Corrosion Behavior of Reinforcing Steel in the Immersed Tube Tunnel (ITT) under Submarine Environment

The corrosion behavior of reinforcing steel in the ITT under a submarine environment was investigated. Electrochemical tests were carried out to separately determine the linear polarization curves and the AC impedance spectra of rebars in the ITT scaled-down models subjected to pressurized seawater erosion, from which key parameters were obtained, including the self-corrosion potential (Ecorr), corrosion current density (icorr), polarization resistance (Rp), concrete resistance (Rc), and charge migration resistance (Rct). The results show that in the process of pressurized seawater erosion, the rebars on the seawater side of the ITT models corroded earlier than the rebars on the cavity side, and it is recommended that anti-chloride ion penetration measures be taken on the surface of the seawater side as a priority in the project. The corrosion rate of rebars on the seawater side was significantly higher than that on the cavity side, and the corrosion rate of rebars on the cavity side increased as the erosion time increased. The corrosion rate of rebars in the ITT models was affected by chloride ions to a greater extent than by oxygen. Furthermore, by regression equation, a linear function between Rp obtained from the polarization curves and Rct obtained from the AC impedance spectra was established.


Introduction
Since the first use of an ITT in the United States in 1910 in the Yerba Buena Island Crossing, more than 150 submarine tunnel projects in the world have adopted ITTs [1]. And in recent years, with the implementation of China's national strategy of "Ocean Power", long-service-life concrete projects, such as the Hong Kong-Zhuhai-Macao Bridge [2], the Shenzhen-Zhongshan Link [3], the Bohai Bay Submarine Tunnel [4], and the Jiaozhou Bay Submarine Tunnel [5] have been completed or are under planning. In addition, there is also the under-construction Fehmarnbelt Tunnel between Germany and Denmark, and these submarine tunnels all use ITTs as a construction option. Driven by the convenience of transportation and the development of sea resources, and secured by high-performance concrete preparation technologies, advanced engineering machines, and construction technologies, the ITT has the advantages of short construction time and high construction quality [6], and has been adopted by more and more submarine tunnel projects.
While under the submarine environment, the durability of ITTs, like other reinforced concrete (RC) structures, is mainly faced with the corrosion of reinforcing steel caused by seawater erosion, which will lead to structural performance degradation and eventual failure [7,8]. Therefore, the builders have put forward high requirements for the durability design of the ITTs, taking into account economics, safety, and long service life. For the corrosion mechanism of RC structures, researchers have found that environmental temperature [7], protective cover [9], and the pH of the concrete pore solution [10] have significant

Design and Fabrication of ITT Scaled-Down Models
Two ITT scaled-down models with lengths, widths, and heights of 300 mm, 300 mm, and 280 mm, respectively, were prepared and named as Model 1 and Model 2. A total of 16 pieces of 8-mm diameter HPB300 bright-round carbon rebars (for which the elemental composition is shown in Table 4) were buried in each model, where 8 pieces each were buried in the inner (near the cavity) and outer (near the seawater) sides, the thickness of the protective layer of the bars was set to 30 mm, and the wall thickness was 100 mm. The rebars to be tested were acid washed, soaked in saturated calcium hydroxide solution for 7 d, and then buried in the mortar together with saturated glycerol electrodes and Ag/AgCl probes. The ends of the rebars were soldered with copper conductors, and finally, the top and bottom of the tunnel models were sealed with epoxy resin. Figure 1a,c shows the detailed arrangement of the tunnel models with the reinforcements, moisture sensors, saturated glycerol electrodes, Ag/AgCl probes, and air inlet and outlet holes. The diameters of the air inlet and outlet holes were 10 mm. The air pump delivered air through the inlet hole into the ITT model's cavity by the 10 mm diameter rubber tube, and then excluded the air from the outlet hole, thus realizing the air circulation in the model's internal cavity. The tunnel models were subjected to standard curing (T = 20 ± 2 • C, RH ≥ 95%) for 21 d, the central 200 mm areas of the 4 sides of each model were retained as the test surface, and the remaining 2 sides and the edge areas of the 4 sides were sealed with 1 mm thickness epoxy resin, as shown in Figure 1b. Then, curing was continued up to 28 d.    Figure 2 shows the fabrication and testing processes of each ITT model, which includes the following: (1) model casting; (2) burying the rebars, saturated glycerol electrodes, and Ag/AgCl probes; (3) using wires to electrically connect the inner and outer rebar; (4) casting the top plate of the model; (5) installing the humidity sensors, as well as the inlet and outlet pipes; and (6) transferring the model into a pressure vessel to perform pressurized seawater erosion. The temperature of the room was controlled at 20 ± 2 °C during the test.

