You are on page 1of 22

Cement and Concrete Research 173 (2023) 107287

Contents lists available at ScienceDirect

Cement and Concrete Research


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

Chloride profiles – What do they tell us and how should they be used?
Klaartje De Weerdt a, *, William Wilson b, Alisa Machner c, Fabien Georget d
a
Department of Structural Engineering, Norwegian University of Science and Technology (NTNU), Norway
b
Department of Civil and Building Engineering, Université de Sherbrooke, Sherbrooke, QC, Canada
c
TUM School of Engineering and Design, Technical University of Munich (TUM), Germany
d
Institute of Building Materials Research, RWTH Aachen University, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Chloride profiles are used to select concrete compositions for new reinforced concrete structures, evaluate the
Chloride condition of existing structures, and to predict their remaining service life, all with regard to chloride-induced
Concrete reinforcement corrosion. They are invaluable tools for improving the sustainability of our infrastructure. We
Corrosion
examine what chloride profiles tell us and how they should be used. Through a series of case studies, we establish
Service life modelling
Diffusion
a list of crucial research questions that provide a direction for future research in this field: What is the contri­
bution of chloride binding to chloride transport? What is the path for chloride transport in concrete? How are our
findings on chloride profiles relevant for reinforcement corrosion? We then discuss the implications of the an­
swers to these questions for engineering and mechanistic modelling practices.

1. Introduction development of new concrete compositions or binders and during the


service life of reinforced concrete structures exposed to external chlo­
Chloride-induced corrosion of steel reinforcement is one of the main rides, such as sea water or de-icing salts. When developing new concrete
degradation mechanisms for reinforced concrete structures worldwide compositions or binders, chloride profiles from for example accelerated
and is therefore a major threat to our global infrastructure [1]. Ever so laboratory tests are used to compare their chloride ingress resistance.
important, the mechanisms of chloride ingress are still not fully During the service life, chloride profiles are taken to determine the
understood. condition of a structure and plan potential repairs or determine the
Although chloride ingress determines the service life of many rein­ remaining service life.
forced concrete structures, their design and maintenance rely largely on Fig. 1 shows a typical chloride profile, giving the total chloride
experience-based prescriptive rules. There is tremendous pressure to content in mass % of dry concrete as a function of depth from the
develop new low-CO2-cements with a variety of compositions and exposed surface. Such profiles are determined by obtaining ground
therefore potentially variable resistance to chloride ingress with regard concrete powder at different depths from the exposed surface and ana­
to chloride diffusivity and chemical interaction with the hydrated lysing its chloride content. There are a number of international stan­
binders. This intensifies the need for models that can predict the dura­ dards describing this process, e.g., NT BUILD 443 [3], ASTM C1556-11a
bility of these new binders. For such models to work accurately, the [4], and EN 12390–11 [5]. Amongst other things, profiles show the
fundamental mechanisms of chloride ingress and chloride-concrete extent to which chlorides have penetrated the concrete at a certain point
interaction need to be known, quantified, and taken into account in in time. The common purpose of chloride profiles is to get an insight into
the models. Moreover, we are now designing structures for much longer to how high the chloride levels are at the reinforcement as a function of
service lives and extending the service lives of existing structures. As we time, and whether these levels can cause corrosion.
need to act fast, we can no longer rely on long-term experimental testing. The left-hand sketch in Fig. 1 illustrates typical features of a chloride
Instead, we urgently need adequate fundamental models based on profile, where a) the height reflects the amount of chloride present in the
experimental input that can be measured relatively rapidly. concrete at a specific depth; b) the width reflects how far the chlorides
Our understanding of chloride profiles will play a pivotal role in this have penetrated into the concrete; and c) the surface phenomena relate
development. Chloride profiles are a key engineering tool during the to the peaking of the profile near the surface.

* Corresponding author.
E-mail address: klaartje.d.weerdt@ntnu.no (K. De Weerdt).

https://doi.org/10.1016/j.cemconres.2023.107287
Received 10 March 2023; Received in revised form 10 July 2023; Accepted 24 July 2023
Available online 14 August 2023
0008-8846/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

These features are affected by a number of parameters. This paper 3. How are our findings on chloride profiles relevant for reinforcement
focuses on the impact of two parameters on each of these features: 1) the corrosion?
exposure conditions, and 2) the binder type. We start with a general
introduction to chloride profiles and then demonstrate the impact of At the end of the paper, we summarize how our findings on chloride
these parameters by referring to specific case studies. To guide the profiles feed into these questions.
reader through the complexity of chloride profiles, we begin with
simplified descriptions and move progressively to more advanced 2. Understanding chloride profiles
characterization and modelling.
This paper only considers chloride ingress from an external source, e. This section focuses on case studies aimed at understanding chloride
g., sea water or de-icing salts, and it considers the components of the profiles. We start by looking at the basics of chloride profiles for readers
unexposed concrete to be “chloride-free”. We will focus mainly on who have not worked with chloride profiles before. In Section 2.2 we
submerged exposure and will not go into drying and wetting or salt look into the impact of exposure on chloride profiles using two experi­
spray exposures. We do this because even in the simple case of mental set-ups. Finally in Section 2.3, we discuss the impact of the
controlled standard bulk diffusion experiments, there are still crucial binder type.
knowledge gaps that need to be filled in before we can widen our focus.
The chloride content in the vicinity of the rebar is key with respect to 2.1. Chloride profile basics
corrosion. The critical chloride content is a crucial parameter when it
comes to service life modelling. The critical chloride content is an even Concrete is a composite material consisting of aggregates and sand
more complex topic than chloride ingress, though it will not be the focus held together by cement paste. With respect to chloride resistance,
of this paper. We will, however, indicate how the findings presented cement paste is the “active part”, and both aggregates and sand can be
here would also impact the critical chloride content. considered “inert”. The total chloride content or “height” of the chloride
This is not a review paper. There are very comprehensive textbooks profile is therefore largely determined by the binding capacity and
on chloride ingress available, such as Shi et al. [6], Tang et al. [7], and amount of the cement paste in the concrete. Similarly, the chloride
Bertolini et al. [8]. Excellent reviews on chloride binding are also ingress depth or “width” of the profile is a measure of the chloride
available [9–11]. penetration resistance of the cement paste, which relates to its porosity
On the basis of simple calculations and a selection of case studies, the and transport properties. Finally, we need to take into account the
aim of this paper is to explain our current understanding of chloride heterogeneity of the concrete, especially near the surface.
profiles, to highlight the research gaps we have encountered, and to
provide insights into how these profiles can be used for engineering 2.1.1. Chloride binding and the height of profiles
purposes as well as for fundamental research aimed at elucidating the Chloride profiles represent the total chloride content, comprising
mechanisms involved in chloride ingress. both free and bound chlorides. When hydrated cement paste comes into
The paper is divided in two main parts: the first focuses on case contact with a chloride-containing solution, some of the free chlorides
studies and the second on modelling. In the first part (Section 2), we start from the liquid phase will become bound. This binding can occur
with a basic explanation of what a chloride profile is followed by a set of chemically or physically. Chlorides are chemically bound by their
case studies focusing on the impact of exposure and binder type on the incorporation into the interlayer of AFm phases [12–17]. In this paper,
shape of chloride profiles. In the modelling part (Section 3), we discuss we will refer to these phases as Friedel's salt solid solutions (FSss) or
the use of chloride profiles as input or output for both engineering and more generally, chloride containing AFm (Cl-AFm). Chlorides can also
mechanistic models. accumulate in the diffuse layer of the C-A-S-H. This accumulation is
The paper evolves around the following reoccurring research usually referred to as physical binding [6,18–23]. In this paper, the
questions: abbreviation C-A-S-H is used to refer to calcium aluminosilicate hydrates
(instead of the more traditionally used C-S-H) to take into account the
1. What is the contribution of chloride binding to chloride transport? aluminate content typically observed in hydrated cement paste [24,25].
2. What is the path for chloride transport in concrete? To understand chloride profiles, we need to know how much of the total

Fig. 1. Sketch showing the total chloride profiles from a concrete element submerged in sea water at Solsvik field station (Norway) for 16 years. The concrete was
made with a Portland composite cement containing 4 % silica fume and 19 % fly ash [2]. On the left, typical features of the chloride profile are marked with letters.
On the right, the two main causes that affect the shape of chloride profiles are marked with numbers. These main causes will be addressed in this paper by means of
case studies.

2
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

chloride is actually bound, and in which phases it is bound. can be slightly overestimated due to the loss of some solid phases when
To illustrate how much of the total amount of chloride can actually drying at 105 ◦ C. In addition, both the stoichiometry of Cl-AFm and the
be bound, we present here a preliminary estimate of the potential composition and chloride content of C-A-S-H can vary greatly, which
contribution of free and bound chlorides to the total chloride content. affects their contribution to the total chloride content.
Note that we consider AFm and C-A-S-H to be the main phases respon­ Although there are some uncertainties in the example presented, a
sible for the chloride binding of hydrated cement paste. Other phases, similar picture has been obtained in other studies (see Fig. 8 in Section
such as calcium hydroxychloride hydrates [26,27] are not considered 2.2 and Fig. 11 in Section 2.3). The general idea of Fig. 2 should
here, and adsorption on surfaces other than C-A-S-H, e.g., portlandite therefore be kept in mind. We discuss this in more detail with respect to
[28], is considered negligible. the exposure conditions (Section 2.2) and the binder type (Section 2.3).
For the estimate, we used the total chloride profile of a PC-FA con­ Finally, it is important to note that the cement content in concrete (e.
crete exposed to chlorides for 16 years submerged in sea water [2] as g., in kg/m3) has a large impact on the potential amount of chloride
shown in Fig. 1. We focused on the chloride content measured in sec­ bound in the concrete. A concrete with a low cement content of, e.g.,
tions between 5 and 10 mm in depth where the total chloride content is 220 kg/m3, has only half the theoretical potential binding capacity of a
about 0.5 wt% of dry concrete. We focused on this depth as we wanted to more cement-rich concrete with 440 kg/m3 cement, and this is without
look at a relatively higher chloride content near the exposed surface, and taking into account the shrinkage or potential differences in water-to-
at the same time avoid the outermost sections (0–5 mm) due to the in­ binder (w/b) ratios. So, when the heights of different chloride profiles
crease in paste content in the concrete close to the exposed surface, are compared, the cement content of the concrete is a crucial parameter.
which is caused by the wall effect and which is further described in What can we learn from these findings with respect to the first key
Section 2.1.3. question raised in the introduction - What is the contribution of chloride
The supplementary data to this paper gives a detailed description of binding to chloride transport? We have seen that a major part of the total
how we calculated the amount of free chloride and the amount of chloride content shown in typical chloride profiles is bound in hydrates,
chloride bound in Friedel's salt and C-A-S-H. The results are shown in whereas only a minor share is found in the pore solution. Chloride
Fig. 2. As input data, we used the suction porosity of the concrete of 15 binding is therefore important for our understanding of chloride pro­
vol% [2], the phase assemblage calculated by thermodynamic modelling files. Based on the information presented here, we cannot yet say any­
based on the binder composition [2], and an estimated Cl/Ca molar- thing on the impact of chloride binding on chloride ingress.
ratio for the C-A-S-H of 0.05 [2,29,30].
What this simplified calculation demonstrates is that a large share of 2.1.2. Chloride ingress and the depth of profiles
the total chloride content is actually not present as free chlorides, but The chloride ingress resistance of concrete is determined by its pore
instead is taken up by solids (Fig. 2). Therefore, to understand chloride volume and, even more importantly, the pore structure of the concrete
profiles, we need to understand the interactions between the chlorides or mortar, which in turn means the pore structure of the cement paste
and the hydration phases. and its interfacial transition zone (ITZ) because the aggregates can be
It should be noted that estimating the free chlorides is not easy considered inert. In principle, the chloride ingress resistance for a spe­
because it relies on an estimation of the porosity, which we assume to be cific binder can be governed by adjusting the w/b ratio of the concrete.
filled with a solution with a chloride concentration similar to sea water This is reflected in regulations and norms when they prescribe a
(i.e., 0.5 mol/L). The chloride concentration in the pore solution at a maximum allowed w/b ratio or a minimum compressive strength range
depth of 10 mm is assumed to be close to the one in the pore solution at for a given exposure condition, e.g., EN 1992 [32], EN 206 [33], and
the exposed surface and to decrease when moving deeper into the CSA A23.1:19 [34]. According to the Power's theory [35,36], a plain
concrete. Thus, assuming a similar concentration as in the exposure Portland cement would theoretically require 42 wt% of water to allow
solution at 10 mm depth is a simplification and will lead to an over­ full reaction, assuming 23 % of non-evaporable water content and 19 %
estimation of the free chloride content. The value of the porosity is of gel water. With a w/b ratio of 0.42 therefore, the cement matrix
highly dependent on the method used to measure it [31]. In this case, the would be filled with hydration products and gel pores, although
samples were dried at 105 ◦ C to determine the pore volume, which we contraction pores would form due to chemical shrinkage. With higher
assumed to be filled with the same chloride concentration as the expo­ w/b ratios, there is excess water, which results in capillary pores which
sure solution. The pore volume (and with it the amount of free chloride) can lead to a considerable decrease in the ingress resistance of the paste
if these capillary pores are connected [37–39]. This is why most w/b
ratios for durable chloride-exposed concrete are in the range of
0.45–0.40 or lower, even though cement hydration in concrete is
generally not complete. For more severe exposure conditions, e.g.,
splash or tidal zone vs. airborne chlorides, or where a longer service life
is required, a lower water-to-binder ratio is usually prescribed. Fig. 3
(left) illustrates the impact of varying the w/b ratio in the lower range of
w/b ratios. However, we should keep in mind that a lower w/b ratio
typically results in a higher cement content (in kg/m3) of the concrete in
order to maintain a similar paste volume in the concrete, thereby leading
to a larger initial environmental footprint for the concrete. Thus, from a
life cycle assessment point of view there will be a balance between the
w/b ratio and the cement content or, in other words, between the service
life and the built-in emissions.
Fig. 3 (right) shows how the binder plays a dominant role in the
chloride ingress resistance of concrete. For the same w/b ratio and
binder content, an increase in the silica fume content from 5 to 10 % has
a similar or even better effect than reducing the w/b ratio from 0.40 to
Fig. 2. Left: the predicted phase assemblage of the cement paste in the PC-FA
concrete (w/b = 0.42) exposed for 16 years submerged in sea water [2]. Right: 0.25 (Fig. 3, left). Section 2.3 focuses on the impact of the binder type
the distribution of the total chloride content between free chlorides in the pore and discusses how certain composite cements form more favourable
solution, chemically bound in the Friedel's salt, and physically bound by the C- pore structures with higher resistance to chloride ingress and can
A-S-H. The detailed calculations are given in the supplementary data. therefore either be used at higher water-to-binder ratios than for

3
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Fig. 3. Sketches based on data from Tang [40] for chloride profiles from concrete exposed submerged in a marine environment for 5 years. Left: binder consisting of
95 % PC and 5 % silica fume (SF) with varying w/b ratios, i.e.,0.25, 0.30, and 0.40 (H5, H1 and H4 in the reference; with a binder content of 550, 500 and 420 kg/m3
concrete). Right: binders consisting of 95 % PC + 5 % SF and 90 % PC + 10 % SF both with a w/b ratio of 0.30 and a binder content of 500 kg/m3 concrete (H1 and
H2 in the reference). Note that the profile is normalized to the binder content using the amount of calcium in the powdered concrete and the calcium content of
the binder.

