Required research for predicting AAR expansion: the Ichikawa theory revisited

Research on alkali–aggregate reaction (AAR) expansion has been ongoing since Stanton's pioneering study in 1940 (Stanton, 1940), with numerous contributions from Rilem technical committees. Although these committees have tried to develop appropriate test methods that reproduce expansion behaviour in the field and are able to assess the safety of concrete mixtures, no satisfactory methods have been introduced. In 2017, the situation of global AAR research activity was described by Sims and Pool (2017). The primary goals of AAR research are to prevent detrimental expansion in new structures and to assess the potential for future expansion in existing ones, along with the associated loss of structural functionality. Before evaluating a final structure's function, it is essential to evaluate AAR expansion, including various mitigation methods. However, there are discrepancies between the evaluation of expansion in the laboratory and experience in the field. In order to improve laboratory tests, basic understanding of the mechanism of AAR expansion is indispensable. In this review paper, the principles of AAR expansion are discussed and the factors that should be considered in expansion tests are evaluated.

The consensus among researchers is that AAR expansion arises from the formation of alkali–silica gel (ASG), which results from the reaction between alkaline pore water in concrete and the silica in reactive aggregates – a chemical process known as the alkali–silica reaction (ASR) (Katayama, 2010; Yamada et al., 2022). Although alternative expansion mechanisms have been proposed (Leemann et al., 2024), the role of ASG remains widely accepted.

Despite this general understanding, several unresolved phenomena persist, particularly concerning the risk assessment of aggregate reactivity, strategies to mitigate AAR expansion and the long-term prediction of expansion in concrete containing reactive aggregates. These challenges highlight the complexity of addressing AAR in both new and existing structures.

Given that AAR expansion is fundamentally linked to ASR, aggregate reactivity has traditionally been assessed using chemical methods (ASTM, 2016; JISC, 2017a). These methods involve immersing aggregates in an alkaline solution and measuring the dissolution of silica. In Japan, where reactive aggregates such as andesite and chert are prevalent, this approach remains widely used. However, the chemical method has been discontinued in North America, as its results often fail to align with the observed expansion behaviour of real structures.

To improve reliability, testing methods that simulate real-world conditions more closely, such as reacting aggregates in mortar or concrete, have been developed. Among these, the concrete prism test (CPT) is considered the most reliable, as it uses aggregate dimensions that are representative of actual applications (AIJ, 2018; ASTM, 2024; Nixon and Sims, 2016a, 2016b). Nonetheless, discrepancies between CPT results and real-world expansion behaviour have been reported (Ideker et al., 2024). Alkali leaching during CPTs has been identified as a significant factor contributing to these inconsistencies, which has prompted the development of mitigation techniques, including the Rilem AAR-10 and AAR-11 methods (Borchers et al., 2021; Rønning et al., 2021). In these methods, the size of concrete prism used is enlarged from 75 × 75 mm × 250 mm to 100  × 100 mm  ×  400 mm, and length readings are carried out without prior cooling of prisms. Several other methods have also been proposed to minimise alkali leaching and ensure sufficient moisture supply (Thomas et al., 2006; Yamada et al., 2021).

Although numerous exposure tests have been conducted to identify CPT conditions that replicate field behaviour (Lindgård et al., 2010; Nguyen et al., 2022; Thomas et al., 2006; Yang et al., 2020), no testing method has provided reliable estimates. Previous studies have introduced the concept of an acceleration factor to determine the extent to which CPT results are accelerated relative to real structures (Kawabata et al., 2014). However, the temperature dependence of ASR is non-linear and varies with each aggregate, necessitating a temperature correction for each aggregate (Ulm et al., 2000). Approaches that fit CPT results obtained under different temperature conditions to expansion curves incorporating temperature effects (Larive, 1998) may provide a basis for replicating the expansion behaviour of exposed specimens (Kawabata et al., 2023). Additionally, these methods hold potential for further development towards predicting structural performance (Comi et al., 2009; Saouma and Perotti, 2006).

