2.1 Materials

The materials used in the assay were: cement CP V - ARI, fly ash, pyrite and limestone filler.

The composition of the cementitious system was performed in laboratory, by replacing 35% of the CP V - ARI, in paste, by fly ash, this amount between 15% and 50%, in paste, established by ^{NBR 5.736 (1991)}, in order to classify the composition as a typical type CP IV cement.

The CP V - ARI cement used had specific mass of 3.13 g/cm³ and specific BET area of 1.07 m²/g. Fly ash, from the combustion of mineral coal from the reservoir of the Valley of the river Tubarão, had specific mass of 1.95 g/cm³ and BET specific area of 1,09 m²/g.

Contaminant added to paste was pyrite (FeS₂), resulting from the beneficiation of mineral coal from the region of Figueira - Paraná. This material, also studied by ^{Pereira (2015)}, has, in its chemical composition, a purity of 94%. The material has specific mass of 4.95 g/cm³ and, after grinding, was classified as a material passing through the sieve with mesh openings of 2.4 mm.

Limestone filler, inert material, was used to compensate the contaminant mass in the pastes, its percentage variable according to the amount of pyrite added to the mixture. The specific mass of carbonaceous material was 2.84 g/cm³.

2.2 Pastes with mineral addictions

The follow-up of reactions due to ISR and consequent implications in the mechanical properties of the hydrated matrix were evaluated by means of study in Portland cement paste. The reason for executing the study in paste was to facilitate the identification of products formed by the degradation mechanism, since the presence of sand and gravel would generate the occurrence of crystalline peaks, of higher intensity, in the X-ray diffractograms, making the identification of compounds formed by ISR difficult.

3.1 Consistency of pastes by Kantro’s mini cone assay

The reference group, without the addiction of pyrite, had the greater spreading among the other groups. Observing Figure 1 it is possible to see that the spread was decreasing with the increase of the addition of contaminating material.

Figure 1 Consistencies of the series of cement pastes tested by Kantro cone test. 

This is according to what was exposed by ^{Katsiadramis et al. (2010)}, because materials of greater granulometry, in this case pyrite (D_max equal to 2.4mm), act as obstacles, increasing the viscosity of the compound. This is contrary to the fine grains, in this case the filler, which act as lubricants increasing the fluidity of the material.

Besides the mentioned physical effect, there is the possibility of a chemical effect in consequence of the interactions of cement particles with pyrite’s ions.

3.2 Specific mass in fresh state

Figure 2 indicates the growth of specific mass during the replacement of the filler by pyrite.

Figure 2 Specific mass in the fresh state of the cement pastes series tested (^{NBR 13278: 2005}). 

In this case a behavior opposite to the one presented in the consistency assay was perceived. As the replacement of the filler by pyrite was being made, the specific mass in the fresh state of the series was increasing. This behavior is due to the fact that the replacement happens between two materials of different specific masses: filler (2.84 g/cm³) and pyrite (4.95 g/cm³). Since a material of lower specific mass is removed and one of greater specific mass is added, the specific mass of the compound is increased. The gradual substitution of filler also collaborated with an increase of the consumption of cement in the paste, in volume. It was perceived that the consumption for the series Ref. (0.0%) was 943.162 kg/m³, while series SO₃ (0.5%), SO₃ (1.0%) and SO₃ (5.0%) have consumptions equal to 945.857 kg/m³,948.588 kg/m³ and 971.820 kg/m³, respectively.

3.3 Resistance to axial compression and to traction in flexion

In Figures 3 and 4 are presented the values of resistance to axial compression for all of the analyzed groups (28 and 84 days).

Figure 3 Axial compressive strength in cement pastes at 28 days (underwater curing condition). 

Figure 4 Axial compressive strength of the cement pastes at 84 days of age (28 days in submerged cure + 56 days of exposure). 

With values obtained at 28 days, it was possible to observe that group reference and 0.5% of SO₃ have resistances to compression of 20 MPa, considered as equal, according to Tukey’s test with 95% of trust. Series 1% and 5% of SO₃, despite the proximity of values and overlap of the error bar, could not be considered as statistically equivalent by the test used. Thus, at the age of 28 days, the series having more resistance was the one of 1% of contamination with SO₃.

The greater values of resistance to axial compression of the series with higher amounts of contamination, SO₃(1.0%) and SO₃(5.0%), may be explained by different factors. First, as previously exposed, those two traces have more consumption of cement (3.03%). This may had contributed for the resistance at the age of 28 days being higher than the others. Another factor that could explain this increase of resistance is the filling of voids in the pastes, by products formed by the internal attack of sulfates. This is according to ^{Araújo (2008)}, because the formation of products coming from the internal attack of sulfates in the voids, fills the spaces, generating gains of resistance in an initial stage. ^{Biczók (1972)} and ^{Ouyang et al. (2014)}, studying the external attack of sodium sulfate, also reported that with the increase of compactness, there is a gain of resistance at an initial stage.

