3.1 Change of performance with time.

The early lifetime of made-made materials, structures and systems is often characterized by a high probability of failure. It takes some time to overcome inevitable teething problems and to reach the required level of maturity and stability. Once that point is reached a ‘quiet’ period follows until we arrive again at a period of increasing probability of failure. Exceeding a certain predefined probability of failure then marks the end of the service life of a structure or system. The high probability of failure in the beginning, the subsequent period of ‘rest’ and the subsequent period of increasing probability of failure can be presented with the bathtub curve (Figure 1a).

Figure 1 Evolution of the probability of failure (bath tube curve; left) and of performance (right) of complex systems (after ^{Van Breugel, 2014}) 

In essence the bathtub curve also applies to our fixed assets, even though it has not very often been used for infrastructure. The length of the period in which the probability of failure is low is of crucial importance for the economic performance of these assets. The bathtub curve suggests that this period is a period during which ‘nothing happens’. It is a period of ‘rest’, or ‘dormant’ period. Assuming that in the period of low probability of failure nothing happens, however, is misleading. If there would really be ‘rest’, what could then be the driving force behind the increase of probability of failure with elapse of time? To illustrate the foregoing reasoning, it may help to put the bathtub curve of Figure 1a upside down, as shown in Figure 1b. On the vertical axis we now put ‘Performance’ instead of the ‘Probability of failure’. After a short period of teething problems the material, structure or system has reached the required (high) level of performance. That is the level at which the material should demonstrate its capacity to meet safety and functional criteria, if possible without intervention for maintenance or repair. It is the period of ‘top-level sport’ for all the basic building blocks, i.e. atoms, molecules and interfaces, from which a material or structure is made. When these basic building blocks give up and leave their position, the period of decay begins. Then ageing has started! These first tiny steps of decay will most probably not be observed at the macroscale immediately.

The moment that the first basic building blocks give up can only be captured with comprehensive and appropriate material models at subsequent levels of observation. Here chemistry, physics, electrochemistry, mechanics and mathematics meet each other and need each other for developing tools for describing and predicting ageing processes at a fundamental level.

3.2 Driving forces behind ageing - A closer look.

Ageing has been defined is a change of performance of a material, structure or system with time. How time per se can result in a change of performance is not easy to understand at first sight. How can a material ‘at rest’ change its performance with time? A closer look at any piece of matter ‘at rest’ tells us that the status of rest only applies to a certain length scale. Going down to the atomic scale the world is in motion all the time! Fundamental entities, i.e. basic building blocks, are continuously moving with a certain probability to leave their position for one that fits them better. This phenomenon takes place in the time domain. It is an inherent feature of matter and lies at the basis of ageing of materials. On top of this inherent feature we see, at different scales, a number of gradients, which may cause the basic building blocks of matter to start moving. Gradients are the driving forces causing changes in the material with elapse of time. Note that at the boundary of any piece of material with its environment gradients do exist. These gradients concern, for example, temperature, humidity and radiation and they may cause changes at the surface of the material.

The foregoing illustrates that a material ‘at rest’ is hardly conceivable. Al lower scales there is motion all the time and a variety of gradients cause the basic building blocks of matter to change their position. In essence, this holds for all materials and systems. Basic building blocks strive for a position (energy level), where they feel more comfortable. By designing materials in a smart way, i.e. by minimizing internal gradients and concentrations of stresses and strains, there will be less reason for basic building blocks to leave their position. Hence, the ageing process will slow down and the service life of materials, structures and systems will increase.

4.1 Performance of bridge decks - An inventory.

In March 2001, the results of a most interesting study were published by ^{Mehta (Mehta and Burrows, 2001}). He analysed the performance of bridge decks of bridges built in four subsequent periods in the twentieth century. The first period was the period before 1930, the second between 1930 and 1950, the third from 1950 to 1980 and the fourth from 1980 to present. The concrete mixtures used for the bridge decks were characterized by the chemical composition and the fineness of the cement. The cements used in the first period, before 1930, had a C₃S content less the 30% and a Blaine surface of 180 m²/kg. Consequently, the rate of hydration was low. The performance of many of the bridge decks made with these cements was quite good.

