Microstructure and Corrosion of Cast Magnesium Alloy ZK60 in NaCl Solution

In this work, the effects of the microstructure and phase constitution of cast magnesium alloy ZK60 (Mg-5.8Zn-0.57Zr, element concentration in wt.%) on the corrosion behavior in aqueous NaCl (0.1 mol dm−3) were investigated by weight-loss measurements, hydrogen evolution tests, and electrochemical techniques. The alloy was found to be composed of α-Mg matrix, with large second-phase particles of MgZn2 deposited along grain boundaries and a Zr-rich region in the central area of the grains. The large second-phase particles and the Zr-rich regions were more stable than the Mg matrix, resulting in a strong micro-galvanic effect. A filiform corrosion was found. It originated from the second-phase particles in the grain boundary regions in the early corrosion period. The filaments gradually occupied most areas of the alloy surface, and the general corrosion rate decreased significantly. Corrosion pits were developed under filaments. The pit growth rate decreased over time; however, it was about eight times larger than the general corrosion rate. A schematic model is presented to illustrate the corrosion mechanism.

The microstructure of Mg alloys, especially their second phases, has an evident impact on their corrosion behavior [3,5]. The second phases of Mg alloys may have a dual role in their corrosion, i.e., a galvanic acceleration effect or a corrosion blocking effect, depending on their quantities and distribution [24][25][26]. In Mg-Al alloys, when the amount of aluminum (Al) is low (e.g., Mg-5Al), the β-phase (Mg 12 Al 17 ) is relatively discontinuous in the Mg matrix and mainly acts as a cathode

Microstructure Characterization
The cast ZK60 samples (10 × 10 × 10 mm 3 ) were sealed with epoxy resin (working area = 1.0 cm 2 ), ground with 2000 grit SiC paper, and polished with 3 µm diamond paste. Then, they were etched with a picric acid solution for metallographic analysis. The metallographic structure of the cast ZK60 sample was observed using a 3D optical microscope (VHX-1000, KEYENCE, Osaka, Japan), SEM (Quanta 200, FEI, Eindhoven, The Netherlands) equipped with EDX (EDAX-Genesis), and transmission electron microscopy (TEM). The TEM sample was firstly mechanically ground to a thickness of about 20 µm, and then, it was ion milled at 4 keV and 4 • , cooled by liquid nitrogen. TEM observations were carried out with an FEI Talos F200X transmission electron microscope (FEI, Portland, OR, USA) operated at 200 kV. The element content and elemental distribution of the cast ZK60 sample were characterized by EDX and an electron probe micro-analyzer (EPMA-8050G, SHIMADZU, Kyoto, Japan), respectively. Phase structure analysis was performed by XRD (X'Pert PRO, PANalytical B.V., Almelo, The Netherlands) using Cu Kα radiation. The scan range of 2θ was from 20 • to 90 • with a scan step of 0.02 • . The XRD pattern was analyzed with the X'Pert HighScore Plus software (2.0, PANalytical B.V., Almelo, The Netherlands). SKPFM (SPM-9700, SHIMADZU, Kyoto, Japan) was used to measure the relative Volta potential differences among different microstructural constituents to show their relative nobility. Meanwhile, the corresponding topography maps of the same area were also obtained.

Electrochemical Tests
All the electrochemical tests were carried out using a CS 350 Corrtest electrochemical workstation (Wuhan Corrtest, Wuhan, China). The working electrode (10 × 10 × 10 mm 3 ) was sealed with epoxy resin (working area = 1.0 cm 2 ), which was ground with 2000 grit SiC paper and rinsed in distilled water and ethanol. A saturated calomel electrode (SCE) and a platinum (Pt) electrode were used as the reference electrode and the counter electrode, respectively. Polarisation curves were generated and electrochemical impedance spectroscopy (EIS) was performed, respectively, at different corrosion times. At free corrosion potentials (E corr ), the EIS tests were performed with an AC voltage amplitude of 10 mV in the frequency range of 100 kHz-0.05 Hz. The EIS results were fitted with the Zview2.0 software. The polarization curves were generated at the scan rate of 0.5 mV s −1 , scanning towards the positive direction. The corrosion current density (i corr , mA cm −2 ) was estimated by the cathodic Tafel extrapolation method, according to [5,44]. The corresponding corrosion rate (P i , mm y −1 ) was converted by the equation [5,45]: P i = 22.06i corr (1) where the corrosion current density i corr is estimated by the Tafel extrapolation of the cathodic branch of the polarization curves, and P i is related to the average corrosion rate. All the electrochemical tests were repeated at least three times in this study.

Weight Loss Tests
The cast ZK60 specimens (20 × 20 × 4 mm 3 ) were ground with 2000 grit SiC paper, rinsed in distilled water and ethanol, dried with cold air, and kept in a vacuum desiccator before the weight loss test. The specimens were corroded in 0.1 M NaCl for different times and then immersed in a CrO 3 solution (180 g L −1 ,~25 • C) for 10 min to remove corrosion products. At least three parallel tests were performed under each test condition. The average weight loss rate of the cast ZK60 alloy (∆W, mg cm −2 h −1 ) can be converted to a general corrosion rate P w (mm y −1 ) using [5,44]: where ρ is the metal density (g cm −3 ). For the cast ZK60 alloy, ρ = 1.8 g cm −3 ; thus, Equation (2) becomes: Materials 2020, 13, 3833 4 of 21

Hydrogen Evolution Tests
Following the work of Shi [46], plug-in specimens were employed to perform the hydrogen evolution test in 0.1 M NaCl at room temperature (~25 • C). Figure 1 presents the schematics of the test system. The cast ZK60 specimens (10 × 10 × 10 mm 3 ) were treated as the above weight loss samples. The evolved hydrogen was collected into a burette, and its volume (V H , mL cm −2 ) was recorded at different times. The hydrogen evolution rate, v H (mL·cm −2 day −1 ), can also be converted to a general corrosion rate, P H (mm y −1 ), using [5,44].
Materials 2020, 13, x FOR PEER REVIEW 4 of 22 recorded at different times. The hydrogen evolution rate, vH (mL·cm −2 ·day −1 ), can also be converted to a general corrosion rate, PH (mm y −1 ), using [5,44]. Figure 1. Schematic diagram of the test system for the hydrogen evolution measurement.

