Hydrogen production through aluminium corrosion in a cement‐based matrix

In France, deep geological disposal is considered for the storage of high and intermediate‐level long‐lived radioactive wastes. For aluminium, the possibility to encapsulate the wastes in a cement‐based matrix is studied. However, cement being an alkaline environment, aluminium can lose its passivity, starts to corrode leading to hydrogen evolution in the infrastructures and generate a possible explosive hazard after decades of storage if hydrogen can accumulate somewhere in the facility. It is therefore necessary to study the corrosion behaviour of aluminium in the different cements considered for the encapsulation to estimate the possible amount of hydrogen that could be generated through corrosion and design the cement capsules accordingly. This work mainly focused on the reaction occurring at the aluminium‐cement interface. Raman spectroscopy did not highlight significant differences in the nature of the corrosion products forming at the cement/aluminium interface, leading to the conclusion that it is not the chemistry of the cement that is the key factor controlling the corrosion rate but rather the physical properties of the cement matrix.


| INTRODUCTION
In France, different disposal facilities are considered for the radioactive waste according to the level of radioactivity of the wastes and the half-life of the main radionuclides.Surface disposal is used for low and intermediate short-lived radioactive waste while deep geological disposal is considered for high and intermediate-level long-lived radioactive waste.
The storage procedure of aluminium radioactive waste selected by ANDRA (french acronym meaning the French Agency for Nuclear Waste Management) consists of the encapsulation of the material in a cement-based matrix before surface or geological storage.Since the cements are alkaline matrices, aluminium depassivation is possible, [1][2][3][4] leading to hydrogen evolution via the cathodic reactions of the corrosion process.This hydrogen evolution raises two safety concerns.The first is the risk of the loss of the confinement of the waste due to cement fracture under internal pressurisation if the gas accumulated in the capsule, for instance at cement heterogeneities or at the cement/ aluminium interface.The second is the risk of hydrogen release out of the capsule that could then locally accumulate in the storage facilities and generate an explosive hazard.
To ensure the safety of the disposal facilities and the integrity of the waste conditioning, the amount of aluminium that is disposed in each waste package must be specified and is limited to mitigate the level of hydrogen produced by aluminium corrosion.
In a previous study, [5] the corrosion resistance of an aluminium alloy (grade 5754) in different cement matrices (CEM I and CEM III/C according to European standards) was studied in different configurations (cure, exposure to controlled relative humidity, resaturation with water, etc.) at room temperature.In each case, hydrogen evolution was monitored to address the actual corrosion rate variation versus time.Different behaviours were highlighted depending on the cement type.In all cases, when the cement was resaturated with water after the cure, a restart of the corrosion process was observed after variable time lags.However, the corrosion tended to stop in CEM III/C while it continued in CEM I, all along the duration of the exposures, leading in some cases to the cracking of the cement matrix.These cracks were attributed to the internal pressure rise, through H 2 production in the corrosion process and/or to the internal mechanical stress generated on the cement by the growth of the corrosion products layers.
To explain the different corrosion rates of aluminium depending on the cement environment, an additional research programme was initiated aiming at studying the evolution of the environment at the metal/cement interface during resaturation.A specific test cell has been developed to allow for electrochemical and local pH measurements at this interface.This article presents the results of the work conducted to date in the frame of this additional programme still in progress.

