Condensed from Paper No. 4 presented at CORROSION/96
Reproduced with permission from NACE International.

In the CO2 corrosion of steels, the bicarbonate ion HCO3- is simultaneously the buffer for carbonic acid, the source of FeCO3 precipitation and the product of the cathodic reaction. In addition to the spatial separation of the production of Fe++ and HCO3-, the galvanic coupling between the steel and cementite layers is also the principal cause of internal acidification in these layers, since the HCO3- ions are then removed from the steel surface by electromigration. This can facilitate the initiation of localized corrosion, by lateral galvanic coupling.

In the general paradigm of the study of the CO2 corrosion of steel, recent developments have revealed the decisive influence of variations in the protectiveness of corrosion layers 1. Thus, corrosion layers containing the same solid components can be either extremely protective 2 or very little so, or can even be corrosive 3. For a C-Mn steel, such as the St 52 grade used in the latter two studies 2,3, the corrosion layers are composed of an insoluble corrosion product, iron carbonate (FeCO3), and/or an undissolved component from the steel, namely cementite (Fe3C). Like the metal, cementite is an electronic conductor. The cathodic corrosion reaction can therefore occur as readily on the cementite as on the surface of the steel itself. This leads to the possibility of galvanic coupling between the steel substrate and the layer of undissolved cementite.

Based on the general electrochemical characteristics of the reduction of H+ ions and the anodic dissolution of iron, the first study on the aggravation of corrosion by layers of undissolved cementite 3 initially led to the conclusion that galvanic coupling between the steel and the cementite was only of a marginal nature, and suggested that local acidification must occur in the aqueous medium trapped within the pores of the solid layer. However, it is now realized 4 that these basic electrochemical properties, which had been observed in strong acid solutions and assumed to be general, cannot in fact be transposed to CO2 media. Under specific conditions of this sort, different orders of reaction for the reduction of H+ and different slopes of the individual Tafel lines lead on the contrary to a situation where it is impossible to distinguish between the effect of galvanic coupling and that of internal acidification 4. It is now necessary to consider that either one or the other may occur, or even both together.

Mechanisms determining protectiveness

In order to remain brief, only the general principles of protectiveness will be recalled 5. In addition to the two anodic and cathodic electrochemical reactions, anodic and cathodic, the system represented by a metal corroding beneath a corrosion layer also involves two or three different transport processes, corresponding to the supply of reactants for the cathodic process and the evacuation of the products of both the anodic and cathodic reactions. The transport between the external corrosive medium and the metal is accounted for mainly or partially by diffusion through the liquid contained in the pores of the corrosion layer, the remainder being due to the permanent growth of the layer in contact with the metal and its permanent redissolution in the external medium. If precipitation and redissolution are the majority processes, the corrosion layer will be called "soluble". Conversely, if diffusion in the liquid phase is the predominant mechanism, the layer will be called "insoluble". Depending on whether the overall process is governed by diffusion of the metal cation or by that of the precipitatable anion, the layers will be designated respectively as "insoluble cationic" (IC) or "insoluble anionic" (IA). Any imposed polarization will obviously greatly modify the interaction between the electrochemistry and the transport mechanism 6. From this standpoint, galvanic coupling between the metal and a conductive corrosion product will already cause a major modification in the mechanisms of protectiveness.

All these diffusion fluxes naturally imply marked gradients in the concentration of the different species involved, particularly those of the metal cation Mn+ and the corresponding precipitatable anion Xm-. This leads to local solubility equilibria inside the layer which can be vastly different from what might be expected in the external medium. In particular, in an IA layer, local depletion of the precipitatable anion Xm- causes a corresponding increase in the local solubility of the metal ion Mn+, and therefore allows higher Mn+ concentration gradients and fluxes, i.e. faster steady state corrosion rates 2,3,5.

