Sulfide Stress Cracking Behavior of a Martensitic Steel Controlled by Tempering Temperature

A medium-carbon Cr–Mo–V martensitic steel was thermally processed by quenching (Q) at 890 °C and tempering (T) at increasing temperatures from 650 °C to 720 °C and the effect of tempering temperature, Tt, on sulfide stress cracking (SSC) behaviors was estimated mainly via double cantilever beam (DCB) and electrochemical hydrogen permeation (EHP) tests and microstructure characterization. The results indicate that the threshold stress intensity factor for SSC, KISSC, increased with increasing Tt. The overall and local H concentration around the inclusions decreased with increasing Tt, due to reductions in the amounts of solute atoms, grain boundaries and dislocations, which effectively prevented SSC initiation. Also, increasing Tt caused an increased fraction of high-angle boundaries, which evidently lowered the SSC propagation rate by more frequently diverting the propagating direction and accordingly restricted SSC propagation. The overall SSC resistance of this Q&T–treated steel was therefore significantly enhanced.

Among these efforts, the correlations between various metallurgical factors and the susceptibility to SSC/HIC have recently received increasing attention [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. Once these steels are exposed to the sour environment, atomic hydrogen (H) can form on the material surface due to the H + →H reaction [11], which depends on environmental and surface conditions. After the H atoms enter the material, these metallurgical factors might have their respective impacts on H mobility or diffusion [10], by acting as reversible (solute atoms, dislocations, grain boundaries, etc.) or irreversible (nonmetallic inclusions, precipitated particles, voids, etc.) traps for H atoms, depending on their binding enthalpies with H [12]  The instantaneous SSC propagation length for each Q&T-treated sample was measured with the DCB test interrupted, according to the NACE (National Association of Corrosion Engineer) TM0177 standard [31] and evaluated as a function of immersing time (t), as shown in Figure 1. This reveals that SSC probably developed in three stages: initiation, rapid and palliative propagation. For the first stage, SSC was not detected for a couple of hours, the inoculation period after the DCB specimen was immersed in the solution, during which the diffusion of H atoms into the crystalline interstices and voids was assumed to be accumulated for initializing the H damage. For the second stage, SSC grew rapidly with the simultaneous assistance of the highly concentrated stress and the accumulated H atoms. For the third stage, SSC grew at a gradual reduced rate and approached a metastable state, where its propagation was almost imperceptible, probably due to the relaxation of concentrated stress and the approximately saturated H atom concentration. The instantaneous SSC propagation length for each Q&T-treated sample was measured with the DCB test interrupted, according to the NACE (National Association of Corrosion Engineer) TM0177 standard [31] and evaluated as a function of immersing time (t), as shown in Figure 1. This reveals that SSC probably developed in three stages: initiation, rapid and palliative propagation. For the first stage, SSC was not detected for a couple of hours, the inoculation period after the DCB specimen was immersed in the solution, during which the diffusion of H atoms into the crystalline interstices and voids was assumed to be accumulated for initializing the H damage. For the second stage, SSC grew rapidly with the simultaneous assistance of the highly concentrated stress and the accumulated H atoms. For the third stage, SSC grew at a gradual reduced rate and approached a metastable state, where its propagation was almost imperceptible, probably due to the relaxation of concentrated stress and the approximately saturated H atom concentration. As also indicated in Figure 1, the steel tempered at 720 °C had both the longest inoculation period for detected SSC and the least SSC propagation length among these Q&T-treated samples. SSC initiated over a longer time and propagated at a decreased length with the increased Tt.
SSC propagation behaviors were also interpreted in terms of the correlation between da/dt and KIapplied, as shown in Figure 2. This indicates that da/dt decreased as a result of decreasing KIapplied. With KIapplied decreased to a certain value, da/dt approached zero, where KISSC was obtained, according to the NACE TM0177 standard [31]. Furthermore, da/dt decreased and KISSC increased with increasing Tt, indicating that the high-temperature tempering retarded SSC propagation and correspondingly enhanced SSC resistance.  As also indicated in Figure 1, the steel tempered at 720 • C had both the longest inoculation period for detected SSC and the least SSC propagation length among these Q&T-treated samples. SSC initiated over a longer time and propagated at a decreased length with the increased T t .