example a concrete with plain Portland cement or be used to provide a Moreover, it should be noted that normalization by the calcium content
longer service life. only works with aggregates that do not contain calcium, e.g., granite
How do these observations feed into the second key question raised aggregate [43]. An alternative method that allows normalization of the
in the introduction - What is the path for chloride ingress in concrete? So chloride content to the paste content would be μXRF-scanning, where
far, we have seen that the cement paste part and the ITZ of the concrete applying filters on elemental maps makes it possible to distinguish be­
determine its chloride ingress resistance and that the w/b ratio and tween aggregates and paste, and thereby plot the chloride ingress
binder type affect the pore volume and, more importantly, the pore normalized to the paste content.
structure or pore connectivity in the cement paste and potentially also The type of surface also seems to affect the chloride ingress depth (x)
the ITZ. through another competing mechanism [42]. Concretes with cast sur­
faces show a tendency to have a slightly improved chloride ingress
2.1.3. Units of chloride profiles and surface phenomena resistance compared to sawn surfaces after 5 years of exposure (see
When we discuss and compare chloride profiles, we should pay Fig. 4). Variations in x are below 5 mm for 0.1 wt% Cl per concrete. This
attention to the units used. Chloride content can be expressed as wt% of has been related to the introduction of defects at the surface during
the concrete or normalized to the binder content and expressed as wt% sawing, which should also be considered in combination with the
of the binder. Both approaches have their drawbacks as discussed in this
section.
When we study concrete cores extracted in the field, the chloride-
exposed surface is typically a cast surface, which means there is a
“wall effect”. At and near a cast surface, a higher paste-to-aggregate
ratio is found than in the bulk concrete, because the concrete has been
cast against a mould or formwork or finished by trowelling. This wall
effect causes fluctuations in the aggregate content, which gradually level
out with increasing distance from the cast surface [41]. Because ag­
gregates can be considered inert (see Section 2.1.1), an increase in the
cement paste content close to the surface leads to a local increase in the
potential binding of chlorides when the chloride content is expressed as
wt% of concrete.
De Weerdt et al. [42] investigated the effect of cast and sawn con­
crete surfaces on the chloride ingress resistance of concretes (daggregate
max = 8 mm) prepared with two different Portland cements and a
Portland fly ash cement (20 wt% fly ash) and exposed for up to 5 years in
a marine splash zone. The total chloride content in the outermost section
(0-1 mm) after 5 years of exposure was >30 % higher for the concretes
with a cast surface than for the concretes with a sawn surface. This was
attributed to the higher paste content in this section for the cast surfaces.
We could take this wall effect into account by normalizing the total
chloride content to the binder or paste content. Tang [40] did this by
analysing both the chloride content and the calcium content of the
concrete powder and using the calcium content as a measure for the
Fig. 4. Chloride profiles after 5 years in the tidal zone for two different con­
paste content. However, this is not very reliable for the outermost sec­
crete compositions both with a w/b ratio of 0.45. PC is plain Portland cement
tions of long-term exposed samples because these sections are also and PC-FA is a blended cement containing about 20 wt% fly ash. The dashed
affected by calcium leaching, as pointed out by Fjendbo et al. [43]. line corresponds to a sawn surface and the solid line to a cast surface [42].

4
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

difference in paste content. 2.2.1.1. Phase changes and chloride binding. Sea water is a multi-ion
solution. Table 1 shows the main elements present in sea water from
2.1.4. Critical chloride content the Trondheim fjord, which has a very similar composition to the
The critical chloride content (Ccrit) or chloride threshold value can be Atlantic Ocean. The presence of these different ions leads to a typical
defined as the chloride content required for the depassivation of the steel elemental zonation in the outermost millimetres of marine-exposed
[44]. It is a crucial parameter for service life modelling, both when used concrete [2,29,30]. Elemental zonation is the formation of zones or
for the design of new structures and for the assessment of existing ones. layers parallel to the surface which are enriched in specific ions (see
In such models, the chloride ingress is typically linked to the chloride Fig. 5). From the surface inwards, we observe first a calcite and brucite
content at the reinforcement and thereby to the critical chloride content. crust on the surface and in cracks, followed by a magnesium- and
Some of our findings related to chloride profiles are therefore also carbonate-enriched zone where magnesium silicate hydrate (M-S-H) and
relevant for the critical chloride concept, though the initiation of calcite can be found; next comes a sulphate-rich zone in which ettringite
chloride-induced corrosion entails far more than a chloride content or is typically observed in voids (e.g., air voids, ITZ, and Hadley grains);
concentration at the reinforcement because it strongly depends on the and finally there is a chloride-enriched zone in which Cl-AFm and
conditions at the steel-concrete interface as pointed out in several re­ chlorides in the C-A-S-H are found [2,29,30].
views [45,46]. Fig. 6 shows the predicted changes in the phase assemblage of
Like chloride profiles, the critical chloride content can be expressed Portland cement paste upon exposure to increasing amounts of NaCl
in various ways, as summarized in a review by Angst et al. [44]. Ccrit can solution (left) and sea water (right) [49]. The phases predicted for sea
be expressed as a total chloride content relative to the concrete mass, or water exposure (Fig. 6 right) agree well with the zonation sketched in
as the total chloride content relative to the binder mass, where the latter Fig. 5. The exposure to increasing amounts of NaCl solution leads to a
allows us to compare critical chloride contents in concrete with different decrease in the total volume of the hydration phases, whereas the
binder contents. changes in the phase assemblage upon exposure to sea water lead to an
From a more mechanistic point of view, the free chlorides in the pore increase in the volume of the solids in the cement paste. The volume
solution could be considered responsible for the corrosion onset, increase is caused by the formation of ettringite due to the reaction of
assuming that the bound chlorides are immobilized and do not the cement paste with sulphates from the sea water. In addition, the
contribute to corrosion initiation. The free chlorides can be expressed surface can be covered with a calcite and/or brucite crust.
relative to the binder or the concrete mass, or they can be expressed as a The elemental zonation at the surface of marine exposed concrete
concentration in the pore solution. To take into account the passivating does not affect the structural integrity of marine structures as it only
effect of the alkaline pore solution (high pH) on the steel reinforcement, affects the outer millimetre of the concrete [2,29,30]. Depending on the
the critical chloride content is commonly reported as the ratio between degree of abrasion the surface is exposed to, i.e., currents and particles,
the chloride ions and the hydroxyl concentration or activity, i.e., [Cl− ]/ parts of the surface can be abraded away. The magnesium rich phases
[OH− ] [44]. This has been regarded as one of the most relevant defi­ are typically observed in cracks and cavities rather than on the surface
nitions for Ccrit. However, determining the pore solution composition in when exposed to an abrasive environment, which could indicate that
the vicinity of the reinforcement is very difficult in practice. they form a slightly weaker zone. The sulphate enriched zone in the
This brings us to the third key question raised in the introduction - So investigated concretes in [2,29,30] did not show signs of deterioration
how are our findings on chloride profiles relevant for reinforcement corro­ associated with sulphate attack.
sion? As for chloride profiles, when we compare critical chloride con­
tents for different concretes, we need to normalize them either to the 2.2.1.2. Chloride ingress. The experimental and modelling results
cement content or to the hydration product content. We have shown that shown above for sea water indicate a filling of pores, so less chloride
a large part of the total chloride content in concrete is adsorbed or bound ingress might be expected in the sea water-exposed mortars than in the
and might not contribute to the initiation of corrosion. More knowledge NaCl-exposed samples. To investigate how the chloride ingress profiles
about chloride binding and its reversibility therefore seems crucial for are affected by these phase changes in the outermost sections, Portland
our understanding of the critical chloride content. This will be discussed cement mortar samples (w/b = 0.40) were exposed by unidirectional
further in Section 2.3. diffusion for 180 days to NaCl solution and sea water with the same
chloride concentration (approx. 0.55 mol/L), which was renewed every
2.2. Exposure solution week [49]. Fig. 7 shows the chloride profiles after 21, 90 and 180 days of
exposure to each solution. The portlandite (CH) profiles were deter­
This section looks at the impact of the exposure solution on chloride mined using thermogravimetric analysis on the same profile ground
profiles. We present two studies: one where the impact of other ions in sections as used for the chloride analysis. The loss of portlandite was
sea water besides chloride was investigated, and the other where the used as a measure of leaching because Portlandite is one of the first
impact of the alkalinity of the exposure solution on the chloride profile hydration phases to dissolve upon leaching (see Fig. 6).
was investigated. Taken together, these two studies explain the typical Contrary to what might be expected, chlorides actually penetrated
shape of chloride profiles in submerged marine-exposed concrete. slightly further for sea water exposure than for NaCl exposure; the
These studies were carried out on mortar samples with sawn sur­ chloride profiles for NaCl exposure lie consistently slightly to the left of
faces, as is common in accelerated laboratory testing. By investigating the profiles for sea water exposure. In addition, slightly higher
sawn surfaces, we eliminate the “wall effect” observed for cast surfaces maximum chloride contents were measured for sea water exposure than
as described in Section 2.1.3. for NaCl exposure.
This means that the filling of (seemingly coarse) porosity by phases
2.2.1. Multi-ion sea water formed due to the additional ions in sea water, such as ettringite, brucite
In the first set-up, chloride profiles in Portland cement mortars and calcite, does not necessarily limit chloride ingress. This is relevant
exposed to sea water were compared with the chloride profiles of a input for the second research question - What is the path for chloride
similar set of mortars exposed to a NaCl solution with the same chloride transport in concrete? It indicates that a reduction in porosity is not suf­
concentration as in the sea water (see Fig. 7). The aim was to investigate ficient to reduce chloride transport. This makes it even more important
whether other ions present in sea water besides chloride would change to gain a better understanding of the pore network, and not just the
the shape of the chloride profile. capillary pores, but also the porosity within the hydrates (i.e., C-A-S-H).
These results also show that NaCl solution can be used to simulate

5
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Table 1
Concentrations of the main elements in sea water from the Trondheim fjord [29], standard sea water [48], and the pore solution of Portland cement (PC) paste
containing 5 % limestone hydrated sealed for 140 days [47], all in [mmol/L].
Ca Cl K Mg Na S C pH Ref.

Sea water Trondheim Fjord 8.8 548 8.9 47 411 26.9 0.2 n.d. [29]
Standard sea water 10.5 566 10.6 55 485 29.3 1.9 7.7 [48]
Pore solution PC paste 1.2 n.d. 532 n.d. 279 27.6 n.d. 13.7 [47]

n.d. = not determined.

related to the difference in their intrinsic or ion-specific diffusion coef­


ficient [51] as well as charged-surface effects [52,53]. The slightly
higher total chloride content and deeper ingress measured upon expo­
sure to sea water than in NaCl exposure might be due to the slightly
higher leaching potential of sea water [54], which is supported by the
slightly lower levels of portlandite near the surface in the case of sea
water, as shown in Fig. 7 (right).
All in all, we can conclude that the predicted increase in the volume
of reaction products near the exposed surface upon sea water exposure
does not seem to have a major influence on the chloride ingress.
Nevertheless, chloride ingress seems to go hand in hand with leaching of
hydroxyl ions. This gave Machner et al. [50,55] the idea of the second
experimental set-up, which is presented below.

2.2.2. Effect of leaching


We came to the hypothesis that leaching has an important impact on
the appearance of total chloride profiles in concrete. This hypothesis is
supported not only by the dominant impact of the pH and the calcium
concentration on chloride binding demonstrated by Hemstad et al. [56],
but also by the fact that exposure to sea water or NaCl does not seem to
cause sodium ions to follow the chlorides into the concrete [2,29,30],
which indicates that chloride ingress has to go hand in hand with hy­
droxyl leaching to maintain electroneutrality.
Fig. 5. Sketch illustrating zonation in sea water-exposed concrete showing the To demonstrate the impact of leaching on the shape of total chloride
typical phases observed in the respective zones [2,29,30]. profiles, mortars were exposed to unidirectional bulk diffusion in a NaCl
solution and in a NaCl solution with KOH (150 mmol/L), referred to here
chloride ingress from sea water in laboratory experiments, as recom­ as NaCl+KOH, both with the same chloride concentration (0.55 mol/L).
mended in current standards, for example EN 12390–11 [5]. The experiments were performed on two sets of mortars prepared with
When we compare the portlandite profiles with the chloride profiles two different composite cements. The first set of mortars was prepared
in Fig. 7, we observe an inverse relationship between the portlandite with a CEM II/C-M(S-LL) cement containing 50 wt% Portland cement,
content and the chloride content. This indicates that chloride ingress 40 wt% ground granulated blast-furnace slag, and 10 wt% limestone
goes hand in hand with hydroxyl and calcium leaching and therefore the filler [55], and the second set with a CEM VI (S–V) cement containing
dissolution of portlandite. This also agrees with the observation that 47 wt% Portland cement, 43 wt% ground granulated blast furnace slag,
sodium does not follow chloride equimolarly into the concrete upon and 10 wt% fly ash [50]. Similar results were found in both sets, but in
chloride ingress [2,29,30,49,50]. The difference in the observed this paper, we focus on this second set of mortars. The reason KOH was
mobility of chloride and hydroxyl ions compared to sodium ions is added to one of the exposure solutions was to limit leaching. The extent

Fig. 6. Predicted volume of phases in Portland cement paste exposed to increasing amounts of NaCl solution (left) or sea water (right) in cm3/100 g cement [49].

6
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

1.2 5

CH [wt.% of dry mortar]


Cl [wt.% of dry mortar]
4
0.9

3
0.6
N_21d
2 S_21d
N_90d
0.3 S_90d
1
N_180d
S_180d
0.0 0
0 5 10 15 0 5 10 15
depth [mm] depth [mm]
Fig. 7. Total chloride profiles (left) and portlandite (CH) profiles (right) of Portland cement mortar exposed in the laboratory to sea water (S) or 0.55 mol/L NaCl
solution (N) for 21, 90 and 180 days, expressed as wt% of mortar dried at 105 ◦ C [49].

of the leaching was investigated by measuring the portlandite content in proportions were calculated based on the amount of C-A-S-H and AFm
the samples as a function of depth from the exposed surface. predicted by thermodynamic modelling and their actual chloride con­
tent (Cl/Si and Cl/Al) as determined by SEM-EDS [50]. We can see that
2.2.2.1. Phase changes and chloride binding. Fig. 8 shows the chloride leaching in the NaCl-exposed sample leads to more chloride binding in
(Cl) and portlandite (CH) profiles of the mortars exposed for 180 days to both the C-A-S-H and AFm than in the sample exposed to NaCl+KOH.
NaCl and NaCl+KOH solutions [50]. The mortar exposed to NaCl+KOH The impact of progressive leaching on chloride binding in cement
had a reduced portlandite content only in the outermost sections (0–3 paste was also investigated by Machner et al. [57] and Hemstad et al.
mm), whereas the NaCl-exposed mortar showed a decrease in por­ [56], who confirmed that moderate leaching leads to an increase in
tlandite content up to depths of 10 mm, which confirmed that the KOH chloride binding due to an enhanced chemical and physical binding,
addition was able to limit calcium leaching from the mortar into the whereas severe leaching (pH <12) leads to a decomposition of the
exposure solution. binding phases and a drop in the binding capacity of the cement paste, as
Moreover, the chloride profiles are dramatically different for the two sketched in Fig. 9. Hemstad et al. [56] related the increase in chloride
exposure conditions. Upon enhanced leaching due to exposure to NaCl, binding when the pH drops from 13 to about 12 to an increase in the Cl/
the total chloride content is three times higher than in the case of Ca ratio of the AFm solid solution and an increase in the amount of AFm
exposure to NaCl+KOH. Since in both cases the mortar is exposed to the formed. When the pH drops further, the amount of AFm decreases due to
same chloride loading, the difference in the maximum total chloride dissolution, which leads to a reduction in binding. But changes in the
content of the exposed samples demonstrates that the chloride load is AFm alone only explain part of the pH dependence of the binding
not the only factor involved in the results. (Fig. 9), which indicates that the C-A-S-H has a similar pH-dependent
To shed light on the changes in chloride binding between the NaCl behaviour. We calculated the pH dependent chloride uptake in C-A-S-
and NaCl+KOH exposure conditions, the changes in the phase assem­ H by assuming that the difference between the total amount of bound
blage of the mortars were investigated. Fig. 8 shows the distribution of chloride and the chlorides bound in AFm, are chlorides physically bound
the chloride between the pore solution, Friedel's salt, and C-A-S-H. These in C-A-S-H (Fig. 9). The increase in chloride adsorption on C-A-S-H with

Fig. 8. Chloride and portlandite profiles for mortars exposed for 180 days to NaCl solution (left) and NaCl+KOH solution (right). The total chloride content is
distributed between the pore solution, Friedel's salt, and C-A-S-H based on thermodynamic modelling and SEM-EDS to determine the actual chloride content in the
Friedel's salt and C-A-S-H [50]. The three zones discussed in the text are indicated with Roman numerals at the top of the graphs.