The case of CO2 corrosion in steels

In this case, the insoluble corrosion product is iron carbonate, for which the solubility product is conventionally written in the form:

[ Fe++ ] [ CO3-- ] = Ks (1)

Nevertheless, in the presence of significant concentrations of carbonic acid and bicarbonate, the effective carbonate ion content becomes of second order compared to these species. Where transport phenomena are concerned, it is never advisable to reason in terms of trace components whose excessively low concentration completely excludes the possibility of direct transport. For example, at a pH of 6, the transport capacity of acidity via the highly mobile H+ (or H3O+) ions is totally negligible compared to the indirect effect of diffusion of tens or hundreds of mM/L of dissolved CO2, even though the mobility of the latter is low and the kinetics of the final hydration of CO2 and dissociation of H2CO3 are slow 4,7. Thus in order to conserve a realistic physical meaning for each solubility product, they will be systematically written in terms of the majority independent species, which can be transported directly. For example, HCO3- ions are enough concentrated for being directly transported, but the CO3-- are mainly formed or removed locally by chemical reaction. From this local equilibrium between HCO3- and CO3--., equation (1) can easily be rewritten in the form

[ Fe++ ] [ HCO3- ] = K2 Ks [ H+ ] = K [ H+ ] (2)

where K2 is the equilibrium constant for the second dissociation of carbonic acid.

At a given pH, the bicarbonate anion then appears "mathematically" in equation (2) as the precipitatable anion leading to the formation of solid iron carbonate, with a corresponding solubility "constant" equal to K [H+]. At the same time, HCO3- naturally remains the anion which buffers the carbonic acid 8, and is in solubility equilibrium with the partial pressure PCO2 :

[ H+ ] [ HCO3- ] = K1 [ CO2 ] = K*1 PCO2 (3)

Finally, HCO3- is also the "reduction product" of CO2. Actually, the mass balance of the oxidizing power, i.e. the whole ensemble involving all transport of acidity-carrying species, their "direct reduction" or the in situ generation of H+ just before reduction (which are equivalent in a mass balance), can be expressed as:

CO2 + H2O + e- H2CO3 + e- H+ + HCO3- + e- 1/2 H2 + HCO3- (4)

The three roles respectively represented by equations (2) to (4) are naturally unseparable, and thus occur simultaneously in the corrosion layer. There is therefore a strong and complex interaction between the local pH, the solubility of iron, the transport of Fe++ and HCO3-, and the respective localization of their electrochemical productions.

Generalization to other metals and acids

From CO2 to H2S : Given the completely parallel situation between CO2 and H2S 9, all that has just been said concerning FeCO3 can be immediately transposed to all ferrous sulfides, FeS or FeS1±e, i.e. in practice to all the sulfides other than pyrite and marcassite. It is sufficient to rewrite equations (2) to (4), where X- represents either HCO3- or HS- :

[ Fe++ ] [ X- ] = K [ H+ ] (5)

[ H+ ] [ X- ] = K1 [ HX ] (6)

HX + e- 1/2 H2 + X- (7)

From CO2 to acetic acid : The mathematics involved in the transposition of equations (2) to (4) to equations (5) to (7) are not related to the diacid character of H2CO3 and H2S, but solely to the valency ratio between Fe++ and X-. In these conditions, equations (5) to (7) can be transposed to all monoacids, and particularly to acetic acid, "HAc", and to all the carboxylic acids of similar pKa known to be present in produced waters, either directly or in the form of the corresponding anions X- 10. The solubility product for ferrous acetate, [ Fe++] .[ Ac- ]2 = Ks, can be reduced back to equation (5), with K1 = Ka and K = (1/[HAc]).Ks/Ka.

The conversion of uniform corrosion into localized attack

In the CO2 corrosion of steels, the maximum transport capacity is always that due to the CO2 itself 3. Furthermore, in uniform corrosion, the anodic and cathodic current densities are equal at all points. Equation (4) consequently indicates that Fe++ and HCO3- are always produced simultaneously, i.e. in the same amounts (in meq/L) and at the same place. This equality maximizes the solubility product of FeCO3 shown in equation (2), and from the resulting precipitation this ensures the greatest possible protectiveness of the deposit.

If for any external reason whatsoever a temporary local anode appears, the local equality between the anodic and cathodic currents would no longer hold. The production of Fe++ in would increase at this local anode and that of HCO3- would decrease. Due to the associated rise in the solubility of iron, the corrosion layer on the anode would become less and less protective. Conversely, on the cathode, the production of Fe++ would decrease and that of HCO3- would increase. The corresponding drop in the solubility of iron would then enhance the protectiveness of the deposit formed on the cathode.

