A number of investigations employed SSRT to compare the HE characteristics of two different (lamellarized and equiaxed) morphologies of a medium-Mn steel microstructure, as summarized in Table 1. In SSRT, relative elongation loss due to H (HE index) is commonly used as a metric to measure a HE sensitivity of a material, and the HE index is given by the following equation: where TE_H-free and TE_H-charged are the total elongations obtained from SSRT of tensile specimens in H-free and H-charged conditions, respectively. Interestingly, three separate investigations had similar interpretations with respect to the relative H-resistance of the two morphologies [35,36,37]. Cameron et al. [35] applied two different processing routes to a medium-Mn steel (Fe–0.01C–9Mn–3Ni–1.4Al, in wt pct) to generate different microstructures. That is, IA of (i) as-quenched and (ii) cold-rolled martensitic microstructures led to lamellarized martensite–austenite and equiaxed ferrite–austenite microstructures, respectively. The SSRT results indicated that both microstructures showed similar degrees of elongation loss (91–92 pct) after H charging, despite the observations that the equiaxed morphology absorbed a much higher H concentration for a given H-charging condition (15.6 wppm in equiaxed vs. 1.87 wppm in lamellarized) and had a higher ultimate tensile strength (1100 vs. 900 MPa). In this regard, Cameron et al. [35] interpreted that the equiaxed microstructure was more resistant to HE, compared to the lamellarized condition.

Han et al. [36] conducted a systematic comparison of the two morphologies, comparing specimens having similar austenite volume fractions and similar H concentrations prior to the tensile tests. Figure 3a,b shows transmission electron microscopy (TEM) micrographs of the lamellarized martensite austenite and equiaxed ferrite austenite microstructures, produced through IA of hot-rolled and cold-rolled specimens, respectively. Figure 3c indicates similar H concentrations of the electrochemically H-charged specimens from the H thermal desorption analysis (TDA) results for the two morphologies. The SSRT results of the H-charged specimens, Figure 3d, indicate that the equiaxed microstructure better resisted HE, which is consistent with the interpretation of Cameron et al. [35]; it should be noted that, again, the equiaxed condition had a higher strength. Shen et al. [37] compared a slightly different combination of microstructures, i.e., a mixture of lamellarized and equiaxed structures having coarse lamellae (10–20 μm in length) and grains (~1.5 μm) versus a mainly fine lamellarized microstructure (lamellae 1–5 μm in length). The study by Shen et al. [37] is also consistent with the studies by Cameron et al. [35] and Han et al. [36], in that the microstructure containing equiaxed grains exhibited a higher resistance to HE. It should be pointed out, however, that the difference in microstructural feature sizes is also considered a critical factor that affects the HE characteristics in addition to the morphological change in the study by Shen et al. [37]. Specifically, Shen et al. [37] reported that as compared to the “larger” mixed microstructure, the “finer” lamellarized microstructure had a larger density of H-trapping interfaces, which accelerated H uptake and, thus, promoted embrittlement particularly at a prolonged H-charging time (24 h). It is also important to note that this hypothesis alone does not clearly explain the microstructure-dependent HE susceptibility for shorter H-charging time conditions (2 and 8 h), which resulted in relatively comparable H concentrations between the different microstructures.

Jeong et al. [10] designed experiments to decouple a variety of secondary factors; in their study, lamellarized and equiaxed conditions had similar strength levels, austenite fractions, and austenite mechanical stabilities (with respect to mechanically-induced martensitic transformation). In addition, Jeong et al. [10] explored a relatively wide range of H concentrations (0.4 to 4.4 ppm). Figure 4a–d summarizes the influence of H concentration on SSRT results for the lamellarized and equiaxed microstructures, produced through different two-stage annealing cycles. Interestingly, these results [10] were clearly different from the three other above-referenced studies [35,36,37] in that the lamellarized condition was less H-sensitive for a similar H content; note in Figure 4c that the H-induced elongation loss is much lower for the lamellarized condition. Particularly for low H concentrations (0.4–1.6 ppm), the H-induced ductility loss of the lamellar structure was negligible (Figure 4c).