Long-term continuous AAR expansion has been observed in some dams, even under low-alkali content conditions (Bérubé et al., 2002; Sims et al., 2012). This has drawn attention to the role of alkali release from aggregates, leading to the establishment of Rilem AAR-8, which evaluates this impact (Menéndez et al., 2021). The Rilem ASR technical committee is currently working on a CPT method to determine an alkali threshold (AT), below which detrimental AAR expansion does not occur (Bavasso et al., 2020). In general, the AT is described by the sodium oxide equivalent (Na₂O_eq) content in concrete. Factoring in alkali contributions from aggregate release may facilitate the development of effective AAR expansion control measures tailored to specific aggregates. However, as long as CPT results remain inconsistent with real-world exposure data, the efficacy of the proposed AT remains uncertain.

Regarding the AT, in Japan, extensive evaluations of 210 aggregate types were conducted during the 1980s using the Japanese mortar bar method (JISC, 2017b), and an AT of 3.0 kg/m³ Na₂O_eq was established (Kuwahara et al., 1989). The first priority of AAR countermeasures in Japan is to control the Na₂O_eq content to be less than this AT. All present, the cement used in Japan is a low-alkali type, and it is possible to achieve this requirement naturally. However, cases have since been reported where this AT has proved ineffective, partially because of the mix proportion pessimum effect of a high proportion of reactive aggregate in the total coarse aggregate (Koga et al., 2013). Koga et al. (2013) detected expansive cracks in some exposed concrete prisms containing 3.0 kg/m³ Na₂O_eq in the field.

Because ASR is a chemical reaction, a minimum pH or AT at which reactive silica ceases to dissolve may theoretically exist, implying the presence of an AT below which AAR expansion will not occur. However, Kawakami et al. (2024) reported that CPTs conducted under various accelerated conditions showed no straightforward positive correlation between alkali content and expansion rate. Instead, a pessimum phenomenon was observed for many aggregates, where expansion rates peaked at intermediate alkali levels.

In summary, most attempts to quantitatively predict long-term AAR expansion, including assessments of whether expansion will occur, have yielded unsatisfactory results. It is thus important to reconsider the basic reason for these mismatches.

In this work, these issues were revisited through the lens of the Ichikawa theory (Ichikawa and Miura, 2007) on the fundamental mechanism of AAR expansion. In simple terms, the Ichikawa theory posits that ASG generated by ASR induces AAR expansion only when constrained. This theoretical framework offers an opportunity to re-examine unresolved phenomena and refine the research questions necessary for accurately predicting long-term AAR expansion. This work does not present original experimental results but rather aims to interpret various phenomena using existing data under the assumption of the Ichikawa theory.

First in this paper, the Ichikawa theory is outlined, followed by a detailed discussion of typical unresolved issues identified from surveyed literature. In selecting references, a policy of focusing on cases in Japan that facilitate understanding of Ichikawa's theory was adopted, while also appropriately referring to overseas cases. This editorial policy was intended to aid readers' logical understanding and ensure consistency in the discussion of the reinterpretation of the AAR expansion mechanism and the examination of unexplained phenomena, which are the objectives of this paper. Finally, the critical research challenges that must be addressed to achieve a quantitative understanding of long-term AAR expansion are described.

The Ichikawa theory proposes that AAR expansion is not solely governed by ASR but is primarily driven by the constraint of ASG generated by ASR and the induced expansion pressure. The extended mechanism underlying AAR expansion based on the Ichikawa theory is illustrated in Figure 1.

Primary ASG forms when reactive silica in the aggregate reacts with an alkaline solution. Unlike single crystals, aggregates are absorbent, containing voids that allow the penetration of alkali solutions (Figure 1(a)). This absorbability is evident from the water absorption of coarse aggregates reported in quality control sheets, even when the aggregate looks to be free of pores.

The penetrating alkali solution originates from the pore water of concrete, which is a lime-saturated solution with a pH of approximately 13.0–13.5. This solution contains dissolved alkali metals and, owing to its high OH⁻ concentration, the Ca²⁺ ion concentration is limited to several millimolar, as determined by the dissolution equilibrium of calcium hydroxide (Ca(OH)₂). The reaction between silica and this alkaline solution forms alkali metal silicate hydrate, or water glass, containing minimal Ca²⁺ ions (Figure 1(b)). The primary ASG is hypothesised to be a fluid sol, referred to as alkali–silica sol (ASS), rather than a gel.

