Computational model of a localized attack in CO2 corrosion in the presence

of acetic acid

J. Amri 1,2, ∗ , E. Gulbrandsen 1, R. P. Nogueira 2

1
Institute for Energy Technology (IFE), NO-1254 Kjeller, Norway
2
UMR5266 and 5631 INP Grenoble-CNRS-UJF, SIMAP and LEPMI,
BP 75, 38402 St Martin d’Hères, France

Summary

In previous work, we have shown that the presence of acetic acid (HAc) may cause
localized attacks in CO2 corrosion. The corrosion rate depends on many species and
equilibria, thus complicating the analysis of the mechanism. A numerical model of steady-
state behaviour of a single localized corrosion attack in CO2-containing media in the presence
of HAc is therefore developed. The objective is to overcome the system complexity and
analyze underlying mechanisms. The model is based on mass-conservation and consists of 2D
axis-symmetric attack in representative working conditions. Electrochemical kinetics
occurring on the working surfaces are described by Tafel equations fitted to recent
experimental data obtained on X65 pipeline steel. The present paper focuses on giving further
information about the chemistry and electrochemistry along the cavity depth. Particular
attention is paid to the effect of HAc concentration on the coupling behaviour between the
attack and the outer surrounding surface.
Keywords: CO2 corrosion; Pipeline steel; Acetic acid; Localized attack; Numerical modelling

1. Introduction

Although a wide research has been devoted to the internal CO2 corrosion of carbon
steel pipelines, this corrosion is still a major concern for oil and gas industry [1, 2]. The
corrosion can be even more severe in the presence of other acidic gases such as hydrogen
sulphide (H2S) or short chain carboxylic acids. Being sometimes present in the co-produced
aqueous phase in large amounts up to several thousand ppm, acetic acid (CH3COOH),
commonly abbreviated as HAc, is frequently the most prevalent among these organic acids.

∗
Corresponding author. Tel.: +47 63 80 63 41; fax: +47 63 80 62 58

E–mail address: [email protected]

1
Results issued from field observations and laboratory studies have shown that HAc causes an
additional contribution to the CO2 corrosion rate [3-5]. Rudimentary understanding however
exists regarding its basic role in the propagation stage of localized attacks. In our recent work
[6], HAc has been identified as a driving factor that triggers and sustains the growth of
localized attacks through galvanic coupling effect. Although common behaviours are
generally observed, the key factors involved in CO2 corrosion in the presence of HAc
influence the propagation mechanism in a complex way. It is therefore worthwhile to develop
a computational model that could overcome this complexity and solve the large set of mass-
conservation equations of tightly coupled and non-linear nature.

2. Conceptual description and assumptions

Theory: brief introduction

In a stagnant electrolyte subject to diffusion and electromigration where the mass-

transport phenomena due to the convective motion may be ignored, the net molar flux of
specie j is given by the Nernst-Planck law:
N j = - D j∇ ( C j ) - u jz jC jF∇ ( V ) (1)

Dj is the effective diffusion coefficient, Cj is the concentration, uj is the mobility, zj is the

charge number, F the Faraday’s constant and V denotes the electrostatic potential in the
electrolyte. The first term on the right hand side represents the diffusion contribution arising
from a concentration gradient while the second term represents the migration contribution
induced by the electric field in the electrolyte. Assuming a mass-conservation condition, the
full transport-reactive phenomena may be fundamentally described by:
∂C j
+ ∇(Nj ) = R j (2)
∂t
Rj is the reaction rate including both production and consumption parts. Substituting Nj in
equation (2) leads to the following general form:
∂C j
= D j∇ 2 ( C j ) + z ju jF∇ ⎡⎣C j∇ ( V ) ⎤⎦ + R j (3)
∂t
The relation between the electrostatic potential and the space charge density ρ in the
electrolyte is given by the Poisson’s equation:
ρ F
∇2 ( V ) = -
ε 0ε r
=-
ε 0ε r
∑z C
j
j j
(4)

ε0 and εr are the vacuum permittivity and the relative permittivity of the electrolyte,
respectively.

Description of the model and assumptions

The model consisted of a single localized corrosion attack in dilute electrolyte. All the
diffusion coefficients were assumed isotropic and constant. A major geometrical

2
simplification was applied by assuming conveniently the localized attack as a perfect
cylindrical hole of finite depth within the range 0-20 mm. Thanks to the cylindrical symmetry,
the model was therefore reduced to two-dimensional pattern. The axial symmetry also
allowed the consideration of only the one-half of the model, thus reducing the resolution time.
We assumed that the natural convective forces are sufficiently large to continuously ensure
the replenishment of species and maintain the equilibrium conditions throughout the bulk
electrolyte. In order to further simplify the model, the calculation domain for the outer
electrode was set within the Nernst diffusion layer of a thickness δ estimated to 300 μm for
natural convection conditions [7].

