]

 30 (GR1)]=115 300

where E0 0:446  0.017V

0 (3Fe(OH)2 :Fe(OH)2 Cl) 509 910  650 cal=mol


5c:

E0 [40 (FeOOH) 0 (Cl )  0 (GR1)]=69 180

where E0 0:440  0.007V

17
0 24 6

Eh: V

metallic ferrite (-Fe) and iron(II) hydroxide (Fe(OH)2) the


former obtained by reduction of GR
j ferric oxide or oxyhydroxide (lepidocrocite, -FeOOH)
obtained by oxidation of GR
j aqueous iron(II) (Fe2+), which appears when GR dissolves
(Genin et al., 2002).
j

-FeOOH
1

20

a
04

06

Fe(OH)2

16
Fe

As shown in Figures 12 and 13, the domain of existence of


chloride GR is confined to a region lying between

02

FeOH

GR(Cl )

12

13
6

Fe
6

The extent of the domain of GR will, in general, increase with


increasing activity of the anion (in this instance, chloride, but
sulfate may also be involved).

19

18

FeOOH

10

0 11
2
4
6

(or 2 134993  2722 J=mol)

The value for the standard chemical potential (average value of


the three processes mentioned above) is determined to be
0 509 500  500 cal/mol (or 2 133 277  2094 J/mol).
3

02

6d:

Fe(

OH

)2

6
8

10
pH

11

12

15
6
14

13

14

Figure 13. EhpH equilibrium diagram of the system iron/chloride


containing solution at 258C for an activity [Cl] 0 .55,
corresponding to a 1 M potassium chloride solution; data for
GR(Cl) Fe4(OH)8Cl.nH2O taken from (Refait et al., 1998)

Because of the presence of additional components, restrictions


have to be placed on the construction and interpretation of 2D
Pourbaix-type diagrams and, in this case, calculations were
restricted to a fixed chloride activity. For example, Figure 13
was constructed for an activity (chloride concentration) similar
to that found in seawater and Genin et al. (2002) note that since
the pH of seawater is about 8 .0 to 8 .5, it can be seen that
GR(Cl) would form easily as a corrosion product of iron since
it would precipitate as soon as the activity of Fe2+ reaches 104.
As explained by Sagoe-Crentsil and Glasser (1993), the supposed

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

81

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

very low solubility of iron and its oxides/hydroxides at high pH


is enhanced by the presence of chloride. The enhancement in
solubility is greatest as the pH decreases. While species mobility
is not an explicit function of solubility, higher solubility does
associate roughly with higher mobility and, as a result, iron
transport will almost certainly be increased by formation of
soluble aqueous chlorocomplexes. Thus, soluble aqueous iron(II)
chlorocomplexes can potentially be mobile in high pHlow Eh
regimes.
Figure 14 shows a calculated Ehchloride concentration stability
diagram at pH 12. GR is stable under reducing conditions and
it acts as a solid sink for chloride ions. Oxidation reactions
leading to destruction of the GR phase will, however, liberate
chloride and enable a cyclic process to occur as the chloride thus
liberated is attracted back to anodic regions where it is free to
reform GR (Figure 15).
It is important to remark on the role of oxygen activity in the
overall process: the differing solubilities may result in iron being
solubilised and transported away from the steel with oxidative
precipitation occurring. In these circumstances, the iron thus
removed and precipitated, often at some distance from the
corroding steel, would be physically unable to contribute to
formation of a passivating film.

08
06
04

Eh: V

02
Iron hydroxide

0
Mag.

02
04

GR1
c

Hydroxide

06

FeCl24H2O

d
Iron

08
5

1
3
2
log(Cl): mol/l

Figure 14. Ehchloride concentration [Cl] stability diagram at


pH 12 (Sagoe-Crentsil and Glasser, 1993). According to SagoeCrentsil and Glasser (1993) The critical boundary conditions for
the change from spinel to GR occur along the line ab. At
chloride concentrations close to the thresholds, the exact value of
which are Eh and [Cl] dependent, the solubility of iron
undergoes a large increase while the nature of the stable solid
also changes from spinel (abbreviated as Mag, for magnetite), to
green rust (GR1)

82

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Oxygen
Electrolyte
Chloride
Chloride

GR(chloride)

-FeOOH

Steel

Figure 15. Sketch of the mechanism by which chloride is liberated


from GR as a result of its oxidation to form akaganeite (FeOOH). The liberated chloride may again be transported to sites
where GR is forming; thus the GR can act as a transient sink

Figure 14 shows a somewhat different view of mineralogical


development and offers an attractive starting point to explain the
role of chloride. In this view, GR has a definite field of stability
and is increasingly stabilised at higher aqueous chloride activity.
However, Figure 14 is based on calculations using a thermodynamic database of uncertain quality.
Additionally, a large body of evidence suggests that the breakdown of passivation in the presence of chloride may be complex.
Akaganeite (-FeOOH) is often observed as a reaction product
and is probably stabilised by chloride substituting in part for
hydroxide. Ferrous hydroxychloride (-Fe2(OH)3Cl) has also been
identified in the corrosion products of iron from archaeological
artefacts (Reguer et al., 2007). This latter phase has not, to the
authors knowledge, been observed in laboratory accelerated
corrosion tests. According to Remazeilles and Refait (2008), the
explanation for this lies in the slow formation kinetics of ferrous
hydroxychloride at pH and concentration equilibrium conditions
intermediate between those of ferrous hydroxychloride and
ferrous hydroxide. The formation of ferrous hydroxychloride is
said to occur by transformation of metastable precursors. These
latter appear to be iron(II) hydroxychloride layer compounds
made of ferrous hydroxide layers alternating with ferrous chloride
layers, structurally similar to ferrous hydroxide, with chloride
ions substituting at random for hydroxide ions (Remazeilles and
Refait, 2008). But if this account is correct, the salt thus formed
does not contain iron(III) and is unlikely to have a structure
associated with a GR-like phase.
The oxidation of ferrous hydroxychloride, in an aqueous solution
of ferrous chloride, like that of ferrous hydroxide (Fe(OH)2) leads
to the formation of chloride GR (GR1), the formation of which
depends on the appropriate concentration of dissolved species.
GR1 is said to oxidise further to lepidocrocite (-FeOOH) in the
presence of a slight excess of dissolved ferrous chloride, or to an
overchlorinated GR1 in highly concentrated solutions (Refait
and Genin, 1997). The end product depends on the Eh and
concentration of dissolved species. The fully oxidised solid
product goethite (-FeOOH) does not form directly from the GR

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

compound. Goethite formation is induced by a dissolved


iron(II) species and the complex akaganeite (FeCl2,aq.-FeOOH)
formation is favoured by an increase of the chloride in the
overchlorinated GR1 precursor. This mechanism is reported to
take place in concentrated (23 M) chloride solutions and at pH
values between 4 and 6. Such conditions could be self-generated
in confined media such as at the interface between iron and its
corrosion products, where chloride ions accumulate (Refait and
Genin, 1997).

conditions and at relatively low chloride concentrations, implies a


shift of the stability fields in equilibrium mineralogy away from
those shown for the ironwateroxygen system even at low
chloride concentration (102 M). However, the quality and
quantity of thermodynamic data on the impact of chloride vary
greatly between the various reports. Phases are reported to occur
and are found in nature for which there is a lack of thermodynamic data. More data are thus needed to determine the
stability of particular phases, on the transition between regimes
and on the relevant kinetics of phase development. The importance of chloride-containing solids would help explain frequent
reports of enhanced concentrations of chloride in the vicinity of
steel, as they are required for the formation of essential iron(II)
as well as chloride and the steelconcrete interface is the only
place where the Eh of Portland cement paste is sufficiently low to
stabilise iron(II). This view would also suggest that breakdown of
the spinel passivating layer occurs with the formation of chloridecontaining solids, including GR.

