Micro-Mechanical Characterisation of Hydrogen Embrittlement and Fatigue Crack Growth Behaviours in Metastable Austenitic Stainless Steels with Microstructure Refinement

Yoji Mine

doi:10.2320/matertrans.MT-MF2022041

Abstract

This article reviews the microstructural evolution in ultrafine-grained and nanotwinned austenitic stainless steels that have been subjected to hydrogen embrittlement (HE) and fatigue cracking. It provides guidelines for the development of high-strength austenitic steels without sacrificing HE and fatigue performance. The author focuses on the hydrogen-induced ductility loss and short fatigue crack growth associated with deformation-induced martensitic transformation, using micro-tension and micro-fatigue testing technologies. In type 304 metastable austenitic stainless steel, the microstructure produced by high-pressure torsion depends strongly on the processing temperature. Nanocrystalline austenite with enhanced strength and moderate ductility can be obtained at a processing temperature of ∼423–573 K, whereas dual-phase microstructures comprising austenite and martensite are formed by processing at room temperature. Introducing ultrafine grains and nanotwin bundles mitigates the hydrogen-induced ductility loss in metastable austenitic steel by controlling the dynamic martensitic transformation. The microstructure refinement also contributes to enhanced resistance to short fatigue crack growth by changing the route of the damage accumulation process via phase transformation and detwinning.

1. Introduction

Weight reduction in structures such as automobiles has become an urgent issue owing to environmental pollution and energy shortages; consequently, structural materials need to be strengthened. Austenitic stainless steels are the choice for use in corrosive environments in the chemical, biomedical, automotive, and energy industries. The major drawback of austenitic stainless steels is their low yield stress (YS).¹⁾ Unfortunately, strengthening of austenitic steels through dispersed precipitation and grain refinement not only reduces the ductility but also increases the susceptibility to hydrogen embrittlement (HE).²^–⁴⁾ By contrast, in face-centred cubic high-entropy alloys, ultra grain refinement and nanotwinning increased the strength and reduced the HE susceptibility.⁵^–⁷⁾ For austenitic stainless steels, a bimodal microstructure comprising of fine- and coarse-grained (CG) domains provides good balance between strength and ductility.⁸^,⁹⁾ Further, Mine et al. reported¹⁰^,¹¹⁾ that an ultrafine-grained (UFG) type 304 stainless steel, which is a mixture of austenite and martensite grains, exhibited enhanced strength and moderate ductility with mitigation of hydrogen-induced ductility loss. Thus, the use of heterogeneous microstructures is a promising strategy for simultaneously improving conflicting mechanical properties.

The social implementation of structural materials places great importance on the fatigue performance, which is dominated by the damage accumulation process at the microscale. Therefore, for bimodal microstructures, it is important to understand the microstructural evolution at fatigue crack tips in each microconstituent. Similarly, local microstructural evolution during loading under hydrogen environments is a crucial factor for HE.¹²^,¹³⁾ In particular, for austenitic stainless steels, because HE susceptibility depends significantly on their austenite stability,¹⁴⁾ control of deformation-induced martensite formed on the microscale is the key to finding solutions to mitigate HE. Recent advances in micro-mechanical testing technologies have enabled us to examine the elementary process of deformation and microstructural evolution during several loading modes on a microscale.¹⁵^–¹⁷⁾ The author and co-workers established micro-tension¹⁸^–²⁰⁾ and micro-fatigue²¹^,²²⁾ testing as well as indentation testing for micro-compression,²³⁾ micro-cantilever bending²⁴⁾ and micro-shearing.²⁰^,²⁵⁾

The author focused on the fact that the primary deformation mechanism varies with the stacking fault energy (SFE) of austenitic steels,²⁶⁾ which results in significantly different microstructures. Type 304 metastable austenitic stainless steel was processed using high-pressure torsion (HPT) and rolling at moderate temperatures to obtain ultrafine austenite grains and nanotwin bundles, respectively. HPT processing enables to impose extremely high strain in a quantitatively controlled manner.²⁷⁾ This article provides a novel perspective for the development of high-strength austenitic steels with high resistance to HE and short fatigue crack propagation based on the analyses of micro-tension and micro-fatigue testing.

2. Microstructure Refinement and Mechanical Properties of Austenitic Stainless Steels Processed Using HPT at Moderate Temperatures

A nanoindentation-based study by Misra et al. revealed²⁶⁾ that different deformation mechanisms were activated in austenitic steels with different SFE values. Strain-induced α′ martensitic transformation dominates in type 301LN steel which exhibits a low SFE (∼15 mJ m⁻²), whereas mechanical twinning prevails in a twinning-induced plasticity steel with a moderately high SFE (∼40 mJ m⁻²). For the type 316L steel with a high SFE (∼60 mJ m⁻²), dislocation gliding increased in addition to mechanical twinning.

Figure 1 displays the electron back-scatter diffraction (EBSD) maps of the microstructures of type 304, 316L, and 310S steels processed using HPT at room temperature.²⁸^,²⁹⁾ According to the Schramm equation,³⁰⁾ the SFE values are estimated to be 14.1, 63.3, and 91.4 mJ m⁻² for type 304, 316L, and 310S, respectively. Type 304 steel contained a high fraction of α′ martensite in which the grain refinement was not attained (Fig. 1(a)). For the type 316L steel, α′ martensite was locally observed and the austenite grains were refined to an applicable extent (Fig. 1(b)). By contrast, a nanocrystalline (NC) austenite microstructure with an average grain size of ∼0.09 µm was obtained for type 310S steel (Fig. 1(c)), where the grain size range <∼0.1 µm for NC; <∼1 µm for UFG.³¹⁾ Therefore, as pointed out by Misra et al.,²⁶⁾ the deformation process, depending on the SFE, determines the microstructure formed after HPT processing in austenitic steels. Notably, the SFE depends not only on the chemical composition of the steel but also on the temperature of the deformation process.³²^,³³⁾

Fig. 1

EBSD maps of microstructures in type 304, 316L, and 310S stainless steels processed using HPT at room temperature for one turn. IQ and IPF represent the image quality and inverse pole figure, respectively.

