Introduction

The demand for metallic materials with an excellent balance between strength and ductility, as well as good corrosion resistance at cryogenic temperatures, is growing rapidly for applications such as liquefied gas storage and transportation, as well as space propulsion [1,2,3]. In this respect, 316L austenitic stainless steels (SS) are widely used in a cryogenic environment due to the excellent combination of strength and plasticity. At room temperature and elevated temperatures, they deform by a combination of deformation twinning and slip. At cryogenic temperatures, they show improved strength which is attributed to transformation-induced plasticity (TRIP) arising from the deformation-induced martensitic transformation (DIMT) [4, 5]. In essence, this is due to the higher strength of α′-martensite compared to its parent γ-austenite phase, as well the load partitioning between the different phases.

There are two martensitic transformation pathways that occur for 316L SS during straining at cryogenic temperatures; pathway I: γ-austenite (face-centred cubic (FCC) structure) → α′-martensite (body-centred cubic (BCC) structure) and pathway II: γ → ε-martensite (hexagonal close-packed (HCP) structure) → α′ [6,7,8]. This reflects the fact that ε can act as an intermediate phase, but it is not a necessary precursor for α′ [9, 10]. The double-shear theory proposed by Olson and Cohen [11] considers the intersection of shear bands as the nucleation site of α′ including a kinetic model dominated by partial dislocation gliding to describe the DIMT process. However, recent reports have shown that not only the intersections of shear bands, but also single-shear bands can be preferential nucleation sites for α′ when only one glide system is activated [9, 12]. Moreover, various forms of shear bands, including slip bands, mechanical twins, stacking faults bundles and ε phase, can arise during deformation and serve as potential nucleation sites for α′ [12,13,14], thereby leading to multiple transformation pathways.

These transformation pathways vary with deformation temperature, straining level and rates, grain orientation, etc. [15,16,17,18,19], dependent upon the γ stability and stacking fault energy (SFE) [16]. It has been found that the martensitic transformation changes from γ → α′ to γ → ε → α′ as the deformation temperature decreases in some austenitic stainless steels or alloys [9, 16, 20, 21]. A previous report [22] has identified that when SFE decreases to approximately 18 mJ/m2, the deformation mechanism changes from twinning to martensitic transformation. For austenitic stainless steel, this transition occurs at temperature close to 173 K. Further lowering the deformation temperature may promote the formation of ε as a nucleation site, which is associated with decreased SFE. The transformation pathways can also differ at various stages of straining, resulting in a variety of nucleation and growth mechanisms of α′ and associated mechanical responses. For example, Sato et al. [23] and De et al. [24] reported that the γ → ε → α′ pathway tended to occur during the early-stage deformation at 223 K in 304 austenitic steel. The grain orientation is another critical factor affecting the martensitic transformation with specific orientation relationships (OR) between the γ and the new phase such as the Kurdjumov–Sachs (K–S) (111)γ//(110)α and [110]γ//[111]α OR and the Pitsch (100)γ//(110)α and [110]γ//[111]α OR [25,26,27,28]. The clarification of the grain orientation and OR in the martensitic transformation process can aid in a better understanding of the transformation mechanism. Nevertheless, the deformation-induced martensitic transformation and the selection of transformation pathways in 316L SS under cryogenic deformation are poorly understood.

Here, we consider 316L SS deformed at a range of cryogenic temperatures, including ultra-low temperatures (15 and 50 K), near the phase transformation transition temperature (173 K) and at room temperature. To this end, we use electron backscattered diffraction (EBSD) and transmission Kikuchi diffraction (TKD) to reveal and classify the complex transformation pathways and their dependence on grain orientation, straining level and local microstructural features including shear bands, stacking faults and mechanical twins. The mechanical response, temperature dependence and orientation relationship of different transformation pathways were also reported.