| EXPERIMENTAL 2.1 | Electrochemical measurement at the interface cement-aluminium
The measurement system implemented for this work is schematically presented in Figure 1.The system consists of an electrode produced with the aluminium grade 5754 and introduced in a cement matrix (CEM I, CEM III/C or CEM V) which is then immersed in the resaturation solution named S1a (laboratory solution specified at ANDRA to assess the interactions of water infiltration in the ground with cementitious materials and the other components in the waste disposal site) at room temperature.The cement compositions were the following: -CEM I: 95% of clinker.
The aluminium sample was machined thick enough to allow the drilling and the installation of a capillary tube used in the reference electrode holder setup to measure the potential at the interface cement-aluminium (Luggin capillary).The amount of cement used was around 40 g per cm² of aluminium.
Other holes have also been considered for carrying out local pH measurements using long micro-probes (Inlab NMR model from Mettler Toledo).The opening of these tubes was positioned at two different distances from the interface (one at the aluminium surface and the other away from the interface to make measurements at the interface or in the cement bulk).
A counter electrode (316 L grid) was also placed around the cement block to perform some electrochemical measurements (polarisation resistance (R p ) and electrochemical impedance spectroscopy [EIS]).A platinum wire was also immersed near the open end of this capillary to allow redox potential measurements.
The aluminium sample 5754 was a disc with 60 mm in diameter and 15 mm in thickness.The mortar was made using the following mass proportions: 28% of cement, 58% of sand and 14% of water (mass proportions).
The cure of the cement was carried out in a polyvinyl chloride (PVC) tank with a diameter of 10 cm.First, a 4 cm layer of cement was introduced at the bottom of the PVC mould before positioning the aluminium sample and filling the full mould with an additional cement layer (4-5 cm).
After 4 days of cure, the cement block was removed from the mould and immersed with the instrumentation system in a sealed PVC tube containing the resaturation solution S1a which composition is given in Table 1.In this part, the cure duration was only 4 days to avoid the drying of the reference electrode while in our previous study, [5] a cure up to 1 month was considered.
The polarisation resistance (R p ) measurements were carried out based on the guidelines given in the ASTM G59 standard [6] (ΔE = ± 20 mV instead of 30 mV, 0.125 mV/s, equilibrating time of 60 s).
EIS measurements were performed at different times after the test start.The impedance diagrams were acquired at the open-circuit potential under potentiostatic control over a frequency range from 100 kHz to 10 mHz with an amplitude of ± 20 mV.The spectra were plotted by recording 8 points per decade.
At the end of the test, the samples were extracted from the cement matrices and cleaned with nitric acid in an ultrasonic tank according to the recommendations of ASTM G1. [7] For aluminium alloys, the best practice is to use 70%v/v nitric acid at room temperature.

| Corrosion product analysis
The samples were prepared in PVC moulds with the same geometry as for the electrochemical measurements.First, a layer of mortar was placed at the bottom of the PVC mould.After a couple of minutes, the aluminium coupons (approximately 30 × 15 mm were placed on the mortar on two perpendicular diameters to be at the same distance from each other.They were then covered with another layer of mortar. After the cure of the cement, the capsules containing the aluminium coupons were removed from the mould and immersed in a container containing the resaturation solution.For each type of cement (CEM I, III/C and V), five cement capsules were prepared to follow the evolution of the corrosion scale after different durations: after the cure (1 month for this part) and after different periods of resaturation (100, 500, 1000 and 3000 h).
At the end of each exposure period, the samples were extracted from the cement matrix and stored under vacuum until Raman analysis of the corrosion products were made.
The analyses were performed using a Raman spectrometer (Horiba-Jobin Yvon Xplora plus model equipped with a TE-cooled CCD camera (1024 × 256 pixels)).The measurements were performed with a 532 nm laser source and magnification ×100 or ×50 (long range) as a function of the thickness and morphology of the corrosion products layer.For each sample, several areas were analysed and several points were recorded in each.A minimum of two samples per condition were analysed.

| pH measurements
To analyse the differences in corrosion rates obtained in the three considered cements, a continuous monitoring of the pH was carried out in each cement matrix.The measurements were made at the surface and at a distance of 1 cm from it, that is, in the cement bulk.The results obtained are shown in Figure 2.
The initial pH was close to 13.5 in the different studied cements, whatever the considered measurement was at the metal/cement interface or in the bulk.This pH was close to the one expected in a solution in equilibrium with the selected cement matrix.
In all cases, the pH tended to progressively decrease.This evolution was anticipated because the corrosion process may lead to the consumption of hydroxyl ions according to one of the following reactions: The pH seems more constant in the CEM V than in the other cements and there is also less difference between the two measurement positions.
According to the potential-pH diagram for aluminium [8] (Figure 3) this decrease of pH tends to promote the passivity of the aluminium.Indeed, with only traces of aluminium cation (10 −6 M) the transition between passivity and activity is around pH = 9.5.However, in a closed system where the cations can accumulate and the concentration increase as in a cement/aluminium matrix, this limit is significantly shifted to a more alkaline environment.For instance, with a concentration of cations of 10 −2 M (dashed lines in Figure 3), the border is then at about pH = 12.5.According to our in situ pH measurements, the passivity of the aluminium coupons is therefore possible after a sufficiently long period of exposure and the alkalinity of the cement matrix does not invalidate the storage concept, provided that the integrity of the capsule is ensured during the whole period of storage.The decrease in pH observed for two of the cements (CEM I and CEM III/C) is followed by a sudden rise in pH at the interface after 2500 h for CEM I and 2800 h for CEM III/C.This sudden rise is explained by the cracking of the matrix and the loss of the specimen confinement with an immediate water ingress which is more alkaline.Cracked cement matrices were indeed observed after the tests as shown in Figure 4.After cracking, the direct contact of the very alkaline resaturation solution with the metal due to the loss of confinement can also trigger the corrosion process.
Given the results of the previous study, [5] cracking of the cement matrix was expected for CEM I but not for CEM III/C.This early rupture of the matrix could be due to the geometry of the casting(which was different from the one previously used to allow for the electrochemical monitoring, and to a lower cement thickness).Similar cracks were reported by Spasova et al. [9] and attributed to the mechanical stress generated in the cement-based matrix by the growth of the corrosion products layer at the metal/ cement interface.Herting and Odnevall [10] also observed cracks in the cement-matrix (CEM I) only after 2 years of exposure but the concrete cylinders containing the samples were immersed in an artificial groundwater with a lower initial pH (8.1 ± 0.2), thus promoting a passive state of the aluminium at the beginning of the exposure.After 2600 h of testing, the CEM V matrix was not cracked (Figure 4c).