7
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Fig. 9. Left: sketch illustrating the pH dependency of chloride binding (total Cl and Cl in AFm and C-A-S-H) [56,62]; Right: sketch illustrating how leaching can result
in a linear chloride binding isotherm [61].

decreasing pH has been attributed to the increasing calcium concen­ and number of ions leaving the sample must equal the charge and
tration in the pore solution, which results in more calcium in the Stern number of ions entering it. Limiting the leaching of hydroxyl anions
layer of the C-A-S-H surface rendering it more positive, allowing the from the sample [50,55] also limits the ingress of chloride anions into
accumulation of more negatively charged chlorides in the diffuse layer the sample. So, to fully understand the impact of leaching and ion bal­
of the C-A-S-H [6,18–23,58–60]. At around pH 12 the chloride in C-A-S- ances in the sample on the chloride ingress rate and depth, we need
H starts to decrease as the system's portlandite buffer is consumed, multi-ion physical transport models that take leaching into account. We
calcium concentrations decrease, and C-A-S-H starts to decalcify discuss these in the modelling section (see Section 3.2.2).
[56,62]. Note that the contributions of C-A-S-H and AFm to this pH How do these findings relate to the second research question - What is
dependent binding behaviour will be dependent on the phase assem­ the path of the chloride transport in concrete? They indicate that the path of
blage and thereby the cement type (Section 2.3). the chlorides is influenced by the mobility of other ions and the need of
Georget et al. [61] proposed engineering models to describe a pH- the system to maintain charge balance. Chlorides move into the con­
dependent binding isotherm (see Fig. 9, right). Binding isotherms are crete, but cations (e.g., Na+) do not equimolarly follow it into the con­
typically described as Freundlich or Langmuir functions which are valid crete. Instead, we observe that chloride ingress goes hand in hand with
for a specific pH of the exposure solution. By taking leaching into ac­ leaching. The charge balance is maintained by hydroxyl ions (OH− )
count, they can turn into a near-linear isotherm. For example, the leaving the concrete as chloride ions enter (Cl− ).
outermost section of a marine-exposed concrete which is exposed to the
highest chloride level (in this case 0.5 M) would also be exposed to the 2.2.2.3. Peaking behaviour near the surface. Chloride profiles of long-
highest level of leaching and therefore the lowest pH. The extent of term chloride-exposed samples generally exhibit a low chloride con­
leaching decreases with increasing depth below the surface, so the pH tent at the surface followed by the maximum chloride content (a peak or
increases and the chloride content in the concrete decreases. The dashed even a plateau). With increasing depth below the exposed surface, a
arrows in Fig. 9 (right) illustrate the movement to higher pH isotherms gradual decrease in chloride content is observed. We refer to this as
with increasing depth in the concrete resulting in an approximately “peaking behaviour”. It can be seen in all the chloride profiles shown so
linear isotherm. This has an important impact when chloride profiles are far (see Figs. 1, 7, and 8).
used for chloride ingress modelling over time, see Section 3.1. For chloride profiles determined on concrete with cast outer surfaces
To put this is in perspective of our third research question - How are (e.g., Fig. 1), this peaking behaviour contradicts what would be ex­
our findings on chloride profiles relevant for reinforcement corrosion? We pected, namely an increase in the total chloride content when
have shown that chloride ingress goes hand in hand with hydroxyl approaching the exposed surface due to an increase in the paste content
leaching. As described in Section 2.1.4, one of the preferred ways of as a result of the wall effect (see Section 2.1.3).
expressing the critical chloride content is the [Cl− ]/[OH− ] ratio. This For concretes exposed to sea water or road salt spraying, or to the
parameter is affected both by the ingress of chlorides, the numerator, splash or tidal zone, the peaking behaviour has been attributed to drying
and by leaching, which decreases the denominator. From a corrosion and wetting, which leads to an accumulation of chlorides and other ions
initiation point of view, this means we must not only take chloride a few mm below the exposed surface [63]. This is a plausible mechanism
ingress into account, but also leaching. for these kinds of exposures, but peaking behaviour is also generally
observed for permanently submerged concrete or mortar, where we
2.2.2.2. Chloride ingress. In addition to changes in the chloride binding attribute it to the varying degrees of leaching near the exposed surface.
due to various leaching conditions, Fig. 8 also shows differences be­ The following three zones can be distinguished (see Fig. 8):
tween the two exposure scenarios in the chloride ingress depth. Due to
the enhanced leaching upon exposure to the NaCl solution, the chlorides – Zone I is the outermost zone which is heavily leached, leading to the
seem to have penetrated deeper into the concrete than with exposure to decomposition of AFm and AFt and the decalcification of the C-A-S-H
NaCl+KOH. This can be explained by the need to maintain the elec­ causing a dramatic reduction in chloride binding. In the case of sea
troneutrality of the sample, as described by the Nernst-Planck equation water exposure, this zone is enriched in carbonates, magnesium,
[7]. Consequently, the leaching of ions from the pore solution of a potentially sodium, and sulphates.
sample into the exposure solution directly affects the extent to which – Zone II is the moderately leached zone, where leaching leads to a
other ions are able to penetrate the sample. This is because the charge moderate reduction in pH and therefore enhanced chloride binding

8
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

in the C-A-S-H and AFm. This zone comprises the maximum of the variability of field conditions, making it possible to focus on the effects
chloride profile. of the binder itself. Six types of binders with three water-to-binder ratios
– Zone III is the zone unaffected by leaching. (w/b = 0.3, 0.4 & 0.5) were investigated: an ordinary Portland Cement
(OPC), a white Portland cement (WPC), a slag-Portland cement (CEM
It should be noted that Zone I, the heavily leached zone, is only a III/A), a 20 wt% fly-ash blend with OPC (20%FA), a limestone-calcined
couple of 100 μm thick, whereas a typical minimum thickness for a clay cement with 50 wt% Portland clinker (LC3–50), and a 20 wt% glass
profile-ground section for a chloride profile is 1 mm. The ground section powder blend with OPC (20%GP).
of 0–1 mm from the exposed surface thus comprises the heavily leached Fig. 10 shows the chloride profiles measured by μXRF after 6 months
Zone I, in which the chloride content is close to zero, but also the of unidirectional bulk diffusion with exposure to a 0.5 M NaCl solution
beginning of Zone 2 in which the chloride content increases (w/b = 0.4) [67]. Considerable variation in chloride ingress depth is
[2,30,49,56]. This means that in a conventional chloride profile the very visible between the systems, which we will investigate in Section 2.3.2,
low chloride content at the surface is not observed, and instead a but the profiles have very similar maximal intensities near the exposed
reduction in the chloride content is observed in the outer section. surface. In Section 2.3.1, we explore whether these really indicate
Changes in the value and location of the maximum chloride con­ similar chloride-binding behaviour in all samples.
centration over time due to leaching has an important impact on the
prediction of chloride ingress using analytical models based on the error 2.3.1. Binding in blended cement pastes
function solution of Fick's 2nd law. We come back to this in Section 3.1. Wilson et al. [65] were able to quantify the distribution of chlorides
These observations give some insight into the first research question - with equilibrium experiments for four of the six binders profiled in
What is the contribution of chloride binding to chloride transport? As we Fig. 10, as shown in Fig. 11. Leaching has a significant impact on the
have seen, the total chloride content at the exposed surface can be rather binding (see Section 2.2), so chloride binding was investigated with two
low and does not reflect the ingress potential, but instead is greatly exposure conditions: a solution of 0.5 M NaCl (pH ≈ 12.5 after exposure)
affected by leaching. Moderate leaching (pH > 12.4) can lead to and a solution of 0.5 M NaCl with 0.3 M NaOH (pH = 13.0 after expo­
enhanced binding, whereas severe leaching (pH < 10.5) leads to a sure). These are sea water-like chloride concentrations, and we should
decrease in the chloride-binding capacity of the paste. These findings note that different behaviour may be obtained with exposure to higher
indicate that chloride binding has a complex impact on the chloride chloride concentrations.
ingress, depending on the degree of leaching, it might first function as a The total chloride content (obtained by dividing the loss of chlorides
sink by accumulating chlorides in hydrates, but eventually it might in the exposure solution by the mass of the sample) is of the same order
function as a source by releasing the chlorides again upon severe of magnitude for the different binders. In line with previous observa­
leaching. Nevertheless, chloride binding has an important function in tions (Section 2.2.2), preventing leaching with an exposure solution
detecting how far the chloride ions have penetrated into the concrete with a higher pH (13.0) led to lower total chloride content (shaded bars
and helping us understand the shape of the chloride profiles. in Fig. 11). Similarly, the water-soluble chloride content (measured with
a methodology adapted for pastes from the ASTM C1218 [66]) is of the
2.3. Effect of the binder composition on chloride profiles same order of magnitude for all binders and decreases with increasing
pH. Strongly bound chlorides were defined as the difference between
The composition of binders affects both the microstructure of con­ total and water-soluble chlorides. Nevertheless, the ratio of strongly
crete and its chloride-binding capacity, so we can expect the chloride bound chlorides over total chlorides differs between binders (e.g., the
profiles to differ in both depth and height for different binders. This ratio for the LC3–50 system is about twice that of the CEM III/A system).
section explores the impact of the binder composition on chloride pro­ Moreover, we observe the effect of the binder on the distribution of
files based on laboratory studies of cement pastes [64,65,67,68]. chlorides between the pore solution, the Friedel's salt solid solution
Working in laboratory conditions avoids the macro-scale effects and the (FSss), and the C-A-S-H. Chlorides in the pore solution were defined as

Fig. 10. Chloride profiles in various cement pastes (w/b = 0.4) after exposure to a 0.5 M NaCl solution for 6 months, as measured by μXRF. The dashed line
represents an arbitrary reference chloride intensity of Cr = 200 counts [67].

9
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Fig. 11. Distribution of chlorides in various pastes (w/b = 0.4) after exposure to 0.5 M NaCl at pH = 12.5 (plain coloured bars) and with 0.3 M NaOH addition to
reach pH = 13 (shaded bars) [65].

the free water content of each paste sample (i.e., the difference in weight concentrations, hydrotalcite which in the presence of magnesium com­
between its saturated surface dry condition and after drying by solvent petes with AFm/AFt for aluminium, Strätlingite with the local depletion
exchange) multiplied by the concentration of the exposure solution at of Portlandite, and so on. Upon exposure to chloride-rich solutions, some
equilibrium. The quantity of chlorides depends mainly on the w/b ratio hydrates can either transform into (or provide alumina for the formation
and only slightly on the binder type. However, the impact of the binder of) Friedel's salt solid solutions (FSss) [12,69]. This all means that the
on chlorides in FSss and on C-A-S-H is more complex, as we will show AFm content in the cementitious system prior to exposure cannot be
below. directly used to predict the FSss content after exposure to chlorides [70].
This is also shown in Fig. 12a with the strongly bound chlorides as a
2.3.1.1. Impact of the binder on chemical binding in FSss. The use of function of FSss content. Higher values are obtained for Al-rich binders,
alumina-rich supplementary cementitious materials (SCMs) in blended i.e., 20%FA and LC3 systems. Interestingly, this is not the case for the
cement pastes leads to increased available alumina in the system, CEM III/A system, despite its relatively high alumina content. This
resulting in higher incorporation of alumina in the C-A-S-H structure behaviour was explained by the higher contents of magnesium and
and in the formation of additional aluminium-containing hydrates. sulphate, which lead to the increased formation of hydrotalcite and
These include hydrates from the AFm family that can chemically ettringite, with little alumina left for FSss. Hydrotalcite formed in Port­
accommodate various anions (e.g., OH− , SO2− 2−
4 , CO3 Cl ) into the
− land cement containing dolomite has been shown to take up chlorides to
interlayer of stable phases (monosulphate, hydroxy-AFm, hemi­ a similar extent as FSss [60], but the hydrotalcite/C-A-S-H mix in this
carboaluminate, monocarboaluminate, Friedel's salt, Kuzel's salt, etc.) or study (i.e., the inner reaction products of slag particles) was found to
their solid solutions, as well as ettringite (AFt) at high sulphate bind chlorides similarly or less than the C-A-S-H itself, i.e., significantly

Fig. 12. (a) Strongly bound chlorides (total minus water-soluble chlorides) vs. FSss (measured by XRD-Rietveld with an external standard), (b) comparison of the
stoichiometry of FSss (χ 2Cl) obtained from XRD-Rietveld and SEM-EDS [65]. The filled symbols indicate values for exposure solutions with pH ≈ 12.5, while the
hollow symbols indicate solutions with pH = 13.

