Critical Assessment 15: Science of deformation and failure mechanisms in twinning induced plasticity steels

Deformation mechanism

The plastic deformation of twinning induced plasticity (TWIP) steels involves the progressive formation of elegant arrays of deformation twins.^1–5 Depending on the crystallographic orientation, different grains can twin on one or more twinning systems or even have no twins.^6–9 The twins normally have a thickness in the range of a few to several hundreds of nanometres and usually transverse the entire austenite grains.^10–13 Coherent twin boundaries can be transparent to the passage of lattice dislocations or act as obstructions, depending on the Burgers vector of the dislocations and the operative slip system, although there will be some weak resistance because the corresponding slip systems in the twin and matrix will be differently stressed.^14–16 There will, therefore, be some pile-ups at the coherent twin boundaries,^4,17,18 leading to an increase in the flow stress by inducing back stress,^4,19–21 similar but weaker than the effect of grain boundary. Detailed transmission electron microscopy (TEM) characterisation has confirmed the existence of such dislocation pile-ups against twin boundaries in circumstances where multiple slip systems operate,^4,22 and TWIP steels have been found to exhibit a strong Bauschinger effect. ¹⁹ Phenomenological models ¹⁹ have been proposed to reproduce the experimental back stress, based on the densities of twin and grain boundaries. While these models agree well with the experimental stress–strain relation and back stress, they may not capture the complete underlying mechanisms. The polarisation of the three-dimensional dislocation structure can also induce substantial back stress, ²³ especially when the dislocation density is very large and dislocation cells are not formed, which are the features of dislocation structure in TWIP steels. The contribution of twins on back stress and flow stress could be overestimated by this model. ¹⁹ Besides, twin boundaries can also enhance the workhardening rate by reducing dislocation mean free path.^19,24–27 Experimental measurements^28–31 of the dislocation density by X-ray diffraction and line profile analysis reveal that the dislocation density of TWIP steels at fracture (with a true strain of 0·4–0·5) is usually in the order of 10 ¹⁵  m^− 2, which may account for more than half of the total flow stress. Various models^{19,24,26,27,32,33} have been proposed to describe the dynamic reduction of dislocation mean free path by twinning, yet the effectiveness of these models needs to be verified with an accurate dynamic dislocation density measurement, which remains a challenge for experiments. Twinning mechanisms based on the dissociation of perfect dislocation into two Shockley partial dislocations, ³⁴ the formation of three-layer stacking fault, ³⁵ the stair rod cross-slip of partial dislocation ³⁶ and the dissociation of perfect dislocation into Shockley and Frank partial dislocations ³⁷ have been applied to explain the formation of twins in TWIP steels. Nevertheless, the operating mechanism is still under debate.

Perhaps not surprisingly, compared to deformation twinning, much less attention has been paid to dislocation slip in the context of TWIP steels. Owing to the low stacking fault energy (SFE), ³⁸ perfect dislocations in TWIP steels tend to dissociate into two Shockley partial dislocations connected by a narrow ribbon of intrinsic stacking fault, ³⁹ which has been confirmed by TEM observation. ⁴⁰ The dislocation cross-slip is strongly inhibited by such extended dislocation core unless this extended core is constricted by external force with the assistance of thermal activation.^39,41 Therefore, dislocations in TWIP steels prefer planar slip, which leads to the formation of planar dislocation structures (such as highly dense dislocation walls^10,42 at small strain and tangle structure at large strain^29,43–45) instead of the heterogeneous cell structure developed by wavy slip in face centred cubic metals with high SFE. With dislocation annihilation suppressed by the resistance to cross-slip, dislocation multiplication in TWIP steels may contribute significantly to the workhardening rate even in the absence of twinning. In fact, the experiments on single crystals of Hadfield steels indicate that the workhardening rate contributed merely by dislocation multiplication can be as high as that of polycrystalline samples with contributions from both dislocations and twins, ⁴⁶ indicating that the contribution of dislocations on the workhardening rate can be large. In other words, in the opinion of the current authors, the contribution of dislocations in the workhardening rate of TWIP steels may have been significantly underestimated in the literature. Stated another way, the contribution of twinning to the workhardening rate of TWIP steels may have been overestimated. More investigations are required in order to distinguish the respective contributions from dislocations and twins on the workhardening rate of TWIP steels.

