3.1. Electrochemical Impedance Spectroscopy
EIS measurements were performed on both coupons at various exposure times in soil at OCP or under CP. In the present case dealing with a resistive medium, a parasitic influence of the reference electrode at high frequency was observed [
32], and
Rs was then determined at 10 kHz, i.e.,
Rs = Re(
Z) at 10 kHz. For the same reason the EIS data obtained at frequencies higher than 10 kHz were not considered and are not shown on the Nyquist plots except when noted.
First,
Figure 2 displays the evolution of
Rs with time for coupons WE
CP1 and WE
CP2 in soil during the whole experiment. During the OCP period (days 1–7),
Rs increases from 1670 to 1865 Ω cm
2 for coupon WE
CP1 and from 1380 to 1815 Ω cm
2 for coupon WE
CP2. This increase in
Rs can be attributed to the changes in soil physicochemical properties. Actually, water tends to flow vertically in the cell, because of gravity, so that a gradient of soil moisture appears with time. This movement of fluid necessarily modifies the soil at the vicinity of the coupons during the first days. Moreover, corrosion products are formed on the metal surface, which also modifies the steel/soil interface.
The application of CP leads to an immediate sharp increase in Rs to ~2250 Ω cm2 for coupon WECP1 and ~3850 Ω cm2 for coupon WECP2. Various changes occur at the steel/electrolyte interface when CP is applied: dissolved O2 is consumed and interfacial pH increases, the electric field induces migration of ions, electrocapillary effects modify the contact angle between liquid phase and metal, etc. It is worth noticing that there is a difference of 1600 Ω cm2 in Rs values obtained between coupons WECP1 and WECP2 after 8 days. This difference can then be attributed to the different applied ECP values set for coupons WECP1 and WECP2, i.e., −1.0 and −1.2 V vs. Cu/CuSO4, respectively. This indicates that lower ECP leads to higher Rs. This initial increase of Rs is, however, followed by a rapid decrease so that at day 14 both Rs values are similar to those measured at OCP. This initial increase of Rs is then a transient phenomenon that was not further studied and is yet to be explained.
However, clear trends are observed after day 21. For coupon WE
CP1,
Rs decreases further before reaching a constant value of 890 ± 50 Ω cm
2 after day 42, indicating that a steady state has been reached. This value is actually ~800 Ω cm
2 lower than the initial value of 1670 Ω cm
2 (day 3). This result is consistent with previous study [
25], where the decrease of
Rs was attributed to an increase of the “wet area” due to electrocapillary effects. A negative shift of potential allows the liquid phase to spread over the steel surface because it decreases the solid–liquid contact angle
θ [
33,
34,
35,
36].
In contrast,
Rs values obtained for coupon WE
CP2 after 21 days take an opposite trend in relation to that observed for coupon WE
CP1, i.e., increase again with exposure time. This increase is not continuous, and some fluctuations are observed, with local minima at days 42 and 70.
Rs reaches a final value of ~2400 Ω cm
2 after 90 days, thus being ~1000 Ω cm
2 higher than the initial value of 1380 Ω cm
2 (day 3). At the CP potential of −1.2 V vs. Cu/CuSO
4, the cathodic process is enhanced, leading to the acceleration of O
2 consumption at the steel/electrolyte interface and more likely to the increase of water reduction rate. The increase of
Rs with time may then be due to the accumulation of hydrogen (H
2) bubbles that form on the steel surface according to Equation (2). This phenomenon would indeed lead to a decrease of the “wet” area (i.e., the area in contact with the liquid phase) and thus to an apparent increase of
Rs [
10,
25]. The cathodic processes taking place at −1.2 V vs. Cu/CuSO
4 are studied and discussed in
Section 3.2.
The Nyquist EIS plots obtained for both coupons before the application of CP are compared in
Figure 3. Both Nyquist diagrams display similar features, though they are more clearly seen for coupon WE
CP2. A first capacitive loop is present at high frequency, followed by a linear part and a second capacitive loop at low frequency. The linear part makes angles of 45° and 50° with Re(
Z) for coupons WE
CP1 and WE
CP2, respectively. This shows the influence of a diffusional process. Previous studies reported that the cathodic reaction of the steel buried in unsaturated soil was indeed partially controlled by the diffusion of O
2 [
10,
24]. Actually, the obtained Nyquist diagrams are characteristic of a bounded diffusion phenomenon, i.e., a finite-length diffusion for a planar electrode. At the beginning of the corrosion process, the corrosion product layer is very thin and should not hinder the diffusion of O
2. The O
2 concentration gradient is then located in the soil itself, as observed in previous work [
37].