Besides the analysis between groups, it was possible, for the age of 84 days, to evaluate the influence of exposition. Again, with the application of the test, it was observed that both exposition conditions had statistically equivalent resistances.

Regarding the series, it was noted that, at this age, series Ref. (0.0%) and SO₃(0.5%) continued indicating statistical similarity, in both conditions of exposition. However, it was also noted statistical equivalence of series SO₃(0.5%) with SO₃(5.0%), in both studied conditions. This similitude between the group of lower contamination and the group of greater contamination does not indicate fall of resistance by the last one, but an increase of resistance that is inferior to the others. During the condition of wetting and drying, for instance, group Ref.(0.0%) had, at 84 days, a gain of approximately 69%, while group SO₃(5.0%), under the same condition, had increase of 38.11%. As much as there is no fall in the obtained value, a change in the behavior of groups is perceived already at the age of 84 days. Series SO₃(1.0%) was the one, again, with the greater value of resistance.

The delay in resistance gain observed to series SO₃(5.0%) is according to results from ^{Pereira (2015)} and ^{Oliveira (2013)}. Thus, it is supposed that the sulfate provided by the initial reactions of oxidation of pyrite changed the hydration kinetics of the cement, having differences in resistances already at 84 days.

In Figure 5 are displayed the values of flexion strength obtained for pastes at 28 days of age. In this case, it was perceived that the reference paste and the pastes with amounts of contamination by SO₃ had statistically equal resistances, between 5 and 6 MPa.

Figure 5 Bending tensile strength of the cement pastes at 28 days of age (submerged curing condition). 

In Figure 6 are displayed results of flexion strength at 84 days of age. It was observed that the standard deviation between samples was high, with coefficients of variation near 20%, which collaborated for Tukey’s test not indicating statistical difference between all samples. Analyzing exposition conditions, at 84 days, it is noted that only series SO₃(1.0%) had statistical difference regarding itself, the lower value being obtained in the cycle condition.

Figure 6 Bending tensile strength of the cement pastes at 84 days of age (28 days submerged cure + 56 days of exposure). 

Besides the statistical equivalence observed between series at the age of 84 days, it was perceived that there was no significant difference between the age of 28 days and the age of 84 days, from all of the studied series. Thus, it is not possible to declare that there was gain or loss of resistance between both ages (28 and 84 days) for all groups.

3.4 Dimensional Variation

In Figure 7 is presented the dimensional variation of the pastes, when submitted to weekly cycles of drying and wetting. The reference paste had the greater retraction between the studied pastes. Retraction happened progressively until the age of 11 days, moment when it stabilized and started to fluctuate as a function of the type of exposition to which it was submitted. The partial substitution of the limestone filler by pyrite, in different amounts, caused the reduction of the retraction of pastes. The use of contaminated material, in different amounts, caused the reduction of the retraction of pastes. The use of contaminated material in amount of 1.0% SO₃ reduced retraction to, approximately, half of the retraction observed in the reference paste. With the amount of 5% of SO₃ practically no retraction was observed, which may be verified at the age of 28 days, moment before the exposition to the drying and wetting cycles. This behavior indicates the beginning of the oxidation of pyrite and the formation of expansive hydrated compounds, to be analyzed in the presentation of the X-ray diffractograms. From 28 days of age, the exposition to the cycles of drying and wetting cycles enabled the dimensional variation (retraction/expansion) of the paste that, in average, was similar to the dimension observed at 28 days, the moment before the procedures of drying and wetting. This behavior indicated that the oxidation reaction of pyrite was not significantly influenced by the variation of exposition condition (until 84 days of age of the tested compounds), which aimed to provide, mainly, oxygen for its occurrence.

Figure 7 Dimensional variation of the test specimens subjected to the wetting and drying cycle at 84 days of age. 

In Figure 8 are presented the results of dimensional variation of pastes when submitted to the initial underwater curing of 28 days, at the tank with continuously aerated water. Dimensional variation until the age of 28 days was equivalent to the one observed in Figure 7. In this case no variations (oscillations) were noted, as in the previous exposition condition. It was noted that, from 28 days of age, there was no significant variation in any of the studied series.

Figure 8 Dimensional variation of the specimens submitted to the aerated tank at 84 days of age. 

Generally, there was a tendency of forming expansive compounds due to the internal attack of sulfates when the contaminated cementitious material was exposed to drying and wetting cycles.

In Figure 9 is presented, comparatively, the dimensional variation of the pastes, at ages of 28, 56 and 84 days, as a function of the SO₃ amount and of the exposition condition after the initial underwater curing in water saturated with lime.

