Effect of Salt Concentration on the Corrosion Behavior of Carbon Steel in CO₂ Environment

The accelerated corrosion of low alloy steel caused by dissolved carbon dioxide (CO₂) has been recognized as a serious problem in the oil and gas industry. It has a significant but difficult to quantify impact on the cost of crude oil and gas production.^1-5  The failure of equipment caused by corrosion often results in great economic loss and catastrophic accidents, which has attracted great consideration.^2,6-8  Like many other forms of corrosion, CO₂ corrosion is a complex phenomenon and can be affected by numerous factors. Literature reviews show that temperature,^1,4,9-10  pH,^8-9,11  partial pressure,^5,10,12  and flow rate^8,13-14  all have significant effects on the CO₂ corrosion process. However, corrosion mechanism in this particular environment is still not well understood.

Kinetics of CO₂ corrosion has been studied using various electrochemical techniques. For example, Ortega-Toledo, et al., studied the Tafel polarization behavior of X80 pipeline steel in 3 wt% NaCl saturated with CO₂ at 50°C.¹⁵  The vertical line for cathodic branch indicated that the corrosion process was under diffusion control. Oxidation peak was observed in the anodic branch. Jawich and his coworkers studied the corrosion rate of C1018 mild steel (UNS G10180)⁽¹⁾ cylindrical electrode in 3 wt% NaCl with CO₂-saturated solution at 40°C by Tafel extrapolation method.¹⁶  Although the cathodic branch results showed that the cathodic reaction was also under diffusion control, no oxidation peak could be found in the anodic branch. Nam’s group investigated the corrosion kinetics of mild steel in 0.01 M (0.06 wt%) NaCl with CO₂-saturated solution at room temperature by potentiodynamic polarization and determined the Tafel slope b_a and b_c of 110 mV/decade and 169 mV/decade in this dilute salt solution, respectively.¹⁷  Liu, et al., performed potentiodynamic polarization tests of Q235 steel exposed in saltwater (about 1 wt% NaCl) saturated with CO₂ at different temperatures.¹  Their results showed that the anodic Tafel slope (b_a) increased, while the cathodic Tafel slope (b_c) decreased with temperature, which indicates that kinetics changed with temperature, and temperature had different effects on cathodic and anodic processes. However, no explanation was given for this temperature dependence.

In recent years, salt concentration effect on CO₂ corrosion of mild steel has also been studied. Nešić’s group focused on the salt concentration effect of NaCl (from 3 wt% to 25 wt%) on CO₂ corrosion mechanisms at 20°C.¹³  They found that CO₂ corrosion rate of carbon steel significantly decreased with increasing NaCl concentration. Han, et al., also reported decreased corrosion rates when salt concentration increased from 0.5 wt% to 20 wt%.¹⁸  They proposed a model showing that cathodic reactions were retarded by sodium chloride but anodic reactions were not significantly affected. Contrary to Han’s model,¹⁸  the potentiodynamic sweep analysis by Nešić’s group showed that both cathodic and anodic processes were retarded by increasing the salt concentration.¹³ 

Corrosion scale is also known to play an important role in CO₂ corrosion. It is generally accepted that the corrosion scale formed at low temperature (below ~60°C) offers minimal protection, with Fe₃C (cementite) predominating the film.^19-21  Jasinski found that the corrosion product at room temperature was Fe₃C, but both FeCO₃ and Fe₃C formed at higher temperature (95°C).⁵  Kinsella, et al., studied the corrosion scale formed on mild steel electrode exposed in 3% NaCl with CO₂ saturation at 30°C.¹⁰  The x-ray diffraction and Fourier transform infrared spectroscopy analyses indicated the chemical compositions consisted of FeCO₃, Fe₃C, and iron oxides. Moreover, it was found that the corrosion rate of carbon steel may be accelerated by the formation of corrosion scale. Dugstad, et al., investigated the mechanism of the iron carbonate film formation and found that galvanic contact between Fe₃C and iron will accelerate iron dissolution by promoting cathodic reaction, as Fe₃C has a much lower overpotential for the cathodic reaction.^22-23  Crolet and his coworkers also found that Fe₃C can act as sites for cathodic reactions.^24-25  The Fe₃C layer separated the anodic and cathodic reactions, and acidification occurred at the anodic sites, which promoted further metal dissolution. Therefore, the presence of Fe₃C could potentially increase the rate of attack by galvanic coupling and local acidification. However, the structure and composition of the corrosion scale from CO₂ corrosion and its effect on further corrosion are still controversial and the effect of chloride on the scale formation is not elucidated.