There should therefore be a threshold beyond which any disturbance in the local balance iA/iK would spontaneously tend to intensify. This eventually leads to a situation where the presence of the galvanic couple enables iron to be produced at the anode without any bicarbonate being formed. Fe++ therefore becomes more soluble, and HCO3- less soluble. Consequently, if the bulk medium is itself buffered by some bicarbonate content, a possible local depletion in HCO3- at the anode, at constant CO2, will locally decrease pH below its value in the bulk. Conversely, the excess HCO3- formed at the cathode will decrease the local solubility of Fe++, and therefore prevent significant amounts of Fe++ from being produced there.

This process would naturally explain the random nature of the nucleation of pits, the difficulty of reproducing mesa attack in the laboratory, and the fact that certain water compositions stabilize such galvanic cells more than others 13. In fact, all depends on whether or not the following chain reaction is triggered 1 :

the presence of a galvanic cell spatially separates the anodic and cathodic reactions,

this separation locally modifies both the water chemistry and the protectiveness of the corrosion layers,

the difference in local protectiveness enhances the galvanic coupling.

In localized corrosion, galvanic coupling occurs in a relatively visible manner 13, since it takes place on a macroscopic scale, in a direction parallel to the surface of the steel. In the case of a layer of undissolved cementite, galvanic coupling between the cementite and the steel will necessarily have similar effects. However, until now, these have remained hidden, since they occur on a microscopic scale, and in a direction perpendicular to the metal surface.

Localization of FeCO3

In the same way as equation (4) represents the solubility of iron carbonate as a function of bicarbonate anion concentration, the precipitation reaction itself can be expressed in the form:

Fe++ + 2 HCO3- FeCO3 + CO2 + H2O (8)

The FeCO3 can thus precipitate not only on the steel, but also directly on the cementite Fe3C, due to the ambient concentration in Fe++ and the additional bicarbonate anions produced on cementite by the cathodic reduction of CO2 . A first consequence of the galvanic coupling between the steel and the cementite is therefore the possibility of FeCO3 precipitation at a certain distance from the steel, whereas in the absence of coupling it can only form on the steel, where the local Fe++ and HCO3- concentrations are a maximum. Among all the corrosion layer morphologies observed 2,3,14, the main difference between protective and unprotective forms is the presence of empty cementite in contact with steel in the protective case, and its absence in the unprotective one . However, in unprotective layers, the thickness of empty cementite does not appear to exert a decisive influence, no more so than does the presence or absence of FeCO3 further away from the steel surface. Similarly, for protective layers, the existence of empty cementite elsewhere than on the steel seems to be of no importance.

It is naturally very difficult to imagine even qualitatively the distribution of coupling currents on the cementite. The continuity and the electrical conductivity of this carbide network are difficult to predict. High resistance is frequently observed in cementite layers, which thus reveal to be poor effective conductors. Similarly, it has been found 5 that stable potential gradients can exist in the aqueous solution which permeates the layers. Finally variations in the extent to which the pores in the cementite are obstructed by FeCO3 can also influence the distribution of current densities. It is therefore impossible, even by numerical modeling, to determine whether the cathodic current density is uniform in the cementite, or whether it increases on approaching the metal. Indeed, the ohmic drops DU within the solution are growing from the metal outwards, but the gradients of the reducible species are in the opposite direction.

Whatever the situation, the presence of empty cementite in contact with the steel can at present only be explained by galvanic coupling which is either sufficiently strong, or whose range is sufficiently long. Conversely, its absence indicates that galvanic coupling is either inexistent or only slight, or occurs over only a very short range. Consequently, unprotective corrosion layers are always associated with significant galvanic coupling between steel and cementite.

Galvanic coupling and internal acidification

Another constant characteristic of unprotective corrosion layers 2,3 is the existence of oblique pseudo-polarization curves, with slopes close to 120 mV/log. The most recent analyses 4 show that this can be explained by the presence of galvanic coupling or by internal acidification, without it being possible to decide between one or the other, or both.