Hibbingite (-Fe2(OH)3Cl) may also be expected to form at high


sodium chloride concentration, according to the solubility data of
Nemer et al. (2011). For the reaction
7:

Fe2 (OH)3 Cl(s) 3H $ 3H2 O 2Fe2 Cl

the solubility constant of ferrous hydroxychloride (Fe2(OH)3Cl(s))


at infinite dilution and 258C was found to be
8:

log10 K 17.12  0.15

and for the reaction


9:

Fe(OH)2 (s) 2H $ 2H2 O Fe2

the solubility constant of ferrous hydroxide (Fe(OH)2) at infinite


dilution and 258C was found to be
10:

log10 K 12.95  0.13

The body of evidence showing the formation of these chloridebearing phases, many of which seem to occur under equilibrium

Sequence

Concentrating on processes at the vicinity of the interface,


several possibilities for binding exist and it is likely that the
mechanisms shown in Table 7 are composition dependent.
Table 7 correlates the changing oxidation state of iron with

Iron oxidation state


Fe (iron)
(0)

In the absence of chloride, many models have been presented


explaining passivation and corrosion in terms of pH and Eh. The
data are relatively good and calculations made using relatively
simple models are approximately in accord with experiment.
However, when chloride is introduced, it interacts in ways that
cannot necessarily be foreseen from models. Experiments, confirmed by natural occurrences, are telling us that a range of
phases, some containing essential chloride and others containing
anions furnished by the cement (e.g. sulfate), are involved in the
corrosion process: the occurrence of phases in the GR family is
cited as example. Archaeological evidence may not be relevant as
the pH conditions may be lower than those prevailing in concrete.

Fe (iron) (II)

Mixed states

Notes
Fe (iron) (III)

Fe (iron)!

Fe(OH)2
!
Fe3O4
! FeOOH (iron(III) oxide-hydroxide),
(ferrous hydroxide)
(iron(II,III) oxide)
Fe(OH3) (ferric hydroxide)
Fe (iron)! -Fe2(OH)3Cl (ferrous
Fe(OH)3(iron(III) oxide-hydroxide),
hydroxychloride)
-FeO(OH) (iron(III) oxide-hydroxide)

Form when chloride is


absent or low
Low chloride: FeO(OH)
(akaganeite) contains some
Cl (chloride)
Iron ! -Fe2(OH)3Cl (ferrous ! Green rust ! Fe(OH)3 (iron(III) oxide-hydroxide), Green rust stabilised by
hydroxychloride)
-FeO(OH) (iron(III) oxide-hydroxide) Cl (chloride), SO42
(sulfate), CO32 (carbonate)

Table 7. Sequence of corrosion product formation showing


changes in oxidation state and chloride concentration; pH is
assumed to be high, c. 1213

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

83

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

chloride concentration. It is apparent from a survey of the


literature that a number of pathways are possible: other
reactions may need to be included in Table 7. The extent of
anion substitutions in the relevant phases, with implications for
their thermodynamic stability, is not at present known so
construction of the diagram is, to some extent, speculative.
Although sequence A in Table 7 operates in the absence of
chloride, it is possible (but unlikely) that the hydroxides thus
formed can incorporate significant chloride. However mechanisms B and C bind chloride and the product phases require
essential chloride for stabilisation.

hypothesised that reduction of water to give hydrogen occurs at


or near the steel/spinel boundary while reduction of water to
hydroxide occurs in or near the iron oxide/cement interface. This
zoned mineralogical layer forms spontaneously (i.e. it is selfassembling) and, if mechanically disrupted, is self-healing. This
duplex layer thickens slowly with time but maintains its
integrity.

A modern picture of the corrosion scenario is summarised with


the aid of Figure 16. In the passive state, a strong oxygen
potential gradient occurs between air, at the left-hand side, and
embedded steel. The Eh gradient (ABB9) is maintained by slow
corrosion at minus several hundred millivolts. The normal (i.e.
chloride-free) product of corrosion is a spinel phase consisting of
iron oxide.
The Eh gradient is thought to steepen through the passivating
layer (note that the diagram is not to scale as the passivating
layer may be only a few microns to tens of microns thick
whereas the cover concrete is typically several centimetres or
more). Its oxygen composition is graded but details of the
gradient are not known. In the absence of chloride, the passivating layer is probably composed of zones of spinel (innermost)
and an outer layer of oxidised (and possibly hydrous) phases
with iron(III). The spinel layer is semiconducting and it is

Passivating
layer
A

PO2 (Eh)

B
Air
Concrete
cover

Steel

Surface

Depth

Figure 16. Schematic diagram showing redox state of passivated


steel in concrete and corresponding redox gradients for concrete
exposed to air. The Eh is fixed at two points, A and B: A is fixed
by the fugacity of oxygen in the standard atmosphere, here
assumed to be air or oxygen-saturated water; B is approximately
fixed by the potential to discharge hydrogen from water at
pH  13. See text for additional information. The diagram is not
to scale and the passivating layer, across which the redox gradient
is steepest, may be only microns thick

84

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

In the presence of chloride, however, the layer mineralogy alters,


as has been discussed in previous sections. Many mechanisms
may be operative, depending on temperature, salinity, pH and
species activities.

Testing and test methods


Simulating conditions at the steelconcrete interface
The interface is, in the authors view, a region of indefinite
thickness including the zone of altered mineralogy but also all
parts of the system in which transport phenomena processes occur
that control the overall reactions. Thus, for example, reactions
physically distant from the steelcement interface, such as
transport of chloride and of oxygen through the cover concrete,
may need to be included in the system. Most studies on the
composition of the passive layer at high pH have, of necessity,
been carried out using simulates. These usually comprise alkaline
solutions of either sodium or calcium hydroxide (or both).
The chemical nature and concentration of the solution as well as
the absence of cement solids may result in differences in
composition and thickness of the film compared with the products
formed on steel in mortar or concrete (Montemor et al., 1998).
The results are appraised by first considering simple simulations
in chloride-free environments. Haupt and Strehblow (1987) used
electrochemical techniques and X-ray photoelectron spectroscopy
(XPS) to characterise the passive layer formed on iron in sodium
hydroxide solutions, finding an inner iron(II)-rich layer and an
outer iron(III)-containing oxide; the latter was said to act as a
barrier to ion migration and it was found to thicken linearly with
the electrode (an iron bar) potential and log time as expected for
a high-field mechanism of oxide growth. The outer layer is first
formed when the soluble corrosion products change from iron(II)
to iron(III) and when the electrode capacity achieves its maximum value. The inverse of the capacity is said to increase
linearly with potential and consequently increasing with iron(III)
oxide thickness (Haupt and Strehblow, 1987).
Oranowska and Szklarska-Smialowska (1981) compared the
passive film formed in saturated calcium hydroxide solutions
under open circuit conditions (no current) and at anodic potentials (externally applied current) by means of electrochemical and
ellipsometric measurements. In the open circuit case, the metaladhering layer grown on the surface of iron was said to exhibit a
refractive index of 1 .43, which the authors (Oranowska and
Szklarska-Smialowska, 1981) claim is composed of iron
hydroxide. At anodic potentials, the grown layer has a refractive

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

index of 1 .7 and is said to be composed of iron(III) oxidehydroxide (FeOOH) and calcium hydroxide.

which are less electrically conductive than iron(II,III) oxide


(-Fe3O4) spinel.

Hugot-Le Goff et al. (1990) identified the products formed on


iron in 1 M sodium hydroxide using in situ Raman spectroscopy
and a rotating split ring disc electrode. A rotating ring disc
electrode consists of a double working electrode which, by
rotation, induces a flux of analyte to the electrode. The central
disc and the ring electrodes are separated by a non-conductive
barrier and connected to a potentiostat. In the split ring
electrode, the half-rings are maintained at independent potentials, which allow for the detection of iron species (Fe2+ and
Fe3+). From the cyclic voltammograms and the Raman spectra
of solids, Hugot-Le Goff et al. (1990) conclude that iron oxide
(Fe3O4) is the main constituent of the passivating film. Products
of anodic oxidation of iron can deposit on this inner layer of
Fe3O4, including Fe(OH)2, Fe(OH)3, -FeOOH, -FeOOH,
-Fe2O3 and Fe3O4 (arising from the transformation of other
compounds). The nature of the product(s) is dependent on the
potential applied and the time of exposure to the sodium
hydroxide solution.

Ghods et al. (2011) used XPS to evaluate the properties of the


passive layer on carbon steel formed in saturated calcium hydroxide solution. They determined the thickness of the film to be
about 4 nm, and this did not change between the second and ninth
day of exposure. They also reported inner and outer film layers to
be mainly formed by iron(II) and iron(III) oxide phases respectively. Miserque et al. (2006) reported a similar thickness (less
than 6 nm) of corrosion products on mild steel when immersed in
simulated pore solutions and characterised by XPS. The differences between these values and those reported by Montemor et al.
(1998), who measured a layer nearly 100 nm thick, are attributed
to the use of the oxygen spectra to measure film thicknesses:
oxygen spectra may lead to an overestimation due to formation of
oxygen-rich precipitates such as calcium hydroxide.

