One of the most studied treatments during which martensite is induced is the continuous cooling treatment under constant stress σ, see Figure 4a. Usually, the initial structure is fully austenitic, either because the steel of study is austenitic or because it has been previously austenitized (high temperature deformation). In this case, if the cooling rate was high enough to avoid ferritic/pearlitic/bainitic transformations, a martensite transformation would happen [49,92,100]. The transformation would be athermal, it would start at a temperature Ms(σ) [49,92,100], where Ms(σ) would be higher than Ms if the deformation mode was uniaxial tension, uniaxial compression or torsion [49,92,100,101,102,103]. In addition, in cases in which the Mα′f temperature was above room temperature, the α′ volume fraction would increase under stress, because of the additional mechanical driving force and (if the applied stress was above the austenite σY) the increase in nucleation sites [49]. In cases in which no plastic deformation was introduced, the addition of the mechanical driving force would not only affect the start of the α′ transformation, but also the selection of specific crystallographic variants [84,104,105,106], especially during the first stages of the transformation [106]. The selected variants would have the highest total driving force in absolute value, as ΔGmech is a function of the position of the habit plane [84,104,105]. Variant selection in α′ has also been reported in transformations on continuous cooling under stresses higher than the σY by Liu et al. and it has been associated to the application of stress rather than to the introduction of dislocations [49]. This assumption would follow the same line of thought of Tamura, who suggested that the strain-induced transformations can be understood only in terms of the applied stress [50,107] or of Chatterjee and Bhadeshia, who reached the same conclusion later [69]. Regarding the martensite scale, Liu et al. reported that, the α′ microstructure obtained by applying 70 MPa to an austenitized medium carbon steel during continuous cooling was more refined than the stress-free α′ microstructure obtained by a continuous cooling. They attributed this refinement to the small applied plastic deformation (<5%), which led to the recrystallization of some prior austenite grains at high temperature and to a dislocation density increase in the non-recrystallized grains. The scale of the α′ strain-induced laths/plates has not been reported in any other study, to the authors’ knowledge.

No work in which multiphase structures with retained austenite were cooled under stress have been found by the authors. Although it would be expected that the transformation of the retained austenite would also be promoted at higher temperatures than the ones at which the stress-free martensite transformation would happen, it is important to consider that the retained austenite may also be subjected to hydrostatic stresses that would lead to the stabilization of the retained austenite.

Martensite can also be induced when a steel is held at a fixed temperature T—always below Md—while an increasing stress is applied, see Figure 4b. When martensite is induced during straining in austenite structures, it is possible that the transformation process is γ→ ε, γ→ ε →α′, or γ →α′ [25], depending on the applied strain and the deformation temperature [25], where the formation of ε is always preceded by the formation of stacking faults [39]. In addition, the application of stress can also lead to the coalescence of α′ or ε, as previously reported [108,109,110,111].

Regarding variant selection, stress/strain induced α′ [23,50,51,52,112], and ε [23,48,52] present this phenomenon. Even when the transformation happens from a fully austenitic structure, there is no agreement about the reasons of the variant selection phenomenon. While some authors claim that variant selection in transformations from a fully austenitic structure may be explained by the mechanical driving force criterion [48,50,51,52], other authors suggest that some transformation sequences lead to local stress fields which make the analysis more complex [23]. The rules governing variant selection in TRIP transformations in multiphase structures—i.e., formation of stress or strain induced phases from retained austenite—have not been deeply studied. Yamashita et al. studied this phenomenon in a microstructure consisting of ferrite, bainitic ferrite, and retained austenite which was subjected to tensile tests. The retained austenite transformed to α′, always governed by variant selection, although the explanation for this variant selection was not always based on the mechanical driving force, but depended on the location of the retained austenite (at the ferrite grain boundaries or inside the ferrite matrix) [112]. The effect of other types of matrices on the selection of variants of α′ formed by TRIP effect in multiphase microstructures is still unclear.