SSC propagation behaviors were also interpreted in terms of the correlation between da/dt and K Iapplied , as shown in Figure 2. This indicates that da/dt decreased as a result of decreasing K Iapplied . With K Iapplied decreased to a certain value, da/dt approached zero, where K ISSC was obtained, according to the NACE TM0177 standard [31]. Furthermore, da/dt decreased and K ISSC increased with increasing T t , indicating that the high-temperature tempering retarded SSC propagation and correspondingly enhanced SSC resistance. The instantaneous SSC propagation length for each Q&T-treated sample was measured with the DCB test interrupted, according to the NACE (National Association of Corrosion Engineer) TM0177 standard [31] and evaluated as a function of immersing time (t), as shown in Figure 1. This reveals that SSC probably developed in three stages: initiation, rapid and palliative propagation. For the first stage, SSC was not detected for a couple of hours, the inoculation period after the DCB specimen was immersed in the solution, during which the diffusion of H atoms into the crystalline interstices and voids was assumed to be accumulated for initializing the H damage. For the second stage, SSC grew rapidly with the simultaneous assistance of the highly concentrated stress and the accumulated H atoms. For the third stage, SSC grew at a gradual reduced rate and approached a metastable state, where its propagation was almost imperceptible, probably due to the relaxation of concentrated stress and the approximately saturated H atom concentration. As also indicated in Figure 1, the steel tempered at 720 °C had both the longest inoculation period for detected SSC and the least SSC propagation length among these Q&T-treated samples. SSC initiated over a longer time and propagated at a decreased length with the increased Tt.
SSC propagation behaviors were also interpreted in terms of the correlation between da/dt and KIapplied, as shown in Figure 2. This indicates that da/dt decreased as a result of decreasing KIapplied. With KIapplied decreased to a certain value, da/dt approached zero, where KISSC was obtained, according to the NACE TM0177 standard [31]. Furthermore, da/dt decreased and KISSC increased with increasing Tt, indicating that the high-temperature tempering retarded SSC propagation and correspondingly enhanced SSC resistance.   Figure 3 shows a typical SSC fracture morphology of the DCB specimen tempered at 700 • C. It can be seen that the fracture surface was covered with a corrosion product film that was rich in Fe and S, as examined by EDS (energy dispersive spectroscopy) analysis, illustrated in Figure 4. The corrosion film was loose and featured microcracks and micropores, probably leading to poor resistance to the migration of HS − or S 2− to the fresh material surface.  Figure 3 shows a typical SSC fracture morphology of the DCB specimen tempered at 700 °C. It can be seen that the fracture surface was covered with a corrosion product film that was rich in Fe and S, as examined by EDS (energy dispersive spectroscopy) analysis, illustrated in Figure 4. The corrosion film was loose and featured microcracks and micropores, probably leading to poor resistance to the migration of HS − or S 2− to the fresh material surface.  The corrosion film was subsequently removed with a solution of dilute hydrochloric acid and hexamethylene tetramine and the real SSC fracture surface is shown in Figure 5. Actually, the SSC fracture was a typical quasi-cleavage cracking, featuring some random secondary cracks underneath the fracture surface, consistent with an early work reported by Dadfarnia et al. [28]. In addition, some intergranular fracture areas were also perceptible on the fracture surface in all the DCB samples.  Figure 3 shows a typical SSC fracture morphology of the DCB specimen tempered at 700 °C. It can be seen that the fracture surface was covered with a corrosion product film that was rich in Fe and S, as examined by EDS (energy dispersive spectroscopy) analysis, illustrated in Figure 4. The corrosion film was loose and featured microcracks and micropores, probably leading to poor resistance to the migration of HS − or S 2− to the fresh material surface.  The corrosion film was subsequently removed with a solution of dilute hydrochloric acid and hexamethylene tetramine and the real SSC fracture surface is shown in Figure 5. Actually, the SSC fracture was a typical quasi-cleavage cracking, featuring some random secondary cracks underneath the fracture surface, consistent with an early work reported by Dadfarnia et al. [28]. In addition, some intergranular fracture areas were also perceptible on the fracture surface in all the DCB samples. The corrosion film was subsequently removed with a solution of dilute hydrochloric acid and hexamethylene tetramine and the real SSC fracture surface is shown in Figure 5. Actually, the SSC fracture was a typical quasi-cleavage cracking, featuring some random secondary cracks underneath the fracture surface, consistent with an early work reported by Dadfarnia et al. [28]. In addition, some intergranular fracture areas were also perceptible on the fracture surface in all the DCB samples.  Figure 3 shows a typical SSC fracture morphology of the DCB specimen tempered at 700 °C. It can be seen that the fracture surface was covered with a corrosion product film that was rich in Fe and S, as examined by EDS (energy dispersive spectroscopy) analysis, illustrated in Figure 4. The corrosion film was loose and featured microcracks and micropores, probably leading to poor resistance to the migration of HS − or S 2− to the fresh material surface.  The corrosion film was subsequently removed with a solution of dilute hydrochloric acid and hexamethylene tetramine and the real SSC fracture surface is shown in Figure 5. Actually, the SSC fracture was a typical quasi-cleavage cracking, featuring some random secondary cracks underneath the fracture surface, consistent with an early work reported by Dadfarnia et al. [28]. In addition, some intergranular fracture areas were also perceptible on the fracture surface in all the DCB samples.

Prior Austenite Grain Observation
The typical prior austenite grain (PAG) morphology of the samples quenched at an identical temperature of 890 °C, waiting for further tempering at differing Tt, is shown in Figure 6. The average PAG size, Dγ, was quantitatively estimated and is summarized in Table 2. As the table states, Dγ remained at almost the same level of 7.5 μm under identical quenching conditions.