Furthermore, in the general models for protectiveness 5,6, the precipitatable anion is considered to come solely from the external medium, by diffusion or electromigration. However, in the CO2 corrosion of steels, the precipitatable anion HCO3- is itself a corrosion product, and is formed directly within the porous layer. Like Fe++, it must therefore be evacuated by diffusion into the external medium, i.e. in the opposite direction with respect to the original models. In this situation, galvanic coupling and acidification are necessarily closely related.

For example, in the sense acidification-coupling, the density and distribution of the coupling current naturally depend on the pH profiles, since the effect of pH on the values of iA and iK on the steel is different, and perhaps also on the values of iK on cementite. The interaction is even more marked in the reverse sense coupling-acidification, since coupling itself appears as a major source of acidification.

Case of a completely empty cementite layer

This simple case has been described in reference 3. The assumption that the concentration of dissolved CO2 is sufficient for its transport capacity to be greater than that of all other species will be conserved here. In the absence of coupling, HCO3- will be produced only on the steel. Because of the similarity of all the diffusion coefficients in water ^* , its removal by diffusion requires dynamic enrichment in HCO3- of the same order as that for Fe++ 5. This leads to an at least hyperbolic decrease in the local solubility Fes and the impossibility of evacuating large iron fluxes.

On the contrary, in the presence of coupling, the nearer to the outer surface the production of HCO3-, the easier it is to remove by diffusion. The dynamic enrichment necessary to evacuate a given flux decreases with the distance between the point of formation and the external medium. Since the total flux remains the same, the HCO3- concentration profile is therefore rapidly leveled. However, this can explain the absence of alkalinization but not internal acidification due to complete disappearance of HCO3-. As before 3, an electromigration effect is necessary to "force" the HCO3- anions out of the layer. In fact, a phenomena of this sort does indeed exist, due to the galvanic coupling. Consider a plane P parallel to the metal at a distance x from the surface. The fluxes of Fe++ and HCO3- ions crossing this plane correspond to ionic currents, for which the same sign convention will be used as for the electrochemical currents:

The flux of Fe++ ions is conservative, i.e. there is no FeCO3 precipitation. Expressed in mA/cm2, the flux through any plane P therefore corresponds to the current iA for the anodic reaction on the steel:

JFe++ = iA (9)

The coupling current iC(x) through this same plane P,i.e.the ionic current closing the circuits of the cathodes on the left of P, is equal to that necessary to cancel out the sum of iA and the overall current iK(x) representing the sum of the cathodic reactions taking place on the right of P (Error! Bookmark not defined.iK(x)Error! Bookmark not defined. < iA):

iC(x) + iK(x) + iA = 0 (10)

In the steady state regime, the species not involved in the reactions have no corresponding fluxes. The net ionic current iC(x) thus corresponds simply to the difference in the fluxes of Fe++ and HCO3-, whose signs are opposite:

iC(x) = JFe++ - JHCO3- (11)

Replacing JFe++ and iC by their corresponding values, and remembering that
iK(x) > - iA, gives:

JHCO3- = iA - iC(x) = 2 iA + iK(x) > iA > | iK(x) | (12)

This signifies that, at any plane in the cementite layer, the simultaneous presence of an ionic coupling current flowing towards the metal and a flux of Fe++ cations flowing towards the external medium causes an electromigrational flux of HCO3- anions, which is not only directed outwards, but is also greater than the rate of production of bicarbonate by all the cathodic reactions taking place between the metal and the plane considered. The HCO3- ions are therefore totally depleted in contact with the metal, leading to internal acidification. The difference between this total flux JHCO3- and the rate of cathodic production Error! Bookmark not defined.iK(x)Error! Bookmark not defined. is therefore a purely chemical diffusion flux, -D grad [HCO3-]. While galvanic coupling is indeed the driving force for acidification, it is its dependence on pH which will define the exact pH profile in the steady state. Precise numerical modeling will therefore not be easy, and must take into account recently published new results concerning the electrochemistry in CO2 media 4.