Scheriau et al. studied³⁴⁾ the relationship between the microstructure and Vickers hardness of 316L stainless steel processed using HPT at temperatures ranging from 77 to 993 K. In the steel with a relatively high SFE, grain refinement was facilitated by dislocation gliding and mechanical twinning at room temperature. As the processing temperature increased, the density of the twins decreased, resulting in an increase in the grain size. When the steel was processed at a cryogenic temperature, an NC austenite microstructure was formed with ε martensite instead of α′ martensite.

Figure 2 shows the EBSD maps and transmission electron microscopy (TEM) images representative of the microstructures of type 304 metastable austenitic stainless steel processed using HPT in the temperature range of 303–573 K.³⁵⁾ The SFE of the steel is determined as 15.2 mJ m⁻² using the Schramm equation.³⁰⁾ An HPT processing at 303 K induced α′ martensite (Fig. 2(a)), of which the fraction was somewhat low compared to that obtained for the type 304 steel exhibiting a SFE of 14.1 mJ m⁻² (Fig. 1(a)). This is attributed to the difference in the deformation-induced martensitic transformation temperatures. The α′ martensite fraction was decreased with increasing processing temperature (Figs. 2(a)–2(c)). When the steel was processed at 373 K, a lamellar microstructure was formed inside a distorted austenite grain with a negligibly low fraction of α′ martensite (Fig. 2(c)). The (111) pole figure (Fig. 2(d)) shows the formation of mechanical twins, where the two-fold (111) pole corresponds to the direction normal to the lamellar plane. The mechanical twinning can be a precursor for α′ martensite formation in a metastable austenitic stainless steel subjected to plastic deformation at room temperature.³⁶^,³⁷⁾ The α′ martensite phase almost disappeared after HPT processing at 373 K, although co-existing with nanotwins in austenite grains in the steel processed at 323 K.

Fig. 2

EBSD maps and TEM bright-field images with SAED patterns of microstructures in type 304 stainless steel processed using HPT in the temperature range of 303–573 K for one turn. PF, RD, and CD represent the pole figure, radial direction, and circumferential direction, respectively.

NC microstructures were obtained using HPT processing in the temperature range of 423–573 K (Figs. 2(e)–2(g)). The selected area electron diffraction (SAED) patterns indicate fine austenite grains with a very low fraction of α′ martensite phase. The austenite grains were elongated in the circumferential direction of the HPT discs with an aspect ratio of ∼2.1–2.3. The average grain sizes were determined as ∼0.12 µm at 423 K, ∼0.09 µm at 473 K, and ∼0.07 µm at 573 K. A close examination revealed bundles of nanotwins in some fine grains (Fig. 2(h)). Similar microstructures were observed in Cu–30% Zn alloy and type 316L stainless steel subjected to multidirectional forging.³⁸^,³⁹⁾ It was concluded that processing at a high strain rate and a low temperature induces mechanical twinning to the alloys with a decreased SFE, which facilitates grain refinement via intersecting of different twin variants.³⁸^,³⁹⁾ For the type 304 steel, which exhibits a low SFE compared to the type 316L steel, a deformation process at a moderate temperature higher than room temperature is required to prevent the α′ martensitic transformation and to activate the mechanical twinning.

Figure 3 shows the dependence of the Vickers hardness on the processing temperature and nominal stress–nominal strain responses of type 304 and 310S steel specimens processed using HPT at varying temperatures.³⁵⁾ The Vickers hardness was measured as ∼480–490 for both steels processed at 303 K, which was approximately three times higher than that of the solution-treated steels (Fig. 3(a)). For type 304, as the processing temperature increased, the Vickers hardness decreased owing to a reduction in the α′ martensite fraction while increasing again above 373 K. The highest Vickers hardness (∼540) was attained for steel processed at 573 K. The increase in Vickers hardness with increasing processing temperature was associated with grain refinement, similar to that of type 310S. The nominal stress–nominal strain curves of the type 304 steels processed at 303–573 K indicated an increase in the ultimate tensile stress (UTS) with increasing processing temperature, whereas the elongation-to-failure did not change significantly (Fig. 3(b)). The stress–strain behaviour exhibited a processing temperature-dependence different from the Vickers hardness behaviour. This is because the Vickers hardness strongly reflects the deformation resistance of the initial microstructure. Type 304 steel processed at 573 K exhibited excellent balance between the strength and ductility: σ_B = ∼1.7 GPa and ε_f = ∼25%.

Fig. 3

(a) Dependence of Vickers hardness on the processing temperature and (b) and (c) nominal stress–nominal strain curves obtained for type 304 and 310S stainless steels processed using HPT in the temperature range of 303–573 K for one turn.

Figure 3(c) shows the nominal stress–nominal strain curves of the type 310S stable austenitic stainless steels processed using HPT at 303 and 473 K, which contained NC microstructures with an average grain size of ∼0.09 and ∼0.06 µm, respectively.⁴⁰⁾ When comparing the 473 K-processed specimens (Figs. 3(b) and 3(c)), the YS and UTS were higher in the type 310S steel than in the type 304 steel. Both steels reached the UTS at a nominal strain of approximately 3%, indicating that the reduced uniform elongation values can be attributed to the NC microstructures. In type 304 steel, the nominal stress–nominal strain curve contains a sustained stress regime after the initial stress drop, which results in an enhanced elongation-to-failure compared with type 310S steel. This distinctive feature did not appear in the stress–strain curve obtained by the tensile testing of the same steel at 473 K (not shown). Therefore, the stress–strain response peculiar to the type 304 steel was presumably related to the deformation-induced α′ martensitic transformation in the UFG microstructure. This will be discussed in detail below.