The use of 1 M sodium hydroxide solutions in corrosion studies


is justified by Joiret et al. (2002) by the fact that the redox
processes in the passive layer are well differentiated and that
sodium hydroxide concentrations are not far from those typically
encountered in the pore fluid of OPC pastes. Using various
electrochemical techniques, including a ring disc electrode, they
point to the influence of the redox potential of the constitution of
the oxide layer on the electrochemical behaviour of the system.
According to their results, the passive film is based on a
magnetite-structured phase (for which spinel is the generic term).
The spinel structure can tolerate a range of iron(II)/iron(III)
ratios, so it can be oxidised and reduced within definite limits
while still maintaining single-phase integrity, its limit of stoichiometry depending on the electrode potential, temperature and
oxygen activity.
Freire et al. (2009) studied the passive films formed on mild
steel in sodium hydroxide and potassium hydroxide solutions by
applying anodic potentials using XPS, auger electron spectroscopy (AES), atomic force microscopy (AFM) and electrochemical impedance spectroscopy (EIS) for product
characterisation. They concluded that the surface film is structured, with iron oxides (Fe2+) in the inner layer and an outer
layer rich in iron hydroxides (Fe3+), the growth of the outer
layer being topotactic.
More realistic formation conditions for the passivating layer
include the use of aqueous solutions made with sodium hydroxide
and/or potassium hydroxide saturated with calcium hydroxide.
Sanchez et al. (2007) applied an anodic potential in saturated
calcium hydroxide solutions with 0 .2 M potassium hydroxide and
concluded that the film becomes more resistive with ageing. The
ageing process is associated with enrichment of iron(III) oxides,

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

The film growth process was followed by Abd El Haleem et al.


(2010) using steel immersed in saturated calcium hydroxide
solutions and measuring the open-circuit potential during the first
4 h. They found a logarithmic dependency between the steadystate potential and the calcium hydroxide concentration:
E a1  b1 log C(CaOH)2 (Figure 17). They also reported a
logarithmic relation between the open circuit potential and the
immersion time (E a2  b2 log t), which leads to the conclusion
that the initial rate of oxide film thickening follows a direct
logarithm law. An increase in the concentration and pH or
temperature led to a decrease in the film growth rate due to its
partial dissolution (Figures 18(a) and 18(b) respectively).
Poursaee and Hansson (2007) measured the time needed to grow
a passive film in steel immersed in mortar or in synthetic pore
solution. This was done by monitoring the corrosion rate until
values considered to correspond to passivity were reached. They
reported that growth of the film required 7 d and 3 d in mortar
and synthetic pore solution respectively. The composition of the
synthetic pore solution used was 9 .17 g NaOH, 31 .4 g KOH,
0 .96 g CaSO4.2H2O, 4 .2 g Ca(OH)2 in 1 .75 l of water.
In summary, there is abundant evidence that a passivating layer
develops spontaneously on steel embedded in cement. The details
of its mineralogical evolution are not completely agreed, but it is
apparent that a spinel-structured phase is an important component
of the film. This film is extremely insoluble at high pH and is
physically adherent both to metal and cement paste. The film
may be of only nm thickness in the early stages of growth, but
nevertheless supports a significant Eh gradient its inner surface
contacting metal has a low Eh (about 600 mV), maintained by
the slow release of hydrogen by reduction of water, whereas the
outer surface of the passivating layer is at higher Eh, such that it
coexists with ferric hydroxide and/or oxyhydroxide(s). Several
phases have been singled out occurring in this outer zone and it
is possible that it does not consist of a single phase but is a
mixture. Simulate aqueous solutions can be used in place of
85

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Calcium hydroxide
10 104 M

50 103 M

10 103 M

200

Initial rate of oxide film thickening:


nm per unit decade of time

25 103 M

50 10

10 102 M
Saturated calcium
hydroxide

250
300

16

15
110

400
450

50

100

150
Time: min
(a)

200

250

300

175
200
225

Efin: mV (SCE)

17

105

500

250
275
300
325
350
50 45 40 35 30 25 20 15 10
log CCa(OH)2: M
(b)

Figure 17. Variation of the open circuit potential of the steel


electrode with time (a) and variation in the steady final potentials,
Efin, of the steel electrode with concentration (b) in naturally
aerated calcium hydroxide solutions; data from Abd El Haleem et
al. (2010)

actual cement and appear to give comparable data on the


evolution of the protective layer.
Chloride concentration gradients in concrete
Chloride diffusion profiles are usually measured either by
determining the total chloride content in the concrete at measured depths and/or by measuring the free chloride content in the
pore solution, perhaps indirectly using a pore fluid expression
step. The methods used are summarised in Table 8. The total
and free chloride contents are usually determined by extraction
and are often subdivided into acid-soluble and water-soluble
chloride. However, only slender evidence exists that this latter
86

18

350

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

log (rate of oxide film thickening): nm/decade of time

Electrode potential: mV (SCE)

19

50 105 M

030

115

120
pH
(a)

125

130

135

028
026
024
022
020
018
016
014
012
010
00030

00031

00032
00033
1/T: K1
(b)

00034

00035

Figure 18. Variation of the initial rate of oxide thickening of the


steel electrode with the pH of the solution (a) and variation of the
logarithm of the initial rate of oxide thickening of steel electrode
with 1/T in naturally aerated 103 M calcium hydroxide solutions
(b); data from Abd El Haleem et al. (2010)

division, and the associated analytical methods used to determine acid- and water-soluble fractions, have physical significance.
The chloride concentration gradients in concrete may be determined from drilled cores cut into slices at selected intervals. The
sampling method restricts the spatial resolution of the technique.
Drilled powders can also be used. Solid slices are ground and the
resulting powders are either dissolved in dilute nitric acid solution
for subsequent chloride concentration measurement or else
pressed into pellets for X-ray fluorescence spectrometry (XRF).
The free chloride content is measured from the concrete either by
applying high-pressure pore fluid expression or by mixing ground
concrete samples with a solvent, nominally acid or water. For
some types of calculation it is desirable to determine chloride in

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Measurement

Sample preparation

Analysis

Total chloride content

Pressed powder
Powder dissolution in dilute nitric acid

X-ray fluorescence spectrometry (XRF)


Titration (potentiometric or colorimetric)

Free chloride content

Pore solution expression followed by analysis


Leaching method in crushed sample

Ion selective electrode measurement (ISE)

Table 8. Methods for determining chloride content in concrete.


Titration and ISE apply to all sample preparations except the
pressed powder, where XRF is used

the cement paste, in which case it is also necessary to determine


the cement content of a concrete.

laborious sample preparation and does not distinguish between


free and bound chloride.

According to Dhir et al. (1990), the acid-soluble chloride content


determination underestimates total chloride (i.e. acid extraction
does not remove all chloride from the powdered concrete
samples). Dhir et al. (1990) also note that different acid extraction techniques, with differences in acid concentration, temperature and contact time with concrete powder, led to different
results. The total chloride content in concrete can, it is claimed,
only be determined by XRF (Dhir et al., 1990).

It would thus appear that chloride diffusion profiles can be


determined accurately using existing techniques.

Other methods to measure chloride profiles have also been


reported. Jensen et al. (1996, 1999) measured chloride ingress
using the scanning capabilities of electron probe microanalysis
(EPMA). This technique proved to be accurate for the determination of chloride profiles on a micron scale with 1001000 times
better spatial resolution than conventional drilling or sawing
techniques. The method is especially suitable for high-performance cement pastes where chloride penetration in non-accelerated
tests may be very slow. EMPA also allows for the characterisation
of potential local routes for chloride ingress, such as cracks, and
along interfacial transition zones (i.e. at pasteaggregates interfaces). Jensen et al. (1996, 1999) recognise that EPMA requires

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

log C b a log C b

11:

where a and b are adsorption constants and Cb and C are the


bound and free chloride contents respectively. At low concentrations (below 0 .05 M) a Langmuir equation fit was applied

12

Bound chloride: mg/g

A round-robin chloride test was carried out in 30 laboratories and


the results of 64 measurements of total chloride content were
reported by Castellote and Andrade (2001a). Three methods were
used for extracting chlorides from the solid sample and six
methods for analysing the solution obtained. According to the
results, the use of any of the methods can lead to rigorous
measurement of the total chloride content. However, the best
reliability was obtained when each laboratory extracted the
chloride according to the method normally used. For the watersoluble chloride content, the round-robin test involved 20 laboratories and 37 measurements (Castellote and Andrade, 2001b).
The values were compared to a target obtained from the pore
expression technique. Again, a conclusion of the test is that free
chlorides in concrete can be determined rigorously using any of
the methods available. XRF was not, however, included in these
tests, which consisted in all cases of an extraction and an analysis
method.