In this research, the effect of NaCl concentration on CO₂ corrosion was examined in a wide range from 0.001 wt% to 10 wt%. Open circuit potentials (OCP) were continuously monitored and instantaneous corrosion rates under freely corroding condition were measured by both linear polarization resistance (LPR) and electrochemical impedance spectroscopy (EIS) techniques. Structure and composition of the corrosion scales after exposure were examined by 3D optical microscope and scanning electron microscope (SEM) equipped with energy dispersive x-ray spectroscopy (EDX). Anodic and cathodic kinetics were studied by potentiodynamic polarization and potentiostatic hold using microelectrodes. Finally, a two-stage corrosion scale formation process was proposed.

Two different grades of carbon steels, C1010 (UNS G10100) and C1018, were utilized depending on their availability. The nominal chemical composition for both steels is listed in Table 1. Microelectrode technique, which has been successfully applied to study corrosion in low conductivity electrolytes such as drinking water,^26-27  biofuel,^28-29  and microbial ecology,³⁰  was used to minimize IR drop in dilute solution.^31-32  Wire electrodes of alloy C1010 (diameter of 500 μm) were used in the potentiodynamic and potentiostatic polarization experiments. The polyimide-coated wires were embedded in two-component epoxy resin and mounted in a PVC holder. The exposed surface area of the wire electrode was approximately 0.002 cm².

Large rectangular electrodes with dimensions of 45 mm × 25 mm × 6 mm (2 in × 1 in × ¼ in) were cut from a bar of C1018 carbon steel and tested under naturally corroding condition. Each electrode was drilled, tapped, and mounted with a PTFE gasket on an electrode holder as described elsewhere.³³  The exposed surface area of the rectangular electrode was approximately 30 cm².

Before testing, all electrodes were wet-polished with silicon carbide (SiC) abrasive paper up to 600 grit, rinsed with acetone, isopropanol, and deionized (DI) water, and Ar dried.

The test solutions were made up from analytical grade sodium chloride and DI water (18.2 MΩ·cm in resistivity) with five concentrations: 0.001, 0.01, 0.1, 1, and 10 wt%. The test solution was purged with research grade CO₂ for at least 2 h in the pre-conditioning vessel before being transferred to the test cell, and the test system was continuously purged with CO₂ at a lower flow rate to ensure saturation throughout the test.

Electrochemical experiments were conducted in either a typical flat cell or ASTM G5 polarization cell (Greene cell).³³  The working electrode was hung approximately in the center of the cell. Counter electrodes were made of platinized niobium and potential was measured with reference to saturated calomel electrode (SCE). To minimize the ohmic potential drop, the reference electrode was connected to the cell externally through a Luggin capillary and positioned by the working electrode as closely as possible (~1 mm).

The following electrochemical methods were performed:

• Tafel polarization experiments were conducted using the C1010 wire electrodes in the flat cell. Before polarization, OCP was measured for 2 h. Then, Tafel polarization curves were recorded at a constant sweep rate of 0.167 mV/s and the scanning range was from −300 mV to 300 mV with respect to OCP.

• Potentiodynamic polarization (anodic and cathodic) experiments were performed in the flat cell setup using the wire electrodes. Potentiodynamic scans were conducted after 2 h OCP delay. A typical anodic scan started at −30 mV_OCP to 1.0 V_SCE, while the cathodic scan started at +30 mV_OCP to −1.0 V_SCE. The scan rate was 0.167 mV/s.

• Potentiostatic hold tests were performed by applying selected constant potentials (−0.2 V_SCE and −0.8 V_SCE) to the wire electrodes for approximately 2 h when steady state current densities were reached.

• LPR measurements were conducted by polarizing the electrode from −10 mV to 10 mV with respect to OCP at a scan rate of 0.125 mV/s.

• EIS measurements were performed at OCP with sinusoidal potential excitation of 10 mV amplitude. The frequency range was from 10 kHz to 10 mHz.

Long-term immersion tests were conducted on the rectangular electrodes in different concentration NaCl solutions (0.001 wt% to 10 wt%) using the ASTM G5 polarization cell. OCP (~45 min), LPR (~3 min), OCP (~2 min), and EIS (~10 min) were performed in sequence every hour, and the total duration for each test was 100 h. Evolution of the corroding surface was recorded periodically with a digital recorder for selected tests. After 100 h immersion, the electrode was taken out of the test cell, and loose and bulky corrosion products were removed by brushing with distilled water and ultrasonic cleaning with acetone. The electrode was then blown dry with Ar.

Instantaneous corrosion current density was determined by the Stern-Geary equation:^13,18,34 

where b_a and b_c are absolute values of the anodic and cathodic Tafel slopes, respectively. Polarization resistance, R_p (Ω·cm²), was determined from both LPR and EIS measurements for every hour during the immersion test.

The instantaneous corrosion current densities (mA/cm²) were converted to corrosion rates in mm/y by Equation (2):³⁵ 

where K₁ = 3.27 mm·g/mA·cm·y, ρ is the density of carbon steel (7.85 g/cm³), and EW is the equivalent weight (27.92 g/eq assuming n = 2 for iron oxidation).

