CORROSION SCIENCE

An Electrochemical Model for Prediction

of Corrosion of Mild Steel
in Aqueous Carbon Dioxide Solutions

S. Nesic,* J. Postlethwaite,** and S. Olsen***

ABSTRACT KEY WORDS: aqueous solutions, carbon dioxide corrosion,

carbon steel, films and film formation, flow, mass transfer,
A predictive model was developed for uniform carbon dioxide modeling, pH, temperature, uniform corrosion, weight loss
(CO2) corrosion, based on modeling of individual electro-
chemical reactions in a water-CO2 system. The model takes INTRODUCTION
into account the electrochemical reactions of hydrogen ion
(H+) reduction, carbonic acid (H2CO3) reduction, direct water Most of the models used in industry to predict car-
reduction, oxygen reduction, and anodic dissolution of iron. bon dioxide (CO2) corrosion rates are “worse case”
The required electrochemical parameters (e.g., exchange
models. The ones in the public domain, such as the
current densities and Tafel slopes) for different reactions
models of deWaard and Lotz1 and Dugstad, et al.,2
were determined from experiments conducted in glass cells.
The corrosion process was monitored using polarization
are semiempirical. Such models do not lend them-
resistance, potentiodynamic sweep, electrochemical imped- selves to extrapolation easily. Moreover, including the
ance, and weight-loss measurements. The model was effects of new parameters is possible only through
calibrated for two mild steels over a range of parameters: correction factors that further obscure the underly-
temperature (t) = 20°C to 80°C, pH = 3 to 6, partial pressure ing corrosion mechanisms and that can have
of CO2 (PCO2) = 0 bar to 1 bar (0 kPa to 100 kPa), and ␻ = dubious interactions once more than one correction
0 rpm to 5,000 rpm (vp = 0 m/s to 2.5 m/s). The model was factor is applied.1 Because of the high cost and long
applicable for uniform corrosion with no protective films duration of corrosion experiments, development of
present. Performance of the model was validated by compar- improved prediction models is required.
ing predictions to results from independent loop experiments.
As part of a long-term project on CO2 corrosion
Predictions also were compared to those of other CO2 corro-
conducted in Norway, a fully theoretical approach
sion prediction models. Compared to the previous largely
empirical models, the model gave a clearer picture of the has been taken by modeling individual electrochemi-
corrosion mechanisms by considering the effects of pH, cal reactions occurring in a water-CO2 system. This
temperature, and solution flow rate on the participating is not a novel approach per se in that such models
anodic and cathodic reactions. have been created for advanced analysis of polariza-
tion scans3-8 and for educational purposes.9 In the
area of CO2 corrosion, Bonis and Crolet briefly men-
Submitted for publication December 1994; in revised form, July tioned one such model, but they have not presented
1995. it in detail.10 They demonstrated performance of the
* Institutt for Energiteknikk, P.O. Box 40, N-2007, Kjeller, Norway.
Present address: Department of Mechanical Engineering, model by comparing it to predictions made with the
University of Queensland, Brisbane, Qld. 4072, Australia. model of deWaard and Milliams, obtaining results of
** University of Saskatchewan, Saskatoon, Canada. Present address:
Institutt for Energiteknikk, P.O. Box 40, N-2007, Kjeller, Norway.
a similar order of magnitude.11 Gray, et al., presented
*** Statoil, Postuttak, 7004, Trondheim, Norway. an electrochemical model as part of an experimental
0010-9312/96/000067/$5.00+$0.50/0
280 © 1996, NACE International CORROSION–APRIL 1996
 CORROSION SCIENCE

study of CO2 mechanisms.12-13 That study was ex-

tended over too broad a pH range, so that all stated
assumptions did not hold (especially for pH > 7,
where protective films probably formed). However,
the overall theoretical approach served as an inspira-
tion for the present work.
Processes to be modeled in a water-CO2 system
are the electrochemical reactions at the metal surface
and the transport processes of all the species in the RE
system, such as hydrogen ions (H+), CO2, carbonic
CO2 gas in
acid (H2CO3), and iron (Fe2+). The present work fo-
CO2 gas out
cuses on the model of the electrochemical reactions
occurring in an acidic solution with dissolved CO2. Luggin cappillary

Transport processes were treated in a simplified way Platina CE

by assuming independent diffusion of the species Rotating cylinder
and by using well-established mass-transfer coeffi-
Water out
cients for the hydrodynamic systems of interest: the
rotating cylinder and pipe flow. A more detailed Temperature probe

approach to the coupled transport of species in a Water in

water-CO2 system will be incorporated in the future.14 pH electrode

Performance of the model was demonstrated by WE
comparing predictions to results from flow loop ex-
periments and results of the CO2 corrosion prediction FIGURE 1. Schematic of the experimental cell.
models of deWaard and Lotz1 and Dugstad, et al.2

EXPERIMENTAL tion of CO2 in the water also was measured in se-

lected experiments. Electrochemical measurements
Equipment were made using a Gamry Instruments† potentiostat
Experiments were conducted at atmospheric with an eight-channel multiplexer connected to a
pressure in a battery of eight glass cells. Gas (CO2 or personal computer.
nitrogen [N2]) was bubbled through the cells continu-
ously. The corrosion process was studied using Material
electrochemical techniques and occasionally checked Two low-carbon steels were tested: St52 (corre-
using weight-loss measurements. A three-electrode sponding to ASTM A537 Grade 1, [UNS K12437])(2)
setup was used (Figure 1). A rotating cylinder elec- which is a typical construction steel and X-65 steel,
trode (RCE) with a speed control unit (0 rpm to which is a typical pipeline steel. Chemical composi-
5,000 rpm)(1) was used as the working electrode (WE). tions of the steels are given in Table 1. The WE was
A concentric platinum ring was used as a counter machined from the parent material into a cylinder
electrode (CE). A saturated silver-silver chloride 10 mm (0.394 in.) in diameter and 10 mm (0.394 in.)
(Ag-AgCl) reference electrode (RE) was connected to in length. The exposed area of the specimen was
the cell externally via a Luggin capillary and a porous 3.14 cm2 (0.487 in.2).
wooden plug. The rotation speed of the WE was con-
trolled using a stroboscope. The pH was monitored Procedure
with an electrode immersed in the electrolyte. Tem- The glass cells were filled with 3 L (0.793 gal) of
perature (t) was followed with a platinum-100 probe electrolyte, which was distilled water + 1 mass%
that also served as an input for the temperature sodium chloride (NaCl). In different experiments, CO2
regulating system, which was a hot plate combined or N2 was bubbled through the electrolyte (min. 60
with a magnetic stirrer. Oxygen concentration was min) to saturate or deaerate the solution. Monitoring
monitored using an Orbisphere† oxygen meter. The of pH and O2 concentration was used to judge when
concentration of Fe2+ was measured occasionally the solution was in equilibrium. When needed,
using a photospectrometric method. The concentra- hydrochloric acid (HCl) or sodium bicarbonate
(NaHCO3) was added to adjust the pH. The tempera-
ture was maintained within ± 1°C in all experiments.
(1)
5,000 rpm for the cylinder corresponded to a peripheral velocity of Before each polarization experiment, the steel
2.61 m/s (0.8 ft/s), a shear stress of 25 Pa, and a Reynolds
number of 26,175. WE surface was polished using 500- and 1,000-grit
(2)
UNS numbers are listed in Metals and Alloys in the Unified silicon carbide (SiC) paper, washed with alcohol,
Numbering System, published by the Society of Automotive
Engineers (SAE) and cosponsored by ASTM.
mounted on the specimen holders, and immersed
†
Trade name. into the electrolyte. The free corrosion potential was