Chloride binding isotherms


Most authors use isotherms of bound chloride against free
chloride (Figure 19) for determining the chloride binding
capacity of concrete. Tang and Nilsson (1993) developed a
method based on adsorption from solution to calculate the
isotherms, assuming that local equilibrium is achieved. The data
can be expressed at high concentrations (above 0 .01 M) by a
Freundlich equation

10
8
6
4
2
0
0

02

04
06
08
Free chloride: mol/l

10

12

Figure 19. Typical plot of bound chloride against free chloride


isotherm. Data obtained by immersing samples in a chloride
solution of known concentration until equilibrium is reached. The
bound chloride content in the sample is deduced from the
difference between the initial and the equilibrium chloride
concentration

87

Advances in Cement Research


Volume 27 Issue 2

12:

Chloride in cement
Galan and Glasser

1
1 1
1

C b kC bm C C bm

where k is the adsorption constant and Cbm is the bound chloride


content at saturated monolayer adsorption. Figure 20 shows the
data fittings for both concentration ranges and the corresponding
equations. The method of Tang and Nilsson (1993) is widely used
for the calculation of isotherms. However, some authors
(Jirickova and Cerny, 2006) consider this method to overestimate
the real chloride binding capacity of concrete as it uses crushed

specimens for the tests and then back-calculates the binding


capacity to the amount of cement. Tang and Nilssons method
would, in this view, be an idealisation of the binding problem,
assuming that chloride equilibrates with every grain of hydrated
cement.
Jirickova and Cerny (2006) performed experiments on small
but not crushed granular cement paste specimens to maintain
the effect of the permeation structure of concrete and compared
the chloride binding isotherms obtained with data from crushed
samples. They conclude that chloride binding capacity of

20
w/c 04 adsorption

log Cb 03864 log C 1412

Bound chloride: mg/g gel

w/c 06 adsorption

R2 0953

10

w/c 08 adsorption
w/c 04 desorption

w/c 06 desorption
w/c 08 desorption

log Cb 03788 log C 1140


R2 0996

0001 0002

0005 001 002


005 01
Free chloride: mol/l solution
(a)

02

05

12
w/c 04 OPC paste
1/Cb 0002438/C 01849

10

w/c 06 OPC paste

R2 0997

1/Cb: g gel/mg chloride

w/c 08 OPC paste

08

06

04

02

100

200
300
1/C: l solution/mol chloride
(b)

Figure 20. Freundlich (a) and Langmuir (b) isotherms of OPC


pastes (Tang and Nilsson, 1993)

88

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

400

Chloride in cement
Galan and Glasser

concrete does depend on the sample size and method of preparation, values obtained for crushed specimens being greater than
for non-crushed concrete.
Castellote et al. (1999a) used electrochemically assisted methods
on non-steady-state chloride migration experiments at constant
voltage (12 V) with 0 .05 M to 1 M sodium chloride solutions for
21 d and with 0 .5 M solutions for 7, 14, 21 and 28 d. The results
obtained were fitted to a BrunauerEmmettTeller (BET) isotherm (Brunauer et al., 1938, 1940) (Figure 21); they extended
the trends to higher concentrations to compare them with
published data on natural diffusion (i.e. not electrically accelerated) (Sergi et al., 1992). However, the isotherms (Figure 22)
show different shapes, with convex isotherms being obtained in
the migration experiments. The data extensions of the isotherms
allowed the authors to find a chloride concentration at which
combined chlorides were similar. The classification of isotherms
proposed by Brunauer, Deming, Deming and Teller (BDDT)
(Brunauer et al., 1940) was applied. The classification differentiates five types of isotherms, the different shape of each type
corresponding to a different adsorption process.
In the migration process, both the shape of the isotherm and the
value of the fitting parameters correspond to the so-called type III
or V isotherms. Type V isotherm has an inflection point at high x
values, as opposed to type III, where the convex character persists
over the range of values shown on the x-axis. This fact, as well as
the proposed existence of a limit in the percentage of combined
chlorides, enabled the authors to conclude that the chloride
isotherm in migration experiments should be of type V. The
differences in the ratio between free and combined chlorides in
migration and diffusion were attributed, at least partly, to a
change in the controlling step of the process, the transport being

Bound chloride: % concrete

012
009
006
003
0
003
0

005

010
015
[Cl]/[Clsat]

020

025

Figure 21. Experimental data of bound chloride against the ratio


between free chloride in the pore solution and chloride
concentration of a saturated sodium chloride solution at constant
temperature; fitting of the data to a BET isotherm from Castellote
et al. (1999a). The dashed curves represent 95% confidence
interval

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

20

Bound chloride: % cement

Advances in Cement Research


Volume 27 Issue 2

15

Obtained from natural diffusion data

10
05
Obtained from migration data

0
0

20

40
60
80
Free chloride: g/l

100

120

Figure 22. Chloride binding isotherms determined from


electrically assisted migration Castellote et al. (1999a) and natural
diffusion experiments Sergi et al. (1992)

much slower in natural diffusion, opposite to what happens in


migration.
Chloride diffusion
Several methods and standards have been proposed in recent
years for measuring chloride transport in concrete, most considering saturated conditions and diffusion-controlled transport.
The methods used to measure chloride diffusion can be classified
following different criteria whether the test is natural or
accelerated, by using electrochemical means, or by the variable
measured (such as chloride concentration, current or conductivity). Yuan and colleagues (Yuan, 2009, Yuan and Santhanam,
2013) proposed a classification, shown in Table 9, of the methods
for which the basic equations are given in Table 2. These
methods (Aashto, 1980; Aenor, 2009; Alexander et al., 1999;
Andrade et al., 2000; ASTM, 2003, 2012; Castellote et al.,
1999b, 2001, 2002; Freeze and Cherry, 1979; Friedmann et al.,
2004; Halamickova et al., 1995; Lu, 1998; Nordtest, 1995, 1997,
1999; Page et al., 1981; Samson et al., 2003; Shi et al., 1999;
Stanish et al., 2000; Streicher and Alexander, 1995; Truc et al.,
2000) are used in conjunction with analytical methods to
determine chloride concentrations presented in Table 8.
Ficks first and second laws, used for describing ideal stationary
and time-dependent concentration regimes respectively, are still
the most widely used relations with which to describe the physics
of diffusion. Because the laws apply to homogeneous media, they
have to be applied to concrete assuming certain boundary
conditions. It is known that this approach is too simplistic, but it
is not known how much error is introduced by using simplifications.
The so-called ponding tests, based on Ficks second law, aim to
calculate numerical diffusion coefficients from the chloride
profile measurements of a sample after direct exposure to a saline
solution. Figure 23 shows two possible geometries for the
ponding test: in the first case, the saline solution is placed directly
on the cement surface allowing chloride to diffuse across the
89

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Theoretical base

Test method

Measurement

Test duration

Reference

Ficks first law


Ficks second law

Steady-state diffusion test


NT Build 443
ASTM C1556
NT Build 355
Trucs method
NT Build 492/Tangs method
Breakthrough time method
Multi-regime method

Chloride flux
Chloride profile
Chloride profile
Chloride flux
Chloride flux
Penetration depth
Breakthrough time
Anolyte conductivity

Several months
.35 d
.35 d
Several weeks
Several days
2472 h
Several weeks
Several days

Samsons method
Friedmanns method
Lus method
Andrades method
Formation factor method
ASTM C1202/Aashto T227
Conductivity test method
Aashto T 259
Water pressure method

Current
Current
Resistivity
Resistivity
Resistivity
Charge passed
Conductivity
Chloride profile
Penetration depth

120 h
Several
Several
Several
Several
6h
Several
90 d
Several

AC impedance method
Integral method

Impedance
Electrical potential

Several minutes
Several days

Page et al. (1981)


NT Build 443 (Nordtest, 1995)
ASTM C1556-11a (ASTM, 2003)
NT Build 355 (Nordtest, 1997)
Truc et al. (2000)
NT Build 492 (Nordtest, 1999)
Halamickova et al. (1995)
Castellote et al. (1999b, 2001)
UNE 8398709 (Aenor, 2009)
Samson et al. (2003)
Friedmann et al. (2004)
Lu (1998)
Andrade et al. (2000)
Streicher and Alexander (1995)
ASTM C1202-12 (ASTM, 2012)
Alexander et al. (1999)
Aashto T 259-80 (Aashto, 1980)
Freeze and Cherry (1979)
Stanish et al. (2000)
Shi et al. (1999)
Castellote et al. (2002)

NernstPlanck
equation

NernstEinstein
equation
Formation factor
Other

weeks
minutes
minutes
minutes
minutes
weeks

Table 9. Summary of test methods for chloride transport in


concrete (Yuan, 2009; Yuan and Santhanam, 2013)

interface, the sides being protected by an impermeable cylindrical


tube. In the second geometry, the sample is immersed in the
solution but all surfaces except that across which chloride will
diffuse are coated with a chloride-impermeable layer. Both
arrangements are basically the same chloride is supposed to
diffuse in one direction.