Figure 9 Dimensional variation versus SO₃ content added in the cement pastes series tested. 

Exposition of the pastes to the drying and wetting cycles (Figure 9-A) evidenced that, for amounts of SO₃ until 1.0%, no influence was verified, even at the same ages, in the dimensional variation observed at 28 days. For the amount of SO₃ of 5.0%, the exposition to the drying and wetting cycles tended to cause expansion in the paste. In the case of exposition to the tank with constantly aerated water (Figure 9-B), independently of the SO₃ amount, there was no dimensional variation regarding the age of 28 days. Those data are indications of the importance of continuing the monitoring of the dimensional variation over time, being possible to generate greater difference among the series of assays at most advanced ages due to the progress of ISR.

The linearization of results presented in Figure 9, by means of the square root of the SO₃ amount, enabled to establish the slope of trend lines for ages 28, 56 and 84 days, representing the coefficient of internal attack by sulfates. In Figure 10 is presented the dimensional variation as a function of the amount of contaminant contained in the pastes and of the exposition condition. Coefficients of determination (R²) of the trend lines were over 0.979, indicating that the linear model used properly represented the behavior observed in the pastes.

Figure 10 Dimensional variation as a function of the contaminant content (SO₃) contained in the cement pastes and the exposure conditions: A - drying and wetting cycles; B - tank with aerated water. 

In Figure 10-A is evidenced that the exposition condition of drying and wetting tended to potentiate the internal attack by sulfates, because there was expansion of pastes over time in contamination of 5.0% SO₃. This behavior was not observed during the exposition to the aerated water tank (Figure 10-B). Besides all, the differences presented are small and the monitoring must continue in order to get information from a large time of exposition to ISR.

3.5 X-ray diffractometry

The analysis of hydrated compounds formed after the different types of exposition, at the age of 56 days, was performed comparatively to the respective samples with 28 days, age when started the exposition of the pastes to the drying and wetting and the submersion in continuously aerated water, with the aim of verifying if those conditions influenced the reaction of pyrite, responsible by the formation of deleterious hydrated compounds in the hydrated matrix.

In Figure 11 are presented the diffractograms of the four cementitious compositions studied at the 28th day of age. Analyzing the diffractograms, it is possible to say that the products formed, independently of the composition of the different pastes, were the same ones, with some differences regarding to calcite, pyrite and ettringite. Calcite (CaCO₃), a constituent of the limestone filler, has main peak (of greater intensity) in the region of 29º 2θ and, thus, as this material was replaced by pyrite (FeS₂), it started to stand out jointly with the peaks of pyrite, located at 33, 37 and 56º 2θ. Phases identified in diffractograms were the mullite, quartz, calcite, ettringite, portlandite, the hydrated calcium monocarboaluminate and the pyrite, in pastes with addiction, as identified by the ICDD.

Figure 11 Diffractograms of cement pastes at 28 days of age in submerged curing (T - Mulite, P - Portlandite, E - Etringite, K - Calcite, M - Calcium monocarboaluminate hydrate, Q - Quartz, I - Pyrite). 

Ettringite, with characteristic peak of higher intensity located at position 9.1º 2θ, was identified in all pastes, however with different intensities. Despite the technique of X-ray diffractometry being qualitative/comparative, the variation of intensity of the characteristic peek may indicate the presence of this phase in greater of smaller amounts. The analysis of this peek evidences that the contamination of the paste with pyrite intensified the formation of ettringite which demands, necessarily, an extra source of sulfate ions besides those made available by the gypsum part of the Portland cement. This behavior constitutes evidence that the pyrite contained in the pastes is taking part in the reactions, in other words, there was oxidation of iron sulphide for the formation of the intermediary phases. The paste with amount of 5.0% of SO₃ had a more pronounced peek of ettringite than the other pastes. These results corroborate with the behavior observed in the dimensional variation of pastes at the age of 28 days. The paste with the higher amount of contaminant (5.0% SO₃) had greater relative dimensional variation, justified by the possible formation of a greater amount of ettringite, being this an expansive phase.

In Figure 12 are presented the diffractograms of the four cementitious compositions, comparing the chemical phases formed at 28 days of age, after underwater curing, with those ones formed at 56 days, for both conditions of exposition (drying and wetting and aerated water tank).

Figure 12 X - ray diffractograms for the analysis of the evolution of RSI comparing the 28 - day cure and the changes from the types of exposure (T - Mulite, P - Portlandite, E - Etringite, K - Calcite, M - Calcium monocarboaluminate hydrate; Q-Quartz, I-Pyrite). 

Phases indicated were the same commented for Figure 12, with the variation of intensity of the peek of ettringite being highlighted, more pronounced for the amount of 5.0% of SO₃.