1 Introduction

In comparison with other areas of research related to construction technology and building materials—such as digital fabrication or environmentally friendly cementitious binder systems—the science and engineering of corrosion of steel in concrete has a history of more than half a century. The advantage of this is that comparatively well documented long-time experience is available for corrosion and the related processes, mainly transport processes through the concrete cover—at least for traditional building materials. Regarding the durability of reinforced concrete, the processes under study are slow, i.e. they typically need time frames of years or decades to manifest. Thus, in this field, long-time documentation is highly valuable.

The downside of the topic having been tackled for several decades, however, is that concepts became deeply rooted in the research community and are generally taken for granted in daily research endeavors. It is common to study certain aspects inherent to the established concepts rather than to fundamentally question their benefit for science and engineering. One of the most prominent examples in this regard, which will also be discussed in more detail in this review, is the concept of the critical chloride content or chloride threshold value for corrosion initiation. Although it is widely known that the predictive power of this conceptual approach used to forecast the time to corrosion initiation in chloride-bearing environments is poor, the author of this article knows of no serious attempts to quit the established concept or replace it with another one.

This paper highlights a number of challenges related to corrosion of steel in concrete. Section 2 discusses societal, economic, technological, as well as educational challenges, and Sect. 3 summarizes the state-of-the-art, highlighting the areas where scientific questions still remain open. Finally, these challenges also present great opportunities in research and education for the coming generations of researchers working in the field of steel corrosion in concrete; these are discussed in Sect. 4.

2 Challenges

2.1 Societal and economic challenges

A wide variety of civil infrastructure is built in reinforced concrete (RC): bridges, tunnels, marine structures, sewer systems, dams, etc. In fact, concrete is the world’s most used man-made material [1, 2]. Many of the built engineering structures are of crucial importance for a smooth and reliable functioning of economy and personal live. Infrastructure is a key factor in the competitiveness of economies; neglecting infrastructure maintenance or infrastructure expansion were recently highlighted to be among the main reasons limiting economic growth and global competitiveness in the case of the United States of America (U.S.) or India, respectively [3].

While there are many examples of excellent durability of reinforced concrete, even under harsh exposure conditions, a number of deterioration mechanisms may limit the service life of RC structures. The by far most common cause of degradation and thus poor durability of RC is corrosion of the reinforcement steel. This was apparent from different reviews, analyzing practical cases of premature deterioration of RC structures, where it was concluded that in 70–90% of the evaluated cases corrosion was the dominant degradation mechanism [4, 5].

Premature degradation not only concerns the safety and quality (traffic jams, infrastructure availability, etc.) of human life, the required repair work also negatively affects the environment due to the increased consumption of energy and materials. Additionally, corrosion of infrastructures causes high costs to the society [6,7,8,9]. In 2002, a comprehensive survey of the economic impact of metallic corrosion in the U.S. was published [6]. It was concluded that the overall direct costs arising from corrosion is 3.1% of the gross domestic product (GDP). With 2016s GDP, this figure becomes $ 575 × 109—annually. Note that the numbers given are only estimates of the direct costs, while indirect costs such as lost productivity, traffic jams, or negative impact on the environment, are not included. As shown in Fig. 1, approximately half of the direct costs estimated in [6] were attributed to corrosion of infrastructures such as highway bridges, gas and oil pipelines, drinking water distribution, sewer systems, and electrical utilities, while the other half is mainly due to corrosion of airplanes, ships, motor vehicles, and industrial manufacturing facilities.

Fig. 1
figure 1

Economic impact of corrosion [6] and other selected major issues in the U.S. Regarding corrosion costs attributable to concrete structures, the dark green area represents the costs for concrete highway bridges and the light green area indicates an extrapolation to other types of concrete structures not covered in [6] (parking garages, retaining walls, tunnels, etc.). The numbers for health issues include direct medical costs related to diagnosis and treatment of health conditions caused by cigarette smoking [10] and obesity [11]. The data for weather and climate disasters [12] include physical damage to residential, commercial, and public buildings and infrastructures, vehicles, agricultural assets, and time element losses (e.g. hotel-costs for loss of living quarters). All data is representative for the years 2010–2011

Full size image

Figure 1 also shows a comparison of the corrosion costs to other issues with large economic impacts such as the annual healthcare spending attributable to cigarette smoking [10] and obesity [11], or the direct costs arising from weather and climate disasters [12]. While these health- and climate-related issues are well recognized as major societal challenges and frequently appear in the media and in political debates in many regions in the world, the tremendous financial burden caused by corrosion receives comparatively little attention, particularly in Europe.

Additionally, the continuous aging of our infrastructures will aggravate this situation in the years to come. In industrialized countries, the age profile of reinforced concrete infrastructures typically reflects a peak in construction activities around 1960–1980. Thus, most of the bridges have an age of 40 or more years and they were designed for a service life of 50 years [13]. Consequently, in the coming decades, the number of bridges that have to be operated beyond their design life will steeply increase. Based on construction dates of Dutch highway bridges and experience about time to corrosion induced damage as well as average repair lives, Polder et al. [14] estimated that the bridges needing repair will in the Netherlands increase over the coming 20 years by a factor of 2–4, and over the coming 40 years even by a factor of 3–6. Thus, the age profile of infrastructures in industrialized countries is expected to lead to an additional increase in corrosion costs over the coming decades.