Pressurized Seawater Corrosion
An analysis of seawater samples obtained along the South China coast revealed a chloride ion concentration of 16.5 g/L; thus, the corrosion medium used in this study was  Figure 2 shows the fabrication and testing processes of each ITT model, which includes the following: (1) model casting; (2) burying the rebars, saturated glycerol electrodes, and Ag/AgCl probes; (3) using wires to electrically connect the inner and outer rebar; (4) casting the top plate of the model; (5) installing the humidity sensors, as well as the inlet and outlet pipes; and (6) transferring the model into a pressure vessel to perform pressurized seawater erosion. The temperature of the room was controlled at 20 ± 2 • C during the test.  Figure 2 shows the fabrication and testing processes of each ITT model, which includes the following: (1) model casting; (2) burying the rebars, saturated glycerol electrodes, and Ag/AgCl probes; (3) using wires to electrically connect the inner and outer rebar; (4) casting the top plate of the model; (5) installing the humidity sensors, as well as the inlet and outlet pipes; and (6) transferring the model into a pressure vessel to perform pressurized seawater erosion. The temperature of the room was controlled at 20 ± 2 °C during the test.

Pressurized Seawater Corrosion
An analysis of seawater samples obtained along the South China coast revealed a chloride ion concentration of 16.5 g/L; thus, the corrosion medium used in this study was (b) (c) Figure 2. Fabrication and testing processes for the ITT models.

Pressurized Seawater Corrosion
An analysis of seawater samples obtained along the South China coast revealed a chloride ion concentration of 16.5 g/L; thus, the corrosion medium used in this study was based on sodium chloride at 10 times the concentration of simulated seawater, i.e., a chloride ion concentration of 165 g/L. To simulate the effect of the seawater pressure on chloride ion transport in the ITTs, the model was placed in a special corrosion vessel, and Materials 2023, 16, 3300 5 of 16 a pressurized device was used to maintain 6.5 atmospheres of pressure inside the vessel (comparable to the seawater pressure in the ITT of the Hong Kong-Zhuhai-Macao Bridge located at 40 m underwater) [36], as shown in Figure 2.

Electrochemical Test Methods
Electrochemical tests were performed at room temperature using a PARSTAT 2273 workstation to measure the self-corrosion potential, linear polarization curves, and AC impedance spectra. Pretest preparation: the ITT models to be tested were removed from the pressurized vessel, placed in a saturated calcium hydroxide solution (where the middle 200 mm test area on the side was just submerged in the solution), and left for 8 h before conducting the electrochemical tests, as well as electromagnetic shielding using a copper mesh cover during the electrochemical tests. The electrochemical test system was a threeelectrode system with a saturated calomel electrode (SCE) as the reference electrode (RE), an external titanium mesh as the auxiliary electrode (AE), and a test rebar as the working electrode (WE). The actual corrosion area of a single test rebar in this experiment was calculated to be 50.24 cm 2 . The length, width, and thickness of AE were 150 mm, 150 mm, and 2 mm, respectively, and its exposed area was 705.56 cm 2 . The exposed area of the AE was about 14.0 times the area of the WE.

Polarization Curves
According to the electrochemical weak polarization theory of metal corrosion, R p can be derived from the slope of the curve for the potential ∆E versus the current density ∆i at the corrosion potential (E−E corr ) (shown in Figure 3): based on sodium chloride at 10 times the concentration of simulated seawater, i.e., a chloride ion concentration of 165 g/L. To simulate the effect of the seawater pressure on chloride ion transport in the ITTs, the model was placed in a special corrosion vessel, and a pressurized device was used to maintain 6.5 atmospheres of pressure inside the vessel (comparable to the seawater pressure in the ITT of the Hong Kong-Zhuhai-Macao Bridge located at 40 m underwater) [36], as shown in Figure 2.

Electrochemical Test Methods
Electrochemical tests were performed at room temperature using a PARSTAT 2273 workstation to measure the self-corrosion potential, linear polarization curves, and AC impedance spectra. Pretest preparation: the ITT models to be tested were removed from the pressurized vessel, placed in a saturated calcium hydroxide solution (where the middle 200 mm test area on the side was just submerged in the solution), and left for 8 h before conducting the electrochemical tests, as well as electromagnetic shielding using a copper mesh cover during the electrochemical tests. The electrochemical test system was a threeelectrode system with a saturated calomel electrode (SCE) as the reference electrode (RE), an external titanium mesh as the auxiliary electrode (AE), and a test rebar as the working electrode (WE). The actual corrosion area of a single test rebar in this experiment was calculated to be 50.24 cm 2 . The length, width, and thickness of AE were 150 mm, 150 mm, and 2 mm, respectively, and its exposed area was 705.56 cm 2 . The exposed area of the AE was about 14.0 times the area of the WE.

