When a steel electrode is corroding in an aerated electrolyte, the potential, called open-circuit potential (OCP), reaches a value somewhere between the potential of Fe^II/Fe equilibrium and that of O₂/H₂O equilibrium. The corrosion process can be accelerated electrochemically by increasing either the potential of the electrode, i.e. the potentiostatic procedure, or the current flowing through the electrode, i.e. the galvanostatic procedure. Such experiments can be devised in the laboratory for monitoring the early stages of the corrosion process. In this article, experiments performed at OCP using E24 steel (98.2% Fe, 0.122% C, 0.206% Si, 0.641% Mn, 0.016% P, 0.131% S, 0.118% Cr, 0.02% Mo, 0.105% Ni and 0.451% Cu) are described. The steel surfaces were polished with silicon carbide (particle size 25 μm), rinsed thoroughly with Milli-Q water and carefully dried. The potential was measured using an Ametek (Princeton Applied Research) 263/A potentiostat system. A saturated calomel electrode (SCE) was used as a reference, but the potential is expressed with respect to the standard hydrogen electrode (SHE) in order to facilitate the comparison with potential–pH equilibrium Pourbaix diagrams.

The kinetics of degradation of steel tends to decrease as the metal is covered by a rust layer and the behaviour of a steel structure dipped in water for years is governed by the properties of this several-millimetre-thick layer. The understanding of the mechanisms at such late stages of the corrosion process requires a detailed analysis of the morphology and composition of the rust layers. The case of a steel structure left for 25 years in the Atlantic Ocean is detailed. The identification of transient unstable phases such as GRs is only made possible if the samples are sheltered immediately from oxygen. Therefore, the corrosion layers, about 10–15-mm thick, scrapped from the metal in the permanently immersed zone, 1 m above the mud line, were immediately placed in acetone, preventing any chemical evolution of the samples for 3–4 months. They were coated with epoxy resin (Struers Epofix^®), sawed up into 2-mm-thick slices, and again coated with resin. Cuts were achieved so that slices were perpendicular to the steel/rust layer interface. More details concerning this procedure can be found elsewhere [25].

Additional information about the mechanisms of formation and transformation of rust can be obtained via the study of the oxidation of a Fe(II) precipitate in aqueous suspension [10–13,15,33,38,40]. The oxidation processes involved in chloro-sulfated media were then studied using the following methodology. FeCl₂⋅4 H₂O and FeSO₄⋅7 H₂O were dissolved in a 100-ml flask of milliQ water. The $\{[{\mathrm{Cl}}^{-}]/[\mathrm{SO}_4{}^{2-}]\}$ ratio was set at 1/12. NaOH was dissolved in another 100-ml flask and two $\{[{\mathrm{Fe}}^{2+}]/[{\mathrm{OH}}^{-}]\}$ ratios were considered, 1 and 0.58. The NaOH concentration was set at 0.4 M, and the temperature at 25 °C. The solutions were mixed, leading to the precipitation of a Fe(II) compound. Magnetic stirring (∼500 rpm) in the open air ensured a progressive homogeneous oxidation of the precipitate and a thermostat controlled the temperature, which was kept at $25\pm 0.5\text{°C}$. Reactions were monitored by recording the pH, measured via a glass electrode, and the potential E of a platinum electrode immersed in solution, using the saturated calomel electrode as a reference (but all potentials in the following refer to the standard hydrogen electrode, SHE).

Different characterisation methods must be combined to identify unambiguously the various components of a rust layer. X-ray diffraction (XRD) can be used along with Raman and/or Mössbauer spectroscopy. Here, analyses performed by XRD, conversion electron Mössbauer spectroscopy (CEMS) and micro-Raman spectroscopy are presented.