The average corrosion rate over 100 h (Δt) exposure was calculated by Equation (3):

All electrochemical measurements were made using a Solartron Modulab^† and were performed at ambient temperature (~20°C) and atmospheric pressure. Every test was repeated at least once to ensure reproducibility.

After long-term immersion, the rectangular electrodes were immediately rinsed with acetone and examined by a HITACHI TM3000^† tabletop SEM equipped with EDX. After final cleaning, both exposed surface and cross section were again examined using SEM/EDX. For cross-section examination, the rectangular electrodes were mounted in epoxy resin and polished with SiC abrasive paper up to 1200 grit to expose the cross section.

Morphology of the corroded surfaces was also examined by 3D optical microscope (ContourGT^† by Bruker).

The long-term immersion test solutions were analyzed for total iron concentration using an inductively coupled plasma-optical emission spectrometer (ICP-OES, Perkin Elmer model P-400^†).

Before corrosion rates were calculated through polarization resistance determined by either LPR or EIS, Tafel polarization was performed to determine b_a, b_c, and therefore B values (Equation [1]) in CO₂-saturated NaCl solutions. The Tafel polarization curves in three salt concentrations are shown as examples in Figure 1: 0.001%, 0.01%, and 0.1%. Figure 1(a) shows the original Tafel polarization curve in 0.001% NaCl test solution, as well as the IR corrected curve. This comparison suggests that ohmic potential drop was minimized by using 500 μm diameter wire electrode in 0.001% NaCl solution, and hence IR drop at higher salt concentration can be neglected. Anodic and cathodic slopes were determined by the Tafel extrapolation method. The slopes of the dashed lines, as marked in Figure 1, represent the corresponding slopes in the anodic and cathodic branches. The anodic slopes (b_a) decrease from 148 mV/decade to 65 mV/decade as NaCl concentrations increase from 0.001% to 10%. For the cathodic branches, one linear region was identified from −0.75 V_SCE to −0.95 V_SCE for solutions containing 0.1% NaCl (Figure 1[c]) or higher. The slope was determined to be more than 1,000 mV/decade, which agrees with previous research findings.^16,36  This suggests that the cathodic reaction is under diffusion control when NaCl concentration is higher than 0.1%. In comparison, two linear regions can be observed from the cathodic branches in 0.001% and 0.01% NaCl solutions. The linear region near OCP has a slope that decreases with decreasing salt concentration (Figures 1[a] and [b]), which may suggest the influence of charge transfer controlled process at the highest dissolved CO₂ concentration as discussed later. Two limits for the cathodic Tafel slope were considered here: b_c = ∞ for diffusion control and b_c = 120 mV/decade for activation control. The corrosion rate calculated by these two approximations can be used as lower and upper limits of the true value when accurate determination of B is unavailable for mixed control.³⁵  When the cathodic reaction is under complete diffusion control, Stern-Geary equation (Equation [1]) can be simplified as (⁠⁠). The anodic Tafel slopes and calculated B values for all salinities are summarized in Table 2. The B values exhibit a decreasing trend for both cases and for diffusion control, B value decreases from 64 mV to 26 mV with increasing NaCl concentration from 0.001% to 10%. These B values of ~20 mV were found to be close to previous studies.^1,15-17  In comparison, the B values for diffusion control case were approximately 1.4 to 2.2 times higher than their counterparts for activation control case. This will set the upper and lower limits for the corrosion rate calculation later.

To study the effect of salt concentration on the corrosion rate of carbon steel with CO₂ saturation, long-term immersion tests were performed using large rectangular C1018 electrodes. Electrochemical techniques, including OCP, LPR, and EIS, were performed sequentially to determine both corrosion potentials and instantaneous corrosion rates. SEM/EDX was used to examine corrosion morphology and identify corrosion products. Total dissolved iron concentrations were analyzed by ICP.

Figure 2 shows the OCP vs. time trend of C1018 electrodes in different concentration NaCl solutions with CO₂ saturation for 100 h. It is clear that the OCP values were more noble but more fluctuating at lower salt concentrations (i.e., 0.001% and 0.01%), probably indicating less stable interface formed in dilute solutions.³⁷  At 0.001%, the OCP was initially at −0.665 V_SCE and gradually decreased to a steady state value of ~0.690 V_SCE at the end of 100 h immersion (25 mV potential changed). In comparison, the OCP change in 10% NaCl solution was only 12 mV (from −0.703 V_SCE to −0.715 V_SCE) over 100 h immersion. In general, all of the OCP values were close to each other, around −0.7 V_SCE, which is in excellent agreement with the literature.^6,15,38 

Figures 3(a) and (b) show examples of LPR curves obtained at various time of C1018 electrodes tested in 0.001% and 1% NaCl with CO₂-saturated solutions, respectively. The good linear relationship between overpotential and current density can be observed and the apparent polarization resistance (R_p^′) was determined as the slope of these curves. It is obvious that the LPR curves of both samples initially show steep slopes, and gradually decrease with immersion time. However, this apparent polarization resistance includes a contribution from solution resistance (R_s), which can cause an underestimation of corrosion rates, especially in dilute solution with higher solution resistance.^39-40  The true polarization resistance (⁠⁠) can be calculated by:

where R_s can be determined by EIS measurements.

