Hydrogen Stress Cracking Behaviour in Dissimilar Welded Joints of Duplex Stainless Steel and Carbon Steel

As the need for duplex stainless steel (DSS) increases, it is necessary to evaluate hydrogen stress cracking (HSC) in dissimilar welded joints (WJs) of DSS and carbon steel. This study aims to investigate the effect of the weld microstructure on the HSC behaviour of dissimilar gas-tungsten arc welds of DSS and carbon steel. In situ slow-strain rate testing (SSRT) with hydrogen charging was conducted for transverse WJs, which fractured in the softened heat-affected zone of the carbon steel under hydrogen-free conditions. However, HSC occurred at the martensite band and the interface of the austenite and martensite bands in the type-II boundary. The band acted as an HSC initiation site because of the presence of a large amount of trapped hydrogen and a high strain concentration during the SSRT with hydrogen charging. Even though some weld microstructures such as the austenite and martensite bands in type-II boundaries were harmless under normal hydrogen-free conditions, they had a negative effect in a hydrogen atmosphere, resulting in the premature rupture of the weld. Eventually, a premature fracture occurred during the in situ SSRT in the type-II boundary because of the hydrogen-enhanced strain-induced void (HESIV) and hydrogen-enhanced localised plasticity (HELP) mechanisms.


Introduction
The demand for stainless steel for marine structures and the oil and gas industry has increased tremendously over the past few years owing to its high corrosion resistivity. In the early 2000s, carbon steel accounted for more than 80% of the steel used in marine pipes, pumps, valve internals, and vessels; however, stainless steel now accounts for more than 50% of the steel used for such purposes [1,2]. In particular, the use of 22-Cr duplex stainless steel (DSS) has recently increased [2,3] because reduces the use of expensive elements and has better yield strength and corrosion resistance compared to austenitic stainless steel (ASS).
With the rising demand for DSS for various applications, the need for dissimilar welded joints (WJs) with carbon steel has increased. Duplex, super duplex, and austenitic stainless steels are commonly used as filler metals in the dissimilar WJs of DSS and carbon steel. However, dissimilar WJs used in the offshore, oil, and gas industries may cause sulphide stress cracking during the mining and transportation of shale gas, and stress corrosion cracking in various stress and corrosion environments, such as marine structures and ships. The propagation of these cracks is generally accelerated by diffusible hydrogen [4][5][6][7][8]. Inappropriate conditions during cathodic protection to prevent corrosion produce a hydrogen environment [9][10][11].
DSS exhibits the most appropriate mechanical properties when austenite and ferrite are approximately 50:50 and relatively weldable. When DSS is welded, however, this balance is easy to break, which increases ferrite fraction. As it increases, the WJ becomes more susceptible to hydrogen and hydrogen stress cracking (HSC). [12][13][14][15]. Low carbon steel normally has good weldability, and the welding of high strength steels produces a softened zone where the strain is concentrated making it vulnerable to HSC initiation. Furthermore, the welding microstructures have high hardness, such as MA constituent and martensite, accelerate hydrogen embrittlement [16][17][18]. In the case of dissimilar-welded joints in DSS and carbon steel, the ferrite phase fraction is increased in the heat affected zone (HAZ) of the DSS and the softened HAZ of high-strength carbon steels. Moreover, a large compositional difference between stainless steel and carbon steel produces the type-II boundary along the fusion line and galvanic coupling in dissimilar WJs [12]. It has been reported that hydrogen-induced cracking occurs owing to the galvanic coupling, sulphide stress cracking and debonding along the fusion line. The increase of ferrite fraction, the softened zone of carbon steel, galvanic coupling, and type-II boundary are known to influence HSC. However, the fracture mechanism of the dissimilar WJs has not been systematically studied. Therefore, further study needs to define the fracture mechanism in the dissimilar WJs, especially under hydrogen and stress environments.
In this study, we investigated the HSC behaviour of transverse WJs of dissimilar metal welds in DSS and carbon steel. Specifically, the effect of the weld microstructure on the initiation of HSC fracture in the WJs was investigated. The HSC behaviour of the WJs with DSS and ASS as filler metals was investigated by carrying out in situ slow strain rate testing (SSRT) with hydrogen charging.

Materials and Welding Parameters
The materials used for dissimilar-metal welding were 2205 DSS and two types of carbon steels: CSD (EQ51, carbon steel for DSS filler metal WJ) and CSA (S460, carbon steel for ASS filler metal WJ). The welding consumables (2.4 mm diameter) used in this study consisted of the 2209 DSS filler metal for the duplex filler (DF) joint and 309 LMo ASS filler metal for the austenite filler (AF) joint. The mechanical properties of the materials used in the dissimilar-metal welding are listed in Table 1. Because 2209 DSS has higher strength than ASS, CSD (with higher strength than CSA) was used for the DF combination to avoid a mismatch between the welding materials and base metals. The AF combination used CSA, which showed a strength comparable to that of the ASS filler metal. The chemical composition (wt.%) of each material is listed in Table 2. The base metals were 14 mm thick, and a single V-groove of 60 • was produced to perform 12-pass manual gas tungsten arc welding. The welding position was 1G, root face was 0 mm, and root gap was 4 mm. Figure 1 shows the schematic of the base metals and groove geometry. Ar with a purity of 99.99% was used as the shield gas and was introduced at a flow rate of 20 L/min. The following welding conditions were used: arc voltage = 8.8-14 V, arc current = 85-253 A, and welding speed = 2.5-15.5 cm/min.