Figure 4 shows the true stress–true strain curves and the corresponding EBSD phase maps of the necked regions in the mid-thickness section, which resulted from interrupted tensile testing of type 304 stainless steel with an NC microstructure processed using HPT at 473 K.⁴⁰⁾ The true stress was determined using the narrowest width of the necked region at that moment. The plateau regime to a true strain of ∼0.08 in the true stress–true strain curves corresponds to the initial stress drop in the nominal stress–nominal strain curves (Fig. 4(a)). Although martensite was almost invisible immediately after the onset of necking (Fig. 4(b)), at a true strain of ∼0.12, where the plateau regime was terminated (Fig. 4(a)), the α′ martensite fraction was increased up to ∼82% (Fig. 4(c)). In the strain hardening regime, (110) fibre texture of α′ martensite evolved (Fig. 4(d)). These findings suggest that in NC type 304 metastable austenitic steel, deformation-induced martensite transformation occurs during the plateau regime in the true stress–true strain curve, followed by strain hardening due to the deformation of the formed martensite in the necked region. This was responsible for the enhanced elongation-to-failure of the NC type 304 steel.

Fig. 4

(a) True stress–true strain curves and (b)–(d) corresponding EBSD phase maps of the necked region for interrupted tensile tests of NC type 304 stainless steel processed using HPT at 473 K. LD and TD represent the loading and transverse directions, respectively.

Figure 5 shows the reduction of area (RA) plotted against the UTS obtained for type 304 steel samples processed in the temperature range of 303–573 K and type 310S steel samples processed at 303 and 473 K.³⁵^,⁴⁰⁾ For comparison, the results for the solution-treated steels are included in the figure. A high RA of approximately 55% remained in the NC type 304 steel. The ductility loss due to grain refinement was lower in type 304 than in type 310S. Thus, it can be concluded that in type 304 steel, dynamic martensitic transformation during tensile loading results in enhanced balance between strength and ductility via grain refinement.

Fig. 5

Relationship between RA and UTS in solution-treated specimens and HPT-processed specimens in the temperature range of 303–573 K for type 304 steel and at 303 and 473 K for type 310S steel. ST represents the solution-treated specimen.

3. Micro-Tensile Characterisation of HE in Metastable Austenitic Stainless Steels with UFG and Nanotwinned Microstructures

3.1 Grain size dependence of HE susceptibility of UFG steel

For stable austenitic steels with ordinary grain sizes, hydrogen-induced ductility loss may be determined by the combination of strengthening and enhanced slip planarity by local hydrogen concentration.⁴¹⁾ In contrast, metastable austenitic stainless steels such as types 301 and 304 undergo severe hydrogen-assisted degradation in the mechanical properties,¹⁴^,⁴²⁾ which can be attributed to deformation-induced martensitic transformations despite different testing conditions.⁴³^–⁴⁷⁾ With regard to the effect of grain refinement on the HE, Macadre et al. reported that the hydrogen-induced ductility loss was mitigated by a fine-grained austenite microstructure with an average grain size of ∼1 µm in Fe–16Cr–10Ni (in mass%) relatively stable austenitic stainless steel.⁴⁸⁾ A micro-tension testing study by the author and co-workers⁴⁹⁾ was employed on type 304 metastable austenitic stainless steel produced through post-HPT annealing to elucidate the role of martensitic transformation in HE of a UFG microstructure. Although the UFG microstructure processed through post-HPT annealing has a wide distribution in the grain size, micro-tension testing can reduce the effect of grain size inhomogeneity because of its small specimen size, that is, a typical characteristic size of 20 µm.⁵⁰⁾

Figure 6 shows the nominal stress–nominal strain curves resulting from micro-tension tests in an uncharged state using UFG type 304 austenitic steel specimens annealed in the temperature range of 873–973 K after HPT processing at 473 K.⁴⁹⁾ The UTS increased with decreasing grain size and reached ∼1.6 GPa for an average grain size of ∼0.14 µm. In the specimens annealed in the temperature range of 873–943 K, necking occurred immediately after the onset of yielding, followed by the maintenance of local deformation at a constant flow stress, which resulted in a moderate elongation-to-failure of ∼35%. By contrast, the 973 K-annealed specimen with an average grain size of ∼0.46 µm exhibited a considerable uniform elongation after the onset of yielding at 770 MPa. A similar transition behaviour determined by the Considère criterion was observed at approximately 1 µm grain size in varied alloys.⁵¹^–⁵³⁾ When the grain size is considerably reduced, the grain boundary cannot act as an effective barrier to the accumulation of dislocations, resulting in reduction of the strain hardening rate.⁵⁴⁾ The specimens annealed at a temperature higher than ∼943 K contained a high fraction of twin boundaries and σ-FeCr precipitates.⁴⁹⁾ Introducing the twin boundaries and secondary phase precipitates to the UFG microstructure can actively contribute to enhancing the strain hardenability.⁵⁵^,⁵⁶⁾ Therefore, the presence of twin boundaries and σ-FeCr precipitates can lower the grain size for the transition of the tensile behaviour to ∼0.4 µm.