The EIS data at OCP were then modelled using the electrical equivalent circuit (EEC) displayed in
Figure 4a. In this EEC,
Rs is the soil electrolyte resistance,
Rt is the charge transfer resistance,
Qdl a constant phase element (CPE) used to represent the double layer capacitance and
Wd is a bounded diffusion impedance. The CPE was used instead of an ideal capacitance because the first attempts to model the experimental data, performed with a capacitance, did not lead to acceptable goodness-of-fit. Actually, CPE is commonly used because it takes into account various effects due to inhomogeneity, porosity, roughness and other non-ideal dielectric properties of the electrode [
38]. The results obtained via this modelling are discussed together with those obtained for the coupons under CP at the end of this section.
The influence of CP can be observed on the Nyquist plots of
Figure 5 and
Figure 6. As explained in
Section 2.3.2, only the data obtained after day 42 are discussed. These Nyquist diagrams consistently exhibit a flattened semi-circle loop, followed by what may be an incomplete second loop at lower frequency. The main flattened semi-circle loop is characterized, like that observed at OCP, by a linear behavior at high frequency. However, the initial linear part of the Nyquist diagrams obtained under CP after day 42 makes an angle of 30° and 25° with Re(
Z) for coupons WE
CP1 and WE
CP2, respectively, as illustrated on day 42 Nyquist plot in
Figure 5 and day 63 Nyquist plot in
Figure 6. This suggests that the steel coupons behave as semi-infinite porous conductive electrodes [
37,
39,
40]. The cathodic process, predominant under CP, would then involve the diffusion of O
2 inside the pores of a conductive mineral film covering the steel surface and its reduction all along the conductive walls of the pores. At OCP, the initial linear part of the Nyquist diagram made an angle of about 45° with the Re(Z) axis (
Figure 3). The impedance, mainly corresponding to a bounded diffusion impedance, then described the diffusion rate of oxygen through a non-conducting porous layer, which in the present case was more likely the soil itself. At day 42, i.e., after 35 days under CP, the impedance describes the diffusion (and reduction) of O
2 in the pores of a conductive layer that formed later under CP.
Furthermore, the influence of the applied potential is also clearly revealed. In the case of the CP potential of −1.0 V vs. Cu/CuSO
4, i.e., coupon WE
CP1 as shown in
Figure 5, the flattened semi-capacitive loop at high frequency of the Nyquist plots is observed to increase in magnitude with exposure time. In the present case, the electrochemical process is mainly controlled by the diffusion of O
2 inside the pores of the mineral film covering the steel surface. Consequently, the gradual increase of the resistance associated with this electrochemical process indicates that the mineral film hinders more and more efficiently the diffusion of O
2.
In the case of excessive CP, i.e., coupon WE
CP2 as shown in
Figure 6, the evolution of the Nyquist plots with exposure time is significantly different. The main difference relates to the evolution of
Rs, whose value increases with time in this case (see also
Figure 2). The diameter of the main flattened loop increases with time between day 42 and day 63, as observed for coupon WE
CP1, but does not seem to increase afterwards.
To obtain additional information and quantify the evolution of the steel/soil interface with time, a modelling of the EIS data obtained for the coupons under CP was achieved. The EEC used for both coupons is displayed in
Figure 4b. In this EEC,
Q1 and
R1 are a CPE and a resistance, respectively, used to model the phenomenon associated with the main flattened capacitive loop present in all cases (
Figure 5 and
Figure 6).
Q2 and
R2 are the elements used to fit the low frequency part of the impedance diagram. Because only a very small part of the corresponding capacitive loop (if one assumes this is really a capacitive loop) is seen, various values could be obtained for
Q2 and
R2 parameters and consequently these parameters are not presented nor discussed. Various fits were performed, with different values for
R2 and
Q2, to study the influence of these parameters on the values obtained for
R1 and
Q1. It was noted that
R1 and
Q1 were not significantly influenced by
R2 and
Q2. The considered EEC led to satisfactory fittings of the experimental data, as illustrated by
Figure 7.