Microstructural Analysis
The WJs were mechanically cross-sectioned, polished, and chemically etched with a 2 vol.% nital solution for the carbon steels and a solution of glycerigia containing 10 mL HNO3, 20 mL glycerol, and 30 mL HCl for the stainless steels. The microstructures and fractographs of the WJs were observed using light optical microscopy and field-emission scanning electron microscopy (FE-SEM). The cross-sectional microstructure of the HSC specimen was examined using electron backscattered diffraction (EBSD) at an acceleration voltage of 18 kV, a working distance of 15 mm, and a step size of 0.05 μm. The raw data were processed using orientation imaging microscopy software (ver. 7.2 of TSL) [19]. To confirm the route of HSC initiation and propagation, the fractured sample was extracted using a focused ion beam (FIB) and analysed using field emission transmission electron microscopy (FE-TEM, Thermo Fisher Scientific, TALOS F200X, Oberkochen, Germany) at 200 kV.

Mechanical Testing and in-situ SSRT
The locations where the tensile specimens were extracted and the Vickers hardness of the WJs was measured are shown in Figure 1. The tensile specimens for the SSRT were prepared in the joint root having no macrodefects and according to ASTM E8 standards. Sub-size specimens were fabricated for the transverse WJs without a notch. To investigate the effect of hydrogen on the strain concentration behaviour of the specimens, digital image correlation (DIC) was performed. The strain distribution of the specimens during the tensile test was observed using a two-dimensional DIC measuring system equipped with a 1/1.7"12 M pixel CCD camera, an ISOCELL bright HMX image sensor (Samsung, Seoul, Korea), and GOM-correlated correction software [20]. The Vickers hardness was measured along the half thickness of the tensile specimen with intervals of 0.2 mm across the

Microstructural Analysis
The WJs were mechanically cross-sectioned, polished, and chemically etched with a 2 vol.% nital solution for the carbon steels and a solution of glycerigia containing 10 mL HNO 3 , 20 mL glycerol, and 30 mL HCl for the stainless steels. The microstructures and fractographs of the WJs were observed using light optical microscopy and field-emission scanning electron microscopy (FE-SEM). The cross-sectional microstructure of the HSC specimen was examined using electron backscattered diffraction (EBSD) at an acceleration voltage of 18 kV, a working distance of 15 mm, and a step size of 0.05 µm. The raw data were processed using orientation imaging microscopy software (ver. 7.2 of TSL) [19]. To confirm the route of HSC initiation and propagation, the fractured sample was extracted using a focused ion beam (FIB) and analysed using field emission transmission electron microscopy (FE-TEM, Thermo Fisher Scientific, TALOS F200X, Oberkochen, Germany) at 200 kV.

Mechanical Testing and In-Situ SSRT
The locations where the tensile specimens were extracted and the Vickers hardness of the WJs was measured are shown in Figure 1. The tensile specimens for the SSRT were prepared in the joint root having no macrodefects and according to ASTM E8 standards. Sub-size specimens were fabricated for the transverse WJs without a notch. To investigate the effect of hydrogen on the strain concentration behaviour of the specimens, digital image correlation (DIC) was performed. The strain distribution of the specimens during the tensile test was observed using a two-dimensional DIC measuring system equipped with a 1/1.7"12 M pixel CCD camera, an ISOCELL bright HMX image sensor (Samsung, Seoul, Korea), and GOM-correlated correction software [20]. The Vickers hardness was Metals 2021, 11, 1039 4 of 15 measured along the half thickness of the tensile specimen with intervals of 0.2 mm across the WJ using a load of 300 g for 15 s. Nano-indentation was performed at intervals of 2.3 µm across the fusion boundary in the weldment using a load of 10 mN.
After polishing the gauge section up to #2400 grit SiC paper, hydrogen was electrochemically pre-charged with a current density of 50 A/m 2 for 48 h in 3% NaCl and 0.3% NH 4 SCN aqueous solution containing As 2 O 3 (0.5 g L −1 ) to prevent the recombination of hydrogen atoms at the specimen surface. On the basis of our previous study, we assumed that the time required for hydrogen to diffuse through the ferrite phase to the centre of the tensile specimen with a thickness of 2 mm would be approximately 5 h [17]. Therefore, considering the difference between the hydrogen charging conditions used in this study and those used in the previous study, pre-charging for 48 h was determined to be sufficient to reach the equilibrium distribution up to the ferrite phase and the interface of the ferrite and austenite phases within the gauge length. The charging solution was renewed after every 24 h, and the charging conditions used during the SSRT were the same as those used in the pre-charging. The SSRT was conducted at a cross-head speed of 2.5 × 10 −5 mm s −1 (nominal strain rate = 10 −6 s −1 ). The schematic of the in situ SSRT is shown in the authors' previous study [17].
To evaluate the HSC sensitivity of the WJs, the relative reduction of area (∆RoA) during the SSRT depending on hydrogen charging was calculated using the following equation where RoA HF and RoA HC denote the reduction of area of the WJs after the SSRT under the hydrogen-free (HF) and hydrogen-charged (HC) environments, respectively. The SSRT was performed at least three times for each condition and the average value was reported.