Martensitic Structure Observations
The PAG of martensitic steel can normally be partitioned into three types of martensitic substructures: packet, block and lath [32,33]. The martensitic packet and block in each Q&T-treated sample were also characterized via SEM and electron back-scattered diffraction (EBSD), as shown in Figures 7 and 8, respectively. The average packet and block size, Dp and Db, were quantitatively determined and are summarized in Table 2. This demonstrates that Dp and Db increased with increasing Tt from 650 °C to 720 °C. The grain boundary misorientation angle (GBMA) might play a significant role in controlling SSC resistance by affecting hydrogen permeability [20,34] and crack propagation [21,26]. Therefore, the GBMA distribution of martensitic substructure in each Q&T-treated sample was also analyzed via EBSD and the fraction of the HAGBs, fGBMA≥15°, is summarized in Table 2.

Prior Austenite Grain Observation
The typical prior austenite grain (PAG) morphology of the samples quenched at an identical temperature of 890 • C, waiting for further tempering at differing T t , is shown in Figure 6. The average PAG size, D γ , was quantitatively estimated and is summarized in Table 2. As the table states, D γ remained at almost the same level of 7.5 µm under identical quenching conditions.

Prior Austenite Grain Observation
The typical prior austenite grain (PAG) morphology of the samples quenched at an identical temperature of 890 °C, waiting for further tempering at differing Tt, is shown in Figure 6. The average PAG size, Dγ, was quantitatively estimated and is summarized in Table 2. As the table states, Dγ remained at almost the same level of 7.5 μm under identical quenching conditions.

Martensitic Structure Observations
The PAG of martensitic steel can normally be partitioned into three types of martensitic substructures: packet, block and lath [32,33]. The martensitic packet and block in each Q&T-treated sample were also characterized via SEM and electron back-scattered diffraction (EBSD), as shown in Figures 7 and 8, respectively. The average packet and block size, Dp and Db, were quantitatively determined and are summarized in Table 2. This demonstrates that Dp and Db increased with increasing Tt from 650 °C to 720 °C. The grain boundary misorientation angle (GBMA) might play a significant role in controlling SSC resistance by affecting hydrogen permeability [20,34] and crack propagation [21,26]. Therefore, the GBMA distribution of martensitic substructure in each Q&T-treated sample was also analyzed via EBSD and the fraction of the HAGBs, fGBMA≥15°, is summarized in Table 2.  Table 2. Quantitative determinations of PAG and martensitic substructure features in Q&T-treated samples at differing T t .

Martensitic Structure Observations
The PAG of martensitic steel can normally be partitioned into three types of martensitic substructures: packet, block and lath [32,33]. The martensitic packet and block in each Q&T-treated sample were also characterized via SEM and electron back-scattered diffraction (EBSD), as shown in Figures 7 and 8, respectively. The average packet and block size, D p and D b , were quantitatively determined and are summarized in Table 2. This demonstrates that D p and D b increased with increasing T t from 650 • C to 720 • C. The grain boundary misorientation angle (GBMA) might play a significant role in controlling SSC resistance by affecting hydrogen permeability [20,34] and crack propagation [21,26]. Therefore, the GBMA distribution of martensitic substructure in each Q&T-treated sample was also analyzed via EBSD and the fraction of the HAGBs, f GBMA≥15 • , is summarized in Table 2. As presented, HAGBs were in the majority in each sample and f GBMA≥15 • increased slightly with increasing T t from 650 to 720 • C.
Materials 2018, 11, x FOR PEER REVIEW 6 of 18 As presented, HAGBs were in the majority in each sample and fGBMA≥15° increased slightly with increasing Tt from 650 to 720 °C.

Martensitic Lath and Precipitate Observations
TEM micrographs of each Q&T-treated sample are presented in Figure 9. As the figure shows, the martensite consisted of regular laths in the minority and irregular ones in the majority. The irregular laths had a curved boundary, which were assumed to derive from the polygonization of prior regular laths in as-quenched samples, due to tempering at a high Tt. In contrast, a small As presented, HAGBs were in the majority in each sample and fGBMA≥15° increased slightly with increasing Tt from 650 to 720 °C.

Martensitic Lath and Precipitate Observations
TEM micrographs of each Q&T-treated sample are presented in Figure 9. As the figure shows, the martensite consisted of regular laths in the minority and irregular ones in the majority. The irregular laths had a curved boundary, which were assumed to derive from the polygonization of prior regular laths in as-quenched samples, due to tempering at a high Tt. In contrast, a small