Figure 1 – Concentration profile for a specie j as a function of the distance from the
metal surface in steady-state diffusion/convection regime. C eq s
j and C j are the equilibrium and

surface concentrations, respectively.

According to Poisson’s equation (4), (ε0εr)-1 ≈ 109 is sufficiently large that any slight
deviation from the electroneutrality induces very large electrical restoring forces [8, 9]. These
forces tend to remove charge gradients on a much faster timescale than those associated with
pure diffusion processes [10]. In order to minimize this effect, we assumed that the local
charge neutrality is fulfilled at any time.

∑z C
j
j j =0 (5)

The boundary settings are particularly important since they describe both the chemical
and electrochemical phenomena occurring at the electrode-electrolyte interface. In this
condition, the molar fluxes are proportional to the current density generated by the cathodic
and anodic reactions at the working surfaces through the Faraday’s law.
i = ia + ic = F ∑ z j N j (6)
j
Figure 2 shows a dimensional description of the model as well as the corresponding
boundary settings.

3
 Figure 2 – Dimensional description of the 2D-axisymmetric attack as well as the
corresponding boundary settings used in the modelling approach. The chamfer line (5) is
included to smooth the straight corner to simplify the numerical grid pattern.

The kinetics of electrochemical reactions occurring on the electrode surface is

essential for determining the corrosion rate of a metal exposed to a corrosive medium. These
reactions strongly depend on the electrode potential E.
E = φ w -φ r (7)

φ w and φ r are the Galvani potentials of the working and the reference electrodes, respectively.
For a corroding electrode (net current = 0) at its open circuit potential Ecor, as well as for IR-
compensated data in polarized conditions, the interfacial electrochemical processes can be
described in terms of current-potential relationships, for example Tafel equations if
appropriate.
⎡ ⎛ E - E cor ⎞ ⎛ E - E cor ⎞⎤ (8)
i = i cor ⎢exp ⎜ 2.3 ⎟ - exp ⎜ - 2.3 ⎟⎥
⎣ ⎝ ba ⎠ ⎝ bc ⎠⎦
icor is the corrosion current density and ba and bc are the anodic and cathodic Tafel slopes,
respectively. To simulate the galvanic coupling effect, the sum of the integrated currents for
the attack and the external surface are required to equal zero. In the finite-elements
calculations, both the potential and the current distributions are rather calculated in the
electrolyte domain according to the configuration (B) of Figure 3. In that case, the electrode is
just represented as a boundary condition describing the current density-potential relationship
at the electrode-electrolyte interface. There is therefore a need for a definition that relates to
the electrolyte domain. One simple definition, which directly relates to experimental
measurements, is the potential V(r,θ,z) of a reference electrode versus the corroding electrode.
V ( r,θ,z ) = φr ( r,θ,z ) - φ w (9)

The electrolyte potential V0 at a given point of the boundary representing the working
electrode surface as well the new open circuit potential Vcor are therefore given by:
V0 = - E (10)

4
 Vcor = - E cor (11)

Here E is the potential of the working electrode versus the reference electrode located at a
given point of the surface (i.e., IR-compensated data). It is worth noticing that the new
electrode polarization ΔV is still the same since:
ΔV = Vcor - V0 = ( - E cor ) - ( - E ) = E - E cor (12)

Figure 3 – Potential definitions according to experimental (A) and modelling (B) approaches,
respectively. RE = reference electrode.

Model application

The electrochemical kinetics were described by Tafel equations fitted to recent

experimental data obtained on X65 pipeline steel in CO2-containing media in the presence of
HAc. According to experimental results, the corrosion potential has been found to increase
monotonically with HAc concentration [6]. In the present model we used the Ecor data
reported in Figure 2 of Ref [6]. A similar fitting of icor - HAc concentration was also used. The
anodic and cathodic Tafel slopes were however found to be fairly independent of HAc
concentration and were estimated to 45 mV/dec and -120 mV/dec, respectively. Since only
proton reduction and iron dissolution were the only electrochemical reactions considered in
this model, the net fluxes of H+ and Fe2+ at the electrode boundaries are given by equation (6);
the fluxes for all other species were hence assumed zero. The simulated corrosive medium
consisted of 10 mM aqueous NaCl containing different amounts of HAc saturated with 1 bar
CO2. Chemical reactions as well as their corresponding equilibrium ( K j ), forward ( k fj ) and
backward ( k bj ) reaction rate constants at 25 °C are listed in Appendix below. In addition to
water, these reactions involve 11 aqueous species. Their respective diffusion coefficients are
also given in Appendix. The system was anaerobic. In this first approach, the formation of
solid corrosion products of iron carbonate (FeCO3) was omitted. This simplification is
supported by the fact that FeCO3 precipitation is very slow at room temperature [3, 11], and
the formation of protective layers takes weeks at such temperatures even under conditions of
artificially elevated pH [11]. The effect of FeCO3 precipitation on the overall electrochemical
kinetics, in particular at elevated temperatures, is nevertheless expected to be the subject for
further model development in the near future.