EIS technique was used to determine both polarization resistance and solution resistance. Figure 4 shows examples of Nyquist and Bode plots for C1018 electrodes in different concentration NaCl solutions (0.001% and 1%) saturated with CO₂. As shown in Figures 4(a) and (c), a similar feature of Nyquist plots can be observed, i.e., one depressed capacitive semicircle at high frequency and one inductive semicircle at low frequency. The capacitive semicircle at high frequency is attributed to the double layer capacitance and the charge transfer resistance, while the inductive semicircle at low frequency is related to the adsorbed intermediate product formed during the corrosion process.^3-4,41  It has been reported that the inductive loops in EIS curves can be attributed to the adsorption of intermediate product, Fe(I)_ad species caused by chloride accelerated mechanism.^42-43  However, this inductive loop was not observed when the test solution was purged with air (not shown here). The corresponding Bode plots (Figures 4[b] and [d]) show that there are two time constants, i.e., a high-frequency peak and a low-frequency valley. The peak with a maximum phase angle increases and shifts to low frequencies with increasing immersion time, which is usually associated with an increase of double layer capacitance.⁴⁴  The aperture of phase angles also increased with increasing immersion time, which could be a result of an improved surface coverage leading to a more capacitive surface film.^10,45 

To quantify the EIS measurements, the impedance plots were fitted with an equivalent circuit shown in the inset of Figure 4(a).^3-4  This model includes the following elements: solution resistance R_s, charge transfer resistance R_t, inductance L with the corresponding inductance resistance R_L, and constant phase element (CPE) representing the double layer capacitance. It can be seen from the Figures 4(a) through (d) that the fitted and measured results match well in both Nyquist and Bode plots. The following equation was used to calculate R_p:

It should be noted that at the presence of low-frequency inductance loop, difference exists between R_t and R_p. R_t can be measured as Z′ when −Z′′ = 0 in the Nyquist plots at intermediate frequencies, while R_p is defined as the difference in impedance between the solution resistance (|Z| when f → ∞) and low-frequency asymptote (|Z| when f → 0).⁴⁶ 

The typical fitting results of the electrochemical parameters for equivalent circuit components are listed in Table 3. For both 0.001% and 1% NaCl solutions, true polarization resistance (by EIS and LPR) decreases over time, while double layer capacitance and n increases over time. Furthermore, R_s contribution is significant and cannot be omitted during LPR measurement in 0.001% NaCl solution. In contrast, its contribution to the apparent polarization resistance is less than 5% in 1% NaCl and thus may be neglected.

According to the fitted results of LPR and EIS, change of R_p with immersion time is provided in Figure 5(a) for all five NaCl concentrations. It can be observed that the R_p decreases gradually as time elapses, and reaches relative steady state toward the end of the test. This change of R_p is probably a result of evolution of the electrode surface, as discussed later.^3,45  Figure 5(b) presents progression of corrosion rate as a function of time under the assumption of diffusion control cathodic kinetics. Corresponding to the decreasing trend of R_p, the corrosion rate gradually increases with time and reaches steady state at the end of 100 h. Further observation shows that R_p and i_corr values exhibit more fluctuation at lower NaCl concentration (0.001% and 0.01%), which may be a result of the unstable interface. The average corrosion rate over 100 h immersion was calculated via Equation (3) and presented in Figures 5(c) (diffusion control) and (d) (activation control). The ends of error bars represent the maximum and minimum values, and the LPR and EIS corrosion rates agree with each other. The results in both Figures 5(c) and (d) indicate the average corrosion rate decreases with increasing salt concentration. The average corrosion rates for diffusion control case are approximately 1.4 to 2.2 times higher than those for activation control caused by the effect of B values. It should also be mentioned that the corrosion rates reported here show a similar magnitude as reported by other researchers.^1,13,15-17 

Test solutions after long-term immersion experiments were collected and total iron concentration was analyzed by ICP. The results shown in Table 4 indicate that total iron concentration (presumably Fe²⁺) decreases with increasing salt concentration, which is consistent with the corrosion rate trend shown in Figures 5(c) and (d). In addition to total iron analysis, solution pH was also recorded after 100 h of immersion (Table 4). All solution pH increased by two units (from ~4 to ~6) and similar pH results have been reported before.⁸ 

All of the results lead to the conclusion that the corrosion rate of C1018 carbon steel increases with time, but decreases with increasing salt concentration under CO₂ saturation condition at room temperature.