| OCP and redox potential monitoring
The use of a reference electrode (saturated sulphate electrode, SSE) positioned near the exposed aluminium surface allowed the monitoring of the corrosion potential.
Its evolution in the three cements considered is shown in Figure 5.The potential measured at the beginning of the resaturation was approximately −1350 mV/SHE for CEM I and CEM V and slightly higher for CEM III/C (−1250 mV/SHE).It increased by about 200 mV during the first weeks of resaturation for CEM III/C and CEM V and remained constant for CEM I.This potential evolution suggests that CEM I corroded uniformly throughout the entire exposure period without significant evolution while a protective corrosion scale formed at the surface of the other two cements, leading to a rise of the potential.
These potential values seem consistent with those presented in the work of Lahalle [11] obtained on an F I G U R E 3 Aluminium potential-pH diagram of aluminium. [8]The drop in potential in CEM III after approximately 1800 h of immersion is attributed again to the cracking of the matrix leading to a direct contact between the aluminium and the resaturation solution.This interaction resulted in higher corrosion of the aluminium sample, similar to the one observed in CEM I.It was not possible to visually correlate the drop in potential with the rupture of the cement because the moulds used were not transparent.This is also the reason why the test was continued after the drop in potential, resulting in corrosion rates than those which would have been measured if the confinement had been maintained.
The evolution of the redox potential (measured using the Pt electrode) of each cement matrix is given in Figure 6.It was anticipated that, if the hydrogen evolution is sufficiently high, a low redox potential would be measured.As shown in Figure 6, the redox potential in the CEM I is very low (−700 mV/SHE) already at the beginning of the exposure, indicating a significant hydrogen production at the surface of the specimens.On the contrary, the redox potential in CEM V remains very high (0 mV/SHE) during the whole duration of the test, suggesting a low rate of hydrogen evolution.Since hydrogen evolution is a product of the cathodic reaction in the corrosion process, the corrosion rate in CEM V should thus be relatively low.
The behaviour of the CEM III/C is intermediate with a drop of potential after about 1200 h of exposure, indicating an increase of the hydrogen evolution after this period of time and suggesting therefore that the corrosion rate significantly increased after this duration, maybe after the cement started to crack.