Silver Decoration to Observe Hydrogen-Trapping Behaviour
The specimen was decorated with silver to observe its trapping behaviour and hydrogen distribution [21,22]. After polishing the specimen to a thickness of 800 µm using #2400 grit SiC paper on one side and a 0.06 µm diamond suspension on the other, specimens were charged with hydrogen from the #2400-polished side for 5 h. Then, a 2 mM K[Ag(CN) 2 ] silver decoration solution was dropped on the opposite side of the specimen at 20 • C for 10 min. The specimen was then cleaned with distilled water, and the silver particles were observed using FE-SEM.  Figure 3 shows the Vickers hardness distribution measured across the WJs along the black-dotted lines. Regardless of the type of the filler metal used, the hardness of the WJs decreased in the following order: DSS > FZ > carbon steel. A hardened HAZ was observed in the narrow area near the fusion line of the DSS side, and a softened HAZ was observed on the carbon-steel side. In particular, softening was more pronounced in the HAZ of CSD than in the HAZ of CSA because of the higher yield strength of CSD, despite the similar chemical compositions of CSD and CSA (Tables 1 and 2). The significant softening in the HAZ of CSD can be attributed to its coarse ferrite grains, as shown in Figure 2f.  Figure 3 shows the Vickers hardness distribution measured across the WJs along the black-dotted lines. Regardless of the type of the filler metal used, the hardness of the WJs decreased in the following order: DSS > FZ > carbon steel. A hardened HAZ was observed in the narrow area near the fusion line of the DSS side, and a softened HAZ was observed on the carbon-steel side. In particular, softening was more pronounced in the HAZ of CSD than in the HAZ of CSA because of the higher yield strength of CSD, despite the similar chemical compositions of CSD and CSA (Tables 1 and 2). The significant softening in the HAZ of CSD can be attributed to its coarse ferrite grains, as shown in Figure Figure 4 shows the engineering stress-displacement curves of the WJs. Figure 4a,b shows the SSRT curves of the WJs with the DF and AF, respectively, in the HF and HC environments. The tensile properties of the WJs with the DF and AF under the HF and HC conditions are listed in Table 3. After the hydrogen charging, the yield strength and tensile strength of both the WJs increased and the displacement decreased. The increased strength of the WJs can be attributed to solid solution strengthening by the oversaturated hydrogen [23][24][25].

Tensile Behaviour of the Transverse WJs with Respect to Hydrogen Charging
The reduction in the displacement of the WJs indicated that hydrogen reduced their plastic deformation. To evaluate the HSC sensitivity of the WJs, their ΔRoAs were measured instead of quantifying the decrease in their displacement. This is because the transverse WJs had a non-uniform microstructure distribution within the gauge section [17]. The ΔRoAs of the WJs with the DF and AF were almost the same (~65%), indicating that hydrogen caused rapid fracturing immediately after reaching the ultimate tensile stress (UTS) with little plastic deformation. In conclusion, both the WJs were sensitive to HSC.   Figure 4 shows the engineering stress-displacement curves of the WJs. Figure 4a,b shows the SSRT curves of the WJs with the DF and AF, respectively, in the HF and HC environments. The tensile properties of the WJs with the DF and AF under the HF and HC conditions are listed in Table 3. After the hydrogen charging, the yield strength and tensile strength of both the WJs increased and the displacement decreased. The increased strength of the WJs can be attributed to solid solution strengthening by the oversaturated hydrogen [23][24][25].  Figure 4 shows the engineering stress-displacement curves of the WJs. Figure 4a,b shows the SSRT curves of the WJs with the DF and AF, respectively, in the HF and HC environments. The tensile properties of the WJs with the DF and AF under the HF and HC conditions are listed in Table 3. After the hydrogen charging, the yield strength and tensile strength of both the WJs increased and the displacement decreased. The increased strength of the WJs can be attributed to solid solution strengthening by the oversaturated hydrogen [23][24][25].

