Factors Influencing the Electrical Properties of Ettringite Binders as Repair Materials

,


Introduction
Binders that produce ettringite as the main hydration product, a.k.a.ettringite binders, are increasingly being used in repair applications because the rapid hydration and strength gain associated with ettringite formation allows for expedited construction timelines [1].A main goal of such repairs may be to extend the service-life of reinforced concrete for which rebar corrosion caused by chloride ingress has caused deterioration and spalling.This issue raises the question: Does the repair material have transport (e.g., diffusion) properties low enough to prevent further degradation and reach the desired servicelife?
Electrical resistivity is emerging as a preferred indicator for transport properties [2].Evaluating the transport properties of ettringite binders is especially important since chloride binding products like Friedel's salt have not been observed in laboratory studies of calcium sulfoaluminate (CSA) systems [3,4].However, the efficacy of these methods has mainly been studied in ordinary portland cement (OPC) systems and studies have found discrepancies between the electrical properties of ettringite systems and actual rebar corrosion tests in mixtures produced with ettringite binders.
For example, Afroughsabet et al. found higher electrical resistivity in a CSA cement than OPC, and this was interpreted as suggesting lower transport properties (e.g., permeability and/or ionic diffusivity) in CSA systems than OCP systems [5].However, the water absorption was higher in CSA and corrosion readily occurred due to lower pore solution alkalinity.Moffatt et al. suggested that the pore solution alkalinity is a reason for higher resistivity in CSA systems, despite poor durability in other tests [3].
Another possible factor to consider in relating electrical resistivity to transport properties is the amount of pore solution in the system, i.e., the degree of saturation.Ettringite binders rapidly consume water during hydration (e.g.internal drying), and this could lead to greater electrical resistivity than an OPC sample that maintains a higher internal relative humidity [6,7].However, methods have been developed to correct for varying levels of saturation in resistivity measurements [8].
The objective of this research was to investigate the role of such factors in the electrical properties so a more accurate interpretation towards transport properties could be obtained.In addition to the experimental observations, thermodynamic calculations and pore partition modelling were used to support the findings and provide insight on how the phases assemblages may have influenced the electrical properties.

Methodology
The cements used in this study were a commercially available calcium sulfoaluminate cement (CSA) and Type III ordinary portland cement (OPC).The oxide compositions of both cements are shown in Table 1.Three mixtures of 100% CSA, 70%OPC+30%CSA, and 100% OPC were prepared with w/c = 0.5 using a vacuum mixer for 90 seconds at 400 rpm, manually scraped to ensure homogeneous mixing, and then repeated.These will be referred to as "OPC", "Blend", and "CSA" mixtures, respectively.The mixtures containing CSA cement included 0.1% citric acid per weight of cement to regulate working time.Cylinders (50 mm diam.x 100 mm height) were cast then sealed and placed on rollers at 50 rpm for 24 hours.After 24 hours, samples were demolded and immediately tested or placed in a freezer for later testing.
Bulk electrical resistance measurements were taken immediately after demoulding using a Giatec RCON 2 device with an AC current of 1 kHz.Further details on performing bulk resistance measurements and calculating resistivity has been presented elsewhere [9,10].The pore solution was then extracted using the same apparatus described in Ref. [11] and method for pastes described in Ref. [12].A maximum pressure of 320 MPa was held constant for 1 minute once attained, before unloading.Solutions were carefully sealed after collection and then analysed using XRF to determine the Na + , K + , SO4 2+ , Al 3+ , and Ca 2+ concentrations following previously developed methods [13,14].The conductivity of the solution was then calculated using Snyder's approach considering all the ionic species mentioned above [15].The samples that were frozen and not tested immediately after demoulding were later used to measure degree of saturation, porosity and thermogravimetric analysis.The degree of saturation and porosity were determined following AASHTO TP 135 [16] with the following exceptions: 1) the oven-drying procedure was performed at 60ºC instead of 105ºC and the mass was checked more frequently to mitigate ettringite decomposition.The drying process was stopped after the percent change in mass loss from the original mass was less than 0.1%/hr.2) specimen dimensions were 50 mm diameter and roughly 30 mm thick.
Prior to thermogravimetric analysis (TGA), samples were crushed using a mortar and pestle and passed over a 75-micron sieve.Isopropanol exchange was then performed to halt the hydration processes and remove most of the free water [17].A 40-45 mg sample was then used in TGA (Q50, TA instruments) with heating rate of 10ºC/min up to 1000ºC in a nitrogen environment.The degree of hydration was calculated from the difference between weight loss at 105ºC and 1000ºC and an assumed 0.25g H2O/g of cement for the mixtures containing OPC [18].The entire weight loss up to 1000ºC and an assumed 0.35g H2O/g of cement was used for the CSA mixture.

