3.1 Materials

In the evaluation, 19 aggregate samples were used (13 of coarse aggregate, 3 of coarse sand and 3 of fine sand), all considered suitable for concrete (^{UNIT, 2005}) from quarries located in different areas of the country (Table 4). Additionally, the mineralogical composition and the alkali reactive components are indicated according to IRAM 1531 (^{IRAM, 2016}) and IRAM 1512 (^{IRAM, 2012}). The samples from ordinal 1 to 6 were samples of coarse aggregate called “granites”, which presented microcrystalline quartz and strained quartz (Qz/mc and Qz/t) as reactive components. Ordinals 7 to 10 were basalt samples and presented altered volcanic glass (Vg) as a reactive component (also sample 10 presented opal). Samples 11 and 12 were metadolomite, and sample 12 presented microcrystalline quartz (Qz/mc). Sample 13 was a pebble with the presence of microcrystalline quartz (Qz/mc) and chalcedony (Ccd) as reactive components. Samples 14 to 19 are fluvial deposit sands with microcrystalline quartz (Qz/mc) as a reactive component (also sample 16 presented chalcedony). In Figure 1, photomicrographs with crossed polarized light of the samples are shown.

Figure 1 Photomicrographs with crossed polarized light of the samples 

The cement used for the mortars and concretes is a Portland cement containing only clinker as the main component, without mineral additions and for general use in concrete (Table 5), identified as CPN 40 according to UNIT 20. (^{UNIT, 2017}).

Mortars and concretes were manufactured using the standard IRAM 1674 (^{IRAM, 1997}) and IRAM 1700 (^{IRAM, 2013}) procedures, respectively. The proportions are shown in Tables 2 and 3.

3.2 Methodology

Three expansion test methods were applied (AMBT, CPT-60 and CPT-38). In the AMBT test, the length of the mortar bars was recorded at 0 days (initial length, prior to immersion in the 1 N NaOH solution at 80 °C), 14 days, and 28 days after immersion. In the accelerated concrete prism test, (CPT-60), the length was recorded initially at 0 days (prior to conditioning at 60 °C and 100% RH) and at 13 weeks. Finally, in the concrete prism, traditional test method (CPT-38), the length was recorded initially at 0 days (prior to conditioning at 38 °C and 100 % RH) and at 52 weeks. For each test method, the expansion E(ti) of each of the bars was determined according to (1):

Eti=100*(Lti-Lt0)/Ln (1)

Where:

E (t_i): expansion to test age t _i

t_i: test age, expressed in days for the AMBT and in weeks for the CPT-60 and CPT-38 test methods.

L_ti: bar length between pins at age ti, in mm

L_t0: bar length between pins at time 0, in mm

Ln: effective bar expansion length, in mm

The expansion at age t _i per aggregate sample was obtained as the mean expansion of three bars of the same mortar tested simultaneously, previously calculating the coefficient of variation per sample to verify the precision between bars according to ASTM C1260 (^{ASTM, 2014}) for the AMBT and according to IRAM 1700 (^{IRAM, 2013}) for concrete prisms.

After calculating the expansion and adopting the limits of Table 1, the degree of reactivity for each sample was determined, and the correlation between the results by applying the three test methods was analyzed. Each sub-sample of aggregate included to evaluate the expansion at three test methods was extracted from the same sample by quartering and fractionation.

4.1 Degree of reactivity according to the applied test method

Table 5 shows the reactivity degree of the aggregates according to the expansion in the mortar bars (AMBT) at 14 and 28 days, the expansion in concrete prisms, accelerated test (CPT-60) at 13 weeks and expansion in concrete prisms, traditional test (CPT-38) at 52 weeks (1 year).

In the samples of coarse “granite” aggregate (sample ordinal from 1 to 6) in the AMBT, mortars 1, 2, 3 and 6 presented an expansion at 14 days of less than 0.10% and at 28 days of less than 0.19%. Therefore, under this test method, these samples are classified as having a degree of reactivity R0 (non-reactive). Granites 4 and 5 exceeded the limit of 0.10% at 14 days, both being classified with the degree of reactivity R1 for this age. However, adopting the expansion at 28 days, sample 4 would be classified as non-reactive (R0) and sample 5 as R1. In the concrete methods CPT-60 and CPT-38, sample 1 presented an expansion of less than 0.08% and 0.04% (limits of both methods, respectively) and was classified as R0 (non-reactive) applying these methods. The concrete prepared with sample 2 and that prepared with sample 6 presented an expansion above the limit, being classified as R1. The concrete with sample 3 presented the highest expansion of all and was classified with a degree of reactivity R2 in CPT-60 and with R1 in CPT-38.

In the basalt samples (sample ordinal from 7 to 10), according to the AMBT method, the expansion at 14 and 28 days classified sample 7 as R1, sample 8 as R3 and R2 (14 and 28, respectively) and samples 9 and 10 as R3. In the last two, expansions that exceeded 1% at 14 days were obtained. However, in the concrete evaluation, basalt samples 7, 8 and 9 were classified as non-reactive (R0), and sample 10 was classified as R1.

In the metadolomite samples (11 and 12), all the evaluation methods classify sample 11 as non-reactive (R0) and sample 12 as R1. The pebble (sample 13) was classified as R1 by all methods.

In the sands, samples 14 and 15 presented the same behavior, in the AMBT and in CTP-38, both were classified with reactivity degree R1 and by CPT-60 as non-reactive (R0). Both samples came from sediments the Santa Lucía river, although they were extracted from different areas.