10
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

less than the chemical binding in FSss. same range as chlorides in porosity). With this in mind, we should
Furthermore, the FSss can vary in stoichiometry depending on the remember that a relatively small fraction of the chloride in C-A-S-H is
binder and exposure conditions. This variation can be described using irreversibly bound [23,58], and most chlorides are water-soluble, which
χ 2Cl, which gives the ratio between the anions in the interlayer of the FSss brings us back to the first unresolved question - What is the contribution of
(χ 2Cl = 2Cl/Ca), where χ 2Cl = 1 reflects the theoretical Friedel's salt chloride binding (the physical binding in the C-A-S-H) on chloride transport?
composition. Monocarboaluminate has previously been considered as
the other endmember of the solid solution (i.e., χ 2Cl = 0) [12,69]. 2.3.2. Chloride ingress in blended cement pastes
However, hemicarbonate is considered here as the other endmember Chloride ions penetrate cementitious materials through the solution
following the study by Georget et al. [71] on the incorporation of guest in the pores, interacting and reacting with surfaces along the way.
ions into the hemicarbonate structure. Wilson et al. [64] investigated properties related to chloride transport
Fig. 12b shows the stoichiometry χ 2Cl of FSss for the various binders for the pastes presented in Section 2.3.1 in terms of porosity, pore
investigated. The stoichiometry was determined using two recently connectivity, pore solution conductivity, bulk electrical conductivity,
developed methods [65,71–73]; one based on XRD-Rietveld performed and the effective diffusion coefficient (Deff) obtained using the mini-
on wet samples (i.e., neither stopped nor dried), and the other using migration method in a constant migration regime after the exhaustion
SEM-EDS combined with phase segmentation EDXIA. The two methods of chloride binding.
yielded similar results: the FSss generally reached χ 2Cl ≈ 0.5, which is The characteristics of the pore network can be captured indirectly
about half that of the theoretical stoichiometry of Friedel's salt. More­ with the formation factor defined as F = σ 0 / σ b [81], where σ 0 is the
over, the increase in the pH of the exposure solution leads to lower χ 2Cl conductivity of the pore solution and σ b is the bulk electrical conduc­
values, which mean less chemical binding under these exposure condi­ tivity. Fig. 13a shows that (by far) the highest inverse formation factor
tions, notably for the LC3–50 and 20%FA systems. This agrees with our (1/F) is for WPC systems and the lowest is for LC3–50 systems, which
previous observations in Section 2.2.2. agrees with the measured Deff (and the penetration depths in Fig. 10).
Finally, the combination of the amount and stoichiometry of FSss Very similar results were obtained with the pore connectivity parameter
leads to the values of chlorides in FSss shown in Fig. 11. Due to the (β = 1 / Fϕ as defined in several papers [82–84]), which takes into ac­
different FSss contents and stoichiometry, the chemically bound chlo­ count both the formation factor and the porosity (ϕ, here the free water
rides are significantly different between systems, with values higher for content). Note that the pore connectivity parameter and the tortuosity
the LC3–50 system and lower for the CEM III/A system. Similar findings parameter (defined in different ways [85–93], one of which is the in­
were reported by Babaahmadi et al. [70] and Shi et al. [74], who verse of the connectivity parameter) should be interpreted as descriptors
investigated the effect of SCMs on the amount of chloride in FSss and C- of the porous network. This includes several phenomena and mecha­
A-S-H for composite cements containing various amounts of metakaolin nisms, such as the effects of the geometry of the pore network, in­
and/or silica fume. teractions between ions and surfaces, and errors from conductivity or
porosity measurements, i.e., the unexplained component of the bulk
2.3.1.2. Impact of the binder on physical binding in C-A-S-H. The physical conductivity once the effects of the pore solution conductivity and the
binding of chlorides occurs in the diffuse layer of the C-A-S-H and can porosity have been removed.
vary significantly depending on the amount of C-A-S-H, its composition, Fig. 13a shows that no general relationship for all blended cement
the other ions in the solution, and the pH [6,18–23,50,56,75,76]. To systems could be found between the effective diffusion coefficient and
calculate the amount of chloride physically bound in C-A-S-H, one needs the characteristics of the porous network (the formation factor and the
to determine the C-A-S-H content, which can be very challenging in pore connectivity parameter), which suggests the presence of the non-
blended cement systems without relying on complex hypotheses (e.g., negligible effect of another parameter in chloride transport.
the amount of water in the C-A-S-H, the incorporation of guest ions, the As described in Section 2.2.2, electroneutrality and ion exchange are
reaction degree of the SCMs, variable composition, or the reactivity of key in the process of chloride ingress and binding. Systems with a higher
SCM particles). Instead of trying to calculate the total C-A-S-H content, concentration of ions in the pore solution (and therefore a higher
the problem was simplified by using the amount of Si in the C-A-S-H. leaching potential) can be expected to have noticeably more chloride
This was calculated using GEMS thermodynamic modelling with input ingress. Pore solution conductivity is a great measure of the ionic
obtained using XRD-Rietveld for quantifying the crystalline phase con­ strength of the pore solution. The binder has a very important impact on
tents and using the EDXIA framework with quantitative SEM-EDS the composition of the pore solution (e.g., its alkali and hydroxyl con­
hypermaps to obtain the C-A-S-H chemical composition [58,65,73]. tents) and therefore on its conductivity [94]. Note that the conductivity
Interestingly, the OPC, CEM III/A and LC3–50 systems showed a decreases with increasing w/b due to dilution. In the pastes investigated,
relatively similar amount of Si in the C-A-S-H, even though the phase the pore solution conductivity for the WPC and LC3–50 systems is about
assemblage and C-A-S-H composition were very different, with Ca/(Si + 3–5 times lower than that of the OPC, 20%FA and 20%GP systems. At
Al) ranging between 1.25 for LC3–50 and 1.9 for the white Portland the same time, the WPC system exhibits the smallest, and the LC3–50
paste. Due to a high degree of reaction and a lower AFm content, the system the largest, chloride ingress depths in Fig. 10. This shows that the
WPC paste had a significantly higher content of Si in the C-A-S-H, role of the conductivity of the pore solution must be considered in
meaning more C-A-S-H is available for physical binding of chlorides than combination with that of the porous network, e.g., with the bulk elec­
in the other pastes. trical resistivity measurement described below, if we want to be able to
The phase segmentation of C-A-S-H from quantitative chemical correlate it to chloride ingress.
mappings [58,73] also provided the Cl/Si ratio of the C-A-S-H after As Fig. 13b shows, a good correlation (R2 = 0.90) was obtained be­
exposure: relatively similar values of Cl/Si ≈ 0.06 were obtained for the tween the diffusion coefficient from the mini-migration test (in a con­
blended cement pastes, whereas higher values of Cl/Si ≈ 0.10 were stant migration regime, after the exhaustion of binding) and the bulk
obtained for ordinary and white Portland cement. These results align conductivity (measured instantly without any chlorides) [64]. This
with previous observations showing that physical chloride binding is means that the chloride diffusion potential can be rapidly estimated
higher for higher Ca/(Si + Al) values [50,77–80]. without chlorides, using a bulk electrical conductivity measurement.
Fig. 11 includes chlorides in the C-A-S-H calculated using the amount This bulk conductivity depends on the characteristics of the pore
of Si in the C-A-S-H and Cl/Si ratios. These chlorides represent the network and the conductivity of the pore solution, so these parameters
majority of the total chlorides in the OPC and WPC pastes, but they make are also the most relevant for the diffusion.
up a much smaller proportion in the CEM III/A and LC3–50 pastes (in the How do these findings feed into the first research question - What is
the contribution of chloride binding to chloride transport? The binders

11
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Fig. 13. Diffusion coefficients for blended cement pastes measured with the mini-migration method (a) as a function of bulk electrical conductivity divided by the
pore solution conductivity (the inverse formation factor), and (b) as a function of bulk electrical conductivity [64].

investigated exhibited different distributions of the bound chloride, see 3.2.2) and to have a very high specific surface area prone to interactions
Fig. 11. Nevertheless, Fig. 14 (adapted from a study with the same pastes with ions in the pore solution, e.g., physical binding. Second, as shown
to be published by Wilson et al. [68]) illustrates that there is a very both in Georget et al. [58] and in Fig. 11, some of the chlorides in the C-
strong correlation between the bulk conductivity at 3 months and the A-S-H are water-soluble (i.e., not strongly bound). Third, although the
penetration depth (Cr = 200 counts) for bulk diffusion experiments after number of pastes investigated is small, a greater ingress depth (Fig. 10)
1 year (R2 = 0.97) and 4 years (R2 = 0.92). This strong relationship has was observed for systems with a higher level of chloride binding in the C-
deep implications for the assessment of chloride ingress resistance. It A-S-H (Fig. 11).
indicates that chloride ingress resistance is not affected by the differ­ Therefore, even if the C-A-S-H surface interactions can contribute to
ences in chloride binding of these systems (i.e., proportions of chemical reducing the diffusion rate compared to the diffusion in the pore solu­
and physical binding), because the total chloride content is similar for all tion, this does not mean that the movement of physically bound chlo­
systems exposed to 0.5 M NaCl solutions. Chloride ingress resistance rides is stopped, as is the case with FSss where the chlorides are
thus appears to be governed by the properties of the porous network and chemically incorporated into the hydrate structure. Since the porous C-
the ionic strength of the pore solution, which can be together well A-S-H generally represents a large part of the cement paste volume, a
characterized with a bulk conductivity measurement. non-negligible portion of chloride ingress could occur through the C-A-
We need to combine various observations to answer the second S-H as further described in the Section 3.2.2.
research question - What is the path for chloride transport in concrete? First, This concept of a contribution to chloride ingress through the C-A-S-
the C-A-S-H is known to hold multiple scales of porosity (see Section H needs further validation. Nevertheless, it reinforces the importance of
studying chloride binding, not only in terms of total chlorides, but also
in terms of physical and chemical binding so that we can better under­
stand both diffusion and the critical chloride content.
Did we gather new insights with regarding to our third research
question - How does our understanding of chloride profiles affect our un­
derstanding of the critical chloride content? In Section 2.1.4, we stated that
one of the preferred ways of expressing the critical chloride content Ccrit
is the [Cl− ]/[OH− ] ratio. However, this assumes that the bound chlo­
rides are completely removed from the pore solution and do not take
part in the corrosion initiation. Our observations from Section 2.3.1
show that far from being removed from the system, some of the chlorides
adsorbed on the C-A-S-H appear to be mobile. And in Sections 2.2.2 and
2.3.1, we showed that bound chlorides can be released when there are
changes in the pH. So, perhaps we should focus on the total chloride
content instead.
Glass and Buenfeld [95] argued along the same lines that bound
chlorides can contribute to the corrosion risk and that other factors
besides the hydroxyl concentration in the pore solution, such as the
buffer capacity and the steel interface, can influence the inhibitive
properties of concrete. They therefore proposed to express Ccrit in terms
of total chloride content over the mass of the binder, which is a unit
similar to that used for the chloride profiles and is much easier to
measure. Similarly, Sergi and Glass [96] proposed to express Ccrit as the
total chloride content divided by the acid neutralization capacity of the
Fig. 14. Sketch of the penetration depth (Cr = 200 counts) for bulk diffusion concrete, which aligns with our observations in Section 2.2.2: the
experiments after 1 year vs. the bulk conductivity at 3 months (for selected binding capacity is pH-dependent and can completely vanish at low pH.
blended cement pastes and water-to-binder ratios), data to be published by Their method is also experimentally more straightforward than
Wilson et al. [68].

12
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

determining the pore solution composition in concrete at the rebar. each based on a large number of assumptions for the chemical model
With regard to the impact of the SCMs, we showed in Section 2.3.1.2 (elements considered, initial phase assemblage, solid solutions, etc.)
that the total amount of hydrates can vary significantly between blended [61,99–105]. They all report excellent results when fitted to selected
cement systems with relatively high SCM replacement levels because of experimental data. This is especially true when the calibration and the
the presence of incomplete SCM reactions or fillers leading to a portion validation set are the same or similar. For diffusion problems like
of the paste remaining unreacted and therefore inert to chlorides. chloride ingress, the diffusion coefficient is the most sensitive parameter
Potentially, this might make normalization of the critical chloride con­ [103,106], i.e., its fitted value is highly dependent on the uncertainty of
tent to the total amount of hydrates more relevant than normalization to the overall model. It is worth noting that the reliability of the model
the total binder content. However, the definition of Ccrit as the total decreases as its complexity increases, i.e., with the number of input
chloride/neutralization capacity ratio [96] is independent of normali­ parameters required.
zation to either paste or hydrates and might therefore be of greater in­ The fitting of total chloride profiles using mechanistic models to
terest for more complex blended cement systems. obtain a diffusion coefficient is therefore misleading due to the un­
certainties in the chemical model. Instead, it is suggested that the
3. Using chloride profiles diffusion coefficient should be determined independently through other
methods. The use of electro-migration methods, as suggested in Section
In the previous section, we described the features of chloride profiles 2.3.2, by other researchers [64,107–109], and in the fib model code
and how concrete composition, exposure and binder type can affect [110,111], is of particular interest due to the relatively short time
them. We now discuss the use of these profiles, in particular with regard required (~2 weeks) compared to natural diffusion (~3 months - 3
to models and simulations, for which they can be used as either input or years).
output. They are input for models used to derive quantitative material
parameters, the most common being the chloride diffusion coefficient. 3.1.2. Can chloride profiles be used to fit engineering properties?
They are output for models that predict profiles used to assess the state Unlike mechanistic models, engineering models have only a single
of the material and the structure, for example to predict the chloride equation with a single variable. This means that fitting these models to
penetration depth at a certain point in time or for comparison with in- chloride profiles can, in theory, provide two reliable parameters: the
situ monitoring. amplitude (the amount of chloride) and the rate of the ingress (or,
In this section, we distinguish between two types of models: mech­ equivalently, the penetration depth).
anistic and engineering. Mechanistic models aim to describe individually Due to the overall conservation of the typical diffusion profile shape,
each contributing mechanism. For chloride ingress, they are reactive Fick's second law is often used to model chloride profiles with the total
transport models in which the diffusion of chloride ions is the main chloride content, C, as the main variable [97,110,112]:
transport mechanism. Some models take into account the transport of
other ions and/or water. The transport of chloride is coupled to a
∂C ∂2 C
= Dapp 2 (1)
chemical model that describes the binding in the solid phases, using ∂t ∂x
either a binding isotherm or more complex thermodynamic/kinetic where Dapp is an apparent diffusion coefficient, t is time, and x is the
models. For each of these mechanisms, a separate mathematical equa­ ingress depth. In constant boundary conditions and with a semi-infinite
tion can be written. Engineering models, in contrast, lump together all the material, the following analytical solution exists, using the comple­
mechanisms in a single equation. This is achieved using strong as­ mentary error function (erfc):
sumptions to simplify the problem (e.g., only diffusion, fixed porosity, ( )
no leaching, etc.) and using lump parameters, such as the apparent x
C(xt) = CS erfc √̅̅̅̅̅̅̅̅̅̅̅ (2)
diffusion coefficient [97]. Unlike mechanistic models, which often 2 Dapp t
require complex numerical solvers, engineering models are simplified to
obtain an analytical solution to the problem. Analytical solutions might where C(x,t) is the total chloride content at a depth x and time t, and the
seem a strong mathematical requirement, but as long as only diffusion is fitting parameters are the apparent diffusion coefficient (Dapp) and the
considered, a large number of initial and boundary conditions can be surface concentration (Cs). Why are we able to fit two parameters in one
solved analytically [98]. The engineering models should strive to be profile? This is because both the height and the depth of the profile can
physically representative, and thus, they should derive from the mech­ be fitted, whereas in mechanistic models, the amplitude is fixed by the
anistic models. However, as detailed in this section, mechanistic models chemical model and the porosity, which are assumed to be fixed prop­
also have their fair share of limitations. Therefore, engineering models erties of the material. However, it is important to assess the physical
often rely on empirical and field validation. meaning of all fitting parameters.
The surface concentration, Cs, has been used as an indication for the
3.1. Profile as input: parameter determination total binding capacity [11]. However, our discussion on the impact of
exposure and more specifically leaching (Section 2.2.2) demonstrates
3.1.1. Can chloride profiles be used to fit mechanistic models? the limitations of this approach. In realistic conditions, the surface
As described above, mechanistic models contain at least two equa­ concentration of chloride profiles does not reach a maximum at the
tions (chloride diffusion and chloride binding) with at least two vari­ surface but slightly deeper. This is referred to as the “peaking behav­
ables (free and bound chlorides). Mathematics requires that two pieces iour”, see Section 2.2.2.3. The first few points of the profile near the
of information are provided to determine the parameters of these exposed surface typically need to be ignored to obtain a good fit with the
equations. In Section 2, we showed that the chemical model is full of erfc function shape, as illustrated on Fig. 15. This means that Cs is only a
uncertainties (C-A-S-H amount, C-A-S-H sorption, AFm solid solution, fitting parameter with no direct physical meaning.
pH and leaching, and so on). So, if we attempt to fit the diffusion coef­ The apparent diffusion coefficient is the parameter responsible for
ficient on the total chloride profile including the uncertain chemical the profile shape [61,97]. A lower apparent diffusion coefficient leads to
model, the fitted diffusion coefficient will be highly unreliable. This is a sharper diffusion front, penetrating less into the sample. Although it
because chloride binding does not modify the overall diffusion profile has the same unit [m2/s] and the same order of magnitude (~1.10− 12
shape, as demonstrated by the successful application of engineering m2/s), the apparent diffusion coefficient should not be confused with the
models such as Fick's second law, which only describe diffusion and not effective diffusion coefficient. The latter is used in mechanistic models
binding (see the next section). and, in theory, it represents only the effect of the pore network geometry
This argument is confirmed by the many reactive transport models,

13
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

( √̅ ) √̅
xc t =a t+b (3)

where xc is the penetration depth for a defined reference chloride con­


tent, t is time, a is a diffusivity coefficient, and b is the offset. The
diffusivity coefficient, a, is similar to the “carbonation” coefficient used
to compare cementitious systems and carbonation conditions [118,119],
though for carbonation, no offset b is considered.
A similar approach is to collapse the profiles collected at different
times for the same system and chloride exposure on to a “master curve”
√̅̅
by plotting along the x/ t axis (see Fig. 16a). This curve makes it
possible to fit a single Dapp (and Cs) for all the profiles independent of the
time of collection. Mathematically, this means we can still use Eq. (2),
√̅̅
but the variable is x/ t instead of x and t.
These approaches provide a good method for increasing the reli­
Fig. 15. Chloride profile fitted with the erfc function to obtain parameters ability of the “diffusivity parameter” that describes the chloride profile,
(Dapp, Cs, and xc). because it uses data from several profiles. These master curves are an
excellent tool for assisting decision-making for the repair of structures or
and the particle-surface interactions. In contrast, the apparent diffusion estimating their remaining service life, e.g., by detecting significant
coefficient is a lumped parameter, representing both the transport co­ deviation in the expected behaviour, which could be attributed for
efficient and the effect of the binding, so its value depends on the example to the occurrence of cracks.
boundary conditions (composition of the exposure solution, tempera­ Of course, the practical difficulty is to obtain several of these profiles
ture, and so on). However, it can be demonstrated that if the binding for the same system, which in this context refers to both the concrete
isotherm is linear, the effective and apparent diffusion coefficients can composition and the exposure, or in other words both the material pa­
be related analytically [61,97]. Such a linear binding isotherm can be rameters and the boundary conditions. Chloride content determination
obtained by assuming pH-dependent isotherms, as illustrated in Fig. 9. is a laborious process. However, the square root approach based on xc
Even though the apparent diffusion coefficient cannot be directly (the penetration depth for a defined reference chloride content) can also
related to intrinsic physical parameters, provided the measurements are be applied to (semi-)quantitative methods, such as silver nitrate spray­
carried out in the same controlled conditions, the apparent diffusion ing [120], EPMA [121] or μXRF [122], which can be much less labour-
coefficient of samples can be compared to provide a classification of intensive than ordinary chloride profiles. There is also a lot of interest in
cementitious systems [113]. For example, this allowed us to successfully developing in-situ methods to measure chloride ingress, such as re­
compare the impact of SCMs on the apparent diffusion coefficient (see sistivity [123–125] and chloride sensors [126,127].
Section 2.3). The main uncertainty for these approaches is the time needed to
The final parameter in Fig. 15 is the penetration depth (xc). This observe a meaningful evolution of the chloride ingress. For the square
depth is found by assuming a reference chloride content and deter­ root approach, this translates into the decision on whether to use an
mining the corresponding ingress depth on the chloride profile through offset b or not (see Fig. 16b). The offset can be attributed to an initial
interpolation. This reference chloride content is just a chosen total period of fast ingress, a fitting artefact, or the effect of leaching (Section
chloride content and is therefore neither directly nor uniquely related to 2.2.2). Further research is needed to better understand the physical
the critical chloride threshold [61,114]. meaning of these parameters.
If chloride profiles are obtained after various exposure times, the
penetration depth can be collected, and then plotted as a function of the 3.2. Profiles as output: monitoring and predictions
( √̅̅)
square root of the exposure time (xc = f t ). As illustrated in Fig. 16b,
the points then align linearly as a function of the square root of time 3.2.1. Engineering models
[61,67,114–117]: Without prior knowledge, the application of engineering models
generally results in highly uncertain predictions. Since they rely on

√̅̅
Fig. 16. The square root analysis, a) Master curve of chloride profiles for one system plotted as a function of x/ t , b) “Critical concentration” plotted as a function of
√̅̅
x vs. t and corresponding trend lines, with and without offset.