The ASS would not generate expansion pressure if it diffuses freely through the porous concrete. Expansion pressure may occur if the viscosity of the ASS is high enough, if its passage through microvoids is slower than its formation rate or if it is physically constrained by another mechanism (i.e. constraint by a reaction rim) (Figure 1(c)). Figure 2 (Yamada et al., 2013) shows an example of ASG filling a crack that extends from an aggregate into the surrounding cement paste. The variation in the calcium/silicon molar ratio within the ASG filling the cracks is noteworthy: the ratio is lower at the deeper regions of the crack and increases as it approaches the cement paste, eventually nearing the calcium/silicon molar ratio of the cement hydrate calcium silicate hydrate (C-S-H) (Katayama, 2010, 2012a).

ASG can form both within the internal voids of aggregates and on their surfaces. The presence of reaction rims (JCI, 2011; Torii and Yamada, 2017) (Figure 3), commonly observed in rapidly reactive aggregates such as andesite, suggests that such aggregates contain fine voids throughout, facilitating the penetration of alkaline solutions.

ASG generated within the voids and on the surfaces of aggregates interacts with the cement paste, converting into C-S-H. As C-S-H is impermeable to ASG but permeable for ions and water, expansion pressure is generated as ASR progresses and ASG accumulates within the aggregate. Depending on the nature of the aggregate and the reaction conditions, ASG formation can occur more rapidly in the pore system than at the rim (Figure 1(c)) or can occur more rapidly at the rim than in the pore system (Figure 1(d)).

Predominant ASG formation within voids causes cracks to initiate at these sites, propagating into the surrounding cement paste and other aggregates, resulting in sharp cracks (Sanchez et al., 2015) (Figure 1(e)). Under severe accelerated conditions with highly reactive aggregates, ASG accumulation at the reaction rim predominates, leading to ‘onion skin’ cracks (Figure 1(f)). Figure 4 (Kawabata et al., 2024) shows an example observed after a CPT. The propagation of ASG-filled cracks occurs at a rate significantly higher than the conversion of ASG into C-S-H.

Once sharp cracks form, the ASG in contact with the cement paste undergoes exchange with Ca²⁺ ions and is converted into C-S-H. If the rate of new ASG formation is slower than the C-S-H conversion rate, the cracks in the aggregate become sealed by C-S-H, as shown in Figure 2. However, if ASR continues internally, the ASG generates additional expansion pressure (Figure 1(g)). In non-reactive aggregates, ASG in contact with the cement paste transforms into C-S-H, starting from the interface. This process can be imagined from the texture shown in Figure 5 (Kawabata et al., 2019).

Conversely, when the rate of ASG formation owing to ASR exceeds the rate of its conversion into C-S-H, the peripheral regions of cracks extending from aggregates may transform into C-S-H, while the central portions of the cracks remain filled with ASG that continues to flow (Figure 1(h)). This fluctuating ASG supply, combined with the repeated conversion of ASG to C-S-H in crack peripheries and its continued passage through crack centres, accounts for the broom-like structures often observed in AAR-expanded concrete. These structures have been documented in polarised light microscopy and scanning electron microscopy (SEM). An example is shown in Figure 6 (Kawabata et al., 2024).

The Ichikawa theory has been questioned on the grounds that the processes described should produce ASS; however, the existence of ASS is not consistently observed. For example, in Figures 2, 5 and 6, certain regions deep within cracks appear devoid of any material. This section examines the conditions under which ASS may not be observed and potential reasons for its absence. It is sure from microstructural observations of AAR-deteriorated concrete that cracks originate within the aggregate and propagate into the cement paste. Hereafter, the formation mechanism of various texture in cracks are explained.

The reaction products of ASR are collectively referred to as ASG. However, as shown in Figure 2, the chemical composition of ASG is heterogeneous. Detailed analyses using SEM and energy-dispersive spectroscopy, as well as electron probe micro-analysis, indicate that the calcium content of ASG increases progressively from the aggregate interior toward the cement paste. Notably, ASG within cracks in the cement paste exhibits a composition identical to that of C-S-H derived from Portland cement (Katayama, 1998, 2008, 2010, 2012a, 2012b).