Figure 7a shows as an example the Nyquist plot for coupon WE
CP1 at OCP (day 7) and
Table 2 gathers the results obtained for both coupons at OCP. It can be seen that the numerical values are similar for both coupons, except for resistance
Rd. In the expression of the bounded diffusion impedance
Wd,
Rd is a scaling factor that depends on the kinetics of the interfacial reaction and the bulk concentration of the electroactive species. The larger value of
Rd for coupon WE
CP2 suggests a lower O
2 concentration. This would indicate a faster O
2 consumption at the vicinity of coupon WE
CP2. This faster evolution may also correlate with the faster evolution of
Rs observed for this coupon between day 1 and day 7 (
Figure 2). In conclusion, both coupons behave quite similarly at OCP, as confirmed by voltammetry measurements (
Section 3.2).
Figure 7b shows as an example the Nyquist plot for coupon WE
CP1 under CP at day 70, and
Table 3 gathers the results obtained between days 42 and 90 for both coupons. First, as observed qualitatively using the Nyquist plots of
Figure 5 and
Figure 6, the resistance
R1 associated with the main capacitive loop increases continuously with time for coupon WE
CP1, to reach a maximum of 1661 Ω cm
2 at day 90. For coupon WE
CP2, after an initial increase between days 42 and 63, where a maximum of 1334 Ω cm
2 is reached,
R1 decreases slightly to end at 1230 Ω cm
2.
It is interesting to note that the
n coefficient of the CPE is low, between 0.52 and 0.60 considering both coupons. This coefficient measures the deviation from an ideal behavior, i.e.,
n = 1 for a true capacitor and
n < 1 for a CPE. However, a “real” CPE is characterized by
n values higher than 0.7, and
n = 0.5 corresponds to a diffusion element. The values of
n obtained here in each case confirm that the cathodic process, i.e., mainly oxygen reduction, is strongly influenced by diffusion. Parameter
R1 can then be considered as a “resistance to diffusion” induced by the mineral layer growing on the steel surface and inside the pores of the soil. For coupon WE
CP1, the continuous increase of this resistance shows that this layer is more and more protective as it grows with time. Simultaneously, the CPE element
Q1 also increases continuously with time. It was observed that the Nyquist plots corresponded to a porous conductive electrode behavior, more likely due to the formation of a porous magnetite layer [
20,
27]. The growth of this conductive layer would increase the overall cathodic surface, thus leading to an increase of the associated capacitance.
Moreover, the increase of R1 for coupon WECP1 is not associated with an increase of Rs, which indicates that this increased “resistance to diffusion” is not in any case associated with a decrease in the active area of the electrode. It can really be attributed to the growth of the porous conductive layer that covers the steel surface and progressively hinders more and more the transport of O2.
In the case of coupon WE
CP2, the initial increase of
R1 is associated with an increase of
Rs that may be due to a decrease of the active area [
25]. However,
Rs increases from 1711 Ω cm
2 to 2142 Ω cm
2 between day 42 and day 63, i.e., an increase of +25%, while
R1 increases from 595 Ω cm
2 to 1334 Ω cm
2, i.e., an increase of +124%. Consequently, the main reason for this initial increase of
R1 is not the possible decrease of the active area but, as for coupon WE
CP1, a real increase of the “resistance to diffusion” associated with the growth of the porous conductive layer. However,
R1 decreases slightly after day 63, while
Q1 decreases as soon as day 42 and fluctuates after day 63. This shows that the protective ability of the porous conductive layer is influenced by a phenomenon that does not take place for coupon WE
CP1. This could be due to the hydrogen evolution reaction that should be more important at the lower potential of −1.2 V vs. Cu/CuSO
4. Hydrogen evolution could damage, through the release of H
2 bubbles, the porous conductive layer. This point is discussed further at the end of
Section 3.3.