Experiment

A commercial type 316L stainless steel (SS) was supplied by Baosteel Company. It was manufactured by hot extrusion and annealing at 1040 ℃ followed by air cooling to room temperature having chemical composition (in weight per cent) listed in Table 1. The as-received specimen is fully austenitic (Fig. 1a), with a grain size ~ 18 µm and a few annealing twins are observed (Fig. 1b). Dog bone-shaped tensile samples (machined by electrical discharge machining) having a gauge length of 42 mm and a diameter of 8 mm with M12 screw-threaded ends were prepared with the longitudinal direction (LD) of the rolled plate along the tensile axis (Fig. 2a). The tensile tests were carried out at 297, 173, 50 and 15 K at a strain rate of 10−4 s−1 on an Instron tensile testing machine having a load capability of ± 100 kN with a cryogenic chamber attached providing a high-vacuum environment (< 10−5 Pa). The deformation temperature was varied using a built-in heater and two liquid helium tubes. The temperature was stabilized for 30 min at the high-vacuum state before tensile loading. An extensometer was used to measure the strain during tensile loading. LD, TD and ND represent longitudinal (loading) direction, transverse direction and normal direction, respectively.

Table 1 Chemical composition of the 316L SS (wt%)
Figure 1
figure 1

a EBSD phase map and b inverse pole figure (IPF) map of the as-received sample for a longitudinal direction–transverse direction (LD–TD) cross-section. Black lines represent high-angle grain boundaries above 10°.

Figure 2
figure 2

Schematic showing a the start, and b the end, of tensile deformation. The areas for the detailed microstructural examination of tensile specimens are shown in c and d

The samples for microstructural observation were prepared by sectioning the fractured tensile specimen along the loading direction (shown in Fig. 2) followed by grinding and polishing. The microstructural evolution of the samples was characterized by a field emission gun scanning electron microscope (SEM, TESCAN Mira3 LC) equipped with an electron backscatter diffraction (EBSD) system from Oxford Instruments. The SEM beam parameters were set to an accelerating voltage of 15 kV, current of 11 nA and the step size of 400 nm. The AZtec EBSD data acquisition software was used to acquire the electron diffraction patterns with a binning of 2 × 2 pixels, 10 bands for each pattern detected and 80 of Hough space resolution.

Transmission Kikuchi diffraction (TKD) was used to study the microstructure arising from the martensitic transformation pathways. The TKD disc shape samples were extracted from regions of the failed specimens along the deformation direction (shown in Fig. 2). They were firstly mechanically ground followed by being punched into discs with 3-mm diameter. The foils were prepared by twin-jet electropolishing using an electrolyte of 5% perchloric acid and 95% methanol in TenuPol-5 at − 35 °C and the setting voltage of 35 V. TKD analysis was carried out on FEI Magellan HR equipped with EBSD detector. The TKD data were collected at 20 kV acceleration voltage, the beam current of 16 nA and the step size of 30 nm, with the sample tilted 20° to the EBSD detector. The EBSD and TKD data processing were performed using the AZtecCrystal software. The strain values of the measured areas were estimated using finite element method (FEM) simulation in ABAQUS. These data were simulated based on the stress–strain data obtained during the tensile tests, corresponding accurately to the actual stress–strain results. The parameters used and simulated results for 15, 50 and 173 K deformation are described in the Appendix.

Results

Mechanical properties

Figure 3 shows the engineering stress–strain curves for 316L SS deformed at 15, 50, 173 and 297 K. The yield strength (YS) and ultimate tensile strength (UTS) of the samples increase as the deformation temperature decreases, rising, respectively, from 660 and 750 MPa at 297 K to 1010 and 1260 MPa at 15 K. The maximum elongation (~ 42%) is recorded at 173 K, surpassing the values observed at 297 K (~ 34%) and 15 and 50 K (~ 27% and ~ 28%, respectively). At 15 and 50 K, there is a slight decrease in stress following yielding, after which it remains relatively constant until engineering strain reaches 0.1, at which point, there is a sharp increase in the strain hardening rate. This rate of increase is notably higher than that observed at other temperatures.