3.1 CEMS

3.2 XRD

3.3 Raman spectroscopy

E–pH Pourbaix diagrams are maps that summarise the thermodynamic information describing the possible routes followed during the corrosion of steels in carbonated aqueous media (Fig. 1). Values of the standard Gibbs energy of formation $\mathrm{\Delta }{G}_{\mathrm{f}}^0$ that are retained for the various species are listed in Table 1 and the equilibrium equations are reported in Table 2. For Fe species, the $\mathrm{\Delta }{G}_{\mathrm{f}}^0$ values were taken from [5,20], or re-computed from solubility products using the value $\mathrm{\Delta }{G}_{\mathrm{f}}^0(\mathrm{Fe}^{2+}{}_{\mathrm{aq}})$ of Table 1. This allowed us to have a complete set of consistent values and equations. The $\mathrm{\Delta }{G}_{\mathrm{f}}^0$ value of FeCO₃ was computed from the solubility product ($-\log K_{\mathrm{so}}$) value of −10.80 given by [8], but values ranging from −10.43 to −11.20 are reported [19]. The other values were taken from Wagman et al. [48]. The diagrams were drawn using an activity of carbonate species equal to 0.1. A first diagram was drawn considering $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$, but ignoring FeCO₃. It is represented in solid lines. The second diagram was drawn considering both phases. Equilibrium reactions involving FeCO₃ are represented in dotted lines. The domain of stability of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ is totally included inside that of FeCO₃, which is delimited by lines (15), (16), (17) and (18). This demonstrates that the hydroxycarbonate $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ is metastable with respect to FeCO₃.

However, it was reported that homogeneous $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ layers could form on iron disks dipped in 0.1 M NaHCO₃ solutions [1]. This experiment was performed again, using a 30-mm diameter E24 steel disk. The OCP of the disk was measured during the 24 h of immersion. It decreased from about $+50{\text{mV}}_{\mathrm{SHE}}$ down to about $-450{\text{mV}}_{\mathrm{SHE}}$, while the initial Fe₂O₃ film formed in air upon the steel surface dissolved. It stabilised then at $-450{\text{mV}}_{\mathrm{SHE}}$ as the corrosion of iron proceeded. The green rust layer obtained after 24 h was analysed by CEMS at room temperature in the inert atmosphere of the gas flow proportional counter. The spectrum (Fig. 2) is composed of three spectral components, two doublets due to a paramagnetic compound and one sextet due to a magnetically ordered compound. The hyperfine parameters of this sextet prove to be typical of α-Fe at room temperature. This is the signal coming from the substrate, detected because the rust layer is porous and does not cover completely the metal. The hyperfine parameters of the doublets $D_1$ ($\delta =1.10\text{mm}{\text{s}}^{\mathrm{-1}}$ and $\mathrm{\Delta }=2.22\text{mm}{\text{s}}^{\mathrm{-1}}$) and $D_3$ ($\delta =0.55\text{mm}{\text{s}}^{\mathrm{-1}}$ and $\mathrm{\Delta }=0.42\text{mm}{\text{s}}^{\mathrm{-1}}$) are characteristic of Fe^II and Fe^III atoms in a GR compound [38]. Since only carbonate species were present in the solution, this GR can only be the hydroxycarbonate of chemical formula $\mathrm{Fe}^{\mathrm{II}}{}_4\mathrm{Fe}^{\mathrm{III}}{}_2{(\mathrm{OH})}_{12}{\mathrm{CO}}_3\cdot 3{\mathrm{H}}_2\mathrm{O}$. The $D_1/D_3$ area ratio, which is practically identical to the Fe^II/Fe^III ratio within the GR, is measured at 1.8 (≡35.7% Fe^III), very close indeed to the awaited value of 2 (≡33.3% Fe^III). Moreover, a slight discrepancy due to the small emission percent comes from ignoring a second ferrous doublet $D_2$, the intensity of which should be (1/3) that of $D_1$. Consequently, the value that is computed must lie between 1.5 and 2. The measured value of the OCP is thus reported in the Pourbaix diagram (Fig. 1). The pH of the 0.1 M NaHCO₃ solution is buffered at about 8.3 by the hydrogenocarbonate ions and the corresponding point is represented by a ringed cross, which is located at the centre of the domain of stability of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$.

