Fresh and hardened properties of concrete
Table 5 shows the fresh and hardened properties of the concrete mixtures. The air content values ranged from 0.9 to 9.0%. Only in two cases (OPC-6.2 and FA35-0.9) were the target air contents not achieved. Loss of air content or the quality of the air void system is not considered in the current study; further studies on this topic are required. Neither the fresh concrete slump nor concrete strength were measured as prior studies showed no correlations between these properties and level of damage . On the contrary, high-strength, OPC-only mixtures showed the worst damage resistance.
The 91-day bulk resistivity values for the concrete mixtures (Table 5) tested in this study were plotted against air content (Fig. 2). Values from a previous study, using the same materials and concretes with similar air contents, but cured for 28 days are also shown in the figure . All mixtures showed increasing bulk resistivity as the curing duration increased from 28 to 91 days. However, the increase in the control mixture was negligible when compared to mixtures with SCMs, consistent with findings from literature which show that the incorporation of SCMs increases bulk resistivity, especially at later ages . This increased bulk resistivity implies that the concrete has a more refined microstructure and reduced conductivity of the pore solution (due to pozzolanic/latent hydraulic reactions that result in greater alkali binding) [19, 30, 31]). The refined microstructure suggests that the concrete would be more resistant to solution ingress (and therefore, calcium oxychloride damage). The 91-day bulk resistivity increased as the SCM replacement increased and was higher for the more pozzolanic fly ash compared to the latent hydraulic slag [30, 31]. The air content did not have a strong impact on the bulk resistivity, except for the FA35 mixture, which showed a sharp increase in the bulk resistivity at the highest air content. Bulk resistivity measurements when curing is done in a moist room are somewhat complex to interpret and ideally these measurements should be done on specimens immersed in limewater or simulated pore solutions, which ensures controlled leaching and degree of saturation . These last two curing choices were deliberately not utilized here because of the potential for reactions between calcium hydroxide that is deposited on the surface and deicing salt solutions that complicate interpretations of damage .
Figure 3 shows a plot of the 91-day formation factor values (calculated using the measured pore solution resistivity) versus the bulk resistivity values. There was a strong linear relationship between the two parameters. Theoretically, because the bulk resistivity is affected by a number of experimental variables, it is often stated that the formation factor is a better parameter to use for durability assessments . In this study, while there is some scatter, formation factor and bulk resistivity were interchangeable. This was because the range of bulk resistivity was much greater than the range of pore solution resistivities (the latter only varied from 0.27 to 0.49 Ohm m, with most values falling in a narrower range as shown in Figure S1). Similar correlations have also been reported in literature . While the formation factor is considered by many to be a more fundamental parameter, its measurement can, depending on the followed procedures and saturation procedure, be complex. Therefore, for mixtures which do not have unusual admixtures or SCMs, the bulk resistivity may be an adequate proxy for the formation factor.
Visual monitoring during exposure
Figure 4 shows the damage progression in Group 1 (Fig. 4a), Group 2 (Fig. 4b), and Group 3 (Fig. 4c) specimens during exposure. The results show that increasing SCM replacement and air content significantly reduce the damage at any given exposure duration, an observation in alignment with literature . The increase in durability due to air entrainment could be linked to the reduction in the degree of saturation due to changed (slower) sorption behavior of the air voids [3, 33, 35]. In addition, the air voids provide ‘space’ which reduces expansion/crystallization pressures associated with the formation of calcium oxychloride and other phases . The increase in durability due to SCMs is because they reduce the amount of calcium oxychloride that forms (Eq. 1) through dilution, pozzolanic, and latent hydraulic reactions [15, 19, 20, 22, 23]. In addition, SCMs at later ages also likely reduce the ingress of the salt solutions due to a continuous refinement of the microstructure from pozzolanic reactions . Finally, increased chloride binding in mixtures with SCM due to Friedel’s salt formation  could also reduce the amount of chloride available to form calcium oxychloride. In all cases, an increase in SCM replacement levels resulted in further damage reduction . In all specimens, mixtures with low/no SCM and air entrainment showed the earliest signs of damage (often in the first 50 days of exposure) and the most rapid damage progression.