Figure 3
figure 3

Engineering stress–strain curves at 15, 50, 173 and 297 K.

Microstructural evolution under deformation at cryogenic temperatures

The microstructural evolution as a function of straining at 15 K is shown in Fig. 4, characterized using EBSD phase and inverse pole figure (IPF) maps. At a strain of 0.03 (Fig. 4a), a considerable amount of α′ is already evident within the γ grains with a trace amount of ε present too, implying that the deformation-induced phase transition has occurred. At 0.07 strain, the martensitic transformation continues and is accompanied by non-uniform deformation which is evidenced by the orientation spread in individual grains (misorientation less than 10°) and subdivision of the grains (misorientation greater than 10°) (Fig. 4b). The amount of the ε increased slightly from 0.03 to 0.13 strain (see the transformed microstructure in Fig. 4a–c), but with further straining to 0.25 (Fig. 4d), any ε has transformed to the α′ phase. Overall, the majority of the γ phase appears to have directly transformed to the α′ phase (Pathway I) in agreement with in situ diffraction experiments conducted by the authors [29]. Initially, much of the α′ forms as complex parallel strips or gridded structure containing numerous high-angle grain boundaries (> 10°) (Fig. 4b). The parallel α′ structures tend to have a similar orientation, which is different to those in the perpendicular direction. In contrast, there are few high-angle grain boundaries in γ grains due to its preferential martensitic transformation instead of dislocation slip and deformation twinning. At the strain of 0.25 (Fig. 4d), the most severe distortion and irregular shapes can be seen within the α′ grains, where fine grains with high-angle grain boundaries are formed. From Fig. 4, we can see that γ grains are prone to undergo γ → ε → α′ or γ → α′ phase transformation with few sub-grain boundaries and little distortion inside the γ grains.

Figure 4
figure 4

EBSD phase (left) and inverse pole figure (IPF) (right) maps for 316L SS samples deformed at 15 K to a 0.03, b 0.07, c 0.13 and d 0.25 engineering strain. γ (red), ε (blue) and α′ (green) areas were all identified in the EBSD maps. Black lines represent high-angle grain boundaries above 10°.

Figures 5 and 6 illustrate the phase and microstructural evolution for the samples deformed at 50 and 173 K, respectively. These results show similar features to those seen for deformation at 15 K in Fig. 4. Again, both γ → ε → α′ and γ → α′ transformation pathways can be observed during the deformation forming complex networked α′ structures (Figs. 5a and 6b). As at 15 K, many α′ grains display elongated a strip-like morphology, which cross perpendicularly.

Figure 5
figure 5

EBSD phase maps (left) and inverse pole figure (IPF) maps (right) for 316L SS samples deformed at 50 K to a 0.05, b 0.13, c 0.20 and d 0.26 engineering strain. γ (red area), ε (blue area) and α′ (green area) were all identified. Black lines represent high-angle grain boundaries above 10°.

Figure 6
figure 6

EBSD phase maps (left) and inverse pole figure (IPF) maps (right) for 316L SS samples deformed at 173 K to a 0.03, b 0.09, c 0.25 and d 0.40 engineering strain. γ (red area), ε (blue area) and α′ (green area) were all identified. Twins are indicated in white boxes. Black lines represent high-angle grain boundaries above 10°.

At 173 K (Fig. 6), some deformation twinning (< 111 > 60°) within the parent γ grains can be observed with straining, as shown in the white boxes in Fig. 6a and c. This indicates that the sample at 173 K deforms by a combination of deformation twinning and martensitic transformation. Twins can arise in areas of stress concentration, which can inhibit the formation and propagation of cracks [30], thereby amplifying the ductility. Furthermore, the amount of criss-crossed α′ seems to be lower at 173 K than that at 15 and 50 K. For deformation at 173 K, the rate of transformation to α′ with straining is slower than when deformed at 15 and 50 K, being insignificant at small strains and with some γ phase retained and several large α′ grains observed even close to the failure strain.