Martensitic Lath and Precipitate Observations
TEM micrographs of each Q&T-treated sample are presented in Figure 9. As the figure shows, the martensite consisted of regular laths in the minority and irregular ones in the majority. The irregular laths had a curved boundary, which were assumed to derive from the polygonization of prior regular laths in as-quenched samples, due to tempering at a high T t . In contrast, a small quantity of regular laths might possess a particular crystallographic orientation with high thermodynamic stability [35] and could survive heavy tempering. Also, as this figure shows, the polygonized irregular martensitic laths evidently increased with elevating T t , at the expense of the regular martensitic laths. The average lath width, W l , was further determined quantitatively and is also summarized in Table 2. W l obviously increased with increasing T t , indicating a widening of the martensitic lath, probably due to the interface migration or the merging of two neighboring laths [36]. quantity of regular laths might possess a particular crystallographic orientation with high thermodynamic stability [35] and could survive heavy tempering. Also, as this figure shows, the polygonized irregular martensitic laths evidently increased with elevating Tt, at the expense of the regular martensitic laths. The average lath width, Wl, was further determined quantitatively and is also summarized in Table 2. Wl obviously increased with increasing Tt, indicating a widening of the martensitic lath, probably due to the interface migration or the merging of two neighboring laths [36]. There still appeared to be a large quantity of rod-like and spherical precipitated particles distributed in the martensitic laths or at their boundaries, as also shown in Figure 9. Meanwhile, the rod-like precipitated particles were spheroidized gradually as the Tt increased. The average equivalent diameter, D and the volume fraction, f, of the precipitates were further estimated quantitatively and are provided as the function of Tt in Figure 10. This shows that both the D and f of precipitates obviously increased with increasing Tt. There still appeared to be a large quantity of rod-like and spherical precipitated particles distributed in the martensitic laths or at their boundaries, as also shown in Figure 9. Meanwhile, the rod-like precipitated particles were spheroidized gradually as the T t increased. The average equivalent diameter, D and the volume fraction, f, of the precipitates were further estimated quantitatively and are provided as the function of T t in Figure 10. This shows that both the D and f of precipitates obviously increased with increasing T t .

Dislocation Estimations
The dislocation density, ρ, of each Q&T-treated sample was determined by XRD pattern, as shown in Figure 11 and summarized in Table 2. The dislocation density decreased from 1.41 × 10 14 to 0.62 × 10 14 m −2 with increasing Tt from 650 °C to 720 °C. The present result is in agreement with a previous work by Masakatsu et al. [37]. They reported that the half-width of the (211) peak in the XRD pattern, indicating the dislocation density, reduced with elevated Tt due to the addition of 0.1 wt % vanadium to the 0.5Cr-0.7Mo steel.  Figure 12 shows the H permeation curves for Q&T-treated samples. The H diffusion coefficient, D0 and the diffusion H concentration, C0, were calculated with Equations (10) and (11) and the results are summarized in Table 3. As the table shows, D0 increased and C0 decreased with increasing Tt, indicating lower H permeability.

Dislocation Estimations
The dislocation density, ρ, of each Q&T-treated sample was determined by XRD pattern, as shown in Figure 11 and summarized in Table 2. The dislocation density decreased from 1.41 × 10 14 to 0.62 × 10 14 m −2 with increasing T t from 650 • C to 720 • C. The present result is in agreement with a previous work by Masakatsu et al. [37]. They reported that the half-width of the (211) peak in the XRD pattern, indicating the dislocation density, reduced with elevated T t due to the addition of 0.1 wt % vanadium to the 0.5Cr-0.7Mo steel.

Dislocation Estimations
The dislocation density, ρ, of each Q&T-treated sample was determined by XRD pattern, as shown in Figure 11 and summarized in Table 2. The dislocation density decreased from 1.41 × 10 14 to 0.62 × 10 14 m −2 with increasing Tt from 650 °C to 720 °C. The present result is in agreement with a previous work by Masakatsu et al. [37]. They reported that the half-width of the (211) peak in the XRD pattern, indicating the dislocation density, reduced with elevated Tt due to the addition of 0.1 wt % vanadium to the 0.5Cr-0.7Mo steel.  Figure 12 shows the H permeation curves for Q&T-treated samples. The H diffusion coefficient, D0 and the diffusion H concentration, C0, were calculated with Equations (10) and (11) and the results are summarized in Table 3. As the table shows, D0 increased and C0 decreased with increasing Tt, indicating lower H permeability.  Figure 12 shows the H permeation curves for Q&T-treated samples. The H diffusion coefficient, D 0 and the diffusion H concentration, C 0 , were calculated with Equations (10) and (11) and the results are summarized in Table 3. As the table shows, D 0 increased and C 0 decreased with increasing T t , indicating lower H permeability. Materials 2018, 11, x FOR PEER REVIEW 9 of 18