Tensile Behaviour of the Transverse WJs with Respect to Hydrogen Charging
The reduction in the displacement of the WJs indicated that hydrogen reduced their plastic deformation. To evaluate the HSC sensitivity of the WJs, their ΔRoAs were measured instead of quantifying the decrease in their displacement. This is because the transverse WJs had a non-uniform microstructure distribution within the gauge section [17]. The ΔRoAs of the WJs with the DF and AF were almost the same (~65%), indicating that hydrogen caused rapid fracturing immediately after reaching the ultimate tensile stress (UTS) with little plastic deformation. In conclusion, both the WJs were sensitive to HSC.  The reduction in the displacement of the WJs indicated that hydrogen reduced their plastic deformation. To evaluate the HSC sensitivity of the WJs, their ∆RoAs were measured instead of quantifying the decrease in their displacement. This is because the transverse WJs had a non-uniform microstructure distribution within the gauge section [17]. The ∆RoAs of the WJs with the DF and AF were almost the same (~65%), indicating that hydrogen caused rapid fracturing immediately after reaching the ultimate tensile stress (UTS) with little plastic deformation. In conclusion, both the WJs were sensitive to HSC.  Figure 5a,b shows the strain distribution within the gauge lengths of the DF and AF WJs, respectively, during the SSRT under the HF condition. The colour variation from blue to red indicated significant strain distribution. Plastic deformation began in the softened HAZ of the carbon steels, which was associated with the low hardness of both the DF and AF (Figure 3). Subsequently, as the plastic deformation progressed further, strain accumulation was observed throughout the carbon steels, which showed relatively low hardness compared to the DSS and FZ. When the AF was used (with a strength lower than that of the DF), the FZ of the ASS was deformed after reaching UTS, unlike that of the DSS. After reaching UTS, necking occurred in the HAZ of the carbon steels, leading to significant deformation, and hence fracture.  Figure 5a,b shows the strain distribution within the gauge lengths of the DF and AF WJs, respectively, during the SSRT under the HF condition. The colour variation from blue to red indicated significant strain distribution. Plastic deformation began in the softened HAZ of the carbon steels, which was associated with the low hardness of both the DF and AF (Figure 3). Subsequently, as the plastic deformation progressed further, strain accumulation was observed throughout the carbon steels, which showed relatively low hardness compared to the DSS and FZ. When the AF was used (with a strength lower than that of the DF), the FZ of the ASS was deformed after reaching UTS, unlike that of the DSS. After reaching UTS, necking occurred in the HAZ of the carbon steels, leading to significant deformation, and hence fracture.  Figure 6 shows the fracture surfaces and cross-sections of the fractured locations of the WJs after the SSRT under the HF and HC conditions. Figure 6a,b shows the fracture surfaces of the WJs with the DF and AF after the SSRT under the HF condition, respectively. A typical necking fracture was observed with severe plastic deformation during the SSRT. However, Figure 6c,d shows the fracture surfaces of the WJs with the DF and AF, respectively, after the in situ SSRT under the HC conditions. As can be observed from the figures, these WJs fractured with little plastic deformation and necking. These were also observed for the cross-sectioned fracture surfaces of the WJs (Figure 6e-h). Under the HF condition, the DF and AF fractured in the HAZ of the carbon steels, as indicated by the DIC results ( Figure 5), and elongated microstructures were observed along the loading direction because of the severe plastic deformation (Figure 6i,j). Under the HC condition, both the DF and AF fractured along the type-II boundary located on the fusion boundary of the carbon steel side, and slightly deformed microstructures were observed ( Figure  6k,l). Figure 7 shows the detailed fractography of the HSC process. Regardless of the filler metal used, the WJs fractured along the type-II boundary (Figure 6k,l). Thus, WJs with different filler metals showed similar fracture morphologies. Some dimples were observed near the centre of the fractured specimen (Figure 7b), and quasi-cleavage (QC) fracture morphology was observed over the entire surface (Figure 7c-f). Specifically, the  Figure 6 shows the fracture surfaces and cross-sections of the fractured locations of the WJs after the SSRT under the HF and HC conditions. Figure 6a,b shows the fracture surfaces of the WJs with the DF and AF after the SSRT under the HF condition, respectively. A typical necking fracture was observed with severe plastic deformation during the SSRT. However, Figure 6c,d shows the fracture surfaces of the WJs with the DF and AF, respectively, after the in situ SSRT under the HC conditions. As can be observed from the figures, these WJs fractured with little plastic deformation and necking. These were also observed for the cross-sectioned fracture surfaces of the WJs (Figure 6e-h). Under the HF condition, the DF and AF fractured in the HAZ of the carbon steels, as indicated by the DIC results ( Figure 5), and elongated microstructures were observed along the loading direction because of the severe plastic deformation (Figure 6i,j). Under the HC condition, both the DF and AF fractured along the type-II boundary located on the fusion boundary of the carbon steel side, and slightly deformed microstructures were observed (Figure 6k,l). QC fracture morphology involved the intergranular (IG) (Figure 7c,e,f) and stair morphologies (Figure 7d,e). QC and IG cracking are the most common fracture morphologies of HSC [26][27][28][29].