3.2. Voltammetry and Estimation of Residual Corrosion Rate
The voltammograms obtained in this study were computer fitted using theoretical kinetic laws. From the EIS experiments, discussed in the previous section, it was observed that the cathodic process, mainly linked to O
2 reduction, was at least partially controlled by diffusion. Therefore, the mathematical modelling of the voltammograms
j(
E) used in the present study, adapted from that reported previously [
10,
11,
25,
27], was described by the following expression:
where
j = overall current density,
jA = anodic current density,
jC = cathodic current density,
jcorr = corrosion current density,
jlim = limiting current density (O
2 diffusion) in A/cm
2,
Ecorr = corrosion potential in V vs. Cu/CuSO
4,
βA = anodic Tafel coefficient and
βC = cathodic Tafel coefficient for O
2 reduction in V
−1.
This model corresponds to an anodic process controlled by charge transfer and a cathodic process partially controlled by diffusion (mixed control) [
10,
24,
25,
29]. The cathodic process is then in this case restricted to O
2 reduction. An additional contribution due to water reduction was envisioned. However, this model implied a very large number of adjustable parameters and the procedure proved unreliable as various solutions could be obtained for the fitting of the experimental curves. It was consequently discarded.
Equation (4) was then used in any case to perform the mathematical modelling, and
Figure 8 shows as an example the computer fitted log|
j| vs.
E voltammogram obtained after 49 days for coupon WE
CP1. In any case, the quality of the fitting proved excellent, which demonstrates that the electrochemical model was adequate. Actually, at
ECP = −1.0 V vs. Cu/CuSO
4, it was indeed expected that the contribution of water reduction was negligible. The current density
jA(
ECP) was extrapolated from the anodic Tafel line log|
jA| vs.
E, as illustrated in
Figure 8. From the extrapolated
jA(
ECP) value, the residual corrosion rate
τrc was estimated via Faraday’s law.
All the fitted parameters for coupon WE
CP1 are listed in
Table 4. To facilitate comparison with published data, the Tafel coefficients
βA,C were converted to Tafel slopes
bA,C using the expression:
The accuracy of the obtained values is about ±10% in any case. For Ecorr, determined for coupon WECP1 directly on the graph and not through the fitting procedure, the accuracy is ±1 mV.
Firstly, the voltammetry measurements performed on day 7 just before the application of CP revealed that the corrosion rate
τcorr was 330 µm yr
−1. This value is consistent with that obtained for a similar saturation level (~70–60% sat.) in a previous work performed in the same artificial soil [
10]. After 42 days of experiment (35 days of CP), the residual corrosion rate
τrc of coupon WE
CP1 was estimated at 19 µm yr
−1. It decreased to 7 µm yr
−1 after 77 days in soil (70 days of CP) and remained constant until the end of the experiment.
Table 4 also shows that under CP,
Ecorr increased from −0.54 vs. Cu/CuSO
4 after 42 days in soil (35 days of CP) to a final value of −0.36 V vs. Cu/CuSO
4 after 90 days in soil (83 days of CP). These values are much higher than the initial
Ecorr value measured at OCP. It illustrates the major changes induced by CP at the steel/electrolyte interface. The main change is the increase in interfacial pH, which itself can induce other effects, e.g., calcareous deposition or passivation of the steel surface.
The differences observed between the Tafel coefficients measured at OCP and those obtained when the coupon is under CP also reveal the strong influence of CP on the steel/soil interface. The application of CP actually led to a significant increase of
bA from 52 mV/decade at OCP to an average value of 440 mV/decade under CP. Similarly,
bC increased from 62 mV/decade to the same average of 440 mV/decade. This phenomenon was already observed in a previous study [
27] conducted in similar soil environment and was attributed to the presence of a magnetite layer. The anodic reaction could involve the Fe(II,III) cations inside the magnetite layer and the dissolution of the oxide, thus varying
bA. Similarly,
bC could vary because O
2 reduction can take place on the magnetite layer and not necessarily on the steel surface. Finally, the limiting current density |
jlim| seems to have decreased slightly with time as the larger (absolute) values are observed at days 42–63 and the lower (absolute) values at days 70–90.
For coupon WE
CP2 polarized at a lower potential of −1.2 V vs. Cu/CuSO
4, the influence of water reduction could be significant. However, O
2 reduction seems to be predominant because the EIS data, similar to those obtained for coupon WE
CP1, revealed the importance of a diffusional process. As explained above, due to an excessive number of adjustable parameters, it proved unreliable to add another cathodic process to the modelling. Therefore, the voltammograms obtained for coupon WE
CP2 were also computer-fitted using Equation (4). Acceptable fittings were achieved, as illustrated in
Figure 9 with the voltammogram obtained after 49 days in soil.