CORROSION–Vol. 52, No. 4 281

CORROSION SCIENCE

TABLE 1
Chemical Composition of the Steels Used for the WE (mass%)
C Mn Si P S Cr Cu Ni Mo Al

X-65 0.065 1.54 0.25 0.013 0.001 0.05 0.04 0.04 0.007 0.041
St52 0.130 1.25 0.35 0.022 0.004 0.12 0.31 0.08 0.02 0.035

TABLE 2 Potentiostatic measurements also were con-

Experimental Conditions ducted in which the potential was kept at a desired
level while the rotational speed was varied. Experi-
Test solution Water + 1 mass% NaCl
Test material Low-carbon steel, St52 and X-65 mental conditions are summarized in Table 2.
Temperature 20°C to 80°C
Pressure 1 bar N2 or CO2 RESULTS AND DISCUSSION
pH 3 to 6
Fe2+ < 1 ppm
More than 50 glass cell experiments were con-
Dissolved oxygen < 20 ppb
Velocity Static to 5,000 rpm ducted. Many experiments were repeated several
Test duration 0.5 h to 48 h times to check reproducibility of the results. To
Sweep rate 0.1 mV/s to 0.2 mV/s clarify the effect of CO2, some experiments were done
Polarization resistance From –5 mV to 5 mV (vs Eoc) in solutions purged with N2 where the pH was ad-
AC impedance ± 5 mV vs Eoc from 1 mHz to 100 kHz
justed by adding HCl. Selected experiments have
Potentiodynamic sweep From –600 mV to 200 mV (vs Eoc)
IR compensation Automatic or manual been presented here to explain the most important
effects.

Effect of CO2
followed immediately after immersion. Depending on Measurements at pH 4 — In Figure 2, potentiody-
the conditions, the potential stabilized within ± 1 mV namic sweeps done on St52 steel are compared for
in 1 min to 10 min. two solutions at pH 4, first for one purged with N2
Polarization resistance (Rp) measurements were (pH adjusted by adding HCl) and the other for one
conducted by polarizing the WE ± 5 mV from the free purged with CO2 (pH adjusted with NaHCO3).(3) Ex-
corrosion potential and scanning at 0.1 mV/s. In perimental results are overlaid with the predictions of
most experiments, automatic IR drop compensation individual reactions generated with the model pre-
was used. In other cases, solution resistance was sented. The solution containing CO2 produced higher
measured independently using alternating current cathodic currents, probably because of the additional
(AC) impedance, and the measured Rp then was cor- cathodic reaction in the direct reduction of H2CO3.11
rected. AC impedance measurements were done by This was investigated using potentiostatic mea-
applying an oscillating potential ± 5 mV around the surements where the potential was kept in the
free corrosion potential to the WE using the fre- limiting current (ilim) region while the rotation speed
quency range 1 mHz to 100 kHz. was varied (Figure 3). The gap between the two
At the end of each experiment, the potentiody- curves even at stagnant conditions confirmed the
namic sweeps were conducted, starting 500 mV to assumption of Schmitt and Rothman15 and Eriksrud
600 mV below and finishing 100 mV to 200 mV over and Søntvedt16 that there is a flow-independent com-
the free corrosion potential. The typical scanning rate ponent of ilim in CO2 solutions that probably is
was 0.1 mV/s to 0.2 mV/s. In some cases, the ca- controlled by a chemical step (i.e., the hydration of
thodic and anodic sweeps were conducted separately CO2 to H2CO3). However, the present work concerned
starting from the free corrosion potential. The ca- the turbulent region comprising all the points above
thodic sweeps sometimes were repeated by sweeping ␻ > 50 rpm. In this region, a linear relationship was
in the opposite direction, without significant differ- observed for ilim vs ␻0.7 in both CO2 and HCl solutions,
ence in the result. In each experiment, the anodic which was to be expected from theory.19 At a given
sweeps were conducted only once for a single WE flow rate, ilim for a CO2 solution tentatively can be
specimen and a given electrolyte (starting from the separated into three components: one relating to the
free corrosion potential) since they altered the speci- diffusion of H+, one to the hydration of CO2, and one
men surface and contaminated the electrolyte with to the diffusion of H2CO3 respectively; in agreement
significant amounts of dissolved iron (Fe2+ > 3 ppm). with the proposals of Schmitt and Rothman.15
Typically, the Fe2+ concentration was kept at < 1 ppm. Figure 2 shows that the H2O reduction was
under charge-transfer control and did not seem to be
(3)
Current is shown throughout this work in A/m2. This is
affected significantly by addition of CO2. The same
convenient as for iron dissolution: 1 mm/y = 1.155 A/m2. was true for the anodic dissolution of Fe2+. The Tafel

282 CORROSION–APRIL 1996

 CORROSION SCIENCE

FIGURE 2. CO2 effect at pH 4, water + 3% NaCl, PCO2 = 1 bar FIGURE 3. Effect of CO2 on ilim obtained with a RCE, pH 4, water +
(100 kPa), t = 20°C, 1,000 rpm, τ = 1.7 Pa, Re = 5,235, and St52 steel. 3% NaCl, PCO2 = 1 bar (100 kPa), t = 20°C, and St52 steel.

slopes of the H2O and Fe2+ lines were ≈ 120 mV/de-

cade and 40 mV/decade, respectively.
Figure 4 shows Rp measurements on St52 steel
at different rotating speeds. The B values used to
obtain the corrosion current (icorr) were calculated by
using the model presented below and typically were
16 mV to 20 mV. There was a flow effect on both
corrosion curves which probably came from the
mass-transfer ilim for H+ reduction. The effect of addi-
tion of CO2 on the corrosion rate was small, as the
reduction of H2CO3 gave a smaller contribution to the
overall corrosion reaction than H+ reduction at pH 4.
Measurements at pH 5 — Similar experiments FIGURE 4. Effect of CO2 on icorr obtained with a RCE, pH 4, water +
were conducted at pH 5. In the solution with HCl at 3% NaCl, PCO2 = 1 bar (100 kPa), t = 20°C, and St52 steel.
pH 5, there were few H+ ions so that a distinct ilim
could not be observed (Figure 5). The dominant ca-
thodic reaction was direct reduction of H2O. Addition
of CO2 produced a clear increase in the overall ca-
thodic current. The anodic line remained unchanged
with a slope of ≈ 40 mV/decade.
The effect of CO2 on icorr at pH 5 was much stron-
ger than at pH 4. The direct reduction of H2CO3 was
the dominating cathodic reaction at pH 5. The contri-
bution by the reduction of H+ was much smaller than
at pH 4. Over the whole range of rotation speeds at
pH 5, icorr was increased two to three times by addi-
tion of CO2 (Figure 6). In this figure, the lines are
nearly horizontal, indicating there was very little
effect of flow. The reason for this was that the domi-
nating cathodic reaction in CO2 solutions was H2CO3 FIGURE 5. CO2 effect at pH 5, water + 3% NaCl, PCO2 = 1 bar
reduction, which was not flow sensitive. The same (100 kPa), t = 20°C, 1,000 rpm, τ = 1.7 Pa, Re = 5,235, and St52 steel.
was true for the HCl solution at pH 5, where direct
water reduction also was not flow sensitive.
tions, the main cathodic reaction was reduction of
Effect of Velocity H+ ions, as shown by Stern.17 The observed flow-
Measurements at pH 3 Without CO2 — In these dependent cathodic ilim represented the diffusion limit
experiments on St52 steel, the solution was purged of H+ ions from the bulk to the metal surface.12,17 For
with N2, and pH was adjusted to 3 by adding HCl. higher overpotentials, the dominant cathodic reaction
Potentiodynamic sweeps for different rotational changes to the direct reduction of water, as shown by
speeds are shown in Figure 7. Under these condi- Stern.17-18 Tafel behavior for the H+ reduction reaction

CORROSION–Vol. 52, No. 4 283

CORROSION SCIENCE

FIGURE 6. Effect of CO2 on icorr obtained with a RCE, pH 5, water + FIGURE 7. Velocity effect in HCl solution at pH 3, water + 3% NaCl,
3% NaCl, PCO2 = 1 bar (100 kPa), t = 20°C, and St52 steel. PN2 = 1 bar (100 kPa), t = 20°C, and St52 steel.