Effect of Type-II Boundary Microstructure on the HSC Behaviour of the WJs
To investigate the effect of the type-II boundary on the HSC of the WJs in detail, their fusion-line regions (Figure 8) on the carbon steel side (red box in Figure 3) were examined using EBSD, energy-dispersive X-ray spectroscopy, and nano-indentation analyses. In the EBSD phase map, the red and green colours denote the face-centred cubic (FCC, austenite) and body-centred cubic (BCC, ferrite or martensite) phases, respectively. The type-II  Figure 7 shows the detailed fractography of the HSC process. Regardless of the filler metal used, the WJs fractured along the type-II boundary (Figure 6k,l). Thus, WJs with different filler metals showed similar fracture morphologies. Some dimples were observed near the centre of the fractured specimen (Figure 7b), and quasi-cleavage (QC) fracture morphology was observed over the entire surface (Figure 7c-f). Specifically, the QC fracture morphology involved the intergranular (IG) (Figure 7c,e,f) and stair morphologies (Figure 7d

Effect of Type-II Boundary Microstructure on the HSC Behaviour of the WJs
To investigate the effect of the type-II boundary on the HSC of the WJs in detail, their fusion-line regions (Figure 8) on the carbon steel side (red box in Figure 3) were examined using EBSD, energy-dispersive X-ray spectroscopy, and nano-indentation analyses. In the EBSD phase map, the red and green colours denote the face-centred cubic (FCC, austenite) and body-centred cubic (BCC, ferrite or martensite) phases, respectively. The type-II