14
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

fitting apparent parameters, they are highly dependent on the condi­ describe how the experimental findings described in Section 2 can
tions for which the calibration data were obtained. The range of material provide input to improve the mechanistic models and to discuss the
(SCMs, w/c, curing, etc.) and boundary conditions (temperature, rela­ further experiments required to this end. To structure this discussion, we
tive humidity, alkalinity, composition exposure solution, etc.) cannot be need to describe the building blocks of a reactive transport model. This is
fully represented. This is especially true when considering the ever- illustrated in Fig. 17.
widening range of new SCMs or the unfortunate consequences of This model describes the flux of matter (ions, water, gas, etc.) that
climate change on for example temperature and sea water composition. contribute to two mechanisms: the transport (i.e., the movement of
However, it is reasonable to utilize a two-step process: calibration mobile species), and the reactions (i.e., the phase changes, such as
and modelling. The calibration step uses at best data from the structures precipitation or evaporation). The former is parameterized by the
at an early age and/or in-situ testing blocks from the same batch, though transport properties (e.g., the diffusion coefficient), while the latter are
often one has to rely on existing data of a similar concrete and exposure. parameterized by the equilibrium conditions and the reaction path (i.e.,
These constrained parameters are then used in the second step to model the local equilibrium and kinetics). To better understand how such pa­
the long-term behaviour of the concrete in these conditions. These rameters are chosen for various physical problems, we recommend
predictions (i.e., baseline behaviour) can then be compared to real reading publications that use a similar formalism for leaching, carbon­
measurements to analyse the actual state of the concrete. The moni­ ation, and chloride ingress [61,132,133]. To simplify our discussion
toring is of course easier with non-destructive methods [126]. In this here, we note the diffusion coefficient D, and the binding term B.
case, the parameters in the model can also be continuously updated We assume that each effect contributes to the overall parameter. For
using further monitoring. example, we could write:
It has been observed that the apparent diffusion coefficient approach
does not work when using early-age experimental data to predict long- D = D0 ̃
daggregates ̃
dITZ ̃
dpaste geometry
̃
dsolution ̃
dcharged_surfaces + Dcracks (5)
term data. Indeed, the apparent diffusion coefficients typically decrease
In most models, the multiplication factors (̃ d) are lumped together in
significantly as a function of time at early ages [128–130], even when
a single complex parameter, like the pore connectivity presented in
the concrete is not subject to mechanical loading and thereby potential
Section 2.3.2, the formation factor [82], or one of the many definitions
crack formation, and kept in a controlled environment. This time-
of tortuosity [134]. However, to correctly take the wide variety of
dependent behaviour is usually attributed to the continuation of hy­
binders into account, each term needs to be understood and modelled
dration and the corresponding microstructural changes. In engineering
separately.
models, such as those used in the DuraCrete guidelines [131], which
A question to answer is how the exposure solution should be
provided the basis for the fib Model Code [110,111] and resulted in the
modelled. Most “fully” mechanistic models use the Nernst-Planck
ISO standard [112], this is commonly taken into account by introducing
equation [135,136]. This equation couples the diffusion of multiple
an aging coefficient, m:
ions to ensure the conservation of the electroneutrality of the pore so­
(t )m lution and the solids. Another simpler method is to consider only the
(4)
0
Dapp (t) = Dapp,0
t chloride diffusion. In this case, ̃
dsolution needs to consider the effect of the
other ions. In particular, it means that D is a function of the chloride
where Dapp,0 is a reference apparent diffusion coefficient at t0 , and m is
concentration (and the exposure solution), and therefore equivalent to
the aging coefficient, which is a fitting parameter.
an apparent diffusion coefficient [135]. The importance of the coupling
It is clear that such aging coefficients are non-physical because this
between the diffusivity of multiple ions is related to our discussion in
mathematical expression is not bounded (i.e., D goes to 0). As such this
Section 2 on the importance of pH and leaching. However, for modelling
empirical approach is limited and prone to over-fitting. It only works
purposes, the first order of magnitude is obtained by quantifying the
because these parameters are fitted to limited data. In particular, it has
diffusion of chloride “on its own”, especially when comparing different
been shown that long-term chloride profiles collapse to a master curve in
binders. The diffusion of the chloride ion is therefore the focus of this
the square root view [61], which indicates a unique apparent diffusion
section.
coefficient at late age (5 to 20+ years). Similarly, the Life 365 model
The second research question – What is the path for chloride transport?
(www.life-365.org) uses an aging coefficient for up to 25 years, after
- concerns the definition of the geometrical contribution of the pore
which the diffusion coefficient is considered constant. Nevertheless, the
structure to the diffusion coefficient, D.
early-age measurements (typically at <1 year) still show significant
We first discuss normalization with respect to the binder content. In
differences from later ages (5, 10 and 20 years) for some systems [61].
modelling terms, this translates into how the aggregates should be
The effects that lead to these differences in early-age investigations need
modelled. Aggregates are considered to be impermeable and not reac­
to be identified and quantified because they are especially important
with slow reacting SCMs such as fly ash, for SCM assessment, and for tive with regard to chlorides. This means that ̃
daggregates is a multiplication
formulation optimization. factor that describes the reduction of the connected pore space, as well
as the increase in geometric tortuosity. For simple geometry, this can be
3.2.2. Mechanistic models relatively easily resolved using a homogenization scheme [137]. Ho­
To discuss mechanistic models in depth would require a review on its mogenization is a mathematical method for modelling the microstruc­
own, so the aim of this section is more limited. Here we just want to ture to predict the effective transport properties. However, concrete is a

Fig. 17. The building blocks of a reactive transport model for chloride binding.

15
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

very heterogeneous and multiscale material, so quantifying this effect is effective diffusion coefficient of cement pastes is close to the diffusivity
much more difficult [138–141]. in C-A-S-H pores in most relevant systems, most models assume that
Aggregates can also have an effect on the diffusion in the binder due liquid diffusion is limited by the gel pores. This agrees with permeability
to the presence of the ITZ. Depending on the conditions, the ITZ has been measurements [39].
described as either more or less permeable than the bulk paste [41,142]. Therefore, we need to estimate the volume fraction of capillary and
There is still no consensus on the impact of the ITZ on the overall ingress, gel pores present, to obtain a first estimate of the effective diffusion
especially since it is difficult to quantify. However, it has been shown coefficient of a cement paste. This is challenging because it is difficult to
that the cement type has a stronger impact on chloride diffusion than the quantify the hydrate phase assemblage, especially the C-A-S-H due to its
ITZ [143]. Cracks and microcracks in the surface of the concrete might variable composition and microcrystalline nature. Moreover, there are
also impact the chloride penetration depth, especially at an early age two populations of C-A-S-H: the high-density C-A-S-H in the inner
[144,145]. product, and the low-density C-A-S-H in the outer product. It is unclear
Another important matter is where can chlorides diffuse in bulk cement how the inner and outer products contribute to the chloride transport.
paste? This requires an understanding of the various pore classes as Do they have different intrinsic diffusion coefficients? The development
determined from proton NMR relaxometry (1H NMR) measurements of models combined with advances in quantification using advanced
[146], see Fig. 18. Capillary pores are the pores between the grains that characterization techniques might answer this question in the future.
are left after the initial packing and the initial water-to-cement ratio. This is especially important when considering blended cement sys­
Interhydrate pores are a subcategory of capillary pores and usually tems, where SCMs lead to changes in the system by reducing the clinker
defined as the pores between the agglomerates of hydration phases (such factor and changing the C-A-S-H quantity, composition, and pore
as the C-A-S-H needles). Gel pores are the intrinsic porosity of the C-A-S- structure. Whether the space is filled with C-A-S-H, capillary pores, or
H and are often described as the result of the poor crystallinity and poor impermeable phases (e.g., AFm or unreacted SCM) will govern the
stacking of the C-A-S-H sheets. The final class is not a pore, but the impact of SCMs on the geometric term of the effective diffusion coeffi­
interlayer space in the C-A-S-H structure, where the water and ions are cient. Advances in knowledge in the domain of hydration, microstruc­
only mobile through bond breaking. 1H NMR measurements are more ture and thermodynamic predictions [149] are therefore highly relevant
accurate than the more commonly used mercury intrusion porosimetry to obtaining quantitative models for the diffusion coefficients.
(MIP) method, which can reliably capture only interhydrate and capil­ For the first research question - What is the contribution of chloride
lary porosity due to the physical limitations of the pressure it uses [147]. binding to chloride transport? - we need to distinguish two contributions
MIP is therefore not a suitable method for analysing the effect of the pore to the chloride profiles: the mobile and the immobile chlorides, i.e. the
structure on the diffusion coefficient. chloride that can contribute to transport, or not, in agreement with the
What is the impact of the various pores on chloride ingress? These pore definition established in Fig. 17 Binding is commonly associated first
classes can be used in models to derive the effective transport properties with immobile species, but this section highlights that there are two
[139,148]. Since these pores have different geometries and different types of binding: strong binding (immobile species), and weak binding
sizes, their intrinsic diffusion coefficients are different by orders of (mobile species).
magnitude. Typically, the intrinsic diffusion coefficient of C-A-S-H gel Chemical binding is assumed to be strong binding, because the
pores is around 10− 11–10− 12 m2/s, a similar order of magnitude to the chlorides are trapped in the interlayer of the AFm crystalline structure.
cement paste effective diffusion coefficient [141]. The diffusion coeffi­ Physical binding in the C-A-S-H can be assumed to be weak binding,
cient for the capillary pores is assumed to be one to three orders of because the most common model for physical binding assumes the
magnitude higher depending on their characteristic pore size. Since the chlorides accumulate in the electrical double layer (EDL), i.e., in the

Fig. 18. Pore classes in cement pastes revealed by Proton NMR Relaxometry and Mercury Intrusion Porosimetry.

16
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

solution. However, a portion (~10–20 %) of the binding in C-A-S-H has subject to heterogeneity effects and the imprecision of the experimental
been shown to be irreversible, i.e., not taken into account in the water- protocols. It is possible to normalize to the binder content, but this is not
soluble test (see Fig. 11) [61]. This could indicate that, even though the straightforward due, for example, to leaching. Chloride profiles also
majority of physical binding should be modelled as weak binding, a tend to exhibit “peaking behaviour” near the exposed surface, which
portion of the bound chloride in the C-A-S-H should be considered as means that the total chloride content is lower at the surface than in the
strong binding. following sections deeper in. We attribute this peaking behaviour to
Strong binding makes two complementary contributions to the leaching because severe leaching at the surface (pH < 10.5) leads a loss
chloride transport. First, it acts as a “storage” term (e.g., a sink initially, of the chloride-binding capacity of the paste, whereas moderate leaching
a source during leaching). This contribution reduces the total flux of (pH > 12.4) slightly deeper in can lead to enhanced binding. Given the
chloride by reducing the amount of free chloride able to move further. variation in paste content and the effect of leaching, it is clear that the
Second, at equilibrium it determines the concentration in the pore so­ total chloride content near the surface is governed by binding and that it
lution and is therefore the driving factor for diffusion. To analyse the does not represent a “driving potential” for chloride ingress.
effect of the strong binding, it is important to determine the equilibrium In contrast, we argued that the central part of the profile is the most
between the pore solution and the solid phases. This is particularly indicative of the material's chloride resistance in a given environment. In
√̅̅
challenging due to the presence of solid solutions, i.e., FSss (see Sections the square root approach to modelling, the profiles overlap on the x/ t
2.2 and 2.3). From the modelling point of view, this means that to view, indicating that a single lump parameter can be used to quantita­
√̅̅
predict diffusion, we need accurate chemical models to determine the tively describe the profiles. The x/ t master curve also enables to
FSss in equilibrium with the solution so that we can predict the strong identify which points at the surface or at the tail of the curve should be
binding. considered in the fitting process as several profiles can be directly
The contribution of the weak binding to chloride ingress is the change √̅̅
compared. The x/ t master curve is also the best option for defining
in the mobility of ions. Due to the presence of charged surfaces, ions are what the surface and tail parts of the curve mean. The lump parameter is
going to be attracted to or repulsed from the EDL [78,79]. This effect is either the apparent diffusion coefficient (the erfc solution of 2nd Fick's
only dominant when the pore size is sufficiently small compared to the √̅̅
law) or the diffusivity coefficient (the slope in the xc vs. t approach).
EDL size (~1 nm) [52,53]. To properly determine the effect of the These lump parameters depend on the exposure conditions, so they are
physically bound chloride on chloride ingress, we need to model the not intrinsic material properties, but they can be used to classify
impact of the charged surface on the diffusion coefficient. More specif­ cementitious systems when measurements are carried out in the same
ically, there are two challenges. Firstly, we need to determine the conditions. The square root approach is more reliable, but more time-
amount of surface and its state (i.e., the density of charges); and sec­ consuming because it requires several profiles.
ondly, the reduction of mobility of ions in the presence of this surface. These engineering models can also be used for predicting chloride
This charged-surface effect needs to be included (̃ dcharged_surfaces ) to enable profiles, though they do require sufficient “early-age” data on similar
the model to predict the impact of changes in the C-A-S-H composition concrete in the same exposure to reliably calibrate the models. When
and densities (i.e., packing) on chloride transport. The C-A-S-H applied correctly, they can be a very valuable engineering tool for the
composition and densities strongly depend on both the presence of any maintenance and repair planning of existing structures. We want to
SCMs (due to their wide range of composition and pozzolanic activity) underline that calibration data for the same exposure is needed, which is
and the exposure conditions (i.e., any leaching of hydroxyl ions). a true challenge when making detailed engineering predictions of
And finally, how does this feed into to our third research question - chloride ingress for a real structure due to the high variability in the
How are our findings on chloride profiles also important for corrosion? exposure conditions both on the structure (e.g. surface orientation and
For engineering models, we need expressions we can easily relate to height) and in time.
total chloride content per binder or concrete. In that respect, the total
chloride content/binder content [95] and the total chloride content/ 3.3.2. Knowledge gaps
neutralization potential [96] are promising. Both can easily be linked to Sufficient “early-age” data is the core of the issue because current
measured chloride profiles and are relatively easy to determine experi­ engineering approaches rely on an aging coefficient, or an offset in the
mentally. The latter also considers variations in the passivation capacity square root approach. The main issue is that these empirical coefficients
of different binders. require the measurement of several profiles to reliably identify these
For mechanistic models, which in theory can predict the pore solu­ parameters. The challenge is due to both the complexity and the
tion composition in exposed concrete through advanced chemical destructive aspects of the current methods of obtaining these profiles.
models that take into account binding and interactions with the envi­ The solution is either to identify the causes of these early-age effects and
ronment such as leaching, the expression [Cl− ]/[OH− ] may be propose a flexible model, or to develop fast non-destructive methods for
preferred. This expression also directly links to corrosion science models measuring the profiles.
and experiments [150].
Considering that the chloride concentration is only one of many 4. Summary and perspectives
factors affecting reinforcement corrosion initiation [45,46], future
mechanistic models should ideally also be able to predict other equally This paper presents how a series of case studies brought us to a better
important factors such as the moisture conditions, availability of oxygen understanding of chloride profiles in concrete, as schematically illus­
and concrete resistivity in the vicinity of the reinforcement, which are trated in Fig. 19. We also reflect on the use of chloride profiles in en­
required to be able to calculate a probability of corrosion initiation. gineering and mechanistic models and highlight the crucial knowledge
gaps encountered. The paper evolves around three central questions, the
answers to which are summarized below.
3.3. Implications for engineering practice