Polarization Curves
According to the electrochemical weak polarization theory of metal corrosion, Rp can be derived from the slope of the curve for the potential ΔE versus the current density Δi at the corrosion potential (E−Ecorr) (shown in Figure 3): The calculated Rp can be used to determine the icorr using the Stern-Geary equation [37]: where ΔE is the applied polarization potential in mV; Δi is the corresponding difference in the current density before and after polarization in µA/cm 2 ; Rp is the polarization The calculated R p can be used to determine the i corr using the Stern-Geary equation [37]: where ∆E is the applied polarization potential in mV; ∆i is the corresponding difference in the current density before and after polarization in µA/cm 2 ; R p is the polarization resistance in Ω·cm 2 ; i corr is the corrosion current density in µA/cm 2 ; B is the Stern-Geary constant (which is taken as 52 mV and 26 mV when the test reinforcement in the concrete is in passivated and rust-activated states, respectively) [35]; and β a and β c are the Tafel constants for the anodic and cathodic polarization processes, respectively.
After the ITT models were eroded by pressurized seawater for 180 d and 365 d, linear polarization curves were generated by performing tests with a potential scan interval of E corr ±10 mV and a scan rate of 0.166 mV/s.

AC Impedance Spectra
After the ITT models were eroded by pressurized seawater for 180 d and 365 d, AC impedance spectra (Nyquist plots) were measured by performing tests at E corr with the scan frequency interval set to 10 mHz-100 kHz and an applied AC voltage of 10 mV. The AC impedance spectra were then calculated by ZSimpWin V3.3 software, and the characteristics of the Nyquist plot with the best fit were used to select the equivalent electric circuit (EEC) shown in Figure 4 for fitting calculations, where the rebars were in passivation and corrosion states corresponding to R s (C c (R c (Q dl (R ct )))) and R s (C c (R c (Q dl (R ct W)))) [38,39], to determine the key electrochemical parameters associated with the corrosion of the rebars.
constant (which is taken as 52 mV and 26 mV when the test reinforcement in the con is in passivated and rust-activated states, respectively) [35]; and βa and βc are the Tafe stants for the anodic and cathodic polarization processes, respectively.
After the ITT models were eroded by pressurized seawater for 180 d and 365 d, l polarization curves were generated by performing tests with a potential scan interv Ecorr ±10 mV and a scan rate of 0.166 mV/s.