In some cases, ASG appears to have shrunk and cracked owing to drying, with subsequent recrystallisation forming crystalline structures known as ‘rosettes’. There are different understandings of the role of rosettes in expansion (Leemann et al., 2024). These rosettes are characterised by voids between grains, as shown in Figure 7, which illustrates an analysis of AAR-induced expansion in mylonite granite from Thailand (Hirono et al., 2016). The figure shows that ASG-filled cracks have undergone rosetting, resulting in the formation of voids between rosette particles. Katayama (2012a) reported instances of extensive ASG rosetting, with no detectable changes in chemical composition, suggesting that rosettes form by recrystallisation accompanied by volumetric shrinkage of ASG. The chemical composition of these rosettes corresponds to solid solutions of shlykovite (KCa[Si₄O₉(OH)].3H₂O) and mountainite (KNa₂Ca₂[Si₈O₁₉(OH)].6H₂O).

The formation of empty, relatively large cracks within aggregates may be attributed to crystallisation pressure during the formation of calcium-rich phases near the cement paste, which could deform the aggregate interior, creating unfilled cracks (Leemann et al., 2024). Alternatively, if ASS serves as the primary precursor to ASG, the absence of filling material could be explained by the loss of fluid ASS during the concrete specimen preparation.

When concrete affected by ASR is cut and kept wet by covering with plastic film, a transparent substance may exude from the aggregate. Figure 8 illustrates the fracture surface of a concrete core containing reactive andesite, where a transparent, gel-like material is visible filling the cracks in the andesite (Torii and Yamada, 2017). Similarly, Figure 9 depicts a cut surface of a concrete core containing reactive granite mylonite (Torii and Yamada, 2017). In this instance, the paste surrounding the granite grains darkened over time. No such discolouration was observed immediately after cutting. This darkening may result from the exudation of ASS from the aggregate, followed by its penetration into the surrounding cement paste.

When uranyl solution is sprayed on such specimens, fluorescence is observed (Igarashi et al., 2016, 2019; Natesaiyer and Hover 1988, 1989; Sanno et al., 2013). Fluorescence occurs only in ASG that has not been fully replaced by calcium. For example, the white area on the fracture surface of the core shown on the left-hand side of Figure 8 does not show fluorescence. Although the handling of uranyl typically requires compliance with regulations governing nuclear materials, it can be safely used in multi-element mixed standard solutions for chemical composition analysis, provided the concentration is sufficient to induce fluorescence.

Figure 10 shows a cut surface of a concrete prism containing alkali-reactive andesite having an alkali content of 5.5 kg/m³, cured at 10°C for a predetermined duration and observed using the uranyl fluorescence method (Kitagawa et al., 2023). Figure 10(a) corresponds to a curing age of 26 weeks, during which cracking had just started, while Figure 10(b) shows the specimen after curing for 62 weeks. The reactive andesite particles, which appear purple under fluorescence, exhibit less fluorescence in the aggregate itself and more in the surrounding paste. Notably, the fluorescence intensity diminished at 62 weeks compared with 26 weeks.

Figure 11 shows the change in fluorescence area for CPT specimens under varying conditions and curing ages (Kitagawa et al., 2023). Fluorescence was more pronounced during the early stages of expansion and decreased after cracking occurred. This trend aligns with a transition from an initial phase of rapid expansion to a period of reduced expansion rate following the onset of cracking. This behaviour reflects the evolution of AAR expansion, where the inflection point (Ben Haha et al., 2007; Dunant and Scrivener, 2010) marks the transition from high initial expansion speed to long-term, low-speed expansion (Kawabata et al., 2018; Yang et al., 2024).

The Ichikawa theory provides a plausible explanation for this phenomenon. The inflection point corresponds to crack initiation, after which ASG under pressure flows outwards through cracks owing to pressure release (Farnam et al., 2015). Upon release, the ASG in the cement paste undergoes calcium displacement, losing fluidity and transforming into C-S-H, which acts as a migration barrier for ASG (as illustrated in Figure 2). Consequently, cores with extended curing durations exhibit reduced uranyl-florescence emitting ASG, consistent with the diminished fluorescence observed over time.

Although ASG or ASS may initially fill cracks within aggregate during the early stages of growth, it can recrystallise into rosettes or C-S-H after the cracks propagate into the cement paste. ASG crystallisation and transformation to C-S-H is associated with volumetric shrinkage, which can result in the formation of hollow cracks within the aggregate. For subsequent expansion to occur, these voids would need to be refilled with ASS, leading to a slower rate of expansion.