FIGURE 8. Velocity effect in HCl solution at pH 4, water + 3% NaCl, FIGURE 9. Velocity effect in CO2 solution at pH 4, water + 3% NaCl,
PN2 = 1 bar (100 kPa), t = 20°C, and St52 steel. PCO2 = 1 bar (100 kPa), t = 20°C, and St52 steel.

was clear only for the highest rotating speed, with a adjusted by adding NaHCO3 (Figure 9). The potentio-
slope of ≈ 120 mV/decade. For higher overpotentials, dynamic sweeps with CO2 (Figure 9) had larger ilim
all three cathodic curves converged onto the same values when compared to HCl solutions (Figure 8) at
line, which was the direct reduction of H2O. The the same pH. Further, the effect of
three anodic curves displayed clear Tafel behavior, rotation speed was much smaller, suggesting that
with a slope of ≈ 40 mV/decade. No effect of rotating there was a flow-insensitive component to ilim. The
speed was noticed on the anodic reaction. ilim value for the reduction of H2CO3 was chemical
Measurements at pH 4 Without CO2 — Potentio- reaction-controlled and insensitive to flow. The flow
dynamic sweeps in a solution purged with N2 at sensitivity observed related to the contribution to ilim
pH 4 (obtained by adding HCl) are shown in Figure 8 by the reduction of H+, which was mass transport-
for St52 steel. The overall shape of the curves was controlled. This could be seen by comparing ilim at
similar to those in the experiments at pH 3 except 1,000 rpm and 4,000 rpm with and without CO2. In
that all ilim values were 10 times lower. This was the CO2 solution, the current increased by a factor of
expected as the bulk concentration of H+ ([H+]b) was 1.8. In the solution without CO2, the curent in-
10 times lower at pH 4. For the lowest rotating speed creased by a factor of 3.5. The anodic dissolution of
(100 rpm), no ilim was distinguishable as the dominat- iron was not affected by the flow.
ing cathodic reaction throughout the measured range Measurements at pH 5 with CO2 — Flow sensitiv-
was the direct reduction of H2O, which was under ity at pH 5 was tested for a solution purged with CO2
activation control. where the pH was adjusted by adding NaHCO3. A
Measurements at pH 4 with CO2 — Subsequent potentiodynamic sweep conducted on a St52 RCE at
experiments were conducted with the same steel in a 1,000 rpm is presented in Figure 10 along with indi-
solution purged with CO2 instead of N2. The pH was vidual cathodic reactions and the theoretical sweep

284 CORROSION–APRIL 1996

 CORROSION SCIENCE

generated using the model presented. The sweep was

halted potentiostatically in the ilim region, and the
velocity (v) was changed from 1,000 rpm to 4,000
rpm and back. Cathodic ilim values were mildly sensi-
tive to flow (factor = 1.3). Pure diffusion control
would have lead to a factor of 2.6 for a fourfold in-
crease in the rotational speed for a rotating cylinder,
where the rate of mass transfer was proportional to
the speed raised to the power of 0.7, as given by the
correlation of Eisenberg, et al.19 Thus, the observed
ilim values were not controlled entirely by diffusion.
Since there were few H+ ions in the solution at
pH 5, the dominant cathodic reaction was H2CO3
reduction. As previously suggested, ilim for this reac-
FIGURE 10. Velocity effect in CO2 solution at pH 5, water + 3% NaCl,
tion was controlled by a slow hydration step, so no PCO2 = 1 bar (100 kPa), t = 20°C, and St52 steel.
flow sensitivity was expected. The small change with
rotation speed shown in Figure 10 could be attrib-
uted to the diffusion of present H+ ions or H2CO3
molecules from the bulk to the surface.

Effect of pH
Solution Without CO2 — The effect of pH is illus-
trated in Figure 11 for steel St52 in a solution purged
with N2 where the pH was adjusted by adding HCl.
The ilim values at 1,000 rpm were reduced proportion-
ately to the H+ concentration.
The position of the Tafel line for H2O reduction
stayed approximately the same over the whole pH
range, with a slope of ≈ 120 mV/decade. This was in
accordance with theory and agreed with the findings
of Gray, et al.13
This analysis showed that the Tafel line for
anodic dissolution of iron maintained the slope of FIGURE 11. pH effect in HCl solution, water + 3% NaCl, PN2 = 1 bar
(100 kPa), t = 20°C, 1,000 rpm, τ = 1.7 Pa, Re = 5,235, and St52 steel.
40 mV/decade over the whole pH range tested. The
exchange current density changed according to:

∂log i0(Fe) effect of changing the pH between 4 and 5 for solu-

=1 (1)
∂pH tions with and without CO2 were not possible as ilim
at pH 5 without CO2 (Figure 11) was very small and
between pH 3 and 4, while it changed very little indistinct.
between pH 4 and 5, which was in agreement with Figure 12 shows pH had a small effect on the
findings of Bockris, et al.20 anodic dissolution reaction for St52 steel. A similar
CO2 Solution — With CO2, the effect of changing conclusion was reached previously by Videm.21 The
the pH from 4 to 5 was to decrease ilim by a factor of absence of a strong pH dependence of the anodic
≈ 3 (Figure 12). This was much less than expected for dissolution of iron in CO2-containing environments
a tenfold decrease in H+ concentration if the cathodic was confirmed independently in another set of
reaction was solely H+ reduction. The reason for the experiments with low-carbon X-65 pipeline steel
threefold change was that the cathodic ilim in CO2 (Figure 13). The measured anodic sweeps are shown
solutions was composed of a chemical reaction- along with individual electrochemical reactions and
limited component and a H+-limiting component. The the theoretical sweep generated with the model. The
concentration of H2CO3 (and thus, the chemical reac- pH had no clear effect on anodic dissolution of iron
tion limited current) did not vary with pH, whereas in CO2 environments over the range of practical
ilim for H+ reduction was proportional to [H+]b. The interest 4 < pH < 6 (pH values outside this range
relationship between pH and ilim without CO2 when are less common for waters found in oil and gas
the cathodic reaction was only H+ reduction is shown production). In any case, the measured effect was
in Figure 11, where a change in pH from 3 to 4 re- far below the one seen previously in HCl solutions
sulted in a decrease by a factor of 10 as was expected (Figure 11) and expected from the mechanism
for a tenfold decrease in [H+]b. Comparisons of the of Bockris.20