AC Impedance Spectra
After the ITT models were eroded by pressurized seawater for 180 d and 365 d impedance spectra (Nyquist plots) were measured by performing tests at Ecorr wit scan frequency interval set to 10 mHz-100 kHz and an applied AC voltage of 10 mV AC impedance spectra were then calculated by ZSimpWin V3.3 software, and the ch teristics of the Nyquist plot with the best fit were used to select the equivalent el circuit (EEC) shown in Figure 4 for fitting calculations, where the rebars were in sivation and corrosion states corresponding to Rs(Cc(Rc(Qdl(Rct)))) and Rs(Cc(Rc(Qdl(Rct [38,39], to determine the key electrochemical parameters associated with the corrosi the rebars. In Figure 4, Rs denotes the solution resistance; in the high-frequency region, the constants are CcRc (where Cc denotes the capacitance of the concrete protective laye Rc denotes the resistance of the concrete protective layer) and QdlRct, considering the homogeneity of both the concrete and the reinforcing steel surface (where the con phase angle element CPE (denoted as Qdl) is used instead of the ideal double electric capacitance); Rct denotes the resistance to electron migration in the reinforcing steel to crete interfacial region, which is affected by the passivation state of the reinforcing surface; and in the low-frequency region, W denotes the Warburg element introduce diffusion control of corrosion of reinforcing steel. Figure 5a shows the Ecorr distribution for the rebars in the outer and inner sides o ITT models. At 180 d of pressurized seawater erosion, the mean Ecorr of the outer reba the ITT models was −434 mV with a variance of 350, and the mean Ecorr of the inner r was −294 mV with a variance of 148. At 180 d of erosion, the outer and inner rebars i ITT models had stable Ecorr distributions and were in similar corrosion states. Comp to the inner rebars, the mean Ecorr of the outer rebars was negative 140 mV and 48% hi indicating a clear tendency for the outer bars toward corrosion by seawater environ erosion. The relationship between Ecorr and the corrosion state of rebars in ASTM C 2015 [40] was used to determine that at 180 d of erosion, the outer rebars had an Eco nificantly lower than −276 mV and a probability of rebar corrosion of more than whereas the inner rebars had an Ecorr of approximately −276 mV, where the probabil In Figure 4, R s denotes the solution resistance; in the high-frequency region, the time constants are C c R c (where C c denotes the capacitance of the concrete protective layer and R c denotes the resistance of the concrete protective layer) and Q dl R ct , considering the nonhomogeneity of both the concrete and the reinforcing steel surface (where the constant phase angle element CPE (denoted as Q dl ) is used instead of the ideal double electric layer capacitance); R ct denotes the resistance to electron migration in the reinforcing steel to concrete interfacial region, which is affected by the passivation state of the reinforcing steel surface; and in the low-frequency region, W denotes the Warburg element introduced by diffusion control of corrosion of reinforcing steel. Figure 5a shows the E corr distribution for the rebars in the outer and inner sides of the ITT models. At 180 d of pressurized seawater erosion, the mean E corr of the outer rebars in the ITT models was −434 mV with a variance of 350, and the mean E corr of the inner rebars was −294 mV with a variance of 148. At 180 d of erosion, the outer and inner rebars in the ITT models had stable E corr distributions and were in similar corrosion states. Compared to the inner rebars, the mean E corr of the outer rebars was negative 140 mV and 48% higher, indicating a clear tendency for the outer bars toward corrosion by seawater environment erosion. The relationship between E corr and the corrosion state of rebars in ASTM C876-2015 [40] was used to determine that at 180 d of erosion, the outer rebars had an E corr significantly lower than −276 mV and a probability of rebar corrosion of more than 90%, whereas the inner rebars had an E corr of approximately −276 mV, where the probability of rebar corrosion could not be determined. Thus, corrosion of the outer rebars in the ITT models was concluded to be significantly stronger than that of the inner rebars. rebar corrosion could not be determined. Thus, corrosion of the outer rebars in the ITT models was concluded to be significantly stronger than that of the inner rebars.  Figure 5b shows that at 365 d of erosion by pressurized seawater, the mean Ecorr of the outer rebars was −461 mV with a variance of 143, and the mean Ecorr of the inner rebars was −380 mV with a variance of 150. At 365 d of erosion, the outer and inner rebars had stable Ecorr distributions and similar corrosion states, where the mean Ecorr of the outer rebars was negative 81 mV and 21% higher compared with the inner rebars, indicating that the outer rebars were more severely corroded than the inner rebars by pressurized seawater erosion. The relationship between the Ecorr of the rebars and the corrosion state in ASTM C876-2015 [40] was used to determine that at 365 d of erosion, Ecorr of both the inner and outer rebars in the ITT models was significantly lower than 276 mV and the probability of rebar corrosion was more than 90%. Figure 6 shows the typical linear polarization curves of the outer and inner rebars in the ITT models at 180 d and 365 d of pressurized seawater erosion. The linear polarization curves for the outer and inner rebars were used to calculate the icorr of the rebars using Equations (1) and (2), and the calculation results were shown in Figure 7.   Figure 5b shows that at 365 d of erosion by pressurized seawater, the mean E corr of the outer rebars was −461 mV with a variance of 143, and the mean E corr of the inner rebars was −380 mV with a variance of 150. At 365 d of erosion, the outer and inner rebars had stable E corr distributions and similar corrosion states, where the mean E corr of the outer rebars was negative 81 mV and 21% higher compared with the inner rebars, indicating that the outer rebars were more severely corroded than the inner rebars by pressurized seawater erosion. The relationship between the E corr of the rebars and the corrosion state in ASTM C876-2015 [40] was used to determine that at 365 d of erosion, E corr of both the inner and outer rebars in the ITT models was significantly lower than 276 mV and the probability of rebar corrosion was more than 90%. Figure 6 shows the typical linear polarization curves of the outer and inner rebars in the ITT models at 180 d and 365 d of pressurized seawater erosion. The linear polarization curves for the outer and inner rebars were used to calculate the i corr of the rebars using Equations (1) and (2), and the calculation results were shown in Figure 7. models was concluded to be significantly stronger than that of the inner rebars.  Figure 5b shows that at 365 d of erosion by pressurized seawater, the mean Ecorr of the outer rebars was −461 mV with a variance of 143, and the mean Ecorr of the inner rebars was −380 mV with a variance of 150. At 365 d of erosion, the outer and inner rebars had stable Ecorr distributions and similar corrosion states, where the mean Ecorr of the outer rebars was negative 81 mV and 21% higher compared with the inner rebars, indicating that the outer rebars were more severely corroded than the inner rebars by pressurized seawater erosion. The relationship between the Ecorr of the rebars and the corrosion state in ASTM C876-2015 [40] was used to determine that at 365 d of erosion, Ecorr of both the inner and outer rebars in the ITT models was significantly lower than 276 mV and the probability of rebar corrosion was more than 90%.     [35], which was in a medium corrosion rate state, and the corrosion rate of the rebars was 0.0053 mm/year. The i corr of the inner rebars ranged from 0.218 to 0.290 µA/cm 2 with a mean value of 0.260 µA/cm 2 and a variance of 0.0004. The mean i corr value of the inner rebars was higher than the critical value of 0.1 µA/cm 2 [35], and the rebars were in a low corrosion rate state with a corrosion rate of 0.0030 mm/year. The average corrosion rate of the outer rebars was 1.75 times higher than that of the inner rebars, i.e., the corrosion rate of the outer rebars was significantly higher than that of the inner rebars in the early process of corrosion in the ITT models.  Figure 7a shows icorr of the outer and inner rebars in the ITT models at 180 d and 365 d of pressurized seawater erosion. At 180 d of erosion, the icorr of the outer rebars in the ITT models ranged from 0.365 to 0.562 µA/cm 2 with a mean value of 0.455 µA/cm 2 and a variance of 0.0038. The mean icorr of the outer rebars was close to 0.5 µA/cm 2 [35], which was in a medium corrosion rate state, and the corrosion rate of the rebars was 0.0053 mm/year. The icorr of the inner rebars ranged from 0.218 to 0.290 µA/cm 2 with a mean value of 0.260 µA/cm 2 and a variance of 0.0004. The mean icorr value of the inner rebars was higher than the critical value of 0.1 µA/cm 2 [35], and the rebars were in a low corrosion rate state with a corrosion rate of 0.0030 mm/year. The average corrosion rate of the outer rebars was 1.75 times higher than that of the inner rebars, i.e., the corrosion rate of the outer rebars was significantly higher than that of the inner rebars in the early process of corrosion in the ITT models. Figure 7b shows icorr of the outer and inner rebars in the ITT models at 365 d of pressurized seawater erosion. At 365 d of erosion, the icorr of the outer bars in the ITT models ranged from 0.395 to 0.560 µA/cm 2 with a mean value of 0.479 µA/cm 2 and a variance of 0.0023, which was in a medium corrosion rate state, and the corrosion rate of the rebars was 0.0056 mm/year. The icorr of the inner bars ranged from 0.298 to 0.385 µA/cm 2 with a mean value of 0.338 µA/cm 2 and a variance of 0.0006, which was in a low corrosion rate state, and the corrosion rate of the rebars was 0.0039 mm/year. At 365 d of erosion, the average corrosion rate of the outer rebars in the ITT models was 1.43 times higher than that of the inner rebars, indicating that the corrosion rate of the outer rebars was still higher than that of the inner rebars. By comparing the corrosion rates of outer rebars and inner rebars in the ITT models, after erosion for 180 d and 365 d, the corrosion rate of outer rebars was always higher than that of inner rebars, and the corrosion rate of outer rebars remained at a higher rate as the corrosion time grew, while the corrosion rate of inner rebars gradually increased.