Effect of Type-II Boundary Microstructure on the HSC Behaviour of the WJs
To investigate the effect of the type-II boundary on the HSC of the WJs in detail, their fusion-line regions (Figure 8) on the carbon steel side (red box in Figure 3) were examined using EBSD, energy-dispersive X-ray spectroscopy, and nano-indentation analyses. In the EBSD phase map, the red and green colours denote the face-centred cubic (FCC, austenite) and body-centred cubic (BCC, ferrite or martensite) phases, respectively. The type-II boundary areas are indicated with white-dashed lines; the FZ was on the left and the carbon steel was on the right. The type-II boundaries of both the DF and AF consisted of a mixture of the austenite and lath ferrite regions (Figure 8a-d). This boundary region had a width of less than 20-100 µm throughout the WJs. In this area, sharp composition gradients were observed in the vicinity of the FZ owing to the differences in the compositions (Cr, Ni, Mo, and Mn) of the stainless steels used as the filler metals and the carbon steels used as the base metals (Figure 8e). The Cr, Ni, Mo, and Mn contents of the boundary of the stainless steel decreased rapidly as it approached the carbon steel side, and the composition gradient became stable, which is assumed to be the austenite and lath ferrite regions, respectively.
Metals 2021, 11, x FOR PEER REVIEW 9 of 15 boundary areas are indicated with white-dashed lines; the FZ was on the left and the carbon steel was on the right. The type-II boundaries of both the DF and AF consisted of a mixture of the austenite and lath ferrite regions (Figure 8a-d). This boundary region had a width of less than 20-100 μm throughout the WJs. In this area, sharp composition gradients were observed in the vicinity of the FZ owing to the differences in the compositions (Cr, Ni, Mo, and Mn) of the stainless steels used as the filler metals and the carbon steels used as the base metals (Figure 8e). The Cr, Ni, Mo, and Mn contents of the boundary of the stainless steel decreased rapidly as it approached the carbon steel side, and the composition gradient became stable, which is assumed to be the austenite and lath ferrite regions, respectively. To analyse the lath ferrite in the boundary in detail, a nano-indentation analysis was conducted (Figure 8f). A significant increase in hardness was observed in the range of approximately 20 μm near the fusion line. The type-II boundary region had a higher indentation value than the FZ of the AF with a complete austenite microstructure. Furthermore, the boundary near the FZ of the DF showed higher indentation value than the FZ of the DF, which showed a duplex microstructure. Thus, the lath ferrite present in the type-II boundary was identified to be the martensite phase.  To analyse the lath ferrite in the boundary in detail, a nano-indentation analysis was conducted (Figure 8f). A significant increase in hardness was observed in the range of approximately 20 µm near the fusion line. The type-II boundary region had a higher indentation value than the FZ of the AF with a complete austenite microstructure. Furthermore, the boundary near the FZ of the DF showed higher indentation value than the FZ of the DF, which showed a duplex microstructure. Thus, the lath ferrite present in the type-II boundary was identified to be the martensite phase. Figure 9 shows the phase and kernel average misorientation (KAM) maps of the cross-section of the fractured locations of the WJs with the DF and AF after the in situ hydrogen charging SSRT, as obtained using EBSD. Both the AF and DF fractured in the martensite region at the type-II boundary (Figure 9a,b). In the KAM map (Figure 9c), the colour variation from blue to red indicated an increase in the lattice distortion. This allowed for an indirect comparison of the dislocation densities and strain distributions of the austenite and martensite regions. The austenite region showed little strain (blue), whereas the martensite region showed relatively high strain (red). In the martensite region, a relatively high strain was applied mainly along the martensite lath boundary (green). In addition, the interface between the martensite and austenite phases showed high strain, ranging from green to red. The martensite lath boundary and the interface between the austenite and ferrite phases are the major trapping sites of diffusible hydrogen [30][31][32]. According to the hydrogen-enhanced localised plasticity (HELP) mechanism, the trapped hydrogen intensified the dislocation concentration at the boundaries [33][34][35]. This was associated with the local strain concentration at the martensite-lath boundary and the interface of the austenite and martensite phases. This strain concentration was induced by the hydrogen trapped during the in situ SSRT.
Metals 2021, 11, x FOR PEER REVIEW 10 of 15 Figure 9 shows the phase and kernel average misorientation (KAM) maps of the cross-section of the fractured locations of the WJs with the DF and AF after the in situ hydrogen charging SSRT, as obtained using EBSD. Both the AF and DF fractured in the martensite region at the type-II boundary (Figure 9a,b). In the KAM map (Figure 9c), the colour variation from blue to red indicated an increase in the lattice distortion. This allowed for an indirect comparison of the dislocation densities and strain distributions of the austenite and martensite regions. The austenite region showed little strain (blue), whereas the martensite region showed relatively high strain (red). In the martensite region, a relatively high strain was applied mainly along the martensite lath boundary (green). In addition, the interface between the martensite and austenite phases showed high strain, ranging from green to red. The martensite lath boundary and the interface between the austenite and ferrite phases are the major trapping sites of diffusible hydrogen [30][31][32]. According to the hydrogen-enhanced localised plasticity (HELP) mechanism, the trapped hydrogen intensified the dislocation concentration at the boundaries [33][34][35]. This was associated with the local strain concentration at the martensite-lath boundary and the interface of the austenite and martensite phases. This strain concentration was induced by the hydrogen trapped during the in situ SSRT. Figure 9d shows the high-magnification image of the cross-section of HSC in the martensite region of the type-II boundary. Voids were observed along the martensite-lath boundary. This is consistent with the observation of strain concentration at the martensite-lath boundary (Figure 9c). This strain explains the dimple fracture observed in the HSC fractography (Figure 7b). To explain the effect of hydrogen on this phenomenon, a hydrogen-enhanced strain-induced void (HESIV) mechanism was suggested, which was observed in our previous study [17]. According to the HESIV theory, voids can be easily formed by the hydrogen at the intersections of hydrogen trapping sites such as carbide interfaces, lath boundaries, and dislocations [17,36,37]. To analyse the causes of the IG fracture and stair morphology observed in the HSC fracture surface (Figure 7), the FE-TEM images of the fracture surfaces were extracted using FIB, as shown in Figure 10a,b. The bright-field TEM image was extracted from the red box of the SEM image, in which the IG and stair fracture occurred along the interfaces of the FCC austenite and lath-type BCC martensite (Figure 10a). In addition, the EBSD phase  Figure 9d shows the high-magnification image of the cross-section of HSC in the martensite region of the type-II boundary. Voids were observed along the martensitelath boundary. This is consistent with the observation of strain concentration at the martensite-lath boundary (Figure 9c). This strain explains the dimple fracture observed in the HSC fractography (Figure 7b). To explain the effect of hydrogen on this phenomenon, a hydrogen-enhanced strain-induced void (HESIV) mechanism was suggested, which was observed in our previous study [17]. According to the HESIV theory, voids can be easily formed by the hydrogen at the intersections of hydrogen trapping sites such as carbide interfaces, lath boundaries, and dislocations [17,36,37].
To analyse the causes of the IG fracture and stair morphology observed in the HSC fracture surface (Figure 7), the FE-TEM images of the fracture surfaces were extracted using FIB, as shown in Figure 10a,b. The bright-field TEM image was extracted from the red box of the SEM image, in which the IG and stair fracture occurred along the interfaces of the FCC austenite and lath-type BCC martensite (Figure 10a). In addition, the EBSD phase map, measured from the cross-section of the HSC fracture surface, also showed that cracks occurred along the austenite and martensite interface (Figure 10c). This was consistent with the strain concentration at the martensite and austenite interfaces in Figure 9c.
Metals 2021, 11, x FOR PEER REVIEW 11 of 15 map, measured from the cross-section of the HSC fracture surface, also showed that cracks occurred along the austenite and martensite interface (Figure 10c). This was consistent with the strain concentration at the martensite and austenite interfaces in Figure 9c. Figure 10b shows the magnified image of the dashed blue box shown in Figure 10a. The right side of Figure 10b shows that the lath-martensite region was responsible for the occurrence of the stair fracture (Figure 7d,e). That is, the overall QC fracture exhibiting stair morphology occurred under HC conditions (Figure 10b) because the strain concentration at the martensite-lath boundary caused HSC initiation simultaneously with IG fracture (martensite and austenite interfaces in Figure 10c).