It is unclear why these results are not consistent with the other studies presented in Table 1, but it might suggest that there can be other factors that affect the severity of HE. For instance, Jeong et al. [10] concentrated on the influence of connectivity (or microstructural percolation) of the austenite on H transport to explain the greater H-resistance of the lamellarized microstructure (Figure 4a–c). Jeong et al. performed electrochemical H permeability measurements and showed that H diffusion was slower in the lamellarized microstructure compared to the equiaxed microstructure. They explained that an austenite grain in the lamellarized microstructure is more likely to be connected to neighboring austenite grains, which effectively obstructs H migration. They suggested that the inhibited percolation through austenite decelerates H accumulation near the crack tip and, thus, positively contributes to H-resistance. The ability of H to percolate through mixed ferrite–austenite microstructures has also been proposed by Fielding et al. [38] as a critical factor in HE-resistance.

On the other hand, Han et al. [36] focused on the fact that the fracture and cracking mechanisms are different in the two conditions. For the lamellarized microstructure, H-induced fracture occurred preferentially along “prior” austenite grain boundaries, whereas fracture in the equiaxed microstructure occurred either across the ferrite grains or along the granular retained austenite grain boundaries. Greater degrees of crack deflection were present in the equiaxed condition, which was interpreted to improve H-resistance, as compared to the lamellarized condition. A more recent study [39] performed three-dimensional (3D) atom probe tomography (APT) of the lamellarized condition, used by Han et al. [36]. The 3D-APT analyses revealed noticeable segregation of C, Mn, and P at prior austenite grain boundaries, which may explain the lamellarized microstructure being more prone to prior austenite grain boundary cracking. The unfavorable influence of the combination of H and segregated elements (e.g., Mn, Si, P, and S) on the severity of HE has been demonstrated in other publications [40,41]. Therefore, control of impurity levels and segregation can also be a factor determining which microstructures better resist HE.

According to the results summarized in Table 1, the effect of microstructural morphology on H absorption is usually not significant for microstructures with a similar amount of austenite (e.g., the studies by Han et al. [36] and Jeong et al. [10]). In contrast, only the study by Cameron et al. [35] was unique, given that the equiaxed morphology absorbed a higher H content, nearly an order of magnitude than that of the lamellarized microstructure, for the same charging condition. The larger H absorption may partly be explained by the higher fraction of austenite, where H is more soluble than in martensite (and ferrite), in the equiaxed microstructure.

Retained austenite is the key constituent controlling activation of the TRIP and/or TWIP effect, which have a pronounced influence on the strain hardening behavior, in medium-Mn steels. Some studies suggest that retained austenite is also beneficial for the resistance of medium-Mn steel to HE because of its high H-trapping capability or its positive contribution to overall ductility (toughness). As an example, Wang et al. [42] investigated the influence of austenite fraction and film thickness, associated with austenite grain size, on H absorption and HE characteristics of a medium-Mn steel (Fe–0.01C–9Mn–3Ni–1.4Al, in wt pct), as shown in Figure 5. Figure 5a shows an electron backscatter diffraction (EBSD) phase map of a fully lath martensitic microstructure in the as-quenched condition. Wang et al. [42] produced two additional microstructures by applying an austenite reversion treatment to the as-quenched martensite. When the steel specimen was austenite reversion-treated at 600 °C with a hold time of 1 h, the microstructure contained austenite with a volume fraction of 10 pct and an average thickness of 200 nm, as shown in the EBSD phase map in Figure 5b. As the hold time at 600 °C increased to 8 h, the austenite fraction increased to 35 vol pct and the average thickness of the austenite films increased to 500 nm (Figure 5c). The H-TDA results by Wang et al. [42] (Figure 5d) indicate that the specimen containing a greater amount of retained austenite absorbs more H for a given H-charging condition because H solubility in austenite is much higher than in ferrite or martensite. The engineering stress–strain curves obtained by SSRT (Figure 5e) show that the H-induced ductility loss became less significant with increasing austenite fraction. Similar to the study by Wang et al. [42], Du et al. [43] produced three different microstructures from a medium-Mn steel (Fe–0.065C–5.45Mn–0.2Si, in wt pct): an as-quenched martensitic microstructure and two intercritically annealed microstructures containing different amounts of retained austenite within the martensitic matrix. Du et al. [43] also presented a beneficial effect of retained austenite with respect to the H-resistance in the medium-Mn steel. Figure 6 shows the SSRT results obtained by Du et al. [43], indicating that with increasing fraction of retained austenite, the H content of the specimen significantly increased, and the H-induced elongation loss became less significant.