3.3.1. Which part of the chloride profile is relevant for engineering models? 4.1. What is the contribution of chloride binding to chloride transport?
Traditionally, the surface (binding capacity) and the tail end (ingress
depth) of a chloride profile are used for engineering models. In fact, We are able to determine chloride profiles thanks to the interactions
however, neither should be used. The profile near the exposed surface is between chlorides and cement hydrates, i.e., chemical binding (in Cl-
likely to be affected by an increased cement content in the concrete due AFm) and physical binding (on C-A-S-H) (see Fig. 19) because only a
to the “wall effect”. The tail-end of the profiles (low chloride content) is small fraction of the chlorides are found in the capillary pores (in the

17
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

two parameters comprise the chloride ingress resistance of paste. In


other words, the chloride penetration resistance can be assessed without
knowledge of binding properties. Such knowledge is crucial to under­
standing the path for chloride transport.
The exposure case studies showed that the ingress of chloride anions
is not followed equimolarly by cations such as sodium. Instead, we found
it was linked to hydroxyl leaching. We demonstrated that limiting
leaching reduced the ingress and binding of chlorides in concrete,
whereas increased leaching promoted them. This also shows how
important pore solution conductivity is for assessing chloride ingress.
When we compared concretes exposed to sea water and NaCl solu­
tion, we observed that with sea water exposure the precipitation of
phases near the concrete surface partially filled larger voids and cracks,
but that this did not limit the chloride ingress compared to NaCl expo­
sure for which no precipitation was observed. This indicates that chlo­
ride ingress is not affected by a reduction in the coarser porosity, but is
governed by transport in the finer pores, i.e., those associated with the C-
A-S-H phase. Similarly, the higher chloride ingress observed in Portland
cement pastes than in blended cements might be associated with their
higher C-A-S-H content.
Mechanistic models describe the transport of chloride ions using the
intrinsic diffusion coefficient, which is linked to the pore structure of the
Fig. 19. Concluding sketch of a chloride profile from a concrete element sub­
cement paste. Porosity can be divided into a number of pore classes
merged in sea water at Solsvik field station for 16 years [2], which illustrates according to their morphology. Diffusivity depends on the mobility in all
our gained understanding compared to Fig. 1. the different types of pores. Although the mobility of chloride in the gel
pores associated with C-A-S-H is less than in larger pores, the contri­
pore solution). bution of the gel pores to chloride transport seems to be crucial because
The exposure case studies showed that the total chloride content near they represent such a large part of the porosity in hydrated cement paste.
the surface is strongly affected by variations in the amount of bound
chloride due to, on the one hand, the “wall effect”, which causes a higher 4.3. How are our findings on chloride profiles relevant for corrosion of the
paste content near the surface, and on the other hand, leaching, which reinforcement?
causes increased binding with moderate leaching (pH > 12.4) or
reduced binding with severe leaching (pH < 10.5). These variations The understanding of chloride profiles described in this paper can
result in the “peaking effect” typically observed near the exposed help us select a meaningful expression for Ccrit for use in estimating the
surface. risk of corrosion from chloride profiles. First, when we compare Ccrit for
This means that the total chloride content near the surface does not different concretes, we should normalize it to the cement paste content
represent a “driving potential” for chloride ingress. This was confirmed or even to the content of hydration products if a considerable part of the
in the binder case study, where the chloride profiles of cement pastes cement has not reacted (e.g., blended cement with high SCM content).
with various SCMs exposed to 0.5 M NaCl all showed very similar total Second, to express Ccrit we need a better understanding of the distribu­
chloride contents at the surface and similar binding capacities under tion of chlorides and their role in corrosion initiation. This issue is
equilibrium conditions, even though the ingress depths varied greatly. strongly related to the understanding of chloride binding, i.e., how the
Nevertheless, chloride binding does affect chloride ingress. This ef­ pore solution interacts with the solid phases and the influence of
fect is best understood by looking at the distribution of chlorides be­ changes in pH (e.g., leaching). Third, the hydroxyl concentration of the
tween the pore solution, the Cl-AFm, and the C-A-S-H. The binder case pore solution and the acid neutralization capacity of the binder are both
study showed that the distribution can vary significantly between measures for the corrosion inhibition potential of the binder. Taking this
binders. The binder impacts the amounts of Cl-AFm and C-A-S-H. For into account, a definition of Ccrit as a ratio of the chloride content vs. the
example, the white Portland cement seems to form significantly more C- inhibition potential seems more appropriate. Combining all these ob­
A-S-H than the blended systems. Moreover, the binder and the pH affect servations, the expression of Ccrit as the total chloride content normal­
the amount of chloride taken up per unit mass of Cl-AFm and of C-A-S-H. ized to the acid neutralization potential might be preferred because it is
For mechanistic models, the total chlorides need to be divided into independent of the binder content, it considers a corrosion inhibition
strongly bound and weakly bound chlorides (see Fig. 19). The binder potential, it is relatively easy to determine, and it can be linked to a
case study showed that the strongly bound chlorides, i.e., chlorides in chloride profile because it uses similar units.
the AFm interlayer and 10–20 % of chlorides in the C-A-S-H, can be Mechanistic models can open up for a much larger range of expres­
considered immobile. In a mechanistic transport model, they would end sions for Ccrit because they can take into account complex chloride-
up in a storage term. The weakly bound chlorides (i.e., most chlorides binding behaviour, leaching, and can predict pore solution composi­
adsorbed on the C-A-S-H) will contribute to the transport, but at a slower tions in the vicinity of the reinforcement.
rate than the free chlorides.
4.4. What next?

4.2. What is the path for chloride transport? We have seen that strongly bound chlorides (Cl-AFm) and “mobile”
chlorides (in the C-A-S-H and in the pore solution) play different roles in
By investigating various blended cement pastes and w/b ratios, we transport and potentially in corrosion. This should be taken into account
showed that the chloride penetration depth after exposure a 0.5 M NaCl in upcoming mechanistic models. However, experimental validation of
solution was strongly correlated with the bulk electrical conductivity. these models will require the use of advanced experimental methods,
Since this value (in the absence of chlorides) combines the influence of some of which exist but are still challenging to use.
both the pore network and the pore solution, we can conclude that these This is particularly important in blended cement systems because

18
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

SCMs affect several parameters related to reactive transport, such as: (1) [14] F.P. Glasser, A. Kindness, S.A. Stronach, Stability and solubility relationships in
AFm phases: part I. Chloride, sulfate and hydroxide, Cement Concr. Res. 29
the amount and the stoichiometry of the Cl-AFm; (2) the amount of C-A-
(1999) 861–866, https://doi.org/10.1016/S0008-8846(99)00055-1.
S-H formed and its composition and spatial distribution; (3) the pore [15] M.Y. Hobbs, Solubilities and Ion Exchange Properties of Solid Solutions Between
solution in equilibrium with the solids; and (4) the pore structure and the Hydroxyl, Chlorine and Carbontrioxide End Members of the Monocalcium
therefore its diffusivity and transport properties. Aluminate Hydrates, PhD thesis, University of Waterloo, Canada, 2001.
[16] H. Pöllmann, Mischkristallbildung in den Systemen 3CaO⋅Al2O3⋅CaCl2⋅10H2O –
This paper has demonstrated the crucial role of C-A-S-H in chloride 3CaO⋅Al2O3⋅CaCO3⋅11H2O und 3CaO⋅Al2O3⋅CaCl2⋅10H2O – 3CaO⋅Al2O3⋅Ca
transport in concrete. It is vital that we obtain a better understanding of (OH)2⋅12H2O, PhD thesis, University of Erlangen-Nuremberg, Germany, 1980.
the impact of differences in C-A-S-H composition and structure on [17] H. Pöllmann, Solid solutions of complex calcium aluminate hydrates containing
Cl− , OH− and CO2− 3 anions, in: 8th International Congress on the Chemistry of
chloride transport properties by using advanced experiments and Cement Rio de Janeiro, 1986, pp. 300–306.
models. Including this understanding in self-sufficient reactive transport [18] C. Arya, N.R. Buenfeld, J.B. Newman, Factors influencing chloride-binding in
approaches by combining the elements discussed above concerning the concrete, Cem. Concr. Res. 20 (1990) 291–300, https://doi.org/10.1016/0008-
8846(90)90083-A.
inherent heterogeneity of concrete is the next big challenge we face. [19] A. Delagrave, J. Marchand, J.-P. Ollivier, S. Julien, K. Hazrati, Chloride binding
capacity of various hydrated cement paste systems, Adv. Cem. Based Mater. 6
CRediT authorship contribution statement (1997) 28–35, https://doi.org/10.1016/S1065-7355(97)90003-1.
[20] O. Wowra, M.J. Setzer, Sorption of chlorides on hydrated cement and C3S pastes,
in: M.J. Setzer, R. Auberg (Eds.), Frost Resistance of Concrete, E & FN Spon,
Klaartje De Weerdt: Conceptualization, Writing – original draft, London, 1997, pp. 147–153.
Writing – review & editing, Visualization. William Wilson: Conceptu­ [21] K. De Weerdt, D. Orsáková, M.R. Geiker, The impact of sulphate and magnesium
on chloride binding in Portland cement paste, Cem. Concr. Res. 65 (2014) 30–40,
alization, Writing – original draft, Writing – review & editing, Visuali­
https://doi.org/10.1016/j.cemconres.2014.07.007.
zation. Alisa Machner: Conceptualization, Writing – original draft, [22] C. Labbez, A. Nonat, I. Pochard, B. Jönsson, Experimental and theoretical
Writing – review & editing, Visualization. Fabien Georget: Conceptu­ evidence of overcharging of calcium silicate hydrate, J. Colloid Interface Sci. 309
(2007) 303–307, https://doi.org/10.1016/j.jcis.2007.02.048.
alization, Writing – original draft, Writing – review & editing,
[23] G. Plusquellec, A. Nonat, Interactions between calcium silicate hydrate (C-S-H)
Visualization. and calcium chloride, bromide and nitrate, Cem. Concr. Res. 90 (2016) 89–96,
https://doi.org/10.1016/j.cemconres.2016.08.002.
[24] S. Barzgar, B. Lothenbach, M. Tarik, A. Di Giacomo, C. Ludwig, The effect of
Declaration of competing interest
sodium hydroxide on Al uptake by calcium silicate hydrates (CSH), J. Colloid
Interface Sci. 572 (2020) 246–256, https://doi.org/10.1016/j.jcis.2020.03.057.
The authors declare that they have no known competing financial [25] J. Haas, A. Nonat, From C–S–H to C–A–S–H: Experimental study and
thermodynamic modelling, Cem. Concr. Res. 68 (2015) 124–138, https://doi.
interests or personal relationships that could have appeared to influence
org/10.1016/j.cemconres.2014.10.020.
the work reported in this paper. [26] I. Galan, L. Perron, F.P. Glasser, Impact of chloride-rich environments on cement
paste mineralogy, Cem. Concr. Res. 68 (2015) 174–183, https://doi.org/
Data availability 10.1016/j.cemconres.2014.10.017.
[27] C. Jones, S. Ramanathan, P. Suraneni, W.M. Hale, Calcium oxychloride: a critical
review of the literature surrounding the formation, deterioration, testing
No data was used for the research described in the article. procedures, and recommended mitigation techniques, Cem. Concr. Compos. 113
(2020), 103663, https://doi.org/10.1016/j.cemconcomp.2020.103663.
[28] Y. Elakneswaran, T. Nawa, K. Kurumisawa, Electrokinetic potential of hydrated
Appendix A. Supplementary data cement in relation to adsorption of chlorides, Cem. Concr. Res. 39 (2009)
340–344, https://doi.org/10.1016/j.cemconres.2009.01.006.
Supplementary data to this article can be found online at https://doi. [29] K. De Weerdt, H. Justnes, M.R. Geiker, Changes in the phase assemblage of
concrete exposed to sea water, Cem. Concr. Compos. 47 (2014) 53–63, https://
org/10.1016/j.cemconres.2023.107287. doi.org/10.1016/j.cemconcomp.2013.09.015.
[30] U.H. Jakobsen, K. De Weerdt, M.R. Geiker, Elemental zonation in marine
References concrete, Cem. Concr. Res. 85 (2016) 12–27, https://doi.org/10.1016/j.
cemconres.2016.02.006.
[31] C. Gallé, Effect of drying on cement-based materials pore structure as identified
[1] U.M. Angst, Challenges and opportunities in corrosion of steel in concrete, Mater.
by mercury intrusion porosimetry: a comparative study between oven-, vacuum-,
Struct. 51 (2018) 4, https://doi.org/10.1617/s11527-017-1131-6.
and freeze-drying, Cem. Concr. Res. 31 (2001) 1467–1477, https://doi.org/
[2] K. De Weerdt, D. Orsáková, A.C.A. Müller, C.K. Larsen, B.M. Pedersen, M.
10.1016/S0008-8846(01)00594-4.
R. Geiker, Towards the understanding of chloride profiles in marine exposed
[32] EN 1992-1-1:2004/A1 Design of concrete structures - Part 1-1: General rules and
concrete, impact of leaching and moisture content, Constr. Build. Mater. 120
rules for buildings, CEN, 2014.
(2016) 418–431, https://doi.org/10.1016/j.conbuildmat.2016.05.069.
[33] EN 206:2013+A2: Concrete Specification, performance, production and
[3] NORDTEST, NT BUILD 443 Concrete, Hardened: Accelerated Chloride
conformity, CEN, 2021.
Penetration, Concrete, Hardened, NORDTEST, Finland, 1995.
[34] CSA A23.1/CSA A23.2. Concrete materials and methods of concrete construction/
[4] ASTM, C1556-11a, Standard Test Method for Determining the Apparent Chloride
Test methods and standard practices for concrete, CSA Group, 2019.
Diffusion Coefficient of Cementitious Mixtures by Bulk Diffusion, ASTM
[35] T.C. Powers, T.L. Brownyard, Studies of the physical properties of hardened
International, 2016.
Portland cement paste, ACI J. Proceedings 43 (1946) 249–336, https://doi.org/
[5] EN 12390-11: Testing hardened concrete - Part 11: Determination of the chloride
10.14359/15301.
resistance of concrete, unidirectional diffusion, CEN/TC 104, 2015.
[36] D.P. Bentz, E.F. Irassar, B.E. Bucher, W.J. Weiss, Limestone fillers conserve
[6] C. Shi, Q. Yuan, F. He, X. Hu, Transport and Interactions of Chlorides in Cement-
cement; part 1: an analysis based on Powers’ model, Concr. Int. 31 (2009) 41–46.
Based Materials, CRC Press, London, 2021.
[37] P. Halamickova, R.J. Detwiler, D.P. Bentz, E.J. Garboczi, Water permeability and
[7] L. Tang, L.-O. Nilsson, P.A. Muhammed Basheer, Resistance of Concrete to
chloride ion diffusion in Portland cement mortars: relationship to sand content
Chloride Ingress – Testing and Modelling, CRC Press, London, 2017.
and critical pore diameter, Cem. Concr. Res. 25 (1995) 790–802, https://doi.org/
[8] L. Bertolini, B. Elsener, P. Pedeferri, E. Redaelli, R.B. Polder, Corrosion of Steel in
10.1016/0008-8846(95)00069-O.
Concrete: Prevention, Diagnosis, Repair, 2nd edition, WILEY-VCH, Singapore,
[38] R.A. Cook, K.C. Hover, Mercury porosimetry of hardened cement pastes, Cem.
2014.
Concr. Res. 29 (1999) 933–943, https://doi.org/10.1016/S0008-8846(99)00083-
[9] H. Justnes, A review of chloride binding in cementitious systems, Nord. Concr.
6.
Res. 21 (1998) 48–63.
[39] W. Vichit-Vadakan, G.W. Scherer, Measuring permeability of rigid materials by a
[10] Q. Yuan, C. Shi, G. De Schutter, K. Audenaert, D. Deng, Chloride binding of
beam-bending method: III, cement paste, J. Am. Ceram. Soc. 85 (2002)
cement-based materials subjected to external chloride environment – a review,
1537–1544, https://doi.org/10.1111/j.1151-2916.2002.tb00309.x.
Constr. Build. Mater. 23 (2009) 1–13, https://doi.org/10.1016/j.
[40] L. Tang, Chloride Ingress in Concrete Exposed to Marine Environment - Field Data
conbuildmat.2008.02.004.
up to 10 Years Exposure, SP Report, SP Swedish National Testing and Research
[11] H. Zibara, Binding of External Chlorides by Cement Pastes, PhD thesis
Institute, 2003, p. 68.
Department of Civil Engineering,, University of Toronto, Canada, 2001.
[41] K.L. Scrivener, A.K. Crumbie, P. Laugesen, The interfacial transition zone (ITZ)
[12] M. Balonis, B. Lothenbach, G. Le Saout, F.P. Glasser, Impact of chloride on the
between cement paste and aggregate in concrete, Interf. Sci. 12 (2004) 411–421,
mineralogy of hydrated Portland cement systems, Cem. Concr. Res. 40 (2010)
https://doi.org/10.1023/B:INTS.0000042339.92990.4c.
1009–1022, https://doi.org/10.1016/j.cemconres.2010.03.002.
[42] K. De Weerdt, M. Bernhardt, M.R. Geiker, O. Skjolsvold, K.O. Kjellsen, Influence
[13] U.A. Birnin-Yauri, F.P. Glasser, Friedel’s salt, Ca2Al(OH)6(Cl,OH)⋅2H2O: its solid
of cast and cut surface on chloride ingress resistance of concrete in marine splash
solutions and their role in chloride binding, Cem. Concr. Res. 28 (1998)
zone, in: 4th International Conference on Sustainable Construction Materials and
1713–1723, https://doi.org/10.1016/S0008-8846(98)00162-8.