Fig. 6

Nominal stress–nominal strain curves obtained from micro-tension testing of UFG type 304 steel annealed in the temperature range of 873–973 K after HPT processing at 473 K in the uncharged state. d represents the average grain size.

Figure 7 shows the effect of hydrogen on the nominal stress–nominal strain response and the fracture morphology in the 973 K-annealed specimens with an average grain size of ∼0.5 µm.⁴⁹⁾ The hydrogen-charged specimen was cathodically charged with hydrogen in aqueous H₂SO₄ before micro-tension testing. The hydrogen charge increases the YS from ∼770 to ∼920 MPa and reduces the elongation-to-failure from ∼60 to ∼40% (Fig. 7(a)). Both the uncharged and hydrogen-charged specimens exhibited a similar uniform elongation of ∼16–17%, whereas hydrogen halved the local elongation. In CG type 304 steel, it was reported⁵⁷^,⁵⁸⁾ that the uniform and local elongations were significantly degraded owing to the premature exhaustion of strain hardenability. Thus, hydrogen-induced loss of uniform elongation was prevented in the 973 K-annealed type 304 steel with an average grain size of ∼0.5 µm. The RA in the hydrogen-charged specimen was determined as ∼57%, which is lower than that in the uncharged specimen (∼89%). This coincided well with the effect of hydrogen on the elongation-to-failure. Equiaxed and elongated dimples were observed in the uncharged specimen (Fig. 7(b)), whereas the fracture surface of the hydrogen-charged specimen was covered with fine dimples and flat facets with sizes similar to those of the average grain size (Fig. 7(c)). Flat facets have been reported to arise from twin boundary separation in metastable austenitic stainless steels with ordinary grain sizes.⁵⁹^–⁶²⁾ The flat facets presumably prevailed in the hydrogen-charged specimen of the 943 K-annealed steel because of its high fraction of twin boundaries.

Fig. 7

(a) Nominal stress–nominal strain curves and (b) and (c) fracture surfaces of the uncharged and hydrogen-charged specimens in type 304 stainless steel annealed at 973 K after HPT processing at 473 K.

Figure 8 shows the effect of hydrogen on the nominal stress–nominal strain response and the fracture morphology in the 943 K-annealed specimens with an average grain size of ∼0.3 µm, which was smaller than the grain size of the transition of the tensile behaviour.⁴⁹⁾ The uniform elongation was limited to a few percents independent of hydrogen charging (Fig. 8(a)) because the plastic instability conditions were readily fulfilled in an early stage of tensile deformation in such an UFG specimen. In contrast, the local elongation was reduced from 30 to 24% upon hydrogen charging. Unlike the specimens in the range greater than the transition grain size, α′ martensite was locally formed in the necked region in the specimens with smaller grain sizes. In the uncharged specimen, a slant fracture region extended at the periphery of the fracture surface, whereas in the fibrous region, micrometre- and grain-sized dimples were observed, which arose from inclusions and grain boundaries, respectively (Fig. 8(b)). The fibrous region, which was significantly rougher, prevailed on the fracture surface of the hydrogen-charged specimens (Fig. 8(c)). To clarify whether hydrogen affects the martensite formation or the nucleation and growth of voids, the author and co-workers investigated the fracture process of the hydrogen-removed specimen after α′ martensite was formed in the necked region in a hydrogen-precharged state (not shown).⁴⁴⁾ The result suggested that hydrogen facilitates the void nucleation and growth process rather than contributes to the martensitic transformation in an early stage.

Fig. 8

(a) Nominal stress–nominal strain curves and (b) and (c) fracture surfaces of the uncharged and hydrogen-charged specimens in type 304 stainless steel annealed at 943 K after HPT processing at 473 K.

Figure 9 shows the effects of hydrogen on the mechanical properties of type 304 austenitic stainless steels with different average grain sizes. The Hall–Petch relationships hold well between the 0.2% proof stress and grain size in the uncharged and hydrogen-precharged states (Fig. 9(a)). The friction stress (σ₀) and Hall–Petch coefficient (k_y) of the uncharged type 304 steel was determined as 176 MPa and 450 MPa µm^1/2, respectively. Hydrogen increased the friction stress by 40% to 245 MPa, whereas the Hall–Petch coefficient remained almost unchanged. This indicates that hydrogen strengthens the matrix via a solid solution effect but does not significantly impact grain refinement strengthening. The relationship between the RA and UTS in the uncharged state reveals that the RA was reduced with increasing UTS in the range greater than the transition grain size, i.e. ∼0.4 µm, whereas it was invariable in steels with smaller grain sizes (Fig. 9(b)). The hydrogen-induced loss of RA was remarkably reduced in the UFG specimens. In other words, grain refinement not only enhanced the balance between strength and ductility but also mitigated HE in metastable austenitic stainless steel. In particular, in a range smaller than the transition grain size, the RA in the hydrogen-charged specimens was maintained at ∼60% while the UTS was 1–1.5 GPa.

Fig. 9

(a) Grain size dependence of YS and (b) relationship between RA and UTS in the uncharged and hydrogen-charged specimens of solution-treated and post-HPT-annealed type 304 steels. R² denotes the correlation coefficient for least-square fitting method. The data for the NT type 304 steel are also included in (b).