There have been detrimental effects of deformation-induced transformation of retained austenite to martensite on HE reported. A study of Shao et al. [44] falls into this category, for example. Figure 7 shows retained austenite volume fraction for a medium-Mn steel (Fe–0.2C–5Mn–0.6Si–3Al, in wt pct) as a function of engineering tensile strain [44]. The data points at zero pct strain in Figure 7 indicate that there is little difference in the initial retained austenite volume fraction for the various IA hold times employed at 750 °C. However, the H-induced elongation losses in SSRT (HE index), presented in Figure 7b, became larger as the hold time increased. Shao et al. [44] explained that for the longer hold time conditions, the average austenite grain size was much larger, leading to a decrease of the mechanical stability during testing (Figure 7a). Shao et al. [44] also showed that austenite grain coarsening due to the longer hold time resulted in the absorption of slightly less H. The reduced uptake of hydrogen was attributed to a decreased area fraction of H-trapping sites associated with phase interfaces in the coarser microstructure.

Zhang et al. [45] investigated the influence of the IA temperature (700–850 °C) on HE of a medium-Mn steel (Fe–0.20C–0.6Si–4.9Mn–3.1Al, in wt pct); in their study, the HE susceptibility increased with increasing IA temperature, which was attributed to decreased austenite stability.

A detrimental effect of mechanically-induced martensite is also highlighted in a study by Wang et al. [46], who investigated the influence of pre-strain on HE of a medium-Mn steel (Fe–0.25C–8.7Mn–2.69Al–0.5Si, in wt pct). Wang et al. [46] reported that pre-straining resulted in an increase in dislocation density and partial transformation of austenite to martensite, leading to a substantial increase in the HE susceptibility. Liu et al. [47] also explored the pre-strain effect on H-induced cracking of medium-Mn steel, emphasizing the importance of austenite mechanical stability to suppress H-induced damage, similar to the study by Wang et al. [46].

It should be noted that, when the annealing conditions are varied, the steel strength, which is also a factor influencing the H-sensitivity, changes, in addition to the microstructure. For example, the studies by Wang et al. [43] and Du et al. [43] clearly showed that the HE susceptibility decreased with increasing retained austenite fraction (Figure 5 and Figure 6). However, in these studies, the specimen with a higher fraction of retained austenite exhibited a reduced yield strength, which may have contributed to the increased H-resistance. In contrast, the study by Shao et al. [44] clearly showed that, as IA time increased, HE became increasingly severe (Figure 7), despite the decreases in yield and tensile strength. Therefore, it was clear, in this study, that the H-resistance of medium-Mn steels was influenced by the mechanical stability of retained austenite.

A few investigations evaluated the HE characteristics of substantially Al- and Si-alloyed medium-Mn steels containing coarse δ-ferrite grains (Table 2). Figure 8a,b shows the microstructures of medium–Mn steels containing lower (1.1 wt pct) and higher Al contents (3.1 wt pct), respectively [48]. The microstructure of the low Al medium-Mn steel exhibited a relatively uniform distribution of UFG austenite and ferrite (Figure 8a), while the high-Al/medium-Mn steel contained coarse δ-ferrite layers in addition to the UFG austenite–ferrite regions (Figure 8b). Figure 8c,d shows the tensile properties of the low-Al and high-Al steels in H-free (air) and H-charged conditions, evaluated by means of SSRT. In the H-free condition, the low Al steel exhibited a high ultimate tensile strength and high work-hardening rate, associated with a pronounced TRIP effect (Figure 8c). Comparatively, the high-Al steel had an increased ductility and a low work-hardening capacity (Figure 8d). Ryu et al. [48] measured austenite fraction of each specimen as a function of the tensile strain and confirmed that the mechanical stability of austenite in the high-Al steel was higher compared to the low-Al steel. The SSRT results of H-charged specimens (Figure 8c, d) indicate that the H-induced loss in ductility was more pronounced in the low–Al/medium-Mn steel containing less stable austenite.