In Group 1, OPC-1.9, OPC-6.2, FA20-1.6, and SL20-1.1 failed after 39, 122, 122, and 100 cycles. Three mixtures with the highest levels of SCM replacement (FA35-4.3, FA35-8.1, and SL35-9.0) did not show any sign of damage at the end of the exposure duration. By comparing Fig. 4b and c, damage caused by the different salts showed the same trends. In Group 2, OPC-1.9, FA20-1.6, and SL20-1.1 failed after 336 days, 600 days, and 600 days, respectively. Five mixtures with the highest SCM levels did not show any damage during exposure. In Group 3, OPC-1.9, OPC-6.2, FA20-1.6, and SL20-1.1 failed after 336 days, 600 days, 600 days, and 600 days, respectively. Six mixtures did not show any damage during exposure. Generally, damage was most severe in Group 1 (20% CaCl2, temperature cycles 5 to 20 °C), and the least severe in Group 2 (20% CaCl2, constant temperature 5 °C). Most crucially, in all exposure conditions, the effects of SCMs and air entrainment were consistent, with both substantially reducing damage. The same observation was made in harsher exposure conditions (25% CaCl2, temperature cycles − 8 to 25 °C), suggesting that the effects of SCMs and air entrainment in improving durability are universal and independent of exposure conditions . The mixtures exposed previously to harsher conditions showed significantly more damage than the mixtures in this study, but it is unclear if that was largely due to the harsher conditions or if it was because those specimens were only cured for 28-days [10, 15]. It is known that both these factors (curing duration and exposure solution concentration) do affect damage significantly [10, 15, 20].
Mass changes during exposure
Figure 5 shows the mass change values for control (Fig. 5a), fly ash (Fig. 5b) and slag (Fig. 5c) Group 1 mixtures. The mass increases as the specimens are exposed to the salt solutions [15, 35]. Despite the scatter, the mass change behavior is roughly linear. In previous work , distinct non-linearity/bi-linear mass change behavior was apparent in mixtures with high SCM and high air content. Sorption in cementitious materials when plotted against the square root of time is known to follow bi-linear behavior with initial and final sorption behaviors being controlled by matrix saturation and air void saturation, respectively [3, 33, 38]. The linear behavior and the low values of mass change could imply that the air voids in high SCM/high air mixtures are unsaturated, due to a more refined microstructure and less aggressive exposure conditions [15, 34]. However, sorption in cementitious materials, even with water, is complex, and affected strongly by specimen preparation and other factors [33, 34, 38, 39]. For high concentrations of salt solutions, due to the chemical interactions affecting transport , advanced reactive transport models are likely needed to fundamentally explain mass change behavior. At any rate, further studies and direct measurements are needed to quantify air void saturation in these systems.
Mass change is also affected by spalling and leaching processes, however, it appears in this case, that solution sorption dominated, as in all cases, the rate of mass change (slope of the curve) was reduced as the air content and SCM replacement level increased. Often, mixtures which showed poor durability showed high values of mass change (5% or greater), and mixtures which showed high durability showed low values of mass change (2% or lower). However, there were several exceptions, and there was no single level of mass change which corresponded to failure. In actuality, as SCM and air content increased, the mass change value that corresponded to failure increased. For the SCMs, this is because there is lower chemical damage potential at higher replacements due to the lower amount of calcium hydroxide [3, 21]. For the air, the reason is because a higher amount of solution needs to be absorbed in higher air content mixtures to reach degrees of saturation comparable to lower air content mixtures [3, 40, 41]. Therefore, the rate of mass change is critical. Because enough specimens have not failed, it is unclear how exactly the mass change at failure depends on SCM and air content. If this relationship could be determined, then based on the rate of mass change, times to failure/service life could be determined through curve fitting . The final measurement is shown as “X” in these figures. This point is not reliable because significant spalling can occur around failure (reflected as a sharp drop in mass), which interferes with mass change measurements. Therefore, the mass change point that is before this final point was used for fitting and developing relationships (as in Table 6).
When comparing fly ash and slag mixtures, the fly ash mixtures had lower mass changes, reflecting their overall lower damage status (Fig. 4a).
Figure S2 and Figure S3 (Supplementary Material) show the mass change values for control, fly ash, and slag mixtures in Group 2 and Group 3, respectively. Results from Group 2 and Group 3 were strikingly similar to those from Group 1: mass change behavior was roughly linear, increase in SCM replacement and air content resulted in lower rate of mass change in all mixtures, and tolerable mass change values increased as the SCM replacement and air content increased. The extent of sorption (mass change) was lower in Group 2 and Group 3 (constant temperature) than in the harsher exposure conditions in Group 1 (low temperature cycles), suggesting that the temperature cycling contributes to an increase in damage. One important difference in Group 3 mixtures is the significant initial reduction in mass in these specimens which was not that obvious in the other specimens. This initial decrease may be linked to calcium hydroxide leaching, which is known to be occur when mixtures are exposed to MgCl2 . Support for this hypothesis comes from the fact that this initial decrease is not as apparent when the SCM replacement increases, which is expected because calcium hydroxide content also decreases as the SCM replacement increases. However, the reduction is almost 1% in some mixtures, which would need a large proportion of the present calcium hydroxide to leach, which may be unrealistic given the time and the specimen size. Therefore, leaching of other phases or gradual damage could also contribute to the reduction. Subsequent increases in masses in these specimens are higher than in Group 2 specimens. The increased leaching likely contributed to a more porous microstructure, which drove greater mass sorption, explaining the somewhat higher damage in Group 3 specimens when compared to Group 2 specimens (Fig. 4).