It has been argued⁶³^–⁶⁵⁾ that dynamic α′ martensite formation is a crucial factor determining the HE susceptibility of metastable austenitic stainless steels. The saturated hydrogen content of martensite was lower than that of austenite by one order of magnitude.⁶⁶⁾ When hydrogen-bearing austenite is transformed to α′ martensite, excess hydrogen can be generated, which corresponds to the difference in the saturated hydrogen content between the two phases. Subsequently, excess hydrogen diffuses out of the α′ martensite phase to the surrounding austenite phase because of its extremely higher diffusivity in the former than in the latter. A finite element method calculation by Wang et al. supported the idea that cracking occurs in the austenite phase neighbouring the formed martensite variants.⁶⁷⁾ Therefore, it is rationalised that excess hydrogen accumulates at the interphase boundary, facilitating cracking. Several variants of α′ martensite were selected in austenite with ordinary grain sizes, whereas UFG austenite was transformed to a single variant from each grain.⁶⁸⁾ In the UFG austenite, α′ martensite variants were separately formed during tensile loading. Thus, excess hydrogen was dispersed, which resulted in a reduced local hydrogen concentration and, in turn, mitigated the hydrogen-induced ductility loss.

3.2 Twin orientation dependence of HE susceptibility of nanotwinned steel

Twin boundaries have garnered considerable attention because of their dual roles in dislocation motion, which dominates the mechanical properties. Twin boundaries contribute not only to the strengthening by hindering the dislocation motion but also to strain accommodation via dislocation gliding parallel to the twin plane.⁶⁹^–⁷¹⁾ For face-centred cubic metals, several researchers reported⁷²^–⁷⁶⁾ that the introduction of twin lamellae with an average spacing on the nanometre scale increases the strength while maintaining moderate ductility. Lu et al. successfully obtained a fine balance between YS and ductility by introducing bundles of nanotwins and recrystallised grains into type 316L stable austenite stainless steel.⁷³^,⁷⁴⁾ It was reported that type 304 metastable austenitic stainless steels with similar bimodal microstructures exhibit an enhanced resistance to HE, which is attributed to the inhibition of strain localisation.⁷⁵^,⁷⁶⁾ Thus, the introduction of nanotwin bundles into metastable austenitic steel is a promising strategy for strengthening and reducing the susceptibility to HE.

Figure 10 displays the EBSD map and TEM image of the microstructure obtained by rolling of the 673 K-preheated type 304 steel plate to a reduction ratio of ∼40% in thickness.⁷⁷⁾ Mechanical twins were observed without α′ martensite phase (Fig. 10(a)). TEM dark-field image and SAED pattern reveal nanotwins and a high density of dislocations (Fig. 10(b)). The average lamellar spacing of the nanotwins was determined to be ∼0.06 µm. Figure 11 shows the effects of the twin plane orientation and hydrogen on the true stress–true strain response obtained for micro-tension testing using single-crystalline, twinned bi-crystalline, and nanotwinned specimens (hereinafter referred to as SC, BC, and NT, respectively) of type 304 metastable austenitic stainless steel.⁶²^,⁶⁴^,⁷⁷⁾ The specimens in which the twin plane is normal and parallel to the loading axis are denoted as N and P, respectively. In the I orientation, the twin plane was inclined at an angle of approximately 45° with respect to the loading axis. Suffixes ‘-U’ and ‘-H’ represent the uncharged and hydrogen-charged specimens, respectively. The true stress (σ_T) and true strain (ε_T) were calculated from the nominal stress (σ) and nominal strain (ε) using σ_T = σ(1 + ε) and ε_T = ln(1 + ε). However, the true stress level is not valid after the onset of necking. In the uncharged state, YS was determined to be ∼760 MPa for NTP, ∼650 MPa for NTN, and ∼540 MPa for NTI. The YS values of the NT specimens were 2.3–3.7 times higher than those of the SC and BC specimens.

Fig. 10

(a) EBSD phase map, (b) TEM dark-field image, and SAED pattern of nanotwin bundles within an austenite grain in type 304 stainless steel processed by rolling to reduction ration of ∼40% in thickness immediately after preheating at 673 K.

Fig. 11

True stress–true strain curves obtained for micro-tension testing of SC, BC, and NT specimens of type 304 stainless steel. In the I orientation (a), the twin plane is inclined at 45° with respect to the loading axis. In the P and N orientations (b) and (c), the twin plane was parallel and normal to the loading axis, respectively. Suffixes ‘-U’ and ‘-H’ represent the uncharged and hydrogen-charged specimens, respectively.

Similar to the SC specimens with a [123] loading axis, the NTI-U specimen exhibited three-stage strain hardening (Fig. 11(a)). In the NTI-H specimen, primary slip and/or mechanical twinning parallel to the twin plane were activated after the onset of yielding, resulting in moderate uniform elongation via crystal rotation. Figure 12 shows the fracture surfaces of the hydrogen-charged SC, BC, and NT specimens.⁶²^,⁶⁴^,⁷⁷⁾ All the uncharged specimens exhibited chisel edge type fracture (not shown).⁶⁴^,⁷⁷⁾ Although the fracture surface of the SC-H specimen was covered with quasi-cleavages (Fig. 12(a)), the NTI-H specimen exhibited shear fracture in addition to quasi-cleavages (Fig. 12(b)). The RA was ∼70% for NTI-H and ∼45% for SC-H, indicating a moderate mitigation of the hydrogen-induced ductility loss in the NTI specimens. Figure 13 shows the EBSD maps obtained after failure of the uncharged and hydrogen-charged NTI specimens. In the NTI-U specimen, α′ martensite was uniformly formed over the whole deformed region (Fig. 13(a)), whereas in the NTI-H specimen, was localised in the vicinity of the fracture surface (Fig. 13(b)). Furthermore, the twin-oriented regions extended over the deformed part of NTI-H, indicating that mechanical twinning occurred parallel to the nanotwin lamellae (Fig. 13(c)). Similar mechanical twinning behaviour was observed in an Fe–19Cr–16Ni austenitic alloy⁷⁸⁾ and a TiAl-based alloy⁷⁹⁾ with nanoscale lamellar microstructures in uncharged states. Thus, similar to the CG microstructures,²⁸⁾ hydrogen inhibits the α′ martensitic transformation in the soft orientation, where the dislocations can glide parallel to the nanotwin lamellar plane, facilitating mechanical twinning. In addition, in final fracture, α′ martensite is locally formed, which results in quasi-cleavages. Nevertheless, the hydrogen-induced ductility loss is mitigated in the NTI specimens owing to retarding the α′ martensite formation and quasi-cleavages.