2.2 Technological challenges

A committee recently established by the National Research Council of the U.S., given the task to identify challenges in corrosion research and engineering, concluded that the high societal costs of corrosion (Fig. 1) directly reflect the severe lack of fundamental understanding of corrosion-related degradation of materials exposed to the environment [15]. Reasons for this include the multidisciplinarity, the multi-scale nature, the absence of pure and well-defined materials, the wide variability of exposure situations, the long time spans, and the huge gap between laboratory and real conditions. The lack of fundamental understanding is also highlighted by the facts that policies to deal with corroding RC structures differ markedly from one country to another (despite the technical similarity of the problem) [16], and that engineering strategies continue relying on past experience and are largely empirical [14,15,16]. The presentations and the panel discussion with recognized experts at the international RILEM workshop on present and future durability challenges for reinforced concrete structures, held in Zurich, Switzerland, in 2012 [16], concluded that there is a strong need for a knowledge-based approach to the issue of predicting the durability of RC.

Being able to forecast the long-term durability performance of RC structures in their actual environments is on the one hand crucial for maintenance planning of existing structures, and on the other hand for designing durable and sustainable new structures (Fig. 2). Although these two challenges arise in all countries, the focus is clearly different among different regions in the world. Industrialized countries typically mainly face the problem of dealing with existing, ageing infrastructures that were built mostly in the second half of the last century, while expanding the current infrastructure constitutes a minor portion in these nations’ construction activity. The current lack of knowledge (Sect. 3) forces engineers to take conservative decisions, i.e. to repair or replace structures relatively early. Considering the expected increase in needed repair works over the coming decades [14], there is certainly an urgent need to quit the current relatively conservative and thus costly approach. This can only be achieved by providing innovative and cost-effective technologies as well as fundamental understanding of corrosion deterioration mechanisms.

Fig. 2
figure 2

Technological challenges related to corrosion of steel in concrete are different in different regions in the world. Industrialized countries (a) are mainly challenged to maintain their stock of existing infrastructures, while emerging countries (b) expanding their infrastructures are challenged to design sustainable and durable new structures, where the ever-increasing diversity in construction materials and processes does not anymore permit relying on empirical long-term experience (c). d Summarizes the technological challenges highlighting the need for a more science-based approach

Full size image

In emerging countries, on the other hand, there is still a large need to expand infrastructures [3], and thus their main challenge lies in the design of durable new structures (Fig. 2). In these regions, a huge positive sustainability impact can potentially be made if proper construction materials are used, i.e. by combining low environmental footprint of the material with long service lives. Research endeavors have over the last decades increasingly focused on developing new cement types with decreased CO2 emissions [1, 17]. Concretes produced with these cementitious materials differ from traditional ones in terms of microstructure and chemistry of the pore liquid. In comparison to Portland cement concretes, they typically have lower pore solution pH, lower pH buffer capacity due to the partial or complete lack of Portlandite, which clearly negatively impairs the protection of the embedded steel from corrosion. On the other hand, these materials may have lower porosity, increased pore tortuosity, or better chloride binding capacity, which slows down chloride transport through the concrete cover and thus may have a beneficial impact on the corrosion performance. The state-of-the-art does, however, not permit reliable performance prognoses of these materials in their actual exposure environments, which clearly handicaps holistic sustainability and life cycle assessments. The poor predictive power [16, 18,19,20,21,22] of current models and accelerated testing can be traced to insufficient theoretical grounds (see Sect. 3).

The overarching problem is that in predicting the actual performance of structures built with modern and innovative construction materials and construction processes, one cannot any longer rely on empirical long-term experience from practice. Such experience is available for “traditional” materials, i.e. mainly Portland cement concretes (perhaps also slag concretes) and cold-work hardened reinforcing steels. For all materials introduced after, say 1990, there is very little or no documentation of long-term performance (Fig. 2). Needless to say, for all materials under development or to be developed in the near future, it appears impossible to gain sufficient experience from practical exposure (tests) before approving these materials for large-scale applications in engineering.

In summary, both for existing and for new structures, the challenge clearly lies in making long-term (~ 100 years) predictions for field performance, based on short-term (~ a few months) laboratory-scale testing. Section 3 summarizes the state-of-the-art regarding the science and engineering of dealing with RC structures exposed to corrosive environments. A scientific, knowledge-based approach to forecast the durability performance, particularly the corrosion behavior, is clearly needed.

2.3 Educational challenges

It is the author’s opinion that the current generation of professionals in civil engineering are generally not sufficiently well trained to adequately tackle the technological challenges described above. Interestingly, different analyses by major institutions such as by the National Research Council of the U.S. [15, 23] and by NACE International [6] came to the conclusion that this applies to the entire field of corrosion challenges (Fig. 1), and not only to corrosion of steel in concrete.

Concrete structures are primarily dealt with by civil engineers, who are typically well trained in assessing the structural (mechanical) behavior, but have limited education in materials science related matters. Assessing the condition of ageing reinforced and pre-stressed concrete structures as well as specifying cost-efficient repair works are not straightforward and generally present situations where no off-the-shelf-solutions can be used. This is because existing structures differ strongly in terms of materials used, depending on the time of construction [24], regional owner policies or available commercial products [16]. It is also generally recognized that exposure conditions and histories may exhibit significant variability, and that a number of degradation mechanisms may be at hand simultaneously. Corrosion of reinforcing steel and particularly of pre-stressing steel is under certain conditions not visible at the concrete surface, at least not in early stages of corrosion propagation; this presents a severe limitation for traditional routine inspections that are typically done visually [25]. Additionally, condition assessment is also not straightforward due to the large number of destructive and non-destructive test methods available [26,27,28], some of which are based on relatively simple principles, and some which have more complex theoretical backgrounds. Selecting the right methods and interpreting the results properly, particularly when it comes to assessing the corrosion condition and making an assessment of the future corrosion evolution (e.g. in terms of corrosion rates), requires profound knowledge and mechanistic understanding of concrete durability and corrosion. Regarding the construction of new structures, on the other hand, if performance-based approaches are to be used in order to favor innovative solutions (Fig. 2), the increasing level of complexity in understanding and judging the durability of different systems poses higher demands on the education of civil engineers than the prescriptive approach of current standards.