CORROSION–Vol. 52, No. 4 285

CORROSION SCIENCE

FIGURE 13. pH effect on the iron dissolution reaction in CO2 solution,

FIGURE 12. pH effect in CO2 solution, water + 3% NaCl, PCO2 = 1 bar water + 3% NaCl, PCO2 = 1 bar (100 kPa), t = 20°C, and static using
(100 kPa), t = 20°C, 1,000 rpm, τ = 1.7 Pa, Re = 5,235, and St52 steel. X-65 steel.

to lower potentials with increasing rotational speed

(Figure 7), so that the performed potentiostatic mea-
surements slightly underestimated the true ilim. To
correct this at higher temperatures (50°C and 80°C),
a potentiodynamic sweep was done and then halted
at a potential clearly in the ilim region. The velocity
then was changed from 1,000 rpm to 7,000 rpm and
back. Agreement with the Eisenberg correlation was
much better. This also indicated that measurement
of the H+ diffusion ilim could be used in the future for
quick characterization of the mass-transfer behavior
of different hydrodynamic systems.
Measurements at pH 4 with CO2 — Subsequent
experiments were conducted where CO2 was bubbled
FIGURE 14. Temperature effect on ilim in HCl solution, pH 4, water + at different temperatures (20°C, 50°C, and 80°C)
1 % NaCl, PN2 = 1 bar (100 kPa), obtained using a RCE. while maintaining the atmospheric pressure in the
cells. However, because of the increasing water vapor
pressure at 50°C and 80°C, the partial pressure of
CO2 (PCO2) in the cells was 0.88 bar (88 kPa) and
Effect of Temperature 0.55 bar (55 kPa), respectively. Despite the lower PCO2
Measurements at pH 4 Without CO2 —␣ Tempera- at higher temperatures, the corrosion rate measured
ture was known to accelerate most of the chemical using Rp measurements increased from 1 mm/y
and electrochemical processes occurring in the (0.039 in./y) at 20°C to 2.5 mm/y (0.098 in./y) at
present system. To distinguish the effect on indi- 50°C and 3 mm/y (0.118 in./y) at 80°C. This was
vidual reactions, experiments at higher temperatures confirmed by looking at the potentiodynamic sweeps
were done first on X-65 steel in a solution purged conducted at pH 5 on static specimen made from
with N2 with pH adjusted by adding HCl. X-65 steel (Figure 15). The total cathodic reaction
The most pronounced effect was the increase in increased with increasing temperature. Since there
the mass transfer-controlled ilim which resulted from were very few H+ ions in the solution, the dominating
an increase in the diffusion coefficient (D) for H+ and cathodic reaction was probably H2CO3 reduction and
a decrease in viscosity at higher temperatures. Mea- direct H2O reduction at high overpotentials.
surements of the temperature dependence for ilim Surprisingly, the anodic reaction shown on
were done at pH 4 using the RCE (Figure 14). At Figure 15 was not accelerated significantly at higher
20°C, the rotation speed was varied while the poten- temperatures. Similar experiments were repeated
tial was held at –0.7 VSHE, which was in the ilim region. many times without obtaining a consistent depen-
When results were compared with the correlation of dence of the anodic dissolution on temperature. In
Eisenberg, et al., for mass transfer to a rotating cyl- independent potentiostatic measurements, an
inder, the agreement was only partially satisfactory.19 Arrhenius-type temperature dependence was ob-
It was suspected that the reason was the shift of ilim tained for the anodic dissolution current of X-65

286 CORROSION–APRIL 1996

 CORROSION SCIENCE

steel only for lower temperatures (20°C to 50°C). For

higher temperatures, the rate of anodic dissolution
of iron decreased, indicating a change in the rate-
determining step or a different reaction mechanism.
A more detailed study of this phenomenon will be
conducted in the future.

PHYSICAL MODEL
The experimental findings led to a mechanistic
model for CO2 corrosion.

Cathodic Reactions
When CO2 is added to an aqueous solution, it is
hydrated to form a weak acid (H2CO3): FIGURE 15. Temperature effect in CO2 solution, water + 1% NaCl,
PCO2 = 1 bar (100 kPa), pH 5, and static using X-65 steel.
CO2 + H2O ↔ H2CO3 (2)

Then, the H2CO3 is dissociated in two steps:

Anodic Reactions
H2CO3 ↔ H + HCO3
+ – In a water-CO2 corroding system, it most often
(3)
has been assumed that the sequence of anodic disso-
lution of iron is the same as it is in other acids,11-12
HCO3 ↔ H + CO3
– + 2–
(4)
as in the pH-dependent mechanism proposed by
Bockris, et al.:20
The reduction of H+:

Fe + OH → FeOH + e
– –
(8)
H +e → H
+ –
(5)

is the most important cathodic reaction in an acidic

→ FeOH+ + e–
RD
solution. It has been shown previously12,17 and con-
FeOH (9)
firmed in this project that, as the current density is
increased, the resistance to the mass transfer of H+,
FeOH → Fe + OH
+ 2+ –
(10)
which causes a drop in the concentration at the
metal surface, results in a deviation from Tafel be- However, the present findings, in agreement with
havior. Eventually, ilim is attained. At higher pH those of Videm,21 did not support this assumption
values, the availability of H+ decreases, and other (> pH 4). It is difficult to propose a new mechanism
competing cathodic reactions become important. for iron dissolution in CO2-containing environments
In CO2 systems at low pH (< 4), H+ reduction at this stage without a more detailed electrochemical
(Equation [5]) is still the dominant cathodic equation study, but it was assumed that the iron dissolution
because of the high concentration of H+. At interme- reaction in CO2-containing solutions proceeds with
diate pH (4 < pH < 6), which was the range of interest little influence of pH.
in this work, in addition to the H+ reduction (Equa-
tion [5]), a new cathodic reaction becomes important,
MATHEMATICAL MODEL
the direct reduction of H2CO3:
To build a predictive model of CO2 corrosion, all of
H2CO3 + e → H + HCO3
– –
(6) the previously described reactions must be quantified.

This additional cathodic reaction often is listed H+ Reduction

as the cause for the H2CO3 “to be more corrosive than For H+ reduction, the cathodic part of the general
a completely dissociated acid at the same pH.”11 The rate equation, is used:36
reaction has a limiting current that is controlled by a
slow chemical step, the hydration of CO2, as dis-
cussed in several other studies.15-16 This ilim is H+ s αcF
i H+ = i0 H+ × × exp – η (11)
insensitive to flow. At currents above ilim for H+ or H+ b RT
H2CO3 reduction, the dominant cathodic reaction
changes to direct reduction of water:18
Equation (11) takes into account the effect of
H2O + e → H + OH
– –
(7) resistance to charge transfer and mass transfer. The

CORROSION–Vol. 52, No. 4 287

CORROSION SCIENCE

surface concentration of H+ ([H+]s) can be determined where PH is the partial pressure of hydrogen.
2

from the mass-transfer equation: Limiting Current Density — The diffusion limiting
component of the total current density appearing in
Equation (13) is:
i H+ = kmF H+ b – H+ s (12)

idlim H+ = kmF × H+ b (19)

where km is the mass-transfer coefficient.
Solving Equations (11) and (12) for [H+]s gives,
km in Equation (19) can be calculated from a
after manipulation, the final current density-vs-
correlation (e.g., straight pipe correlation of Berger
voltage relationship for H+ reduction:22-23
and Hau):24