Corrosion Morphology Characterization
The etched metallographic specimens were placed in 0.1 M NaCl, where the solution thickness on the sample surface was about 1 mm, to observe the corrosion development in the initial period (0-1 h) in situ using a 3D optical microscope (KEYENCE VHX-1000). The corrosion morphologies of the cast ZK60 samples corroded in 0.1 M NaCl for different times were measured by SEM and with the 3D optical microscope. The cross-section corrosion morphologies and the elemental distributions were analyzed by EPMA (EPMA-8050G). Figure 2 shows the optical and the back-scattered electron (BSE) SEM micrographs of the cast ZK60 alloy. The microstructure of the cast ZK60 alloy was composed of an α-Mg phase and large second-phase particles, which were mainly deposited along the grain boundaries (Figure 2a,b). According to [47,48], the main components of these particles are Mg and Zn and may be MgZn2 and MgZn. The XRD pattern of the cast ZK60 alloy in Figure 3 only displays the presence of MgZn2. No discernable diffraction peaks from MgZn were detected in this work, which suggests no MgZn phase in the studied alloy; also, there is no Zr detected, maybe due to its low content in the test sample. However, the XRD patterns cannot give accurate structural information for the second phase. Therefore, the crystal structures of the abovementioned second phases were characterized using TEM in detail as follows.

Corrosion Morphology Characterization
The etched metallographic specimens were placed in 0.1 M NaCl, where the solution thickness on the sample surface was about 1 mm, to observe the corrosion development in the initial period (0-1 h) in situ using a 3D optical microscope (KEYENCE VHX-1000). The corrosion morphologies of the cast ZK60 samples corroded in 0.1 M NaCl for different times were measured by SEM and with the 3D optical microscope. The cross-section corrosion morphologies and the elemental distributions were analyzed by EPMA (EPMA-8050G).

Results and Discussion
3.1. Microstructure of the Cast ZK60 Alloy Figure 2 shows the optical and the back-scattered electron (BSE) SEM micrographs of the cast ZK60 alloy. The microstructure of the cast ZK60 alloy was composed of an α-Mg phase and large second-phase particles, which were mainly deposited along the grain boundaries (Figure 2a,b). According to [47,48], the main components of these particles are Mg and Zn and may be MgZn 2 and MgZn. The XRD pattern of the cast ZK60 alloy in Figure 3 only displays the presence of MgZn 2 . No discernable diffraction peaks from MgZn were detected in this work, which suggests no MgZn phase in the studied alloy; also, there is no Zr detected, maybe due to its low content in the test sample. However, the XRD patterns cannot give accurate structural information for the second phase. Therefore, the crystal structures of the abovementioned second phases were characterized using TEM in detail as follows.
To further verify the second phase in the as-cast ZK60 alloy, Figure 4 presents the TEM micrograph of the as-cast ZK60 alloy. Block-shaped and globular second-phase particles can be observed in Figure 4. The blocky phase has a size of about 500 nm, and the globular phase is about 100 nm. Figure 4 also presents the corresponding selected area electron diffraction (SAED) patterns of the second phase. The SAED patterns proved that the two second-phase particles were MgZn 2 [49]. No MgZn phase was found in the as-cast ZK60 alloy.  To further verify the second phase in the as-cast ZK60 alloy, Figure 4 presents the TEM micrograph of the as-cast ZK60 alloy. Block-shaped and globular second-phase particles can be observed in Figure 4. The blocky phase has a size of about 500 nm, and the globular phase is about 100 nm. Figure 4 also presents the corresponding selected area electron diffraction (SAED) patterns of the second phase. The SAED patterns proved that the two second-phase particles were MgZn2 [49]. No MgZn phase was found in the as-cast ZK60 alloy.  To further verify the second phase in the as-cast ZK60 alloy, Figure 4 presents the TEM micrograph of the as-cast ZK60 alloy. Block-shaped and globular second-phase particles can be observed in Figure 4. The blocky phase has a size of about 500 nm, and the globular phase is about 100 nm. Figure 4 also presents the corresponding selected area electron diffraction (SAED) patterns of the second phase. The SAED patterns proved that the two second-phase particles were MgZn2 [49]. No MgZn phase was found in the as-cast ZK60 alloy. The BSE-SEM micrograph in Figure 2b indicates the nonuniform distribution of the alloying elements in the cast ZK60 because the brighter areas contain more elements of higher atomic weight than the darker regions [50]. Figure 5 presents the area distribution of the alloying elements in the cast ZK60, proving that the center area of the grains with light color was richer in Zr and Zn (Zr-rich region) than the neighboring darker zones (grain boundary region). Here, the "grain boundary region" of the as-cast ZK60 is denoted as the areas between the Zr-rich region, i.e., the dark areas in Figure 2b. This uneven distribution of the alloying elements in the cast ZK60 alloy may cause the inhomogeneous electrochemical activity resulting in the micro-galvanic corrosion [51]. Figure 6 presents the SKPFM maps of the cast ZK60 sample, which clearly show that the second phase had the highest potential. Furthermore, the Volta potential profiles (Figure 6c) along the line A and line B indicated that the central region of the grains (i.e., Zr-rich region) exhibited higher potential than those of the grain boundary regions, owing to the higher Zr and Zn contents in the center of grains and higher Mg content in the grain boundary regions ( Figure 5). Thus, the second-phase particles and the central region of the grains should be more stable than the Mg matrix in grain boundary regions and more likely to become cathodes in the micro-galvanic cells. A second-phase particle occurred in the grain area (Figure 6a), which is consistent with Figure 2a, so the micro-galvanic corrosion may also have been initiated in the grains.  The BSE-SEM micrograph in Figure 2b indicates the nonuniform distribution of the alloying elements in the cast ZK60 because the brighter areas contain more elements of higher atomic weight than the darker regions [50]. Figure 5 presents the area distribution of the alloying elements in the cast ZK60, proving that the center area of the grains with light color was richer in Zr and Zn (Zr-rich region) than the neighboring darker zones (grain boundary region). Here, the "grain boundary region" of the as-cast ZK60 is denoted as the areas between the Zr-rich region, i.e., the dark areas in Figure 2b.This uneven distribution of the alloying elements in the cast ZK60 alloy may cause the inhomogeneous electrochemical activity resulting in the micro-galvanic corrosion [51].  Figure 6 presents the SKPFM maps of the cast ZK60 sample, which clearly show that the second phase had the highest potential. Furthermore, the Volta potential profiles (Figure 6c) along the line A and line B indicated that the central region of the grains (i.e., Zr-rich region) exhibited higher potential than those of the grain boundary regions, owing to the higher Zr and Zn contents in the   Figure 7 shows the ΔW (mg·cm −2 ·h −1 ) and Pw (mm y −1 ) values of the cast ZK60 alloy immersed in 0.1 M NaCl for different times (t). The Pw at 24 h was the largest (3.5 ± 0.1 mm y −1 ), which is similar to that of a cast ZK60 in 0.9% NaCl (4.6 ± 0.6 mm y −1 ) reported by Merson et al. [52]. Then, it gradually decreased by approximately half to a relatively stable value (1.4~1.7 mm y −1 ) over 72-96 h. The decrease in Pw over 72-96 h may have been related to the increase in the corrosion product layer Mg(OH)2 on its surface [53].  Figure 8 presents the hydrogen evolution test results for the cast ZK60 alloy to show the change in its corrosion rate in detail, in which vH is the differentiation of the VH-t curve in Figure 8a and PH is calculated by Equation (4). The VH at 24 h ( Figure 8a) was about 0.7 ± 0.03 mL cm −2 , which is higher than that of an extruded ZK60 (~0.5 mL cm −2 ) [54]. The PH value ( Figure 8b) increased with time in the initial corrosion period (0-1.5 h) and then decreased over 1.5-48 h; at last, it increased again over 48-72 h. These results imply that there exist different corrosion stages in the periods of 0-2 h, 2-48 h, and 48-72 h, which may be related to the change in the alloy surface condition. According to Song [24], the total volume of hydrogen collected should equal the total amount of metal lost, and both the weight loss and hydrogen evolution tests were reliable methods. Even though the hydrogen evolution rate PH we measured is slightly lower than the weight loss corrosion rate Pw, which may be caused by the difference in the test methods, they generally show similar change tendencies.  Figure 7 shows the ∆W (mg cm −2 h −1 ) and P w (mm y −1 ) values of the cast ZK60 alloy immersed in 0.1 M NaCl for different times (t). The P w at 24 h was the largest (3.5 ± 0.1 mm y −1 ), which is similar to that of a cast ZK60 in 0.9% NaCl (4.6 ± 0.6 mm y −1 ) reported by Merson et al. [52]. Then, it gradually decreased by approximately half to a relatively stable value (1.4~1.7 mm y −1 ) over 72-96 h. The decrease in P w over 72-96 h may have been related to the increase in the corrosion product layer Mg(OH) 2 on its surface [53].  Figure 7 shows the ΔW (mg·cm −2 ·h −1 ) and Pw (mm y −1 ) values of the cast ZK60 alloy immersed in 0.1 M NaCl for different times (t). The Pw at 24 h was the largest (3.5 ± 0.1 mm y −1 ), which is similar to that of a cast ZK60 in 0.9% NaCl (4.6 ± 0.6 mm y −1 ) reported by Merson et al. [52]. Then, it gradually decreased by approximately half to a relatively stable value (1.4~1.7 mm y −1 ) over 72-96 h. The decrease in Pw over 72-96 h may have been related to the increase in the corrosion product layer Mg(OH)2 on its surface [53].  Figure 8 presents the hydrogen evolution test results for the cast ZK60 alloy to show the change in its corrosion rate in detail, in which vH is the differentiation of the VH-t curve in Figure 8a and PH is calculated by Equation (4). The VH at 24 h (Figure 8a) was about 0.7 ± 0.03 mL cm −2 , which is higher than that of an extruded ZK60 (~0.5 mL cm −2 ) [54]. The PH value ( Figure 8b) increased with time in the initial corrosion period (0-1.5 h) and then decreased over 1.5-48 h; at last, it increased again over 48-72 h. These results imply that there exist different corrosion stages in the periods of 0-2 h, 2-48 h, and 48-72 h, which may be related to the change in the alloy surface condition. According to Song [24], the total volume of hydrogen collected should equal the total amount of metal lost, and both the weight loss and hydrogen evolution tests were reliable methods. Even though the hydrogen evolution rate PH we measured is slightly lower than the weight loss corrosion rate Pw, which may be caused by the difference in the test methods, they generally show similar change tendencies.  Figure 8 presents the hydrogen evolution test results for the cast ZK60 alloy to show the change in its corrosion rate in detail, in which v H is the differentiation of the V H -t curve in Figure 8a and P H is calculated by Equation (4). The V H at 24 h (Figure 8a) was about 0.7 ± 0.03 mL cm −2 , which is higher than that of an extruded ZK60 (~0.5 mL cm −2 ) [54]. The P H value ( Figure 8b) increased with time in the initial corrosion period (0-1.5 h) and then decreased over 1.5-48 h; at last, it increased again over 48-72 h. These results imply that there exist different corrosion stages in the periods of 0-2 h, 2-48 h, and 48-72 h, which may be related to the change in the alloy surface condition. According to Song [24], the total volume of hydrogen collected should equal the total amount of metal lost, and both the weight loss and hydrogen evolution tests were reliable methods. Even though the hydrogen evolution rate P H we measured is slightly lower than the weight loss corrosion rate P w , which may be caused by the difference in the test methods, they generally show similar change tendencies.  Figure 9 presents the polarization curves of the cast ZK60 alloy after immersion in 0.1 M NaCl for different times and the changes in its corrosion rates (icorr and Pi) with time. All the polarization curves showed the typical features of activation-controlled processes [55,56]. A so-called breakdown potential (Ebreak) occurred in the anodic polarization curves (except t = 2 h), owing to the breakdown of the oxide film on the alloy [57], as shown in Figure 9b. Ecorr ( Figure 9a) and Ebreak ( Figure 9) moved positively in the period of 0-24 h and then became negative again, while Pi and icorr ( Figure 9b) displayed a similar change with corrosion time to that of PH as shown in Figure 8b, which is consistent with previous reports [38,58]. The icorr and Pi values in Figure 9b are much smaller than the Pw ( Figure  7) and PH (Figure 8b) values. Similar results from other Mg alloys have been discussed in detail in [44]. However, they displayed a similar change tendency. All these results also suggest that there may be different corrosion stages in the corrosion periods of 0-2 h, 2-24 h, and 24-72 h.  Figure 10 presents the EIS of the cast ZK60 alloy immersed in 0.1 M NaCl for different times and the change in Rp (polarization resistance) as a function of time. All the Nyquist plots display two capacitive loops in the high-frequency region and an inductive loop in the low-frequency region, which is similar to the EIS features of other Mg alloys [59]. The capacitive loops are related to the processes in the surface film and the electrical double layer [60], and the inductive loop should be attributed to the initiation of the localized corrosion on the surface of the cast ZK60 alloy according to [61][62][63]. Based on these EIS features, Figure 11 presents an equivalent circuit to fit the EIS results in Figure 10a Figure 9 presents the polarization curves of the cast ZK60 alloy after immersion in 0.1 M NaCl for different times and the changes in its corrosion rates (i corr and P i ) with time. All the polarization curves showed the typical features of activation-controlled processes [55,56]. A so-called breakdown potential (E break ) occurred in the anodic polarization curves (except t = 2 h), owing to the breakdown of the oxide film on the alloy [57], as shown in Figure 9b. E corr (Figure 9a) and E break (Figure 9) moved positively in the period of 0-24 h and then became negative again, while P i and i corr (Figure 9b) displayed a similar change with corrosion time to that of P H as shown in Figure 8b, which is consistent with previous reports [38,58]. The i corr and P i values in Figure 9b are much smaller than the P w ( Figure 7) and P H (Figure 8b) values. Similar results from other Mg alloys have been discussed in detail in [44]. However, they displayed a similar change tendency. All these results also suggest that there may be different corrosion stages in the corrosion periods of 0-2 h, 2-24 h, and 24-72 h.  Figure 9 presents the polarization curves of the cast ZK60 alloy after immersion in 0.1 M NaCl for different times and the changes in its corrosion rates (icorr and Pi) with time. All the polarization curves showed the typical features of activation-controlled processes [55,56]. A so-called breakdown potential (Ebreak) occurred in the anodic polarization curves (except t = 2 h), owing to the breakdown of the oxide film on the alloy [57], as shown in Figure 9b. Ecorr ( Figure 9a) and Ebreak ( Figure 9) moved positively in the period of 0-24 h and then became negative again, while Pi and icorr ( Figure 9b) displayed a similar change with corrosion time to that of PH as shown in Figure 8b, which is consistent with previous reports [38,58]. The icorr and Pi values in Figure 9b are much smaller than the Pw ( Figure  7) and PH (Figure 8b) values. Similar results from other Mg alloys have been discussed in detail in [44]. However, they displayed a similar change tendency. All these results also suggest that there may be different corrosion stages in the corrosion periods of 0-2 h, 2-24 h, and 24-72 h.  Figure 10 presents the EIS of the cast ZK60 alloy immersed in 0.1 M NaCl for different times and the change in Rp (polarization resistance) as a function of time. All the Nyquist plots display two capacitive loops in the high-frequency region and an inductive loop in the low-frequency region, which is similar to the EIS features of other Mg alloys [59]. The capacitive loops are related to the processes in the surface film and the electrical double layer [60], and the inductive loop should be attributed to the initiation of the localized corrosion on the surface of the cast ZK60 alloy according to [61][62][63]. Based on these EIS features, Figure 11 presents an equivalent circuit to fit the EIS results in Figure 10a Figure 10 presents the EIS of the cast ZK60 alloy immersed in 0.1 M NaCl for different times and the change in R p (polarization resistance) as a function of time. All the Nyquist plots display two capacitive loops in the high-frequency region and an inductive loop in the low-frequency region, which is similar to the EIS features of other Mg alloys [59]. The capacitive loops are related to the processes in the surface film and the electrical double layer [60], and the inductive loop should be attributed to the initiation of the localized corrosion on the surface of the cast ZK60 alloy according to [61][62][63]. Based on these EIS features, Figure 11 presents an equivalent circuit to fit the EIS results in Figure 10a [58,59,64,65]. R s is the solution resistance. CPE f and CPE dl represent the constant phase elements (CPE) for the surface film and the electrical double layer, respectively. R f and R ct represent the surface film resistance and charge-transfer resistance, respectively. R L and L represent equivalent resistance and inductance to describe the low-frequency inductance. It should be noted that R ct is the parallel of the charge-transfer resistance of the anodic process and the cathodic process (R ct,a and R ct,c ) at E corr [63]. The fitting curves are also displayed in Figure 10a, and the fitting parameters are listed in Table 2.

EIS Measurements
Materials 2020, 13, x FOR PEER REVIEW 10 of 22 the surface film resistance and charge-transfer resistance, respectively. RL and L represent equivalent resistance and inductance to describe the low-frequency inductance. It should be noted that Rct is the parallel of the charge-transfer resistance of the anodic process and the cathodic process (Rct,a and Rct,c) at Ecorr [63]. The fitting curves are also displayed in Figure 10a, and the fitting parameters are listed in Table 2.   According to [59,66,67] and the results in Table 2, the fitting curves can be extrapolated to the zero-frequency limit to obtain the polarization resistance (Rp) at different times. Therefore, the Rp-t curve was obtained as shown in Figure 10b. Rp decreased with time over 0-2 h, indicating corrosion acceleration, and then, Rp clearly increased between 2 and 24 h before decreasing slightly over 24-72 h. In this case, the change in Rp with t is consistent with that of Pi (Figure 9b). The Rt and Rf values in Table 2 show the same change tendencies as those for Rp. The change in the CPEdl, CPEf, RL, and L with t should be closely related to the surface layer of the cast ZK60 alloy [59]. The specific changes in these EIS results are discussed later. the surface film resistance and charge-transfer resistance, respectively. RL and L represent equivalent resistance and inductance to describe the low-frequency inductance. It should be noted that Rct is the parallel of the charge-transfer resistance of the anodic process and the cathodic process (Rct,a and Rct,c) at Ecorr [63]. The fitting curves are also displayed in Figure 10a, and the fitting parameters are listed in Table 2.   According to [59,66,67] and the results in Table 2, the fitting curves can be extrapolated to the zero-frequency limit to obtain the polarization resistance (Rp) at different times. Therefore, the Rp-t curve was obtained as shown in Figure 10b. Rp decreased with time over 0-2 h, indicating corrosion acceleration, and then, Rp clearly increased between 2 and 24 h before decreasing slightly over 24-72 h. In this case, the change in Rp with t is consistent with that of Pi (Figure 9b). The Rt and Rf values in Table 2 show the same change tendencies as those for Rp. The change in the CPEdl, CPEf, RL, and L with t should be closely related to the surface layer of the cast ZK60 alloy [59]. The specific changes in these EIS results are discussed later.  According to [59,66,67] and the results in Table 2, the fitting curves can be extrapolated to the zero-frequency limit to obtain the polarization resistance (R p ) at different times. Therefore, the R p -t curve was obtained as shown in Figure 10b. R p decreased with time over 0-2 h, indicating corrosion acceleration, and then, R p clearly increased between 2 and 24 h before decreasing slightly over 24-72 h. In this case, the change in R p with t is consistent with that of P i (Figure 9b). The R t and R f values in Table 2 show the same change tendencies as those for R p . The change in the CPE dl , CPE f , R L , and L with t should be closely related to the surface layer of the cast ZK60 alloy [59]. The specific changes in these EIS results are discussed later.

Corrosion Morphology of the Cast ZK60 Alloy
To observe the real-time corrosion development in the initial period (0-1 h), Figure 12 presents the in situ corrosion images of the etched sample after immersion in 0.1 M NaCl for different times. In the initial period (20 min, Figure 12a), there was no apparent corrosion, with a limited number of H 2 bubbles observed on the surface, implying a slight corrosion state. This slight corrosion should be related to the protective film on the alloy surface impeding the corrosion in the initial period [68]. Over 40-60 min (Figure 12b,c), some dark threads appeared on the surface, which lengthened with time and displayed filiform-like corrosion characteristics. In this period, H 2 bubbles constantly evolved at or near the leading edges of the dark threads, leaving cloudy trails behind them. The magnified picture in Figure 12d suggests that the corrosion filaments seem to traverse within the grain boundary regions and do not extend toward the center of the grains in the initial stage of expansion, which is further verified later. These corrosion characteristics are consistent with those of other Mg alloys in similar corrosion media and should be related to their microstructure features [50,65,69].

Corrosion Morphology of the Cast ZK60 Alloy
To observe the real-time corrosion development in the initial period (0-1 h), Figure 12 presents the in situ corrosion images of the etched sample after immersion in 0.1 M NaCl for different times. In the initial period (20 min, Figure 12a), there was no apparent corrosion, with a limited number of H2 bubbles observed on the surface, implying a slight corrosion state. This slight corrosion should be related to the protective film on the alloy surface impeding the corrosion in the initial period [68]. Over 40-60 min (Figure 12b,c), some dark threads appeared on the surface, which lengthened with time and displayed filiform-like corrosion characteristics. In this period, H2 bubbles constantly evolved at or near the leading edges of the dark threads, leaving cloudy trails behind them. The magnified picture in Figure 12d suggests that the corrosion filaments seem to traverse within the grain boundary regions and do not extend toward the center of the grains in the initial stage of expansion, which is further verified later. These corrosion characteristics are consistent with those of other Mg alloys in similar corrosion media and should be related to their microstructure features [50,65,69].   (Figure 13b) were surrounded by a small number of black corrosion products, implying that the corrosion filaments may have started from the surrounding areas of these second-phase particles. Most of the central regions of the grains, i.e., the Zr-rich areas with a light color (Figures 2b and 5), were still uncorroded. When t = 2 h (Figure 13c), only a small number of second-phase particles occurred in the corroded areas, implying that most of them were removed due to the dissolution of their surrounding Mg matrix. Some broad corrosion areas in Figure 13 show that the corrosion could develop to the central area of grains with increasing corrosion time. Moreover, the corrosion filaments in Figure 13c,d show an apparent corrosion depth, implying that the corrosion also developed under the black corrosion product layer.   [70] mainly occurred on the "darker zones" in grain boundary regions (see Figures 2b and 5) and encompassed the second-phase particles. Some second-phase particles in the boundary regions (Figure 13b) were surrounded by a small number of black corrosion products, implying that the corrosion filaments may have started from the surrounding areas of these second-phase particles. Most of the central regions of the grains, i.e., the Zr-rich areas with a light color (Figures 2b and 5), were still uncorroded. When t = 2 h (Figure 13c), only a small number of second-phase particles occurred in the corroded areas, implying that most of them were removed due to the dissolution of their surrounding Mg matrix. Some broad corrosion areas in Figure 13 show that the corrosion could develop to the central area of grains with increasing corrosion time. Moreover, the corrosion filaments in Figure 13c As shown in Figures 12 and 13, the etched samples were used to observe the origination and developmental features of the corrosion filaments on the cast ZK60. To avoid the influence of the etching treatment on the corrosion process, Figure 14 presents the secondary electron (SE) SEM images of the raw alloy sample after immersion in 0.1 M NaCl for 0.5 and 2 h. The corrosion morphologies in Figure 14a,b also exhibit the characteristics of filiform corrosion, similar to those in Figure 12b. The polishing scratches on the surface of the test samples do not appear to influence the initiation and extension of the corrosion filaments, suggesting that they are controlled by the microstructure of the cast ZK60 alloy. When t = 2 h, the corrosion images in Figure 14c,d display features similar to those in Figure 13c and some small corrosion pits occur in the corrosion areas, which may have been the result of the loss of the second phase-particles. The raw cast ZK60 sample also displayed the filiform corrosion features in the early corrosion period (0-2 h). As shown in Figures 12 and 13, the etched samples were used to observe the origination and developmental features of the corrosion filaments on the cast ZK60. To avoid the influence of the etching treatment on the corrosion process, Figure 14 presents the secondary electron (SE) SEM images of the raw alloy sample after immersion in 0.1 M NaCl for 0.5 and 2 h. The corrosion morphologies in Figure 14a,b also exhibit the characteristics of filiform corrosion, similar to those in Figure 12b. The polishing scratches on the surface of the test samples do not appear to influence the initiation and extension of the corrosion filaments, suggesting that they are controlled by the microstructure of the cast ZK60 alloy. When t = 2 h, the corrosion images in Figure 14c,d display features similar to those in Figure 13c and some small corrosion pits occur in the corrosion areas, which may have been the result of the loss of the second phase-particles. The raw cast ZK60 sample also displayed the filiform corrosion features in the early corrosion period (0-2 h). Figure 15 presents the secondary electron (SE) SEM images of the cast ZK60 alloy after immersion in 0.1 M NaCl for different times to observe the corrosion development over a long period (12-72 h). The corrosion filaments gradually extended to the whole surface of the test samples after 24 h, and the number and depth of the corrosion pits increased with time. These results further prove that with increasing corrosion time, the corrosion of the cast ZK60 alloy can develop toward the central area of grains and toward the depth of the alloy matrix, which is discussed below. Materials 2020, 13, x FOR PEER REVIEW 13 of 22

Characterization of the Corrosion Pits on the Cast ZK60 Alloy
To characterize the development of corrosion pits, the deepest pit on the cast ZK60 sample was selected after immersion in 0.1 M NaCl for different times, and their 3D morphologies and depth profiles were measured, as shown in Figure 16. The density and depth of the corrosion pits increased over 12-72 h, and the size of the pit mouth was generally larger than the pit depth, which is similar to the corrosion morphologies in Figure 15. The corresponding pit depth at different corrosion times was employed to calculate the average penetration rate (P depth , mm y −1 ), which is also presented in Figure 17. The pit depth increased with time, but the fastest P depth occurred in the period of 12-24 h (~24 mm y −1 ) before gradually decreasing to~16.6 mm y −1 in the period of 24-72 h.
to the corrosion morphologies in Figure 15. The corresponding pit depth at different corrosion times was employed to calculate the average penetration rate (Pdepth, mm y −1 ), which is also presented in Figure 17. The pit depth increased with time, but the fastest Pdepth occurred in the period of 12-24 h (~24 mm y −1 ) before gradually decreasing to ~16.6 mm y −1 in the period of 24-72 h. Figure 18 is the cross-sectional EPMA pattern of a corrosion pit on the cast ZK60 sample immersed in 0.1 M NaCl for 72 h. The corrosion pit was covered by the corrosion product, and the second-phase particles were encompassed in it. Only Mg, O, and Cl elements were found in the corrosion product layer, which should result from Mg(OH)2 and Cl − in the test solution. Moreover, a large number of Cl − through the whole corrosion product layer suggests that Cl − could penetrate the product layer easily and may become enriched in the bottom of the corrosion pit to propagate the pit corrosion [71]. The second-phase particles in the corrosion pit imply that they may be related to the formation of corrosion pits, which is discussed later.    Figure 18 is the cross-sectional EPMA pattern of a corrosion pit on the cast ZK60 sample immersed in 0.1 M NaCl for 72 h. The corrosion pit was covered by the corrosion product, and the second-phase particles were encompassed in it. Only Mg, O, and Cl elements were found in the corrosion product layer, which should result from Mg(OH) 2 and Cl − in the test solution. Moreover, a large number of Cl − through the whole corrosion product layer suggests that Cl − could penetrate the product layer easily and may become enriched in the bottom of the corrosion pit to propagate the pit corrosion [71]. The second-phase particles in the corrosion pit imply that they may be related to the formation of corrosion pits, which is discussed later.

Corrosion Reactions on the Cast ZK60 Alloy
A thin oxide film can be formed on the surface of the cast ZK60 alloy in a moist atmosphere at room temperature according to Reactions (5)- (6), which may have a bilayer structure with an inner layer of MgO and an outer layer of Mg(OH)2 [72,73].
Because MgO and Mg(OH)2 are both relatively soluble in water according to Reactions (7)-(8) [3,74], where Ksp is the solubility product constant, MgO will be gradually dissolved and converted to Mg(OH)2 when they are immersed in NaCl solution.

Corrosion Reactions on the Cast ZK60 Alloy
A thin oxide film can be formed on the surface of the cast ZK60 alloy in a moist atmosphere at room temperature according to Reactions (5) and (6), which may have a bilayer structure with an inner layer of MgO and an outer layer of Mg(OH) 2 [72,73].
Because MgO and Mg(OH) 2 are both relatively soluble in water according to Reactions (7) and (8) [3,74], where K sp is the solubility product constant, MgO will be gradually dissolved and converted to Mg(OH) 2 when they are immersed in NaCl solution.
Mg(OH) 2 = Mg 2+ + 2OH − K sp = 10 −11 (8) In this case, the MgO/Mg(OH) 2 oxide film formed on the cast ZK60 alloy was partly protective in a neutral NaCl solution. The anodic and cathodic partial reactions of the corrosion process can be written as Reactions (9) and (10), respectively, and the corrosion product is formed as Reaction (11) [75].

Initiation of Corrosion on the Cast ZK60 Alloy
Based on the microstructure of the cast ZK60 alloy, Figure 19 presents a schematic model to illustrate the initiation and development of the corrosion on the cast ZK60 alloy. As discussed above, a partly protective MgO/Mg(OH) 2 oxide film will be formed on the cast ZK60 alloy in neutral 0.1 M NaCl (Figure 19a). Because the large second-phase particles and the Zr-rich region in the grains were more stable than the Mg matrix in grain boundary regions, as shown in Figure 6, the oxide film in these regions should be more stable than that in their adjacent grain boundary regions (i.e., darker areas in Figure 2b), which was denoted as a stable and active oxide film in Figure 19a [76,77].

Initiation of Corrosion on the Cast ZK60 Alloy
Based on the microstructure of the cast ZK60 alloy, Figure 19 presents a schematic model to illustrate the initiation and development of the corrosion on the cast ZK60 alloy. As discussed above, a partly protective MgO/Mg(OH)2 oxide film will be formed on the cast ZK60 alloy in neutral 0.1 M NaCl (Figure 19a). Because the large second-phase particles and the Zr-rich region in the grains were more stable than the Mg matrix in grain boundary regions, as shown in Figure 6, the oxide film in these regions should be more stable than that in their adjacent grain boundary regions (i.e., darker areas in Figure 2b), which was denoted as a stable and active oxide film in Figure 19a [76,77].
Compared to the Mg matrix in the grain boundary regions, those with stable oxide film act as micro-galvanic cathodes. Under the attack of Cl -, the film around the second phase could be easily broken due to the higher numbers of imperfections [78]. Therefore, the micro-galvanic corrosion was initiated in the areas around the large discontinuous second-phase particles in the grain boundaries, owing to the strong galvanic effect between these second-phase particles and the Mg matrix in the grain boundaries (Figure 13b). It should be noted that there were some second-phase particles deposited in the grains (Figures 2a and 6), and therefore, the corrosion may also be initiated in the grains. However, we did not observe this phenomenon, which may be due to the stable oxide film in the grains (Zr-rich region) and the light galvanic effect between them.  Compared to the Mg matrix in the grain boundary regions, those with stable oxide film act as micro-galvanic cathodes. Under the attack of Cl − , the film around the second phase could be easily broken due to the higher numbers of imperfections [78]. Therefore, the micro-galvanic corrosion was initiated in the areas around the large discontinuous second-phase particles in the grain boundaries, owing to the strong galvanic effect between these second-phase particles and the Mg matrix in the grain boundaries (Figure 13b). It should be noted that there were some second-phase particles deposited in the grains (Figures 2a and 6), and therefore, the corrosion may also be initiated in the grains. However, we did not observe this phenomenon, which may be due to the stable oxide film in the grains (Zr-rich region) and the light galvanic effect between them.
After the corrosion initiation (Figure 19b), the oxide film near the second-phase particles in the grain boundaries cracked first. However, there were no black Mg(OH) 2 products observed in this beginning period (Figure 12a), because they were dissolved in the NaCl solution (Equation (5)). After the concentration of Mg(OH) 2 reached saturation in the NaCl solution, black Mg(OH) 2 precipitation occurred (Figure 12b).

Growth of Corrosion Filaments on the Cast ZK60 Alloy
Because the Zr-rich area within the grains was more stable than the Mg matrix in grain boundary regions, the corrosion preferentially propagated in the grain boundary regions in the initial corrosion period (Figure 19c), displaying the characteristics of filiform-like corrosion, as shown in Figures 12 and 13. According to [69,70], the front edges of the corrosion filaments acted as intense anodes via Equation (6), while the dark tracks behind the anodes acted as locally activated cathodes, where cathodic Equation (7) occurred. In NaCl solution, Cl − may be enriched around the anode areas to attack and break the oxide film [28]. By contrast, the activated cathodes behind the anodes provide the driving force for the development of the corrosion filaments. Then, the precipitates of Mg(OH) 2 cover the surface of the active cathodes behind the anodes to make them gradually turn into inert cathodes.
The second phase in the grain boundaries and the Zr-rich region within the grains have an essential influence on the origination and propagation of the corrosion filaments on the cast ZK60. These filiform-like corrosion characteristics of the cast ZK60 were quite different from those of Mg-3Zn and Mg-8Li alloys reported previously [65,71], in which the corrosion filaments seem not to be affected by the second phases and can originate from the center of the grains and spread to various directions. These differences should be ascribed to their different microstructures, especially the quantity and size of the second phase and the distribution of alloying elements in these alloys.
As the corrosion time was increased, the active "dark areas" along the grain boundaries were gradually covered with a black Mg(OH) 2 layer ( Figure 19c) and potentially became cathodic areas [79]. This change would impede the anodic corrosion process in these areas and make their potential positive, reducing and even eliminating the potential difference between these areas and the Zr-rich regions. Therefore, the corrosion filaments could gradually develop to the central area of grains (Figure 19d) to form some broad corrosion areas, as shown in Figures 13 and 14. Additionlly, the black Mg(OH) 2 layer was porous and imperfect and could not prevent the entry of the NaCl solution ( Figure 18). Thus, the anodic process still existed beneath the Mg(OH) 2 layer, developing an apparent corrosion depth under the corrosion filaments, as shown in Figures 13 and 14. When t > 24 h, most of the alloy surface was gradually covered with the Mg(OH) 2 layer (Figure 15), and finally, it would be covered completely (Figure 19e,f).

Development of Corrosion Pits on the Cast ZK60 Alloy
With the development of the corrosion filaments, the corrosion pits will be formed when the Mg matrix surrounding the large second-phase particles is corroded completely (Figure 19e). Some small corrosion pits occurred in the corrosion filaments when t = 2 h (Figure 14c,d), which may be due to the loss of the second-phase particles when the corrosion product was removed. The visible corrosion pits occurred after 12 h of immersion ( Figure 15), and their number and depth increased with time over 24-72 h, in which P depth was about one magnitude order higher than those corrosion rates (P w , P H , and P i ), as shown in Figures 15-17. The high P depth of the cast ZK60 alloy indicates a severe localized corrosion state and warrants further investigation.
The development of the corrosion pits was related to the occluded environment under the corrosion filaments, covered by the second-phase particles and Mg(OH) 2 precipitate. The solution in the occluded corrosion pit would be alkalized to increase the pH to around 10.5 through Reaction (5). At the same time, Cl − would migrate in due to the charge neutralization and become enriched at the bottom of the pit (Figure 18). These effects would accelerate the anodic process in the corrosion pits and make them develop deeply [80].
According to the above discussion, the microstructure of the cast ZK60 alloy had an essential influence on the corrosion initiation and propagation in NaCl solution. Thus, it is necessary to improve its corrosion resistance by microstructure modification, which will be studied later.

Change in the Corrosion Rate in Different Corrosion Periods
The P H , P i , and R p values of the cast ZK60 alloy generally showed different change tendencies in different immersion periods (Figures 8, 9b and 10b). In the early corrosion period (0-2 h), the corrosion rates (P H and P i ) increased with time (Figures 8b and 9b), while R t , R f , and R p decreased (Table 2 and Figure 10b). In this period, the filiform corrosion was initiated and developed quickly on the cast ZK60 alloy without visible corrosion pits (Figures 12-14), corresponding to the states in Figure 19a-d. In this case, with the development of the corrosion filaments, the broken oxide film increased steadily to accelerate the anodic and cathodic processes, as shown in Figure 9, and made the corrosion rates increase. The increase in the broken oxide film may be the reason why no E break occurred in the anodic polarization curve when t = 2 h (Figure 9). Furthermore, the increase in the broken oxide film enlarged the reactive areas on the cast ZK60 surface to increase CPE dl and CPE f , meanwhile decreasing R L and L in this period (0-2 h, Table 2).
The second corrosion period was determined to be within 2-24 h, according to the above discussion about the corrosion development on the cast ZK60 alloy. In this period, the corrosion filaments gradually occupied most of the active oxide film, and a small number of corrosion pits occurred and developed quickly (Figures 15 and 16), which corresponds to the state in Figure 19e. Because most of the active oxide film was corroded and covered with a black Mg(OH) 2 precipitate, the anodic process, Reaction (9), was inhibited significantly to make E corr more positive, and E break occurs again with a more positive value ( Figure 9). In this case, the corrosion rate (P H and P i ) decreased significantly (Figures 8b and 9b), while R t , R f , and R p increased ( Table 2 and Figure 10b). The difference is that P i and 1/R p reached the lowest value at t = 24 h, but P H decreased persistently until t = 48 h, while P w as shown in Figure 7 displayed the highest value at t = 24 h, which may be ascribed to the differences among these test methods. In this period, the decrease in the broken oxide film may also result in a clear increase in R L and L and decrease in CPE dl (Table 2). Meanwhile, CPE f firstly increased and then decreased, which may have been due to the increase in the area and thickness of the corrosion product layer. Even though very few corrosion pits occurred in this period (Figures 15 and 16), their P depth increased significantly and reached a large value of about 24 mm y −1 (Figure 17), which was much larger than the general corrosion rates of P w , P H , and P i in this period (Figures 7, 8 and 9b). Therefore, the pitting corrosion became more critical than the filiform corrosion in this period.
Over 24-72 h, most of the surface of the cast ZK60 was gradually covered with Mg(OH) 2 layer, but the number and depth of the corrosion pits increased (Figures 15-17), corresponding to the state in Figure 19f. Because of the increase in corrosion pits, the anodic process was accelerated again. In this case, the E break and E corr of the cast ZK60 alloy became negative, and P i increased again (Figure 9), while R t , R f , and R p decreased (Table 2 and Figure 10b). However, P w and P depth decreased with time in this period (Figures 7 and 17), indicating that the increase in the corrosion product layer still impeded the growth rate of the pitting depth and the general weight loss, possibly due to the alkalizing effect of the cathodic reaction and the low solubility of Mg(OH) 2 . In this period, the increase in CPE f may have been due to the dissolution of the corrosion product layer reducing its thickness; meanwhile, the decrease in R L and L may have been related to the acceleration of the pitting corrosion process, similar to in the processes on passive metals [61]. In general, pitting corrosion became more severe in this period and dominated the corrosion behavior of the cast ZK60 alloy. In this case, P depth should be adequate to describe the corrosion rate of the cast ZK60 alloy in this period.

1.
The microstructure of the cast ZK60 alloy was composed of an α-Mg phase and large second-phase particles (MgZn 2 ), which mainly deposited along the grain boundaries, and a Zr-rich region existed in the central area of the grains. The grain boundaries and their adjacent regions (noted as grain boundary regions) had relatively higher Mg contents.