5
3. Results and discussions

Unless otherwise noted, all calculations were performed on a single attack with non-
reactive and electrically insulating sidewalls in simulated electrolytes of 10 mM aqueous
NaCl and 600 ppm HAc saturated with 1 bar CO2 at 25 °C. The initial pH was set to 7. Since
no moving boundaries are involved here, each calculation was therefore carried out
independently in a static geometry at a given depth.
The predicted composition of the electrolyte down the attack at a given depth is shown
in Figure 4. Large quantities of Fe2+ are generated by the metal dissolution with
concentrations exceeding more than 150 mM deeper into the attack. The increasing positive
charges of iron down the attack are balanced by a large migration of Cl- from the bulk
electrolyte and the formation of HCO3- and Ac-, thus preserving the local charge neutrality
throughout the cavity. The relatively high concentration observed for HCO3- results from a
high dissociation rate of H2CO3. According to the Law of Mass Action, the consumption of
carbonic acid boosts the forward reaction of carbon dioxide hydration. On the other hand, the
excess of bicarbonate anions leads to a subsequent increase of CO32- concentration through
the dissociation reaction of HCO3-.

Figure 4 – Predicted concentration profiles along the cavity depth in 10 mM NaCl, 600 ppm
HAc and 1 bar CO2 at 25 °C.

The other important feature of the steady-state behaviour of the attack is related to the
variation of undissociated HAc concentration between the bulk electrolyte and the cavity. As
depicted in Figure 4, the predicted results clearly show a depletion of HAc inside the attack.
This depletion is not only due to the consumption of protons by the cathodic reaction at the
attack bottom but also to mass-transport limitations induced by geometric constraints as
illustrated by Figure 5. Shallow attacks are indeed found to be more easily fed by mass-
transport of HAc than deep ones. The acid depletion inside the attack induces a concentration
gradient in respect with the bulk, thus giving birth to a HAc differential concentration cell.

6
 Figure 5 – Predicted concentration profiles of undissociated HAc along the cavity for three
fixed depths (in mm) in 10 mM NaCl, 600 ppm HAc and 1 bar CO2 at 25 °C.

The depletion of HAc inside the attack is an important fact, since the corrosion
potential decreases with decreasing HAc concentration, as reported in our previous work [6].
The differential cell thus establishes a potential difference so as the attack bottom is
anodically polarized by the outer large surface. This result is in good agreement with the
experimental findings, where HAc was clearly found to act as a factor that triggers and
sustains the growth of the localized attack through galvanic coupling effect. As the depletion
goes on deeper into the attack, CO2 corrosion becomes predominant, thus inducing a local
alkalinization as shown by the predicted pH profile in Figure 6.

Figure 6 – Predicted pH profile along the cavity depth in 10 mM NaCl, 600 ppm HAc and 1
bar CO2 at 25 °C.

7
 The predicted distribution of the coupling current between the attack bottom and the
outer large surface, which is the sum of the individual current densities integrated over the
area of the exposed surface, is depicted in Figure 7. The model reveals that the current reaches
a maximum at 12 mm. The calculated result corroborates the experimental ones obtained with
the help of zero resistance ammeter and artificial pit electrodes. In fact, typical distributions of
the coupling current flowing between the actively corroding attack bottom and the outer large
surface clearly showed the existence of a transition depth at 8 mm below which the current
increases and beyond which it sharply drops off. It is also worth noticing that the magnitude
of the calculated current is in fair agreement with experimental data [6].

Figure 7 – Predicted distribution of the coupling current along the cavity depth in 10 mM
NaCl, 600 ppm HAc and 1 bar CO2 at 25 °C.

