RILEM TC 258-AAA Round Robin Test: Alkali release from aggregates and petrographic analysis. Critical review of the test method AAR-8

1. INTRODUCTION

The investigation developed for decades, regarding the mechanisms of alkali-aggregate reactions, indicates that some aggregates or mineral phases used in concrete can leach significant quantities of alkali to the interstitial solution over time (11. Berube, M.A.; Duchesne, J.; Dorion, J.F.; Rivest, M. (2002) Laboratory assessment of alkali contribution by aggregates to concrete and application to concrete structures affected by ASR. Cem. Concr. Res. 32 [8], 1215-1227. https://doi.org/10.1016/S0008-8846(02)00766-4.
, 22. Santos Silva, A.; Fernandes, I.; Ferraz, A.R.E.; Soares, D. (2017) Can certain alkali minerals explain the slow reactivity of granitic aggregates in dams?. in Swelling Concrete in Dams and Hydraulic Structures - DSC 2017. Edited by Alain Sellier. Étienne Grimal. Stéphane Multon. Eric Bourdarot. ISTE. pp. 93-105.
). To reduce the risk of alkali-silica reaction (ASR) developing in the concrete, the total alkali content of the concrete should be controlled for concrete exposed to wet environments. Therefore, it is necessary to control the alkali content of all components of the concrete, including the additives and water, but especially to limit the alkali content of the cement, and/or to use specific amounts of supplementary cementing materials (33. Stark, D.; Sydney, D. (1993) Eliminating or minimizing alkali-silica reactivity. National Research Council. Washington. U.S.A. Report SHRP-C-343. Retrieved from https://onlinepubs.trb.org/Onlinepubs/shrp/shrp-c-343.pdf.
).

Most rocks commonly used as aggregates contain minerals with Na and/or K. Plagioclase, K-feldspar, and K/Na-feldspar exist in a large number of aggregates, but feldspathoids, mica, clay minerals, zeolites, amphibole, volcanic glass, and unwashed marine sands contain alkalis. Aggregates containing these minerals have been investigated owing to the suspicion that they might increase the alkali content of the concrete (4-84. Grattan-Bellew, P.E.; Beaudoin, J.J. (1980) Effect of phlogopite mica on alkali-aggregate expansion in concrete. Cem. Concr. Res. 10, 789-797.
5. Gillot, J.E.; Rogers, C.A. (1994) Alkali-aggregate reaction and internal release of alkalis. Mag. Concr. Res. 46, 99-112.
6. Menéndez, E.; Prendes, N.; Márquez, C.; Aldea, B. (2012) Evaluation of granitic aggregate behaviour in relation with the alkaline extraction and compositional change in their phases. 14th ICAAR, Austin, EEUU, Ma 2012. Paper 052011
7. Menéndez, E.; Ruiz, S.; García-Rovés, R. (2016) Alkali release from aggregates; contribution to ASR. Const. Mat. 169 [4], 206-214. https://doi.org/10.1680/jcoma.15.00072.
8. Bérubé, M.A.; Duchesne, J.; Dorion, J.F.; Rivest, M. (2002) Laboratory assessment of alkali contribution by aggregates to concrete and application to concrete surfaces affected by alkali-silica reactivity. Cem. Concr. Res. 32, 1215-1227. https://doi.org/10.1016/S0008-8846(02)00766-4.
).