Table 6 shows the slope of the mass change curves multiplied by 100, based on linear fitting for all Groups 1 (SLP). The values are shown for all groups using a common unit of days. The final mass change (FMC) for the Groups and the damage state at the end of the test is also shown. As SCM replacement and air content increased, SLP reduced, and damage at the end of the test also reduced. At equivalent damage levels, FMC (or the tolerable mass change before failure) increased as SCM replacement and air content increased.
Bulk resistivity changes
Bulk resistivity values over time for the control, fly ash and slag mixtures in the Group 1 specimens are shown in Fig. 6. The bulk resistivity behavior is the opposite of the mass change behavior, with the bulk resistivity decreasing over time. However, the behavior is clearly non-linear; prior work has shown power-law fits the bulk resistivity evolution . The bulk resistivity behavior in this case appears to follow either power-law or a bi-linear behavior. The level of scatter in the measurements makes fitting somewhat challenging. Regardless, most mixtures show a point at which the rate of change of bulk resistivity changes. The bulk resistivity at this point, similar to a nick point , increases with SCM replacement and air content. At any given cycle, the bulk resistivity of the concretes increases with air content and SCM replacement, consistent with findings from Figs. 4 and 5. This trend continues until failure or the end of experiment. Previously, based on a comparison of resistivity with visual damage, it was suggested that a bulk resistivity threshold of 4 Ohm m corresponded to failure . Not enough specimens failed in these exposure conditions to confirm or deny this threshold value (or if the threshold of bulk resistivity also depended on SCM replacement and air entrainment). Regardless, specimens with higher initial bulk resistivities showed lower mass loss and experienced lower visual damage at any given number of cycles.
Figure S4 and Figure S5 in the Supplementary Material show the bulk resistivity values for Group 2 and Group 3 specimens. A detailed discussion is not presented here as the conclusions from these specimens are the same as those in Group 1 specimens. The CaCl2 caused a greater decrease in bulk resistivity (57% on average for all mixtures) in comparison to MgCl2 (36% on average for all mixtures) possibly linked to its greater conductivity at 20% concentration .
After 150 cycles/600 cycles, all remaining mixtures were taken out of the solution and the cylinders were capped and tested for compressive strength. Results are shown in Table 7. Specimens with severe damage had 3.0 to 4.2 MPa strength (average 3.8 MPa), those with moderate damage had 6.4 to 23.0 MPa strength (average 11.2 MPa), those with minor damage had 11.2 to 23.5 MPa strength (average 16.5 MPa), and those with no damage had 11.4 to 32.1 MPa strength (average 18.2 MPa). The increasing strengths with reduced damage suggest that damage classification broadly works, although there is clearly scatter and some overlap between classifications. On average, strengths were 14.2 MPa, 17.1 MPa, and 15.3 MPa for Group 1, Group 2, and Group 3 mixtures, which suggests that Group 1 exposure conditions were the harshest and Group 2 conditions were the least harsh. Post-exposure, the strongest mixtures were mixtures with 35% slag or fly ash, largely independent of the type of exposure.
Table 8 shows the estimated solution absorption values for all specimens at the end of testing. Specimens which absorbed the highest amount of solution inevitably failed (OPC-1.9), whereas specimens which absorbed low amounts of solution (SL-9.0) showed limited or no damage at the end of testing. Similar to the strength after exposure, the solution absorption depended strongly on the SCM replacement and air content. Because of variabilities in both parameters and the complexities associated with the measurements, the correlation between the strength and absorption was rather poor (not shown).
