1 INTRODUCTION

Early deterioration of reinforced concrete structures is one of the most pressing problems in the construction industry and in transport infrastructure projects. In the coastal areas of China, about 50% to 70% of structures serving these industries need to be repaired within 10 to 30 years (Fan, 2009). In Western China, the service life of mortar in salt lakes is only 2 to 3 years (Yu, 2004). It has been estimated that corrosion costs account for 3.5% of gross national product (Hoar, 1971), so China lost about 2000 billion RMB in 2014 as a consequence of corrosion (Hou et al., 2012). Saltinduced damage and carbonation are the main reasons for degradation of concrete structures in these harsh environments (Brown et al., 2000; Castro et al., 2000; Poupard et al., 2006; Jin et al., 2010; Cerqueira et al., 2012). Depassivation and corrosion of rebar can lead to the localized formation of porous oxide layers at the steel/concrete interface. The transformation of metallic iron into oxides is accompanied by an increase in volume of two to six times, leading to stress generation and cracks in concrete surfaces (Zhao et al., 2013). Oxygen and aggressive ions easily pass through crevices to accelerate corrosion. As a result, irreversible damage to reinforced concrete, such as concrete cracking, capacity loss, and deterioration of the structure, occurs (Williamson and Clark, 2000; Tapan and Aboutaha, 2011; Bilcik and Holly, 2013).

In an environment containing both chloride and sulfate salts, the mechanism of deterioration of reinforced concrete may become even more complex because of the interaction between chloride ions (Cl^-) and sulfate ions (SO₄^2-). Previous studies have been carried out on the performance of concrete exposed to composite solutions of chloride and sulfate ions and the consequent corrosion behavior of rebar in the concrete (Al-Amoudi et al., 1995; Tumidajski and Chan, 1996; Jin et al., 2007). Al-Tayyib and Shamim Khan (1991) reported that sulfate ions are corrosive to reinforced bar, but not as strongly as chloride ions: the corrosion rate of chloride-contaminated concrete specimens was 23%-35% higher than that of sulfatecontaminated specimens. However, Cheng et al. (1990) and Al-Tayyib et al. (1998) provided contrasting data, indicating that sulfate ions could change the mechanistic parameters of surface film more significantly and the corrosive effect of the aggressive species followed the order SO₄^2->Cl^-. Jarrah et al. (1995), Saleem et al. (1996), and Dehwah et al. (2002) reported that sulfate concentration strongly affected the corrosion rate, but no systematic effect was found for the initial corrosion period. For reinforced concrete in ocean and salt lake environments, calcium hydroxide produced by cement hydration increases the pH value of concrete above 12.5. The penetration of sulfate and chloride ions into a rebar surface is affected by calcium hydroxide. When carbonation exists at the same time, some calcium hydroxide converts to calcium carbonate, which reduces the pH. Corrosion behavior of rebar in such environments is therefore more complicated. To study the influence of materials and environmental factors on composite salt-induced corrosion of rebar, Pradhan (2014) evaluated the effect of composite ions, type of cement, and water/concrete (w/c) ratios on the corrosion of steel in concrete. This study indicated that corrosion behavior of reinforced concrete was influenced by cement type, w/c ratio, and chloride ion concentration, except in the presence of sulfate ions. The corrosion behavior of reinforced concrete in a sulfate-chloride environment showed that the range of the passive zone was larger in the presence of both chloride and sulfate ions compared with that in the presence of chloride ions alone. The presence of sulfate ions may mitigate the occurrence of pitting corrosion caused by chloride ions. Sulfate ions may also reduce the expansion pressure induced by the corrosion products and decrease the fraction of macro defects in the concrete (Jin et al., 2015; Shaheen and Pradhan, 2015). From these studies, it can be concluded that the interaction of sulfate and chloride ions on corrosion of reinforced bar is not unambiguously clarified.

The penetration of these ions into concrete to induce corrosion of rebar is a long-term process that is affected by various factors, including the nature of the materials, ion concentrations, and, particularly, the reaction of sulfate ions with hydrated cement. It is therefore necessary to investigate the interactions between sulfate ions, chloride ions, and cement hydration on the electrochemical behavior of rebar.