Thermodynamic modelling
Thermodynamic modelling of the studied systems is performed using the GEMS3K [19] software coupled with the CEMDATA v18 database [20].GEMS3K performs thermodynamic modelling by determining the phase assemblage of a cementitious system that minimizes its Gibbs free energy.The GEMS/CEMDATA framework can be used to calculate the molar amounts of solid, aqueous, and gaseous products of reactions and the activities of ions in the pore solutions at thermodynamic equilibrium.During calculations, the formation of some of the carbonate-ettringite, hydrotalcite, and hydrogarnet phases were blocked based on evidence from the literature showing that these phases are frequently not observed to form in substantial quantities in cementitious systems at ambient temperatures (less than 60°C) [21,22].
The thermodynamic calculations were performed at the measured degrees of hydration of OPC and CSA.The kinetics of the OPC clinker reactions were simulated using the modified Parrot and Killoh model [23].It was assumed that C2S and C4AF reactions in CSA followed the same kinetics as the corresponding phases in the OPC mixture.Other CSA reactions were assumed to be proportional to the degree of hydration of CSA.
The pore partitioning model (PPM) [24], which synergistically combines thermodynamic calculations with the Powers-Brownyard model [25], to determine the amount of gel water and capillary water in the studied mixtures systems.

Results and Discussion
The results of this study will be discussed in three sections.First, the electrical resistivity measurements are presented followed by the determined physical properties.Lastly, the physical properties are used to help interpret the electrical properties of the different binders.

Electrical Resistivity Measurements
The resistivity measurements for samples immediately after sealed curing are summarized in Table 2.The CSA mixture had the highest bulk resistivity and pore solution resistivity.The OPC and Blend mixtures had similar pore solution resistivities and their bulk resistivity was less than half that of the CSA mixture.In support of these findings, microstructural development after 24 hours in CSA cements can be expected to be higher than that of OPC [7] and the pH of CSA cements has been reported to

Physical Properties
To better understand the resistivity measurements, the degree of saturation, porosity, and degree of hydration were measured.Oven-drying at 60ºC revealed that the CSA sample had the lowest moisture loss as shown in Fig. 1.The OPC sample had the highest moisture loss and required slightly longer than 48 hours to reach less than 0.1%/hr change in percent mass loss.The drying process could be a potential source of error if ettringite decomposition occurred.However, all the samples contained sulphate in the pore solution at one day, and literature suggests that 60ºC does not cause excessive drying of ettringite if sulphate is present in the system [30,31].Fig. 1 shows that samples were removed from the oven as soon as the rate of mass loss had decayed to help prevent ettringite dehydration.The resulting duration of two days at 60ºC is a method that has previously been used in literature for CSA samples prior to porosimetry [4].However, it is acknowledged that the drying at 60ºC could result in lower moisture loss and lower porosity measurements than conventional drying process at 105ºC.This is also indicated in Fig. 1 where the moisture loss curves still show a positive slope at 48 hours, particularly for the OPC containing mixtures, indicative of incomplete drying.
Fig. 2 shows the results of TGA with major DTA peaks labelled for the CSA sample.The OPC and Blend samples exhibited much lower DTA peaks in the range of ettringite and this range may have still contained residual evaporable water after the isopropanol exchange.The Blend sample displayed a lower calcium hydroxide peak in the range of 400-450ºC than the OPC sample, and similarly appeared to show lower peaks for the peaks seen in the CSA sample.
The degree of hydration was calculated from the TGA data and reported with the degree of saturation and porosity in Table 3. Commensurate with its higher degree of hydration, the CSA mixture had the lowest porosity and saturation.The OPC and Blend mixtures showed similar results for all physical properties of Table 3.

