Corrosion products at 80°C
Geochemical and reactive transport modeling provides a reasonably accurate estimation of the mineralogy and chemical conditions at the steel interface, considering factors such as temperature rise and diffusion within the system. These geochemical parameters, which are difficult to measure directly, were effectively taken into account. Although the mineralogy remains relatively stable throughout the process, the chemistry of the porewater undergoes some changes, yet remains predominantly highly alkaline overall. Regarding the redox conditions induced by the concrete itself, it is currently uncertain. However, considering the anaerobic nature of the setup and the presence of slag, it is expected that reducing conditions would be prevalent. Introducing metallic iron at trace level in the modeling of the initial geochemistry gives a calculated redox potential (Eh, SHE) ranging from − 600 mV (with HS/SO4 couple) to -750 mV (without HS/SO4) at a pH of 11.9 and a temperature of 80°C.
Figure 10 shows some possible corrosion products that may form at the steel/concrete interface for the calculated chemical conditions. The Fe – O system points out the competition between goethite (Fe(III) oxyhydroxide) and magnetite (Fe(III)-Fe(II) oxide) according to the redox potential. It is worth mentioning that hematite will replace goethite in the same diagram if included in the system. If the system is further complexified by introducing Ca and Si, silica-hydrogarnet can prevail on magnetite at high alkaline pH but oxidizing to mildly reducing redox potential. Fe(II)-silicates corrosion products such a greenalite can only precipitate for pH lower than 10.5 at 80°C and reducing conditions.
Magnetite was clearly identified as the main corrosion product by micro-Raman. The SEM-EDX analyses in Fe and O of corrosion products are also consistent with magnetite. Magnetite is usually the main product under anoxic corrosion of the carbon steel [5, 14, 16, 24]. No other corrosion products as Fe-silicate or Fe-sulfide were identified by this technique. Magnetite remained stable over the 12 months and did not convert to iron silicates and/or iron sulfides as it has been observed in cementitious/bentonite grout at lower pH [25]. Hematite has been identified in most crossed-sections, usually close to the steel surface and always in relatively low quantity compared to magnetite. It is possible that hematite was present since the beginning as mill scale [13], or else due to the transformation of early oxic corrosion products during the preparation of the experimental set-ups and then heating to 80°C.
Fe-siliceous hydrogarnet CFSH is one the most stable iron phase in hydrated cements [14, 22]. SEM-EDX spectrum 1 of Table 4 indicates a mixed signature between concrete and iron oxide elemental composition. The analysis may correspond to a Fe-siliceous hydrogarnet, but it is most probably an intrusion of concrete in a crack within the steel. It seems that in the present system magnetite formed more quickly on the steel surface than CFSH formation controlled by Ca and Si diffusion from the concrete. Chomat et al. have also seen Fe-enriched layers at the concrete interface with steel [16]. Pally et al. have observed hydroandratite (CFSH) deposits on the iron plate at 80°C but in their experimental set-up diffusion could have been facilitated by the direct contact of the steel with a synthetic solution [18].
Corrosion mechanism and rate
Microstructural characterization shows strong similarities in the corrosion patterns in 3, 7 and 12-months samples. Magnetite was found all along the steel surface although the formation of 20–60 µm long Fe-oxide ingrowths were also identified. The literature states that the carbon steel surface is likely to be passivated due to the formation of a stable magnetite film [5, 16, 25]. This is in line with the short transient stage of negative corrosion potential measured in the three cells for 50 days at maximum. The occurrence of a uniform passive corrosion mechanism is, therefore, likely.
A mean corrosion rate was derived from observed thicknesses during SEM analysis and from the obtained corrosion potential curve (estimating that the altered layer was formed only in the first 50 days). In this study the estimated corrosion rate is in order of magnitude of 10 µm/y. Similar values of early corrosion rates have been found for CEM I and the corrosion rates usually decrease exponentially with time to reach rate smaller than 0.01 µm/y within a year [17]. The long-term corrosion rate could not be determined in the present experiments but was clearly much lower than 10 µm/y. Temperature increase seems to be inoperative for the long-term corrosion rates [17].
There was no trace of any localized corrosion mechanisms. The moderated aqueous concentrations in chloride (500–1000 mg/L) had no effect, in agreement with that did not find any consequences of high concentrations of chloride on the corrosion of steel under highly alkaline anoxic conditions [17].