Saline
solution
Saline
solution

Concrete
specimen

Concrete
specimen

The chloride penetration depth values are usually fitted to a


modified solution of Ficks second law, which allows calculation
of a diffusion coefficient. This calculation and the Ficks law
solution used to fit the data are dependent on the experimental
conditions of the tests. The preparation of the samples before
immersion and the chloride concentration recommended by two
of the standards used to measure these conditional diffusion
coefficients, ASTM C1556 (ASTM, 2003) and NT Build 443
(Nordtest, 1995), are summarised in Table 10.
The accelerated methods use electric fields to force chloride
migration through concrete. The concrete thus acts as a diffusion
membrane between a chloride-containing solution and an initially
basic or neutral but chloride-free solution (Figure 24). No attempt
is made to balance the ionic potentials of the two solutions. Most
accelerated tests apply the NernstPlanck equation to calculate
diffusion coefficients considering either steady-state (Aenor,
2009; Nordtest, 1997) and/or non-steady-state conditions (Castellote et al., 2001; Nordtest, 1999). Modifications of Ficks second
law are also reported. Halamickova et al. (1995) used a modified
version of Ficks second law including a migration term

13:



@C
@ 2 C zFE @C
D

@t
@x2
RT @x

Figure 23. Experimental set-ups for ponding test for measuring


chloride diffusion in concrete

Although Yuan (2009) classifies the Halamickova equation


90

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Curing before test

Chloride solution for test

Immersion in saturated calcium hydroxide 165 g/l sodium chloride: solution


solution at 238C until constant mass is
replaced after 5 weeks, 21258C
achieved
(average 238C)
Table 10. Conditions for measuring diffusion coefficients
according to ASTM C1556 (ASTM, 2003) and NT Build 443
(Nordtest, 1995)

Cathode

Chloride-rich
solution

Concrete

Anode

Chloride-free
solution

Figure 24. Experimental set-up for a chloride migration test; V is


DC voltage delivered from an external source

(Halamickova et al., 1995) in the NernstPlanck block, it also


involves Ficks law.
The multi-regime method reported by Castellote et al. (2001) is
based on the measurement of the conductivity in the anodic
compartment of the cell and conversion of the value thus obtained
to chloride concentration. The steady-state diffusion coefficient
can be then calculated using a modified NernstPlanck equation
(Andrade, 1993). For the non-steady-state diffusion coefficient
the authors give equations related to the time-lag (intersection of
straight line of chloride steady-state flux with the x-axis) and the
equivalent time between migration and diffusion. This latter is
based on the time-dependent theoretical solution to ionic mass
transport through a uniform membrane under the influence of a
uniform electric field, as derived by Keister and Kasting (1986).
The expression for the ratio of the voltage-enhanced time-lag,
(), to the passive time-lag, (0), is

14:

()
6
2 [  coth(=2)  2]
(0) 

where

15:

zeV
kT

V being the voltage drop across the membrane and k Boltzmanns


constant.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

The differences between these accelerated tests lie mainly in the


experimental set-up and in the variable measured from which to
calculate the diffusion coefficient, for example, using the chosen
equation, chloride concentration at the anode, chloride flux,
conductivity, current, and so on; all have been proposed as
parameters on which to base calculations.
The use of the diffusion coefficient obtained from certain experimental data to obtain and predict new data has often led researchers
into unwarranted extensions of the data: the numerical values
necessary for predictions determined under a particular set of
conditions cannot necessarily be extrapolated to other conditions.
The lack of consistency in results obtained by different investigators and different methods has led to a round-robin test on
methods for determining chloride transport parameters in concrete. This exercise was carried out in 27 laboratories using four
concrete mixes and 13 test methods (Castellote et al., 2006). The
methods used were divided into four types diffusion, migration,
resistivity and colorimetric. However, the authors of the present
paper note that colorimetric is a method of analysis and is not a
method for controlling diffusion. Different set-ups were compared (diffusion cell, immersion test, ponding, migration cell,
etc.), the same device often being used for different methods. The
methods were evaluated by four indicators
trueness (related to the deviation to target or average value)
precision (repeatability and reproducibility)
j relevance (quality of information)
j convenience (handling).
j
j

According to the data presented, the better test methods in terms


of trueness are NT Build 443 (Nordtest, 1995), natural diffusion
by ponding (Whiting, 1981) and the multi-regime method for
determining steady-state coefficients (Castellote et al., 2001). The
better precision is obtained by ASTM C1202 (ASTM, 2012). The
methods with the highest relevance appear to be the natural
diffusion (ponding) test (Whiting, 1981), the resistivity method
(Andrade et al., 2000) and the multi-regime method (Castellote et
al., 2001). Finally, the method classified as the most convenient is
the resistivity method described by Andrade et al. (2000).
The authors of the round-robin test report (Castellote et al., 2006)
consider that the analysis cannot be linked to the ability of the
91

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

tests to be used in models of prediction for long term performance. The prediction through modelling is out of the scope of
present exercise.

meantime, a more in-depth analysis of existing test methods is


needed to improve their reliability. Probably, any of the techniques described will be sufficient to rank concretes in merit
order but are insufficient to support lifetime prediction.

The reliability of these methodologies for making long-term


predictions is thus still not resolved.
The writers also express concern about the predictive value of
these tests. Those that use natural diffusion are bound to be
influenced by concentration differences which, in turn, are
artefacts of the experimental conditions imposed by test specification. For example, considering Figure 24, the ionic potentials of
the two solutions, chloride-rich and the initially chloride-free
solution, will almost certainly differ greatly in ionic strengths.
This difference generates an osmotic pressure between the two
solutions that, in turn, drives water from the pure water compartment through the cement and into the saline compartment in
order to equalise osmotic pressures. If chloride diffuses, it will
have to do so against the osmotic drive. The osmotic pressure
will, in general, not remain constant but will decrease with time
as soluble cement components dissolve in the pure water
compartment and as chloride breakthrough is achieved. And, of
course, during the early stages, reactions leading to pore blocking
can occur between each of the solutions and the cement paste;
the nature and consequences to permeation properties will tend to
be asymmetric as cement near the anode and cathode compartments are exposed to quite different regimes of precipitation and
dissolution.
These differences are in part related to aqueous compositions and
in part artefacts of the cell design with the result that, as reaction
continues, the reaction path is influenced by factors such as the
mass of cement and volumes of solution in the compartments and
so on. The fundamental principle that chloride migrates is
achieved, but the numerical values are strongly influenced by
highly artificial factors imposed by the selection of test conditions.
In electrically assisted methods, typically using tens of volts
potential difference, several unwanted electrolysis reactions (e.g.
liberation of chlorine, hydrogen and oxygen gases) may occur.
Joule heating may also occur as a consequence of electrical
losses. The impacts of these side reactions on test reproducibility
and results are largely unknown but are bound to affect the
numerical values obtained.
Many relevant parameters of these tests are undefined by the
specifications and it is concluded that the existing methods
suffer from inadequate analysis of the experimental design
insufficiently specify the conditions of the test
j have potential for systematic error
j show no proven relationship to actual field conditions.
j

Finally, it is important to remark that the modelling of chloride


ingress has evolved considerably in recent years, from the first
models that considered single ion transport and saturated conditions to the recent models focusing on multi-ionic approaches
and permitting analysis of diffusion in unsaturated conditions.
Samson et al. (2005) modelled ionic transport relations in
unsaturated porous materials including gradients in the electrochemical potential and the moisture content using the homogenisation technique. In this approach, the transport equations
are first written at the microscopic level (i.e. at the scale of the
pore). The equations are then integrated over a representative
element volume (REV), the size of which depends on the intrinsic
properties of the material: in concrete, the size of the REV
depends on the maximum diameter of the aggregate particles.
The equations thus obtained consist of time-dependent equations
for both the concentration of ionic species within the pore
solution and the moisture content within the pore space.