The critical ion concentration is very important for predicting the service life of reinforced concrete in saline environments (Lollini et al., 2015). The critical ion concentration for steel corrosion varies according to concrete and steel type and environmental factors (Shi et al., 2012). In this paper, the critical ion concentration was investigated by studying the corrosion of reinforced concrete in simulated pore solutions with different concentrations and types of corrosive ions. Concrete comprises a heterogeneous microstructure, which will affect the penetration velocity of the ions and result in uncertainty in the concentrations of sulfate and chloride ions around the reinforced bar. Many researchers have studied reinforced bar corrosion in simulated concrete pore solutions (Abd El Aal et al., 2009; Liu et al., 2014). In this paper, a comprehensive investigation was carried out, including the electrochemical behavior and corrosion processes of reinforced bar in Ca (OH)₂ solution and cement grout containing aggressive salts NaCl and Na₂SO₄ for different immersion times. The proposed interactions between sulfate ions, chlorine ions, and calcium hydroxide on steel corrosion provide the basis for design of high-durability concrete and corrosion control.

2 EXPERIMENTAL 2.1 Materials and specimen preparation

Cylindrical specimens were made of carbon steel with a diameter of 10 mm and length of 5 mm. The chemical composition of the steel was (wt. %) 0.22 C, 0.30 Si, 0.65 Mn, 0.05 S, 0.045 P, and the remainder Fe. A copper wire lead was soldered to one end of the sample. The flank and end surface with the thick copper wire were sealed by epoxy resin, leaving an exposed area of about 0.785 cm² as the working electrode. The exposed surface was sequentially polished with SiC water polishing papers with grits of 400#, 800#, and 1500#. Finally, the steel specimen was ultrasonically degreased in acetone, washed in distilled water, and then dried at room temperature.

2.2 Electrochemical experiments

Totally, six sets of electrolyte compositions were selected to simulate corrosive concrete pore solutions: (1) saturated Ca (OH)₂ solution with 0%, 0.06%, 0.12%, 0.20%, 0.23%, 0.29% (wt.%) NaCl; (2) 0%, 0.03%, 0.06%, 0.09%, 0.12% (wt.%) NaCl solution at pH 11.5; (3) 0.25%, 0.345%, 0.5%, 1.0% (wt.%) Na₂SO₄ solution; (4) saturated Ca (OH)₂ solution with 0.25%, 0.5%, 1.0% (wt.%) Na₂SO₄; (5) saturated Ca (OH)₂ solution with 0.29% NaCl and 0.25%, 0.5%, 1.0% (wt.%) Na₂SO₄; (6) 0.25%, 0.345%, 0.5%, 1.0% (wt.%) Na₂SO₄ with cement grout (15 g Chinese standard 52.5 R (I) Portland cement plus 400 g water). The pH of solutions (1), (4), and (5) was controlled by saturated Ca (OH)₂ at a value of 12.54; the pH values of solutions (3) and (6) were 7.0 and 12.85, respectively. For solution (2) (Ca (OH)₂ mixed with different concentrations of NaCl), the pH was reduced to 11.5 by bubbling through CO₂-enriched air to simulate carbonated concrete (Paredes et al., 2012).

Prior to measurement, the rebars were immersed in saturated Ca (OH)₂ solution for 7 days to promote a passive film to simulate field conditions (Poursaee and Hansson, 2007). Three identical rebar samples were placed in each solution. Electrochemical experiments were performed in a classical threeelectrode cell, assembled with carbon rebar as the working electrode, a platinum foil of 2.0 cm×2.0 cm as the counter-electrode, and a saturated calomel electrode provided with a Luggin capillary as the reference electrode. The electrochemical measurements were performed at room temperature using a Princeton Versa STAT 3 potentiostat/ galvanostat (Princeton Applied Research, Oak Ridge, United States). EIS measurements were carried out in the frequency interval from 100 kHz to 10 mHz at steady open-circuit potential with a root mean square amplitude of 10 mV. The EIS data were analyzed by ZsimpWin software (Echem software, USA). The polarization resistance (R_p) and the passive film resistance (R_f) were calculated according to equivalent-circuit simulations. The instantaneous corrosion rate (i_corr) of the rebar was obtained from the polarization resistance via the Stern-Geary Eq.1 (Montemor et al., 2003):