Effect of Trapped Hydrogen on the HSC Fracture of the WJs
To investigate the effect of diffusible hydrogen on the type-II boundary, the hydrogen-charged DF specimen was Ag decorated, as shown in Figure 11. The image shows a type-II boundary area denoted by dashed white lines with the FZ of the DF on the right and the HAZ of the carbon steel on the left. The white particles on the surface of the specimen were Ag particles because the specimen was only polished without etching for the decoration. The Ag particles formed because hydrogen was charged from the other side of the observed surface and diffused through the ferrite phase, reaching the observed surface and reacting with the decoration solution. The region where the particles were observed in the FZ was consistent with the ferrite region of the duplex microstructure in the FZ of the DSS (Figure 8a). The hydrogen diffused and trapped in ferrite of FZ is associated with a larger yield strength for the hydrogen-charged DF joint than that for the hydrogencharged AF joint ( Figure 4 and Table 3). The white band observed in the area marked with dashed white lines corresponds to the martensite region of the type-II boundary. This suggested that a high density of Ag particles was observed because martensite has many  Figure 10b shows the magnified image of the dashed blue box shown in Figure 10a. The right side of Figure 10b shows that the lath-martensite region was responsible for the occurrence of the stair fracture (Figure 7d,e). That is, the overall QC fracture exhibiting stair morphology occurred under HC conditions (Figure 10b) because the strain concentration at the martensite-lath boundary caused HSC initiation simultaneously with IG fracture (martensite and austenite interfaces in Figure 10c).