Effect of Tempering Temperature on H Permeability
C0, indicating H permeability, decreased with increasing Tt, as shown in Figure 12 and Table 3. Actually, C0 is dominantly dependent on the density of reversible H traps, NT, on the assumption that the irreversible traps could be saturated by the H atoms [38]. According to previous works [39][40][41][42], H traps such as solute atoms, GBs and dislocations were generally regarded as reversible traps, due to their bonding energy being lower than 60 kJ/mol. Among these reversible traps, the solute atoms were assumed to decrease with increasing Tt, due to increased precipitation, as shown in Figures 9 and 10. Accordingly, the number of reversible traps provided by the solute atoms, NT-SA, also decreased with increasing Tt. However, the precipitates in this Q&T-treated Cr-Mo-V-Ti-B martensitic steel were highly diversified and complicated in their structure [43] and hence it is rather intricate to discern between the atoms consumed for these precipitates and the residual solute atoms. Quantification of NT-SA therefore remains challenging. Accordingly, the NT for these Q&T-treated samples at differing Tt were quantified mainly as functions of the GBs (PAG, martensitic packet/block/lath boundaries, etc.) and the dislocations, respectively, to better understand their effects on C0.
The H trap density provided by the GBs, NT-GB, was estimated, with the grain assumed to be an approximate sphere and counted as n per cubic meter, using the following equation [44]: where VG and SG are the volume and surface area of one grain, respectively; the factor of 1/2 is due to one grain plane shared by two grains; xa is atomic distance, 0.26 nm; xGB is the width of GB, about 10 times that of xa. Equation (1) can then be simplified as:

Effect of Tempering Temperature on H Permeability
C 0 , indicating H permeability, decreased with increasing T t , as shown in Figure 12 and Table 3. Actually, C 0 is dominantly dependent on the density of reversible H traps, N T , on the assumption that the irreversible traps could be saturated by the H atoms [38]. According to previous works [39][40][41][42], H traps such as solute atoms, GBs and dislocations were generally regarded as reversible traps, due to their bonding energy being lower than 60 kJ/mol. Among these reversible traps, the solute atoms were assumed to decrease with increasing T t , due to increased precipitation, as shown in Figures 9 and 10. Accordingly, the number of reversible traps provided by the solute atoms, N T-SA , also decreased with increasing T t . However, the precipitates in this Q&T-treated Cr-Mo-V-Ti-B martensitic steel were highly diversified and complicated in their structure [43] and hence it is rather intricate to discern between the atoms consumed for these precipitates and the residual solute atoms. Quantification of N T-SA therefore remains challenging. Accordingly, the N T for these Q&T-treated samples at differing T t were quantified mainly as functions of the GBs (PAG, martensitic packet/block/lath boundaries, etc.) and the dislocations, respectively, to better understand their effects on C 0 .
The H trap density provided by the GBs, N T-GB , was estimated, with the grain assumed to be an approximate sphere and counted as n per cubic meter, using the following equation [44]: where V G and S G are the volume and surface area of one grain, respectively; the factor of 1/2 is due to one grain plane shared by two grains; x a is atomic distance, 0.26 nm; x GB is the width of GB, about 10 times that of x a . Equation (1) can then be simplified as: where d G is the grain size. This equation indicates that the N T-GB is in inverse proportion to d G , by which N T-GB for these Q&T-treated samples were estimated and are listed in Table 4.
The H trap density provided by the dislocations, N T-dis , was estimated by: where ρ dis is dislocation density and the factor of 5 is assumed to be the ratio of the dislocation width to x a . The N T-dis for these Q&T-treated samples were calculated and are also listed in Table 4.  As Table 4 shows, the estimated N T-MP , N T-MB and N T-ML provided by the martensitic packet, block and lath boundaries decreased with their increasing size and accordingly their decreasing number of boundaries, whereas N T-PAG remained at almost the same level. As the martensitic lath boundaries normally occupied all the PAG and the martensitic packet/block boundaries, N T-ML contained N T-PAG , N T-MP and N T-MB . In general, the PAG and the martensitic packet/block boundaries were confirmed to be HAGBs [45] and correspondingly the effective reversible traps [34], whereas the lath boundaries in as-quenched martensitic steels were normally accepted as LAGBs with MA ≤ 10 • [32], probably contributing little to the C 0 [34]. However, according to the EBSD analyses in Figure 8 and the actual TEM observations in Figure 9, the HAGBs surrounding the irregular laths constituted the major fraction of the martensitic lath boundaries due to their polygonization via heavy tempering, essentially supplying much to the C 0 . In contrast, the remnant LAGBs encircling the regular laths comprised a minor fraction of the martensitic lath boundaries, still contributing little to the C 0 . Thus, the final N T-GB was calculated approximately by combining the N T-ML (in Table 4) and the fraction of HAGB, f GBMA≥15 • , essentially for the polygonized irregular martensitic lath (in Table 2), that is, Table 4.
Consequently, the amount of reversible H traps provided by the various GBs, the dislocations and even the solute atoms all decreased with increasing T t and therefore the overall diffusion H concentration (C 0 ) decreased markedly, as shown in Table 3.