where B is the Tafel slope. Anodic Tafel slopes have been found to vary in different solutions (Chang et al., 2008). To develop an appropriate procedure for obtaining authentic corrosion rates, potentiodynamic polarization experiments were therefore performed to determine the Tafel slopes of rebar in the different solutions. The anodic and cathodic Tafel slopes, b_a and b_c, respectively, were obtained by curve-fitting. The values of B were calculated by Eq.2, as shown in Table 1:

2.3 Surface characterization

Surface morphologies of the rebar were visually observed and the microstructure analyzed by scanning electron microscopy (SEM, KYKY-2800B, KYKY, Beijing, China).

3 RESULT AND DISCUSSION 3.1 Pure NaCl solution

EIS diagrams were determined after the reinforced bars had been immersed in saturated Ca (OH)₂ with different concentrations of NaCl for 1 h: the Nyquist plots are given in Fig. 1. It is evident that all plots showed a single capacitive loop, with the radii reducing with increasing concentration of chloride ions, indicating that the protective ability of the passive film was gradually undermined by the chloride ions.

The R(Q(R(QR))) equivalent circuit chosen to fit the experimental plots is shown in Fig. 2a. The R_s value at high frequency represents the resistance between the electrolyte and rebar, and the time constants at high frequency (Q_f and R_f) reflect information pertaining to the passive film. Q_dl and R_p at low frequency are closely related to the corrosion reaction on the surface of the steel. R_p can be obtained from EIS data, from which the corrosion current (i_corr) can be calculated from Eq.1, as shown in Fig. 3.

Because the passive film may have been heterogeneous, a constant phase-angle element Q in the equivalent circuit was used to represent capacitance. Q is expressed as ω^-n/Y₀ (cosnπ/2+jsinnπ/2), where Y₀ and n are the constant and exponent, respectively, ω is the angular frequency in rad s^-1 (ω=2πf) and j²=-1 is an imaginary number (Nishikata et al., 1997; Zou et al., 2011). In Fig. 2, Rs represents the solution resistance, R_f and Q_f represent the resistance and capacitance of the passive film, respectively, R_ct and Q_dl represent the polarization resistance and double-layer capacitance, respectively, and Z_w represents Warburg resistance (Cheng et al., 2004). To facilitate comparison with other literature, R_p replaces R_ct in the following text (Liu et al., 2014)

The rebar was immersed in solution (2) for 1 h: the Nyquist plots are shown in Fig. 4a, while the calculated polarization resistances and i_corr of the rebar in different chloride concentrations are depicted in Fig. 4b. The capacitive loop radii in the EIS plots obviously reduce with increasing NaCl concentration at both pH 12.54 and pH 11.5. The impedance decreases sharply and the corrosion current increases by a factor of about 10 with decreasing pH of the chloride solution.

Figures 3 and 4b indicate that i_corr was stable at low chloride concentrations due to the intact passive film on the rebar during the first hour, but increased with increasing NaCl concentration. When rebar was immersed into chloride solution at pH 11.5, the relationship between i_corr and NaCl in the corroding solution can be represented by a simple quadratic polynomial. By calculating the derivative of Eq.3, the breakpoint of i_corr can be obtained when the derivative value is equal to zero:

The critical NaCl concentration for the initiation of pitting corrosion at pH 11.5 was 0.006 417% (1.1 mmol/L).

An exponential function fitted the experimental data for rebar immersed into saturated Ca (OH)₂ with NaCl. Exponential functions (4) and (5) were used to depict the relationship between i_corr and the concentration of NaCl.

NaCl solution at pH 11.5:

Saturated Ca (OH)₂ with NaCl:

where i_corr is corrosion current density (μA/cm²) and C_NaCl is the concentration of NaCl in the corroding solution (wt. %).