Fig. 12

Fracture surfaces of the hydrogen-charged SC, BC, and NT specimens of type 304 austenitic stainless steel.

Fig. 13

EBSD maps of deformed microstructures after failure in micro-tension testing of (a) NTI-U and (b) and (c) NTI-H.

The BCP specimens exhibited a three-stage strain hardening, whereas hydrogen reduced the elongation-to-failure (Fig. 11(b)). The activation of the primary slip systems dominated in both the parent and twinned crystals, which induced α′ martensite variants with their habit planes parallel to the activated slips, finally resulting in quasi-cleavages (Fig. 12(c)). The YS of NTP-U was 3.7 times higher than that of BCP-U (Fig. 11(b)). For the NTP specimens, the plateau regime disappeared in the stress–strain curves. In the NTP-H specimen, an enhanced RA resulted from longitudinal cracking due to the twin boundary separation induced by hydrogen (Fig. 12(d)).

In both the BCN and NTN specimens, hydrogen increased the flow stress but reduced the elongation-to-failure (Fig. 11(c)). The BCN-H specimen exhibited hydrogen-induced twin boundary separation (Fig. 12(e)), which significantly degraded its strength and ductility compared to the BCN-U specimen (Fig. 11(c)). In the NTN-H specimen, hydrogen increased the UTS, despite the reduced ductility. Because the twin plane is oriented perpendicular to the loading axis, the resolved shear stress for the twin plane owing to the external stress should be negligible in the N orientation. Ueki et al. reported that a deformation step parallel to the nanotwin lamellar plane appeared at the onset of yielding.⁷⁷⁾ Rèmy interpreted⁸⁰⁾ that shear deformation parallel to the twin plane could be generated by the interaction of dislocations with twin boundaries. A flat-faceted feature was observed with three-directional linear steps at an intersection angle of 60° on the BCN-H fracture surface (Fig. 12(e)). For NTN-H, terraced facets were formed with one-directional steps, which corresponded to the intersection between the twin plane and primary slip planes in the parent and twinned crystals (Fig. 12(f)). Therefore, the introduction of nanotwin boundaries restricts the number of slip systems, which participate in the hydrogen-induced twin boundary separation. Further, a thermal desorption analysis study of Fe–18Mn–1.5Al–0.6C austenitic steel revealed⁸¹⁾ that the activation energies for hydrogen desorption from dislocations and mechanical twin boundaries were estimated as ∼35 and ∼62 kJ mol⁻¹, respectively. This suggests that twin boundaries can be an effective trap site for hydrogen. Therefore, the presence of nanotwin boundaries can disperse the local hydrogen concentration due to dynamic α′ martensitic transformation. As shown in Fig. 9(b), the relationship between RA and UTS reveals that the NT microstructures as well as the UFG microstructures can contribute to strengthening while mitigating the hydrogen-induced ductility loss in metastable austenitic stainless steel.

4. Enhanced Resistance to Fatigue Crack Growth in Austenitic Stainless Steels through Grain Refinement and Introducing Nanotwin Bundles

For metastable austenitic stainless steels, such as type 304, the crack closure effect due to transformation-induced plasticity enhances the extrinsic resistance to fatigue crack growth, thereby decreasing the growth rate of long cracks.⁸²⁾ The understanding of intrinsic resistance to short crack growth is significant for the development of high-strength steels with high fatigue performance; however, this has not yet been revealed. As mentioned above, metastable austenite is locally transformed to martensite ahead of the crack tip.⁸³^,⁸⁴⁾ In addition, Furukane and Torizuka concluded that austenite stability against deformation-induced martensitic transformation was increased through grain refinement.⁸⁵⁾ Therefore, phase transformation is anticipated to change the route of the damage accumulation process that dominates the fatigue crack propagation mechanism. Thus, in metastable austenitic stainless steel, grain refinement can significantly influence the growth rate of short cracks sensitive to the microstructure.⁸⁶^,⁸⁷⁾

The author and co-workers established micro-fatigue testing technology using miniature compact-tension (CT) specimens designed for detailed examination of the elementary process of microstructure-sensitive crack propagation.²¹^,²²^,⁸⁸⁾ Figure 14 shows the setup of a micro-fatigue testing instrument and a miniature CT specimen with typical dimensions (0.05 mm thick and 1 mm wide), which was scaled down to a few percent of a conventional CT specimen. The crack in the miniature CT specimen is presumed to be a mechanically short crack⁸⁹⁾ with a reduced crack closure effect.

Fig. 14

(a) Setup of micro-fatigue testing instrument and (b) miniature CT specimen 0.05 mm thick and 1 mm wide.