The first laboratory test to evaluate alkali release by aggregates was proposed by Van Aardt and Visser (99. Van Aardt, J.H.P.; Visser, S. (1977) Calcium hydroxide attack on feldspars and clays: Possible relevance to Cement-aggregate reactions. Cem. Concr. Res. 7, 643-648. https://doi.org/10.1016/0008-8846(77)90046-1.
), who studied the behavior of feldspars and clay minerals in the presence of calcium hydroxide solutions. The authors observed that the reaction in an aqueous suspension at 39 °C allowed the precipitation of tetracalcium aluminate hydrates, free alkalis, alkali silicates, and hydrogarnets, depending on the type of feldspar. It was also observed that the reaction of calcium hydroxide with sericite, shale, and some igneous rocks, originated from the hydrogarnets, hydrated calcium silicates, and free alkalis when the materials are boiled at 95 °C (99. Van Aardt, J.H.P.; Visser, S. (1977) Calcium hydroxide attack on feldspars and clays: Possible relevance to Cement-aggregate reactions. Cem. Concr. Res. 7, 643-648. https://doi.org/10.1016/0008-8846(77)90046-1.
, 1010. Van Aardt, J.H.P.; Visser, S. (1977) Formation of hydrogarnets: Calcium hydroxide attack on clays and feldspars. Cem. Concr. Res. 7, 39-44. https://doi.org/10.1016/0008-8846(77)90006-0.
).

Several studies have been conducted regarding the release of alkalis from various minerals over time when exposed to a high-pH environment, such as concrete. The rate and amount of alkali released also depends on the temperature, local porosity, and permeability of the cement paste. Additionally, the alkali release is affected by the diverse properties of the aggregates, such as the particle size, lithological composition, mineral grain size, mineral modal content, total alkali content, and bonding of mineral constituents. Furthermore, the degree of alteration, spatial microstructure, fabric or texture, as well as the internal porosity and permeability of the aggregate particles must be considered (11-1311. Bérubé, M.A.; Fournier, B. (2004) Alkalis releasable by aggregates in concrete: significance and test methods. In: Tang, M. and Deng, M. (editors): Proceedings of the 12th International Congress on Alkali-Aggregate Reactions in Concrete, Beijing, China, pp. 17-30.
12. Lu, D.; Zhou, X.; Xu, Z.; Lan, X.; Tang, M.; Fournier, B. (2006) Evaluation of laboratory test method for determining the potential alkali contribution from aggregate and the ASR safety of the Three-Gorges dam concrete. Cem. Concr. Res. 36, 1157-1165. https://doi.org/10.1016/j.cemconres.2006.01.004.
13. Drolet, C.; Duchesne, J.; Fournier, B. (2017) Validation of the alkali contribution by aggregates to the concrete pore solution. Cem. Concr. Res. 98, 10-23. https://doi.org/10.1016/j.cemconres.2017.04.001.
). Release of the respective ions depends on the crystallographic structure of the minerals and the bonding strength between the ions within their structure.

The main principle behind the tests suggested in the recent decades refers to the immersion of the aggregate sample in saturated lime solutions, with or without excess solid lime, or in alkaline solutions (NaOH, KOH) to evaluate the alkali contribution to the concrete from the respective aggregate. However, as of yet no agreement has been reached regarding the amount of alkali released and the procedure for estimating them.