The predicted current-depth peak however seems not as sharp as that found
experimentally. This may be ascribed to the fact that measured electrochemical data
implemented in the present model are limited to pH ≤ 5.5 and PCO2 = 1 bar. Implementation of
data for higher pH and lower PCO2 would probably give a sharper current-depth peak. The
model nevertheless helps us identify a need for further experimental data within the predicted
ranges. At 2 mm depth, however, the pH is only slightly above 5.5 and the partial pressure of
CO2 is practically 1 bar. This gives good agreement between predicted and experimental
results as illustrated by Figure 8.

8
Figure 8 – Predicted (■) and measured (□) coupling currents versus the initial concentration
of HAc for an attack of 2 mm depth in 10 mM NaCl and 1 bar CO2 at 25 °C. The figure also
shows predicted data for 10 mm depth (●).

4. Conclusion

A flexible and predictive model of the steady-state behaviour of a single attack is

developed and applied to pipeline steel in CO2/HAc corrosion. A promising agreement is
obtained between predicted and experimental results. This agreement gives strong and
complementary evidence about the major role played by HAc in the growth mechanism of
localized attacks through galvanic coupling effect. The model also predicts the existence of a
transition depth for the current-depth distribution, which is consistent with experimental
observations. The overestimation of the coupling current particularly for deeper attacks is
however likely to be due to the restricted range of experimental data implemented in the
model. The predicted results nevertheless help us design further experiments in order to get
missing data required to develop the model. An extension of the model is still ongoing.

Acknowledgments

The present work, including a Ph.D. grant for J. Amri, is set within the framework of
the project “Stabilisation of the Top-of-the-Line Corrosion Rate” funded by The French
Norwegian Foundation, Total, and ConocoPhillips. One of the authors (J. Amri) is indebted to
Dr. B. Malki from SIMAP Laboratory (France) for his technical assistance and the
stimulating discussions on corrosion modelling.

9
Appendix

Chemical reactions

Water dissociation: H 2 O U H + + OH - ; K w = 10-8 mol2 m -6 ; k bw = 7.85×107 mol-1 m3 s -1

Acetic acid dissociation: HAc U H + + Ac - ; K ac = 1.75×10 -2 mol m -3 ; k ac

f
= 3.20×105 s -1

Carbon dioxide dissolution: CO2 ( g ) U CO2 ( aq ) ; Ksol = 34 mol m-3 bar -1

Carbon dioxide hydration: CO2 ( aq ) + H2 O U H2 CO3 ; K hyd = 2.58×10-3 ; k fhyd = 3×10−2 s-1

Carbonic acid dissociation: H 2 CO3 U HCO 3- + H + ; K ca = 0.156 mol m -3 ; k fca = 7.80×10 6 s -1

Bicarbonate anion dissociation: HCO 3- U CO 32- + H + ; K bi = 4.67×10 -8 mol m -3 ; k fbi = 6.07 s -1

Diffusion coefficients (units: 10-9 m2 s-1)

Na+ Cl- H+ OH- Fe2+ HAc Ac- CO2 H2CO3 HCO3- CO32-
1.3 2 9.3 5.3 0.7 1.2 1.1 1.96 2 1.1 0.92

Reference

1. R. Nyborg, CORROSION/2002, Paper no. 02233, (Houston, TX: NACE

International, 2002).
2. B.M. Kermani, A. Morshed, Corrosion 59 (2004), 659-683.
3. E. Gulbrandsen, K. Bilkova, CORROSION/2006, Paper no. 06364, (Houston, TX:
NACE International, 2006).
4. Y.M. Gunaltun, D. Larrey, CORROSION/2000, Paper no. 71, (Houston, TX:
NACE International, 2000).
5. M. W. Joosten, J. Kolts, J. W. Hembree, M. Achour, CORROSION/2002, Paper
no. 02294, (Houston, TX: NACE International, 2002).
6. J. Amri, E. Gulbrandsen, R.P. Nogueira, Electrochemistry
Communications, Volume 10, Issue 2, February 2008, Pages 200-203.
7. S. Nesic, B. F. M. Pots, J. Postlethwaite, N. Thevenot, Journal of Corrosion
Science and Engineering, Vol. 1, paper 3 (1995).
8. S. M. Sharland, Corrosion Science 27, 289 (1987).
9. Turnbull, Modelling aqueous corrosion from individual pits (…), NATO ASI
Series, Vol. 266 (1993).
10. S. M. Sharland and P. W. Tasker, Corrosion Science 28, 603 (1988).
11. A. Dugstad, R. Nyborg, M. Seiersten, “Flow assurance of pH stabilized wet gas
pipelines”. CORROSION/2003, Paper No. 03314, San Diego, 16-20 March 2003.

10