| Polarisation resistance measurements
A 316 L stainless steel counter electrode was positioned all around the surface of the cement, rendering possible electrochemical investigations, mainly polarisation resistance measurements (Figure 7).However, these measurements are strongly influenced by the cement layer surrounding the sample.Indeed, the polarisation resistance measurement characterises the resistance of the whole electrochemical system, including the resistance of the electrolyte.In the present case, our electrolyte consists of the layer of cement saturated with water and its resistance may be higher than that of an ionic solution conventionally used in electrochemical tests applied to corrosion.
Despite these limitations, the study of the evolution of the polarisation resistance can give semi-quantitative information useful for the analysis of the corrosion of aluminium in the different cements.
For CEM I, the polarisation resistance is less than 2000 Ω. cm² and tends to decrease throughout the test (900 Ω. cm² after 3 months).The measured R p is very low compared with the one obtained with the two other cements, suggesting a higher corrosion rate of the aluminium provided that the resistance of the electrolyte does not significantly affect the R p .It will be shown in the next sections that it is not the case but that the corrosion rate is indeed the highest in this cement.This result seems also consistent with the OCP evolution of aluminium in this matrix (Figure 5).
In the CEM III and CEM V experiments, the evolution of R p is completely different.The polarisation resistance tends to increase continuously during the test (before cracking for the CEM III), with values reaching more than 15,000 Ω. cm² after several months.In CEM III, the R p suddenly dropped after about 1700 h highlighting the cracking of the cement matrix, the direct contact of the aluminium with the resaturation solution and therefore, an increase of the corrosion rate.The increase of R p versus time again suggests a decrease of the corrosion rate in these two cements over time, again provided that the electrolyte resistance does not significantly affect the polarisation resistance measurement.
One can also note that the polarisation resistance at 3000 h in the CEM III after the cement matrix failure is of the same order of magnitude as the one measured in CEM I, which is in agreement with the fact that the cracking of these two cements was observed after the tests.
The average corrosion rates obtained from the gravimetric measurements are presented in Table 2.
Comparison with the results obtained in a previous work [5] is not possible due to the different cure conditions used.In the tests presented in this study, the cure phase of the matrix was around 4 days to ensure that the pH electrode could not dry out, whereas in the previous study, it lasted 1 month.Nevertheless, the highest corrosion rate was measured after exposure in CEM I, (about 1.5 mm/ year after 4 months of immersion).For CEM III, the value obtained was strongly impacted by the prolonged contact between the aluminium and the resaturation solution following the cracking of the matrix, thus the loss of the encapsulation of the aluminium.In CEM V, the corrosion rate was around 250 μm/year for an immersion period of 110 days.This average value seems in agreement with a study of Fujiwara et al. [12] showing that the corrosion rate of the alloy 1070 P in a Portland cement can be as high as 1 mm/year at the beginning of the curing time but drops down to only some μm/year after about 2 months.