Effect of H Permeability on SSC Initiation
A certain number of inclusions with a large size of more than 3 µm were still present in this experimental steel, although it was prepared by melting in an electric furnace and purifying in a vacuum degassing furnace. There appeared mainly two types of complex inclusions in each DCB sample: the massive Al 2 O 3 ·SiO·CaO and the slender Al 2 O 3 ·CaO·TiO particles, as shown in Figure 13a,b. The slender one was typically located in a corrosion pit where SSC was initiating. This observation was in agreement with previous work by Schiapparelli et al., in which HIC was triggered at a non-deformable inclusion of Al 2 O 3 [17]. In addition, the local H concentration, Cσ, of the highly stressed regions in the DCB sample can be estimated using the following equation [46]: where Co is the diffusion H concentration, σh is the hydrostatic stress, VH is the partial molar volume of H in iron (2.1 × 10 −6 m 3 /mol), R is the gas constant and T is the ambient temperature. As this equation indicates, Cσ decreased as the result of decreased C0. When the H-induced stress concentration exceeded the atomic binding force, SSC eventually initiated. Susceptibility to SSC initiation can be estimated by the incubation time for SSC origination, ti, which was attained by fitting the crack propagation length (Δa) vs. the time curve, as demonstrated in Figure 1. The optimum imitated Equations (5)-(7) for the samples tempered at 650, 700 and 720 °C, respectively, as follows: In addition, the local H concentration, C σ , of the highly stressed regions in the DCB sample can be estimated using the following equation [46]: where C 0 is the diffusion H concentration, σ h is the hydrostatic stress, V H is the partial molar volume of H in iron (2.1 × 10 −6 m 3 /mol), R is the gas constant and T is the ambient temperature. As this equation indicates, C σ decreased as the result of decreased C 0 . When the H-induced stress concentration exceeded the atomic binding force, SSC eventually initiated. Susceptibility to SSC initiation can be estimated by the incubation time for SSC origination, t i , which was attained by fitting the crack propagation length (∆a) vs. the time curve, as demonstrated in Supposing ∆a = 0, the t i for the Q&T-treated samples, tempered at 650, 700 and 720 • C was 1.5, 2.2 and 10.4 h, respectively, indicating a lowered susceptibility to SSC initiation with increasing T t .
Therefore, the increasing T t led mainly to martensitic lath polygonization and widening, providing a decreased amount of HAGBs and accordingly reversible H traps, inducing a lowered overall and local diffusion H concentration (C 0 and C σ ), leading to a depressed stress concentration around the inclusions and simultaneously a prolonged incubation time and finally decreased susceptibility to SSC initiation.

Effect of GBs on SSC Propagation
To further understand the effect of GBs with certain crystallographic orientations on crack propagation, secondary microcracks underneath the SSC fracture surface of DCB samples were observed and the corresponding misorientation angles between grains adjacent to the crack were analyzed by EBSD, as shown in Figure 14 and Table 5. LAGB/HAGB is generally defined as the grain boundary with a misorientation angle of 2-15 • />15 • [47], indicated by the white/black line. It can be seen that the microcrack propagated unimpededly through the LAGBs within grains 1, 2 and 3, indicating that the LAGBs could not hinder the SSC from propagating. In contrast, the microcracks diverted direction when they came across the HAGBs of grains 2 and 3 and finally arrested also at the HAGBs of grains 3 and 4.
Supposing Δa = 0, the ti for the Q&T-treated samples, tempered at 650, 700 and 720 °C was 1.5, 2.2 and 10.4 h, respectively, indicating a lowered susceptibility to SSC initiation with increasing Tt.
Therefore, the increasing Tt led mainly to martensitic lath polygonization and widening, providing a decreased amount of HAGBs and accordingly reversible H traps, inducing a lowered overall and local diffusion H concentration (C0 and Cσ), leading to a depressed stress concentration around the inclusions and simultaneously a prolonged incubation time and finally decreased susceptibility to SSC initiation.

Effect of GBs on SSC Propagation
To further understand the effect of GBs with certain crystallographic orientations on crack propagation, secondary microcracks underneath the SSC fracture surface of DCB samples were observed and the corresponding misorientation angles between grains adjacent to the crack were analyzed by EBSD, as shown in Figure 14 and Table 5. LAGB/HAGB is generally defined as the grain boundary with a misorientation angle of 2-15°/>15° [47], indicated by the white/black line. It can be seen that the microcrack propagated unimpededly through the LAGBs within grains 1, 2 and 3, indicating that the LAGBs could not hinder the SSC from propagating. In contrast, the microcracks diverted direction when they came across the HAGBs of grains 2 and 3 and finally arrested also at the HAGBs of grains 3 and 4.  Table 5. Effects of HAGBs on SSC propagation behavior shown in Figure 14. Hence, HAGB could play a significant role in hindering SSC propagation, whereas LAGB failed to provide an evident barrier, showing the two different impacts on crack propagation. According to previous observations [48], the change in operating slip system was essential for the GB to act as a barrier for the dislocation movement when the crack came across the GB. For the HAGB, its own  Table 5. Effects of HAGBs on SSC propagation behavior shown in Figure 14. Hence, HAGB could play a significant role in hindering SSC propagation, whereas LAGB failed to provide an evident barrier, showing the two different impacts on crack propagation. According to previous observations [48], the change in operating slip system was essential for the GB to act as a barrier for the dislocation movement when the crack came across the GB. For the HAGB, its own operated slip system might induce a dislocation-free zone formation in the adjacent grain and accordingly lead to a significantly diverted direction of the crack. In contrast, LAGB could only yield a relatively low retardation of crack propagation, due to the minor difference in orientation between the two neighboring grains.
The increasing T t brought about martensitic lath polygonization and widening to a higher degree, providing an increased fraction of HAGBs and also an increased mean GBMA, restraining SSC propagation by lowering their propagation rate (Figure 2), finally resulting in an elevated K ISSC value.