Linear Polarization Curves
The analysis presented above shows that although both the outer and inner rebars were corroded at 180 d of erosion, the corrosion rate of the outer rebars was 1.75 times that of the inner rebars. This phenomenon occurred because the outer side of the ITT model was in direct contact with seawater, which obeys Fick's second law of diffusion while penetrating into concrete, such that chloride ions continuously accumulated at the outer rebar surface and reached the critical content of 0.05% (mass percentage) [35] for the occurrence of corrosion earlier than the inner rebar. The corrosion of the rebars in concrete structures in a marine environment is an electrochemical process [11][12][13]41]: after the dense passivation film on the surface of the rebars is destroyed, the Fe substrate acts as the anode and loses electrons to generate Fe 2+ , whereby the substrate is dissolved and consumed; O2 and H2O in the solution on the surface of the reinforcing steel act as the cathode and gain electrons to generate OH − ; at the same time, the Cl − ions that have penetrated into the concrete from the eroding environment preferentially combine with Fe 2+ to form  Figure 7b shows i corr of the outer and inner rebars in the ITT models at 365 d of pressurized seawater erosion. At 365 d of erosion, the i corr of the outer bars in the ITT models ranged from 0.395 to 0.560 µA/cm 2 with a mean value of 0.479 µA/cm 2 and a variance of 0.0023, which was in a medium corrosion rate state, and the corrosion rate of the rebars was 0.0056 mm/year. The i corr of the inner bars ranged from 0.298 to 0.385 µA/cm 2 with a mean value of 0.338 µA/cm 2 and a variance of 0.0006, which was in a low corrosion rate state, and the corrosion rate of the rebars was 0.0039 mm/year. At 365 d of erosion, the average corrosion rate of the outer rebars in the ITT models was 1.43 times higher than that of the inner rebars, indicating that the corrosion rate of the outer rebars was still higher than that of the inner rebars. By comparing the corrosion rates of outer rebars and inner rebars in the ITT models, after erosion for 180 d and 365 d, the corrosion rate of outer rebars was always higher than that of inner rebars, and the corrosion rate of outer rebars remained at a higher rate as the corrosion time grew, while the corrosion rate of inner rebars gradually increased.
The analysis presented above shows that although both the outer and inner rebars were corroded at 180 d of erosion, the corrosion rate of the outer rebars was 1.75 times that of the inner rebars. This phenomenon occurred because the outer side of the ITT model was in direct contact with seawater, which obeys Fick's second law of diffusion while penetrating into concrete, such that chloride ions continuously accumulated at the outer rebar surface and reached the critical content of 0.05% (mass percentage) [35] for the occurrence of corrosion earlier than the inner rebar. The corrosion of the rebars in concrete structures in a marine environment is an electrochemical process [11][12][13]41]: after the dense passivation film on the surface of the rebars is destroyed, the Fe substrate acts as the anode and loses electrons to generate Fe 2+ , whereby the substrate is dissolved and consumed; O 2 and H 2 O in the solution on the surface of the reinforcing steel act as the cathode and gain electrons to generate OH − ; at the same time, the Cl − ions that have penetrated into the concrete from the eroding environment preferentially combine with Fe 2+ to form soluble FeCl 2 , which accelerates the dissolution process of the iron matrix and increases the corrosion rate of the rebars.
During pressurized seawater erosion, the measured relative humidity (RH) of the air inside the cavity of the ITT models was maintained at approximately 97%, i.e., the seawater completely soaked through the tunnel wall, and the outer and inner rebar surfaces of the ITT models were in a water-saturated state. Considering the environmental conditions, the linear polarization curves show that the outer and inner rebars of the tunnel model were under anodic polarization control and that during the process of Cl − transport from the outer to the inner side of the tunnel, the Cl − content was higher on the surface of the outer rebars than on that of the inner rebars, de-passivation of the passivation film and rusting occurred for the outer rebars earlier than for the inner rebars, and the corrosion rate of the outer rebars was significantly higher than that of the inner rebars. At 180 d of erosion, both E corr and i corr of the inner rebars showed that the rebars were likely to have been de-passivated and had started to rust, but the surface Cl − content was slightly higher than the critical value of 0.05% [35] and the corrosion rate of the rebars was low.
Through the analysis above, it can be seen that in ITTs, the reinforcing steel on the seawater side is weaker in durability than that on the cavity side. Therefore, in order to prolong the service life of ITTs, it is recommended to give priority to anti-corrosion methods for the reinforcing steel on the seawater side in the durability design of ITTs, such as applying polyurethane to the concrete surface on the seawater side to prevent seawater ingress [42], applying electrochemical rehabilitation to remove the sources of corrosion [43], and using cathodic protection [24,25] or surface coating epoxy [24,26] for the reinforcing steel on the seawater side. Given the large scale and high construction cost of ITT, the results of this work provide a theoretical basis for the adoption of a partial durability improvement method for the structure. Compared with the currently commonly used durability improvement method for the whole structure, the partial durability improvement method for the structure is more targeted and can reduce the construction and maintenance costs. Figure 8 shows the measured Nyquist plots of the ITT models after 180 d and 365 d of erosion in pressurized seawater. The Nyquist plots of each rebar exhibit an incomplete semicircular capacitance arc in both the high-and low-frequency regions and possibly a lowfrequency tail. The diameters of the high-and low-frequency capacitance arcs correspond to the resistance of the concrete protective layer (R c ) and the resistance to electron migration in the rebar-concrete interfacial region (R ct ), and the low-frequency tail represents the material diffusion W for the corrosion reaction on the rebar surface. The Nyquist plots obtained at 180 d and 365 d of erosion were compared: the diameters of the high-frequency capacitive arcs for both the inner and outer rebars exhibited only a small increase with erosion age, indicating that R c increased slowly, which originated from the continuous hydration of the concrete protective layer and the slow increase in the resistance of the concrete protective layer; by comparison, the diameter of the low-frequency capacitive arc exhibited a significant decrease with erosion age, indicating R ct continuously decreased, the passivation film on the rebar surface was destroyed, and the corrosion rate of the rebars increased significantly.