| EIS
A few EIS measurements at the corrosion potential were also carried out.EIS makes it possible to decorrelate different electrochemical phenomenon according to their time constant such as double-layer capacitance, charge transfer resistance, adsorption of intermediates or mass transport.It may also make it possible to evaluate different morphological aspects such as those of corrosion products (e.g., porosity evolution, microporosity characteristic).Such quantitative results are required to build an equivalent circuit (which is sometimes complex) to model the impedance that is in good agreement with experimental data and physical processes occurring.This analysis was not made yet because it requires a dedicated research work but qualitative analyses allow for extracting valuable results.
Figure 8 presents the results obtained on CEM I.The Nyquist presentation is considered in which the X-axis is showing the real part of the impedance and the Y-axis the opposite of the imaginary one.
In CEM I, three time constants are observed, with a capacitive loop at the lowest and highest frequency ranges and an inductive one at intermediate frequencies.The low-frequency loop is ascribed to the corrosion process.The diameter of this loop decreases from about 1000 ohm.cm² after 7 days to only 200 ohms.cm²after 73 days suggesting an increase of the corrosion rate.
Inductive loops are generally the signature of an adsorption process, but these phenomena can also appear capacitive. [13]The nature of the process generating the inductive loop observed in the CEM I cement has not been found yet.If the adsorption process is associated with the charge transfer phenomenon, adsorption has a lower time constant than charge transfer.In the present case, adsorption has a higher time constant.One hypothesis may be that adsorption is associated with cathodic reactivity in parallel with anodic reactivity described by a simple capacitive loop.It is then inconsistent with a layer of insulating corrosion products.This inductive loop is particularly well seen in this cement but not in the others as shown later in this section.
The high capacitive loop can be associated with the dielectric effect of the cement barrier.The diminution of the diameter of the high-frequency loop associated with the decrease of the high-frequency limit can be associated with the hydration of the cement with the resaturation solution (i.e., the water phase is progressively filling the pores of the cement).Indeed, in the Nyquist representation, the high-frequency limit corresponds to the resistance of the electrolyte (R e ).Here, this limit decreases from about 600 ohms.cm² after 7 days of exposure to about 200 ohms.cm²after 73 days indicating an increasing conductivity of the cement barrier over time.
The impedance spectra obtained with CEM III/C are completely different (Figure 9).After a short period of immersion (7-20 days), three time constants can again be observed.An inductive component is visible at a high frequency that could be due here to adsorption phenomena modifying the impedance response of the cement.The two capacitive loops observed at intermediate and low frequencies can be ascribed to the capacitive effect of the cement (and its hydrated pores) and to the Faradaic process (corrosion), respectively.
From 35 days of exposure, only one-time constant remained observable.After this period of time, the cement was probably fully hydrated and the Faradaic process was probably the only one visible.Considering the diameter of the loop at low frequency for short periods of exposure and the one of the loop remaining after long exposure, one can observe an increase of the loop from about 4000 ohms.cm² after 7 days to about 10,000 ohms.cm² after 73 days.This loop being ascribed to the Faradaic process, this evolution suggests a decrease of the corrosion rate.
Finally, considering CEM V (Figure 10), three-time constants are also visible for short durations with one inductive loop, again at very high frequency and two capacitive loops at intermediate and low-frequency ranges.After a sufficiently long period of time, here from 81 days, the data at low frequencies evolve linearly, following a slope at about 45°which suggests a mechanism limited by mass transfer.
In CEM V, one can ascribe the capacitive loop having a characteristic frequency of some Hz to the Faradaic reaction.The diameter increases over time (500 ohms.cm² after 10 days and 6000 ohms.cm² after 110 days) indicates that the corrosion rate was slowing down over time.This cement is also different from the two others regarding the highfrequency limit of the diagram, corresponding to the electrolyte resistance, which tends to increase (from 1200 ohms.cm² after 10 days to 6400 ohms.cm² after 110 days) suggesting a decrease of the cement conductivity and therefore possibly a kind of sealing of its pores according to a phenomenon that is unknown yet.
To summarise the result, the following table presents the transfer resistance estimated from the Nyquist plots for different durations of exposure in the three cements considered.Such evaluation remains qualitative.The transfer resistance decreases over time in CEM I, indicating an increase of the corrosion rate while it increases in the two other cements indicating better protectivity (Table 3).
Even without a fully quantitative analysis, EIS results demonstrate strong differences in terms of degradation mechanisms from one cement to another.

| Corrosion product analysis
After different periods of exposure in the three cements resaturated with solution S1a described in Table 1, some specimens were not descaled to allow for analysis of the corrosion products using Raman spectroscopy.

| CEM I
No obvious aluminium corrosion products were detected on the samples withdrawn after the cure (using Raman analysis).After 100 h of exposure, it was possible to observe Bayerite and Gibbsite as presented in Figure 11.[16][17] It should be noted that the most intense signal for Gibbsite was observed around 3430 cm −1 instead of 3520 cm −1 .According to Rodgers, [16] the relative intensity of the signals in aluminium hydroxide can be influenced by the crystallographic orientation of the sample which can play a role in the protective nature of this type of layer.Even though a mixture of Gibbsite and Bayerite was observed, it should be noted that Gibbsite was mainly identified.Nearly the same compounds were observed after 500 h of ageing with a difference in the observation of isolated, well-crystallised Gibbsite.It was only after 1000 h of exposure that Nordstrandite appeared which is another aluminium polymorph hydroxide.This new compound was never found alone and always coexisted with Bayerite or Gibbsite.It is characterised by Raman shift around 3490, 3653 and 3620 cm −1 with the signal at 3565 cm −1 being the most intense. [16]fter 3000 h of ageing, all three polymorphs of aluminium hydroxides were observed, namely Gibbsite, Bayerite and Nordstrandite (Al(OH) 3 ), see Figure 12.

| CEM III/C
It should be noted that less corrosion products were available for analyses on the CEM III/C coupons compared with those exposed to the CEM I. Furthermore, the analysis by Raman spectroscopy was affected by strong signal perturbations due to fluorescence effects.This interference from fluorescence was also noted by Potgieter-Vermakk et al. [18] for this type of cement.
After the cure and 100 h of ageing, no aluminium corrosion products were observed in Raman.16][17] It was only after 3000 h of ageing that more defined and intense signals were obtained allowing observation of Bayerite and Gibbsite (around 3420, 3542 and 3655 cm −1 for Bayerite and 3524 and 3617 cm −1 for Gibbsite [14][15][16][17] ).The results highlighted that the corrosion product volume tends to increase between 1000 and 3000 h of ageing, which is in agreement with the cracking observed after 1800 h in the other part of this study.