Figure 15 shows the EBSD maps of UFG microstructures of type 310S and 304 stainless steels processed by post-HPT annealing.⁹⁰⁾ As shown in Figs. 15(a)–(c), the average grain size of UFG austenite was measured as 0.25 and 0.58 µm for type 310S steel (denoted as 310AUS25 and 310AUS58, respectively) and as 0.81 µm for type 304 steel (denoted as 304AUS81). For comparison, in type 304, a UFG martensite microstructure with an average grain size of 0.25 µm (denoted as 304MAR25) was obtained through two-step HPT processing (Fig. 15(d)). A UFG austenite microstructure was obtained using HPT processing at 573 K, followed by annealing at 923 K, which was subsequently processed by HPT at room temperature, resulting in a UFG martensite microstructure with a fraction of 92.4% (Fig. 15(e)). Figure 16 shows the relationship between the fatigue crack growth rate (da/dN) and stress intensity factor range (ΔK) for the UFG specimens.⁹⁰⁾ Overall, the fatigue crack growth rates of the UFG specimens were less than those for CG specimens (not shown).⁹⁰⁾ In 310AUS25 and 310AUS58, the crack growth rate fluctuated in the low ΔK regime and as the ΔK increased, the fatigue crack growth resistance curves, which coincided with each other, exhibited a stable crack growth. The 304AUS81 specimen exhibited high resistance to fatigue crack growth compared with the 310AUS25 and 310AUS58 specimens, although the grain sizes differed from one another. The growth rate for the short crack in 304AUS81 was slightly higher than the extrapolated value of the plots for the long crack in 304AUS51. The long-crack data were obtained using a CT specimen 0.4 mm thick and 10 mm wide.

Fig. 15

EBSD maps of UFG microstructures of type 310S and 304 stainless steels processed through post-HPT annealing. AUS and MAR represent the austenite and martensite phases, respectively.

Fig. 16

Fatigue crack growth resistance curves for UFG specimens of type 310S and 304 stainless steels.

Figure 17 shows the EBSD grain reference orientation deviation (GROD) map and scanning electron microscopy backscattered electron (SEM BSE) images of the deformed microstructure around a crack path in 310AUS58. In the CG specimens, the strain intensity decreased continuously with increasing distance from the crack surface (data not shown). For 310AUS58, the deformed grains are discretely distributed (Fig. 17(a)), whereas some undeformed grains are adjacent to the crack path (Fig. 17(b)). This indicates that both grains favourably and unfavourably oriented for dislocation gliding owing to the external stress were mixed around the crack path in the UFG specimens. A similar microstructural evolution was observed around the short fatigue crack growing in UFG interstitial-free steel processed through accumulative roll bonding.⁹¹⁾ The elastic stress field at the crack tip not only injects dislocations out of the crack tip but also multiplies dislocations at the grain boundaries.⁹²⁾ Furthermore, crack branching was repeatedly observed in the crack path of the UFG specimens (Fig. 17(b)). This indicates that the grains unfavourable for dislocation gliding inhibit continuous crack growth and facilitate strain localisation into favourably oriented grains, resulting in crack branching and resultant plastic dissipation in fatigue crack propagation in the UFG microstructure.

Fig. 17

(a) EBSD GROD map and (b) SEM BSE image of the microstructure around the fatigue crack profile in 310AUS58.

Figure 18 shows the EBSD phase map, GROD map, and SEM BSE image of the microstructure around the crack path in the 304AUS81 specimen. Although α′ martensite was formed in the narrow region adjacent to the crack path, some variants were isolated away form the crack path (shown by arrowheads in Fig. 18(a)). In addition, the degrees of strain accumulation and plastic deformation varied for each grain (Figs. 18(b) and 18(c)). Most α′ martensite grains neighbouring the crack path exhibited the same crystallographic orientation on both sides of the crack (Fig. 18(a)). This indicates that the crack grew inside the α′ martensite variants formed ahead of the crack tip. For CG type 304 steel specimens, the crack extends in large α′ martensite variants with almost the same orientation, because the resolved shear stress of the external stress determines the variant selection of α′ martensite. In contrast, in the UFG type 304 steel, each austenite grain ahead of the crack front was transformed to a single variant of α′ martensite, resulting in crack growth in α′ martensite grains with different crystallographic orientations. Considering that the crack growth rate in 304MAR25 almost comprising of α′ martensite grains was similar to that in 310AUS25 (Fig. 16), it can be concluded that the grain refinement contributes to the enhanced resistance to short fatigue crack growth by changing the route of damage accumulation process via the single-variant martensitic transformation.

Fig. 18

(a) EBSD phase map, (b) GROD map, and (c) SEM BSE image of the microstructure around the fatigue crack profile in 304AUS81.

Figure 19 shows the fatigue crack growth resistance curves of the CG and NT specimens of type 304 stainless steel.⁹³⁾ Overall, the fatigue crack growth rates of the NTP and NTN specimens, which had notches parallel and normal to the twin plane, respectively, were lower than those of the CG specimen with a (111) notch. The fatigue crack growth rate was not dependent on the twin plane orientation, while fluctuating in the low ΔK regime. Figure 20 shows a TEM bright-field image and schematic of the deformed microstructure extracted from the crack-tip region in the NTN. The crack propagated in the nanotwin bundle and α′ martensite variants were formed ahead of the crack tip. The austenite region with the parent crystal orientation was spread adjacent to the formed martensite region, which contained partially dissociated nanotwin lamellae and a high density of dislocations. An analysis using different diffraction vectors revealed partial dislocations with a Burgers vector of (a/6)$[\bar{1}21]$ on the $(11\bar{1})$ plane.⁹³⁾ When partial dislocations glide towards the twin boundary, they dissociate into (a/6)$[\bar{2}11]$ partial dislocations that can glide on the (111) twin plane and (a/6)[110] sessile stair-rod dislocations at the intersection of the $(11\bar{1})$ and (111) planes, as expressed below:

\begin{equation} \frac{a}{6}[\bar{1}21]\to \frac{a}{6} [\bar{2}11] + \frac{a}{6} [110]. \end{equation}

(1)