Interpreting the Electrical Properties
The most accessible electrical property for cement/concrete is the total sample resistivity.However, this measurement is influenced by a number of factors including: the degree of saturation in the sample (i.e.amount of pore solution), the conductivity of the pore solution, the amount of porosity in the sample, and -more intriguingly -the connectivity of the pore solution in the pore network, which is closely related to the transport properties.Aside from the connectivity, all these factors are also associated with the degree of hydration in a system.Fig. 3 shows that the bulk resistivity is proportional to the degree of hydration for the samples in this study, despite the differing binder chemistries.While this observation supports other research that has shown resistivity is an effective tool for monitoring hydration [31,32], it also suggests that the resistivity is significantly Originally derived for geological applications, the formation factor is used to decouple the pore solution resistivity, from resistivity of the porous material so that only microstructural properties of the pore network (i.e., porosity and pore connectivity) remain as the influencing factors.With this approach, formation factor can be linked to transport properties of the porous material such as its permeability/absorption capacity [33] and ionic diffusivity [34,35].
Eq. 1 calculates the formation factor at the given saturation, which will be denoted as FF*.
The formation factor at a completely saturated state can be theoretically calculated from Eq. 2 [8] and is denoted here as FFsat.
The q term in Eq. 2 is a correction factor for the nonlinearity of change in solution conductivity with saturation and can be calculated from the ionic strength of the solution using Eq. 3, where G is a conductivity factor assumed to be 0.4 (mol/L) 1/2 and IM is the ionic strength of the solution [8].Using the formation factor at complete saturation (FFsat) the connectivity of pores (β) can be determined from: which indicates that FFsat is a quantity that is only a function of the properties of the pore network, hence, can be used to relate to transport properties of the material.FFsat can be experimentally determined by taking resistivity measurements on a fully saturated sample, but this was not a preferred method for the present study since the vacuum saturation procedure would have significantly altered and extended the hydration after only one day of sealed curing.
Figure 4 shows a comparison of FF* and FFsat to evaluate the effect of accounting for varying degrees of saturation in the different systems.The formation factor prior to correcting for saturation, FF*, displays the same general trend as the bulk resistivity measurements with the CSA mixture exceeding the OPC containing mixtures.However, the FFsat value for the CSA sample is smaller than that of the OPC mixture.This implies the lower free water content of the CSA samples was a primary factor in its higher resistivity measurement.In support of this observation, the percent mass loss during drying shown in Fig. 1 was roughly twice as high for the OPC containing mixtures compared to the CSA mixture.Similarly, the CSA mixture only released 0.01 mL/g of sample during the pore solution extraction procedure and the OPC and Blend mixtures released around 0.02 mL/g of sample.Greater self-desiccation has been well documented in CSA hydration and is a practical advantage for some building applications where "fastdrying" is desired [6].Drying due to evaporation during sample storage and handling resulted in negligible mass loss.The CSA mixture demonstrates that varying hydration processes can result in vastly different amounts of free water, which can greatly influence the electrical resistivity.
The ramifications of the lower degree of saturation in the CSA sample are seen in the calculated connectivity of the pores, β, which is reported in Table 4 alongside the formation factor results.The connectivity is highest in the CSA system followed by the blended system.This could be attributed to the varying hydration products in each system that determine the resulting morphology and tortuosity of the microstructure.Given that ettringite is the most abundant hydration product in CSA according to the TGA data in Fig. 2 and literature [27], the high connectivity of the CSA mixture could indicate that ettringite forms a more connected pore network than binders containing C-S-H.
Microscopy of the needle-like morphology of ettringite has been shown in prior research and discussed as a point of concern for transport properties [36].The enhanced connectivity of the CSA binder shown in this paper also corroborates studies that have measured higher diffusion of ions in CSA binders [3,5].However, the Long-term durability studies have shown that CSA samples can perform well in seawater environments for over 14 years [37].The use of seawater instead of NaCl in corrosion tests has been shown to yield better results for CSA since the Ca +2 and SO4 -2 ions can replenish decomposed ettringite.These observations suggest that later age testing is necessary to fully understand the pore refinement of ettringite binders using electrical properties or other methods.