In addition, the interaction between the dislocation and interstitial carbon atom should be another important mechanism to be investigated. The interstitial carbon atoms in TWIP steels can perform short range diffusion and cause lattice distortion^39,47 to the austenitic matrix. If the moving speed of carbon atoms is comparable to that of dislocations,^47–49 dynamic strain aging may occur after a critical strain and cause serrated plastic flow with the propagation of Portevin–Le Chatelier bands^50–53 and negative strain rate sensitivity.^54,55 Otherwise, if the mobile carbon atoms cannot follow dislocations, the dislocations will move in a jerky fashion by overcoming these carbon induced obstacles with thermal activation, ⁵⁶ which will lead to the positive strain rate sensitivity.^29,57–62 The chemical composition,^47,49,63 temperature^52,64,65 and strain rate,^52,54,66 which can modify the relative velocity of the carbon atoms and dislocations, can determine the occurrence of dynamic strain aging (DSA). While the relative contribution of DSA to the workhardening of TWIP steels is still controversial in the literature, one quantitative estimation ⁶⁷ based on modelling indicated that DSA accounts for no more than 20 MPa of the total flow stress ( < 3). Compared to the strong solid solution hardening of carbon (187 MPa wt-^− 1 increase in yield stress), ⁶⁸ the substitutive alloying elements such as Mn (Ref. 68) and Al (Ref. 69) only have minor effect on the yield stress. Microalloying elements including Ti, Nb and V have been be utilised to form nanometre sized carbides in the austenitic matrix by thermomechanical processing, e.g. cold rolling followed by controlled heat treatment. ²⁸ The strengthening coefficients on yield stress of these microalloying elements as carbides are 1380 MPa wt-^− 1 for Ti addition ≤ 0·1 wt-, 187 MPa wt-^− 1 for Nb and 530 MPa wt-^− 1 for V addition ≤ 0·4 wt- respectively. ²⁸ Detailed TEM observation reveals that dislocations (either perfect or partial) can bypass these carbides according to the Orowan mechanism, indicating that the strengthening effect should depend on the average distance of the carbides. ⁷⁰ Besides, it is also found that the nanometre sized deformation twins can bypass the carbides. ⁷⁰ While the carbides do not have significant effect on the workhardening rate at small strain, they may decrease the workhardening rate by suppressing the twinning kinetics at high strain level. ²⁸

The SFE of the austenitic matrix, which depends mainly on the chemical composition and temperature,^{38,60,71–73} plays a crucial role in controlling the performance of the deformation mechanisms of TWIP steels. It is generally accepted that deformation twining can occur during plastic deformation when SFE is within the range of 20–40 mJ m^− 1.^23,38,74,75 Beyond this range, twinning will be suppressed, and there will be a transition of deformation mechanism to either martensitic transformation plus dislocation slip (SFE < 20 mJ m^− 2)^38,64,72,76 or only dislocation slip (SFE>40 mJ m^− 2).^75,77,78 Within the range of SFE where deformation twins can form, increasing SFE will suppress the twinning kinetics.^63,69,79 Since the quantitative measurement of twin density by experiment is still a challenge, a reliable twinning kinetics model, i.e. twin population vs the plastic strain, does not exist up to now, and the effect of SFE on twinning kinetics has primarily been evaluated qualitatively based on the electron microscopic observation. One explanation for the suppressive effect of high SFE on twinning kinetics is the positive dependence of the critical twinning stress, i.e. the stress at which twinning initiates, on SFE. Various models based on different twinning mechanisms^26,34,46,80 and some phenomenological models^39,42,81,82 have been proposed to formulate the critical twinning stress as a function ofSFE. Yet, none of these models has been directly proven by experiments, probably due to the challenge in detecting the twin initiation during plastic deformation when the twin population is very limited. The SFE is also believed to control the final morphology and substructure of the deformation twins,^22,80,83 yet, more experimental evidences are required to support this proposition. The reader is referred to an extensive review by Christian and Mahajan on the subject of twin nucleation. ¹⁸ In addition, SFE is also an important factor for dislocation kinetics since it determines the dissociation distance of perfect dislocations and thereby the energy barrier for cross-slip.^23,39 A lower SFE should suppress the dislocation annihilation, which leads to a more effective accumulation of dislocations during plastic deformation and higher workhardening rate.^23,29 In addition, SFE also has the effect on the interaction between the carbon atoms and the stacking faults, which may explain the suppression of dynamic strain ageing by the addition of Al. ⁴⁷ For the interaction between partial dislocations and carbides via Orowan mechanism, a higher SFE can effectively increase the critical stress for the bypassing process. ⁷⁰ For TWIP steels, the chemical composition and grain size are two aspects that can be manipulated in the industry to modify the mechanical properties. In general, the chemical composition is designed in order to have sufficient stability of austenite matrix, a reasonable SFE for controlling the deformation behavior and a required YS enhancement by solid solution and precipitation. ¹ The carbon content as an important chemical addition has contradictory effects on twinning. In addition to retard twinning by increasing SFE, the carbon atoms can promote dislocation pile-up and thereby local stress concentration to aid deformation twinning.^68,84 While grain refinement is an effective method to increase the yield strength of TWIP steels, ¹ it can decrease the workhardening rate by suppressing the twinning kinetics. ² This suppressive effect may be explained by the carbon segregation to the grain boundaries, which increase the local SFE and thereby retard the formation of twins. ⁸⁵