The preparation and interpretation of observational samples must consider the potential loss of ASS during processing. Unless samples are processed and observed in a frozen state, ASS may shrink upon drying, resulting in the generation of voids. Even with meticulous handling, this phenomenon cannot be entirely avoided. If the Ichikawa theory is accurate, understanding the mechanisms of AAR expansion necessitates recognising the possibility that observable features represent only part of the phenomenon, and that unobservable elements may play a critical role.

The CPT is widely regarded as a more reliable method for accelerated expansion testing than various mortar bar tests because the CPT is much closer to the actual condition of concrete containing coarse aggregate (Bavasso et al., 2020; Sirivivatnanon et al., 2023; Yamada et al., 2014). The underlying assumption of the CPT is that an increase in alkali content and temperature accelerates both ASR and AAR expansion. However, the degree of acceleration varies depending on the properties of the aggregate. Highly reactive aggregates exhibit a greater acceleration of ASR under such conditions, whereas less reactive aggregates respond to a lesser extent. Consequently, the results of CPTs conducted under consistently accelerated and sufficiently wet conditions often do not align with the expansion behaviour observed in exposure conditions, where moisture is supplied intermittently, such as through rainfall.

A hypothetical example is now used to explain why CPT results do not match the exposure conditions. At an acceleration temperature of 60°C, which is higher than the average temperature of the exposure environment, the expansion rate of aggregate A is ten times higher and that of aggregate B is five times higher. The expansion observed by the CPT after 1 year at 60°C corresponds to expansion in the field after 10 years for aggregate A and after 5 years for aggregate B. In this case, when the expansion measured by the CPT after 1 year is compared with the expansion in the field after 5 years, the values for aggregate B are the same whereas, for aggregate A, the values in the field are smaller than those measured by CPT. In general, different aggregates are expected to have different temperature dependencies of expansion rate. Therefore, the relationship between the expansion rate of different aggregates measured under the same accelerated test conditions at a given age and the expansion at exposure will be different for each aggregate.

The pursuit of ‘appropriate’ CPT conditions that align simply with field expansion behaviours may be an unattainable goal. Attempting to identify a universally valid solution appears to be an exercise in wishful thinking, akin to seeking a theoretically impossible outcome without acknowledging the limitations of the method.

In response to these challenges, the authors developed a model that accounts for the effects of temperature fluctuations and rainfall. Results from alkali-wrapped CPTs (AW-CPT) conducted at varying temperatures demonstrated the ability to reproduce field expansion behaviour of concrete blocks of the same composition and dimensions exposed in three regions of Japan (cold, warm and intermediate climates) with reasonable accuracy (Kawabata et al., 2023). From an engineering perspective, however, conducting extensive preliminary studies under a wide range of conditions for each specific material composition is impractical. This situation underscores the need for simplified methodologies to ensure that materials will not undergo AAR expansion. Nonetheless, predicting long-term AAR expansion based on CPTs conducted under a single set of conditions remains an exceptionally complex task owing to the underlying mechanisms of AAR expansion. Achieving accurate long-term predictions requires a more nuanced understanding of these mechanisms and the influence of environmental variables.

The above scenario assumes that the mechanism by which ASR is converted into AAR expansion during accelerated curing is consistent, specifically during the early stages of AAR expansion, prior to the expansion's inflection point. However, the reality is more complex. Expansion behaviour at low temperatures involves certain phenomena that cannot be explained by extrapolating accelerated conditions, which are higher than those in the actual environment, to lower temperatures (Kawabata et al., 2024; Kitagawa et al., 2023).