Thermodynamic Modelling
Hydration modelling using GEMS was performed to predict the hydrate assemblage of each binder system, and these results are shown in Fig. 5.The amount of ettringite in each binder according to modelling ranked as expected: CSA > Blend > OPC.C-S-H was the most abundant phase in the Blend system, but less C-S-H and CH was formed than in the OPC sample.The phases seen here generally agree with the TGA results in Fig. 2, although more straetlingite is proposed in place of gibbsite.The GEMS results were then used as input in the pore partitioning model [24] to calculate the volume fractions of unreacted binder, gel solids, gel water, capillary water, and chemical shrinkage.These results are shown in Fig. 6.Ignoring chemical shrinkage, the pore volume can be calculated as the sum of gel and capillary water and was 37%, 33% and 36% for the OPC, CSA, and Blend mixtures, respectively.The modelled porosity differs from the experimental results in Table 3, and this could be attributed to experimental error in the drying process, which may not have removed all gel water present in the system.The effectiveness of vacuum saturation in the porosity procedure could also be a potential source of error.Regardless of the discrepancy in porosity values, the model results agree with the general trends seen in the experimental results for porosity and degree of saturation.
The model results show less capillary water and a greater amount of chemical shrinkage in the CSA mixture, which agrees with the lower degree of saturation determined for the mixture.This suggests that there was less fluid present in the CSA samples that would act as a conductive medium during electrical measurements.The large amount of gel water in the CSA mixture was associated with ettringite and likely did not have a major contribution to the electrical conductivity of the samples.Therefore, accounting for saturation is a necessary step for normalizing results between different binders since the OPC and Blend mixtures intrinsically had a greater amount of capillary water according to Fig. 6.
These modelling results broadly support the notion that the higher resistivity measured in the CSA system was due to the lack of saturation in the pore network rather than decreased connectivity of the pore network.The GEMS results corroborate the TGA results in showing ettringite as the main hydrate phase in the CSA mixture, and less abundant in the OPC and Blend mixtures.The connectivity measurements in Table 4 increase in relation to the amount of ettringite reported by GEMS, and decrease with respect to the amount of C-S-H.This finding suggests that blended systems with OPC hydration phases may have durability benefits over pure CSA systems.

Conclusions
The measured electrical resistivity of a CSA binder was found to be higher than that of OPC at one day of hydration.While the pore solution resistivity was also greater for CSA, this did not fully account for the greater total resistivity.The degree of saturation was a significant factor for the total resistivity and normalizing the formation factor to a completely saturated state resulted in a higher determined connectivity for the CSA and Blend mixtures.The hydrate assemblages from thermodynamic modelling support the observation that ettringite binder phases may have higher pore connectivity than OPC phases regarding transport phenomena, however later age testing is warranted before any conclusions on the longterm durability of specific ettringite binders can be made.

Fig. 3 .
Fig. 3. Comparison of bulk resistivity and degree of hydration.
tested samples at one day of hydration since later age pore solution extraction was unfeasible for the CSA sample using sealed curing.

Table 1 .
Oxide compositions of OPC and CSA cements.Specific gravities (S.G.) also shown, which were used in the PPM.

Table 4 .
Formation factor and connectivity data.