| CEM V
As for CEM III/C, no aluminium corrosion products were observed after the cure for CEM V coupons.After 100 h of ageing, aluminium corrosion products like Gibbsite or Bayerite were clearly identified as shown in Figure 14.It is interesting to note that Gibbsite and Bayerite can be identified separately on the contrary to the coupons exposed in CEM I.After 500 h of exposure, well-crystallised Gibbsite was observed.A hydrated carbonated species was also identified, characterised by signals around 1079 and 3585 cm −1 .This compound most probably came from the cement matrix carbonatation.In addition to this compound, elemental carbon mostly in an amorphous state was also observed at 1343 and 1595 cm −1 , peaks that are characteristic of graphite and disorder bands. [18]This type of compound is frequently encountered in type V cements.After 1000 h of ageing, the well-crystallised Gibbsite already seen after 500 h of exposure is still observed however a mixture of Bayerite and Nordstrandite is also present.After 3000 h of ageing, only the different polymorphs of aluminium hydroxide can be observed with a mixture of Gibbsite and Nordstrandite or isolated Gibbsite or isolated Bayerite as depicted in Figure 15.

| DISCUSSION
The physico-chemical monitoring described in this article highlighted a decreasing trend of the pH down to average values that could be low enough to promote aluminium passivity in a closed system where aluminium cations can accumulate.When the integrity of the cement is ensured, (i.e., if the cement does not crack), low corrosion rates are measured in CEM III/C and CEM V, with a trend to observe a decreasing rate versus time.The behaviour of aluminium in the cement CEM I was completely different with high corrosion rates measured, confirming previous results already published. [5]These evolutions of the corrosion rate were confirmed by polarisation resistance measurement, redox potential monitoring and electrochemical spectroscopy.
For sufficiently long duration of exposure, cracking of the cement capsules was observed in some cases.These physical damages are probably due to hydrogen production/release at the surface of the cement, as well as the growth of the corrosion product layer generating mechanical stresses at the aluminium/cement interface where the crack initiates.When such failure occurs, the corrosion rate of the coupons F I G U R E 13 Raman spectra between 3000 and 4000 cm −1 for CEM III after 500 h (B: Bayerite).
F I G U R E 14 Raman spectra between 3000 and 4000 cm −1 for CEMV after 100 h.increases by one order of magnitude according to R p measurements, due to the direct contact of the metallic coupons with the resaturation solution.
The mechanical resistance of each cement-matrix could be another hypothesis of the cracking.Indeed, the resistance in tension of the CEM I and CEMV after 28 days are 58.5 and 54 MPa, respectively [19,20] while the one of the CEM III/C is only 40 MPa. [21]oreover a geometric defect in the cement-matrix could also be responsible for the cracking, which could be confirmed through additional experiments with other specimen geometries.
The EIS experiments conducted in this work were only analysed from a qualitative standpoint but clearly demonstrated that the corrosion mechanism involved in each cement matrix is different.Further EIS investigations are still needed to analyse the data and to propose a detailed corrosion mechanism associated with each cement.It must be the focus of a dedicated research.
The Raman analysis conducted on the corrosion scale forming in the aluminium grade 5754 exposed to the three different cements identified the same types of corrosion products.The main difference between the three considered cements considered is the resaturation duration required to form the corrosion products.For CEM I and CEM V the Bayerite and the Gibbsite, which are the main corrosion products observed, were detected after 100 h of resaturation while for the CEM III, 3 or 4 months were needed to obtain the same corrosion product layers.
These results tend to support the hypothesis of a limited influence of chemistry compared with physical properties of the cement matrix.The decrease of the corrosion rate seems to be the consequence of the limited access of the water to the aluminium surface (the presence of a corrosion product layer) and not strictly due to a passivity of the metal.In all three cases, the metal is exposed to the same kind of alkaline environment and it is therefore the local saturation of the solution with cations at the aluminium-cement interface that determines the rate of corrosion.Indeed, in addition to water consumption and a limited water resaturation, the local accumulation of aluminium in solution close to the interface could modify the corrosion kinetics through an enlargement of the potential range in which the material remains passive.When cracking occurs in the cement-matrix, a sudden increase of the corrosion rate is observed.This is related to a direct contact between the resaturation water (coming through the cracks) and aluminium, and a loss of the confinement afforded by the cement.
The CEM I cement with a large capillary porosity, is the one providing the lower barrier effect to water ingress among all the tested cements.Blended cements such as CEM III/C and CEM V present much more refined microstructures and therefore much more restricted mass transfer properties (gas and water permeation).Thus it could possibly cause a delay in the water resaturation and, as a consequence, delayed the initiation of the corrosion.
From these observations, the aluminium corrosion seems to be mainly controlled by the rate of water ingress to the aluminium-cement interface.