Figure 3 summarizes groups of professionals needed in order to address the challenges summarized in Fig. 2 on different levels. In view of the large societal and economic impact of RC corrosion (Fig. 1), there is no doubt about the urgent need for experts in the field of corrosion of steel in concrete (in Fig. 3 termed “RC corrosion experts”). Ideally, this group of professionals has profound understanding about corrosion and concrete materials science that is universal enough to be applied to non-standard situations, i.e. situations where purely experience based approaches cannot be used. This group of professionals should also have sufficient background to write codes, standards and technical rules to ensure durability for modern and future commercial products and construction technologies, guarantee proper applications of condition assessment methods, etc. Ideally, these corrosion experts have a university degree in civil engineering, materials science, chemistry, metallurgy, or similar, followed by thorough and prolonged study of a corrosion subject such as in the form of a PhD thesis. The current situation, however, is that there are only relatively few professionals who had this sort of training. Generally, many of the professionals taking the role of “RC corrosion experts” have often no or very little formal education in corrosion, but gained experience through on-the-job-training [23].

Fig. 3
figure 3

Different groups of professionals urgently needed in order to tackle the grand challenge of RC infrastructure ageing in industrialized countries and to ensure the design and construction of durable infrastructures in emerging regions. The education ideally needed for the different groups of professionals is also indicated, and the arrows illustrate the roles that these professionals may take in educating other groups. On the right, the current situation is summarized, which contrasts with the ideal situation and thus highlights the educational challenges that industrialized countries typically face. The figure was developed based on Ref. [23]. RC = reinforced concrete

Full size image

This can be traced, among other reasons, to the limited opportunities for specialized education in RC corrosion, which directly results from the decreasing number of university professors, capable of offering research positions in the field and in mentoring and training PhD students (in Fig. 3 termed “RC corrosion scientist”). For example, this may be illustrated by the number of technical university professors, carrying out scientific research and teaching the issue of reinforcing steel corrosion in concrete at their institutions, which in the German speaking part of Europe (Germany, Austria, Switzerland) has decreased from 7 in the early 2000s to 3 by today—due to retirements, followed by replacements of faculty members with expertise in other fields. It is thus the author’s opinion that there is a need for creating more faculty positions in the field of corrosion and concrete durability. To enhance transfer of knowledge to engineering, this should preferably be in engineering rather than materials science departments. Additionally, this would ensure a minimum level of corrosion education of future generations of civil engineers as the topic of concrete durability and corrosion could be integrated better in curricula at B.Sc. and M.Sc. level, which is currently rarely the case [23].

It is considered important that not only so-called experts, but also the broad workforce in civil engineering has an improved awareness of corrosion and durability problems, as this is generally needed for consulting, tendering, design, etc. This group of professionals is in Fig. 3 termed “consulting engineer/owner dealing with RC infrastructures”, which intends to indicate that not only civil engineers in general, but also owners of public or private assets will benefit from increased durability knowledge. Increased awareness would also allow this group of professionals to take more educated decisions about whether or not to involve more specialized corrosion experts in certain situations. The same applies also the last group of professionals shown in Fig. 3, i.e. those working primarily on-site, for instance inspectors (routine inspections on behalf of public or private owners), technicians carrying out certain types of condition assessment measurements, but also site managers—as it is well known that execution has a pronounced influence on the achieved durability [29].

In summary, it is the authors opinion that currently, there is an insufficient number of adequately trained professionals to deal with the grand challenges related to the durability of RC infrastructures. This can be explained by the decreasing and currently low—with respect to the tremendous societal and economic relevance—number of faculty positions at universities capable of leading research projects and educating professionals in the field of corrosion in RC. These positions are not only needed to provide the scientific basis in order to address the technological challenges such as the questions related to novel cementitious materials (see Sect. 2.2), but also to provide more opportunities for training of RC corrosion experts and to integrate the topic of concrete durability and corrosion in the education of civil engineers at BSc and MSc level.

3 State-of-the-art

3.1 General remarks about corrosion of steel in concrete

Concrete is a porous solid that contains an alkaline liquid phase. In traditional concrete made with Portland cement, the pH is buffered well above 12.5. This is the undisputed reason why embedded carbon steel, as such thermodynamically unstable in the presence of water, is protected from corrosion [28, 30, 31]. The protection mechanism is the formation of a passive film that reduces the rate of oxidation to a technically negligible level.