The objective of RILEM TC 219-ACS ‘Alkali-aggregate reactions in concrete structures’ (2007-2014) was to establish and validate a number of general international ASR performance-based testing concepts. To maintain relevance to the field structures when testing samples in the laboratory, it is crucial that the basic conditions do not change to an extent where the correlation of performance ranking under the two different conditions is not maintained. As Lindgård et al. confirmed (1414. Lindgård, J.; Andiç-Çakır, Ö.; Fernandes, I.; Rønning, T.F.; Thomas, M.D.A. (2012) Alkali-silica reactions (ASR): Literature review on parameters influencing laboratory performance testing. Cem. Concr. Res. 42, 223-243. https://doi.org/10.1016/j.cemconres.2011.10.004.
), several parameters may influence the results obtained in an accelerated performance test compared to the field behavior of concrete structures. These parameters are the type of aggregate, type of binder, mix design, and conditions during pre-storage and ASR exposure. Furthermore, the degree of alkali leaching depends strongly on the storage conditions and mineralogy (22. Santos Silva, A.; Fernandes, I.; Ferraz, A.R.E.; Soares, D. (2017) Can certain alkali minerals explain the slow reactivity of granitic aggregates in dams?. in Swelling Concrete in Dams and Hydraulic Structures - DSC 2017. Edited by Alain Sellier. Étienne Grimal. Stéphane Multon. Eric Bourdarot. ISTE. pp. 93-105.
, 33. Stark, D.; Sydney, D. (1993) Eliminating or minimizing alkali-silica reactivity. National Research Council. Washington. U.S.A. Report SHRP-C-343. Retrieved from https://onlinepubs.trb.org/Onlinepubs/shrp/shrp-c-343.pdf.
, 99. Van Aardt, J.H.P.; Visser, S. (1977) Calcium hydroxide attack on feldspars and clays: Possible relevance to Cement-aggregate reactions. Cem. Concr. Res. 7, 643-648. https://doi.org/10.1016/0008-8846(77)90046-1.
, 1010. Van Aardt, J.H.P.; Visser, S. (1977) Formation of hydrogarnets: Calcium hydroxide attack on clays and feldspars. Cem. Concr. Res. 7, 39-44. https://doi.org/10.1016/0008-8846(77)90006-0.
, 14-1814. Lindgård, J.; Andiç-Çakır, Ö.; Fernandes, I.; Rønning, T.F.; Thomas, M.D.A. (2012) Alkali-silica reactions (ASR): Literature review on parameters influencing laboratory performance testing. Cem. Concr. Res. 42, 223-243. https://doi.org/10.1016/j.cemconres.2011.10.004.
15. Constantiner, D.; Diamond, S. (2003) Alkali release from feldspars into pore solutions. Cem. Concr. Res. 33, 549-554. https://doi.org/10.1016/S0008-8846(02)01001-3.
16. Yujiang, W.; Guiyu, Y.; Ming, D.; Mingshu, T.; Duyou, L. (2008) The use of thermodynamic analysis in assessing alkali contribution by alkaline minerals in concrete. Cem. Concr. Res. 30, 353-359. https://doi.org/10.1016/j.cemconcomp.2007.03.003.
17. Yujiang, W.; Min, D.; Mingshu, T. (2008) Alkali release from aggregate and the effect on AAR expansion. Mater. Struct. 41, 159. https://doi.org/10.1617/s11527-007-9227-z.
18. Soares, D.; Santos Silva, A.; Mirão, J.; Ramos, V.; Fernandes, I.; Menéndez, E. (2015) Assessment of alkalis released by aggregates. Contribution to the alkalinity increase and AAR development in concrete. Second International Dam World Conference. Lisbon. LNEC. pp. 10.
), and the pre-storage conditions might further influence the rate and extent of alkali leaching significantly. Considering this comparison, Lindgård et al. (1414. Lindgård, J.; Andiç-Çakır, Ö.; Fernandes, I.; Rønning, T.F.; Thomas, M.D.A. (2012) Alkali-silica reactions (ASR): Literature review on parameters influencing laboratory performance testing. Cem. Concr. Res. 42, 223-243. https://doi.org/10.1016/j.cemconres.2011.10.004.
) found that specimen size clearly has a large impact on expansion, and this effect can be largely ascribed to the leaching of alkalis from smaller specimens. In their study, they also found that the impact of alkali leaching was lower for larger concrete prisms, although it was still significant. However, they concluded that none of the currently available or commonly used test methods met all criteria for an ideal performance test (1414. Lindgård, J.; Andiç-Çakır, Ö.; Fernandes, I.; Rønning, T.F.; Thomas, M.D.A. (2012) Alkali-silica reactions (ASR): Literature review on parameters influencing laboratory performance testing. Cem. Concr. Res. 42, 223-243. https://doi.org/10.1016/j.cemconres.2011.10.004.
).