Effect of Trapped Hydrogen on the HSC Fracture of the WJs
To investigate the effect of diffusible hydrogen on the type-II boundary, the hydrogencharged DF specimen was Ag decorated, as shown in Figure 11. The image shows a type-II boundary area denoted by dashed white lines with the FZ of the DF on the right and the HAZ of the carbon steel on the left. The white particles on the surface of the specimen were Ag particles because the specimen was only polished without etching for the decoration. The Ag particles formed because hydrogen was charged from the other side of the observed surface and diffused through the ferrite phase, reaching the observed surface and reacting with the decoration solution. The region where the particles were observed in the FZ was consistent with the ferrite region of the duplex microstructure in the FZ of the DSS (Figure 8a). The hydrogen diffused and trapped in ferrite of FZ is associated with a larger yield strength for the hydrogen-charged DF joint than that for the hydrogen-charged AF joint ( Figure 4 and Table 3). The white band observed in the area marked with dashed white lines corresponds to the martensite region of the type-II boundary. This suggested that a high density of Ag particles was observed because martensite has many lath boundaries and dislocations, which act as trapping sites for diffusible hydrogen, compared to the HAZ of the carbon steel.
tration gradient. The hydrogen-charged surface had high hydrogen concentration, and hydrogen diffused into the interior of the metal with low hydrogen concentration. Austenite exhibited high hydrogen solubility and very low hydrogen diffusivity at room temperature, which resulted in low hydrogen concentrations even after a long charging time [38,39]. Ferrite, conversely, has low hydrogen solubility and high hydrogen diffusivity; hence, it exhibited sufficient hydrogen concentration at room temperature. Under the hydrogen charging conditions used in this study, the diffusible hydrogen was not trapped inside the austenite of the FZ. This resulted in the diffusion of hydrogen from the ferrite (high hydrogen content) to the austenite (low hydrogen content). Thus, there was a higher hydrogen concentration at the austenite and ferrite interfaces of the FZ. Similarly, hydrogen was diffused from the carbon steel HAZ composed of ferrite to the FZ, which had a duplex microstructure. In addition, the HAZ of the carbon steel showed a polygonal ferrite microstructure without a lath phase; therefore, there were not many trapping sites for the diffusible hydrogen (Figure 2f,i). This is why the diffusible hydrogen was concentrated near the type-II boundaries between the HAZ of the carbon steel and the austenite of the FZ and was trapped predominantly in the martensite band. In conclusion, in this study, an in situ hydrogen-charging SSRT was carried out to analyse the HSC characteristics of dissimilar-metal welds of DSS and carbon steel. The diffusible hydrogen was prevented from escaping to the deformed microstructures during the tensile testing and additional hydrogen was trapped. The DIC results, shown in Figure 5, indicated that the local strain was concentrated in the softened HAZ of the carbon steel at the beginning of tensile testing, indicating the concentration of dislocations in In this study, the HAZ of the carbon steels, which consisted only of the BCC ferrite phase that easily diffuses hydrogen, had more hydrogen. However, Figure 11 shows that a small number of Ag particles were present in the HAZ of the carbon steel. A high density of Ag particles was observed in the interdendritic ferrite region within the austenite region of the FZ and the martensite region of the type-II boundary. This was probably attributed to the austenite in this boundary area, which affected hydrogen diffusion and trapping behaviour. The driving force for hydrogen diffusion into a metal is the hydrogen concentration gradient. The hydrogen-charged surface had high hydrogen concentration, and hydrogen diffused into the interior of the metal with low hydrogen concentration. Austenite exhibited high hydrogen solubility and very low hydrogen diffusivity at room temperature, which resulted in low hydrogen concentrations even after a long charging time [38,39]. Ferrite, conversely, has low hydrogen solubility and high hydrogen diffusivity; hence, it exhibited sufficient hydrogen concentration at room temperature. Under the hydrogen charging conditions used in this study, the diffusible hydrogen was not trapped inside the austenite of the FZ. This resulted in the diffusion of hydrogen from the ferrite (high hydrogen content) to the austenite (low hydrogen content). Thus, there was a higher hydrogen concentration at the austenite and ferrite interfaces of the FZ. Similarly, hydrogen was diffused from the carbon steel HAZ composed of ferrite to the FZ, which had a duplex microstructure. In addition, the HAZ of the carbon steel showed a polygonal ferrite microstructure without a lath phase; therefore, there were not many trapping sites for the diffusible hydrogen (Figure 2f,i). This is why the diffusible hydrogen was concentrated near the type-II boundaries between the HAZ of the carbon steel and the austenite of the FZ and was trapped predominantly in the martensite band.
In conclusion, in this study, an in situ hydrogen-charging SSRT was carried out to analyse the HSC characteristics of dissimilar-metal welds of DSS and carbon steel. The diffusible hydrogen was prevented from escaping to the deformed microstructures during the tensile testing and additional hydrogen was trapped. The DIC results, shown in Figure 5, indicated that the local strain was concentrated in the softened HAZ of the carbon steel at the beginning of tensile testing, indicating the concentration of dislocations in the HAZ of the carbon steel. Because dislocations are important trapping sites for diffusible hydrogen, the content of the diffusible hydrogen increased in the HAZ of the carbon steel during the in situ SSRT [17,[40][41][42][43]. However, the HAZ of the carbon steel was mainly composed of polygonal ferrite, which did not interact with hydrogen; thus it avoided HSC. Conversely, the strain and diffusible hydrogen concentrated in the HAZ of the carbon steel at the beginning of the tensile testing increased the trapping of diffusible hydrogen in the adjacent high-hardness martensite region. As a result, the type-II boundary became the initiation site for HSC fractures.

Conclusions
In this study, we investigated the effect of the microstructure of dissimilar-metal welds of DSS and carbon steel on their HSC behaviour. The main findings of the study are summarised as follows: 1.
The increase in the ferrite fraction of the DSS HAZ did not affect the fracture mechanism of the dissimilar-metal welds under HF and HC conditions. The softened HAZ of the carbon steel consisting of large polygonal ferrite grains was indirectly associated with the fracture of the welds under HF conditions. 2.
The WJs had almost the same ∆RoAs regardless of the filler metal used. Thus, no significant difference was observed in HSC sensitivities of the WJs with the AF and DF.

3.
Both the DF and AF WJs had type-II boundaries consisting of the austenite and martensite band regions along the fusion line of the carbon steel side, where s HSC fracture occurred.

4.
The martensite band of the type-II boundary trapped a large amount of diffusible hydrogen owing to the presence of a large number of hydrogen trap sites and the hydrogen gradient between the austenite and ferrite regions.

5.
The type-II boundary caused an IG fracture at the interface of the austenite and martensite regions (a strong hydrogen trapping site) and a stair morphology fracture along the martensite-lath boundary. Eventually, a premature fracture occurred during the in situ SSRT in the type-II boundary because of the HESIV and HELP mechanisms.
Dissimilar-metal welds having a duplex phase with austenite and ferrite exhibit various microstructures that are susceptible to hydrogen trapping and HSC fracture. Even though some weld microstructures such as the austenite and martensite bands in type-II boundaries are harmless under normal HF conditions, they can have a negative effect in a hydrogen atmosphere, resulting in premature rupture of DSS and carbon steel welds. To prevent this, a material design to reduce or eliminate the generation of type-II boundaries in the FZ of the carbon steel side must be developed.

Informed Consent Statement: Not applicable.
Data Availability Statement: The raw/processed data required to reproduce these findings cannot be shared at this time, as the data also forms a part of an ongoing study.