Conclusions
Chloride is not now allowed in concretes involving steel
reinforced construction. However, chloride is present in many
service environments where it migrates into concrete. Concrete
itself is little damaged by chloride unless its concentration
exceeds several molar, although portlandite solubility is enhanced
relative to pure water even at low (0 .1 M) concentrations.
However, the paste mineralogy of calcium aluminate hydrates
(AFm) is affected and Friedels salt stabilises at a few tens of
millimolar chloride.
Steel in concrete is normally well passivated by the cement
environment and a chemically reducing environment forms spontaneously in the vicinity of steel. This passivation, however,
breaks down in the presence of chloride. The transition between
regimes of passivation and corrosion is the subject of debate.
Does it depend on chloride content alone or on the ratio of
hydroxide/chloride? And should a distinction be made between
free and bound chloride? The state or condition of the steel
cement interface is complicated by the involvement of other
cement components (e.g. oxygen partial pressure and the activities of other anions such as sulfates and carbonates) so
laboratory simulations based on simplistic considerations give
numerical values that cannot be expected to be universally
applicable. The proposed concept of thresholds dividing regimes
of corrosion from passivation is likely to be fuzzy.

The search for meaningful yet accelerated tests continues. In the


92

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

The migration of chloride through cover concrete is a complex


process both the physical quality of the matrix and the binding
of chloride into cement components are responsible for the
observed diffusion. This diffusion is not strictly Fickian: charge

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

balances have to be conserved in the course of diffusion and it is


desirable to use a NernstEinsteinPlanck model.

salts: a study of cation specific effects. British Corrosion


Journal 21(1): 4953.
Andrade C, Alonso C, Arteaga A and Tanner P (2000) Methodology
based on the electrical resistivity for the calculation of
reinforcement service life. Proceedings of 5th CANMET/ACI
International Conference on Durability of Concrete,
Barcelona, Spain, pp. 899915.
Angst U, Elsener B, Larsen CK and Vennesland (2009) Critical
chloride content in reinforced concrete a review. Cement
and Concrete Research 39(12): 11221138.
Arya C and Xu Y (1995) Effect of cement type on chloride
binding and corrosion of steel in concrete. Cement and
Concrete Research 25(4): 893902.
ASTM (2003) ASTM C1556: Standard test method for
determining the apparent chloride diffusion coefficient of
cementitious mixtures by bulk diffusion. ASTM International,
West Conshohoken, PA, USA.
ASTM (2012) ASTM C1202-12: Standard test method for
electrical indication of concretes ability to resist chloride ion
penetration. ASTM International, West Conshohoken, PA,
USA.
Balonis M, Lothenbach B, Le Saout G and Glasser FP (2010)
Impact of chloride on the mineralogy of hydrated Portland
cement systems. Cement and Concrete Research 40(7): 1009
1022.
Birnin-Yauri UA and Glasser FP (1998) Friedels salt,
Ca2Al(OH)6(Cl, OH) .2H2O: Its solid solutions and their role
in chloride binding. Cement and Concrete Research 28(12):
17131723.
Brown P and Bothe J Jr (2004) The system CaOAl2O3CaCl2
H2O at 23  28C and the mechanisms of chloride binding in
concrete. Cement and Concrete Research 34(9): 15491553.
Brunauer S, Emmett PH and Teller E (1938) Adsorption of gases
in multimolecular layers. Journal of the American Chemical
Society 60(2): 309319.
Brunauer S, Deming LS, Deming WE and Teller E (1940) On a
theory of the van der Waals adsorption of gases. Journal of
the American Chemical Society 62(7): 17231732.
Castellote M and Andrade C (2001a) Round-robin test on
chloride analysis in concrete part I: analysis of total
chloride content. Materials and Structures 34(243): 532556.
Castellote M and Andrade C (2001b) Round-robin test on
chloride analysis in concrete part ii: analysis of water
soluble chloride content. Materials and Structures 34(244):
589598.
Castellote M, Andrade C and Alonso C (1999a) Chloride-binding
isotherms in concrete submitted to non-steady-state migration
experiments. Cement and Concrete Research 29(11): 1799
1806.
Castellote M, Andrade C and Alonso C (1999b) Modelling of the
processes during steady-state migration tests: quantification
of transference numbers. Materials and Structures 32(217):
180186.
Castellote M, Andrade C and Alonso C (2001) Measurement of
the steady and non-steady-state chloride diffusion coefficients

Existing test methods are found not to be realistic and accelerate


different parameters at different rates: accelerated tests are liable
to introduce artefacts.
In theory, it would be possible to optimise the resistance to
diffusion of chloride through concrete, giving rise to a chlorideproof concrete. However, the presence of cracks in real materials
and the need to optimise concrete properties against a wide range
of diffusion-related degradation processes (e.g. sulfate ingress
and sulfate-induced deterioration) mean that multiple strategies
designed to protect steel may be needed chloride attack cannot
be considered in isolation. Nevertheless, poor fundamental understanding of a number of key processes involving chloride and a
lack of generic data on the various aspects of corrosion handicap
efforts to improve performance and provide quantitative lifetime
assessments.

Acknowledgement
The work reported here was funded through research grant
number ENG016 RGG0593, awarded by the Gulf Organization of
Research and Development (GORD), Qatar.

REFERENCES

Aashto (American Association of State Highway and


Transportation Officials) (1980) T 259-80: Standard method
of test for resistance of concrete to chloride ion penetration.
Aashto, Washington, DC, USA.
Abd El Haleem SM, Abd El Aal EE, Abd El Wanees S and Diab A

(2010) Environmental factors affecting the corrosion


behaviour of reinforcing steel: I. The early stage of passive
film formation in Ca(OH)2 solutions. Corrosion Science
52(12): 38753882.
Aenor (2009) Concrete Durability, Test Methods, Measurement of
Chloride Diffusion Coefficient in Hardened Concrete,
Multiregime Method. Aenor, Madrid, Spain.
Alexander MG, Ballim Y and Mackechnie JR (1999) Concrete
durability index testing manual. Research Monograph 4
Department of Civil Engineering, University of Cape Town,
Cape Town, South Africa.
Alonso C, Andrade C, Castellote M and Castro P (2000) Chloride
threshold values to depassivate reinforcing bars embedded in
a standardized OPC mortar. Cement and Concrete Research
30(7): 10471055.
Alonso C, Castellote M and Andrade C (2002) Chloride threshold
dependence of pitting potential of reinforcements.
Electrochimica Acta 47(21): 34693481.
Andrade C (1993) Calculation of chloride diffusion coefficients
in concrete from ionic migration measurements. Cement and
Concrete Research 23(3): 724742.
Andrade C and Page CL (1986) Pore solution chemistry and
corrosion in hydrated cement systems containing chloride

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

93

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

in a migration test by means of monitoring the conductivity


in the anolyte chamber. Comparison with natural diffusion
tests. Cement and Concrete Research 31(10): 14111420.
Castellote M, Andrade C and Alonso C (2002) Accelerated
simultaneous determination of the chloride depassivation
threshold and of the non-stationary diffusion coefficient
values. Corrosion Science 44(11): 24092424.
Castellote M, Andrade C, Kropp J et al. (2006) Round-robin test
on methods for determining chloride transport parameters in
concrete. Materials and Structures 39(294): 955990.
Chatterji S (1978) Mechanism of the CaCl2 attack on Portland
cement concrete. Cement and Concrete Research 8(4): 461
467.
Csizmadia J, Balazs G and Tamas FD (2001) Chloride ion binding
capacity of aluminoferrites. Cement and Concrete Research
31(4): 577588.
Damidot D, Birnin-Yauri UA and Glasser FP (1994)
Thermodynamic investigation of the CaOAl2O3CaCl2H2O
system at 25C and the influence of Na2O. Il Cemento 4: 243
254.
Delagrave A, Marchand J, Ollivier J, Julien S and Hazrati K (1997)
Chloride binding capacity of various hydrated cement paste
systems. Advanced Cement Based Materials 6(1): 2835.
Dhir RK, Jones MR and Ahmed HEH (1990) Determination of
total and soluble chlorides in concrete. Cement and Concrete
Research 20(4): 579590.
Dhir RK, El-Mohr MAK and Dyer TD (1996) Chloride binding in
GGBS concrete. Cement and Concrete Research 26(12):
17671773.
Dilnesa BZ, Lothenbach B, Le Saout G et al. (2011) Iron in
carbonate containing AFm phases. Cement and Concrete
Research 41(3): 311323.
Drissi SH, Refait P, Abdelmoula M and Genin JMR (1995) The
preparation and thermodynamic properties of Fe(II) Fe(III)
hydroxide-carbonate (green rust 1); Pourbaix diagram of iron
in carbonate-containing aqueous media. Corrosion Science
37(12): 20252041.
Eglinton M (2003) Resistance of concrete to destructive agencies.
In Leas Chemistry of Cement and Concrete, 4th edn.
(Hewlett P (ed.)), Chapter 7. Butterworth-Heinemann,
Oxford, UK, pp. 299342.
Elakneswaran Y, Nawa T and Kurumisawa K (2009)
Electrokinetic potential of hydrated cement in relation to
adsorption of chlorides. Cement and Concrete Research
39(4): 340344.
Florea MVA and Brouwers HJH (2012) Chloride binding related
to hydration products: part i: ordinary Portland cement.
Cement and Concrete Research 42(2): 282290.
Freeze RA and Cherry JA (1979) Groundwater. Prentice Hall,
Englewood Cliffs, NJ, USA.
Freire L, Novoa XR, Montemor MF and Carmezim MJ (2009)
Study of passive films formed on mild steel in alkaline media
by the application of anodic potentials. Materials Chemistry
and Physics 114(23): 962972.
Friedmann H, Amiri O, At-Mokhtar A and Dumargue P (2004)