Figure 6 shows the surface morphology of corrosion scale formed on the steel surface in the CO₂-saturated solutions with different concentrations of NaCl. This was performed immediately after the experiment was completed and the electrode was only rinsed with acetone and blown dry. An interesting phenomenon can be observed: dark area of the surface increases with increasing salt concentration. As shown in Figure 6(a), the electrode surface from 0.001% solution shows dark spots which distributed uniformly, while the electrode from 0.1% solution (Figure 6[c]) presents some insular-like dark area on the surface. As to the 10% NaCl exposed electrode in Figure 6(e), the surface is mostly covered with dark area. The corresponding higher magnification SEM images are shown in Figures 6(b), (d), and (f). Figure 7 shows the EDX analysis results corresponding to the bright area A and dark area B in Figure 6(f). As it shows, the Fe peak in area B is much lower than that in area A. The higher content of Na and Cl in area B also suggest the preferred adsorption of chloride and formation of intermediate corrosion product at localized areas.

After removing loose corrosion products by ultra-sonication method (refer to experimental procedure), surface morphology of the C1018 carbon steel electrode was again examined by SEM/EDX and the results are shown in Figures 8(a) through (c). The corrosion morphology is more uniform at low salt concentration (0.001%, Figure 8[a]) than at high salt concentration (10%, Figure 8[c]). In other words, surface roughness after corrosion increases with increasing salt concentration. This was further confirmed by optical and 3D microscopic images in Figure 9.

Figures 8(d) through (f) show the cross-sectional morphology of C1018 carbon steel after final cleaning. A zig-zag interface can be observed between the substrate and a corrosion product layer. As seen in Figures 8(d) and (e), the remaining corrosion product layer is approximately 15 μm and 8 μm thick after immersion in 0.001% and 0.1% NaCl, respectively. The scale thickness ratio seems to agree reasonably well with the average corrosion rate ratio shown in Figures 5(c) and (d), taking into account the difference in total dissolved iron results (Table 4). These corrosion layers are relatively compact. In contrast, the corrosion product layer thickness in 10% NaCl is only about 5 μm or less. This corrosion product layer is discontinuous and some holes can be seen underneath the scale, as shown in Figure 8(f). This agrees with the model proposed by Palacios and his coworkers.⁴⁷  EDX was performed on area A in Figure 8(d) and the results indicate this layer is composed of relatively carbon-rich material. The atomic ratio of carbon to iron is close to 1:3, which suggests the existence of iron carbide (Fe₃C), in agreement with published research.^5,10  Furthermore, the detection of 8 at% oxygen suggests that some oxygen-containing compounds such as iron carbonate (FeCO₃) and/or oxide may also exist in this corrosion product layer.

Cathodic and anodic polarization experiments were performed using microelectrodes to help explain the fact that corrosion rates measured by LPR and EIS decreased with increasing NaCl concentration under CO₂-saturated condition.

Anodic polarization results (Figure 10[a]) show that anodic kinetics is charge transfer controlled at low overpotential, and mass transport/ohmic potential drop controlled at high overpotential. Both the charge transfer region and the anodic limiting current density in the anodic polarization curves increase with increasing salt concentration. Figure 10(b) is a plot of current densities at selected overpotential vs. NaCl concentration. The fitting results indicate a good linear relationship between the current densities and salt concentration. As shown in the table (inset of Figure 10[b]), the slopes are close to 0.5 for overpotential of 300 mV and above, i.e.:

For comparison, a slope of 0.3 was found for the overpotential at 200 mV. This difference should be attributed to different kinetic processes between lower and higher overpotential.

Potentiostatic hold experiments were performed to study the steady state current at selected applied potential (E_app = −0.2 V_SCE) and the results are shown in Figure 11(a). It can be seen that the current density remained relatively constant throughout the 2-h period, except for the 0.001% case. The steady state current density increased with increasing salt concentration. It is interesting to note that there was a significant drop of current density from 2.6 mA/cm² to 0.2 mA/cm² for approximately 0.6 h in 0.001% NaCl solution, which may suggest the formation of complete corrosion product layer or salt film on the electrode.³²  The fact that the current density returned to almost the same level as before may result from breakdown of this corrosion product layer or salt film. The final current densities at the end of 2 h were plotted against NaCl concentration and presented in Figure 11(b). A linear relationship between log (i) and log [Cl⁻] can be observed, and the slope is 0.5, agreeing well with the potentiodynamic results. This also confirms that steady current density was achieved during potentiodynamic polarization.