| CONCLUSIONS
This work aimed at analysing the corrosion behaviour of an aluminium grade 5754 encapsulated in different cement matrices.It complements work already published [5] and focuses on the processes occurring at the cement/aluminium interface and the physico-chemical evolutions of the environment at this location.Three cement types were studied: CEM I, CEM III/C and CEM V.The following conclusions can be reached: • If the integrity of the cement is ensured, the best protectivities are obtained in the following order: CEM III/C > CEMV > CEMI.This trend was observed through electrochemical monitoring and weight loss measurements and is exactly the one that was anticipated from a previous study. [5] After a sufficiently high incubation time, cracking of the CEMI and CEMIII/C was observed.From previous work, this failure was anticipated for the former but not for the latter.It is possible that the failure of CEMIII/C cement is due to the change of the specimen geometry compared to the first study. [5] EIS analyses highlighted different corrosion mechanisms depending on the cement considered.• In the previous study, [5] a hypothesis explaining the different behaviour of the cements was based on the nature of the corrosion products forming at the cement/aluminium interface.The advanced characterisations conducted in this work through Raman spectroscopy did not allow support for this hypothesis.On the contrary, similar corrosion products for all cements were identified • Corrosion of the aluminium in the cement-matrix seems to be limited by the water access at the metal interface.The corrosion rate increased initially during the cure due to the presence of water after casting.With the aluminium corrosion and the water consumption, the water availability decrease led to a decrease of the corrosion rate until the cracking of the cement matrix occurred.When cracking occurs, a high corrosion rate is observed due to easy water ingress at the aluminium-cement interface.

F
I G U R E 2 pH evolutions in the different cement matrices (CEM I, CEM III/C et CEM V) in resaturation conditions.
I G U R E 4 Cement matrix after the resaturation tests (a: CEM I, b: CEM III/C, c: CEM V). [Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 5 Open circuit potential at the interface aluminiumcement in the three cement matrices (CEM I, CEM III et CEM V). aluminium sample immersed in a cement matrix based on CEM I.

F I G U R E 6
Redox potentials (Pt electrode) in the different cement matrix.F I G U R E 7 Polarisation resistances in the different cements.

T A B L E 2 F I G U R E 8
Corrosion rates calculated from the weight loss measurements.Evolution of the Nyquist impedance spectra obtained in CEM I matrix over time.

F I G U R E 9
Evolution of the Nyquist impedance spectra obtained in CEM III matrix over time.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 10 Evolution of the Nyquist impedance spectra obtained in CEM V matrix over time.[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 3 Approximate values of transfer resistance measured with EIS in the different cements, based on the diameter of the loop ascribed to the Faradaic process.

F
I G U R E 11 Raman spectra of aluminium coupons between 3000 and 4000 cm −1 removed from CEM I after 100 h (B: Bayerite, G: Gibbsite).[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 12 Raman spectra between 3000 and 4000 cm −1 for CEM I after 3000 h.[Color figure can be viewed at wileyonlinelibrary.com]In addition to Bayerite and Gibbsite, signals from a hydrated species were observed at 3420, 3630 and 3687 cm −1 .It was difficult to firmly attribute these signals as they could correspond to the first stage of aluminium hydration or could originate from the hydration of the cementitious matrix.Additionally, no other signals at lower Raman shift (between 100 and 2000 cm −1 ) which might help to get a correct assignment were seen.

F
I G U R E 15 Raman spectra between 3000 and 4000 cm −1 for CEMV after 3000 h (1: Gibbsite and Nordstrandite or Gibbsite, 2: Bayerite and Nordstrandite or Bayerite).[Color figure can be viewed at wileyonlinelibrary.com] Chemical composition of the solution used for resaturation.
T A B L E 1