The cements used in the second period were ground to a Blaine fineness between 180 and 300 m²/kg. The construction and building technology used for the bridge decks was similar to those used in the first period. The authors report that the bridge decks built in the second period were less durable than those built before 1930.

The structures that were built between 1950 and 1980 appeared to have more durability problems than those built before 1950. The cements used in this period had a fineness up to 400 m²/kg and a C₃S content beyond 60%. With the aim to get a denser and more durable concrete the w/c ratio was lower than in the first two periods. The higher C₃S content and the higher fineness of the cement had increased the early strength of these mixtures. This made it possible to build faster. This, however, had resulted in a higher probability of early-age thermal cracking and, on top of that, higher autogenous shrinkage of the low water-cement ratio mixtures. The higher proneness to early-age (micro)cracking was the most plausible reason for durability problems at later ages.

In the fourth period the tendency to go for higher strengths continued. Generally, this was realized by using mixtures with a low w/b ratio. The use of low w/b mixtures further increased the risk of cracking. For bridge decks, moderate strengths between 30 and 45 MPa were found. Among 29 bridge decks the cracking in 44 MPa bridge decks was twice that in 31 MPa bridge decks.

4.2 Mix design and proneness to ageing.

Mehta’s study of the performance of bridge decks illustrates how the pressure from the market to build faster has created a demand for mixtures with a high early strength. This was possible by using finer cements with a higher C₃S content. The price of this, however, was a higher probability of early-age cracking of the bridge decks.

For realizing slenderer and more elegant structures a higher final strength is required. High strength is attainable by reducing the w/b ratio. The use of (super) plasticizers has made it possible to reduce the w/b ratio of concrete mixtures to values even below 0.2. With these low w/b mixtures dense concretes are obtained with low permeability. This is considered good for the concrete’s durability. At the same time, however, we see an increase in the concrete’s proneness to (micro)cracking, mainly because of increased autogenous shrinkage.

Another reason for a higher cracking risk of high strength and ultra-high strength concretes are the high temperatures that occur because of high cement contents. By optimizing the particle packing of the aggregate fractions the amount of cement, and hence the peak temperatures, can be reduced. A low cement content is also considered positive from the sustainability point of view (lower carbon footprint of the fresh concrete mixture). A low cement content, however, also has a drawback. A low cement content reduces the inherent self-healing capacity of the concrete. From the self-healing point of view a not too low cement content and the use of ‘old’, coarsely ground cement is favorable. This partly explains the outcome of Mehta’s study that old bridge decks performed better than newer ones. In the terminology of this paper we would say that the old concrete mixtures with coarse cements with a low C₃S content were less prone to ageing than modern mixtures with finely ground cements with a high C₃S content.

5.1 Shrinkage - Influencing factors.

Several mechanisms have been proposed as possible causes of autogenous shrinkage and/or contributing factors. The most commonly reported mechanisms are capillary tension (in the range of high internal relative humidity), changes in the disjoining presses (in medium range relative humidity) and changes of surface tension of the solid gel particles. A common parameter in all these mechanisms is the internal relative humidity. With progress of the hydration process the water in the mixture is gradually consumed and the relative humidity drops. This so-called ‘internal drying’ goes along with an increase in the capillary pressure in the pore water and, when the relative humidity decreases further, changes in the disjoining pressure. Consequently, the volume of the cement paste decreases, which is known as autogenous shrinkage.

Shrinkage of the drying paste is restrained by the aggregate particles in the mixture. Whether the restraint of autogenous shrinkage strains will cause (micro)cracking depends on the size and stiffness of the aggregate particles and on the time dependent properties (creep, relaxation) of the hardening paste. How internal curing of concrete mixtures can be used to prevent a drop of the relative humidity, and hence of autogenous shrinkage, will be discussed in the next section.