Considering the different H-sensitivity between the alloys, Ryu et al. [48] suggested that the detrimental effect of strain-induced martensite is because the transformation leaves the inherited H in a more mobile state. We interpret that the increase in strength due to the austenite to martensite transformation also increases vulnerability to HE. Interestingly, two other studies (by Wang et al. [49] and Shao et al. [44]) on the HE of high Al– and Si–alloyed medium-Mn steels reported observations similar to those of Ryu et al. [48]. That is, retaining a sufficient amount of austenite may benefit the HE resistance of medium–Mn steels, but suppression of the mechanically–induced martensitic transformation is often more critical; a similar finding was described in Section 2.2. The studies by Wang et al. and Shao et al. are also summarized in Table 2.

The beneficial effect of coarse δ–ferrite grains or a bimodal grain size distribution on the mechanical behavior in the H–free condition is well known [17,51]. That is, it has been suggested that there is enhanced strain partitioning in coarse, soft δ-ferrite grains, and thus, γ in the UFG region is stabilized, until larger strains are reached during deformation. This strain-partitioning behavior is believed to positively affect overall ductility and toughness. However, the direct role of δ–ferrite with respect to H-resistance is not clear in the studies summarized in Table 2. More often than not, the overall HE characteristics of a medium-Mn steel appear to be affected more significantly by the austenite fraction and mechanical stability rather than by the presence of δ–ferrite. However, there is some evidence that δ–ferrite layers might play a role in cracking or influence the fracture path in H testing. Wang et al. [49] reported that H-induced cracks are often arrested by the δ–ferrite layer. Warm rolling at an IA temperature is reported to provide a refined microstructure and significantly improved the H–resistance of a medium–Mn steel containing δ–ferrite. Zhang et al. [50] considered the refined δ–ferrite layers in a warm-rolled specimen to be beneficial, hypothesizing that the soft δ–ferrite exhibits low HE susceptibility, resulting in an effective crack termination mechanism. The contribution of the refined δ–ferrite layers in a warm-rolled specimen to deflection of the H–induced cracks is highlighted in Figure 9 [50]. Scanning electron microscopy (SEM) fractographs compare the path and morphology of the H-induced cracks in medium-Mn steels after conventional IA and after intercritical warm rolling. The crack in the intercritically annealed steel was slightly deflected at the δ-ferrite regions (Figure 9a). In the warm-rolled microstructure, the fine lamellar structure with δ–ferrite layers resulted in delamination-type fracture; the crack that encountered the δ–ferrite layer was significantly deflected and propagated in the rolling direction, rather than in the transverse direction (Figure 9b).

Wang et al. [49] suggested a different, but interesting view on the role of the δ–ferrite, i.e., when a large fraction of δ–ferrite exists in a medium-Mn steel microstructure, C and Mn are expected to be enriched in the non-δ-ferritic (mostly martensitic) regions in the as-hot rolled condition, i.e., before an austenite reversion treatment. Therefore, it can be assumed that, in a δ–ferrite-containing microstructure, higher amounts of austenite-stabilizing alloying elements, such as C and Mn, are partitioned to austenite during intercritical treatment, further enhancing the stabilization of austenite.

In addition to its role in δ–ferrite stabilization, Al additions significantly increase SFE, which is associated with mechanical stability of austenite. This dual role of Al may be important in the future design of a H-resistant ferrite–austenite microstructure with increased mechanical stability of austenite. There has been considerable discussion of the interaction between Al and H in “high” Mn TWIP steel [52,53,54]. In summary, Al has been found to slow the diffusion of H in a high-Mn TWIP steel, which leads to the suppression of H uptake and embrittlement [53]. It was also suggested that an Al₂O₃ oxide layer that forms at the surface of TWIP steel interferes with H absorption and, thus, prevents the H uptake [54].