Evolution of phase volume fraction

The evolution of γ-FCC, α′-BCC and ε-HCP phase volume fractions as a function of strain is quantified in Fig. 7. It is clear that no deformation-induced martensitic transformation occurs during the deformation at 297 K. The martensitic transformation with straining increases with decreasing temperature, leaving just 11.91%, 18.05% and 36.69% of γ phase at ~ 0.25 engineering strain at 15, 50 and 173 K, respectively. Besides, the rate of transformation with strain to the α′ increases with decreasing temperature, with the level or transient ε also increasing. The volume fraction of the ε initially increases with strain, but subsequently decreases and transforms into the α′ upon reaching its peak at all cryogenic temperatures. Upon decreasing temperature from 173 to 15 K, the fraction of transient ε rises peaking at 9.11%, 6.64% and 4.01% at 15, 50 and 173 K at engineering strains of approximately 0.13 at 15 and 50 K and 0.25 at 173 K, respectively. These results are in general agreement with in situ neutron diffraction studies mapping the changes in the three phases with strain [29].

Figure 7
figure 7

Evolution of γ-FCC (squares), ε-HCP (hexagons) and α′-BCC (diamonds) phase as a function of engineering strain 15, 50, 173 and 297 K in the 316L SS, obtained by EBSD measurement. The error bars represent the standard deviation in phase volume fraction.

Grain orientation dependence of transformation

Figure 8a shows the phase map of a region of the sample deformed to an engineering strain of 0.03 at 15 K, while the orientations of several γ grains that undergo γ → ε transformation are marked with white letters A–E, and their crystallographic orientations with transforming planes and directions are plotted in Fig. 8b. This reveals that the ε phase has an OR close to the Shoji–Nishiyama (S–N) orientation relation (111)γ//(0001)ε and [110]γ//[11–20]ε [31], to the γ matrix. More generally, as shown in Fig. 9, among all grains undergoing the γ → ε martensitic transformation, more than 90% of the adjacent γ/ε boundaries have an S–N OR within a deviation of 5°.

Figure 8
figure 8

a EBSD phase map showing γ-austenite, α′- and ε-martensite as red, green and blue for sample deformed at 15 K. Black lines represent high-angle grain boundaries above 10°. b Orientations of 5(A, B, C, D and E) example austenite grains (left) selected from the EBSD map that underwent the γ- to ε-martensitic transformation alongside their corresponding ε phase (right) illustrated, showing transforming planes and directions.

Figure 9
figure 9

The relative frequency distribution of the deviation (in degrees) of γ/ε boundaries from the S–N orientation relationship (OR) at 15 K.

Figure 10 shows the ε phase orientation distribution maps for the sample deformed to 0.13 strain at 15 K which is the condition having the highest ε content (see Fig. 7). At different strains and deformation temperatures, the developed texture distribution of ε is similar, and therefore, to avoid repetition, only an example of the pole figures of ε at 15 K strain is shown. All pole figures are presented using equal-area projection showing that the centre of the pole figure corresponds to the longitudinal axis. It is clearly seen that the strong and preferential orientation of the grains is developed from these poles, which is a direct consequence of the cryogenic plastic deformation. Figure 11 exhibits the texture intensity evolution of (0001)ε pole with the angle from the longitudinal direction at cryogenic temperatures. All (0001)ε poles exhibit strong fibre symmetry and mainly orientated at the angle of approximately 50°–65° to the longitudinal direction, with a maximum pole density of about 55°. Figure 9 shows that the OR between γ and ε follows (111)γ//(0001)ε during the γ → ε transformation for samples deformed at cryogenic temperature. This indicates that at all cryogenic temperatures, γ grains with the (111)γ plane normal at about 50°–65° from the loading direction are more prone to the γ → ε martensitic transformation. Tao et al. [27] have reported that deformation-induced martensitic transformation in 304L SS shows a grain orientation dependence, indicating that the (111) planes of γ grains oriented at 10° and 50° from the loading direction are more prone to γ → ε transformation. It is worth noting that in 50°–65° region, the texture intensity is the highest at 15 K, indicating a stronger distribution of (111)γ planes in this direction at 15 K, implying that the sample under lower-temperature deformation prefers to undergo γ → ε transformation.