The fact that the evolution of autogenous shrinkage appears to be strongly correlated with a drop in the internal relative humidity does not mean that the magnitude of autogenous shrinkage can be related directly to the relative humidity. The type of cement has turned out to be an important parameter as well (^{Tazawa and Miyazawa, 1997}). In the very early stage of hydration some types of cement exhibit swelling. This observation is of utmost importance if it comes to the interpretation of shrinkage measurements. Researchers should be aware of the fact that, particularly in the early stage of the hydration process, measured shrinkage strains are the net result of simultaneous expansion and shrinkage mechanisms. In those cases, attributing measured shrinkage strains to one single mechanism will lead to completely wrong conclusions about the underlying mechanisms and, consequently, to wrong measures to prevent or mitigate autogenous shrinkage.

From this brief overview, we learn that a number of parameters affect the magnitude of autogenous shrinkage. By manipulating these parameters, the consequences of autogenous shrinkage can be mitigated and hence the susceptibility of concrete mixtures to ageing. In the next sections emphasis will be on autogenous shrinkage and internal curing of HPC and how internal curing reduces autogenous shrinkage.

5.2 Autogenous shrinkage in C55/65 mixtures and internal curing.

As indicated in the previous section, autogenous shrinkage of low w/c mixtures can be reduced by internal curing. Internal curing can be accomplished by adding water-saturated lightweight aggregate (LWA) particles to the concrete (^{Zhutovsky et.al., 2001}). When the internal RH drops, the water stored in the LWA particles is released to the drying matrix, thus maintaining the RH at a relatively high level. A similar effect can be achieved with mixed-in super absorbing polymers (SAP), a technology promoted by ^{Jensen (2013}) and subject of the RILEM committee 225-SAP (^{Mechtcherine and Reinhardt, 2012}).

In the following results of studies on autogenous shrinkage of concrete and the effect of internal curing will be presented. First test results on autogenous shrinkage of concrete mixtures C55/65 are discussed, followed by results obtained with mixtures C28/35 and C35/45.

5.2.1 Background of the study. In the eighties and nineties of the past century the use of High Performance Concrete (HPC) with target strength C55/65 was considered for several concrete bridges in The Netherlands. The prevailing Dutch design code did not require designers to consider autogenous shrinkage of those mixtures. However, the owner of the bridges, the Dutch Ministry of Transportation, required a check of the overall performance of the mixtures, including a check of the autogenous shrinkage and the effectiveness of internal curing for mitigating the risk of early-age cracking.

5.2.2 Mixture design and test specimen. Four mixtures were tested with w/b ranging from 0.34 to 0.39. The mix compositions are given in Table 2. In mixture I, 60 kg limestone powder was used, while the amount of cement was reduced by the same amount. In mixture IV, 25% of the coarse aggregate was replaced by water-saturated lightweight aggregate, Liapor F10. The autogenous shrinkage was measured on sealed specimens, 100×100×400 mm³.

5.2.3 Measured autogenous deformation and evaluation.

The autogenous deformations of the mixtures are presented in Figure 2. The measurements started after 1 day. This implies that the very first part of the autogenous deformation was not recorded. This was not considered a problem, since the aim of the test series was to quantify how autogenous deformation would affect drying shrinkage. In the practice drying shrinkage will generally not start during the first day after casting. For the purpose of this study it was appropriate, therefore, to measure only the autogenous deformation after 1 day.

Figure 2 Autogenous deformation mixture I to IV. 20ºC. Measurements starting after 1 day (^{Van Breugel et.al., 2000}) 

The shrinkage curves of the mixtures I, II and III show that a greater part of autogenous shrinkage occurs in the first few days after mixing. But even after 28 days autogenous shrinkage still continues. From 28 to 91 days the autogenous shrinkage of the mixtures I, II and III varies from 70 to 90 (m/m. Replacing 25% of the dense aggregate by water-saturated lightweight aggregate particles was sufficient to eliminate autogenous shrinkage of this paste. Obviously, the internal curing by using saturated lightweight aggregate particles (Liapor F10, 4-8 mm) is very effective.