1 1 1
= + (13) kml
i H+ iα H+ idlim H+ Sh = = 0.0165 × Re0.86 × Sc0.33 (20)
D

where idlim(H+) is the diffusion limiting current density or from a rotating cylinder correlation of Eisenberg,
defined below (Equation [19]) and i␣(H+) is the activa- et al.:19
tion (charge transfer) current density in the absence
of resistance to mass transfer:
Sh = 0.0791 × Re0.7 × Sc0.356 (21)
η
–b
iα H+ = i0 × 10 c (14) or any other correlation for the given flow geometry.
H+
The temperature effect on D can be found from
the Stokes-Einstein equation:
Tafel Slope — The cathodic Tafel slope (bc) for H+
reduction appearing in Equation 14 is:
T µref
D = Dref × × (22)
2.303RT Tref µ
bc = (15)
αcF
For the reference temperature tref = 20°C, the
20
According to Bockris, et al., for H reduction, + dynamic viscosity of water is µ ref = 1.002 kg/(m s),25
␣c = 0.5 giving bc = 0.118 V at 25°C. This agreed well and the diffusion coefficient for H+ ions is Dref(H+) =
with the present findings. 9.31 x 10–9 m2/s.26
Exchange Current Density — From the present Water density as a function of temperature was
experiments at the reference temperature (tref ) = 20°C calculated as:13
at pH 4, it was determined that the exchange current
ρ = 1,152.3 – 0.5116 × T (23)
0(H+) ≈ 5 x 10
density for hydrogen ion reduction is iref –2

2
A/m . The pH dependence adopted from Bockris,
et al.:20 and water viscosity:25

∂logi0 H+ 1.3272 20 – t – 0.001053 20 – t

2
= –0.5 (16) µ = µref × 10 (24)
∂pH T + 105

agreed well with the experimental findings. H2CO3 Reduction

The temperature dependence of the exchange It was suggested previously that H2CO3 reduction
current density was modeled with an Arrhenius-type can be under activation control or under chemical
relation: reaction control for higher overvoltages. In a similar
∆H
derivation to the one shown for H+ reduction (Equa-
i0 1 1
tions [11] through [13]), the current-vs-voltage
= e– R T–T
ref (17)
iref
0 relationship is obtained:

The enthalpy of activation for the H+ reduction 1 1 1

reaction was found to be ∆H(H+) ≈ 30 kJ/mol. = + r
iH iα H CO i (25)
2CO3 2 3 lim H CO
Reversible Potential — The reversible potential for 2 3

H+ reduction (Erev[H+]) can be calculated as:

where irlim(H2CO3) is the chemical reaction limiting
2.303RT 2.303RT current density defined below (Equation [29]), and
Erev H+ = – pH – log PH (18) i␣(H2CO3) is the activation (charge transfer) current
F 2F 2

288 CORROSION–APRIL 1996

 CORROSION SCIENCE

density in the absence of chemical reaction resis- layer. Details on this correction will be published in
tance: the future.
The bulk concentration (b) of dissolved carbon
η
–b dioxide ([CO2]b) can be obtained from PCO2:
iα H = i0 H × 10 c (26)
2CO3 2CO3

CO2 = kdCO × PCO (30)

b 2 2
Tafel Slope — From the experiments, bc for
H2CO3 reduction in Equation (26) was found to be
The following equation was used to model the
similar to that for H+ reduction (≈ 120 mV/decade at
Henry’s constant (kdCO2) as a function of tempera-
20°C), which was in agreement with deWaard and
ture:29
Milliams.11
Exchange Current Density — For tref = 20°C, the
exchange current density for H2CO3 reduction in kdCO = 0.0454 1.6616 – 5.736 × 10–2t +
2
Equation (26) has a similar value as the one for H+
1.031× 10–3t2 – 9.68 × 10– 6t3 + 4.471 (31)
o(H2CO3) ≈ 6 x 10
reduction iref –2
A/m 2. Further, it can be
shown that i0(H2CO3) is proportional to the H2CO3 con- × 10– 8t4 – 7.912× 10– 11t5
centration and the H+ concentration:

– 0.5 The diffusion coefficient for H2CO3 in water

i0 H ∝ H2CO3 × H+ (27)
2CO3 (DH2CO3) is calculated from a reference value (1.3 x
10–9 m2/s at 25°C)12,28 and the Stokes-Einstein
Thus, the pH dependence of the exchange current Equation (22). The equilibrium constant for the CO2
density is: hydration reaction is Khyd = 2.58 x 10–3 and does not
change with temperature in the domain of interest.31
∂logi0(H The forward reaction rate for the CO2 hydration
2CO3) reaction (Kfhyd) is a function of temperature.30-31 An
= 0.5 (28)
∂pH attempt was made to use the equation proposed by
Stachewski,30 but low values for the hydration rate
The temperature dependence of the exchange were obtained for higher temperatures (50°C and
current density was modeled the same as for H+ 80°C). A more detailed search showed that rather
reduction according to Equation (17). The activation large variations on the value of kfhyd at high tempera-
enthalpy for this reaction was found to be similar to tures exist in the literature. Therefore, keeping the
the one for H+ reduction ∆H(H2CO3) ≈ 30 kJ/mol. same form of the equation, new coefficients were
Reversible Potential — The two reactions, the determined that gave a much better agreement
reduction of H2CO3 and H+, have the same reversible with the flow-independent, chemical reaction rate-
potential given by Equation (18). controlled limiting currents measured in CO2
Limiting Current Density — The CO2 hydration solutions at higher temperatures.
reaction limiting current density irlim(H2CO3) can be ob-
tained as:27-28 11,715
kfhyd = 10169.2 – 53.0 log T –
T (32)
0.5
irlim H2CO3 = F × CO2 × DH K kf (29) Water Reduction
b 2CO3 hyd hyd
Since water molecules are present in unlimited
quantities at the metal surface, it can be assumed
This equation first was derived by Vetter and is
that the reduction rate of H2O is controlled by the
strictly valid only for stagnant solutions.27 When it charge-transfer process and, hence, pure Tafel
was rederived from principles for flowing solutions, a behavior:
similar expression was obtained but with a multiplier
that accounted for the effect of flow. This multiplier η
–b
is significantly different from unity only when the iH = i0 H × 10 c (33)
thickness of the mass-transfer boundary layer is of 2O 2O

the same order of magnitude as the reaction layer. It

is difficult to give a general rule when this is the case Since the reduction of H2O and H+ are equivalent
since it is affected by many factors. As a first ap- thermodynamically, in the sense that they have the
proximation, it can be said that the Vetter’s Equation same Erev at a given pH, Erev for reduction of H2O is
(29) is correct at higher temperatures (T > 20°C) and given by the same Equation (18) as for H+ reduc-
low velocities (v ≤ 1 m/s [3.2 ft/s]) when the mass- tion.12,34 From the experiments at tref = 20°C, it was
transfer layer is much thicker than the reaction determined that the exchange current density for