According to Jozwiak-Niedzwiedzka et al. (1919. Jozwiak-Niedzwiedzka, D.; Gibas, K.; Glinicki, M.A. (2017) Petrographic identification of reactive minerals in domestic aggregates and their classification according to RILEM and ASTM recommendations, Roads and Bridges - Drogi i Mosty 16, Warsaw, Poland, pp. 223-239.
), the classification of aggregates by petrographic methods in terms of the likelihood of alkali-aggregate reactions focuses mainly on the identification of rocks and forms of silica that are known to be reactive, that is, chert, novaculite (form of chert) volcanic glass, cristobalite, tridymite, and opal. Thin sections can also be analyzed for the size of quartz crystals to clearly determine the ASR susceptibility of the aggregate. Different deleterious materials are sought depending on the rock type, namely, microcrystalline quartz in granite, volcanic glass in volcanic rock, cryptocrystalline quartz in carbonate rock, and deformed particles of quartz, especially strained quartz, in the case of metamorphic rocks (1919. Jozwiak-Niedzwiedzka, D.; Gibas, K.; Glinicki, M.A. (2017) Petrographic identification of reactive minerals in domestic aggregates and their classification according to RILEM and ASTM recommendations, Roads and Bridges - Drogi i Mosty 16, Warsaw, Poland, pp. 223-239.
).

RILEM AAR-8 is a test method intended to assess the potential amount of alkalis released by aggregates using an alkali solution saturated with calcium hydroxide. A draft of the RILEM AAR-8 was prepared in 2013 in the scope of TC 219-ACS, and a round robin test (RRT1) was carried out using five aggregate samples and five laboratories. The potential amounts of sodium and potassium ions released by aggregates were then measured by the immersion of the aggregate particles, whose sizes were further reduced by grinding them to a sand size fraction (0.125 to 4.00 mm), in NaOH (K release) and KOH (Na release) solutions at elevated temperatures. The solutions should be in contact with excess (solid) calcium hydroxide, as this is the situation with concrete in practice and can potentially increase alkali leaching through ion exchange with Ca²⁺ (88. Bérubé, M.A.; Duchesne, J.; Dorion, J.F.; Rivest, M. (2002) Laboratory assessment of alkali contribution by aggregates to concrete and application to concrete surfaces affected by alkali-silica reactivity. Cem. Concr. Res. 32, 1215-1227. https://doi.org/10.1016/S0008-8846(02)00766-4.
, 1111. Bérubé, M.A.; Fournier, B. (2004) Alkalis releasable by aggregates in concrete: significance and test methods. In: Tang, M. and Deng, M. (editors): Proceedings of the 12th International Congress on Alkali-Aggregate Reactions in Concrete, Beijing, China, pp. 17-30.
, 20-2320. LCPC (1993) Essai de granulat - Détermination des alcalins solubles dans l’eau de chaux. Méthode d’essai nº 37, LCPC, Paris, France. pp. 13.
21. Berube, M.A.; Duchesne, J.; Rivest, M. (1996) Alkali contribution by aggregates to concrete. 10th ICAAR. Melbourne. Australia. pp. 899-906.
22. Berube, M.A.; Fournier, B. (2002-b) Alkalis releasable by aggregates in concrete: significance and methods. Reference Report RILEM/TC-ARP/02/12 distributed at the 2002 meeting. Reykjavik. Iceland. pp. 15.
23. Menéndez, E. (2013) Extraction of alkalis from aggregates. Workshop RILEM/TC-ACS-219. Lisbon. Portugal. Unpublished oral presentation.
).

However, the conclusions of RRT1 pointed out that there was a wide variability between laboratories, and stricter information regarding the procedure needed to be included. The research was further carried out within the scope of RILEM TC 258-AAA, Avoiding Alkali Aggregate Reactions in Concrete - Performance Based Concept (2014-2020). The performance of the test by five laboratories showed more coherent results, and an improved revised draft of the test (2018) was used in the second RRT (RRT2), for which five samples were then selected.