A direct method for determining chloride diffusion coefficient


by using migration test. Cement and Concrete Research
34(11): 19671973.
Galan I, Perron L and Glasser FP (2014) Impact of chloride-rich
environments on cement paste mineralogy. Cement and
Concrete Research, in press.
Geiker M, Nielsen EP and Herfort D (2007) Prediction of chloride
ingress and binding in cement paste. Materials and Structures
40(4): 405417.
Genin J-R, Olowe AA, Refait P and Simon L (1996) On the
stoichiometry and Pourbaix diagram of Fe(II)Fe(III)
hydroxy-sulphate or sulphate-containing green rust 2: an
electrochemical and Mossbauer spectroscopy study.
Corrosion Science 38(10): 17511762.
Genin J-R, Refait PH and Abdelmoula M (2002) Green rusts and
their relationship to iron corrosion; a key role in microbially
influenced corrosion. Hyperfine Interactions 139140(14):
119131.
Ghods P, Isgor OB, Brown JR, Bensebaa F and Kingston D (2011)
XPS depth profiling study on the passive oxide film of carbon
steel in saturated calcium hydroxide solution and the effect of
chloride on the film properties. Applied Surface Science
257(10): 46694677.
Glasser FP, Kindness A and Stronach SA (1999a) Stability and
solubility relationships in AFm phases. Part I. Chloride,
sulfate and hydroxide. Cement and Concrete Research 29(6):
861866.
Glasser FP, Tyrer M, Quillin K et al. (1999b) The Chemistry of
Blended Cements and Backfills Intended to Use in
Radioactive Waste Disposals. Environment Agency, Bristol,
UK.
Gode K (2010) Service life of two 100 year old concrete bridges
in Latvia. Scientific Journal of Riga Technical University
11(3): 1320.
Halamickova P, Detwiler RJ, Bentz DP and Garboczi EJ (1995)
Water permeability and chloride ion diffusion in Portland
cement mortars: Relationship to sand content and critical
pore diameter. Cement and Concrete Research 25(4):
790802.
Haupt S and Strehblow H (1987) Corrosion, layer formation, and
oxide reduction of passive iron in alkaline solution: a
combined electrochemical and surface analytical study.
Langmuir 3(6): 873885.
Henocq P, Marchand J, Samson E and Lavoie JA (2006) Modeling
of ionic interactions at the C-S-H surface application to
CsCl and LiCl solutions in comparison with NaCl solutions.
Proceedings of 2nd International RILEM Symposium on
Advances in Concrete Through Science and Engineering,
Quebec City, Canada.
Hirao H, Yamada K, Takahashi H and Zibara H (2005) Chloride
binding of cement estimated by binding isotherms of
hydrates. Journal of Advanced Concrete Technology 3(1):
7784.
Hoar TP (1967) The production and breakdown of the passivity of
metals. Corrosion Science 7(6): 341355.

94

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Hosokawa Y, Yamada K, Johannesson BF and Nilsson LO (2006)

Influence of limestone on the hydration of Portland cements.


Cement and Concrete Research 38(6): 848860.
Lu C, Jin W and Liu R (2011) Reinforcement corrosion-induced
cover cracking and its time prediction for reinforced concrete
structures. Corrosion Science 53(4): 13371347.
Lu X (1998) Rapid determination of the chloride diffusivity in
concrete. In Proceedings of the Second International
Conference on Concrete under Severe Conditions, Troms,
Norway (Gjrv OE, Sakai K and Banthia N (eds)), pp. 1963
1969.
Luo R, Cai Y, Wang C and Huang X (2003) Study of chloride
binding and diffusion in GGBS concrete. Cement and
Concrete Research 33(1): 17.
Macphee DE and Cao HT (1993) Theoretical description of
impact of blast furnace slag (BFS) on steel passivation in
concrete. Magazine of Concrete Research 45(162): 6369.
Mammoliti LT, Brown LC, Hansson CM and Hope BB (1996) The
influence of surface finish of reinforcing steel and pH of the
test solution on the chloride threshold concentration for
corrosion initiation in synthetic pore solutions. Cement and
Concrete Research 26(4): 545550.
Manera M, Vennesland and Bertolini L (2008) Chloride
threshold for rebar corrosion in concrete with addition of
silica fume. Corrosion Science 50(2): 554560.
Matschei T, Lothenbach B and Glasser FP (2007a)
Thermodynamic properties of Portland cement hydrates in the
system CaOAl2O3SiO2CaSO4CaCO3H2O. Cement and
Concrete Research 37(10): 13791410.
Matschei T, Lothenbach B and Glasser FP (2007b) The AFm
phase in Portland cement. Cement and Concrete Research
37(2): 118130.

Models for chloride ion bindings in hardened cement paste


using thermodynamic equilibrium calculations. Proceedings
of 2nd International RILEM Symposium on Advances in
Concrete Through Science and Engineering, Quebec City,
Canada.
Hugot-Le Goff A, Flis J, Boucherit N, Joiret S and Wilinski J

(1990) Use of Raman spectroscopy and rotating split ring


disk electrode for identification of surface layers on iron in
1M NaOH. Journal of the Electrochemical Society 137(9):
26842690.
Hussain SE, Rasheeduzzafar Al-Musallam A and Al-Gahtani AS

(1995) Factors affecting threshold chloride for reinforcement


corrosion in concrete. Cement and Concrete Research 25(7):
15431555.
Jackson DC (1996) Great American Bridges and Dams. Wiley,
New York, NY, USA.
Jensen OM, Coats AM and Glasser FP (1996) Chloride ingress
profiles measured by electron probe micro analysis. Cement
and Concrete Research 26(11): 16951705.
Jensen OM, Hansen PF, Coats AM and Glasser FP (1999) Chloride
ingress in cement paste and mortar. Cement and Concrete
Research 29(9): 14971504.
Jirickova M and Cerny R (2006) Chloride binding in building
materials. Journal of Building Physics 29(3): 189200.
Joiret S, Keddam M, Novoa XR et al. (2002) Use of EIS, ringdisk electrode, EQCM and Raman spectroscopy to study the
film of oxides formed on iron in 1 M NaOH. Cement and
Concrete Composites 24(1): 715.
Jones FE (1962) Hydration of calcium aluminates and ferrites.
Proceedings of 4th International Symposium on Chemistry
of Cements. National Bureau of Standards Monograph 43,
US Department of Commerce, Washington, DC,
pp. 205242.
Jones MR, Macphee DE, Chudek JA et al. (2003) Studies using
27Al MAS NMR of AF m and AF t phases and the formation
of Friedels salt. Cement and Concrete Research 33(2):
177182.
Jovancicevic V, Bockris JO, Carbajal JL, Zelenay P and Mizuno T

(1986) Adsorption and absorption of chloride ions on passive


iron systems.. Journal of the Electrochemical Society
133(11): 22192226.
Keister JC and Kasting GB (1986) Ionic mass transport through a
homogeneous membrane in the presence of a uniform electric
field. Journal of Membrane Science 29(2): 155167.
Kruger J (1976) Passivity and its Breakdown on Iron and
Iron-base Alloys. NACE, Houston, TX, USA.
Liang M, Jin W, Yang R and Huang N (2005) Predeterminate
model of corrosion rate of steel in concrete. Cement and
Concrete Research 35(9): 18271833.
Lin LF, Chao CY and Macdonald DD (1981) Point defect model
for anodic passive films 2. Chemical breakdown and pit
initiation. Journal of the Electrochemical Society 128(6):
11941198.
Lothenbach B, Le Saout G, Gallucci E and Scrivener K (2008)

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Miserque F, Huet B, Azou G, Bendjaballah D and LHostis V