This issue is addressed in the Ichikawa theory. Altering the accelerating conditions changes the texture of the expansion formed by ASG owing to ASR, depending on the reactivity of the aggregate. Specifically, if the initial crack results from the formation of a phase that impedes the migration of ASG produced by the substitution of calcium in the cement paste, under severe accelerated conditions, the rate of ASG formation is faster. Once cracks open, however, the texture formation required to constrain the subsequent production of ASG does not occur in time. Consequently, while ASR progresses and ASG forms, ASG only flows out through cracks once they are fully formed, leading to less enhancement of AAR expansion. A common example of this phenomenon is the underestimation of expansion for rapidly expanding aggregates when tested under severe accelerated conditions (Kawakami et al., 2024). Figure 12 illustrates the expansion behaviour measured in the CPT using a mixture of 30% highly reactive andesite as coarse aggregate, for varying accelerated temperatures and alkali contents. In the figure, the first number in the key indicates the temperature (in °C) and the second number is the alkali contents (in kg/m³). While higher temperatures initially induced greater expansion, the long-term expansion rate may be greater at lower temperatures. Notably, under the most moderate accelerated conditions (20°C, 2.5 kg/m³ alkali content), the expansion after 78 weeks exceeded that observed at 40°C and 60°C for the same alkali content.

On the other hand, for less reactive aggregates, even under accelerated conditions, the rate of ASG formation can be balanced by the ASG containment reaction owing to calcium substitution. As a result, even with the same ASR rate and quantity of ASG produced, AAR expansion is expected to be greater. While accelerated tests can assess the reactivity of a typical reactive aggregate in a given region, they do not necessarily allow for the evaluation of a less reactive, general-purpose aggregate.

The degree of acceleration in CPT expansion has an upper limit based on the reactivity of the aggregate, particularly when attempting to replicate field expansion behaviour. The conclusion is that more accelerated conditions lead to earlier convergence of the expansion rate at a smaller value (Yang et al., 2024). Therefore, overly accelerating the testing process should be avoided, especially when making long-term expansion predictions.

The pessimum phenomenon of AAR expansion typically refers to a compositional pessimum, which depends on the aggregate composition. In this phenomenon, a rapidly expanding aggregate, when mixed with a non-reactive aggregate, exhibits a maximum rate of expansion at a mixing ratio of less than 100%, such as 5–30% of the rapid-reactive aggregate. An aggregate that does not display this extreme pessimum phenomenon is referred to as a late-expanding aggregate (Katayama, 2014). However, the pessimum mixing ratio has been shown to vary depending on the accelerated conditions of the test method and the age of the material being evaluated (Yamada et al., 2022). The pessimum mixing ratio may be smaller when evaluating long-term curing ages, suggesting that indoor tests with shorter evaluation times may overestimate the pessimum ratio, even in the CPT.

Figure 13 shows the expansion behaviours of concrete containing various amounts of highly reactive andesite and different alkali contents, evaluated at different ages. This experiment was carried out to determine the pessimum composition of this aggregate. The aggregate used was reactive andesite, the same aggregate as used in Figures 10–12. The expansion was evaluated according to the Japanese mortar bar test (JISC, 2017a).

At an alkali content of 3.0 kg/m³, the pessimum proportion was 60% at 9 days, but it moved towards 10–40% with increasing curing age (Figure 13(a)). It is possible to conclude that the pessimum expansion would become less with greater age. The situation was very different when the alkali content was increased to 4.0 kg/m³ – the pessimum composition was clearly 20–30% (Figure 13(b)). In the greater ages, the pessimum proportion decreased with a reduction in alkali content. A conceptual diagram illustrating the mechanism behind the occurrence of the pessimum proportion is shown in Figure 14 (Kawabata and Yamada, 2017). The key to explaining the compositional pessimum effect is the extra volume of pores originally contained in the aggregate. For AAR expansion, extra pores must be filled with ASG. If the volume of reactive aggregate is large, the pH decreases more before AAR expansion starts owing to the ASR, which is an alkali-consuming expansion. Of course, if the alkali content is sufficient, AAR expansion will be more in the case of a higher composition of reactive aggregate because the alkali consumption is negligible to fill the pores before expansion starts. But by decreasing the alkali content, there is a pessimum value of reactive aggregate composition to show the maximum expansion, and the pessimum reactive aggregate composition decreases with a decrease in alkali content.