When NaCl is added to the 0.1 M NaHCO₃ solution, the green rust layer resulting of the immersion at OCP of iron disks is still composed of the hydroxycarbonate, even for NaCl concentration of 4 M [1]. This illustrates the well-known stability of the hydroxycarbonate structure with respect to any other form of GRs and in particular those obtained with monovalent anions [26,27,37]. But the results are different if the corrosion process is accelerated electrochemically. Galvanostatic experiments were performed on iron electrodes dipped in a 0.6 M NaHCO₃ + 0.5 M NaCl solution of pH about 8.3 [16]. The corrosion product was a mixture of 35% $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ with 65% FeCO₃. In this case, siderite FeCO₃ forms, whereas it is not obtained at the OCP. Similarly, the dissolution of E24 steel electrodes anodically polarised in a 0.1 M NaHCO₃ + 0.02 M NaCl solution mainly leads to FeCO₃ [42]. The difference between the behaviour at OCP and the behaviour under anodic polarisation is likely a consequence of the influence of $\{[\mathrm{Fe}^{2+}{}_{\mathrm{aq}}]/[{\mathrm{OH}}^{-}]\}$ and $\{[\mathrm{HCO}_3{}^{-}+\mathrm{CO}_3{}^{2-}]/[{\mathrm{OH}}^{-}]\}$ concentration ratios. If they are small, Fe(OH)₂ can precipitate and is rapidly and totally transformed into $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$. An increase of these ratios should favour FeCO₃ at the expense of Fe(OH)₂. At the OCP, the reduction of O₂ produces two OH⁻ ions when the dissolution of iron produces one Fe²⁺. This should favour $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$. When an anodic polarisation is applied, the reduction of O₂ does not match the production of Fe²⁺ ions, thus favouring FeCO₃.

Ten- to fifteen-millimetre-thick rust layers formed on steel sheet piles during 25 years proved to be composed of three main layers [40]. The inner one, close to the metal substrate, is essentially made of magnetite. The intermediate one is composed of iron(III) oxyhydroxides. Finally, in the external one, the closest to the interface with seawater, the sulphated form of GRs proved to be the main constituent (Fig. 3). A Raman spectrum of this $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ is displayed (Fig. 4). In agreement with the previous Raman studies of GRs [6,7,21,45], it is composed of two intense peaks at 430 and $508{\text{cm}}^{\mathrm{-1}}$, that were attributed to Fe^II–OH and Fe^III–OH stretching, respectively [6,7]. The weaker band at $260{\text{cm}}^{\mathrm{-1}}$ was more rarely mentioned [21]. Chemical analyses demonstrated that the GR contained elements Fe, O and S, indicating that it was the hydroxysulphate [40]. However, the ${\mathrm{Cl}}^{-}/\mathrm{SO}_4{}^{2-}$ molar ratio in seawater is about 19. This illustrates once more the affinity of the GR structure for divalent anions. Note that in seawater the main carbonate species is $\mathrm{HCO}_3{}^{-}$ but $\mathrm{SO}_4{}^{2-}$ is overwhelming, since $\{[\mathrm{SO}_4{}^{2-}]/[\mathrm{HCO}_3{}^{-}]\}$ molar ratio is about 12. This explains why $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ forms even though $\mathrm{CO}_3{}^{2-}$ produces a better stability of the layered structure of GRs [26,27,37].