Figure 10
figure 10

Pole figures of the ε-HCP martensite at 15 K after undergoing deformation to an engineering strain of 0.13.

Figure 11
figure 11

The texture intensity distribution of (0001)ε pole relative to longitudinal direction at a 15, b 50 and c 173 K after undergoing deformation to an engineering strain of 0.13, 0.13 and 0.25, respectively.

Discussion

Microstructure of transformation pathways

From the results above, it shows that the martensitic transformation plays a crucial role in the cryogenic deformation process and exhibits different transformation behaviours and microstructures along different transformation pathways, including ε-martensite, twins, α′-martensite, etc. Therefore, we employed TKD to investigate the microstructure and mechanisms of nucleation and growth of martensite at high resolution for different phase transformation pathways. These observations were conducted in regions with limited plastic deformation (e.g. 0.03 and 0.07 strain) to avoid experimental errors associated with grain distortion arising from severe plastic deformation.

Figures 12 and 13 exemplify the transformed microstructure of 316L SS at 15 K with 0.03 strain, showing the γ → ε → α′ phase transformation pathway at nanoscale level using TKD. The phase distribution map in Fig. 12a shows the co-existence of all three phases where the ε phase, shown as shear bands in this phase map, has a strip-like structure that has nucleated within a γ grain. From the corresponding IPF map (Fig. 12b), it is clear that all the ε grains in this region have a similar orientation having transformed from the same parent γ grain. Furthermore, the newly formed α′ grains having a lenticular and blocky morphology are distributed within the ε band. In the same ε band, newly formed α′ grains nucleated and grown independently from each other but stayed in the domain of ε bands. In addition, it was also found that several α′ grains coarsen (white arrows in Fig. 12a). This occurs because they themselves act as nucleation sites triggering further deformation-induced martensitic transformation due to the increasing local stress. These microstructural features suggest that the ε bands and α′ sites can provide the nucleation sites for the transformation of the new α′ phase after which the initially separated α′ grains gradually grow.

Figure 12
figure 12

TKD maps showing the γ → ε → α′ (pathway II) for straining at 15 K to 0.03 strain. a Phase distribution map showing γ-austenite, ε- and α′-martensite as red, blue and green, respectively. b IPF map. c Orientation relationships distribution map, Shoji–Nishiyama (S–N) and Kurdjumov–Sachs (K–S) are represented by green and red lines. d Representative schematic of grain orientation in the unit cell of γ, ε and α′ showing orientation and transforming plane. e Pole figures of S–N and K–S ORs between γ, ε and α′.

Figure 13
figure 13

a TKD maps showing the γ → ε → α′ (pathway II) martensitic transformation with orientation relationships map of the representative grown α′-martensite and other phases obtained at 15 K with 0.03 strain. Shoji–Nishiyama (S–N), Kurdjumov–Sachs (K–S) and Pitsch are represented by green, red and cyan. b Pole figures of Pitsch OR between γ and grown α′.

Figure 12d shows the representative grain orientations of the γ, ε and α′ unit cells along with transforming plane and direction while Fig. 12e shows the nature of the boundaries among these three phases follow the S–N OR (represented by green lines in Fig. 12c). The α′ grains display the Kurdjumov–Sachs (K–S) (111)γ//(110)α and [− 110]γ//[111]α OR (red lines in Fig. 12c) to the parent γ.

With the increasing local stress, α′ grains grow out of the ε bands. As shown in Fig. 13, the OR between the coarsened α′ and γ changes from K–S to Pitsch (001)γ//(− 101)α and [110]γ//[111]α (see orientation relationship maps) represented by the cyan lines in the OR distribution map for the selected area. The change of OR process may be achieved by a rotation of (110)α around [111]α, resulting in (001)γ//(− 101)α, and the rotation is attributed to the continuous lattice bending [32, 33].