In the present work, the original version of the method is presented, as well as the limitations observed are discussed, explaining the need for a revised methodology. Five aggregates were tested and analyzed under an optical microscope and scanning electron microscope to identify the minerals that compose each of the samples. In addition, chemical analysis of the bulk rock was carried out to compare with the results obtained in the RRT-1 of AAR-8 laboratory tests. It should be highlighted that the outcome of AAR‐8 is indicative of the potential maximum amount of alkalis that can be released by the aggregates. It should be noted that there is still no verification of the actual alkali release under field conditions; therefore, the method cannot be used at this stage to limit the total alkali content in concrete. The verification of the method requires the assessment of concrete structures in which these aggregates have been used and knowledge of the alkali content in the original concrete is conclusive.

2. MATERIALS AND METHODS

2.1. Aggregate samples

 ⌅

Table 1 lists the five aggregates used in this test, indicating lithology and provenance. Some of these aggregates are known to be reactive in the field structures. These correspond to four samples of crystalline rocks of igneous and metamorphic origin and one sample of polymictic sand containing a large variety of igneous volcanic lithologies. Samples weighing 500 g were sieved to obtain the granulometric distributions presented in Table 2. The aggregates were then analyzed in five research institutes: Université Laval in Canada, SINTEF in Norway, IETcc-CSIC in Spain, LNEC in Portugal, and RSESpA in Italy. The last laboratory (Italy) analyzed only three of the aggregates tested: cataclasite, granodiorite, and granite.

Table 1.  Designation and provenance of the samples and information of the laboratories that performed the tests.

Type of aggregate	Phonolite	Cataclasite	Granodiorite
Supplying laboratory	Université Laval	SINTEF	IETcc-CSIC
Country	Canada	Norway	Spain
Aspect
Laboratory participated in RRT1	Yes	Yes	Yes
Type of aggregate	Granite	Sea dredged, basalt	-
Supplying laboratory	LNEC	MANNVIT	-
Country	Portugal	Island
Aspect
Laboratory participated in RRT1	Yes	No

Table 2.  Grain size distribution used in the RILEM AAR-8 tests.

Grain size distribution	Percentage of mass
4 mm - 2 mm	10 %
2 mm - 1 mm	20 %
1 mm - 500 µm	20 %
500 µm - 250 µm	25 %
250 µm - 125 µm	15 %
<125 µm	10 %

3. RESULTS AND DISCUSSION

In this section, the results of the physical properties, chemical and mineralogical composition, and petrographic analyses are presented for each sample, as well as the results of the AAR-8 tests performed.

3.3. Mineralogy, petrography, and SEM/EDS

 ⌅

X-ray diffraction (XRD) and petrographic analysis showed that phonolite is composed of nepheline, Na-feldspar, K-feldspar, and other minor silicate minerals; cataclasite contains Na-feldspar, K-feldspar, and quartz; while granodiorite and granite contain Na-feldspar, K-feldspar, quartz, traces of mica (illite), and muscovite. On the other hand, polymictic sea-dredged sand contains particles of variable lithology, dominantly corresponding to basalt. It is mainly composed of Na-feldspar and Ca-feldspar (plagioclase) and contains pyroxenes, such as diopside and augite, which specifically belong to the clinopyroxene group.

In Figure 1, the dominant texture and the main minerals that compose the rocks are observed. The sample that exhibited the strongest signs of alteration was granodiorite, with dull large K-feldspars and sericite formed on plagioclase crystals (Figure 1c). The rock also showed strain features in both quartz and feldspar crystals. The crystals in the granite sample were finer and the rock was slightly altered, with the smallest generation of plagioclase crystals showing argillization. The sea-dredged sand contains variable lithologies, mainly in terms of grain size and content of volcanic glass (Figure 1e and 1f). Palagonite is abundant in some sand grains. Phonolite and cataclasite are fine-grained, while the latter exhibits intense deformation. Phonolite contains abundant nepheline, which is a feldspathic mineral. Finally, the type of feldspar composing the rock is difficult to distinguish in cataclasite on this scale. As the aggregates were very fine-grained, samples of phonolite, cataclasite, and sea-dredged sand were studied using SEM/EDS to identify some of the minerals. Figure 2 presents an example of the SEM/EDS analyses performed, highlighting the importance of this method in the identification of alkali-bearing minerals.