Additional experiments were performed in the laboratory in order to study the competition between GR(Cl⁻) and $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$. The oxidation of a precipitate obtained by mixing solutions of FeCl₂⋅4 H₂O, FeSO₄⋅7 H₂O and NaOH was studied. The ${\mathrm{Cl}}^{-}/\mathrm{SO}_4{}^{2-}$ molar ratio was set at 12, whereas two {[Fe²⁺]/[OH⁻]} ratios, 0.58 and 1, were considered. The redox potential E and pH versus time curves are displayed in Fig. 5. Those obtained for {[Fe²⁺]/[OH⁻]}=0.58 are typical of a two-stage reaction involving the formation of a GR compound as an intermediate product. The first stage elapses from $t=0$ to $t_{\mathrm{g}}$ and corresponds to the formation of a GR from the initial Fe^II compound, the second one elapses from $t_{\mathrm{g}}$ to $t_{\mathrm{f}}$ and corresponds to the oxidation of the GR into FeOOH phases releasing Fe^II and anions into solution. Large E and pH variations of around $t_{\mathrm{g}}$ and $t_{\mathrm{f}}$ testify of the disappearance of an initial phase [38,39]. In contrast, the curves obtained for {[Fe²⁺]/[OH⁻]}=1 indicate that the reaction involves a supplementary stage, a first one ending at $t_{\mathrm{g}1}$, a second one at $t_{\mathrm{g}2}$ and the third one at $t_{\mathrm{f}}$.

Intermediate compounds were analysed by XRD. Patterns are displayed in Fig. 6. The GR found at $t_{\mathrm{g}}$ for {[Fe²⁺]/[OH⁻]}=0.58 is identified as $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ (Fig. 6a). The three main diffraction lines at $2\theta =9.25°$ (1.1 nm), 18.69° (0.55 nm) and 28.22° (0.367 nm) are characteristic of a hydroxysulphate XRD pattern and correspond to (001), (002) and (003) lines of the trigonal structure [46]. Some lines of lepidocrocite γ-FeOOH, end-product of the oxidation process and faint lines of $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ of other indices than $(00\ell )$ are also detected. The abnormal intensity of the $(00\ell )$ lines is due to preferential orientation of $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ micro-crystallites. Thus, in this case, where $\{[{\mathrm{Cl}}^{-}]/[\mathrm{SO}_4{}^{2-}]\}=12$ and {[Fe²⁺]/[OH⁻]}=0.58, $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ forms instead of GR(Cl⁻) even though $\mathrm{SO}_4{}^{2-}$ is in minority, a consequence of the affinity of GRs for divalent anions. The XRD pattern of the product obtained at $t_{\mathrm{g}2}$ for {[Fe²⁺]/[OH⁻]}=1 is almost identical to the previous one (Fig. 6b). $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ is also obtained here. However, the XRD pattern of the product obtained at $t_{\mathrm{g}1}$ is mainly composed of the diffraction lines of GR(Cl⁻) (Fig. 6c). They are denoted R₁. Two of them, at $2\theta =12.92°$ (0.795 nm) and 25.98° (0.398 nm), are extremely intense and correspond to (003) and (006) lines of the conventional hexagonal cell of the rhombohedral structure [38]. Therefore, the first reaction stage corresponds to the formation of GR(Cl⁻) from the initial precipitate, the second stage to the oxidation of GR(Cl⁻) into $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ and the last stage to the oxidation of $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ into γ-FeOOH. The diffraction lines of $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$, denoted R₂, are seen, together with those of GR(Cl⁻), as a result of the oxidation of GR(Cl⁻). Similarly, the two main diffraction lines of γ-FeOOH, denoted L, can be noticed.