The quality of the fittings was, however, not as good as that achieved for coupon WE
CP1. In
Figure 9, it can be seen that a slight shift of
Ecorr proved necessary to obtain this fitting.
Figure 10 shows a focus on the zone where the discrepancy between experimental and fitted curves is the highest.
Figure 10a relates to coupon WE
CP2 and shows some important discrepancies around
Ecorr and in the
EIRfree region extending from −0.60 to −0.65 V vs. Cu/CuSO
4.
Figure 10b relates to coupon WE
CP1 and shows that the modelling is in this case excellent. This shows that the electrochemical model considered in both cases is completely valid for coupon WE
CP1 but is only an approximation for coupon WE
CP2. This can be attributed to a significant contribution of water reduction for WE
CP2 polarized at −1.2 V vs. Cu/CuSO
4.
The fitted parameters obtained for coupon WE
CP2 are listed in
Table 5. Due to the lower goodness-of-fit, the accuracy is in this case about ±20%, as high as ±50% for the parameters linked to the cathodic reaction, i.e.,
bC and
jlim. As already demonstrated in other sections of the study, the application of CP after the first 7 days of the experiment induced significant changes on the steel/electrolyte interface. The evolution of
Ecorr,
bA and
bC observed for coupon WE
CP1 are also observed for coupon WE
CP2.
Ecorr is increased by the application of CP and reaches −0.369 V vs. Cu/CuSO
4 at day 90. Tafel slopes
bA and
bC are also increased by CP, to average values of 510 mV/decade and 660 mV/decade, respectively. The standard deviation on the
bA values is 60 mV/decade (±12%), which shows that the accuracy of the anodic Tafel slope is correct, i.e., not too strongly influenced by the imperfect modelling of the cathodic reaction. This is a crucial point because the estimation of the residual corrosion rate is primary linked to the anodic Tafel slope
bA. By comparison, the standard deviation for the
bC values is 285 mV/decade (±43%). The error on
jlim is also important, and the values vary around an average of −5 × 10
−4, with no apparent link with the polarization time. This can also be attributed to the influence of water reduction.
Similarly, the variations of the residual corrosion rate over time are different for WE
CP1 and WE
CP2.
Figure 11 thus presents the evolution of
τrc with exposure time in soil for both coupons. Firstly, it can be clearly observed that the application of CP led to a significant decrease in steel corrosion rate. Between days 42 and 70, the residual corrosion rate reaches values between 18 and 22 µm yr
−1 for coupon WE
CP1 and between 8 and 15 µm yr
−1 for coupon WE
CP2, while the corrosion rates were estimated at 330 µm yr
−1 and 370 µm yr
−1, respectively, at OCP on day 7. This result is consistent with what could be expected: the lower the potential, the lower the anodic component of the overall current density (as illustrated by the anodic Tafel lines of
Figure 8 and
Figure 9).
After day 70, two opposite trends are observed. For coupon WE
CP1, polarized at a correct protection potential,
τrc decreases to reach a value below 10 µm yr
−1, i.e., 7 µm yr
−1 at days 77, 84 and 90. This decrease of the residual corrosion rate to such low values was already observed and was attributed to the progressive growth of a protective mineral layer on the steel substrate [
23,
24,
25,
26]. In contrast, for coupon WE
CP2 polarized at a lower potential,
τrc slightly increases after day 63, to reach 15 µm yr
−1 at day 77 before to decrease slightly to 12 µm yr
−1 at day 90. In conclusion, the average final (days 77–90) value is 7 µm yr
−1 for coupon WE
CP1 and 13 µm yr
−1 for coupon WE
CP2. The overprotection did not lead to an increased effectiveness of CP and seems to have even decreased this effectiveness. It can be assumed that the overprotection, which increases significantly water reduction and then hydrogen evolution, can damage the protecting mineral layer that covers the steel surface. This assumption also explains why, after an initial decrease to 8 µm yr
−1,
τrc increased later on.