As shown in Figure 12, the cathodic polarization curves show a vertical line in higher overpotential range, indicating the cathodic process is under mass transport control, probably resulting from H⁺ and/or H₂CO₃ reduction.^48-49  The current densities decrease with increasing NaCl concentration, suggesting the cathodic reaction kinetics is retarded by higher Cl⁻ concentration.

Constant potential of −0.8 V_SCE was applied to the C1010 wire electrodes for 2 h in different concentration NaCl solutions. The results are shown in Figure 13(a). The same trend as the potentiodynamic experiments was observed: cathodic limiting current density decreases with increasing salt concentration. The current densities extracted from potentiostatic polarization (steady state after 2 h) and cathodic potentiodynamic polarization (at −0.8 V_SCE) were plotted vs. NaCl concentration and presented in Figure 13(b). Similar results were obtained by both techniques.

For carbon steel corrosion, the overall anodic reaction can be simplified as:^4,10,34 

However, this reaction may involve many steps. It is generally accepted that the first step is the adsorption of water molecules on the steel surface via:^50-51 

In the absence of chloride, the anodic dissolution can proceed in the following steps in acid media:⁵² 

Considering this mechanism and Equation (11) being the rate determining step, an anodic Tafel slope of 120 mV/decade can be determined.⁵²  This Tafel slope has been reported in several phosphate-, borate-, or carbonate-containing aqueous systems.⁵³ 

However, in the presence of the chloride, competitive adsorption between hydroxyl and chloride occurs and chloride adsorption will tend to prevail at high chloride activity⁵²  through the following steps:

The overall reaction proceeds following Equation (12) and an anodic Tafel slope of 60 mV/decade can be determined assuming Equation (13) is the rate determining step. This Tafel slope has been reported by several authors for their results in acidic solutions with chloride.⁵²  There are other theories hypothesizing the role of chloride in the iron dissolution process, including the direct adsorption of chloride onto the surface without the initial adsorption of water molecules.⁵³ 

As shown in Table 2, the experimentally determined anodic Tafel slope decreases from ~150 mV/decade to ~60 mV/decade as the NaCl concentration increases from 0.001% to 10%. Tafel slope of ~60 mV/decade was obtained for salt concentration of 1% and 10%, which suggests the dominant adsorption of chloride (Equation [9′]) at 1% NaCl and above in CO₂-saturated electrolyte. When the salt concentration decreased to 0.1% (i.e., 0.017 M Cl⁻), anodic Tafel slope increased slightly to ~70 mV/decade. At low salt concentration (0.01% and 0.001%), higher Tafel slopes were obtained, which suggests the effect of hydroxyl adsorption (Equation [9]) in the competitive adsorption process. The anodic Tafel slope of ~150 mV/decade was higher than the theoretical prediction under hydroxyl adsorption mechanism, which may be attributed to the ohmic potential drop, mass transport through corrosion scale, or complication from other ions/reactions.

Nobe, et al., examined the Cl⁻-accelerated mechanism vs. OH⁻-accelerated mechanism for iron dissolution in acidic media.^54-55  Under the assumption of Tempkin adsorption behavior for chloride-containing intermediates, the anodic dissolution rate for the Cl⁻-accelerated mechanism can be obtained as follows:

where i_a,Cl is the anodic current density attributed to chloride-accelerated mechanism, k_a,Cl is the anodic reaction rate constant under this mechanism, F is Faraday constant, φ is potential, R is ideal gas constant, and T is absolute temperature.

At constant pH and potential, a reaction order of 0.5 with respect to Cl⁻ can be obtained. This is in agreement with the anodic polarization and potentiostatic hold experiments reported here. Similar results have been reported for iron dissolution in concentrated acidic chloride solutions. For example, McCafferty, et al., conducted polarization for iron in concentrated HCl and obtained a reaction order of 1 between anodic current density vs. [Cl⁻].⁵²  However, the opposite trend was observed under naturally corroding condition, i.e., the corrosion rate decreases with increasing salt concentration. Therefore, it is unlikely that the corrosion process is under anodic control.

Lastly, regarding bicarbonate ions, it has been reported that bicarbonate does not affect the rate of the first oxidation step (Equation [10])⁵³  and its effect may be considered secondary. However, the aggressive nature of bicarbonate may be explained through the following reaction steps:⁵³ 

The participation of bicarbonate in the overall reaction may also account for the Tafel slope deviation from 120 mV/decade.

The competitive adsorption can also be inferred from the results under naturally corroding condition. Higher chloride, as well as sodium, was found in patches on the surface after exposure (Figure 7). Thinner and discontinuous corrosion scale was formed through the competitive adsorption of chloride (Figure 8). It is also interesting to note that the adsorption of chloride possibly retarded the overall corrosion rate under naturally corroding condition, especially at the beginning (Figure 5[b]).