CORROSION–Vol. 52, No. 4 289

CORROSION SCIENCE

H2O reduction is i0(H2O) ≈ 3 x 10–5 A/m2. No systematic As shown by West, the relationship between Erev
dependence of the exchange current density for this and i0 for the anodic dissolution of a metal is:36
reaction was found for 3 < pH < 6. The temperature
sensitivity was modeled according to Equation (17), i0
and a similar activation enthalpy was found as for Erev – Elrev = b log (37)
the previous two reactions: ∆H(H2O) ≈ 30 kJ/mol. The il0
Tafel slope for H2O reduction was found to be the
same as that for H+ reduction (≈ 120 mV/decade Thus, if the value of i0 is known at any concen-
at 20°C). tration of ferrous ions, from experimental data for a
given electrolyte, the Tafel line can be plotted which
Oxygen Reduction will apply regardless of the concentration of ferrous
Oxygen reduction was included in the model to ions in solution.
enable estimation of the effect of any oxygen pres- In the present experiments, no dependence of
ence in CO2 systems (laboratory or field) on the the iron dissolution rate on Fe2+ concentration was
corrosion rate. Since Erev for oxygen reduction is found, in agreement with the above. This finding was
much higher than for the other mentioned reactions, used in the model to construct the curve for Fe2+
pure mass-transfer control can be assumed for this dissolution. Since it was shown experimentally that
reaction in the electrode potential region of interest. the anodic line for both steels was not affected sig-
The diffusion limiting current density for oxygen nificantly by pH over the pH range 4 to 6, it was
reduction, O2 + 4H+ + 4e = 2H2O, is: sufficient to extract from the experiments the i0(Fe)
value for one reference potential, and the anodic
idlim(O ) = 4kmF × O2 (34) dissolution line was defined.
2 b
The activation enthalpy for this reaction was
found to be ∆H(Fe2+) ≈ 40 kJ/mol for lower tempera-
km can be determined from a correlation relating tures (20°C to 50°C). As previously mentioned, at
the Sherwood, Reynolds, and Schmidt numbers, as higher temperatures (> 50°C) this trend was reversed,
shown above for H+ reduction. For tref = 20°C, the so no firm conclusion on the effect of temperature on
diffusion coefficient for O2 is Dref(O2) = 2.09 x 10–9 m2/s.26 iron dissolution could be reached at the present
stage.
Anodic Dissolution of Iron The steel type had the largest effect on anodic
In the present experiments, the corrosion of two
dissolution. Very different i0(Fe) values were found
low-carbon steels was studied. For both steels, the
for the two steels tested (Figure 16). At a reference
anodic dissolution of iron at the corrosion potential
potential Erev = –0.488 V in the case of the St52
(and up to 200 mV above) was under activation con-
steel, the exchange current density was found to be
trol. Thus, pure Tafel behavior can be assumed close
i0(Fe) ≈ 0.1 A/m2, while for the X-65 steel, it was an
to the corrosion potential:
order of magnitude higher at i0(Fe) ≈ 1 A/m2. Such
η
variations in the values of i0 between different steels
i(Fe) = i0(Fe) × 10b a (35) was not unexpected.

Tafel Slope — The anodic Tafel slope (ba) for an- IMPLEMENTATION OF THE MODEL
odic iron dissolution is:
The present electrochemical model was imple-
mented in Microsoft Excel 5.0 for Windows† to exploit
2.303RT the user friendly interface and the advanced graphi-
ba = (36)
αaF cal capabilities. The built-in equation solver is
somewhat slow, so use of other programming lan-
According to Bockris, et al.,20 ␣a = 1.5 giving guages can be beneficial.
ba = 40 mV at 25°C, which agreed very well with the The model requires as input: temperature, pH,
experimental findings. PCO2, oxygen concentration, type of steel, and the
Exchange Current Density — The concentration flow geometry. Pipe flow or a rotating cylinder flow
of ferrous ions in solution does not affect the can be selected. For pipe flow, the inputs are water
dissolution kinetics of iron in the absence of film velocity and pipe diameter, while for rotating cylinder
formation.32,35 The Tafel line for the anodic dissolu- flow, the inputs are the rotating speed and cylinder
tion is not affected by the rise in the concentration of diameter.
ferrous ions required for their transport away from Once the input parameters are determined, the
the anode.32 To construct the anodic dissolution po- program generates an updated graph with the indi-
larization line for iron, the exchange current density vidual and total cathodic and anodic curves. The
is required along with ba. intersection of the total cathodic curve with the

290 CORROSION–APRIL 1996

 CORROSION SCIENCE

anodic curve gives the corrosion potential (Ecorr) by

solving:

i H+ + i H +iH + i O = i Fe (38)
2CO3 2O 2

icorr is calculated from the anodic curve Equation

(35) and the known Ecorr. It is possible to display the
theoretical polarization curve (sweep) generated by
Equation (38) or, by using the same equation, the
theoretical E-vs-i curve for small overpotentials often
used to determine Rp can be displayed. The slope
dE/di at the corrosion potential can be calculated by
numerically differentiating Equation (38). This infor-
mation can be used for calculating the theoretical B
FIGURE 16. Comparison of the anodic dissolution lines for the X-65
value since icorr is known: and St52 steels in CO2 solution, water + 1 % NaCl, t = 20°C, PCO2 =
1 bar (100 kPa), pH 5, and static.
dE
B= × icorr (39)
di E corr

The model enables numerous other possibilities

including: combinations of individual and total
polarization curves, overlaying theoretical and experi-
mental sweeps, and Rp plots.

MODEL VALIDATION
Performance of the model was validated by com-
paring the predictions with results from independent
glass loop experiments. Experimental data were ob-
tained for flow through a straight pipe, dpipe = 15 mm
(0.59 in.) at v = 2 m/s (6.56 ft/s) and PCO2 = 1 bar
(100 kPa) with no protective films present. A more
detailed description of these experiments has been
given elsewhere.33 Comparisons for t = 20°C, 50°C, FIGURE 17. Comparison of the model predictions with results from
and 80°C and pH 4, 5, and 6 are shown in Figure 17. independent glass loop experiments with X-65 steel in water + 1 %
Reasonably good agreement was obtained over the NaCl at v = 2 m/s (6.56 ft/s), τ = 12 Pa, PCO2 = 1 bar (100 kPa), and
whole range. In some experiments (which typically Re = 30,000.
lasted a few days), the corrosion rate varied as films
formed at higher temperatures.33
Subsequently, predictions made with the present made with the models of deWaard and Lotz and
model were compared with the experimentally based Dugstad, et al., except at pH < 3.5. In this range, the
models of deWaard and Lotz1 and with the model of H+ reduction becomes the dominant reaction as pre-
Dugstad, et al.2 The latter two models are semi- viously shown. The models of deWaard and Lotz and
empirical, and each is calibrated with a different set Dugstad, et al., are not valid in this domain.
of experimental data with somewhat different steels.
Predictions with the present model were made by Effect of Flow Rate
using i0(Fe) = 0.1 A/m2 determined for the St52 steel A comparison of the flow effect predictions made
used in the present experiments. with different models is presented in Figure 19 for pH
4, t = 20°C, PCO2 = 1 bar (100 kPa). Agreement of the
pH Effect present model with the deWaard and Lotz model was
The predictions of the pH effect on the CO2 nearly complete except at very low velocities. This
corrosion were made for the case of a solution satu- was a result of the velocity corrections implemented
rated with 1 bar (100 kPa) CO2, t = 20°C, v = 1 m/s in the latest version of the deWaard and Lotz model,
(3.2 ft/s), dpipe = 25 mm (0.98 in.) giving Re = 25,000. through the so-called “resistance model,” which re-
A comparison of the predictions made with different sembles in form the present Equations (13) and (25).
models is presented in Figure 18. The present predic- However, the deWaard and Lotz model lacks the
tions were somewhere in between the predictions physical background presented here, which is com-

CORROSION–Vol. 52, No. 4 291

CORROSION SCIENCE

FIGURE 18. Comparison of predicted pH effect on the corrosion rate

for the case of a solution saturated with PCO2 = 1 bar (100 kPa), t = FIGURE 19. Comparison of predicted velocity effects on the corrosion
20°C, v = 1 m/s (3.2 ft/s), τ = 3.1 Pa, dpipe = 25 mm (0.98 in.), and Re rate for the case of a solution saturated with PCO2 = 1 bar (100 kPa),
= 25,000. pH 4, t = 20°C, and dpipe = 25 mm (0.98 in.).