(2006) X-ray photoelectron spectroscopy and electrochemical


studies of mild steel FeE500 passivation in concrete
simulated water. Journal De Physique IV(136): 8997.
Monteiro PJM, Gjorv OE and Mehta PK (1985) Microstructure of
the steelcement paste interface in the presence of chloride.
Cement and Concrete Research 15(5): 781784.
Montemor MF, Simoes AMP and Ferreira MGS (1998) Analytical
characterization of the passive film formed on steel in
solutions simulating the concrete interstitial electrolyte.
Corrosion 54(5): 347353.
Morris W, Vico A and Vazquez M (2004) Chloride induced
corrosion of reinforcing steel evaluated by concrete resistivity
measurements. Electrochimica Acta 49(25): 44474453.
Nemer MB, Xiong Y, Ismail AE and Jang J (2011) Solubility of
Fe2(OH)3Cl (pure-iron end-member of hibbingite) in NaCl
and Na2SO4 brines. Chemical Geology 280(12): 2632.
Nielsen EP, Herfort D and Geiker MR (2005) Binding of chloride
and alkalis in Portland cement systems. Cement and Concrete
Research 35(1): 117123.
Nightingale ER Jr (1958) Poised oxidationreduction systems: A
quantitative evaluation of redox poising capacity and its
relation to the feasibility of redox titrations. Analytical
Chemistry 30(2): 267272.
95

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Nordtest (1995) Concrete, Hardened: Accelerated Chloride

ionic diffusion coefficients on the basis of migration test


results. Materials and Structures 36(257): 156165.
Samson E, Marchand J, Snyder KA and Beaudoin JJ (2005)
Modeling ion and fluid transport in unsaturated cement
systems in isothermal conditions. Cement and Concrete
Research 35(1): 141153.
Sanchez M, Gregori J, Alonso C et al. (2007) Electrochemical
impedance spectroscopy for studying passive layers on steel
rebars immersed in alkaline solutions simulating concrete
pores. Electrochimica Acta 52(27): 76347641.
Sergi G, Yu SW and Page CL (1992) Diffusion of chloride and
hydroxyl ions in cementitious materials exposed to a saline
environment. Magazine of Concrete Research 44(158): 6369.
Shi M, Chen Z and Sun J (1999) Determination of chloride
diffusivity in concrete by AC impedance spectroscopy.
Cement and Concrete Research 29(7): 11111115.
Shi X, Xie N, Fortune K and Gong J (2012) Durability of steel
reinforced concrete in chloride environments: an overview.
Construction and Building Materials 30: 125138.
Stanish KD, Hooton RD and Thomas MDA (2000) Testing the
Chloride Penetration Resistance of Concrete: A Literature
Review. Department of Civil Engineering, University of
Toronto, Toronto, Ontario, Canada, FHWA contract
DTFH61-97-R-00022.
Streicher PE and Alexander MG (1995) A chloride conduction
test for concrete. Cement and Concrete Research 25(6):
12841294.
Suryavanshi AK and Narayan Swamy R (1996) Stability of
Friedels salt in carbonated concrete structural elements.
Cement and Concrete Research 26(5): 729741.
Suryavanshi AK, Scantlebury JD and Lyon SB (1995) The binding
of chloride ions by sulphate resistant Portland cement.
Cement and Concrete Research 25(3): 581592.
Suryavanshi AK, Scantlebury JD and Lyon SB (1996) Mechanism
of Friedels salt formation in cements rich in tri-calcium
aluminate. Cement and Concrete Research 26(5): 717727.
Tang L and Nilsson LO (1993) Chloride binding capacity and
binding isotherms of OPC pastes and mortars. Cement and
Concrete Research 23(2): 247253.
Tang L and Utgenannt P (2009) A field study of critical chloride
content in reinforced concrete with blended binder. Materials
and Corrosion 60(8): 617622.
Truc O, Ollivier JP and Carcasse`s M (2000) New way for
determining the chloride diffusion coefficient in concrete
from steady state migration test. Cement and Concrete
Research 30(2): 217226.
Veto D (2010) For the centenary of Hungarys first reinforced
concrete church in Rarosmulyad. Periodica Polytechnica
Architecture 41(1): 3541.
Whiting D (1981) Rapid measurement of the chloride
permeability of concrete. Public Roads 45(3): 101112.
Wowra O and Setzer MJ (2000) About the interaction of chloride
and hardened cement paste. Proceedings of 2nd International
RILEM Workshop on Testing and Modelling the Chloride
Ingress into Concrete, Paris, France, pp. 311.

Penetration. Nordtest, Espoo, Finland.


Nordtest (1997) Concrete, Mortar and Cement Based Repair
Materials: Chloride Diffusion Coefficient from Migration Cell
Experiments. Nordtest, Espoo, Finland.
Nordtest (1999) Concrete, Mortar and Cement-based Repair
Materials: Chloride Migration Coefficient from Non-steadystate Migration Experiments. Nordtest, Espoo, Finland.
Oranowska H and Szklarska-Smialowska Z (1981) An
electrochemical and ellipsometric investigation of surface
films grown on iron in saturated calcium hydroxide solutions
with or without chloride ions. Corrosion Science 21(11):
735747.
Page CL (2009) Initiation of chloride-induced corrosion of steel
in concrete: role of the interfacial zone. Materials and
Corrosion 60(8): 586592.
Page CL, Short NR and El Tarras A (1981) Diffusion of chloride
ions in hardened cement pastes. Cement and Concrete
Research 11(3): 395406.
Pourbaix M (1966) Atlas of Electrochemical Equilibria in
Aqueous Solutions. Pergamon Press, New York, NY, USA.
Poursaee A and Hansson CM (2007) Reinforcing steel passivation
in mortar and pore solution. Cement and Concrete Research
37(7): 11271133.
Rasheeduzzafar FH, Al-Saadoun SS and Al-Gahtani AS (1992)
Corrosion cracking in relation to bar diameter, cover, and
concrete quality. Journal of Materials in Civil Engineering
4(4): 327342.
Refait P and Genin J-R (1993) The oxidation of ferrous hydroxide
in chloride-containing aqueous media and Pourbaix diagrams
of green rust one. Corrosion Science 34(5): 797819.
Refait P and Genin J-R (1997) The mechanisms of oxidation of
ferrous hydroxychloride -Fe2(OH)3Cl in aqueous solution:
the formation of akaganeite vs goethite. Corrosion Science
39(3): 539553.
Refait P, Abdelmoula M and Genin J-R (1998) Mechanisms of
formation and structure of green rust one in aqueous
corrosion of iron in the presence of chloride ions. Corrosion
Science 40(9): 15471560.
Reguer S, Neff D, Bellot-Gurlet L and Dillmann P (2007)
Deterioration of iron archaeological artefacts: micro-Raman
investigation on Cl-containing corrosion products. Journal of
Raman Spectroscopy 38(4): 389397.
Remazeilles C and Refait P (2008) Formation, fast oxidation and
thermodynamic data of Fe(II) hydroxychlorides. Corrosion
Science 50(3): 856864.
Revie RW (2011) Uhligs Corrosion Handbook. Wiley, Hoboken,
NJ, USA.
Sagoe-Crentsil KK and Glasser FP (1993) Green rust, iron
solubility and the role of chloride in the corrosion of steel at
high pH. Cement and Concrete Research 23(4): 785791.
Sakashita S, Nakayama T and Ibaraki N (1999) Relation between
the diameter of steel wire and the corrosion rate in NaCl
aqueous solution. Corrosion Engineering 48(8): 514519.
Samson E, Marchand J and Snyder KA (2003) Calculation of
96

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

Advances in Cement Research


Volume 27 Issue 2

Chloride in cement
Galan and Glasser

Yonezawa T, Ashworth V and Procter RPM (1988) Pore solution

and Architecture, Central South University, Changsha,


China.
Yuan Q and Santhanam M (2013) Test methods for chloride
transport in concrete. In Performance of Cement-Based
Materials in Aggressive Aqueous Environments (Alexander
M, Bertron A and Belie N (eds)). Springer, Dordrecht, The
Netherlands.

composition and chloride effects on the corrosion of steel in


concrete. Corrosion 44(7): 489499.
Yuan Q (2009) Fundamental Studies on Test Methods for the
Transport of Chloride Ions in Cementitious Materials. PhD
thesis, Department of Structural Engineering, Ghent
University, Ghent, Belgium; School of Civil Engineering

WHAT DO YOU THINK?

To discuss this paper, please submit up to 500 words to


the editor at journals@ice.org.uk. Your contribution will
be forwarded to the author(s) for a reply and, if
considered appropriate by the editorial panel, will be
published as a discussion in a future issue of the journal.

Downloaded by [] on [01/12/16]. Copyright ICE Publishing, all rights reserved.

97

Vous aimerez peut-être aussi