The sand samples 16 and 19 presented a very similar behavior in the expansion, being classified with reactivity degree R1, except in the expansion at 28 days in AMBT of the sample 16, which was slightly less than 0.19% what classifies it as non-reactive. Both samples were from the same origin (Rio Negro sediments).

Samples 17 and 18 were classified as non-reactive (R0) except for CPT-38 in sample 17 which exceeded the expansion of 0.04 % (at 52 weeks) and was classified as R1.

Adopting the CPT-38 test method as the most reliable and considering the non-coincident results obtained according to the other methods (“no” cells of Table 6), it is observed that erroneous interpretations known as false-positives and false-negatives are reached. False-positives are identified as reactive by the method used, while the expansion in CPT-38 is less than 0.04% (which is associated with a non-deleterious in-service behavior). False-negatives, on the other hand, are aggregates of negligible expansion due to the test method used but which present an expansion greater than 0.04% in the CPT-38 (associated with deleterious in-service behavior).

Analyzing the results that did not match with those obtained in the CPT-38, in the AMBT at 14 days, eight samples yielded erroneous results (42% of the samples), and considering the expansion at 28 days, ten results were erroneous (53%). In the coarse aggregates, the false-negatives of this method belonged to the granite samples (CM, MA, LC, DA) that presented microcrystalline quartz (Qz/mc) as a reactive component. In the sand samples, it was in AGRN where a false-negative was observed (with Qz/m as a reactive component). Regarding the false-positives of the AMBT, they were in four basalt samples of the five samples tested. Both false-positives and false-negatives of this test method have been reported in the literature (Madsen et al., 2018; Torres, 2014; Garber et al., 2005).

In the expansion in the concrete prism, accelerated test method CPT-60, six samples yielded results that did not match with the traditional test method (32% of the samples). Five correspond to false-negatives, two in coarse granite aggregate (LC and LS) and three in sands (AGP, AGPP and AFT). All these samples presented microcrystalline quartz (Qz/mc) and strained quartz (Qz/s) as reactive constituents; for these, the limit of 0.08% should be adopted instead of 0.04% (Table 1).

If the limit of 0.04% was considered instead of 0.08%, there would be three non-coincident results, samples PR48, MA and AFJ (PR48 and AFJ, both exceeded the limit of 0.04% in CPT-60 but not in CPT-38).

4.2 Correlation between expansion results

In Figure 2, the expansion in the AMBT was plotted at 14 and 28 days of testing vs. the expansion in CPT-38, which allowed us to observe the weak correlation between both test methods. Both the points close the abscissa axis and the points close the ordinate axis show results that deviate from the correlation. The first ones are due to granite samples (with Qz/mc and Qz/t as reactive components) and the latter to basalt samples (with the presence of Vv). The results in basalts reported by Madsen et al. (2018) were included, observing that two of the three samples presented weak correlation (points close to the ordinate axis).

ASTM C 1778 standard (ASTM, 2020) defines a zone on the chart for reactive aggregates where there is good correlation between both tests. Among the reactive aggregates, only three remained in the zone of good correlation (AGRN, AFRN, CNC), and only three aggregates were classified as non-reactive by both methods (AFJ, PR48 and CNC1).

Therefore, the mortar bar test method (AMBT) resulted in low reliability to evaluate the potential reactivity of the aggregate, although it is considered adequate to evaluate the uniformity of an aggregate quarry or to evaluate the ability to inhibit the reaction of a preventive measure taken (ASTM, 2020).

Figure 2 Expansion at 14 and 28 days in AMBT vs. expansion at 52 weeks in CPT-38.  

In Figure 3, the expansion of the CPT-60 accelerated test method vs. the expansion using the traditional CPT-38 test method was plotted, including all the samples evaluated. An adequate correlation between the two was obtained (R² = 0.78). The number of samples per type of aggregate was limited, so it was not possible to make a correlation line for each type. The data reported in the literature grouped by reactive constituents (Qz/mc, Qz/t, Vv and others) were added to the same graph.

The aggregates with the presence of volcanic glass (Vv) (identified as basalts) presented an adequate correlation both in own and in the literature data (Madsen et al. 2018 y ^{Falcone et al. 2012}).

^{Falcone et al. (2012}) showed that the aggregates with the presence of Qz/mc and Qz/t were grouped closer to the ordinate axis (points in black outline, unfilled), and therefore, they recommended a limit of 0.08% for aggregates with these reactive constituents in the CPT-60 methodology (Table 1). However, our own results did not show a similar trend for this type of aggregate.

Figure 3 Expansion at 13 weeks in CPT-60 vs. expansion at 52 weeks in CPT-38. 

The CPT-60 method is relatively new, with limited evidence of correlation with the rest of the methods and of correlation with the be havior of the in-service aggregate. A greater number of studies should be available to analyze whether the limit of 0.08 % in the CPT-60 at 13 weeks for aggregates with the presence of microcrystalline or strained quartz is correlated with the limit of 0.04 % in the CPT-38 at 52 weeks. In this sense, there is another method of analysis proposed by RILEM (Nixon and Sims, 2016). In this analysis, by applying the mathematical model known as the Kolmogorov-Avrami-Mehl-Johnson model (abbreviated KAMJ model), the kinetic parameters of the reaction (ln (k) and M) specific to the material under study and to the expansion methodology used are obtained. These parameters are linked to the expansion limits of Table 1, and the reactivity of the sample is obtained (^{Gowripalan and Sirivivatnanon, 2017}; ^{Mangialardi, 2002}; ^{Vila et al., 2020}; ^{Johnston and Fournier, 2000}).