Corrosion can be triggered by the ingress of chloride—e.g. due to exposure to seawater or road deicing salts—and/or by carbonation of the concrete [28]. In his work published in 1982, Tuutti [32] proposed the well-known schematic service life model of a RC structure; a modified version of it is shown in Fig. 4. This conceptual model distinguishes two stages: the initiation stage describes the transport of the corrosive agents through the concrete to the steel until conditions are met for corrosion initiation; the propagation stage describes the process of corrosion and accumulation of related damage until a limit state and thus the end of the service life is reached. Note that corrosion onset is not considered an instant, but rather a period of time [33,34,35,36]. Particularly for chloride-induced corrosion, it is well known that the stable (self-sustaining) growth of anodic sites requires a number of conditions to be fulfilled, such as sufficient supply of chloride ions from the surrounding concrete [35, 37]. Thus, at early stages of corrosion initiation, the steel may repeatedly repassivate, and further accumulation of chlorides is needed to promote active and stable corrosion. Once corrosion propagates, the kinetics may be affected by a number of factors such as pore clogging due to precipitation of corrosion products or other solid phases and self-healing effects in the case of corrosion in cracked zones [38], that would both reduce the corrosion rate, or by concrete cracking and spalling that would accelerate corrosion. This leads to a non-linear corrosion propagation as schematically shown in Fig. 4. The instantaneous corrosion rate is also affected by seasonal or other changes in exposure environment, such as varying moisture conditions [28]. The limit states for corroding RC structures are the loss of steel sectional area or corrosion-induced cracking of the concrete; both can affect the load bearing capacity and structural integrity, and thus the safety of a structure.

Fig. 4
figure 4

Schematic illustration of the life of a reinforced concrete structure, distinguishing the initiation stage, corrosion onset, and the corrosion propagation stage (adapted from Tuutti [32]). Green and red areas indicate where the state-of-the-art is well advanced, and where fascinating research questions still remain open, respectively

Full size image

In traditional research, initiation of corrosion due to chlorides and due to carbonation are usually treated separately [28]. Measuring and modeling the penetration of the so-called carbonation front or of chlorides through the concrete has over the last decades received considerable research attention. Reliable methods are available in international standards and allow the quantification of these transport processes through the concrete cover both in laboratory specimens and in engineering structures [28, 39,40,41,42]. Additionally, predictive models of different degrees of complexity—ranging from fully empirical to sophisticated, mechanistic models—were proposed and some of them provide reasonably accurate predictions, both for carbonation [43,44,45,46,47] and for chloride transport [42, 44, 48,49,50]. For more detailed information about reactive transport modeling it is here referred to the cited literature. However, models and methods to predict the penetration of the carbonation front and of chlorides through the concrete cover are here considered relatively well developed, particularly in comparison to the state-of-the-art related to corrosion initiation and corrosion propagation, as reviewed in the next sections.

3.2 Initiation of reinforcement corrosion

In the case of carbonation of the concrete, it is normally assumed that corrosion initiates as soon as the carbonation front reaches the reinforcing steel, i.e. once the pH value of the concrete pore solution approaches neutrality [28]. To predict corrosion onset in the case of chloride-induced corrosion, it has been the concept for many years to define a certain chloride threshold value or critical chloride content (Ccrit) as the threshold chloride concentration associated with corrosion onset. Over the last 60 years considerable research efforts have been made to determine the value and/or the distribution of Ccrit, as reviewed in Ref. [51]. The literature results, however, scatter over several orders of magnitude, and thus offer a poor basis for service life modeling, particularly considering that Ccrit is one of the most decisive model input parameters. One reason for the literature scatter is that it is experimentally not straightforward to determine Ccrit; in fact, the experimental methodology has long been subject to controversial debate in the literature [51]. This was also reflected within the work of RILEM technical committee 235-CTC [52].

The conceptual understanding of Ccrit has over many years been that different values for Ccrit may be experimentally measured for different cement types, steel grades, etc. However, numerous studies have indicated that effects not systematically related to the materials under study play a major role in corrosion initiation. In fact, the susceptibility to corrosion was found to strongly depend on the local conditions at the steel–concrete interface. These often called “defects” are recognized to offer preferable corrosion initiation sites. Examples are air voids or other macroscopic voids [53, 54], the bleed-water zone at the underside of reinforcement steel [55, 56], or cracks [57]. Regarding the role of voids in corrosion initiation, contradictory results were published in the literature, i.e. air voids at the steel surface were found to present preferential sites for corrosion initiation [53, 54, 58], while the contrary was observed in other experiments [34, 59]. In this regard, the moisture conditions within the concrete pore system have recently been identified to play a crucial role [24, 60], but without doubt, more research is needed to clarify the influence of voids on corrosion initiation. Other factors in corrosion initiation are the steel surface condition, e.g. the presence of mill-scale or rust layers [61, 62]. Also metallurgic features such as lattice defects or inclusions are well known to influence initiation of pitting corrosion [63]. With respect to metallurgy, recent studies shed light on the importance of the steel microstructure—as originating from different production processes such as cold-work hardening or thermal strengthening—on the corrosion susceptibility [64,65,66], a factor that is generally ignored, but may in fact be highly relevant since most of the reinforcing steel from which long-term field experience is available was manufactured with a different process than the steel used today.

While the importance of various local “defects” and “irregularities” at the steel–concrete interface is widely recognized, their relative influence is not well understood. As a result, they are not accounted for in current predictive models. A review [24] of documented properties of steel–concrete interfaces in laboratory specimens and in engineering structures highlighted that the steel–concrete interface is complex, prone to significant variability from one case to another, and that it may be difficult to mimic conditions representative for engineering structures in laboratory specimens. In recognition of this, a test method was recently suggested to measure the statistical distribution of Ccrit in samples taken from engineering structures and to assess the effect on corrosion initiation of various local characteristics of the steel–concrete interface that can be found in structural concrete [67].