Figure 1.  Optical photomicrographs (under plane polarized light N// and crossed polarized light NX) of samples tested: (a) phonolite, (b) cataclasite, (c) granodiorite, (d) granite, (e) and (f) particles of basalt in the sea-dredged sand.

Figure 2.  SEM image of the cataclasite and X-ray element maps of Na and K, showing the abundance of the minerals containing alkalis (albite and K-feldspar).

3.4. Alkali extraction by AAR-8 procedure, problems during the test

 ⌅

3.4.1. Na₂O_eat 26 and 52 weeks

 ⌅

The average percentages of Na₂O, K₂O, and Na₂O_e (in mass %) released from each aggregate were calculated from the results obtained by all the five participating laboratories. The results are expressed as a function of time for both exposure temperatures, 38 °C and 60 °C. Table 6 shows the variation in the release of Na₂O_e with respect to the temperatures, comparing the results for 60º and 38 °C for each sample, and the time of testing, comparing the results after 52 weeks with those obtained at 26 weeks.

Table 6.  Na₂O_e concentration (mass %) measured at 26 and 52 weeks of testing.

Aggregate	Temperature	Na2Oe (26 weeks)	Increase from 38 to 60 °C (%) at 26 weeks	Na2Oe (52 weeks)	Increase from 38 to 60 °C (%) at 52 weeks	Increase from 26 to 52 weeks (%)
Phonolite	38 °C	0.39	43.6	0.57	29.8	46.2
	60 °C	0.56		0.74		32.1
Cataclasite	38 °C	0.11	0.0	0.17	-11.8	54.5
	60 °C	0.11		0.15		36.4
Granodiorite	38 °C	0.07	42.9	0.11	0.0	57.1
	60 °C	0.10		0.11		10.0
Granite	38 °C	0.04	150.0	0.07	57.1	75.0
	60 °C	0.10		0.11		10.0
Sea dredged, basalts	38 °C	0.41	12.2	0.46	21.7	12.2
	60 °C	0.46		0.56		21.7

3.4.2. Na₂O, K₂O and Na₂O_eat 38 °C and 60 °C

 ⌅

Figure 3 exhibits the curves showing the evolution of alkali release. The rates of alkali release always tended to decrease with time, but after 52 weeks of testing, the results still did not reach a plateau, and the aggregates continued to release alkalis.

Figure 3.  Curves obtained from the average of results of all laboratories test of alkalis released in 52 weeks at 38 °C and 60 °C for each aggregate.

For the majority of the samples, there was a high release of Na⁺ than K⁺, except for granodiorite and granite. Figure 3 also shows that phonolite and sea-dredged sand are the two aggregates that released the highest amounts of alkalis, whereas granodiorite and granite released the least, with values that reach a maximum of 0.11% Na₂O_e at one year. Finally, cataclasite released slightly more alkali than the last two lithologies, but the values still remained low.

This Figure 3, shows that all the aggregates studied release alkalis in the solution. For most of the aggregates tested, storage at 60 °C allowed the release of a larger quantity of alkalis than at 38 °C, except for the cataclasite at 38 °C (26 weeks) and the granodiorite at 60 °C (52 weeks). Cataclasite showed anomalous behavior at 60 °C, for which the release was slightly higher at 38 °C than at 60 °C. The temperature did not seem to be an important factor for this aggregate. However, the granite sample is the one that showed the lowest values of alkali release, mainly at 26 weeks. The effect of temperature was the most severe, with an increase of 150% from 38 °C to 60 °C at 26 weeks and of 57.1% after 52 weeks.