AC Impedance Spectra
To investigate the effect of the erosion age on the corrosion behavior of the inner and outer rebars of the ITT models, the EEC shown in Figure 4 was used to fit the Nyquist plots with an erosion age of 180 d, i.e., Figure 8a,c, and the results obtained from the fitting were listed in Table 5, according to which changes in the key parameters for electrochemical corrosion of the ITT models subjected to seawater erosion were analyzed. Figure 9 shows the concrete resistance R c , resistance to electron migration R ct , and double-layer capacitance Q dl in the rebar−concrete interfacial region and the dispersion coefficient n obtained by fitting the Nyquist plots of the inner and outer rebars with the EEC. As shown in Figure 9a, at 180 d of erosion, the mean R c of the outer rebars in the ITT models was 15.2 kΩ·cm 2 with a variance of 9.896, while the mean R c of the inner rebars was 20.8 kΩ·cm 2 with a variance of 12.888. The R c of the outer rebars was 26.9% smaller than that of the inner rebars. The R c of the outer rebars was smaller than that of the inner rebars due to the difference in the test current path between the WE and AE, and the electrode test showed a "concrete protective layer" effect, i.e., the thickness of the concrete protective layer of the outer rebars was smaller than that of the inner rebars, and the contribution of the concrete to the polarization resistance was smaller for the outer rebars than for the inner rebars. To investigate the effect of the erosion age on the corrosion behavior of the inner and outer rebars of the ITT models, the EEC shown in Figure 4 was used to fit the Nyquist plots with an erosion age of 180 d, i.e., Figure 8a,c, and the results obtained from the fitting were listed in Table 5, according to which changes in the key parameters for electrochemical corrosion of the ITT models subjected to seawater erosion were analyzed. Table 5. Key parameters of the Nyquist plots fitted by EEC.  As shown in Figure 9a, at 180 d of erosion, the mean R c of the outer rebars in the ITT models was 15.2 kΩ·cm 2 with a variance of 9.896, while the mean R c of the inner rebars was 20.8 kΩ·cm 2 with a variance of 12.888. The R c of the outer rebars was 26.9% smaller than that of the inner rebars. The R c of the outer rebars was smaller than that of the inner rebars due to the difference in the test current path between the WE and AE, and the electrode test showed a "concrete protective layer" effect, i.e., the thickness of the concrete protective layer of the outer rebars was smaller than that of the inner rebars, and the contribution of the concrete to the polarization resistance was smaller for the outer rebars than for the inner rebars. Figure 9b shows that the mean R ct of the outer rebars in the ITT models at 180 d of erosion was 39.9 kΩ·cm 2 with a variance of 21.012, and the mean R ct of the inner rebars was 83.5 kΩ·cm 2 with a variance of 41.144. The mean R ct of the outer rebars was only 48% of that of the inner rebars, corresponding to a significant difference between the two values. In a marine environment, the surface passivation film of the rebar in concrete is destroyed under erosion by seawater penetration; the iron matrix exposed to a pore solution with a high concentration of chloride ions is gradually dissolved; the resistance of harmful ions to migration to the surface of the iron matrix decreases; the resistance of the electrons lost by the iron matrix to reach the concrete through the ion interfacial zone via the rebars decreases; and the corrosion rate of the rebars increases [9]. In this study, the R ct of the outer rebars was found to be significantly lower than that of the inner rebars. For at 180 d of erosion, the chloride ion concentration on the surface of the outer rebars had exceeded the corrosion threshold of 0.05% [35], and the passivation film was damaged, resulting in the significant reduction of resistance to migration of electrons lost by the iron matrix to the concrete, and the rebars were corroded. At the same time, for the inner rebars, the passivation film was relatively intact, and no corrosion or only slight corrosion may have occurred. Figure 9c shows that at 180 d of erosion, the mean Q dl of the outer rebars in the ITT models was 174/10 10 Ω/cm 2 ·s n with a variance of 81.684, and the mean Q dl of the inner rebars was 272/10 10 Ω/cm 2 ·s n with a variance of 177.340. Figure 9d shows that the mean n value of the outer rebars was 0.46 with a variance of 0.0026, which is lower than the critical value for steel corrosion of 0.6 [44], and the mean n value of the inner rebars was 0.80 with a variance of 0.0020, which is significantly higher than 0.6. During the erosion process, the passivation film of the rebars was destroyed and the surface unevenness increased significantly. The results for the electrochemical response process show that the n values decreased to approach and drop below the passivation critical reference value of 0.6 and the corrosion rate of the rebar continuously increased. The stable n value above 0.6 shows that a rebar is in a passivated state and corrosion is inhibited. The n values of the outer rebars in the ITT models were all below 0.6, indicating the occurrence of corrosion and a reduction in the surface flatness of all the bars, whereas all the n values of the inner rebars were above 0.6, indicating that the rebars may have remained in a passivated state.  Macias [45] found that R ct could be approximately equal to R p under certain conditions, but these two key parameters were different in most cases. Based on the data for the 32 sets of ITT models, the R ct data obtained from the AC impedance spectra were regressed against the R p data obtained from the linear polarization curves, and the results are shown in Figure 10.