$\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ proved to have a well defined composition of $\mathrm{Fe}^{\mathrm{II}}{}_4\mathrm{Fe}^{\mathrm{III}}{}_2{(\mathrm{OH})}_{12}{\mathrm{SO}}_4\cdot 8{\mathrm{H}}_2\mathrm{O}$ [15,39,46]. The average oxidation number of iron is then +2.33. In contrast, it was observed that the composition of GR(Cl⁻) varied continuously [38] as the oxidation advanced. Starting from a compound with formula $\mathrm{Fe}^{\mathrm{II}}{}_3{\mathrm{Fe}}^{\mathrm{III}}{(\mathrm{OH})}_8\mathrm{Cl}\cdot n{\mathrm{H}}_2\mathrm{O}$, chloride and Fe^III contents increased, up to an approximate composition of $\mathrm{Fe}^{\mathrm{II}}{}_{2.2}{\mathrm{Fe}}^{\mathrm{III}}{(\mathrm{OH})}_{6.4}\mathrm{Cl}\cdot n^{\prime }{\mathrm{H}}_2\mathrm{O}$. The average oxidation number of iron varied then from +2.25 to +2.31. We may propose that in the presence of $\mathrm{SO}_4{}^{2-}$ anions, the enrichment in Cl⁻ and Fe^III is replaced by a transformation of $\mathrm{Fe}^{\mathrm{II}}{}_3{\mathrm{Fe}}^{\mathrm{III}}{(\mathrm{OH})}_8\mathrm{Cl}\cdot n{\mathrm{H}}_2\mathrm{O}$ into $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ and obviously sulphate anions deliver a phase that is more stable than that obtained with chloride anions. The differences observed between the oxidation processes at {[Fe²⁺]/[OH⁻]}=1 and 0.58 may result from the nature of the initial precipitate. When {[Fe²⁺]/[OH⁻]} is small, close to 0.5, the value corresponding to the stoichiometric conditions of formation of Fe(OH)₂, the initial precipitate is indeed Fe(OH)₂ or close to Fe(OH)₂ [35]. When {[Fe²⁺]/[OH⁻]} is larger, Fe(II)-hydroxychlorides may form [35]. At {[FeCl₂]/[NaOH]}=2.5, the initial precipitate was identified as β-Fe₂(OH)₃Cl [36]. This compound gets oxidised the way Fe(OH)₂ does into $\mathrm{Fe}^{\mathrm{II}}{}_3{\mathrm{Fe}}^{\mathrm{III}}{(\mathrm{OH})}_8\mathrm{Cl}\cdot n{\mathrm{H}}_2\mathrm{O}$, and we may suppose that, even in the presence of sulphate ions, the oxidation of such an hydroxychloride leads to GR(Cl⁻). The oxidation of β-Fe₂(OH)₃Cl into $\mathrm{Fe}^{\mathrm{II}}{}_3{\mathrm{Fe}}^{\mathrm{III}}{(\mathrm{OH})}_8\mathrm{Cl}\cdot n{\mathrm{H}}_2\mathrm{O}$ could well be a solid-state reaction.

We demonstrated previously that in marine environments the dissolution of iron in seawater should lead to $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$. Some severe cases of corrosion are due to the concomitant presence of sulphate-reducing bacteria (SRB) and of $\mathrm{GR}(\mathrm{SO}_4{}^2)$ among the corrosion products [14,30]. The recent paper that described carefully the sequence of rust stratification on steel sheet piles at the anoxic level of sea mud suggests that it is due to bacterial reduction of previously formed lepidocrocite where large hexagonal $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ crystals are clearly identified (Fig. 3) [40]. We thus propose that $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ forms firstly due to the reduction of γ-FeOOH in the outer layer by ubiquitous dissimilatory iron-reducing bacteria (DIRB). Then, the zone enriched in $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ is favourable to the colonisation of the interface by other micro-organisms, the sulphate-reducing bacteria (SRB). SRBs then induce formation of H₂S, which is an acidification of the environment, which results in a drastic increase of iron degradation. Microbially influenced corrosion of steels would then be a two-step process involving DIRB that form $\mathrm{GR}(\mathrm{SO}_4{}^{2-})$ followed by the reduction of this sulphate reservoir by SRBs.