Materials and Heat Treatment
The 28CrMo48VTiB martensitic steel used for this work was melted in an electric furnace, continuously cast with argon shielding and thermally pieced-rolled-stretched into a seamless tube 177.8 × 10.36 mm (diameter × thickness). The chemical compositions are presented in Table 6. The samples with dimension 300 × 100 mm were cut from the steel tube, reheated at 890 • C for 20 min and water quenched and consequently tempered at differing T t of 650, 700 and 720 • C for 60 min.

Estimations of Mechanical Properties and SSC Susceptibility
The conventional tensile properties of each Q&T-treated sample were determined according to the ASTM (American Society for Testing Material) Standard A370-10 [49]. Each sample was machined along the longitudinal direction of the tube, with geometry and dimensions as shown in Figure 15 and tensioned in a CMT5305 Instron machine at room temperature. Two tensions for each sample were performed and their average values are reported. The hardness for each sample was also determined in a Rockwell hardness tester using a load of 1500 N. For precision, the hardness at 6 different positions in each sample were measured and their average values are reported.
Materials 2018, 11, x FOR PEER REVIEW 13 of 18 operated slip system might induce a dislocation-free zone formation in the adjacent grain and accordingly lead to a significantly diverted direction of the crack. In contrast, LAGB could only yield a relatively low retardation of crack propagation, due to the minor difference in orientation between the two neighboring grains.
The increasing Tt brought about martensitic lath polygonization and widening to a higher degree, providing an increased fraction of HAGBs and also an increased mean GBMA, restraining SSC propagation by lowering their propagation rate (Figure 2), finally resulting in an elevated KISSC value.

Materials and Heat Treatment
The 28CrMo48VTiB martensitic steel used for this work was melted in an electric furnace, continuously cast with argon shielding and thermally pieced-rolled-stretched into a seamless tube 177.8 × 10.36 mm (diameter × thickness). The chemical compositions are presented in Table 6. The samples with dimension 300 × 100 mm were cut from the steel tube, reheated at 890 °C for 20 min and water quenched and consequently tempered at differing Tt of 650, 700 and 720 °C for 60 min.

Estimations of Mechanical Properties and SSC Susceptibility
The conventional tensile properties of each Q&T-treated sample were determined according to the ASTM (American Society for Testing Material) Standard A370-10 [49]. Each sample was machined along the longitudinal direction of the tube, with geometry and dimensions as shown in Figure 15 and tensioned in a CMT5305 Instron machine at room temperature. Two tensions for each sample were performed and their average values are reported. The hardness for each sample was also determined in a Rockwell hardness tester using a load of 1500 N. For precision, the hardness at 6 different positions in each sample were measured and their average values are reported. The susceptibility to SSC in each sample was estimated with the DCB test, according to the NACE TM0177 standard, Method D [31]. The standard DCB specimen is illustrated in Figure 16. A double-taper wedge with certain thickness was selected to provide the desired arm displacement and load by inserting into the DCB specimen. All DCB tests were performed with their specimens immersed in the same standard NACE TM0177 solution A, consisting of 5.0 wt % NaCl and 0.5 wt % CH3COOH dissolved in distilled water saturated with 1 atm H2S. The pH value of the solution did not exceed 3 at the start and 4 at the end of the test. The tests were performed for 2 weeks at ambient The susceptibility to SSC in each sample was estimated with the DCB test, according to the NACE TM0177 standard, Method D [31]. The standard DCB specimen is illustrated in Figure 16. A double-taper wedge with certain thickness was selected to provide the desired arm displacement and load by inserting into the DCB specimen. All DCB tests were performed with their specimens immersed in the same standard NACE TM0177 solution A, consisting of 5.0 wt % NaCl and 0.5 wt % CH 3 COOH dissolved in distilled water saturated with 1 atm H 2 S. The pH value of the solution did not exceed 3 at the start and 4 at the end of the test. The tests were performed for 2 weeks at ambient temperature, during which the instantaneous applied stress intensity factor, K Iapplied and the threshold stress intensity factor, K ISSC , were obtained using Equations (8) and (9): where P is the final equilibrium wedge lift off load, determined from an abrupt change in the slope of load-displacement curve; a i /a is the instant/final crack length of SSC for estimating K Iapplied /K ISSC , obtained by subtracting 6.35 mm from the distance between the slot end and the mean position of the crack front; h is the arm height; and B and B n are the DCB specimen and web thickness, respectively. Three parallel DCB tests were performed for each Q&T-treated sample and the average values are reported.  In order to further understand SSC crack initiation and propagation behaviors occurring in the DCB tests and their dependence on metallurgical aspects, the crack propagation length, Δa, as a function of immersing time, t, was measured by the NACE TM0177 standard, Method D [31]. Then the crack propagation rate, da/dt, was obtained by seeking derivation of the Δa vs. t curves. Finally, da/dt as a function of the corresponding KIapplied was described.