In the transformed microstructure shown in Fig. 14, no ε is observed. Instead, α′ grains of granular or lenticular shape appear to nucleate directly inside the γ grains. These form parallel features where all the grains are of similar orientation. The deformation-induced α′ nucleates on the formed α′ through an autocatalytic process, causing the α′ to become coarser, which then merges with the adjacent α′ into strips [12]. Smaller α′ grains merge into strips, interlacing with the nucleated granular α′ (white arrows in Fig. 14b). K–S OR and Pitsch OR (represented by red and cyan in Fig. 14c) predominate [25, 26]. The newly nucleated α′ and γ follow K–S OR, which becomes the Pitsch OR as the α′ grains continue to grow [34]. According to the previous reports [9, 12, 13], this may be a process completed by stacking fault bundles acting as shear bands, resulting in a γ → α′ transformation pathway. The distinction from α′ nucleation with ε sites lies in the fact that these pathways are driven by different grain orientations. Although the occurrence of stacking faults was not directly observed in our study, the nucleation features of α′ in this case are similar to those in γ → ε → α′ transformation.

Figure 14
figure 14

TKD maps showing the γ → α′ (pathway I) for straining at 15 K to 0.03 strain. a Phase distribution map showing γ-austenite and α′-martensite as red and green, respectively. b IPF map. c Orientation relationships distribution map, Kurdjumov–Sachs (K–S) and Pitsch are represented by red and cyan. d Representative schematic of grain orientation in the unit cell of γ and α′ showing orientation and transforming plane. e Pole figures of K–S OR between γ and α′. f. Pole figures of Pitsch OR between γ and α′.

We also observed the γ → α′ martensitic transformation in which α′ can nucleate at the grain boundaries of γ as shown in Fig. 15. The α′ grains have irregular ovoid shapes of different sizes. We did not observe shear bands as nucleation sites of α′ in this case. The nucleated α′ grains in this case tend to have the same grain orientation (Fig. 15b). In this case, α′ has the Nishiyama–Wassermann (N–W) OR (Fig. 15e), i.e. (111)γ//(110)α and [011]γ//[001]α, with the parent γ, as shown by the purple lines in Fig. 15c. Figure 15d shows the schematic of grain orientation in the unit cell of this transformation with N–W OR showing transforming plane and direction. This type of α′ nucleation at γ grain boundaries with N–W OR seems to have a different transformation mechanism. He et al. [33] also found a similar pathway with N–W OR in the deformation-induced martensitic transformation of austenitic steels and pointed out that this process is achieved by single-shearing, which involve the release of the lattice strain. Karewar et al. [35] found the 19.28° shear along the [11–2]γ and [0–11]α direction causes this change from γ → α′ using the method of molecular dynamics simulations. At present, there are only a few experimental studies of this direct phase transformation.

Figure 15
figure 15

TKD maps showing α′ (pathway I) nucleation at γ grain boundaries measured at 0.07 strain at 15 K. a Phase distribution map showing γ and α′ as red and green, respectively. b IPF map. c Orientation relationships distribution map, Nishiyama–Wassermann (N–W) and Pitsch are represented by purple and cyan. d Representative schematic of grain orientation in the unit cell of γ and α′ showing orientation and transforming plane. e Pole figures of N–W OR between γ and α′.