19
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Technologies, Las Vegas, 2016, pp. 1561–1568, https://doi.org/10.18552/2016/ [69] A. Mesbah, C. Cau-dit-Coumes, F. Frizon, F. Leroux, J. Ravaux, G. Renaudin,
SCMT4D181. A new investigation of the Cl− –CO32− substitution in AFm phases, J. Am. Ceram.
[43] S. Fjendbo, K. De Weerdt, H.E. Sørensen, M.R. Geiker, When and how should Soc. 94 (2011) 1901–1910, https://doi.org/10.1111/j.1551-2916.2010.04305.x.
chloride profiles be calibrated for paste fraction? Nord. Concr. Res. 66 (2022) [70] A. Babaahmadi, A. Machner, W. Kunther, J. Figueira, P. Hemstad, K. De Weerdt,
1–18, https://doi.org/10.2478/ncr-2021-0021. Chloride binding in Portland composite cements containing metakaolin and silica
[44] U. Angst, B. Elsener, C.K. Larsen, Ø. Vennesland, Critical chloride content in fume, Cem. Concr. Res. 161 (2022), 106924, https://doi.org/10.1016/j.
reinforced concrete — a review, Cem. Concr. Res. 39 (2009) 1122–1138, https:// cemconres.2022.106924.
doi.org/10.1016/j.cemconres.2009.08.006. [71] F. Georget, B. Lothenbach, W. Wilson, F. Zunino, K.L. Scrivener, Stability of
[45] U.M. Angst, M.R. Geiker, A. Michel, C. Gehlen, H. Wong, O.B. Isgor, B. Elsener, C. hemicarbonate under cement paste-like conditions, Cem. Concr. Res. 153 (2022),
M. Hansson, R. François, K. Hornbostel, R. Polder, M.C. Alonso, M. Sanchez, M. 106692, https://doi.org/10.1016/j.cemconres.2021.106692.
J. Correia, M. Criado, A. Sagüés, N. Buenfeld, The steel–concrete interface, Mater. [72] S. Sui, W. Wilson, F. Georget, H. Maraghechi, H. Kazemi-Kamyab, W. Sun,
Struct. 50 (2017) 143, https://doi.org/10.1617/s11527-017-1010-1. K. Scrivener, Quantification methods for chloride binding in Portland cement and
[46] U.M. Angst, M.R. Geiker, M.C. Alonso, R. Polder, O.B. Isgor, B. Elsener, H. Wong, limestone systems, Cem. Concr. Res. 125 (2019), 105864, https://doi.org/
A. Michel, K. Hornbostel, C. Gehlen, R. François, M. Sanchez, M. Criado, 10.1016/j.cemconres.2019.105864.
H. Sørensen, C. Hansson, R. Pillai, S. Mundra, J. Gulikers, M. Raupach, [73] F. Georget, W. Wilson, K.L. Scrivener, Edxia: microstructure characterisation from
J. Pacheco, A. Sagüés, The effect of the steel–concrete interface on chloride- quantified SEM-EDS hypermaps, Cem. Concr. Res. 141 (2021), 106327, https://
induced corrosion initiation in concrete: a critical review by RILEM TC 262-SCI, doi.org/10.1016/j.cemconres.2020.106327.
Mater. Struct. 52 (2019) 88, https://doi.org/10.1617/s11527-019-1387-0. [74] Z. Shi, M.R. Geiker, B. Lothenbach, K. De Weerdt, S.F. Garzón, K. Enemark-
[47] K. De Weerdt, M.B. Haha, G. Le Saout, K.O. Kjellsen, H. Justnes, B. Lothenbach, Rasmussen, J. Skibsted, Friedel’s salt profiles from thermogravimetric analysis
Hydration mechanisms of ternary Portland cements containing limestone powder and thermodynamic modelling of Portland cement-based mortars exposed to
and fly ash, Cem. Concr. Res. 41 (2011) 279–291, https://doi.org/10.1016/j. sodium chloride solution, Cem. Concr. Compos. 78 (2017) 73–83, https://doi.
cemconres.2010.11.014. org/10.1016/j.cemconcomp.2017.01.002.
[48] F.J. Millero, R. Feistel, D.G. Wright, T.J. McDougall, The composition of standard [75] J. Tritthart, Chloride binding in cement II. The influence of the hydroxide
seawater and the definition of the reference-composition salinity scale, Deep-Sea concentration in the pore solution of hardened cement paste on chloride binding,
Res. I Oceanogr. Res. Pap. 55 (2008) 50–72, https://doi.org/10.1016/j. Cement Concr. Res. 19 (1989) 683–691, https://doi.org/10.1016/0008-8846(89)
dsr.2007.10.001. 90039-2.
[49] K. De Weerdt, B. Lothenbach, M.R. Geiker, Comparing chloride ingress from [76] Z. Shi, M.R. Geiker, K. De Weerdt, T.A. Østnor, B. Lothenbach, F. Winnefeld,
seawater and NaCl solution in Portland cement mortar, Cem. Concr. Res. 115 J. Skibsted, Role of calcium on chloride binding in hydrated Portland
(2019) 80–89, https://doi.org/10.1016/j.cemconres.2018.09.014. cement–metakaolin–limestone blends, Cem. Concr. Res. 95 (2017) 205–216,
[50] A. Machner, M.H. Bjørndal, H. Justnes, L. Hanžič, A. Šajna, Y. Gu, B. Bary, M. Ben https://doi.org/10.1016/j.cemconres.2017.02.003.
Haha, M.R. Geiker, K. De Weerdt, Effect of leaching on the composition of [77] J.J. Beaudoin, V.S. Ramachandran, R.F. Feldman, Interaction of chloride and
hydration phases during chloride exposure of mortar, Cem. Concr. Res. 153 CSH, Cem. Concr. Res. 20 (1990) 875–883, https://doi.org/10.1016/0008-8846
(2022), 106691, https://doi.org/10.1016/j.cemconres.2021.106691. (90)90049-4.
[51] W.M. Haynes, Handbook of Chemistry and Physics, 94th ed, CRC Press/Taylor [78] Y. Elakneswaran, A. Iwasa, T. Nawa, T. Sato, K. Kurumisawa, Ion-cement hydrate
and Francis, 2013. interactions govern multi-ionic transport model for cementitious materials, Cem.
[52] Y. Yang, R.A. Patel, S.V. Churakov, N.I. Prasianakis, G. Kosakowski, M. Wang, Concr. Res. 40 (2010) 1756–1765, https://doi.org/10.1016/j.
Multiscale modeling of ion diffusion in cement paste: electrical double layer cemconres.2010.08.019.
effects, Cem. Concr. Compos. 96 (2019) 55–65, https://doi.org/10.1016/j. [79] S. Yoshida, Y. Elakneswaran, T. Nawa, Electrostatic properties of C–S–H and C-A-
cemconcomp.2018.11.008. S-H for predicting calcium and chloride adsorption, Cem. Concr. Compos. 121
[53] K. Ferjaoui, Nanoscale Modelling of Ionic Transport in the Porous C-S-H Network, (2021), 104109, https://doi.org/10.1016/j.cemconcomp.2021.104109.
PhD thesis, EPFL, Switzerland, 2022, https://doi.org/10.5075/epfl-thesis-9955. [80] Y. Zhou, D. Hou, J. Jiang, L. Liu, W. She, J. Yu, Experimental and molecular
[54] K. De Weerdt, E. Bernard, W. Kunther, M.T. Pedersen, B. Lothenbach, Phase dynamics studies on the transport and adsorption of chloride ions in the nano-
changes in cementitious materials exposed to saline solutions, Cem. Concr. Res. pores of calcium silicate phase: the influence of calcium to silicate ratios,
165 (2023), 107071, https://doi.org/10.1016/j.cemconres.2022.107071. Microporous Mesoporous Mater. 255 (2018) 23–35, https://doi.org/10.1016/j.
[55] A. Machner, M. Bjørndal, A. Šajna, N. Mikanovic, K. De Weerdt, Impact of micromeso.2017.07.024.
leaching on chloride ingress profiles in concrete, Mater. Struct. 55 (2021) 8, [81] G.E. Archie, The electrical resistivity log as an aid in determining some reservoir
https://doi.org/10.1617/s11527-021-01730-w. characteristics, Trans. AIME 146 (1942) 54–62, https://doi.org/10.2118/
[56] P. Hemstad, A. Machner, K. De Weerdt, The effect of artificial leaching with HCl 942054-g.
on chloride binding in ordinary Portland cement paste, Cem. Concr. Res. 130 [82] R. Spragg, C. Qiao, T. Barrett, J. Weiss, 11 - Assessing a concrete's resistance to
(2020), 105976, https://doi.org/10.1016/j.cemconres.2020.105976. chloride ion ingress using the formation factor, in: A. Poursaee (Ed.), Corrosion of
[57] A. Machner, P. Hemstad, K. De Weerdt, Towards the understanding of the pH Steel in Concrete Structures, Woodhead Publishing, Oxford, 2016, pp. 211–238.
dependency of the chloride binding of Portland cement pastes, Nord. Concr. Res. [83] E.J. Garboczi, Permeability, diffusivity, and microstructural parameters: a critical
58 (2018) 143–162, https://doi.org/10.2478/ncr-2018-0009. review, Cem. Concr. Res. 20 (1990) 591–601, https://doi.org/10.1016/0008-
[58] F. Georget, C. Bénier, W. Wilson, K.L. Scrivener, Chloride sorption by C-S-H 8846(90)90101-3.
quantified by SEM-EDX image analysis, Cem. Concr. Res. 152 (2022), 106656, [84] W.J. Weiss, R.P. Spragg, O.B. Isgor, M.T. Ley, T. Van Dam, Toward performance
https://doi.org/10.1016/j.cemconres.2021.106656. specifications for concrete: linking resistivity, RCPT and diffusion predictions
[59] Q. Zhu, L. Jiang, Y. Chen, J. Xu, L. Mo, Effect of chloride salt type on chloride using the formation factor for use in specifications, in: D.A. Hordijk, M. Luković
binding behavior of concrete, Constr. Build. Mater. 37 (2012) 512–517, https:// (Eds.), High Tech Concrete: Where Technology and Engineering Meet,
doi.org/10.1016/j.conbuildmat.2012.07.079. Proceedings of the 2017 Fib Symposium, Held in Maastricht, the Netherlands,
[60] A. Machner, M. Zajac, M. Ben Haha, K.O. Kjellsen, M.R. Geiker, K. De Weerdt, June 12–14, 2017, Springer International Publishing, Cham, 2018,
Chloride-binding capacity of hydrotalcite in cement pastes containing dolomite pp. 2057–2065.
and metakaolin, Cem. Concr. Res. 107 (2018) 163–181, https://doi.org/10.1016/ [85] M.A.B. Promentilla, T. Sugiyama, T. Hitomi, N. Takeda, Quantification of
j.cemconres.2018.02.002. tortuosity in hardened cement pastes using synchrotron-based X-ray computed
[61] F. Georget, W. Wilson, T. Matschei, Long-Term Extrapolation of Chloride Ingress: microtomography, Cem. Concr. Res. 39 (2009) 548–557, https://doi.org/
An Illustration of the Feasibility and Pitfalls of the Square Root Law, Cem. Concr. 10.1016/j.cemconres.2009.03.005.
Res. 170 (2023) 107187, https://doi.org/10.1016/j.cemconres.2023.107187. [86] P. Panchmatia, J. Olek, N. Whiting, Electrical conductivity of concrete: the effects
[62] K. De Weerdt, Chloride binding in concrete: recent investigations and recognised of temperature, saturation and air content, Indian Concr. J. 89 (2015) 17–25.
knowledge gaps: RILEM Robert L’Hermite Medal Paper 2021, Mater. Struct. 54 [87] K.Y. Ann, S.I. Hong, Modeling chloride transport in concrete at pore and chloride
(2021) 214, https://doi.org/10.1617/s11527-021-01793-9. binding, ACI Mater. J. 115 (2018), https://doi.org/10.14359/51702194.
[63] M.D.A. Thomas, J.D. Matthews, Performance of pfa concrete in a marine [88] N. Neithalath, J. Jain, Relating rapid chloride transport parameters of concretes
environment––10-year results, Cem. Concr. Compos. 26 (2004) 5–20, https://doi. to microstructural features extracted from electrical impedance, Cem. Concr. Res.
org/10.1016/S0958-9465(02)00117-8. 40 (2010) 1041–1051, https://doi.org/10.1016/j.cemconres.2010.02.016.
[64] W. Wilson, F. Georget, K. Scrivener, Unravelling chloride transport/ [89] P. Yang, G. Sant, N. Neithalath, A refined, self-consistent Poisson-Nernst-Planck
microstructure relationships for blended-cement pastes with the mini-migration (PNP) model for electrically induced transport of multiple ionic species through
method, Cem. Concr. Res. 140 (2021), 106264, https://doi.org/10.1016/j. concrete, Cem. Concr. Compos. 82 (2017) 80–94, https://doi.org/10.1016/j.
cemconres.2020.106264. cemconcomp.2017.05.015.
[65] W. Wilson, J.N. Gonthier, F. Georget, K.L. Scrivener, Insights on chemical and [90] Y. Dhandapani, M. Santhanam, Investigation on the microstructure-related
physical chloride binding in blended cement pastes, Cem. Concr. Res. 156 (2022), characteristics to elucidate performance of composite cement with limestone-
106747, https://doi.org/10.1016/j.cemconres.2022.106747. calcined clay combination, Cem. Concr. Res. 129 (2020), 105959, https://doi.
[66] C1218/C1218M-20 ASTM. Standard Test Method for Water-Soluble Chloride in org/10.1016/j.cemconres.2019.105959.
Mortar and Concrete, ASTM International, 2020. [91] G.A. Narsilio, R. Li, P. Pivonka, D.W. Smith, Comparative study of methods used
[67] W. Wilson, F. Georget, K.L. Scrivener, Xcr/√t as an indicator of the Resistance to estimate ionic diffusion coefficients using migration tests, Cem. Concr. Res. 37
against Bulk Chloride Diffusion, Synercrete’23 - Interdisciplinary Approaches for (2007) 1152–1163, https://doi.org/10.1016/j.cemconres.2007.05.008.
Cement-Based Materials and Structural Concrete: Synergizing Expertise and [92] F.P. Glasser, J. Marchand, E. Samson, Durability of concrete — degradation
Bridging Scales of Space and Time, Greece, 2023. phenomena involving detrimental chemical reactions, Cem. Concr. Res. 38
[68] W. Wilson, F. Georget, K.L. Scrivener, Manuscript Under Preparation, 2023. (2008) 226–246, https://doi.org/10.1016/j.cemconres.2007.09.015.