Kawabata and Yamada (2017) applied the model proposed by Furusawa and Uomoto (Furusawa and Uomoto 1992; Furusawa et al., 1994; Uomoto and Furusawa, 1992) to quantitatively reproduce the pessimum composition. This model provides a mechanistic understanding of the phenomenon and quantitative estimation by introducing some empirical parameters. The UFO model explains the compositional pessimum mechanism as follows. ASG generated by ASR fills the voids in the aggregate and the excess ASG after filling the voids generates expansion pressure. The conversion of excess ASG into AAR expansion is treated as an empirical parameter. ASR, being an alkali-consuming reaction, decreases the alkali (OH⁻) concentration in the liquid phase as ASG is produced. The reaction ceases when the alkali concentration falls below a critical threshold, at which point ASR cannot continue. If too much reactive aggregate is present, it consumes alkali to fill the aggregate voids before ASG can begin to expand. Conversely, with less reactive aggregate, there are fewer voids to be filled, so the same amount of ASG produced will result in more ASG contributing to expansion. This balance determines the aggregate ratio that maximises expansion under constant alkali content conditions.

However, this model for predicting AAR expansion is problematic because it represents a constant that converts the quantity of ASG into AAR expansion. It does not quantitatively account for the ASG constraint described in the Ichikawa theory – specifically, the rate at which the ASR rate is controlled by calcium substitution and the resulting constraint on ASG formation.

The Ichikawa theory predicts various factors possibly showing pessimum expansion. As explained earlier, too much acceleration results in less expansion than mild acceleration, and acceleration is effective at the early age before cracking but less so after cracking. It is obvious that all accelerating factors, such as alkali content and curing temperature, will show pessimum effects and the pessimum behaviours are different depending on the type of aggregate and the age of evaluation.

If the rate of ASG formation and the onset of its constraint can be effectively modelled, AAR expansion may be predicted quantitatively. Although reaction rates at low temperatures are difficult to measure, an Arrhenius plot could potentially extrapolate high-temperature test results to lower temperatures. The main challenge lies in reproducing the constraining conditions. This is crucial for observing the microstructural degradation associated with AAR expansion. However, the rate-limiting reaction process that restricts ASG migration, as well as the reaction rate of ASG conversion to C-S-H, remain areas that are not fully understood.

Bérubé and co-workers (Bérubé and Fournier, 2004; Bérubé et al., 2004a, 2004b) pointed out the possibility of an increase in alkali content by alkali release from specific aggregates. Although the alkali content in dam concrete should be low because of the very low cement content for low heat generation, the detected soluble alkali content from cores after long years operation can be considerably high. For example, in one of the surveys conducted by Bérubé et al. (2002), the design alkali content of the Juillet dam was 1.7 kg/m³, whereas the water-soluble alkali content confirmed from the core was approximately 2.1–3.5 kg/m³. The contribution of alkali release to AAR expansion is complex. Alkali release does not occur during the initial period (Yamada et al., 2024), but rather over the long term. This suggests that the alkali concentration does not decrease owing to ASG formation, as proposed in the UFO model. In this case, the pH of the liquid phase either remains constant or increases over time. Reproducing these conditions in accelerated tests is not straightforward.

The Ichikawa theory predicts that expansion will continue slowly over the long term, without subsiding. The slowly produced ASG would not be lost from the system; instead, it would be constantly constrained by the cement paste and contribute to further expansion as it accumulates. Therefore, in a low-temperature environment, a late-expansive aggregate with a slow reaction rate and components that cause alkali release may exhibit significant long-term expansion, even if no expansion is observed during accelerated testing.

At this moment, there is no answer for an appropriate test method, but at least it would be reasonable to develop some new procedures on the assumption of the possible mechanism explained here.

As discussed in this section, alkali release is a phenomenon by which the alkali content gradually increases with the age of concrete. Conversely, with the addition of supplementary cementitious materials, the alkali content decreases with age, except for water-soluble alkalis in supplementary cementitious materials. Therefore, these effects should not be considered in the final alkali balance between solution and solid after long aging, but they should be considered in the time change of pH of solution and the texture formation process.

Based on the Ichikawa theory that the generation of ASS or ASG by ASR itself does not cause expansion, but that expansion only occurs when ASG is confined within the aggregate, various questions related to AAR expansion were reconsidered. The following is a summary of the results, along with the authors’ opinions on the challenges of applying current testing methods in the field.

ASS is easily lost during the preparation of thin sections and samples for SEM observation. Methods for preventing the cutting surface from drying out and visualising ASG have been introduced, as well as observation methods using the uranyl acetate fluorescence method. Furthermore, as ASR progresses, ASG decreases. Without understanding the AAR expansion mechanism, even detailed petrological analysis of degraded textures will not reveal the actual phenomena.