Deformation twins were only found in the sample deformed at 173 K to provide the nucleation sites for the small lenticular α′, as shown in Fig. 16. The α′ can form inside the twins or along the twin boundaries. The transformation microstructure in this case exhibits a remarkable resemblance to that depicted in Fig. 12 with ε as the intermediate phase. Comparing with deformation at 15 and 50 K (with SFE of 4.9 and 7.8 mJ/m2, respectively), the increased SFE (17.7 mJ/m2) at 173 K has led to the γ → γtwins → α′ transformation pathway. Tian et al. [9] found that the α′ can form within individual shear bands in the form of twins with the help of ε in austenitic Fe–Cr–Ni alloys. In our results, no ε was observed accompanying the twins within the same shear band. This is attributed to the increase in deformation temperature, which thermodynamically favours mechanical twins, enabling them to act directly as nucleation sites for α'. In Fig. 16c, the α′ grain obeys K–S OR (red line) to the γ grain, while when α′ is growing to break through the twinning boundaries, the orientation relationship changes to Pitsch OR (cyan). Such a change in orientation relationship is consistent with the growth of α′-martensite at the ε bands in Fig. 13. Although ε can also be nucleation site at 173 K with the similar structure, the amount of ε is less than that at 15 and 50 K.

Figure 16
figure 16

TKD maps at 173 K showing the γ → γtwins → α′-martensitic transformation obtained at  0.03 strain. a Phase distribution map and insert of twin boundaries image as blue. b IPF map. c Orientation relationships distribution map, Kurdjumov–Sachs (K–S) and Pitsch are represented by red and cyan. d Representative schematic of grain orientation in the unit cell of γ and α′ showing orientation and transforming plane. e Pole figures of K–S OR between γ and α′. f Pole figures of Pitsch OR between γ and α′.

The formation of stacking faults, ε-martensite and twins

The phase transformation pathway and the preferred nucleation sites for α′ exert a profound influence on the resulting mechanical properties. Figure 17 shows the formation of stacking faults, ε (HCP structure) and twins from the γ (FCC structure) by the glide of Shockley partial dislocations. The original stacking sequence for FCC structure is labelled as ABCABCABC. When a Shockley partial dislocation is activated on the (111) plane, all layers above that plane will move by a Burger vector of 1/6 < 11\(\overline{2 }\)>. This progress will lead to a stacking sequence of ABCACABC and create a stacking fault. After this, activation of a second Shockley partial dislocation gliding on the second neighbouring plane will form HCP stacking (ABCACACABC). Therefore, stacking faults formed by Shockley partial dislocations could lead to the HCP crystal structure. A stacking fault in the FCC structure will produce two HCP layers and further introduction of Shockley partial dislocations in every other plane will thicken the HCP layer, thereby growing the HCP phase. The difference with formation of the twins is that after introducing a stacking fault, the Shockley partial dislocation glides on both second and third neighbouring planes [36]. This process forms a three-layer twin (ABCBACABC). The ε can be an intermediate phase in the martensitic transformation, which leads us to believe that stacking faults and twins can play the same role as transition stage of martensitic transformation at early deformation. This is also consistent with the Bogers–Burgers model, where α′ can nucleate within various forms of shear bands based on the stacking fault [37, 38]. Rafaja et al. [39] found that stacking faults and twins associated with ε as nucleation sites for deformed-induced α′ in a TRIP steel at room temperature, which is achieved through the glide of partial dislocations on the (111)γ planes.

Figure 17
figure 17

Schematic showing the formation of the stacking faults (SF), HCP phase and twins. Red, blue and green balls represent atoms corresponding to layers A, B and C. The stacking faults are formed from the FCC phase by introducing Shockley partials dislocation with Burger vector 1/6 < 11\(\overline{2 }\)> on neighbouring two (111) planes. The HCP phase is formed by the gliding of Shockley partials dislocation on every other (111) plane or by two consecutive stacking faults. The twins are formed by gliding of Shockley partials dislocation on neighbouring (111) planes.

This can explain why the fraction of ε is not prominent during the phase transformation, because twins and stacking faults can provide sites for α′ in the nucleation with a similar mechanism, forming γ → α′ transformation instead of γ → ε → α′ transformation. This can also be confirmed by the nucleation and growth characteristics of α′ and the related orientation relationship between each phase during the martensitic transformation. In Figs. 12, 14 and 16, α′ grains nucleate within individual shear bands with similar features of deformed microstructure: the separate lenticular α′ nucleate within the shear band where they each act as sites and gradually grow in autocatalytic progress. All α′ formed within the same shear band exhibit the same grain orientation. As for the orientation relationship, the α′ formed in the shear bands follows the K–S OR with the parent γ at the beginning of nucleation and changes to Pitsch OR when it grows out of the shear bands into the γ. This result is consistent with the previous research showing that the OR between the α′ and γ changes during the growth process of α′ [33, 40, 41].