The poor predictive power of the concept of Ccrit may be illustrated with help of results obtained with this test method from a more than 40 year old Swiss road tunnel located in the Alps [59, 68]. Concrete cores of diameter 150 mm, containing a segment of reinforcing steel, were drilled from the ceiling of the tunnel concrete lining that was still free from corrosion, and subsequently tested in the laboratory. The obtained results yielded a probability distribution of Ccrit, that was for the present purpose fitted with a lognormal distribution (parameters λ = E(lnCcrit) = − 0.21, ɛ2 = var(lnCcrit) = 0.74, and first two moments μ = E(Ccrit) = 0.91% chloride by weight of cement, σ2 = var(Ccrit) = 0.48% chloride by weight of cement). This was based on 11 concrete cores all sampled within a concrete area of less than 2 m2, thus presumably uniformly exposed and from the same batch of construction materials.

Figure 5 shows the results of a simple prediction of the time to corrosion initiation (all assumptions given in the figure caption). It is apparent that the predicted time to corrosion initiation is dispersed over large time spans. Corrosion—though with a relatively low probability—may start after a few years, but that it may also take many hundreds or even thousands of years without corrosion. On the one hand, this can be explained by the fact that the statistical distribution of Ccrit, in this case measured on an engineering structure, spans a much wider range than the distributions observed in laboratory specimens (e.g. the data reported in Refs. [69, 70]). On the other hand, this is a direct result of the fact that the increase in chloride concentrations in concrete tends to slow down with time, which thus amplifies any uncertainties in Ccrit when converted into time to corrosion. In other words, it is the unfortunate combination of the two facts, (1) the wide variability of realistic Ccrit and (2) the slow increase in chloride concentration at the cover depth after a few decades, which together lead to huge dispersions in predicted service life, such as of the sort of Fig. 5. Figure 5 also illustrates the strong sensitivity of this model approach to input parameters: even small uncertainties in transport coefficients or the cover depth disperse the predicted time to corrosion over as much as one order of magnitude.

Fig. 5
figure 5

Example of a prediction of the time to corrosion initiation for realistic statistical distribution of Ccrit, illustrating the dispersion of the prediction over a wide range and thus highlighting the severe limitations of the current conceptual approach to forecast chloride-induced steel corrosion in concrete. A simple chloride transport model [42, 69] was used to predict C(x,t), i.e. the chloride content at depth x and time t, the equation of which is displayed in the figure, and with the following assumptions for the calculations represented by the solid line: apparent chloride diffusion coefficient (at time to) Da,o = 10−12 m2/s, reference time to = 28 days, ageing exponent n = 0.4, cover depth x = 40 mm, chloride content at surface Cs = 3% chloride by cement weight. The time to corrosion was calculated as the time until the predicted chloride content at the cover depth equals Ccrit. Ccrit was treated as stochastic variable, the distribution of which was obtained from measurements with a novel test method [67] applied to 11 samples taken from within a small concrete surface area (< 2 m2) of a > 40 years old engineering structure [59, 68]. The shaded area illustrates the strong sensitivity of the chloride transport model to the input parameters. The dashed curves result from making the following changes: decreasing/increasing Da,o by a factor of 2, or decreasing/increasing the ageing exponent by 25%, or decreasing/increasing the cover depth by 30%

Full size image

Ideally, to facilitate the interpretation and application in engineering practice, a predictive model should exhibit a much steeper probability distribution than the curves shown in Fig. 5. The fact that the prediction is strongly dispersed indicates that the parameter, on which the prediction is essentially based (chloride), does not have a dominating influence on the actual phenomenon of corrosion initiation. In other words: reducing the problem to a mere question of chlorides seems not to provide adequate predictions. Other, also or even more relevant parameters, are overlooked. The common way to deal with this unfortunate situation is to define an acceptable probability of corrosion, rather than to look for better ways to predict corrosion initiation. However, the selection of a specific value for this acceptable failure probability is not straightforward, particularly not for localized corrosion, as has been discussed in more detail in Refs. [71, 72].

As an additional problem complicating the use of the concept of Ccrit for service life predictions, it has been experimentally shown that the specimen size affects the susceptibility to pitting corrosion [73,74,75]. This was explained by the increasing probability for the presence of preferable initiation sites with increasing specimen size. Nevertheless, this “size effect” of corrosion in concrete directly raises the question of the “correct size to be used in service life modeling”, i.e. the question of extrapolating small laboratory-scale results to large, real-structure dimensions. This has been addressed in Ref. [71], where it was discussed that the selection of the size used in corrosion predictions should ideally be based on structural considerations, taking into account the load carrying behavior of the actual structural member. A suggestion was also made how Ccrit can be translated from one size into another [71, 75]. Finally, taking into account the size effect in corrosion predictions will also give a better understanding of the physical meaning of the acceptable failure probability.

Nevertheless, after now 60 years of intense research, one may conclude that the poor understanding and predictive power of the current concept of Ccrit call for a new approach. Reducing the issue of corrosion initiation to a question of chlorides only is an oversimplification of reality and does not permit reliable performance forecasting.