5.3 Autogenous shrinkage of traditional concrete mixtures C28/35 and C35/45.

The high autogenous shrinkage of mixture C55/65, much higher than expected, was reason enough to start an investigation on the autogenous shrinkage of traditional concrete mixtures with w/b ratios between 0.44 and 0.50, strength classes C35/45 and C28/35. In a preliminary study on autogenous shrinkage of concrete mixtures with w/c ≈ 0.45 ^{Van Cappellen (2009)} found that particularly at early ages the autogenous shrinkage of concrete mixtures made with blast-furnace slag cement developed faster than that of OPC-mixtures. At 200 days the difference was not very large anymore. Van Cappelle’s study was continued by ^{Mors (2011)} for mixtures made with two types of aggregate, i.e. limestone and quartz. The mixture compositions are given in Table 3. Figures 3 and 4 show the autogenous shrinkage of the traditional mixtures T (0.50) and T (0.44) made with quartz aggregate and the mixtures N (0.50) and N (0.46) made with limestone aggregate. The shrinkage curves convincingly show that also mixtures with w/b in the range from 0.44 to 0.5 exhibit substantial autogenous shrinkage. More importantly, also these mixtures exhibit ongoing autogenous shrinkage at ages beyond 28 days, the age at which the concrete is generally assumed to have reached a high degree of maturity already!

Figure 3 Autogenous shrinkage of traditional mixtures T(0.50) and T(0.44). Quartz aggregate. w/c = 0.5 and 0.44 (^{Mors, 2011}; ^{Van Breugel et.al., 2013}) 

Figure 4 Autogenous shrinkage of mixtures N(0.50) and N(0.46). Limestone aggregate. w/c = 0.46 and 0.50 (after ^{Mors, 2011}; ^{Van Breugel et.al., 2013}) 

5.4 Autogenous Shrinkage, Drying Shrinkage and Design Codes.

In the past experimental studies on drying shrinkage of concrete have often been carried out on 28 days old specimens and the measured shrinkage strains were mostly interpreted as drying shrinkage. From the autogenous shrinkage curves presented in the previous sections we have to conclude, however, that after 28 days autogenous shrinkage cannot be ignored, also not for mixtures with water-cement ratios higher than 0.4. For those mixtures, the contribution of autogenous shrinkage to the measured shrinkage strains in drying specimens has often been neglected. This means that in the past many drying shrinkage tests might have been misinterpreted. A substantial part of the strains measured on drying specimens should have been attributed to autogenous shrinkage. In recent updates of design codes autogenous shrinkage is now explicitly mentioned, also for traditional mixtures with w/b > 0.4. In the new EuroCode 2 and the Japanese code autogenous shrinkage is considered also for mixtures in strength classes < C55/65. For mixtures with w/b 0.44 - 0.50, however, these codes still underestimated the autogenous shrinkage, at least for the tested mixtures and cement types considered in the previous sections. Figure 5 shows autogenous shrinkage according to both the EuroCode 2 and the JSCE-Code, together with the measured autogenous shrinkage of normal strength concretes C28/35 (T(0.50)). Both the measured autogenous shrinkage and the curve after correction for small moisture loss through the sealant are presented. The autogenous shrinkage according to EuroCode 2 is presented for C28/35 and C35/45 mixtures, i.e. mixtures with strengths similar to the measured strength of the mixtures considered in this paper. In both cases the autogenous shrinkage according to EuroCode 2 is about 30% of the measured autogenous shrinkage. Underestimation of autogenous shrinkage by EuroCode 2 was also observed by ^{Darquennes et al. (2012}). The predictions with the Japanese code are closer to the measured values, but still underestimate the measured autogenous shrinkage.

Figure 5 Comparison of measured autogenous shrinkage with predictions with the EuroCode 2 and the Japanese Code (after ^{Mors, 2011}). 

5.5 Shrinkage and Ageing.

Autogenous and drying shrinkage go along with the evolution of internal stresses. In both cases the cement paste is the shrinking component. All shrinkage-induced stresses are subject to relaxation. Relaxation of stresses, however, is not ‘for free’. It requires the re-structuring of smallest building blocks of matter. In other words: the material ages! As far as autogenous shrinkage strains are concerned it has been proposed that the observed long-term deformations might be creep strains following the elastic shrinkage strains exerted by the capillary forces in the pore water. Also, these creep deformations are not ‘for free’, but require re-structuring of the material’s elementary building blocks: The material ages! For quantitative analysis of creep and relaxation ^{Wittmann (1977}) has applied the activation energy concept, which is considered the most appropriate approach for fundamental research of ageing phenomena of cement-based materials.