Electrochemical H Permeation Test
The H permeation behavior in each Q&T-treated sample was quantitatively estimated with the EHP test at 298 K using the classical Devanathan-Stachursky technique. The electrolyte for H charging was a solution of 0.5 mol/L H2SO4 and 0.25 g/L CH4N2S, while the electrolyte for H expanding was an aqueous solution of 0.2 mol/L NaOH. The H charging time might depend on the purpose. The sample for the H permeation estimation was 0.8 mm thick, with the nickel-plated surface acting as the anode side. The current was measured when the background current decreased and reached a steady value of 0.1 μm/cm 2 or lower.
The diffusion coefficient of H, D0 (cm 2 /s), was calculated using the following equation: where L is the specimen thickness (cm) and t0.63 is the time lag for 0.63 saturation of permeation current density at the steady state (s). The reversible H concentration, C0 (ppm), was estimated using the following equation: In order to further understand SSC crack initiation and propagation behaviors occurring in the DCB tests and their dependence on metallurgical aspects, the crack propagation length, ∆a, as a function of immersing time, t, was measured by the NACE TM0177 standard, Method D [31]. Then the crack propagation rate, da/dt, was obtained by seeking derivation of the ∆a vs. t curves. Finally, da/dt as a function of the corresponding K Iapplied was described.

Electrochemical H Permeation Test
The H permeation behavior in each Q&T-treated sample was quantitatively estimated with the EHP test at 298 K using the classical Devanathan-Stachursky technique. The electrolyte for H charging was a solution of 0.5 mol/L H 2 SO 4 and 0.25 g/L CH 4 N 2 S, while the electrolyte for H expanding was an aqueous solution of 0.2 mol/L NaOH. The H charging time might depend on the purpose. The sample for the H permeation estimation was 0.8 mm thick, with the nickel-plated surface acting as the anode side. The current was measured when the background current decreased and reached a steady value of 0.1 µm/cm 2 or lower.
The diffusion coefficient of H, D 0 (cm 2 /s), was calculated using the following equation: D 0 = L 2 6t 0.63 (10) where L is the specimen thickness (cm) and t 0.63 is the time lag for 0.63 saturation of permeation current density at the steady state (s). The reversible H concentration, C 0 (ppm), was estimated using the following equation: where i ∞ is the steady-state atomic H permeation current, µA/cm 2 and F is Faraday's constant, 9.6485 × 10 10 µA·s/mol.

Microstructure Observation
The as-quenched sample at 890 • C was etched in a supersaturated picric acid aqueous solution for optical microscopy (OM) observations of the prior austenite grain (PAG). Each Q&T-treated sample was etched in 4% nital for martensitic packet observation in a Hitachi S-3400N scanning electron microscope (SEM). The electron back-scattered diffraction (EBSD) map was obtained by a SUPRA550 field emission scanning electron microscope equipped with a TSL (Tinysoft Statistical analysis Language) orientation imaging microscopy system to reveal the martensitic block substructure and the misorientation angles between the adjacent martensitic substructure boundaries. SSC inducing the fractured surface and the secondary crack in a cross-section of the DCB specimen after each trial were also examined by SEM and EBSD, respectively. The crystallography orientation information of the martensitic packet/block structures next to the secondary crack and their correlations were determined; Channel 5 software from Oxford-HKL was used for the orientation data processing. Each EBSD sample was polished using 0.04 µm colloidal silica slurry in a vibratory polisher for 1 h. The characteristics of the martensitic lath and the distribution, shape and chemical compositions of precipitated particles were determined by examining thin foils or carbon extraction replicas using a JEM-2010 high-resolution transmission electron microscope.
The martensitic structure size for each Q&T-treated sample was quantitatively estimated by averaging at least 200 PAGs, 200 packets, 150 blocks and 100 laths from the OM, SEM, EBSD and TEM micrographs, respectively. For estimating the equivalent diameter (D o ) and volume fraction (f ) of precipitated particles of different shapes, the rod-like particles were considered with the spherical particles and D o was regarded as the particle size. Then f was calculated with the following equation: where N is the number of particles per area, S 0 is the specific area for estimation and S is the particle area. The dislocation density of each Q&T-treated sample was determined by quantitative X-ray diffraction (XRD) analysis. XRD spectra were obtained by scanning over a scanning angle (2θ) range of 40-110 • and at a step size of 0.02 • in a Rigaku D/max-2500/PC diffractometer (Rigaku, Tokyo, Japan). In order to correct the XRD line broadening resulted from the instrument, the XRD pattern of a standard sample of Si powder which was annealed at 300 • C for 24 h was measured. This diffraction line broadening correction was performed with the data processing via the software of Jade 5 (5.0, Jade Software, Christchurch, New Zealand).