3.3 Propagation of reinforcement corrosion

3.3.1 Quantitative models to predict the propagation stage (corrosion kinetics)

A variety of models have been proposed to predict the corrosion rate of steel in concrete, as reviewed e.g. in Refs. [76, 77]. Some models are simply based on an empirical correlation of the corrosion rate to a relatively easily measureable parameter such as the concrete resistivity, but they were shown to have a poor predictive power [19]. More refined approaches involve finite element or boundary element modeling by relying on mathematical expressions for the anodic and cathodic partial reactions as well as for the ohmic current flow in the electrolyte [78,79,80]. The polarization behavior of the partial reactions is typically described with Butler–Volmer equations (or simplified, e.g. linearized, relationships) and the corresponding coefficients are generally considered time-invariant. Nevertheless, it is well known that the kinetics both of the anodic and the cathodic partial reactions can significantly change when conditions at the steel–concrete interface change. Examples are the influence of local changes in electrolyte chemistry (pH, oxygen concentration, etc.) or half-cell reactions becoming controlled by mass transport [35, 81, 82]. Causes for this may be that the concrete pore system close to the anodic sites becomes blocked by precipitated corrosion products or that the supply of chloride ions to the corrosion pit gets restricted by a limited reservoir of available chloride (release of bound chlorides) [35, 56, 81]. Rigorous descriptions of the transport and precipitation of corrosion products in the pore system and the resulting transport limitations that in turn affect the corrosion kinetics are currently not available. This lack of knowledge presents a serious barrier to developing mechanistic models for the corrosion propagation stage.

3.3.2 Corrosion-induced concrete cracking

Precipitation of corrosion products in the concrete pores may give rise to expansive stresses and lead to concrete cracking and spalling. Cracks are undesirable for numerous reasons, e.g. because they promote further ingress of species that may trigger additional deterioration mechanisms. Corrosion-induced cracking is thus a common limit state in service life considerations. Additionally, cracks play a major role in routine inspections of RC structures, which are typically based on visual assessment and where cracks are the only considered indicator of internally ongoing corrosion. However, there are documented cases of reinforcement corrosion that led to structural collapse (due to loss of steel sectional area), but without any outwardly visual signs of such as cracks or rust stains [25]. It is thus essential to know the conditions under which corrosion leads to cracking and to be able to predict at what stage of the corrosion process this occurs.

A well-known reason for delayed cracking is the presence of chlorides, as these strongly increase the solubility of corrosion products [83] and therefore permit migration and diffusion through the pore system and away from the steel. Thus, for chloride-induced corrosion, precipitation of rust, if any, will thus only occur either distant from the corroding spot (preferably in the internal parts of a structural member [25]) and under conditions not leading to expansive stresses. Precipitation probably also only occurs after relatively high amounts of iron have been released into the concrete pore system, i.e. late in the propagation stage, where already considerable loss of steel cross section has occurred. If corrosion occurs at lower chloride concentrations, i.e. due to carbonation, precipitation of corrosion products is likely to occur earlier, much closer to the steel surface and under conditions that can lead to expansive stresses. This is also favored by the uniform corrosion morphology, which reduces the concrete volume that can accommodate corrosion products. Nevertheless, the fact that already very small amounts of chloride have a pronounced effect on the precipitation of corrosion products [83] implies that the role of chloride should not be neglected in modeling cracking for carbonation-induced corrosion. Even in the absence of chloride exposure, concrete normally contains small amounts of chloride that may enhance the solubility of corrosion products and permit to some extent diffusion and migration away from the steel. Further indication that at least minor amounts of chloride play a role also in carbonation-induced corrosion is the fact that—considering the classical Pourbaix diagram [84] of iron in the absence of chloride—steel would still be passive at the potentials typical for aerated concrete [28] down to pH values as low as 5 or 6, which contradicts all field observations.

A recent analysis [20] of existing models to predict corrosion-induced concrete cracking has indicated that none of the numerous proposed models can reliably predict the time-to-cracking. Typically, the models provided relatively accurate predictions for the experimental results they were fitted to. However, if they are used to model the time-to-cracking for data of other studies, the predictions were found to be poor. One of the reasons is that these models are calibrated with accelerated laboratory tests that are neither representative for real conditions nor comparable to other laboratory tests. Moreover, they use strongly simplified input data to take into account the corrosion rate. As there is a general agreement that the concrete pore structure can accommodate a certain amount of precipitated corrosion products before expansive pressure is exerted [85, 86], predictive models typically include a parameter termed “porous zone” or “diffusion zone” which describes an idealized cylindrical void around the steel bar that first needs to be filled with corrosion products before expansive pressure arises [20]. The expansive pressure that finally leads to concrete cracking is usually explained simply by the volume increase of (precipitated) corrosion products by a factor of ca. 2–7 with respect to the originally solid iron [85]. However, several investigations have shown that the corrosion products are deposited around the steel, but that they also penetrate the concrete matrix and precipitate in the pore system far from the steel [87,88,89].

In short, the conceptual understanding forming the state-of-the-art for modeling corrosion-induced concrete cracking is from a rather global perspective and does not address the fundamental processes in sufficient detail. For instance, the relative kinetics of the rate at which ferrous ions are released at the steel surface (corrosion rate) with respect to the rate of transport into the matrix is crucial in determining the location of precipitation. Under conditions of accelerated testing, e.g. by using impressed currents, the rate of release of ferrous iron is typically several orders of magnitude higher than rates observed in reality. This means that local (super)saturation around the reinforcement steel bar, an important condition for mechanical damage [90, 91], will be reached faster under accelerated corrosion since there is less time for the released ferrous ions to move away than under natural conditions.

Indeed, it is well known that precipitates will only exert expansive stresses if they are forced to grow under confined conditions, such as against pore walls [90,91,92,93,94]. Additionally, this growth must occur under supersaturated conditions, which dictate the crystallization pressure at the pore scale. Finally, the macroscopic stress will only arise once a certain volume fraction of pores is filled by the crystallization compound. The common approach simply based on a volume expansion is thus a strongly simplified view of the problem and it is suggested that future research endeavors to predict corrosion-induced concrete cracking take into account the dual thermodynamic and poromechanical nature of expansive pressure and macroscopic stress.

3.4 Condition assessment of engineering structures

3.4.1 General remarks

Many considerable improvements were made over the past decades with respect to non-destructive test methods for reinforced and pre-stressed concrete structures. Various well-proven test methods are available such as for measuring concrete cover depths or for locating reinforcing steel and tendons even at large depths [95], or for the quantification of durability indicating parameters such as the specific electrical concrete resistivity [96]. The main challenges in assessing the condition of existing concrete structures are the detection of corrosion and the quantification of the rate of corrosion, as well as the condition of pre-stressing steel, particularly in post-tensioned structures inside ducts.

3.4.2 Detecting corrosion initiation

Actively corroding steel in concrete can generally be reliably distinguished from passive steel by means of potential measurements [97]. There exists a number of different national and international standards for potential mapping on concrete structures [98,99,100], and although they differ in philosophy, they generally permit locating the zones where corrosion is actively on-going. It is considered a major success that this method was introduced to engineering practice and forms today, in many countries, an important part of assessing the condition of concrete structures, providing the basis for the design and scheduling of maintenance works.

A major factor driving the costs of the application of potential mapping and other non-destructive test methods in practice is the limited accessibility to structures, e.g. the undersides of bridge girders (Fig. 6), which demands costly equipment and laborious procedures to perform measurements on the respective concrete surfaces. It is expected that novel approaches utilizing drones and robots [101] will over the coming years solve this problem and thus greatly reduce the cost of potential mapping and other non-destructive condition assessment methods, thereby further promoting their application (compare Sect. 3.4.5).

Fig. 6
figure 6

Example of the inspection (here: potential mapping) of the underside of a bridge girder, illustrating that costly equipment and laborious procedures are needed to perform the measurements on the concrete surface of interest. This is one of the main factors driving the costs of condition assessments. Future use of robots and drones will considerably reduce the time and equipment needed to apply non-destructive test methods on engineering structures and thus enhance the quality (e.g. spatial resolution) and quantity of information about the (corrosion) condition of structures

Full size image

3.4.3 Instantaneous corrosion rate measurements

A major challenge presents the non-destructive, reliable determination of the instantaneous corrosion rate, particularly in localized corrosion, where relatively high corrosion rates can occur due to the galvanic element formed between small anodic and comparatively large cathodic areas. Many electrochemical measurement procedures exist [7], most of them relying on the polarization resistance, Rp. The so-called linear polarization resistance method (LPR) makes use of the inverse proportionality between Rp and the instantaneous corrosion rate, known as the Stern–Geary equation [102], which was derived under several assumptions, one of them being the uniform corrosion morphology [103]. Despite this, the LPR method has since the late 1970s been widely applied to macro-cell corrosion of steel in concrete both in research and practice [104,105,106,107,108], and even commercial devices are available. Given the sometimes large discrepancies between the results of the LPR method and other techniques (e.g. gravimetric measurements), researchers have been concerned with a variety of possible error sources. One of the most frequently addressed issue is that of “current confinement” [107,108,109] where—with help of refined instrumentation (guard rings)—it is attempted to manipulate the flow of current through the concrete in order to control the “polarized steel surface area”, knowledge of which then supposedly allows conversion of the determined corrosion current into a current density (~ corrosion rate). The benefit of confinement techniques, however, is subject to controversial discussions in the literature [107].

Surprisingly, little attention seems however to have been given to the fact that the Stern–Geary equation is applied to pronounced situations of localized corrosion, although it is explicitly only valid for uniform corrosion. This fundamental discrepancy was discussed in recent publications by the present author [110, 111]. While a relationship between Rp and the corrosion rate may still exist, it must be acknowledged that this is a purely empirical one, which can be traced to the fact that well-defined mixed potentials according to theory of Wagner and Traud [103] do simply not exist in macro-cell corrosion and thus, the position of the reference electrode in measuring setup, and hence the recorded signal, becomes arbitrary [110, 111]. Further research should thus aim at developing alternative methods to quantify the corrosion rates, preferentially based on sound corrosion theory. If empirical approaches are to be used, it is considered crucial to provide more detailed information about the conditions under which such methods provide adequate results.

The need for methods allowing to reliably quantify the rate of local loss of steel sectional area in corroding concrete structures will rise over the coming years and decades. This is due to the increase in the fraction of engineering structures starting to corrode and needing repair. Reliable data on corrosion rates will permit scheduling maintenance works in a more refined manner than with today’s relatively conservative approaches. This will be needed in order to dilute the expected increase in structures needing repair simultaneously [14] and thus diminish the economic impact of infrastructure aging.

3.4.4 Pre-stressed and post-tensioned concrete

Finally, a major challenge is the assessment of the condition of steel tendons in pre-stressed and post-tensioned structures. Poor grouting of ducts (i.e. the presence of voids), chloride-containing grouts, or perforated ducts permitting penetration of corrosive species from the environment to the steel strands are well-known factors that may seriously affect the corrosion condition of steel tendons [28]. While advances have been made with respect to non-destructive inspection methods [112,113,114], the reliable assessment of pre-stressed and post-tensioned concrete structures still poses difficulties in many practical situations.