Number of Rebars
The least squares method was used to obtain the following linear regression equation: The coefficient of resolvability of the regression equation was 0.984, and the linear ratio of R ct to R p was 1.228, indicating a significant linear relationship. In addition, Figure 10 shows that R p was always greater than R ct and that the linear ratio between the two was in the range of 1.008 to 2.036, which is close to the results of Zhu's study [46] on the electrochemical corrosion of reinforcing steel. As shown in Figure 9a, at 180 d of erosion, the mean Rc of the outer rebars in the ITT models was 15.2 kΩ·cm 2 with a variance of 9.896, while the mean Rc of the inner rebars was 20.8 kΩ·cm 2 with a variance of 12.888. The Rc of the outer rebars was 26.9% smaller than that of the inner rebars. The Rc of the outer rebars was smaller than that of the inner rebars due to the difference in the test current path between the WE and AE, and the electrode test showed a "concrete protective layer" effect, i.e., the thickness of the concrete protective layer of the outer rebars was smaller than that of the inner rebars, and the contribution of the concrete to the polarization resistance was smaller for the outer rebars than for the inner rebars. Figure 9b shows that the mean Rct of the outer rebars in the ITT models at 180 d of erosion was 39.9 kΩ·cm 2 with a variance of 21.012, and the mean Rct of the inner rebars was 83.5 kΩ·cm 2 with a variance of 41.144. The mean Rct of the outer rebars was only 48% of that of the inner rebars, corresponding to a significant difference between the two values. In a marine environment, the surface passivation film of the rebar in concrete is destroyed under erosion by seawater penetration; the iron matrix exposed to a pore solution with a high concentration of chloride ions is gradually dissolved; the resistance of harmful ions to migration to the surface of the iron matrix decreases; the resistance of the electrons lost by the iron matrix to reach the concrete through the ion interfacial zone via the rebars decreases; and the corrosion rate of the rebars increases [9]. In this study, the Rct of the outer rebars was found to be significantly lower than that of the inner rebars. For at 180 d of erosion, the chloride ion concentration on the surface of the outer rebars had exceeded the corrosion threshold of 0.05% [35], and the passivation film was damaged, resulting in the significant reduction of resistance to migration of electrons lost by the iron matrix to the concrete, and the rebars were corroded. At the same time, for the inner rebars, the passivation film was relatively intact, and no corrosion or only slight corrosion may have occurred.  Figure 9c shows that at 180 d of erosion, the mean Qdl of the outer rebars in the IT models was 174/10 10 Ω/cm 2 ·s n with a variance of 81.684, and the mean Qdl of the inner re bars was 272/10 10 Ω/cm 2 ·s n with a variance of 177.340. Figure 9d shows that the mean value of the outer rebars was 0.46 with a variance of 0.0026, which is lower than the critica value for steel corrosion of 0.6 [44], and the mean n value of the inner rebars was 0.80 wit a variance of 0.0020, which is significantly higher than 0.6. During the erosion process, th passivation film of the rebars was destroyed and the surface unevenness increased signif icantly. The results for the electrochemical response process show that the n values de creased to approach and drop below the passivation critical reference value of 0.6 and th corrosion rate of the rebar continuously increased. The stable n value above 0.6 shows tha a rebar is in a passivated state and corrosion is inhibited. The n values of the outer rebar in the ITT models were all below 0.6, indicating the occurrence of corrosion and a reduc tion in the surface flatness of all the bars, whereas all the n values of the inner rebars wer above 0.6, indicating that the rebars may have remained in a passivated state.
Macias [45] found that Rct could be approximately equal to Rp under certain cond tions, but these two key parameters were different in most cases. Based on the data for th 32 sets of ITT models, the Rct data obtained from the AC impedance spectra were regresse against the Rp data obtained from the linear polarization curves, and the results are show in Figure 10. The least squares method was used to obtain the following linear regression equa tion: The coefficient of resolvability of the regression equation was 0.984, and the linea ratio of Rct to Rp was 1.228, indicating a significant linear relationship. In addition, Figur 10 shows that Rp was always greater than Rct and that the linear ratio between the two wa in the range of 1.008 to 2.036, which is close to the results of Zhu's study [46] on the elec trochemical corrosion of reinforcing steel.