Moreover, it is well accepted that the temperature-dependent stacking fault energy has a strong influence on the transformation pathway (Fig. 7). At 15 K, the maximum volume fraction of ε measured was significantly higher than at 50 and 173 K, increasing by 94% and 172%, respectively. As the temperature decreases, more γ in the sample undergoes the γ → ε → α′ transformation. Since stacking faults can act as the precursor of ε, stacking fault energy (γSFE) can be defined to be the Gibbs free energy required to form a platelet of ε of the thickness of only two atomic layers based on the model proposed by Olson and Cohen: \({\gamma }_{{\text{SFE}}}=2\rho\Delta {G}^{\gamma \to \varepsilon }+2\sigma\), where \(\rho\) is the molar surface density along (111) planes, \(\Delta {G}^{\gamma \to \varepsilon }\) is the molar Gibbs energy of the γ → ε transformation and \(\sigma\) is the surface energy of the interface γ/ε (J/m2) [42, 43]. There is positive correlation between stacking fault energy and energy of this phase transformation. Besides, according to [4, 42, 44,45,46,47], the stacking fault energy decreases with dropping temperature in austenite stainless steels, leading to larger separation of partial dislocations. Therefore, the decreased temperature results in the curtailed energy required for the formation of stacking faults and ε, which, in turn, makes the phase transformation accelerated, and ε is more likely to be generated. By contrast, the sample deformed at 173 K exhibited twins from the start of the straining process, some of these twins served as the nucleation site for the initial α′ as displayed in Figs. 6a and 16. In our previous work [29], the SFE was measured as ~ 17.7 mJ/m2 at 173 K. According to [4, 42, 48], twinning becomes the main deformation mechanism when SFE is between 18 and 45 mJ/m2, and phase transformation is more favourable when SFE is less than 18 mJ/m2. Therefore, the fact that SFE is 17.7 mJ/m2 at 173 K promotes the likelihood of initiating α′ embryos within the twinned grains.

Conclusions

In our work, we have investigated the microstructural features of different transformation pathways in deformation-induced martensitic transformation of 316L SS at 15, 50 and 173 K. The transformation pathways depend on temperature and grain orientation with the corresponding nucleation and growth characteristics discussed. The main conclusions are as follows:

  1. 1.

    The excellent strength of 316L SS observed for all the cryogenic temperature deformed samples is mainly attributed to the deformation-induced martensitic transformation. For deformation at 173 K, 42% elongation is achieved due to the contribution from twinning.

  2. 2.

    The γ → α′ transformation pathway (I) predominates at all cryogenic temperatures, while the γ → ε → α′ pathway (II) became increasingly important with decreasing temperature. This is related to the underlying α′ formation mechanism of the temperature-dependent ε, stacking faults and twins as intermediate structures for the martensitic transformation.

  3. 3.

    γ and ε follow the S–N OR of (111)γ//(0001)ε and [110]γ//[11–20]ε. γ grains with (111) plane normal at an angle between about 50°and 65°from the loading direction are prone to the γ → ε → α′ martensitic transformation.

  4. 4.

    At 15 and 50 K, the α′ can nucleate in shear bands with the help of ε and stacking faults, while at 173 K, twins can also serve as sites for α′ nucleation, all of which exhibit comparable microstructures of transformation: separate lenticular α′ nucleate within the shear band and gradually grow through an autocatalytic progress.

  5. 5.

    The orientation relationship between γ and α′ is related to the nucleation and growth of α′: The γ and newly nucleated α′ follow the K–S OR and then change to Pitsch OR as the α′ grows. Another transformation involves irregular oval α′ nucleating directly at grain boundaries of γ with N–W OR.