According to Menéndez et al. (66. Menéndez, E.; Prendes, N.; Márquez, C.; Aldea, B. (2012) Evaluation of granitic aggregate behaviour in relation with the alkaline extraction and compositional change in their phases. 14th ICAAR, Austin, EEUU, Ma 2012. Paper 052011
), the potential contribution of alkalis is released from the granodiorite-generated products of the alkali-silica reaction in the concrete of a dam. The analysis of the granodiorite from the quarry used for the manufacture of the dam concrete with different extraction solutions showed that solutions (Ca(OH)₂ sat.) and (0.1M NaOH + 0.6M KOH + Ca(OH)₂ Sat.) were the most efficient; however, the alkali potential extraction could have been overestimated. In fact, the proposed rock samples were tested under conditions that did not strictly reproduce the field concrete conditions. Additionally, the particles tested are smaller than those applied in the manufacture of concrete with a larger specific surface area, which can in turn lead to higher reactivity.

The order of alkali release was as follows: phonolite > dredged sand (mainly basalt) > cataclasite > granodiorite > granite.

3.4.3. Limitations of the method

 ⌅

The results of the RRT1 pointed out some limitations of the method and showed some aspects that might be owing to the dispersion of the values obtained. An explanation for the variable results was needed to improve the method. It was verified that the method was not exactly the same in all five laboratories. In Table 7, the amount of excess calcium hydroxide (Ca(OH)₂) used in the alkaline solutions to obtain hydrated lime saturated solutions is presented, along with the different acidification procedures applied at the laboratories involved.

Table 7.  Procedures for Ca(OH)₂ addition and for acidification procedures applied at the five laboratories.

Participant	Added Ca(OH)2 (g/L)	Acidification
Université Laval (Canada)	0.5	0.5 mL HNO3/10 ml alkaline solution
SINTEF (Norway)	2	20 drops HNO3/10 ml alkaline solution
IETcc-CSIC (Spain)	2	3 drops HNO3/10 ml alkaline solution
LNEC (Portugal)	12	Dilution 1:20 of alkaline solution in HNO3 1 %
R.S.E.S.p.a (Italy)	20	Concentrated nitric acid. until pH<2

Furthermore, there was induration of the solids over time. Agglomeration of the particles could be due to the presence of 10% of the sample mass smaller than 125 μm. Although the samples were stirred twice a week, the energy applied to break up the agglomerations may vary among laboratories. Moreover, the fact that the sample was of solid (0.5 kg) and alkaline solution (2 L) contributed to the compaction of the particles owing to the difficulty of homogenization.

By analyzing the results obtained during RRT1, it was clear that the method would need to be improved to obtain a test that safeguards repeatability and reproducibility.

4. CONCLUSIONS

Five samples of aggregate were characterized from the physical, chemical, and mineralogical points of view, including petrographic analysis and the performance of a new suggested method, AAR-8, to extract alkalis using an alkali solution saturated with calcium hydroxide. The test performed at five independent laboratories demonstrated that the release of Na₂O_e increased with time; however, the curves did not level out after 52 weeks of exposure, even though the rate of release significantly decreased over time. There was also a slight increase in alkali release with increasing temperature.

It is concluded that the mineral composition of the samples, as well as the size of the crystals, the existence of cracks, and the texture of the rocks, influence the exposure to the solution used in the tests. On the other hand, the sample that releases a higher amount of alkalis (phonolite) also presents a much higher alkali content in bulk rock analysis compared to other samples; however, the much higher content in the alkalis of phonolite is not proportional to the substantially higher release in alkalis. This result demonstrates that the release of alkalis depends on other factors, such as the stability of the mineral or glass in which the alkalis are bound, and the texture of the rock. The sea-dredged sand, although presenting a low content of alkalis in the bulk rock analysis, has a relatively high release of Na₂O, which suggests that the alkalis contained within the volcanic glass are much more easily released than those contained in crystalline minerals. In addition, some seawater may have remained after washing, contributing to the release observed in the tests. Finally, the investigation showed that the potential for alkali release is more closely related to the mineralogy of the sample than to the alkali content determined by bulk rock chemical analysis, which demonstrates the importance of properly identifying the components of the aggregates that have significant release potential, such as feldspathoids and volcanic glass.

From the results obtained in this testing campaign, several changes are proposed to be made to the test in order to improve the method: