Effect of the microstructure on the corrosion behavior of dissimilar friction stir-welded 304 austenitic stainless steel and Q235 low-carbon steel joints

To investigate the effect of the microstructure on the corrosion behavior of the dissimilar friction stir welded (FSW) joint between 304 stainless steel (SS304) and Q235 low-carbon steel, the microstructure of the joint in this work was characterized by optical microscopy, scanning electron microscopy (SEM), and electron backscatter diffraction. The corrosion behavior in different zones of the joint was evaluated by the electrochemical tests, and the corresponding corrosion morphologies were illustrated via SEM and laser confocal scanning microscopy. According to the results, plenty of low-angle grain boundaries (LAGBs) and a low proportion of twin boundaries (TBs) deteriorated the corrosion resistance on the SS304 side of the joint. The corrosion products of the SS304 side mainly included γ-Fe2O3, FeCrO4, and Cr2O3, while those of the Q235 steel side were α-Fe2O3 and α-FeOOH. The corrosion mechanism in the stir zone (SZ) was galvanic corrosion between proeutectoid ferrite and pearlite on the Q235 steel side, during which the austenite remained uncorroded, whereas the proeutectoid ferrite, pearlite, and bainite were severely corroded. The above results indicated that the uniform distribution of mixed structures and a small area proportion of austenite (cathode) would improve the corrosion resistance in the SZ.


Introduction
Owing to the low cost and good corrosion resistance of carbon steel and stainless steel, respectively, their dissimilar steel welded structures have been widely used in the chemical, petroleum, natural gas, nuclear energy, power generation industries, among others [1][2][3][4]. The conventional fusion welding methods for carbon steel and stainless steel mainly include tungsten inert gas welding (TIG) [5,6], laser welding [7], and electron-beam welding [8]. However, these methods tend to produce defects such as solidified cracks, pores, and the coarsening of second phases during welding, resulting in the poor mechanical performance of the resulting joints [2,9]. To prevent the formation of defects in solidified structures caused by conventional fusion welding technologies, solid-state welding provides an ideal high-quality welding method for dissimilar steels [10].
Friction stir welding (FSW) is a solid-state welding technology for aluminum alloys that was invented by the British Welding Research Institute in the early 1990s [11,12]. In the FSW process, a high-speed rotating stirring tool is inserted into the seam of weld plates until the shoulder is in close contact with the material to be welded. The material is softened under the action of frictional heat, while the tool's stirring action causes plastic flow and mixing of materials at the joint, thereby achieving solid-state welding between materials [13]. Compared with fusion welding, FSW requires the lower welding temperature and shorter welding time [14]. Therefore, when FSW is performed on dissimilar steels, the use of a lower heat input prevents the formation of defects in the solidified structures and the burning of the alloying elements which usually take place in conventional fusion welding technologies. The refinement, homogenization, and densification of the microstructure during welding significantly enhance the joint strength [15][16][17][18].
In the past few decades, FSW has become a common method to join steel and other light dissimilar metals such as aluminum [19,20]. Many studies have shown [21][22][23][24][25] that adjusting and optimizing the FSW parameters for aluminum/steel joints can dramatically reduce the formation of brittle Fe x Al y intermetallic compounds in the joint structure, which significantly enhances the joint's tensile strength [26]. Meanwhile, there are great differences in the physical and chemical properties of aluminum/steel materials, so their corrosion behaviors have also been investigated. In particular, Seo et al [27] studied the corrosion properties of dissimilar FSW joints composed of 6061 aluminum alloy and HT590 low-carbon alloy steel. Their results indicated that the corrosion resistance of the HT590 steel base metal of the joint was better than that of the 6061 aluminum alloy; therefore, the corrosion resistance of the weld zone was between those of the two base materials. Mahto et al [28] reported the corrosion properties of dissimilar metal FSW joints composed of AA6061-T6 aluminum alloy and AISI304 stainless steel. They found that as the grain size of the joint decreased, its corrosion resistance decreased accordingly. The corrosion resistance of the two base metals and the joint's weld zone were consistent with the results of Seo et al [27].
In recent years, researchers have begun to explore and develop methods for the FSW of dissimilar metals such as carbon steel and stainless steel to prepare well-formed joints without obvious defects. The joints underwent tensile fracture on the base material side of carbon steel, showing good mechanical properties [15,29]. Jafarzadegan et al [3] conducted the FSW of Type 304 stainless steel and ST37 low carbon steel, and investigated the microstructure and mechanical properties of the joint. According to their findings, dynamic recrystallization (DRX), phase transformation, and grain refinement occurred in stainless steel and low carbon steel, which enhanced the hardness and tensile strength of the joint. Rahimi et al [30] used FSW to join dualphase stainless steel and low alloy steel, and found that the peak temperature significantly affected the microstructure evolution. Specifically, strain-induced continuous dynamic recrystallization and discontinuous dynamic recrystallization resulted in the formation of the fine-grained structure [29]. However, due to the different microstructure potentials in various regions of dissimilar steel FSW joints, galvanic cells formed, and the grain boundaries had divergent characteristics. These factors affected the joint's corrosion resistance, resulting in failure that reduced the service life of the welded structure. Nevertheless, the studies reporting the influence of the microstructure in each zone of carbon steel and stainless steel on the corrosion behavior of FSW joints are still scarce. In view of the above, a FSW joint composed of 304 stainless steel (SS304) and Q235 lowcarbon steel was taken in this research to elucidate the aforementioned failure phenomena. Special attention was paid to the microstructural effects on the corrosion mechanisms and the quantitative analysis of corrosion degree in each micro-zone of the joint.

FSW experiment
In this study, hot-rolled SS304 and Q235 steel plates with dimensions of 100×60×3 mm 3 were used for FSW butt welding. The plates were placed on the advancing side (AS Q235 ) and retreating side (RS 304 ), respectively. The chemical compositions of the two materials are listed in table 1. FSW was performed on a modified vertical milling machine with a rotation speed of 475 rpm and a welding speed of 47.5 mm min −1 , which enabled one to obtain a workpiece with the sound appearance and appropriate mechanical characteristics. A WC-Co tool had a shoulder of 16 mm diameter, tapered pin diameters of 5.3 and 3.0 mm, and a pin length of 2.7 mm. The tilt angle was 2 deg, at which the friction pressure was 20 kN.

Microstructure characterization
After FSW, the metallographic specimens were cut perpendicular to the joint's welding direction (WD). After mechanical polishing, aqua regia and nitric acid-ethanol solution were used to etch the cross-sections of SS304 and Q235 steel, respectively. Optical microscopy (OM) and scanning electron microscopy (SEM) were applied to observe the microstructure after corrosion. Electron backscatter diffraction (EBSD) image acquisition was completed with a fully-automatic HKL-EBSD system connected to a field-emission SEM.

Electrochemical tests
The joint was processed into square pieces with dimensions of 8 mm×8 mm×3 mm. The sampling location, sample dimensions, and division of each area are shown in figure 1. The joint was divided into five macro-zones: SS304 and Q235 steel base materials (BM 304 and BM Q235 ), the retreating side (RS 304 ), the stir zone (SZ), and the advancing side (AS Q235 ), where the RS 304 , SZ, and AS Q235 each included two micro-zones. RS 304 was composed of a heat-affected zone (HAZ 304 ) and a thermomechanically-affected zone (TMAZ 304 ) on the retreating side. The SZ was presented by SZ 304 on the retreating side and SZ Q235 on the advancing side. AS Q235 included TMAZ Q235 and HAZ Q235 on the advancing side. The test surface of the macro-zones of electrochemical samples was the WD-TD surface with a scan area of 0.64 cm 2 . The samples were exposed to 3.5 wt% NaCl solution at room temperature, which contained deionized water and analytically pure NaCl. The open-circuit potential, potentiodynamic polarization, and electrochemical impedance tests were carried out by means of a three-electrode system. The auxiliary electrode was a platinum plate, the reference electrode was 1 M KCl Ag/AgCl, and the working electrode was the sample. A CST520 electrochemical workstation was used for tests. During the tests, the working electrode was first immersed in the test solution for 1 h, and the potentiodynamic polarization measurements were performed after the open-circuit potential (OCP) stabilized. The time for collecting the polarization curves was 1 h. The scan rate was 0.3333 mV s −1 , and the scan range was −0.5 V versus OCP∼+1.0 V versus OCP. The scanning frequency range of electrochemical impedance spectroscopy (EIS) was from 100 kHz to 5 mHz, and the disturbance signal was 5 mV.

Corrosion morphology characterization
SEM was used to examine the corrosion morphology of the above zones after the joint was corroded. A laser confocal scanning microscope (Keyence vk9710) was employed to determine the surface roughness characteristics and to quantitatively assess the corrosion resistance of each zone by measuring the root mean square (RMS) roughness. The corrosion products in the zones were qualitatively analyzed with a Raman spectrometer (LabRAM HR Evolution) coupled with an Ar + laser (532 nm wavelength) within a spectral range of 100-2000 cm −1 . Figure 2 displays the microstructures of the BM 304 and BM Q235 zones. The main phase in BM 304 was austenite (γ) with coarse grains whose average size was 10.9 μm (including twins, TBs) (figure 2(a)). The fraction of lowangle grain boundaries (LAGBs) was 14% (except for TBs) (figure 2(b)). Many TBs (25%) could be observed inside the γ grains, which are marked with black arrows in figures 2(a) and (b). In addition, the microstructure of BM Q235 mainly included ferrite and pearlite. Pearlite was distributed on the ferrite's grain boundaries, with an area proportion of 15.6% (figure 2(c)). Pearlite was composed of eutectoid ferrite (EF) and cementite (Fe 3 C), showing a predominantly lamellar texture. According to figure 2(d), the lamellar pearlite had an irregular morphology, with an interlamellar spacing of 0.7-0.8 μm, and Fe 3 C existed as intermittent rods or granules. Figure 3 depicts the microstructure on the RS 304 side of the FSW joint (i.e., the SS304 side). After undergoing thermal cycling during welding, the γ structure in HAZ 304 recovered, and the grains were coarsened, with an average grain size of 11.3 μm (Zone b in figure 3(a)). Compared with the BM 304 , the LAGBs proportion increased ( figure 3(b)). This can be related to the dislocation rearrange under high temperature and the formation of lowangle grain boundaries (LAGBs) within the major grains [31]. Figure 3(b) shows a decrease in the TBs proportion (marked with a black arrow). TMAZ 304 (Zone c in figure 3(a)) exhibited plastic deformation, mainly due to the action of shouldering. As a result, grains were elongated and to enrich the average size of 6.5 μm, and some of them revealed the emergence of DRX [32]. Compared with the HAZ 304 , the LAGBs proportion decreased. This was because the LAGBs were transformed into high-angle grain boundaries (HAGBs) as the deformation proceeded, and the microstructure with small grains was developed [33]. Meanwhile, as seen in figure 3(c), the TBs proportion increased (marked with a black arrow). Figure 4 shows the SZ microstructure of the FSW joint. Due to the rotation during FSW, the mixed textures in the center of the weld exhibited two morphologies. In the left region (SZ 304 ), SS304 was rolled into Q235 steel to form a river-like morphology. The structure was composed of face-centered cubic (FCC) γ, body-centered cubic (BCC) proeutectoid ferrite (PF), and pearlite. The distribution of the mixed texture was non-uniform, and the area fraction of γ phase was large (figures 4(a) and (b)). On the right region (SZ Q235 ), the Q235 steel structure underwent a phase transformation, forming PF and bainite (B), which were afterward mixed with γ phase in the SS304 of Q235 steel. The mixed texture was more uniformly distributed, and the area proportion of γ phase was low (figure 4(c)). Figure 5 depicts the microstructure of AS Q235 (i.e., the Q235 low-carbon steel side) of the FSW joint. TMAZ Q235 was composed of PF and pearlite. Here, PF includes bulk ferrite (BF), grain boundary ferrite (GF), and acicular ferrite (AF). Compared with BM Q235 , pearlite grains were much coarser, with an area proportion of 55% ( figure 5(a)). The interlamellar spacing in pearlite decreased, and the distribution became more uniform, showing an interlamellar spacing of 0.4-0.5 μm ( figure 5(b)). HAZ Q235 was composed of ferrite and pearlite. Compared with BM Q235 , the area proportion of pearlite increased slightly to 17.7% ( figure 5(c)). Also, pearlite in HAZ Q235 had the smallest interlamellar spacing of 0.2-0.3 μm and was uniformly distributed (figure 5(d)).

Corrosion behavior
3.2.1. Open-circuit potential and potentiodynamic polarization curves Figure 6 displays the open-circuit potentials and potentiodynamic polarization curves acquired at each macrozone of the joint sample in 3.5 wt% NaCl solution. It can be seen from figure 6(a) that BM 304 had the highest open-circuit potential, followed by RS 304 , and finally by SZ, BM Q235 , and AS Q235 , which had similar values. This means that BM 304 and RS 304 were less active compared to SZ, BM Q235 , and AS Q235 . In a word, BM 304 and RS 304 had the nobler corrosion potentials as negative electrodes, while SZ, BM Q235 , and AS Q235 had the lower  potentials as positive electrodes, hence a galvanic coupling effect could have formed in the case of couples of welded joints [34]. According to figure 6(b), BM 304 and RS 304 exhibited anodic passivity during the corrosion process, whereas the cathodic part of SZ, BM Q235 , and AS Q235 had the typical characteristics of a diffusioncontrolled oxygen reduction reaction [35,36]. Table 2 shows the Tafel fitting results of the potentiodynamic polarization curves from each macro-zone of the joint. It can be seen that the corrosion current of BM 304 was the lowest, indicating the best corrosion resistance. The corrosion current then increased in a sequence of RS 304 , SZ, BM Q235 , and AS Q235 , which meant there was an increase in their corrosion rates and the deterioration in corrosion resistance.   Figure 7(a) depicts the EIS Nyquist plots of the joint's macro-zones in 3.5 wt% NaCl solution. It can be seen from the figure that the impedance spectra have a time constant, and there are capacitive reactance lines elated to charge transfer occurring on the electrode surface in each macro-zone of the joint [37]. The capacitive reactance lines of BM 304 and RS 304 have the larger diameters and both show an arc shape, which indicates that these zones possess similar corrosion resistance mechanisms. The diameters of the high-frequency zone of the capacitive reactance arcs of SZ, BM Q235 , and AS Q235 sequentially decreased compared to those of BM 304 and RS 304 . Their  shape changed from an arc to a semicircle, indicating that BM 304 had the best corrosion resistance, followed by RS 304 , and the corrosion resistance then decreased in a sequence of SZ, BM Q235 , and AS Q235 . This variation is consistent with the polarization curves. Figures 7(b) and (c) display the Bode magnitude plots and Bode phase plots of the joint's macro-zone, respectively. Both plots exhibit three distinct regions: low-frequency, mid-frequency, and high-frequency regions. In figure 7(b), the low-frequency region (10 -2 -10°Hz) represents the polarization resistance of the sample. The polarization resistance of the macro-zones followed the order of BM 304 >RS 304 >SZ>BM Q235 >AS Q235, which is consistent with the polarization curves. In the mid-frequency region (10°-10 4 Hz) ( figure 7(c)), the phase angles of BM 304 and RS 304 , as well as those of SZ, BM Q235 , and AS Q235 , were similar. This indicates that the microstructures of BM 304 and RS 304 , as well as those of SZ, BM Q235 , and AS Q235 , were identical, meaning there were similar electrical double-layer characteristics. The high-frequency region (10 4 -10 5 Hz) was attributed to the resistance of the test solution. Figures 7(d) and (e) respectively show the equivalent circuit diagrams used to simulate the EIS data of the SS304 and Q235 sides. In the equivalent circuit diagram, R s denotes the solution resistance; Q dl stands for the constant-phase element (CPE) of the double-layer capacitor; R ch is the charge-transfer resistance. The constantphase elements reflect the capacitive behavior of non-ideal interfaces. The impedance of the constant-phase element is expressed as follows [38]:

EIS test
where Y 0 denotes the scale factor; j 2 =−1 and w=2πf; n is the dispersion coefficient, which is related to the surface non-uniformity. If n=0, the constant-phase element represents a resistor R=Y 0 −1 ; if n=1, the constant-phase element refers to a capacitor C=Y 0 . The EIS parameters of each macro-zone of the joint can be calculated using the circuit diagrams from figures 7(d) and (e), and the results are listed in table 3. It can be seen from the table that the EIS parameters of each macro-zone were significantly different. The charge-transfer resistance R ch had the largest difference, where a higher R ch indicated better corrosion resistance [38]. Moreover, the corrosion resistance of each macro-zone changed as follows: BM 304 >RS 304 >SZ>BM Q235 >AS Q235 . This result is also consistent with the measured polarization curves. In addition, given the least Chi-square (χ 2 ) value, the experimental EIS data show excellent agreement with an equivalent circuit corresponding to R s (Q dl R ch ). Figure 8 displays the corrosion morphology evolution of each micro-zone of the FSW joint. Figures 8(a)-(c) depict the corrosion morphologies of BM 304 , HAZ 304 , and TMAZ 304 , respectively. Different degrees of pitting corrosion occurred in the above three zones on the sample's surface, revealing pits with porous lacy covers due to the accumulation of Cl --rich salt solution therein. In a word, Clions penetrated the surface film to the metal/ film interface, causing a local damage to the film and forming tiny pits [39]. After the initiation of such pits, metal ions at the edges of the pits entered the salt solution faster, reducing its local concentration below the critical value, and passivation occurred. In contrast, the active dissolution at the bottom of pits continued until the intact surface film surrounding the pits was destroyed and new pits formed. However, metal ions diffused rapidly through the new corroded pores, which again decreased the local concentration of the salt solution and induced passivation at the edges. Therefore, during the film destruction and passivation at the pit edges, the pits continued to grow, causing the film to be destroyed repeatedly before finally forming corrosion pits with porous lacy covers [40][41][42][43].

Corrosion morphology
It can be seen from figure 8 that the area of the pit with a lacy cover in BM 304 was small, indicating less corrosion and, thus, better corrosion resistance. The pit with a lacy cover in RS 304 had a larger area, meaning there were more serious corrosion and poorer corrosion resistance. This was because the LAGBs proportion of RS 304 (HAZ 304 and TMAZ 304 ) significantly increased after FSW (figures 3(b) and (c)), which resulted in a higher dislocation density. Balusamy et al [44] showed that an increase in dislocation density provides more active sites, which will accelerate corrosion [45]. In addition, the microstructure characterization results showed that, compared with BM 304 , the proportions of TBs in HAZ 304 and TMAZ 304 were significantly lower. Husain et al  [46] reported a decrease in the proportion of TBs and the tendency of the sample to corrode. Therefore, the increase in the LAGBs proportion and the decrease in the TBs proportion were the main reasons why RS 304 had worse corrosion resistance than BM 304 . In turn, HAZ 304 included a higher LAGB proportion and a smaller TB content than TMAZ 304 , indicating that the former one had worse corrosion resistance than the latter one. Figures 8(d) and (e) show two different corrosion morphologies in the SZ. One is a river-like morphology ( figure 8(d)) with a more serious corrosion, which is consistent with the structure and morphology shown in the OM images of SZ 304 ( figure 4(a)). Because of the poor distribution uniformity of the PF-pearlite mixed texture in the γ phase of SS304 and Q235 steel, as well as the higher corrosion potential and the larger area proportion of γ phase compared to those of PF and pearlite, strong galvanic corrosion occurred between γ as the cathode and PF and pearlite as the anodes, resulting in the river-like morphology. PF and EF in pearlite were prone to corrosion, but γ and Fe 3 C in pearlite remained after corrosion. The other morphology is craquelure ( figure 8(e)), indicating relatively light corrosion because the microstructure in SZ Q235 was composed of PF, B, and γ. Since the mixed texture is more uniformly distributed and the area proportion of γ as the cathode is small ( figure 4(c)), galvanic corrosion was relatively weak [47]. The craquelure morphology means that B and PF were preferentially corroded in sequence. According to figure 8(f), TMAZ Q235 was seriously corroded, revealing coarse intergranular cracks and fine intragranular cracks. The microstructure of TMAZ Q235 was composed of GF, AF, BF, and pearlite, and the area proportion of pearlite was relatively large (55%) ( figure 5(a)). The formation of corrosion cracks in TMAZ Q235 was due to galvanic corrosion. In the entire TMAZ Q235 , GF, AF, and BF served as the anodes, and pearlite was the cathode, while EF and Fe 3 C in pearlite were the anode and cathode, respectively. Therefore, the anodes (GF, AF, BF, and EF in pearlite) were preferentially corroded [48,49], which exposed the remaining Fe 3 C cathode to the solution. Intergranular cracks formed due to the corrosion of GF and BF, whereas small intragranular cracks emerged owing to the corrosion of AF and EF. During corrosion, on account of the high area proportion of pearlite, the reaction area of Fe 3 C as the cathode increased upon prolonging the corrosion time, which promoted galvanic corrosion and caused the corrosion rate within the zone to increase. As a result, TMAZ Q235 experienced severe corrosion [50]. Figure 8(g) displays the corrosion morphology of HAZ Q235 . The non-uniformly distributed ferrite cavities (marked with a white circle) surrounded with various corrosion products can be seen in the figure. The morphology shows that the corrosion of HAZ Q235 was mainly due to the dissolution of ferrite, and galvanic corrosion was secondary corrosion. Most corrosion products were produced since HAZ Q235 had a lower area proportion of pearlite than the adjacent TMAZ Q235 , and the entire surface was equivalent to an anode. Accordingly, a galvanic pair formed with the TMAZ Q235 as the cathode [35], which accelerated corrosion in HAZ Q235 . Due to the larger cathode area of TMAZ Q235 , Fe 2+ diffused relatively easy, leading to an increase in the Fe 2+ concentration on the surface of HAZ Q235 . As a result, HAZ Q235 generated more corrosion products on the surface of Fe 3 C and also on the surface of ferrite. Sun et al [50] showed that corrosion products can prevent the solution from further entering the surface of a substrate; therefore, corrosion products improved the corrosion resistance of HAZ Q235 . Figure 8(h) depicts the corrosion morphology of BM Q235 . Although corrosion occurred on ferrite, the corroded surface was relatively flat, meaning that corrosion was shallow. Additionally, there were fewer corrosion products (highlighted with a white elliptical area), which were mainly located on the pearlite. The morphology characterization data show that BM Q235 had the better corrosion resistance. The reason is that the area proportion of pearlite in BM Q235 was low (15.6%), so the cathode reaction area was small when corrosion occurred, and its galvanic corrosion effect was weak. In addition, the formation of corrosion products at pearlite inhibited the further dissolution of EF between the pearlite lamellae, which decreased the corrosion rate and endowed BM Q235 with better corrosion resistance. Compared with TMAZ Q235 and HAZ Q235 , BM Q235 had the lowest area proportion of pearlite and the best corrosion resistance. In turn, TMAZ Q235 had the highest area proportion of pearlite and the worst corrosion resistance. Finally, HAZ Q235 had the corrosion resistance characteristics between the above two zones. In addition, according to the micromorphology of pearlite in the three zones and the electrochemical corrosion results, the interlamellar spacing and distribution uniformity of pearlite did not obviously affect the corrosion resistance.
To quantitatively compare the corrosion degree of micro-zones in the joint, a laser confocal scanning microscope was used to determine their roughness characteristics by finding the corresponding RMS values. The results are shown in figure 9. The RMS value of each micro-zone varied in the order of BM 304 <TMAZ 304 <HAZ 304 < SZ Q235 <SZ 304 <BM Q235 <HAZ Q235 <TMAZ Q235 . Therefore, the corrosion resistance of the micro-zones followed a sequence of BM 304 >TMAZ 304 >HAZ 304 >SZ Q235 >SZ 304 >BM Q235 >HAZ Q235 >TMAZ Q235 , which is consistent with the corrosion morphology characterization.

Corrosion products
In this section, Raman spectroscopy was used to characterize the corrosion products in each micro-zone on the sample's surface, and the results are shown in figure 10. Because the surface Raman spectra of BM 304 , RS 304 , and SZ 304 were similar, we only tested and analyzed the corrosion products inside and outside BM 304 pits on the SS304 side. In the Raman spectrum of the outer region of the BM 304 pit ( figure 10(a)), the peaks of α-FeOOH, γ-Fe 2 O 3 , Cr 2 O 3 , and FeCrO 4 were at 248 cm −1 , 378 cm −1 [35], 530 cm −1 , and 680 cm −1 , respectively [51,52]. The presence of Cr oxide in the Raman spectrum indicates that the external area of the BM 304 pit was passivated, which is consistent with the corrosion morphology data. The peaks at 1340 cm −1 and 1590 cm −1 are the D and G bands of carbon atoms [53], which represent disordered carbon and graphitized carbon, respectively. Generally, the relative intensity ratio (I D /I G ) is used to express the crystallinity of carbon materials, where a smaller ratio represents fewer defects in the carbon material and a higher graphitization degree [54]. The I D /I G ratio in the Raman spectrum of the outer region of the BM 304 pit was relatively high (0.84), indicating that there were fewer carbon defects in the outer region and that the hexagonal crystal structure was retained [55]. In the Raman spectrum of the interior of the BM 304 pits ( figure 10(b)), there were two oxide peaks with a weak intensity, associated with γ-Fe 2 O 3 (378 cm −1 ) and FeCrO 4 (680 cm −1 ) [52]. The I D /I G ratio in the spectrum (0.86) exceeded that of the pit exterior. These results indicate that more C defects were present inside the pit than outside and that corrosion was more obvious.
In the Raman spectrum of SZ Q235 on the Q235 steel side (figure 10(c)), the peaks at 212, 271, and 583 cm −1 were attributed to α-Fe 2 O 3 , and the weak peak at 378 cm −1 was assigned to γ-Fe 2 O 3 [35], whereas I D /I G was 0.85. According to the Raman spectra of TMAZ Q235 , HAZ Q235 , and BM Q235 (figures 10(d)−(f)), the corrosion products in AS Q235 mainly included α-Fe 2 O 3 , α-FeOOH, and γ-FeOOH. The peaks of α-Fe 2 O 3 appeared at 280, 293, 473, and 589 cm −1 ; those of α-FeOOH were at 217, 390, and 480 cm −1 ; that of γ-FeOOH was at 657 cm −1 [35]. Within the Raman shift range of 100-700 cm −1 , TMAZ Q235 showed strong characteristic peaks, revealing the formation of abundant corrosion products. The Raman peaks of HAZ Q235 and BM Q235 were relatively weak, indicating few corrosion products. Moreover, the I D /I G values of the three zones were 1.18, 0.88, and 0.86, respectively, meaning that the largest amount of carbon defects and the most serious corrosion occurred in TMAZ Q235 . The corrosion degrees of HAZ Q235 and BM Q235 decreased successively, which is consistent with the corrosion morphology results. Figure 11 displays a schematic diagram of the potential polarization corrosion process within the SZ mixed zone of the FSW joint in 3.5 wt% NaCl solution. The left side of the SZ mixed zone was SZ 304 , whose composition was a mixture of PF, pearlite, and γ, with a river-like morphology ( figure 4(a)). On the right side, there was SZ Q235 , containing a mixture of PF, B, and γ ( figure 4(c)). Because γ is electrochemically more stable than B, PF, and pearlite, the potential of the former one was more positive, making it difficult to corrode. Thus, when γ was the cathode, while PF, pearlite, and B were the anodes, galvanic pairs formed ( figure 11(a)).

Corrosion mechanism of the SZ mixed zone
On the left side, due to the non-uniform mixing of the components, there was a large potential difference, and the cathodic reaction area was large, so that PF and EF in the pearlite were corroded preferentially. The corrosion was serious, and γ and Fe 3 C remained after corrosion. On the right side, B and PF were preferentially corroded in turn. Shallow surface corrosion pits were mainly formed by the PF corrosion, and honeycomb cracks emerged due to the corrosion of B ( figure 11(b)). Upon prolonging the corrosion time, the Fe 2+ content increased, and corrosion products such as a little iron oxides were formed (figure 11(c)) according to the reaction equations below [47,56,57]:    (2) and (3) exhibit the dissolution of steel and reduction of oxygen. Equations (4)- (6) show the formation of iron oxide, while those (7) and (8) indicate the appearance of hydroxyl-containing compounds. A higher concentration of Cl − promoted the emergence of FeOOH [56]. The formations of iron(III) oxide (Fe 2 O 3 ) and iron oxyhydroxide (FeOOH) are consistent with the Raman spectroscopy data.

Conclusions
1. Dissimilar steel FSW joints composed of 304 stainless steel and Q235 low carbon steels were divided into five macro-zones, namely BM 304 , RS 304 , SZ, AS Q235 , and BM Q235 . Among them, RS 304 , SZ, and AS Q235 each included a pair of micro-zones, videlicet HAZ 304 and TMAZ 304 , SZ 304 and SZ Q235 , TMAZ Q235 and HAZ Q235 , respectively. The microstructures in the BM 304 , HAZ 304, and TMAZ 304 were mainly presented by γ phase. SZ 304 was mainly composed of γ, PF, and pearlite, and those in SZ Q235 were mainly a mixture of γ, PF, and B. TMAZ Q235 comprised PF and pearlite, while HAZ Q235 and BM Q235 were formed by ferrite and pearlite.
2. The open-circuit potentials, potentiodynamic polarization curves, and electrochemical impedance spectra of the joint's macro-zones showed that the corrosion resistance of the joints followed a sequence of BM 304 > RS 304 >SZ>BM Q235 >AS Q235 . The characterization results on the corrosion morphology of the joint's microzones were consistent with the laser confocal scanning microscopy data, and the corrosion resistance varied in the following order: BM 304 >TMAZ 304 >HAZ 304 >SZ Q235 >SZ 304 >BM Q235 >HAZ Q235 >TMAZ Q235 .
3. The corrosion mechanisms of BM 304 , HAZ 304 , and TMAZ 304 were referred to pitting corrosion, while those of SZ 304 and SZ Q235 corresponded to galvanic corrosion between mixed structures. Meanwhile, the corrosion of TMAZ Q235 was mainly the galvanic corrosion of PF and pearlite, while the corrosion of HAZ Q235 and Figure 11. Schematic diagram of the potential polarization corrosion process through the SZ mixed zone in 3.5 wt% NaCl solution.
BM Q235 was predominately the dissolution of ferrite and weak galvanic corrosion of ferrite and pearlite. The corrosion products inside and outside the BM 304 pit included γ-Fe 2 O 3 , FeCrO 4 , and Cr 2 O 3 , while those of SZ Q235 , TMAZ Q235 , HAZ Q235 , and BM Q235 were mainly α-Fe 2 O 3 and α-FeOOH.