Two different behaviors were also observed via EIS, in particular for the evolution of
Rs over time (
Figure 2). For coupon WE
CP1, the
Rs value remained constant after day 42, which indicates that the steel/soil interface has reached a steady state. Conversely, for coupon WE
CP2,
Rs, after an initial decrease until day 21, tended to increase regularly with time. This increase of
Rs could be associated with a slight decrease of the “wet” or “active” area of the steel electrode [
10,
11,
25,
26]. This decrease of the active area could be due to the accumulation of H
2 bubbles on the surface. The release of such bubbles from time to time would then damage the protective film and lead to the observed increase of
τrc.
Similarly, the diameter of the main capacitive loop present on the Nyquist diagrams, associated with the “resistance to diffusion”
R1, was observed to increase continuously with time for coupon WE
CP1 (
Figure 5 and
Table 4). This also illustrates the increasing protective ability of the mineral layer forming on the steel surface. The increase of the “resistance to diffusion” was less important in the case of coupon WE
CP2 (
Figure 6 and
Table 4) and stopped after day 63 (it even slightly decreased). This result confirms the lesser protectiveness of the layer and is consistent with the increase of
τrc observed after day 63.
It must finally be recalled that the method used here, based on voltammetry, may significantly modify the electrode surface and thus may introduce some side effects and bias. For instance, the steel surface may be passive at
ECP, due to the increase of the interfacial pH, and depassivate during the voltammetry experiments. The extrapolation of the anodic Tafel line, which is actually obtained mainly from the potential region around
Ecorr, would then lead to an overestimation of
τrc. This depassivation phenomenon may, however, take a time longer than that required for the voltammetry experiment. The time required for that depassivation once CP is interrupted was reported to vary with the size of the electrode, from 1 h for a 5 × 5 mm electrode to 13 h for a 30 × 30 mm electrode [
19]. It may, however, depend on the nature of the soil, the composition of the electrolyte present in the soil, the moisture content, the applied potential, etc.
Voltammetry also involves a charging current due to capacitive effects, and the measured current is then the sum of the faradaic current and the charging current [
41,
42]. When the applied potential is increased linearly with time, at is the case for the polarization curves modelled to estimate the residual corrosion rate, the positive charge density stored at the metal surface increases. In this case, the charging current flows in the direction of the anodic faradaic current [
42]. Then, the measured current is necessarily higher than the anodic faradaic current. Because the charging current is proportional to the scan rate [
42], the discrepancy between both currents increases with the scan rate, and for that reason a small scan rate of 0.2 mV/s was considered. In any case, as demonstrated previously [
26], these capacitive effects also lead to an overestimation of the residual corrosion rate.
Increasing the potential from ECP to OCP decreases the cathodic reaction rate and the O2 consumption rate so that the steel/soil interface may progressively be enriched in dissolved O2. This modification also leads to an additional increase in the corrosion rate, i.e., the anodic component of the current density, which leads to an overestimation of the residual corrosion rate.
In conclusion, it must be noted that the method based on voltammetry more likely overestimates the residual corrosion rate. It may, however, give fruitful information when it is used to compare between two situations, for instance, two different applied potentials as done here. Moreover, the results given by voltammetry can be confronted to those given by EIS, and in the present case both methods led to consistent conclusions. Finally, when residual corrosion rates lower than 10 µm yr−1 are obtained using voltammetry, it can be reasonably assumed that the actual residual corrosion rate is indeed lower than this threshold of 10 µm yr−1, which is considered to define an “efficient” CP.
3.3. Characterization of the Mineral Layer Formed on the Steel Coupons
µ-Raman spectroscopy analysis was performed on the surface of the coupons and on the surface of the soil layer in contact with the steel surface. The same phases were identified in both cases, indicating that the corrosion products not only covered the steel surface but also tended to fill the pores of the soil. The analysis of coupon WE
CP1 revealed various corrosion products, and two spectra are presented as examples in
Figure 12. Spectrum (a) is very similar to that of magnetite (Fe
3O
4), which is characterized by an intense pic at 663–676 cm
−1, and two smaller peaks at 300–320 cm
−1 and 510–550 cm
−1 [
43,
44,
45]. However, in our case, the first peak is found at 365 cm
−1 and its intensity, as well as that of the peak at 525 cm
−1, is unusually strong. Actually, spectrum (a) also shares typical features with the spectrum of maghemite γ-Fe
2O
3. Magnetite and maghemite are structurally similar and may be both present. The formation of non-stoichiometric magnetite Fe
3-xO
4, an intermediate between Fe
3O
4 and γ-Fe
2O
3 [
46], is also very likely.