Because the critical concentration of chloride ion for the initiation of pitting corrosion was 1.1 mmol/L at pH 11.5, the initial corrosion current (i_corr) was 190.8 μA/cm², based on Eq.4. According to the calculation results for i_corr, the critical concentration of chloride ions in saturated Ca (OH)₂ solution is 38.3 mmol/L.

3.2 Pure Na₂SO₄ solution

Corrosion product could be observed after the rebar had been immersed in 0.25%-1.0% Na₂SO₄ solution for 1-2 h. EIS diagrams for rebar immersed in solutions with 0.25%, 0.5%, and 1.0% Na₂SO₄ for 1 h are presented in Fig. 5. Only a single capacitive loop was recorded for the steel subjected to the 0.25% and 0.5% Na₂SO₄solutions; in 1.0% Na₂SO₄, however, a high-frequency loop and a low-frequency straight segment were observed in the Nyquist plot. Two equivalent circuits, R(Q(R(QR))) and R(Q(R(Q(RW)))), were employed to simulate these plots, as shown in Fig. 2. The best fit was obtained using the R(Q(R(QR))) circuit for rebar in 0.25% Na₂SO₄ solution, but the R(Q(R(Q(RW)))) circuit was more suitable for fitting the impedance data for rebar in 1.0% Na₂SO₄ solution. The above-mentioned equivalent circuits showed similar suitability for the 0.5% Na₂SO₄ solution. It is evident that as the Na₂SO₄ concentration in the corrosion solution increased, the corrosion gradually changed from a charge transfer controlled process to a diffusion controlled process.

According to the fitted results for the equivalent circuits, R_p was introduced into the Stern-Geary equation to calculate i_corr. The values of R_p and i_corr, calculated from EIS plots for rebar in Na₂SO₄ solution, are shown in Fig. 6. It is evident that i_corr increases with increasing concentration of Na₂SO₄, Eq.6 is proposed:

The calculated critical concentration of sulfate ions in corrosion solution for initiation of pitting corrosion (i_corr=190.8 μA/cm²) is 95 mmol/L, which is about 2.5 times of that of the chloride ion. This implies that, for the initiation of pitting corrosion, sulfate ions are less aggressive than chloride ions at the same molar concentration. This can be attributed to the different penetration velocities and ionic structures of the two ions. The radius of the chloride ion is 0.018 1 nm, but that of sulfate ions is 0.295 nm, which is 6.3 times higher. The penetration rate of sulfate ions into the passive film is therefore slower than that of chloride ions. Furthermore, the passive film on the rebar exhibits behavior typical of an n-type semiconductor. The chloride ion can react with an oxygen vacancy via a Mott-Schottky pair-type reaction to generate cation/oxygen vacancy pairs; the oxygen vacancies will, in turn, react with additional chloride ions at the film or solution interface to generate yet more cation vacancies. It can be therefore concluded that the generation of cation/oxygen vacancies is autocatalytic (Zhao et al., 2013). In contrast, the sulfate ion consists of a central sulfur atom surrounded by four equivalent oxygen atoms in a tetrahedral arrangement and its ligand attachment is via one or two oxygen atoms, as shown in Fig. 7. Because only oxygen atoms can be bound by other atoms, these oxygen atoms prevent SO₄^2- from reacting with oxygen vacancies (V_o^••) via a Mott-Schottky pair-type reaction to generate cation/ oxygen vacancy pairs. These few cation/oxygen vacancies cannot efficiently gather metal vacancies at the Fe/passive film interface during the corrosion process. The passive film on the rebar can therefore grow in Na₂SO₄ solution during the corrosion process: generation and undermining of the passive film proceed simultaneously. The relatively slow penetration rate of SO₄^2- and the continuously generating passive film would consequently reduce the corrosion rate of steel for the same concentrations of SO₄^2- and Cl^- ions. These observations can explain why the critical molar concentration for sulfate ions is higher than that for chloride ions for activating pitting corrosion.