For carbon steel corrosion in CO₂-saturated NaCl solutions, there are two main cathodic reactions among other possibilities: hydrogen evolution reaction (HER) and direction reduction of H₂CO₃.^37,56-58 

These species are closely related through the following equilibria:

As a weak acid, H₂CO₃ serves as an additional source of H⁺ ions that enables HER to proceed at a higher rate than in a solution of strong acid at the same pH.⁵⁸  The direction reduction of H₂CO₃ can contribute to an increase in corrosion rate beyond the limit of HER. It has been postulated that H₂CO₃ adsorbs and reacts on the electrode surface and slow hydration of CO₂ is the rate determining step.⁵⁸  As shown in Table 4, pH is almost constant for the salinity range from 0.001% to 1% but decreases slightly for 10%. Therefore, direction reduction of H₂CO₃ through Equation (18) may be responsible for the fact that cathodic current increases with decreasing salt concentration (Figure 13). To better understand this process, solubility of CO₂ in different NaCl concentrations was calculated using OLI Stream Analyzer^† (software) and is shown in Figure 14. As the salt concentration increases from 0.1% to 10%, the dissolved CO₂ concentration decreases significantly from 0.032 M to 0.023 M. This is in agreement with the cathodic polarization and potentiostatic results in Figure 13. However, noticeable current decrease is observed in the range of 0.001% to 0.1% salinity, even though software calculation suggests minimal difference in CO₂ solubility. One possible reason for this discrepancy is that hydration of CO₂, a rate determining step, may occur faster at lower salinity. Another possible reason is that Fe₃C can form galvanic couple with iron as it has a much lower overpotential for the cathodic reactions than iron.^22,59  The fast corrosion rate in lower salt concentration solution exposes more Fe₃C on the surface, which further promotes the cathodic reduction reaction and increases corrosion rate by galvanic coupling. To summarize, although there are many factors affecting the overall corrosion rate of carbon steel in CO₂-saturated solution, it is mainly controlled by the cathodic reactions which are closely related to the solubility of CO₂.

Based on this discussion, Evans diagram for carbon steel corrosion in CO₂-saturated NaCl solution is proposed in Figure 15. NaCl concentration affects both anodic and cathodic processes. In 10% NaCl solution, chloride adsorption dominates the anodic reaction with Tafel slope of ~60 mV/decade and the cathodic reaction is under diffusion control with the lowest CO₂ concentration. For comparison, in 0.001% NaCl solution, hydroxyl adsorption dominates the anodic reaction with Tafel slope of ~120 mV/decade and cathodic reaction is under diffusion control or mixed control as a result of the highest solubility of CO₂. The deviation of anodic Tafel slope from theoretical prediction may be attributed to factors such as ohmic potential drop, thicker scale formation (Figure 8[d]), and/or competition by other anions (e.g., Cl⁻ and HCO₃⁻). Figure 15 also explains the higher OCP (Figure 2) and higher corrosion current density (Figure 5) observed for lower salt concentration.

As shown in Figure 5(a), polarization resistance decreased quickly in the beginning (first 30 h) and with further increase of immersion time (next 70 h), the decrease of polarization resistance slowed down and reached a relative stable value. Sun, et al., also reported the same phenomenon.³  Figure 16 shows the images of the corroding electrode surface recorded at different immersion time. It is seen that the gas bubbles gradually grew on the surface and almost disappeared with increasing immersion time to 30 h. According to observations, the corrosion occurred initially within the area of these gas bubbles, as the shining surface gradually turned dark and gray when the adsorbed bubbles disappeared. Over time, the adsorbed gas bubbles gradually covered the entire surface and the bright shining surface became completely dark and gray. This occurred approximately after 30 h of immersion and coincided with the time corrosion rates reaching steady state (Figure 5). It should be pointed out that this is a process that the corroded area gradually increases. Therefore, the higher polarization resistance at the initial time may result from the overestimated corrosion area. The identity of the bubbles is unclear, and it could be either the adsorbed reactant H₂CO₃ or product H₂ from cathodic reaction. These adsorbed bubbles may also be associated with the inductance loop found in the EIS measurement at low frequency (Figures 4[a] and [c]).

The stability of interface during the corrosion process can be further confirmed by the n values when fitting the CPE elements in EIS results. As shown in Figure 17, n gradually increases to nearly 1 with increasing immersion time, which is near ideal capacitive behavior. After 30 h, almost all n values are above 0.9 except for 0.001% case, indicating a stable and uniform corrosion layer could be formed after 30 h in these environments.^3,60  This coincides with the results in Figure 16 when the gas bubbles disappeared after 30 h immersion. In addition, the low n values in the 0.001% case may be related to the large fluctuation in corrosion rates (Figure 5[b]).