Figure 20 for pH 5, v = 1 m/s (3.2 ft/s), PCO2 = 1 bar

(100 kPa). From 20°C to 40°C, predictions of the
present model were close to the predictions of the
deWaard and Lotz model. For temperatures > 40°C,
the discrepancy increased. The Dugstad, et al., model
predicted higher corrosion rates than other two mod-
els. The temperature dependence curve produced
with the Dugstad model had a different character as
it included a decrease of the corrosion rates at higher
temperatures due to protective film formation. The
same correction factor present in the deWaard and
Lotz model was responsible for the difference be-
tween the present predictions and the deWaard and
Lotz model at higher temperatures.
FIGURE 20. Comparison of predicted temperature effects on the
corrosion rate for the case of a solution saturated with PCO2 = 1 bar Effect of PCO2
(100 kPa), v = 1 m/s (3.2 ft/s), τ = 3.1 Pa, pH 5, dpipe = 25 mm The effect of PCO2 is demonstrated by comparing
(0.98 in.), and Re = 25,000. the three models for pH 4, t = 20°C, v = 1 m/s
(3.2 ft/s). All three models have a similar power law
dependence between 1 bar (100 kPa) and 10 bar
pensated by a calibration constant in their resistance (1,000 kPa) CO2 with the exponent close to 0.7
model. The Dugstad, et al., model showed a similar (Figure 21). The main difference is at PCO2 < 1 bar
flow dependence as the other two models, but at a (100 kPa), where the present model did not predict
higher level. It was shown later that the reason is any correlation (because H+ reduction dominated)
probably that different steels were used in different while the models of deWaard and Lotz and Dugstad,
studies. It was demonstrated in the present study et al., extrapolated the same power law dependence
that very different rates of anodic dissolution were into this region. It was interesting that the present
obtained with apparently not-so-different steels. With model reproduced the experimentally observed 0.7
a choice of a different i0(Fe) in the present model that power law dependence directly from the fundamental
corresponded better to the steel used in the study of electrokinetic equations.
Dugstad, et al., a much better agreement between
the two models could be achieved. CONCLUSIONS
Effect of Temperature ❖ In the HCl solutions purged with N2 at pH 3 and 4,
The predicted temperature dependence of the the dominant cathodic reaction was H+ reduction
corrosion rate made with the present model is com- except at very low rotation speeds and static condi-
pared with the predictions of the other two models in tions, when it changed to H2O reduction. Tafel

292 CORROSION–APRIL 1996

 CORROSION SCIENCE

behavior for H+ reduction could be observed only at

high rotation speed (slope 120 mV/decade). Diffusion
limiting currents were observed and could be pre-
dicted by using mass-transfer correlations such as
the correlation of Eisenberg, et al.,19 for a rotating
cylinder. At currents beyond the limiting current and
especially for pH 5 and higher, direct reduction of
H2O became the dominant cathodic reaction, which
was under activation control (slope 120 mV/decade).
Anodic dissolution of iron followed Tafel behavior for
small overpotentials (slope 40 mV/decade) and was
not flow sensitive. The pH dependence agreed well
with the mechanism of Bockris, et al.,20 up to pH 4.
The corrosion rate at 20°C and pH 4 was under mixed
control (activation-diffusion) and was flow sensitive.
FIGURE 21. Comparison of predicted PCO2 effect on the corrosion
For high rotation speeds, it became activation con-
rate for the case of a solution at pH 4, t = 20°C, v = 1 m/s (3.2 ft/s),
trolled. At pH 5, the corrosion rate was controlled by τ = 3.1 Pa, dpipe = 25 mm (0.98 in.), and Re = 25,000.
direct reduction of H2O and was not flow sensitive.
❖ In CO2 solutions, it was confirmed that there was
an additional cathodic reaction: direct reduction of
H2CO3. At pH 4, both H+ and H2CO3 reduction were of CO2 corrosion prediction models. The present model
similar magnitude at low rotation speeds. For higher performed similarly to the latest version of the model
rotation speeds, reduction of H+ dominated. At pH 5, of deWaard and Lotz. This was not surprising since
reduction of H2CO3 was the dominant cathodic reac- the main operating assumption of the original
tion. Limiting currents for this reaction were found to deWaard and Milliams model, charge-transfer con-
be under chemical reaction-control and nearly flow trolled H2CO3 reduction, also is included in the
insensitive. Reduction of H2O was not affected by the present model. The so-called “resistance model” built
presence of CO2. This reaction was under activation into the latest deWaard and Lotz model resembles in
control and was not flow or pH sensitive. The pH had form some of the present theoretical equations. The
very little effect on anodic dissolution of iron in CO2 predictions presented here were quite different from
environments over the range of practical interest the ones made with the empirical model of Dugstad,
4 < pH < 6. icorr at 20°C and pH 4 was under mixed et al.2 Some latest simulations have shown that their
control (activation-diffusion chemical reaction) and results also can be simulated by selecting the anodic
flow sensitive. At pH 5, icorr was under mixed control exchange current density appropriate for the steel
(activation-chemical reaction) and flow insensitive. used in their experiments, which was unknown pre-
❖ The model of electrochemical reactions occurring viously.
in a water-CO2 system takes into account the follow- ❖ Compared to all previous models, the present
ing cathodic reactions: H+ reduction, H2CO3 reduc- theoretical model gives a much clearer picture of the
tion, direct water reduction, and oxygen reduction. corrosion mechanisms and of the effect of key pa-
One anodic reaction was accounted for: iron dissolu- rameters. Most of the important constants in the
tion. The data generated by the model can be used to: present model, which were determined experimen-
— Predict the corrosion rate; tally, are physically meaningful. Being theoretical,
— Determine the governing corrosion mechanism the model is open for inclusion of new concepts such
(activation, diffusion, reaction, or mixed control); as a more detailed treatment of the transport pro-
— Conduct sensitivity tests with respect to the cess, which could lead to predictions of protective
input parameters of temperature, pH, velocity, and film formation or inclusion of an inhibitor effect in a
PCO2; physically correct way.
— Help understand and explain data from elec-
trochemical measurements such as potentiodynamic ACKNOWLEDGMENTS
sweeps and Rp measurements; and
— Obtain the theoretical B value in the case of This work was supported financially by Statoil
multiple cathodic reactions. and the Norwegian Council for Scientific Research.
Performance of the model was validated by com-
paring the predictions with results from independent NOMENCLATURE
loop experiments. Good agreement was obtained for
t = 20°C, 50°C, and 80°C and pH 4, 5, and 6. ba, bc anodic and cathodic Tafel slope (V/decade)
❖ Predictions made with the present model were [CO2]b bulk concentration of dissolved carbon
compared with performance of the other best-known dioxide (mol/m3)

CORROSION–Vol. 52, No. 4 293

CORROSION SCIENCE

D diffusion coefficient (m2/s) 8. K.R. Trethewey, J.S. Keenan, “Microcomputer-Based Corrosion