As proposed by Zhu et al.,⁹⁴⁾ if the stair-rod dislocations redissociate into partial dislocations that can glide in the opposite direction on the (111) twin plane and initial partial dislocations on the $(11\bar{1})$ slip plane,

\begin{equation} \frac{a}{6}[110]\to \frac{a}{6} [2\bar{1}\bar{1}] + \frac{a}{6} [\bar{1}21], \end{equation}

(2)

detwinning can be caused by movement of the twin boundary towards the twin crystal. In addition, the EBSD analysis⁹³⁾ revealed that the α′ martensite variants were formed with their habit plane parallel to the twin plane: (111)_P/T//(011)_M and $[0\bar{1}1]_{\text{P/T}}{\mathrel{/\!/}}[\bar{1}1\bar{1}]_{\text{M}}$, where subscripts P, T, and M represent austenite parent crystal, twin crystal and martensite crystal, respectively. Considering the double shearing mechanism⁹⁵⁾ for phase transformation, the α′ martensite variant can be induced by the $[\bar{1}\bar{1}2]$ first shear on the (111) twin plane. Chen et al.⁹⁶⁾ performed tensile testing of NT 304 steel in a TEM environmental cell and reported that detwinning was followed by the nucleation and growth of martensite variants. Therefore, dislocations are deduced to accumulate at sessile dislocations left at the prior twin boundaries, which induced α′ martensite variants with their habit plane parallel to the twin plane. The fatigue crack propagates in the α′ martensite formed ahead of the crack tip by interaction of dislocations with twin boundaries. Thus, the retardation of damage accumulation through microstructural evolution ahead of the crack tip results in an enhanced resistance to fatigue crack growth in type 304 metastable austenitic stainless steel.

Fig. 19

Fatigue crack growth resistance curves for CG and NT specimens of type 304 stainless steels.

Fig. 20

TEM bright-field image and schematic of the deformed microstructure in the NTN specimen.

5. Concluding Remarks

This article addresses coupled metallographic examination with micro-mechanical testing of metastable austenitic stainless steels with focus on the effect of deformation-induced martensitic transformation on the HE and fatigue performances. The conclusions can be summarised as follows:

(1) In type 304 metastable austenitic stainless steel, the microstructure evolved during HPT processing depended significantly on the processing temperature. An NC austenite microstructure was obtained using HPT processing at temperatures higher than 423 K via intersecting mechanical twinning.
(2) Type 304 metastable austenite and type 310S stable austenite exhibited a high UTS of 1.6–1.8 GPa for NC microstructures in which the uniform elongation was limited to ∼3% via grain refinement. For the type 304 steel, a plateau regime appeared in the true stress–true strain response immediately after the onset of necking. This is attributed to α′ martensitic transformation, which results in a significant local elongation.
(3) The grain size of the transition in tensile deformation behaviour was approximately 0.4 µm for the UFG microstructures in type 304 steels processed using HPT and subsequent annealing. A YS of ∼770 MPa, a UTS of ∼1 GPa, and an elongation-to-failure of ∼60% were attained for the UFG microstructure, with an average grain size of 0.5 µm.
(4) The dependence of YS on the grain size held the Hall–Petch relationships with a friction stress of 175 MPa and a Hall–Petch coefficient of 450 MPa µm^1/2 for UFG type 304 steel in the uncharged state. The hydrogen charge increased the friction stress by 40% but did not affect the grain refinement strengthening. The hydrogen-induced ductility loss was mitigated by grain refinement, while maintaining the high strength of the UFG microstructure. This is because generation of excess hydrogen owing to dynamic α′ martensitic transformation is dispersed by grain refinement.
(5) The YS values of the single-variant specimens of NT type 304 stainless steel were determined as 540–760 MPa depending on the twin plane orientation. The hydrogen-charged NT specimen with the soft orientation exhibited a moderate uniform elongation by retardation of α′ martensitic transformation and twinning parallel to the twin boundaries. In the hard orientations, where the twin boundaries were parallel and perpendicular to the loading axis, the hydrogen-induced twin boundary separation contributed positively and negatively to ductility, respectively, while increasing the UTS. Shear deformation parallel to the twin plane plays a determining role in the mitigation of HE in NT type 304 stainless steel.
(6) The fatigue crack growth rate in the UFG microstructure was reduced more in type 304 metastable austenite than in type 310S stable austenite. In the UFG metastable austenite, single-variant α′ martensite transformation ahead of the fatigue crack tip changed the route of damage accumulation, resulting in the enhanced resistance to short fatigue crack growth.
(7) The fatigue crack growth rate in nanotwin bundles was reduced independently of the twin plane orientation in NT type 304 stainless steel. Detwinnig and subsequent α′ martensite transformation were caused by interaction of dislocations with the nanotwin boundaries, resulting in the reduced crack growth rate.

These findings suggest that introducing ultrafine grains and nanotwin bundles is effective for mitigation of the hydrogen-induced ductility loss and enhanced resistance to short fatigue crack growth through controlling the dynamic martenstic transformation. Plastic deformation processing of metastable austenitic steels at controlled moderate temperatures can be the promising strategy for obtaining sustainable materials.

Acknowledgment

The author is indebted to R. Oura, A. Matsushita, M. Fujiura (Kumamoto University), and Dr. S. Ueki (Kyushu University) for their valuable assistance in conducting the micro-tension and micro-fatigue testing experiments and to Prof. Z. Horita (Kyushu University) and Prof. K. Takashima (Kumamoto University) for their insightful discussions. This study was supported in part by Grants-in-Aid for Scientific Research (B) JP19H02464 and (A) JP20H00311 from the Japan Society for the Promotion of Science (JSPS).

REFERENCES

Corresponding author

Register with J-STAGE for free!