Case of a layer of cementite partially filled with siderite

The inverted Fe++ and HCO3- concentration profiles are perfectly able to maintain a lower degree of saturation in FeCO3 within the layer than in the outside solution. This explains why CO2 corrosion can continue in a medium saturated in corrosion product 1, and even why only the corrosion of steels by CO2 can lead to high FeCO3 supersaturations without the use of artificial means. However, with these profiles, the solubility limit of FeCO3 can also be exceeded at some point within the layer. During the resulting precipitation transient, the two species Fe++ and HCO3- are removed from their respective fluxes in equal amounts (in meq/cm2). It is also possible that this precipitation might not occur within the pores, but directly on the surface of the cementite, due to local alkalinization by the cathodic reduction of CO2.

By locally isolating the cementite from the water, and by reducing the porosity of the corrosion layer, the precipitation of FeCO3 thus causes an overall decrease in the galvanic coupling, and above all, profoundly modifies its spatial distribution. In particular, coupling with an empty external cementite layer is probably highly attenuated by the screen of ohmic drop represented by the region obstructed with FeCO3. The empty outer layer is therefore theoretically not very dangerous, in agreement with actual observations.

Conversely, equations (10) and (11) show that the process of internal acidification begins by electromigration whenever there is a deficit of cathodic reaction on the steel. There is therefore no need for large " diffusion distances". A thin layer of empty cementite in contact with the steel is thus sufficient to cause at least the onset of internal acidification. Only layers obstructed by FeCO3 directly in contact with the metal can be protective, and again this corresponds to what is found in practice.

These considerations explain the existence of multiple steady states 2, together with the sometimes erratic influence of the iron content in the corrosive medium:

If the iron content in the test medium is high right from the moment of immersion of freshly polished specimens, FeCO3 can precipitate on the metal, and the layer is protective. If the iron concentration subsequently falls and a certain amount of external redissolution of FeCO3 exposes an outer layer of empty cementite, this has no effect on the overall protectiveness of the corrosion layer.

If, on the other hand, the iron content of the medium becomes high only after an initial phase of corrosion leading to the formation of an empty cementite layer, then internal acidification prevents further precipitation of FeCO3 in contact with the metal, even though the outer part of the layer becomes obstructed. The layer is then unprotective, and even enormous iron supersaturations cannot subsequently render it protective 2.

Lateral stability of galvanic coupling

Uniform corrosion does not necessarily leave the surfaces concerned rigorously planar. Random fluctuations of various types occur, due to a number of causes (segregations, inclusions, crystal texture, turbulence, fretting, etc.). However, in general, after a small disturbance, a stable system reverts naturally to its initial state. There is then an average statistical compensation of all past disturbances at any given point. Conversely, a disturbance may take on an amplitude which is no longer negligible compared to the scale on which the system is naturally regulated. In the presence of multiple steady states 2 the system path can then incorporate a number of irreversible "branches".

For example, in an empty layer of porous cementite, a local undulation in the metal dissolution front will not change the uniform nature of the corrosion. In contrast, gradual clogging by FeCO3 can reduce the thickness of the empty part to a value close to the depth of the undulation. In addition to the galvanic coupling and electromigration flux strictly perpendicular to the initial metal surface, parallel components then arise. This leads to the possibility of spontaneously initiating the lateral galvanic coupling. In the same way as spinodal decomposition in alloys is a precursor to the precipitation of separate phases, lateral fluctuations of the galvanic coupling between steel and cementite can give rise to the macroscopic coupling involved in localized corrosion.

Here again, the proposed mechanism is in perfect agreement with the observation of the initiation of mesa attack by intense nucleation of pits 15, with the presence of empty cementite layers on the corroded "canyons" and sealed cementite layers on the unattacked plateaux.

In the long saga of the study of CO2 corrosion in steels 1, the consideration of galvanic coupling between the steel and and a layer of undissolved cementite provides the missing link in many mechanistic models. It explains the possibility of local depletion of precipitatable anion in spite of the absence of any precipitation (C instead of IA layers). It also explains how bicarbonate can be produced within the corrosion layer itself, by cathodic reduction of CO2, and then forcibly removed by electromigration. It shows how unprotective layers can even become corrosive, due to internal acidification 3. Finally, it accounts for the decisive influence which slight traces of free acetic acid can have on the corrosivity of a medium, by becoming concentrated to almost pure acetic acid in contact with the metal.