3.4.5 Structural health monitoring and automated inspection

A number of sensors have been developed over the past years to measure durability-relevant parameters on engineering structures. This includes embeddable reference electrodes [115], sensors for chloride concentrations [116,117,118,119] and pH values [120,121,122] of the concrete pore solution, and sensors for the measurement of concrete resistivity or galvanic currents [123]. Although the use of such sensors in the context of monitoring has early been demonstrated to provide valuable data [124, 125], there is to-date no widespread application of these sensors in engineering practice. Increasing their use as embedded probes in engineering structures in the context of structural health monitoring would, combined with the advances made in information technology, data transmission, etc., would permit to generate large amounts of valuable durability-related data. Semi-automated systems would be needed to process and pre-analyse the data in order to provide useful information for engineers. It is believed that such approaches would permit to better asses the condition of infrastructures, namely by following the evolution over time of different corrosion-relevant parameters.

As mentioned above, also the inspections of structures performed at regular time-intervals would benefit strongly from automation, namely through utilizing drones and robots [101, 126, 127]. Drones may be used for recording images of the concrete surface (without touching the structure), followed by automated image analyses, e.g. for crack detection. Additionally, if able to touch the concrete, climbing robots and flying drones may also perform electrochemical and other non-destructive measurements on structures, which is of particular advantage in difficult to access areas of structural members (an example is shown in Fig. 6).

The author believes that automating electrochemical measurements combined with image-based inspections of engineering structures will considerably reduce the costs related to the conventional approach of condition assessment and corrosion inspection. This is because flying drones and climbing robots render heavy equipment such as lifting platforms and mobile inspection trucks obsolete. Furthermore, they reduce the personnel involved in the measurement, and speed up the measurement by reducing much of the manual work (marking the grid, documenting visual appearances, etc.), which reduces the costs further. In addition to reducing costs, manpower as well as measuring and processing time, the use of automated monitoring and inspections will also reduce the negative impact on the availability of the road networks for the end-user.

Finally, automating inspections will also make it easier to apply non-destructive testing in practice. This will favor its integration in routine inspections of reinforced concrete structures. In contrast to conventional routine inspections relying essentially on visual assessments, this permits the detection of corrosion in early stages, i.e. when the corrosion damage is still small and can be repaired more cost-efficient than at later stages. This will lead to additional cost-savings for the owners, such as public owners of infrastructures, and thus in the end for society.

4 Concluding remarks and opportunities

This paper has highlighted a number of societal, economic, technological, and educational challenges related to corrosion of steel in concrete. On the one hand, the enormous financial impact of infrastructure corrosion was illustrated, which, on the other hand seems to be inadequately balanced by educational and scientific research activities. Based on this, as well as on a review of the state-of-the-art, several opportunities for research and education were identified as outlined below:

  • Societal and economic opportunities The tremendous societal relevance of the progressive infrastructure ageing and corrosion presents a unique opportunity to industrialized countries for maintaining or improving their competitiveness. Mastering the here highlighted technological challenges better in the future than today would permit making significant financial savings and therefore free resources, which will increasingly be bound in infrastructure maintenance, for other societally relevant purposes.

  • Educational opportunities The practically non-existent formalized corrosion education of civil engineers and other professionals working with reinforced concrete infrastructures presents an opportunity for universities and other educational institutions as well as professional associations to play a lighthouse role in this field.

  • Technological opportunities The main technological challenges (Fig. 2) are on the one hand to design durable new structures, i.e. to reliably predict the long-term corrosion performance of RC structures in their actual environments, particularly for modern materials and in the absence of long-term experience. On the other hand, there is an ever-increasing need to cost-effectively maintain existing, ageing RC structures, i.e. to abandon conservative, experience-based decision taking and instead move towards knowledge based selection and design of repair actions. Opportunities are:

    1. 1.

      Favoring innovative solutions—both for the maintenance of existing and the design of new structures—through abandoning prescriptive approaches and instead use performance-based approaches. To reach this aim, however, substantial improvements in service life modeling still need to be made.

    2. 2.

      Embracing the technology of structural health monitoring by installing durability sensors on structures, combined with semi-automated analysis of the large amount of produced data, presents a major opportunity for refined condition assessments of structures, but also gaining experience with modern systems in different exposure conditions.

    3. 3.

      Innovative inspection methods: A major potential is seen in enhancing routine inspections by enriching them with the application of non-destructive test methods on structures with help of flying drones and climbing robots.

  • Scientific opportunities The following issues (Fig. 4) present major areas where improvements of the state-of-the-art are not only urgently needed, but will also permit making the most significant impact in tackling the highlighted technological challenges:

    1. 1.

      A major research opportunity is seen in investigating the steel–concrete interface and in understanding the role of the many different local characteristics in corrosion initiation. This is currently addressed in RILEM TC 262-SCI [128], created in 2014, but more research efforts are certainly needed to make the leap forward needed in this regard.

    2. 2.

      Recent advances in reactive transport modeling and poromechanics present fascinating opportunities for tackling the problem of predicting corrosion-induced concrete cracking and for predicting corrosion propagation in concrete, i.e. taking into account effects such as concentration polarization of the steel, pore clogging and crack self-healing in the concrete, etc.

    3. 3.

      Further improvements are still to be made with respect to non-destructive condition assessment methods, particularly with respect to reliably determining corrosion rates on engineering structures and assessing the corrosion state of pre-stressed and post-tensioned concrete structures.