3.3 Saturated Ca (OH)₂ solution with Na₂SO₄

When rebar was immersed into saturated Ca (OH)₂ solution with various concentrations of Na₂SO₄, the existence of Ca (OH)₂ reduced the corrosion rate of rebar. Typical Nyquist diagrams for different immersion times are plotted in Fig. 8. During the first 3 hours of immersion in saturated Ca (OH)₂ solution with 0.25% Na₂SO₄, only a single capacitive loop was observed. With increasing immersion time, a lowfrequency straight segment began to appear. When the concentration of Na₂SO₄ was 0.5%, the Nyquist plots were characterized by both a high-frequency loop and a low-frequency straight segment. As the Na₂SO₄ content increased to 1.0%, two loops appeared in the Nyquist plots for the rebar/corrosion solution interface, which could be attributed to a thick layer of corrosion products covering the steel surface. Figure 9 shows the fitted results using the equivalent circuits R(Q(R(QR))) and R(Q(R(Q(RW)))) for rebar immersed in saturated Ca (OH)₂ solution with 1.0% Na₂SO₄. It can be seen that R(Q(R(Q(RW)))) is more suitable than R(Q(R(QR))), which indicates that the corrosion rate is mainly controlled by a diffusion process.

In comparison with the results of Fig. 6, the calculated R_p increased by 10 to 20 times in Fig. 10, while i_corr reduced sharply. There are two reasons: first, the increase of pH from 7.0 to 12.54 in the corrosion solution could promote the generation of passive film; second, the rebar was covered with gypsum formed by the reaction between Ca (OH)₂ and Na₂SO₄, which inhibited the penetration of sulfate ions into the passive film.

As shown in Fig. 8, the radii of the high-frequency capacitive loops reduced initially, then increased, and then again decreased with increasing corrosion time. According to the fitted EIS results, i_corr of the steel increased with increasing Na₂SO₄ concentration in saturated Ca (OH)₂. In general, the value of i_corr initially increased with increasing corrosion time, then decreased, and then again increased until the onset of pitting corrosion.

The evolution of i_corr with corrosion time may be attributed to two reaction processes simultaneously taking place on rebar in saturated Ca (OH)₂ with Na₂SO₄. The first is that sulfate ions penetrated into the passive film, which was undermined, thereby accelerating the corrosion; the second is that Na₂SO₄ reacted with Ca (OH)₂, resulting in the formation of gypsum, which deposited on the surface of the passive film and could block the penetration of sulfate ions and oxygen, thereby suppressing corrosion. Competition exists between these two processes. The morphology of steel corroded in saturated Ca (OH)₂ solution with Na₂SO₄ is shown in Fig. 11.

According to the evolution of i_corr, the corrosion process can be divided into three stages. During the first stage, i_corr increased with time. Because formation of a gypsum layer is a slow process and was not continuous at the initial time, sulfate ions penetrated into the passive film and easily induced corrosion. During the second stage, i_corr decreased with time. The thickness of the oxide layer increased and the gypsum layer gradually deposited on its surface. These two layers would prevent the diffusion of sulfate ions and oxygen to the passive film. During the third stage, i_corr increased rapidly. The transformation of metallic iron into oxide rust is accompanied by a volume increase by a factor of four to six. The increase in the amount of rust in the inner layer would result in expansion pressure, which would cause cracking of the gypsum layer and thereby allow large amounts of sulfate ions and oxygen to penetrate into the block layer.

3.4 Composite solution

Rebar was immersed into saturated Ca (OH)₂ solution containing 0.29% NaCl and Na₂SO₄: EIS diagrams for different immersion times are shown in Fig. 12. Similar results for different times are apparent in 0.25% and 0.5% Na₂SO₄ solutions: both Nyquist plots contain two loops. When 1.0% Na₂SO₄ was added, however, the Nyquist plot showed two loops and a low-frequency short straight segment. A R(Q(R(Q(RW)))) circuit was used to fit the experimental data.