Raman spectrum (b) displays the main characteristic peaks of goethite (α-FeOOH) with an intense peak at 388 cm−1 and three smaller peaks at 300 cm−1, 547 cm−1 and 692 cm−1. The broad peak at 692 cm−1 is, however, unusually intense, which suggests that magnetite (main peak at 663–676 cm−1) is also present in this case.
The analysis of coupon WE
CP2 gave similar results. The obtained Raman spectra (
Figure 13) confirmed that corrosion products did form. Spectrum (a) revealed the presence of lepidocrocite γ-FeOOH, clearly identified by its two main peaks at 247 cm
−1 and 379 cm
−1 [
43]. However, broad vibration bands are also visible at 708 cm
−1, 370 cm
−1 and 525 cm
−1. They correspond to ferrihydrite, a poorly crystallized and poorly ordered hydrated Fe(III)-oxyhydroxide [
47]. The main peak of magnetite is visible at 671 cm
−1 as a shoulder of the main band of ferrihydrite. The presence of magnetite could be demonstrated more clearly via the analysis of other zones of the steel surface, as illustrated by spectrum (b).
It must be noted that CaCO3 phases could not be detected. First, it is possible that the conditions required for the formation of a calcareous deposit were not met. Secondly, µ-Raman spectroscopy is a local characterization technique, and if CaCO3 formed as a minor component, for instance, in the pores of the magnetite layer, it could have been missed. In any case, the decrease with time of the residual corrosion rate for coupon WECP1 can be attributed to the formation of a thin layer of corrosion products.
In conclusion, for both WE
CP1 and WE
CP2 coupons, the steel surface proved to be covered by a layer mainly composed of magnetite and various Fe(III) oxyhydroxides. The predominance of magnetite, already observed in previous studies [
20,
27], can be attributed to the increase of the interfacial pH [
48] induced by the cathodic reactions (see Equations (1) and (2)). Moreover, the cathodic polarization can also induce the reduction of Fe(III) to Fe(II) and thus the reduction of FeOOH compounds into magnetite.
The EIS results, which showed a behavior typical of conductive porous electrode, can then clearly be attributed to the presence of the magnetite-containing layer. Magnetite is an electronic conductor with low resistivity [
49], and consequently O
2 molecules can be reduced on the walls of the pores of this layer. The term “magnetite layer” may not actually be appropriate. Magnetite was also identified in the pores of the soil at the very vicinity of the steel surface. The “porous electrode” behavior observed via EIS might not be due, strictly speaking, to a porous magnetite layer but rather to a small layer of soil where the pore walls could be covered with interconnected magnetite particles. This would explain how the small amount of magnetite resulting from the residual corrosion process could give rise to this porous electrode behavior.
The higher protective ability of the layer formed at −1.0 V vs. Cu/CuSO
4 is not linked to its composition because the corrosion products identified for coupon WE
CP2 polarized at −1.2 V vs. Cu/CuSO
4 proved rather similar. It is therefore due to differences in terms of porosity, compactness and/or adhesion properties. The layer formed at −1.0 V vs. Cu/CuSO
4 is more protective because it has fewer defects, which can be attributed to the fact that water reduction is not active. Actually, the release of H
2 bubbles at −1.2 V vs. Cu/CuSO
4 may have finally damaged this layer, which would explain the increase of the residual corrosion rate observed after day 63 (
Figure 11). Water molecules are present in the pores of the corrosion product layer and at the surface of the steel. H
2O reduction can then take place directly at the metal surface, and H
2 molecules can then accumulate on this surface, finally forming H
2 bubbles. If trapped between the metal and the protective layer, the bubbles would induce mechanical stresses and eventually cracks in the layer. The H
2 bubbles would then be released in the soil through these cracks, and this H
2 flow would moreover remove some magnetite particles from the metal surface, move them and finally set them on the pore walls of the soil.