The mechanical response that is observed when a fully austenitic structure is subjected to tension has been assessed in many austenitic steels. To better understand how a stress–strain curve changes under the influence of the appearance of martensite and why those changes occur, one can use Figure 5, where there are three stress–strain and their corresponding strain hardening rate-strain curves which correspond to three different steels which are fully austenitic before testing. The austenite σY of those three steels are the same at the testing temperature. Note that the steel B seems to yield at a lower stress, although this is only an effect of a stress-induced transformation, as is explained subsequently. Because of the different thermodynamics of the steels, steel A does not undergo any SIT/DIT while being deformed, steel B does undergo a SIT and steel C undergoes a DIT. Using the strain hardening rate-strain curve enables to notice much subtler abnormalities. As can be observed, if the transformation is SIT/DIT (steel B/C), the stress–strain curve takes a sigmoidal curve. After reaching σcrit−SIT and σcrit−DIT, a SIT/DIT starts and the stress–strain behavior is no longer as would expected. From this point onwards, the martensite fraction increases as the stress to which the austenite is subjected is higher [107,113,114], mainly as a function of the strain and not of the temperature [115], until it reaches a limit, see Figure 3. There are several effects that compete when a SIT/DIT takes place during loading and they can be studied by looking at the strain hardening rate–strain curves in Figure 5b. The first of them is called dynamic softening and it is a product of the transformation working as a competing deformation mechanism. It is associated with the fact that the SIT/DIT transformation strain contributes to the total strain, while the stress increases more slowly [116,117] and it usually predominates at low strain values [10]. In some cases, this phenomenon has also been associated to the formation of ε in the early stages [118], as this phase has been proved to be almost ideally plastic [119]. Dynamic softening is characterized by a rapid decrease of the strain hardening rate, as happens for steels B and C in Figure 5b at low strain values. For higher strains, one can observe that the second effect, called static hardening, is predominant. Static hardening is characterized by a strong increase of the strain hardening rate due to the presence of martensite which acts as a reinforcing phase [116,117]. It has been previously reported that, during the static hardening effect, the martensite volume fraction can increase up to a given amount, called percolation threshold [120], and that, up to this point, the austenite stress level is similar to the macroscopic stress [63]. Once the martensite volume fraction is higher than the percolation threshold, the martensite forms a percolating cluster which extends through the whole structure [120]. In this stage, the strain hardening rate further increases until it reaches its maximum value and the austenite stress level deviates from the macroscopic stress level, suggesting that the material is acting as a composite [10,63]. Finally, once the volume fraction of martensite is close to the unity, the material behavior starts to resemble the martensite behavior and, hence, the strain hardening rate starts to decrease [63]. This final hardening has also been attributed to the activation of strain hardening mechanisms which involve a high dislocation density or a complex state of stress, among others [10]. Note that, although all the mentioned effects are predominant at different stages, it could be possible that some of them were not noticeable during the deformation process. These concepts can be extrapolated to the mechanical behavior of a multiphase structure if one understands that the initial microstructure has already a composite-like behavior [26], as would happen for an austenite structure in which α′ has formed by TRIP effect in a high fraction, i.e., after the percolation threshold has been overcome. However, in this case, the retained austenite may be more stable against deformation depending on several factors: (a) its chemical stability, which depends on its chemical composition [121]; (b) in the case that the matrix phase has formed by a displacive transformation, the stability of the retained austenite increases because of the hydrostatic stresses that the matrix exerts on this retained austenite, as reported in Section 3.2.2. [28]; and (c) if the matrix phase is α_B, the amount of dislocations introduced during the transformation in the retained austenite mechanically stabilize it [122].

Therefore, if one wants to control the mechanical properties of a microstructure which contains retained austenite or that is fully austenitic and that is expected to undergo TRIP effect, one must control several factors. First, as mentioned, the austenite stability is one of the most important factors controlling the mechanical properties of these structures. If the austenite stability is low, a SIT or DIT starts at very low strain values and the material yield strength and elongation are lower, although their ultimate tensile strength is high [23,24,123]. If the austenite is too stable, the TRIP effect may never occur, hence, its benefits cannot be exploited [123]. If the austenite stability is high enough, the ultimate tensile strength of the material is not so high, although its elongation is higher [24]. In addition, it has been reported that, in an austenitic steel with rather stable austenite, the formation of martensite—specially α′, rather than ε—during straining leads to higher uniform elongations [25] because of the suppression or delay of the necking phenomenon [124]. Previous studies have shown that the uniform elongation can be maximized if it is made sure that the induced transformation continues until the latter stage of deformation or the onset of necking [107,115,125] and that the transformation rate is slow, as the dislocations have time to be accommodated and the local stresses can be suppressed [23,115]. The combination of high strength and high ductility can be obtained at temperatures close to Md at which the DIT starts for high strains, before the necking effect takes place, enabling its suppression [24]. A similar effect has been reported in multiphase microstructures [126,127,128].

Second, the strain rate can also affect the mechanical behavior, depending on whether the specimen is deformed under isothermal or under adiabatic conditions, on the product phase and on the deformation temperature. Generally, under isothermal conditions, an increase of the strain rate leads to an increase of the final fraction of stress or strain induced martensite [6,129]. However, if the strain rate is too high, there is a transition from isothermal to adiabatic material behavior, which makes the SFE increase [9,120,129,130] and the driving force for the transformation decrease (in absolute value) [131]. Therefore, using higher strain rates would decrease the volume fraction of stress/strain induced martensite as the strain rate is higher [9,131,132,133,134]. In multiphase microstructures, the effect of the strain rate follows similar trends [135,136,137].