Other phases can also be obtained from GRs in more specific conditions. Akaganeite forms from GR(Cl⁻) in solutions containing a large excess of $\mathrm{Fe}^{2+}{}_{\mathrm{aq}}$ and Cl⁻ [36]. Ferrihydrite was also reported to form from GR(Cl⁻), when the Fe concentration is very low and the oxidation kinetics fast [41]. Phosphate species modify considerably the oxidation process of aqueous suspensions of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$. For instance, in a first article, the final product was unambiguously identified as ferrihydrite [3], but more recently it was concluded that this product was a ‘ferric green rust’, that is a Fe^III compound characterised by a layered structure similar to that of GRs [23]. The existence of such ‘ferric green rusts’ demonstrates clearly that Fe^II cations can be oxidised in situ and that the structure of GRs can sustain up to 100% Fe^III [17,34,41]. In order to compensate for the increase in positive charge, the oxidation of Fe^II is accompanied by a deprotonation of the OH⁻ ions of the brucite-like layers. Ferric GRs can be obtained by oxidation of GRs with hydrogen peroxide [4,34,41], electrochemical polarisation of iron [21] or aerial oxidation of dried GR layers [1]. The formation of the ferric $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ was reinvestigated here in the case of steel corrosion.

First, a $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ layer was grown on a steel electrode left at OCP in a 0.1 M NaHCO₃ solution, as described here in part 3. The sample was removed from the solution and left in the dry atmosphere of the laboratory. The deep-green corrosion layer turned progressively to brown and was analysed one month later by XRD. Secondly, a 10-ml solution containing approximately 30% H₂O₂ was added to an aqueous suspension of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ prepared with the procedure described previously [3]. The suspension turned immediately to brown and the precipitate was filtered, dried to powder and analysed by XRD. Both XRD patterns are presented in Fig. 7.

The two most intense lines visible in the pattern of the dried rust layer are those of the substrate α-Fe. Lines of goethite α-FeOOH are also observed, together with three other lines that can be attributed to the ferric GR. The main one corresponds to an interplanar distance of 0.735 nm ($2\theta =14°$), that is the distance $d_{003}$ between interlayers. The occurrence of this peak in the pattern of the ‘ferric GR’ definitively proves that the sequence among hydroxide and anionic layers is conserved in the structure. The decrease of $d_{003}$ from 0.750 nm in $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ [13] to 0.735 nm in ‘ferric GR’ is attributed to the decrease of the ionic diameter of iron species, from 0.156 nm for Fe^II to 0.129 nm for Fe^III [44]. Two other peaks are visible, at 0.3675 nm ($2\theta =28.2°$) and 0.254 nm ($2\theta =41.3°$).

The pattern obtained by oxidation of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ with hydrogen peroxide is composed of five main lines that can all be attributed also to ‘ferric $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$’. Assuming that the ferric GR keeps the pyroaurite structure of $\mathrm{GR}(\mathrm{CO}_3{}^{2-})$ [2], it is computed that the observed diffraction lines are indeed associated with intense lines of the GR structure. The parameters of the conventional hexagonal cell are determined at $a=0.301\text{nm}$ and $c=2.205\text{nm}$. The corresponding Miller index (hkl) was then used to identify the ‘ferric GR’ diffraction lines on the patterns of Fig. 6. However, it is not possible to exclude that ferrihydrite forms together with the ferric GRs if the conditions of corrosion are inhomogeneous as it is surely in actual cases. As a matter of fact, the most poorly ordered form of ferrihydrite, the so-called two-line ferrihydrite, is characterised by an XRD pattern made of two broad lines at $2\theta \sim 42°$ and 74°. Intense lines of the ferric GR are also located in these angular regions and may mask those of ferrihydrite.

The determining role of GRs in the corrosion processes of ferrous alloys in neutral and alkaline media is now well settled. Fe^II–III hydroxysalts constitute a major step before the formation of the various components of rust and may control partially the properties of the rust layers formed on steel. Therefore, the positive role of various corrosion inhibitors commonly used to prevent degradation of steels in neutral or alkaline media are partially explained by the influence those species may have upon the formation and/or transformation of GRs. Phosphates, nitrites and nitrates are examples of such inhibitors.