6.1 Materials design.

Ageing is an inherent feature of materials. Solutions for ageing problems require, therefore, interventions at fundamental materials level. For coping with ageing problems, two approaches are conceivable, i.e. the preventive approach and the reactive approach.

In the preventive approach the focus is on designing homogenous materials with as few as possible internal gradients, stress concentrations and interfaces. For heterogeneous materials, like concrete, this is a big challenge. When going through the subsequent length scales, from (sub-) nano to meso level, concrete behaves as a complex system. To state it differently, concrete is a ‘product of the mind’ (^{McCarter, 2009}), of which the properties are determined by the properties of the individual components and of the interfaces between them. Some of these components - in fact all! - change with time, and so do the properties of the interfaces. This makes heterogeneous materials susceptible to ageing.

In the reactive approach the heterogeneity of the material, and hence the internal concentrations of stresses and strains and the occurrence of internal damage and ageing, are considered a matter of fact. If ageing is unavoidable indeed, self-healing could be a solution for ageing problems. When dealing with concrete the presence of still unhydrated cement particles provides an inherent self-healing capacity. In this respect concrete made with a coarse cement is considered favourable to concrete made with a fine cement. ^{Mehta’s (2001}) observation that old bridge decks, built with coarse cement, have performed better than the younger ones built with finer cements, could be explained, at least in part, by the role of self-healing in the older structures. The modern trend to, firstly, use finer cement to speed up the rate of strength gain and, secondly, reduce the amount of cement to reduce the CO₂ footprint of concrete, may work out negative on the material’s resistance against ageing! In these cases, a comprehensive life cycle analysis is needed for weighing all the pros and cons of modern trends in concrete mix design.

6.2 Ageing and design codes.

For designing and realizing concrete structures design codes are indispensable. From the numerous buildings and fascinating construction works realised in the past a high degree of maturity of these codes can be inferred. In section 5.4 we have seen, however, that currently used prescriptive codes fall short in describing long-term performance, i.e. shrinkage, of concrete structures. In this respect, it is interesting to reflect on the recent tendency to switch from prescriptive codes to performance-based codes. The question is whether it is to be expected that with this switch ageing issues will be considered more appropriately and will become part of an integral design approach for concrete structures. Strictly speaking the change from prescriptive to performance-based codes is a return to the origin of the building profession. In the ancient past the whole building process was in the hands one person: the builder. The builder had the integral responsibility to meet all the safety and functional criteria set by the owner. How the builder managed to meet the owner’s criteria was not prescribed in detail. This was all considered the builder’s expertise and responsibility. In his classical book on building technology Vitruvius (85-20 BC) stated that ideally the whole building process should be in the hands of one person. When Vitruvius wrote his book, a few years BC, he noticed already that this ideal situation was no longer tenable. The building process became too complicated and one single person could not be an expert in all areas of the building process. Gradually the builder had to share his responsibility with others. This situation started the emerge of certificates etc., and, later on, prescriptive codes. The user of these documents could be held responsible for a correct interpretation of and compliance with the codes, but not for the content of the codes.

Prescriptive codes can be judged as the ultimate consequence of a process of increasing fragmentation of the building process and, more importantly, of the vision that everything, including quality, is engineerable. The huge sustainability problems we face today, however, illustrates that this vision has lost most of its convincing power. Prescriptive codes, even the most detailed ones, are necessary, but insufficient for guaranteeing quality and/or sustainability. Prescriptive codes deal with materials properties with the primary goal to provide the designer with data needed for designing safe structures. Any change of performance of the materials with time is considered a time dependent property without addressing the cause of these changes.

With performance-based codes the building process has been given back to the builder, including the challenge to accomplish (long-term) quality criteria and sustainability goals. The builder’s freedom to decide how to meet these criteria and goals may stimulate the builder to invest in fundamental research of traditional and new building materials and in innovative design concepts. Furthermore, performance-based codes, in combination with new DBFM (Design-Built-Finance-Maintenance) contracts, will also force the builder to focus on both the short-term and long-term performance, i.e. ageing, of materials and structures. For that purpose, the builder will need reliable predictive models, including models for quantifying the rate of ageing processes and the consequences thereof.