Modeling Applied to Polarization Curves,” in Computer Model-
Erev reversible potential (V) ling in Corrosion, ed. R.S. Munn, ASTM STP 1,154 (West
F Faraday constant (96,490 coul/equiv.) Conshohocken, PA: ASTM, 1992), p. 113.
[H+]s, [H+]b surface and bulk H+ concentration (mol/m3) 9. J.M. Costa, M. Vilarassa, “Microcomputer Programs for Corro-
sion Education,” in Proc. U.K. Corrosion’88, vol. 3 (1988): p. 99.
i current density (A/m2) 10. M.R. Bonis, J.L. Crolet, “Basics of the Prediction of the Risks of
ia activation component of the total current CO2 Corrosion in Oil and Gas Wells,” CORROSION/89, paper
no. 466 (Houston, TX: NACE, 1989).
density (A/m2) 11. C. deWaard, D.E. Milliams, Corrosion 31 (1975): p. 131.
i0 exchange current density (A/m2) 12. L.G.S. Gray, B.G. Anderson, M.J. Danysh, P.G. Tremaine,
d
i lim diffusion limiting current density (A/m2) “Mechanism of Carbon Steel Corrosion in Brines Containing
r
Dissolved Carbon Dioxide at pH 4,” CORROSION/89, paper no.
i lim chemical reaction limiting current den- 464 (Houston, TX: NACE, 1989).
sity (A/m2) 13. L.G.S. Gray, B.G. Anderson, M.J. Danysh, P.R. Tremaine,
“Effect of pH and Temperature on the Mechanism of Carbon
km mass-transfer coefficient (m/s)
Steel Corrosion by Aqueous Carbon Dioxide,” CORROSION/90,
kfhyd forward reaction rate for the CO2 hydra- paper no. 40 (Houston, TX: NACE, 1990).
tion reaction (l/s) 14. S. Nesic, “Prediction of Transport Processes in CO2 Corrosion,”
Progress in the Understanding and Prevention of Corrosion,
kdCO2 Henry’s constant, (mol/[m3 bar]) Proc. 10th Europ. Corros. Cong., vol. 1 (London, England:
Khyd equilibrium constant for the CO2 hydra- Institute of Metals, 1993), p. 539.
tion reaction 15. G. Schmitt, B. Rothman, Werkst. Korros. 28 (1977): p. 816.
16. E. Eriksrud, T. Søntvedt, “Effect of Flow on CO2 Corrosion
l characteristic length, pipe or cylinder Rates in Real and Synthetic Formation Waters,” in Advances in
diameter (m) CO2 Corrosion, vol. 1. Proc. CORROSION/83 Symp. CO2 Corro-
sion in the Oil and Gas Industry, eds. R.H. Hausler, H.P.
[O2]b bulk concentration of oxygen (mol/m3) Goddard (Houston, TX: NACE, 1984), p. 20.
PCO2 partial pressure of CO2 gas (bar) 17. M. Stern, J. Electrochem. Soc. 102 (1955): p. 609.
P H2 partial pressure of hydrogen gas, bar 18. P. Delahay, J. Amer. Chem. Soc. 74 (1952): p. 3,497.
19. M. Eisenberg, C.W. Tobias, C.R. Wilke, J. Electrochem. Soc.
R universal gas constant (8.3143 J/(mol K) 101 (1954): p. 306.
Re = ␳␯l/µ Reynolds number 20. J.O’M. Bockris, D. Drazic, A.R. Despic, Electrochim. Acta 4
Sc = µ /␳D Schmidt number (1961): p. 325.
21. K. Videm, “Fundamental Studies aimed at Improving Models for
Sh = kml/D Sherwood number Prediction of CO2 Corrosion,” in Progress in the Understanding
t temperature (°C) and Prevention of Corrosion, Proc. 10th Europ. Corros. Cong.,
vol. 1 (London, England: Institute of Metals, 1993), p. 513.
T absolute temperature (K) 22. V.S. Bagotzky, Fundamentals of Electrochemistry (New York,
v velocity (m/s) NY: Plenum Press, 1993), p. 141.
␣c , ␣a apparent transfer coefficients 23. A.J. Bard, L.R. Faulkner, “Electrochemical Methods (New York,
NY: John Wiley and Sons, 1980), p. 110.
∆H enthalpy of activation (J/mol) 24. F.P. Berger, K.-F. F.-L Hau, Int. J. Heat Mass Trans. 20 (1977):
␩ = E – Erev overpotential (V) p. 1,185.
µ viscosity (kg/m-s) 25. Handbook of Chemistry and Physics, 66th ed. (Boca Raton, Fl:
CRC Press Inc., 1985), p. F-37.
␳ density (kg/m3) 26. P.W. Atkins, Physical Chemistry, 2nd ed. (Oxford, England:
Oxford University Press, 1982), p. 905.
REFERENCES 27. K.J. Vetter, Electrochemical Kinetics, Theoretical Aspects,
Sections 1, 2, and 3 of Electrochemical Kinetics: Theoretical
and Experimental Aspects (translation from German) (New
1. C. deWaard, U. Lotz, “Prediction of CO2 Corrosion of Carbon York, NY: Academic Press, 1967), p. 235.
Steel,” CORROSION/93, paper no. 69 (Houston, TX: NACE, 28. T. Hurlen, S. Gunvaldsen, R. Tunold, F. Blaker, P.G. Lunde, J.
1993). Electroanal. Chem. 180 (1984): p. 511.
2. A. Dugstad, L. Lunde, K. Videm, “Parametric Study of CO2 29. IUPAC, Chemical Data Series, No. 21, “Stability Constants for
Corrosion of Carbon Steel,” CORROSION/94, paper no. 14 Metal Ion Complexes, Part A: Inorganic Ligands” (Elmsford, NY:
(Houston, TX: NACE, 1994). Pergamon Press, 1982).
3. E.L. Pye, M.L. Pye, F.H. Brock, “Pol-Plot: Basic Electrode Soft- 30. D. Stachewski, Chemie-Ing.-Techn. 41 (1969): p. 1,111.
ware, Simple to Complex,” CORROSION/88, paper no. 102 31. D.A. Palmer, R.V. Eldik, Chem. Rev. 83 (1983): p. 651.
(Houston, TX: NACE, 1988). 32. J. Postlethwaite, Electrochim. Acta 15 (1970): p. 1,847.
4. G. Rochini, “Extension of the Stern and Geary’s Method 33. S. Nesic, G.T. Solvi, J. Enerhaug, “Comparison of the Rotating
for Calculating of Electrochemical Parameters with Inter1 Cylinder and Pipe Flow Tests for Flow Sensitive CO2 Corrosion,”
Code,” CORROSION/88, paper no. 103 (Houston, TX: NACE, CORROSION/95, paper no. 130 (Houston, TX: NACE, 1995).
1988). 34. H. Kaesche, Metallic Corrosion (Houston, TX: NACE, 1985), p. 113.
5. G. Rochini, “Determination of the Validity Interval of Linear 35. W. Lorenz, K. Heusler, “Anodic Dissolution of Iron Group Met-
Approximation Method with Inter1 Code,” CORROSION/88, als,” in Corrosion Mechanisms, ed. F. Mansfeld (New York, NY:
paper no. 105 (Houston, TX: NACE, 1988). Marcel Dekker, 1987).
6. J.G. Hines, Brit. Corros. J. 18 (1983): p. 10. 36. J.M. West, Electrodeposition and Corrosion Processes (Van
7. F. Mansfeld, Corrosion 29 (1983): p. 397. Nostrand, 1964), p. 36.

294 CORROSION–APRIL 1996