The reasons why the expansion trends in laboratory tests, accelerated expansion tests and field exposure do not match was considered. Although ASR can be accelerated by high temperatures and alkali boosting, the constraint reaction of ASG – that is, the C-S-H conversion reaction of ASG owing to contact with cement paste – cannot be accelerated. Therefore, the accelerated ASR tests in the laboratory do not accelerate AAR expansion.

Let us turn now to the causes of various types of pessimism. Normally, the term 'pessimum phenomenon' refers to the phenomenon in which expansion reaches its maximum at a certain value below 100% when highly reactive aggregates and non-reactive aggregates are mixed together. However, there are also pessimum values for the alkaline content and the temperature at which expansion reaches its maximum and, moreover, the values of all these pessimum phenomena depend on the age of the material being evaluated.

These various pessimum phenomena can be predicted by evaluating the reactivity of the aggregate using the former ASTM chemical method (ASTM, 2016) and applying it to a certain reaction model. Therefore, AAR expansion may be able to be explained, in principle, by the reaction rate of the aggregate.

Alkali release from aggregates is a phenomenon in which the amount of alkali slowly increases over the long term. Considering the Ichikawa theory, ASR will continue to occur under conditions where the initial amount of alkali is low, and there will be sufficient time for ASG to be converted to C-S-H, so it will show greater expansion than in tests where the initial amount of alkali was increased.

An attempt has been made to explain various unresolved phenomena related to AAR expansion, based on the Ichikawa theory, by considering the discrepancies between the phenomena observed in reality and in accelerated conditions that induce AAR expansion. In this context, the following research questions should be addressed in future investigations.

• Given the substantial evidence supporting the idea that AAR expansion is caused by ASS, detecting ASS – a fluid with minimal calcium content – is crucial to validating the Ichikawa theory.

• Even if ASS is recognised as the primary substance responsible for AAR expansion, it may be lost during sample preparation. Therefore, when observing AAR expansion textures, it is essential to account for sample preparation conditions that preserve ASS. Alternatively, researchers must consider the potential loss of ASS when interpreting observed textures.

• Research on current engineering challenges, such as the discrepancy between CPT and exposure expansions, accelerated testing conditions that do not induce AAR expansion and the impact of alkali release from aggregates, should focus on solutions grounded in the understanding that no straightforward answers exist. Rather than repeatedly testing accelerated methods that could match exposure conditions, a more comprehensive approach is needed.

• Various AAR expansion predictions do not always reproduce reality, because AAR expansion is not simply caused by ASR. Understanding that promoting ASR by temperature or total alkali content does not necessarily promote AAR expansion shows that current AAR expansion test methods may not be sufficiently reliable. At the very least, ASR expansion testing should not be overly promoted.

• Ideally, if the rates of ASR and the constraint of ASG, as well as the changes in crack structure resulting from this balance, can be quantified, it would enable predictions of long-term AAR expansion in practical situations.

Various unresolved issues related to concrete expansion caused by AAR were explored using the Ichikawa theory as a mechanistic framework. According to the Ichikawa theory, ASS produced by ASR interacts with cement paste to form C-S-H. This interaction allows water and alkalis to pass through while sealing out the ASS, thus enabling ASR to persist and the expansion pressure to accumulate as the amount of ASS in the aggregate increases. ASG encompasses both the ASS and the compositional changes that occur when ASS interacts with cement paste.

The Ichikawa theory provides an explanation for various phenomena, including the difficulty of detecting ASS in tests, the inconsistency between accelerated tests and expansion rates under exposure conditions, and the roles of pessimum effects and alkali release in promoting long-term expansion. Ultimately, the expansion process is governed by the balance between the rate of ASS formation and the rate of the substitution reaction with calcium, which encapsulates the ASS.

Future research on AAR expansion should build on the insights provided by the Ichikawa theory, with an emphasis on conducting degradation microstructural analysis. Additionally, a quantitative evaluation of the fundamental rates predicted by the theory – the rate of aggregate dissolution and the rate of C-S-H formation that constrains ASS – along with the development of an expansion model that integrates these factors, is essential for advancing the understanding of AAR expansion.