Based on these analyses, it can be inferred that the corrosion process may be divided into two stages: (I) initial corrosion scale formation, and (II) corrosion after a stable and uniform corrosion layer formed. In the first 30 h of immersion (Stage I), corrosion rate kept increasing (Figure 5[b]), electrode surface turned dark and gray from bright and shiny, and large gas bubbles diminished by visual observation (Figure 16). In the next 70 h of immersion (Stage II), corrosion rate reached and maintained a high level, electrode surface was completely covered by scale, and gas bubbles disappeared visually. It should also be pointed out that corrosion rates, as well as OCPs, converge over time after the local environment formed at the interface and the bulk chemistry has less influence on the corrosion process.

A comparison between Figures 8(e), 8(f), and 5 suggests that the corrosion product layer thickness is proportional to corrosion rate. The thickest scale was found in 0.001% NaCl solution, which exhibited the highest corrosion rate. EDX results suggest that this scale possibly consists of Fe₃C and FeCO₃. During the corrosion process, dissolution of iron leaves behind Fe₃C on the corroded surface. FeCO₃ may form through the participation of HCO₃⁻ in the oxidation step of FeOH_ads (Equations [15] and [16]), or precipitation reactions as follows:^56-57,61 

The formation of iron carbonate depends on the local concentration of Fe²⁺ and CO₃²⁻/HCO₃⁻/H₂CO₃. The fact the thickest scale was found from the most dilute solution may be explained by the higher concentration and higher transference number of CO₃²⁻/HCO₃⁻/H₂CO₃ under this condition. These FeCO₃ precipitated under the Fe₃C layer or anchored between the Fe₃C frameworks are difficult to remove even by ultrasonic cleaning.⁴⁷  The compact corrosion layer may retard the diffusion of Fe²⁺, thus increasing the local supersaturation of Fe²⁺ and facilitating the formation of more FeCO₃. In contrast, in concentrated chloride solution (10% NaCl), this scale is thinner and discontinuous (Figure 8[f]), leaving some porous areas that are high in chloride concentration (Figure 7[d]). This can be attributed to lower CO₂ solubility, lower transference number of CO₃²⁻/HCO₃⁻/H₂CO₃, and more importantly, preferred adsorption of Cl⁻.^62-63 

Finally, an interesting phenomenon was found in the CO₂-saturated NaCl solutions in which the corrosion rate increases with decreasing salt concentration in a wide range from 0.001% to 10 wt%. This appears to be different from the corrosion behavior of carbon steel in naturally aerated salt water, where a maximum corrosion rate was found around 3 wt%.^64-65  Corrosion rate decreases with either decreasing or increasing salt concentrations for different reasons: low conductivity at low salt concentration and low oxygen solubility at high salt concentration.^64-65  The corrosion rates of carbon steel were also measured in air-saturated NaCl solution in the same concentration range (from 0.001 wt% to 10 wt%) and a maximum corrosion rate was found around 1% in air-saturated NaCl solution (not shown here). Different corrosion behaviors in these two atmospheres may be attributed to the different solubility of CO₂ and O₂ (from air) at room temperature: ~0.03 M for CO₂ (Figure 14) vs. ~0.0003 M for O₂ (~10 ppm⁶⁴ ). High solubility of CO₂ itself and associated high corrosion rates in this environment (at least one order of magnitude higher than the case of oxygen) will further increase solution conductivity.

• In CO₂-saturated NaCl solutions, corrosion rate decreases with increasing salt concentration ranging from 0.001 wt% to 10 wt% at room temperature. Highest corrosion rate, thickest corrosion scale, and highest dissolved total iron was found in the 0.001% NaCl solution after long-term (100 h) exposure.

• The anodic reaction kinetics is dominated by chloride adsorption in high concentration chloride solution (e.g., 10%) and hydroxyl adsorption in dilute solution (e.g., 0.001%). A reaction order of 0.5 was determined with respect to chloride concentration.

• The cathodic reaction is under diffusion control and/or mixed control depending on the salt concentration, and the limiting current density increases with decreasing salt concentration (increasing CO₂ concentration).

• The overall reaction rate of carbon steel in CO₂-saturated NaCl solution is under cathodic reaction control and the solubility of CO₂ is considered to have significant impact.

• Surface roughness increases in high chloride solution as a result of the competitive adsorption between hydroxyl and chloride, while more uniform corrosion occurs at lower salt concentration through dominant hydroxyl adsorption mechanism.

• The corrosion behavior can be divided into two main stages: (I) initial formation of corrosion scale, and (II) continuous corrosion after the formation of a stable and uniform corrosion layer. Hydroxyl, chloride, and CO₂ compete during the corrosion scale formation. After a relatively stable scale forms, the corrosion rate remains steady state.