20
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

[93] T. Sanchez, P. Henocq, O. Millet, A. Aït-Mokhtar, Coupling PhreeqC with electro- [122] M. Khanzadeh Moradllo, B. Sudbrink, Q. Hu, M. Aboustait, B. Tabb, M.T. Ley, J.
diffusion tests for an accurate determination of the diffusion properties on M. Davis, Using micro X-ray fluorescence to image chloride profiles in concrete,
cementitious materials, J. Electroanal. Chem. 858 (2020), 113791, https://doi. Cem. Concr. Res. 92 (2017) 128–141, https://doi.org/10.1016/j.
org/10.1016/j.jelechem.2019.113791. cemconres.2016.11.014.
[94] A. Vollpracht, B. Lothenbach, R. Snellings, J. Haufe, The pore solution of blended [123] R. du Plooy, G. Villain, S. Palma Lopes, A. Ihamouten, X. Dérobert, B. Thauvin,
cements: a review, Mater. Struct. 49 (2016) 3341–3367, https://doi.org/ Electromagnetic non-destructive evaluation techniques for the monitoring of
10.1617/s11527-015-0724-1. water and chloride ingress into concrete: a comparative study, Mater. Struct. 48
[95] G.K. Glass, N.R. Buenfeld, The presentation of the chloride threshold level for (2015) 369–386, https://doi.org/10.1617/s11527-013-0189-z.
corrosion of steel in concrete, Corros. Sci. 39 (1997) 1001–1013, https://doi.org/ [124] K. Reichling, M. Raupach, N. Klitzsch, Determination of the distribution of
10.1016/S0010-938X(97)00009-7. electrical resistivity in reinforced concrete structures using electrical resistivity
[96] G. Sergi, G.K. Glass, A method of ranking the aggressive nature of chloride tomography, Mater. Corros. 66 (2015) 763–771, https://doi.org/10.1002/
contaminated concrete, Corros. Sci. 42 (2000) 2043–2049, https://doi.org/ maco.201407763.
10.1016/S0010-938X(00)00050-0. [125] M. Fares, G. Villain, S. Bonnet, S. Palma Lopes, B. Thauvin, M. Thiery,
[97] G. Sergi, S.W. Yu, C.L. Page, Diffusion of chloride and hydroxyl ions in Determining chloride content profiles in concrete using an electrical resistivity
cementitious materials exposed to a saline environment, Mag. Concr. Res. 44 tomography device, Cem. Concr. Compos. 94 (2018) 315–326, https://doi.org/
(1992) 63–69, https://doi.org/10.1680/macr.1992.44.158.63. 10.1016/j.cemconcomp.2018.08.001.
[98] J. Crank, The Mathematics of Diffusion, 2nd edition, Clarendon Press, 1979. [126] M. Torres-Luque, E. Bastidas-Arteaga, F. Schoefs, M. Sánchez-Silva, J.F. Osma,
[99] O. Truc, J.P. Ollivier, L.O. Nilsson, Numerical simulation of multi-species Non-destructive methods for measuring chloride ingress into concrete: state-of-
diffusion, Mater. Struct. 33 (2000) 566–573, https://doi.org/10.1007/ the-art and future challenges, Constr. Build. Mater. 68 (2014) 68–81, https://doi.
BF02480537. org/10.1016/j.conbuildmat.2014.06.009.
[100] M. Geiker, E.P. Nielsen, D. Herfort, Prediction of chloride ingress and binding in [127] Y. Seguí Femenias, U. Angst, F. Caruso, B. Elsener, Ag/AgCl ion-selective
cement paste, Mater. Struct. 40 (2007) 405–417, https://doi.org/10.1617/ electrodes in neutral and alkaline environments containing interfering ions,
s11527-006-9148-2. Mater. Struct. 49 (2016) 2637–2651, https://doi.org/10.1617/s11527-015-0673-
[101] J. Marchand, E. Samson, Predicting the service-life of concrete structures – 8.
limitations of simplified models, Cem. Concr. Compos. 31 (2009) 515–521, [128] P.S. Mangat, B.T. Molloy, Prediction of long term chloride concentration in
https://doi.org/10.1016/j.cemconcomp.2009.01.007. concrete, Mater. Struct. 27 (1994) 338–346, https://doi.org/10.1007/
[102] O.P. Kari, Y. Elakneswaran, T. Nawa, J. Puttonen, A model for a long-term BF02473426.
diffusion of multispecies in concrete based on ion–cement-hydrate interaction, [129] K. Stanish, M. Thomas, The use of bulk diffusion tests to establish time-dependent
J. Mater. Sci. 48 (2013) 4243–4259, https://doi.org/10.1007/s10853-013-7239- concrete chloride diffusion coefficients, Cem. Concr. Res. 33 (2003) 55–62,
3. https://doi.org/10.1016/S0008-8846(02)00925-0.
[103] S. Pradelle, M. Thiéry, V. Baroghel-Bouny, Comparison of existing chloride [130] D. Boubitsas, L. Tang, P. Utgenannt, Chloride Ingress in Concrete Exposed to
ingress models within concretes exposed to seawater, Mater. Struct. 49 (2016) Marine Environment -Field Data up to 20 Years’ Exposure, SP Swedish National
4497–4516, https://doi.org/10.1617/s11527-016-0803-y. Testing and Research Institute, 2014.
[104] V.Q. Tran, A. Soive, V. Baroghel-Bouny, Modelisation of chloride reactive [131] S. Engelund, General Guidelines for Durability Design and Redesign, vol 15 of
transport in concrete including thermodynamic equilibrium, kinetic control and Document BE95–1347/R, DuraCrete: Probabilistic performance based durability
surface complexation, Cem. Concr. Res. 110 (2018) 70–85, https://doi.org/ design of concrete structures, European Union Brite EuRam III, CUR, Gouda,
10.1016/j.cemconres.2018.05.007. Netherlands, 2000.
[105] O.B. Isgor, W.J. Weiss, A nearly self-sufficient framework for modelling reactive- [132] F. Georget, J.H. Prévost, B. Huet, Reactive transport modelling of cement paste
transport processes in concrete, Mater. Struct. 52 (2019) 3, https://doi.org/ leaching in brines, Cem. Concr. Res. 111 (2018) 183–196, https://doi.org/
10.1617/s11527-018-1305-x. 10.1016/j.cemconres.2018.05.015.
[106] D. Conciatori, É. Grégoire, É. Samson, J. Marchand, L. Chouinard, Sensitivity of [133] F. Georget, J.H. Prévost, B. Huet, Impact of the microstructure model on coupled
chloride ingress modelling in concrete to input parameter variability, Mater. simulation of drying and accelerated carbonation, Cem. Concr. Res. 104 (2018)
Struct. 48 (2015) 3023–3036, https://doi.org/10.1617/s11527-014-0374-8. 1–12, https://doi.org/10.1016/j.cemconres.2017.11.008.
[107] C. Andrade, Calculation of chloride diffusion coefficients in concrete from ionic [134] M.B. Clennell, Tortuosity: a guide through the maze, Geol. Soc. Lond., Spec. Publ.
migration measurements, Cem. Concr. Res. 23 (1993) 724–742, https://doi.org/ 122 (1997) 299–344, https://doi.org/10.1144/GSL.SP.1997.122.01.18.
10.1016/0008-8846(93)90023-3. [135] S. Chatterji, Transportation of ions through cement based materials. Part 1
[108] O. Truc, J.P. Ollivier, M. Carcassès, A new way for determining the chloride fundamental equations and basic measurement techniques, Cem. Concr. Res. 24
diffusion coefficient in concrete from steady state migration test, Cem. Concr. Res. (1994) 907–912, https://doi.org/10.1016/0008-8846(94)90010-8.
30 (2000) 217–226, https://doi.org/10.1016/S0008-8846(99)00232-X. [136] E. Samson, J. Marchand, J.J. Beaudoin, Describing ion diffusion mechanisms in
[109] E. Samson, J. Marchand, K.A. Snyder, Calculation of ionic diffusion coefficients cement-based materials using the homogenization technique, Cem. Concr. Res. 29
on the basis of migration test results, Mater. Struct. 36 (2003) 156–165, https:// (1999) 1341–1345, https://doi.org/10.1016/S0008-8846(99)00101-5.
doi.org/10.1007/BF02479554. [137] S. Torquato, Random Heterogeneous Materials: Microstructure and Macroscopic
[110] fib, Model Code for Service Life Design, Fib Bulletin 34, CEB-FIB, 2006. Properties, Springer-Verlag, 2002.
[111] fib, Fib Model Code for Concrete Structures 2010, Ernst & Sohn2013. [138] B. Bary, A polydispersed particle system representation of the porosity for non-
[112] ISO 16204, Durability — Service Life Design of Concrete Structures, 2012. saturated cementitious materials, Cem. Concr. Res. 36 (2006) 2061–2073,
[113] S. Sui, F. Georget, H. Maraghechi, W. Sun, K. Scrivener, Towards a generic https://doi.org/10.1016/j.cemconres.2006.07.001.
approach to durability: factors affecting chloride transport in binary and ternary [139] E. Stora, B. Bary, Q.-C. He, On estimating the effective diffusive properties of
cementitious materials, Cem. Concr. Res. 124 (2019), 105783, https://doi.org/ hardened cement pastes, Transp. Porous Media 73 (2008) 279–295, https://doi.
10.1016/j.cemconres.2019.105783. org/10.1007/s11242-007-9170-z.
[114] S. Fjendbo, H.E. Sørensen, K. De Weerdt, M.R. Geiker, The square root method for [140] K. Bourbatache, O. Millet, A. Aït-Mokhtar, O. Amiri, Modeling the chlorides
chloride ingress prediction—applicability and limitations, Mater. Struct. 54 transport in cementitious materials by periodic homogenization, Transp. Porous
(2021) 61, https://doi.org/10.1617/s11527-021-01643-8. Media 94 (2012) 437–459, https://doi.org/10.1007/s11242-012-0013-1.
[115] S.L. Poulsen, H. Sørensen, Chloride Ingress in Old Danish Bridges, 2nd [141] R.A. Patel, Q.T. Phung, S.C. Seetharam, J. Perko, D. Jacques, N. Maes, G. De
International Congress on Durability of Concrete (ICDC), New Dehli, India, 2014. Schutter, G. Ye, K. Van Breugel, Diffusivity of saturated ordinary Portland
[116] S.L. Poulsen, H. Sørensen, U. Jönsson, Chloride ingress in concrete blocks at the cement-based materials: a critical review of experimental and analytical
Rødbyhavn marine exposure site—status after 5 years, in: 4th International modelling approaches, Cem. Concr. Res. 90 (2016) 52–72, https://doi.org/
Conference on Service Life Design for Infrastructures (SLD4), Delft, the 10.1016/j.cemconres.2016.09.015.
Netherlands, 2018. [142] A.T. Horne, I.G. Richardson, R.M.D. Brydson, Quantitative analysis of the
[117] H. Justnes, M.O. Kim, S. Ng, X. Qian, Methodology of calculating required microstructure of interfaces in steel reinforced concrete, Cem. Concr. Res. 37
chloride diffusion coefficient for intended service life as function of concrete (2007) 1613–1623, https://doi.org/10.1016/j.cemconres.2007.08.026.
cover in reinforced marine structures, Cem. Concr. Compos. 73 (2016) 316–323, [143] A. Leemann, R. Loser, B. Münch, Influence of cement type on ITZ porosity and
https://doi.org/10.1016/j.cemconcomp.2016.08.006. chloride resistance of self-compacting concrete, Cem. Concr. Compos. 32 (2010)
[118] V.G. Papadakis, C.G. Vayenas, M.N. Fardis, A reaction engineering approach to 116–120, https://doi.org/10.1016/j.cemconcomp.2009.11.007.
the problem of concrete carbonation, AICHE J. 35 (1989) 1639–1650, https:// [144] H.S. Wong, M. Zobel, N.R. Buenfeld, R.W. Zimmerman, Influence of the
doi.org/10.1002/aic.690351008. interfacial transition zone and microcracking on the diffusivity, permeability and
[119] A. Leemann, F. Moro, Carbonation of concrete: the role of CO2 concentration, sorptivity of cement-based materials after drying, Mag. Concr. Res. 61 (2009)
relative humidity and CO2 buffer capacity, Mater. Struct. 50 (2017) 30, https:// 571–589, https://doi.org/10.1680/macr.2008.61.8.571.
doi.org/10.1617/s11527-016-0917-2. [145] Z. Wu, H.S. Wong, N.R. Buenfeld, Influence of drying-induced microcracking and
[120] V. Baroghel-Bouny, M. Dierkens, X. Wang, A. Soive, M. Saillio, M. Thiery, related size effects on mass transport properties of concrete, Cem. Concr. Res. 68
B. Thauvin, Ageing and durability of concrete in lab and in field conditions: (2015) 35–48, https://doi.org/10.1016/j.cemconres.2014.10.018.
investigation of chloride penetration, J. Sustain. Cement-Based Mater. 2 (2013) [146] A.C.A. Muller, K.L. Scrivener, A.M. Gajewicz, P.J. McDonald, Densification of
67–110, https://doi.org/10.1080/21650373.2013.797938. C–S–H measured by 1H NMR relaxometry, J. Phys. Chem. 117 (2013) 403–412,
[121] O.M. Jensen, A.M. Coats, F.P. Glasser, Chloride ingress profiles measured by https://doi.org/10.1021/jp3102964.
electron probe micro analysis, Cem. Concr. Res. 26 (1996) 1695–1705, https:// [147] E. Berodier, J. Bizzozero, A. Muller, Mercury intrusion Porosimetry, in:
doi.org/10.1016/S0008-8846(96)00158-5. K. Scrivener, R. Snellings, B. Lothenbach (Eds.), A Practical Guide to

21
K. De Weerdt et al. Cement and Concrete Research 173 (2023) 107287

Microstructural Analysis of Cementitious Materials, CRC Press, 2016, [149] K. Scrivener, T. Matschei, F. Georget, P. Juilland, A. Kunhi-Mohammed,
pp. 419–444. Hydration and Thermodynamics, ICCC Bangkok, Cement and Concrete Research
[148] H. Ranaivomanana, J. Verdier, A. Sellier, X. Bourbon, Prediction of relative Special Issue, 2023 (submitted 2023).
permeabilities and water vapor diffusion reduction factor for cement-based [150] U.M. Angst, O.B. Isgor, C.M. Hansson, A. Sagüés, M.R. Geiker, Beyond the
materials, Cem. Concr. Res. 48 (2013) 53–63, https://doi.org/10.1016/j. chloride threshold concept for predicting corrosion of steel in concrete, Appl.
cemconres.2013.02.008. Phys. Rev. 9 (2022), 011321, https://doi.org/10.1063/5.0076320.

22

You might also like