Conclusions
In this paper, the electrochemical corrosion behavior of reinforcing steel in the IT models was investigated through indoor simulated pressurized seawater erosion test and the following main conclusions were obtained.

•
In the ITT models, the corrosion of the rebars on the seawater side happened earlie Figure 10. The relationship between R p and R ct of the rebars in ITT models.

Conclusions
In this paper, the electrochemical corrosion behavior of reinforcing steel in the ITT models was investigated through indoor simulated pressurized seawater erosion tests, and the following main conclusions were obtained.

•
In the ITT models, the corrosion of the rebars on the seawater side happened earlier than the rebars on the cavity side, and the corrosion rate of the rebars on the seawater side was significantly higher than that of the rebars on the cavity side during the 365 d of erosion. The corrosion rate of the seawater side of the rebars maintained a high corrosion rate, while the corrosion rate of the cavity side of the rebars gradually increased with the growth of corrosion time.

•
The electrochemical corrosion rates of the rebars on both the seawater side and the cavity side of the ITT models were controlled by anodic polarization, which indicated that chloride ion transport played a key role in the corrosion of the rebars, while oxygen transport had a small effect on the corrosion of the rebars.

•
There was an obvious linear relationship between R p obtained by the linear polarization curves and R ct obtained by the AC impedance spectra in the reinforced concrete ITT models under the submarine environment, and the linear scale factor of 1.228 was obtained from the regression equation for both.

•
Considering the reinforcing steel on the seawater side in the ITT is weaker in durability than that on the cavity side, it is recommended in the durability design to give priority to anti-corrosion methods for the reinforcing steel on the seawater side in order to prolong the service life of the ITT, such as applying polyurethane to the concrete surface on the seawater side to prevent seawater ingress, and using cathodic protection or surface epoxy for the reinforcing steel on the seawater side. Given the large size and high construction cost of the ITT, the results of this work show that it is feasible and more reasonable and economical to adopt durability enhancement methods for parts rather than for the entirety of the structure, while the corresponding service life assessment methods and life cycle cost analysis deserve further study.