Carbide and nitride precipitates containing molybdenum (Mo), vanadium (V), niobium (Nb), and titanium (Ti) have long been known to be effective in mitigating H-induced fracture of steels. However, the influence of micro alloy precipitates has not been broadly explored in medium-Mn steels. Park et al. [55] evaluated the HE resistance of medium-Mn steels micro alloyed with Nb, Ti, and V compared to a reference steel (Fe–6Mn–0.08C, in wt pct). They reported that the effects of the microalloying additions on the H-resistance were insignificant for the investigated test conditions. They explained that the insignificant effects of the precipitates were due to the two-phase nature of the microstructure, i.e., micro alloy carbides formed primarily in the ferritic regions, but the H atoms are expected to be trapped preferentially in the austenite, where H is more soluble than in ferrite. It is an important observation that, in a duplex microstructure, carbides and nitrides are mostly present in the ferrite, rather than the austenite, and may not actively interact with H atoms in the microstructure.

Cu is not a carbide/nitride former but can also precipitate in steel, contributing to precipitation strengthening. Li et al. [56] investigated precipitation behavior and HE in Cu–free (Fe–7Mn–2.5Ni–1.5Al–0.01C, in wt pct) and Cu-added medium-Mn steels (Fe–0.01C–7Mn-–2.5Ni–1.5Al–1.5Cu, in wt pct). In processing these two steels, Li et al. [56] used two-stage annealing; the first IA was performed at 630 °C for 1 h and the second step involved “tempering” at 500 °C. The hold times in the “tempering” step used for the Cu–free and Cu-added steels were 5 and 2 h, respectively. Figure 10a,b shows the 3D–APT results for the ferritic and austenitic regions in the Cu-added medium-Mn steel, indicating that Cu, as opposed to the microalloying example above [55], precipitated primarily in the austenite, rather than in the ferrite, despite the increased solubility of Cu in austenite compared to ferrite. In contrast, the APT map of the ferritic region (Figure 10a) shows non-uniformity in the Ni signal; TEM selected area diffraction analysis by Li et al. [56] revealed that the Ni-enriched regions in the ferrite are associated with B2 NiAl precipitates. Figure 10c shows the influence of Cu on the HE characteristics, evaluated by SSRT. The Cu-added steel exhibited a smaller elongation loss due to H, indicating a higher resistance to HE compared to the Cu-free steel. It should be noted that the interpretation of the primary role of the Cu precipitates with respect to the material’s H-resistance is different from the “conventional” mechanism from a viewpoint of precipitates being strong H-traps. The authors explained that the improved H–resistance in the Cu-added steel was due to increased strain compatibility between austenite and ferrite, resulting from the Cu precipitation. Li et al. [56] employed nano-indentation to support this view. Figure 10d shows the average indentation hardness of the ferrite and austenite phases in the Cu–free and Cu–added steels. In both the Cu-free and Cu-added steels, the ferrite was relatively hard (average nano-hardness > 8 GPa), possibly due to the NiAl associated precipitation strengthening of ferrite. In the Cu-free steel, the average hardness of ferrite (~10 GPa) is much harder than that of austenite (~5 GPa). In the Cu–added steel, however, Cu precipitation strengthening led to a significant increase in the average hardness of the austenite, which reduces the difference in hardness between the two constituent phases and, thus, improves the strain compatibility. The possible detrimental effect of strain incompatibility between neighboring constituent phases on the H–induced cracking of a medium–Mn steel was also discussed in a recent publication by Sun et al. [57]. It should be noted that, in the study by Li et al. [56], the improved H–resistance was noted despite the fact that the mechanical stability of the austenite was lower in the Cu-added steel. This study is one of only a few examples indicating that HE was less significant in the presence of a more pronounced TRIP effect.

Yoo et al. [58] reported that additions of 1–3 wt pct Cu improved the H–resistance of a duplex (ferrite austenite) lightweight steel (Fe–0.8C–15Mn–7Al, in wt pct). In their study, the Cu additions promoted the formation of Cu–rich B2 phase primary along the austenite grain boundaries, which was interpreted to trap H atoms and slow H transport during deformation. Cu as a solute is also known to increase SFE of austenite; for example, in a study by Choi et al. [59], the addition of 1–2 wt pct Cu resulted in a transition from a TRIP to a TWIP deformation mechanism in a high Mn austenitic–based steel. Yoo et al. [58] also emphasized the role of Cu in increasing the SFE of austenite and promoting wavy glide of dislocations as low SFE and planar slip are generally thought to be detrimental to H-resistance in austenite. It should be pointed out that the alloy investigated in the study by Yoo et al. [58] was relatively highly alloyed, which resulted in a high SFE of the austenite that limited α′- or ε-martensitic transformation, in the steel microstructure designed to have a mixture of ferrite and austenite. Finally, Cu additions (1–2 wt pct) have been reported to enhance the HE resistance of a fully austenitic, high Mn TWIP steel (Fe–17Mn–0.8C, in wt pct) [60]. In this case, where Cu is not likely precipitated, it is more likely that the improvement in the H-resistance was associated with the increased SFE.