In earlier work, we stated that mixture bulk resistivity multiplied by air content could be used to predict concrete resistance to calcium oxychloride damage . Because the exposure conditions in this study are less harsh and the mixtures are more durable due to the longer curing, limited number of specimens failed. Therefore, it was considered that predicting time to reach a failure classification would not be reliable. Thus, we evaluated whether the bulk resistivity multiplied by air content could predict the number of days taken for Group 1 mixtures to reach a damage classification of Moderate (Fig. 7). Broad trends did not change when using Group 2 or Group 3 mixtures, or when using the damage classification of Minor instead of Moderate. The figure shows a moderate correlation between bulk resistivity multiplied by the air content and the time required to reach Moderate damage, confirming that this could be a performance-based specification against damage. Another way of looking at these results is to consider the relationship between damage state at the end of testing and the product of bulk resistivity and air content (Table 9). While these is clear separation between mixtures showing no damage (average value of 2005 for the product of bulk resistivity and air content) and mixtures showing failure (average value of 178), there is some overlap in the values for minor, moderate, and severe damage. Because bulk resistivity does increase with air content for some mixtures, it is possible that the weights for these parameters need to be reduced. These attempts by using the square roots for bulk resistivity and air content are also shown in Table 9. Of these parameters, BR*Air content0.5 works the best in differentiating mixtures with different levels of damage. Other parameters were also tested and using bulk resistivity and air content appears to be reasonably, but not completely, effective in differentiating mixtures which show minor, moderate, and severe damage. Prediction accuracy could be further increased by considering the quality of the air void system, or by accounting for varied degree of saturation.
As results for the three Groups were similar, durable mixtures in Group 1 exposure were identified and are assumed to be representative. This was done using two conditions—one was damage status of none/moderate/minor at the end of testing and the second condition was that compressive strength post-exposure was larger than 10 MPa. As shown in Table 10, both conditions result in the same mixtures, with two exceptions (OPC-8.5 and SL20-7.0). This provides further validation that the damage classification is relatively robust. If we consider both conditions need to be met, the durable mixtures are FA20-5.1, FA20-7.0, FA35-0.9, FA35-4.3, FA35-8.1, SL35-4.5, SL35-9.0. The results are broadly similar to those from an earlier study , where durable mixtures were those with 20% SCM and 8% air and 35% SCM and more than 4% air. Because of the longer curing duration employed here, the fly ash shows better performance than the slag. Therefore, the durable mixtures were: 20% fly ash, 5.1% air or more; 35% fly ash, 0.9% air or more; 35% slag, 4.5% air or more. For freeze–thaw resistance, less than 5% air is sub-optimal, therefore, the durable mixtures can be considered to be 20% or more fly ash and 5% or more air; 35% or more slag and 5% or more air.
Based on our results, we explain the mechanisms of mitigation for SCMs and air.
SCMs have two damage mitigation mechanisms. The first is through a reduction in the rate of solution ingress, apparent when considering the lower slopes in the mass change curves (Table 6), which occurs in almost all cases when SCMs are used. The reason why SCMs reduce sorption rates is linked to the microstructural densification of the cementitious matrix, which is apparent from their increased bulk resistivity values, especially as curing duration increases (Fig. 2)—this is an expected consequence of the pozzolanic reaction . The second mitigation mechanism is through a reduction of damage potential, driven by reductions in the calcium hydroxide and calcium oxychloride contents [7, 20, 23]. The outstanding performance of certain mixtures, for example, Group 1 FA35-0.9 (Table 7), which show only minor levels of damage at the end of testing provides evidence for this hypothesis. Considering this mixture does not have entrained air, and the deicing salt solution is clearly being absorbed, the minimal damage observed in this mixture cannot be linked exclusively to a reduction in solution sorption, but it must be also connected to the inherently low amounts of calcium hydroxide present in the system. Indeed, beyond certain SCM replacement limits, no calcium oxychloride forms, and damage is not observed .
The air mitigates damage by reducing the degree of saturation at any given time. This is a consequence of the slower sorption associated with greater air void contents . The greater total porosity due to the air volume will also reduce the degree of saturation at equivalent amount of solution absorption. This mitigation mechanism of air is reflected in the slower rate of mass gain in concrete mixtures with greater air contents (Fig. 5a and Table 6). Table 6 shows that increasing air by 4% roughly halves the slope of the mass gain curves. The second mechanism through which the air acts is by reducing crystallization/expansive pressures through provision of space for pressure relief. A similar mechanism is suggested for air reducing freeze–thaw damage . This mechanism is supported by the finding that specimens with higher air contents fail at higher levels of mass gain. Specifically, Table 6 shows that Group 1 OPC-1.9 failed at 2.6% mass gain whereas Group 1 OPC-6.2 failed at 7.5% mass gain. The OPC-6.2 mixture needed to absorb three times as much solution as the OPC-1.9 mixture to fail. These mixtures have no SCMs, and differing mass change kinetics do not affect mass change at failure. Therefore, the higher mass gain tolerance of the OPC-6.2 mixture is likely because the air is allowing for greater solution sorption. This finding does not conclusively demonstrate that the air reduces pressures, however, it does provide support for the hypothesis.