Fracture mechanism

The fracture mechanisms of TWIP steels have not yet been intensively studied in the literature, compared to the investigation on the workhardening mechanism. Some recent studies demonstrate that TWIP steels show rather different fracture and damage behaviour, compared to other ferritic and austenitic steels.^1,86,87 Understanding the fracture mechanisms of TWIP steels may be beneficial to avoid sudden slant fracture, unsatisfactory hole expansion performance, ⁸⁸ delayed fracture^89,90 and fatigue failure.^91,92 Twinning induced plasticity steels show, in general, ductile fracture under tension, consisting of nucleation, growth and coalescent of dimples, despite the observation of quasi-cleavage crack due to martensitic transformation. ⁹³ Recently, in situ three-dimensional X-ray tomography experiments have been carried out to study the morphological evolution of dimples during tensile tests of TWIP steels.^86,87,94 These results show that the primary voids nucleate and grow (>2 μm) along tensile direction during uniform deformation.^86,87 However, the volume fraction of these primary voids is < 0·002, ⁸⁷ which is too small to be the main mechanism causing fracture. These primary voids have very low growth rate due to the constant stress triaxiality value.^86,95 Fine dimples ( < 2 μm) are found around large primary voids in the fracture surface, which are regarded as secondary voids. The facture failure during tensile test of TWIP steels could be caused by the sudden intensive nucleation and growth of secondary voids.^96,97 This mechanism could explain why the post-elongation during tensile test of TWIP steel is, in general, very small. Unfortunately, the evolution kinematic of secondary voids is still unclear due to the resolution limitation of X-ray tomography experiments. The nucleation sites for the fine voids could be at the intersections of twins as observed by in situ and ex situ TEM experiments.^98,99 Besides the mechanism of void formation and growth, the other important parameter affecting the fracture behaviour is the fracture toughness, which has been rarely studied in the literature, ¹⁰⁰ which requires further investigation.

Fatigue mechanism

Compared to the extensive studies on the monotonic deformation behaviour, there is a dearth of knowledge on the cyclic deformation behaviour of TWIP steels, and yet, this is essential for the application. It was reported that the TWIP steels exhibit a fatigue limit (defined as the stress amplitude for a fatigue life of 2 × 10 ⁶ cycles) generally around yield strength,^101–104 which is not superior compared to the fatigue property of other austenitic steel grades.^105,106 One possible explanation for this phenomenon should be the absence of deformation twinning during cyclic loading, which, however, is not well understood due to contradictory experimental evidences in the literature. While it was repeatedly reported that no deformation twinning occurs during cyclic loading,^{101,103,104,107} exceptions had been found in two TWIP steels fatigued at a stress amplitude ∼100 MPa higher than the corresponding YS with both deformation twins and stacking faults formed during cyclic loading.^92,106 Furthermore, careful electron backscatter diffraction characterisation revealed that the grains with < 111> orientation paralleled to the loading axis are favoured for deformation twinning during cyclic tensile loading. ⁹² To increase the YS by either prestraining or grain refinement is an effective method to improve the fatigue limit in TWIP steels.^108,109 During the cyclic loading, TWIP steels may display cyclic hardening followed by subsequent softening. ¹⁰⁶ Such behaviour has also been reported in austenitic stainless steels and explained as the increase in total dislocation density followed by rearrangement of dislocations. ¹¹⁰ Microstructure characterisation reveals that the intersections of slip bands, grain boundaries and annealing twin boundaries are the favourable sites for the crack nucleation.^92,101 While cracks initiate relatively early during cyclic loading, i.e. within 20 of the fatigue life, the propagation rate is rather slow, which leads to a long fatigue life. ¹⁰⁶

Summary and perspectives

Some important perspectives emerge from the assessment presented thus far, which we hope defines the way forward. 1

The workhardening behaviour of TWIP steels should, to a significant extent, be controlled by dislocation–twinning interaction, dislocation–dislocation interaction and dislocation–solid atom/precipitate interaction. It is important to distinguish the respective contributions of these three mechanisms on the workhardening rate in order to appreciate the origin of the exceptional strain hardening properties of TWIP steels. This requires a combination of experimental and modelling work. Reliable experimental methods should be developed for the quantitative measurement of the dislocation density and twin volume fraction. In addition, a detailed constitutive model should be built to capture the deformation mechanisms and accurately reproduce the experimental dislocation and twinning kinetics as well as the stress–strain relation. In particular, a physically based twinning kinetics model is sorely missed.

Acknowledgements

MXH acknowledges the financial support from the National Science Foundation of China (grant no. 51301148), Research Grants Council of Hong Kong (project nos. HKU719712E and HKU712713E) and Seed Funding Programme for Basic Research of HKU (grant no. 201409176053). The authors are also grateful to Professor H. K. D. H. Bhadeshia for his invitation and helpful comments.