Table 3 summarizes several studies related to processing- and microstructure-dependent fracture surface appearance and crack initiation sites of H–embrittled, medium–Mn steels. In general, different phase boundaries and grain boundaries in medium-Mn steel microstructures may serve as crack initiation sites; microstructural morphologies and constituents strongly influenced the H–induced cracking and fracture of medium–Mn steels. A few studies [36,44] observed H–induced cracking at/near mechanically–induced martensite. The studies by Han et al. [36], Wang et al. [49], and Zhang et al. [50] indicated that the lamellarized austenite–martensite microstructure is often subject to H–induced cracking along the prior austenite grain boundaries. However, this type of fracture is not observed in an equiaxed ferrite–austenite microstructure which does not preserve its prior austenite grain structure. Instead, according to studies by Han et al. [36] and Shao et al. [44], the fracture surfaces of the equiaxed microstructure often display dimples filled with fine granular features. Examples of these features are indicated by yellow arrows in Figure 11a, the appearance of which is different from “typical” ductile micro voids indicated by white arrows in Figure 11a. A detailed image of a fine granular feature is shown in Figure 11b. Han et al. [36] and Shao et al. [44] interpreted the fine granular features to be associated with a region of either equiaxed austenite or mechanically–induced martensite. In particular, Han et al. [36] attributed the granular features to intergranular cracking along equiaxed austenite grain boundaries. Shao et al. [44] showed that the Mn content of the granular features was higher than that of surrounding microstructure implying they are associated with austenite (or mechanically–induced martensite) since austenite tends to be more enriched in Mn than ferrite or martensite in these steels.

When δ–ferrite is retained in a medium–Mn steel microstructure, H–induced cracks form preferentially in UFG regions rather than in the δ–ferrite layers [49,50]; examples are shown in Figure 12a–c. The interface between the δ-ferrite layer and UFG region is also a preferential cracking site (Figure 12c). Wang et al. [49] interpreted the δ-ferrite to be beneficial in terms of resisting H–induced cracks in a medium–Mn steel microstructure because the coarse δ–ferrite is the softest phase in the microstructure and softer phases are often less sensitive to HE. While the observations of Wang et al. [49] are important from a microstructural engineering perspective, they also highlight complexity in understanding and predicting the HE behavior of a multiphase steel. Figure 12 compares H–induced cracks that formed in a medium–Mn steel [49] and a 2205 duplex stainless steel [61]. In contrast to the medium–Mn multiphase steel (Figure 12a), cracks in the 2205 duplex stainless steel initiated in ferrite and arrested in austenite (Figure 12b); ferrite is softer than austenite in 2205 duplex stainless steels according to literature [62]. It is possible that “magnitude” of hardness of ferrite, relative to other phases, is a factor determining whether cracks are impeded in ferrite in these microstructures. Another factor contributing to the difference in cracking behavior may be the UFG microstructure in the medium–Mn steel. That is, the UFG microstructure contains an increased area fraction of grain boundaries and phase interfaces, which are potential H-trapping sites and crack initiation sites. It should also be noted that a bimodal grain size distribution is expected to increase strain incompatibility, and a few studies [56,57] have emphasized the possible detrimental effect of strain incompatibility in a microstructure with respect to the material’s H–resistance. This detrimental effect may be more significant in the medium–Mn steel microstructure containing UFG austenite and martensite, in addition to coarse ferrite grains, as compared to the duplex stainless-steel microstructure. Considering the complicated nature of crack initiation in multiphase, fine–grained microstructures, there are still remaining questions regarding the role of δ–ferrite in HE of medium–Mn steels.