Table 2 lists the electrochemical parameters obtained by fitting the EIS diagrams for rebar exposed to saturated Ca (OH)₂ with the composite salt solution for different times. Because sulfate ions increase the electrical conductivity of the electrolyte, R_s, analyzed in the high-frequency region, decreased with time, and then tended to reduce with increasing Na₂SO₄ concentration. The R_f reduced with increasing immersion time and the value of Z_w tended to increase with time. The R_p of rebar in the composite solution (solution (5)) was smaller than that in saturated Ca (OH)₂ with Na₂SO₄ (solution (4)). The value of R_p increased first, then decreased with time. Synergistic corrosion by sulfate and chloride ions accelerated the deterioration of passive film. With addition of NaCl, the solubility of gypsum in the composite solution increased by a factor of 2 to 3 (Madgin and Swales, 1956); the thickness of the deposited layer of gypsum decreased accordingly, thereby accelerating the corrosion. The R_p of the rebar in the different solutions was determined from the equivalent circuit simulations of the EIS data and i_corr values were calculated from Eq.1. The order of corrosion in the simulated electrolytes was: solution (2) > solution (1) > solution (3) > solution (5) > solution (4).

3.5 Cement grout with sulfate solution

Figure 13 presents EIS diagrams for rebar in cement grout with different concentrations of Na₂SO₄ solution. Two equivalent circuits, R(Q(R(QR))) and R(Q(R(Q(RW)))), were used to simulate the plots. R_p was calculated and introduced into the Stern-Gary equation to calculate i_corr. The evolutions of R_p and i_corr are shown in Fig. 14.

The i_corr values increased with increasing immersion time except for 0.25% Na₂SO₄ solution. The i_corr trends were the same as for solution (4), although the values of i_corr were about 8 to 10 times higher. The trends in evolution of R_f in solutions (6) and (4) were identical, as shown in Fig. 15. It is concluded that using a saturated solution of Ca (OH)₂ with Na₂SO₄ to simulate sulfate-contaminated cement grout is feasible. The relatively high i_corr value and low passive film resistance may be attributed to the lower pH of cement grout compared with that of saturated Ca (OH)₂, and which weakened the protective capacity of the passive film.

4 CONCLUSION

(1) Corrosion of rebar in chloride solution was controlled by charge transfer, and the corrosion current increased with increasing NaCl concentration and decreasing pH. The critical chloride ion concentrations for activating pitting corrosion in saturated Ca (OH)₂ solution and in chloride solution at pH 11.5 were 38.3 mol/L and 1.1 mol/L, respectively. Carbonation decreased the pH, which not only accelerated the corrosion rate, but also reduced the critical chloride concentration.

(2) The corrosion rate increased with an increase of Na₂SO₄ concentration in solution. The process was diffusion controlled. The relatively slow penetration velocity of sulfate ion and continuous formation of passive film yielded a higher critical concentration of sulfate ions than that of chloride ions for activating pitting corrosion.

(3) The presence of Na₂SO₄ in Ca (OH)₂ in solution reduced the corrosion rate of rebar significantly. The process was mainly diffusion controlled. Two processes-penetration of sulfate ions into undermined passive film and reaction of Na₂SO₄ with Ca (OH)₂ to form gypsum deposition on the surface of rebar-proceeded simultaneously, which resulted in fluctuation of i_corr at the early stage, but a rapid increase at the later stages.

(4) When rebar was immersed into a composite solution containing both sulfate and chloride ions, overlapping effects accelerated the corrosion process, which was mainly controlled by diffusion. Chloride ions increased dissolution of gypsum and reduced the thickness of the gypsum layer, thereby weakening its ability to block corrosive ions. The i_corr in composite solution increased by 10 times, compared with that of rebar in saturated Ca (OH)₂ with Na₂SO₄.

(5) The effect of Na₂SO₄ on EIS of rebar immersed in cement grout was similar to that in saturated Ca (OH)₂ with Na₂SO₄. Maintaining the alkalinity of concrete and reducing the concentration of sulfate ions around the reinforced concrete is helpful in reducing the risk of steel corrosion in concrete and delaying the onset of corrosion.

(6) In future work, more complex synthetic concrete pore solutions, including KOH, NaOH, and Ca (OH)₂, will be evaluated, and the interaction effects of K⁺ and Na⁺ with SO₄^2- and Cl^- on rebar corrosion will be studied. These results can provide the basis for design of high durability concrete and corrosion control.