Recent Progress in Understanding the Nano/Micro-Mechanical Behavior of Austenite in Advanced High Strength Steels

: Advanced high strength steels (AHSS) are developed to reduce vehicle weight without sacriﬁcing passenger safety. The newly developed AHSS frequently incorporates the austenite as the intrinsic component with large amount and good stability, which is realized by carefully designed alloying elements and thermo-mechanical processing. To explore the great potential of austenite in enhancing the strain hardening behavior of AHSS, detailed information on the mechanical behavior of single austenite grain is a prerequisite, which can be collected by a small-scale test. The present work reviews the recent progress in understanding the nano/micro-mechanical behavior of austenite in varied AHSS. Three different plasticity modes including dislocation plasticity, martensitic transformation, and deformation twinning can be observed in the austenite grains during small-scale tests, given proper stacking fault energy and crystal orientation. The remaining issues concerned with the nano/micro-mechanical behavior of austenite are discussed. The present review advances the general understanding of the nano/micro-mechanical behavior of austenite grains in AHSS, which may shed light on the precise austenite engineering with the development of new AHSS, realizing the dream of high-performance steels at low cost.


Introduction
The advanced high strength steels (AHSS) are developed to overcome the trade-off between the reduction of vehicle weight and the increase of passenger crashworthiness [1]. The ultrahigh strength of AHSS allows the design of thinner structural components and lighter automobiles while protecting the passengers from the anti-intrusion events [2]. The good ductility of AHSS ensures the large energy absorption during the central collision, the easy fabrication of complex structural components, and the observed deformation before the sudden failure [2].

Nano/Micro-Mechanical Behavior of Austenite
Different from the single dislocation plasticity in ferritic phases, the plastic deformation of the austenite in AHSS during small-scale test is realized by the multiple plastic modes, including the dislocation plasticity, martensitic transformation, and deformation twins. For dislocation plasticity in the face-centered cubic (FCC) austenitic phase, there are 12 available slip systems including the 4 possible <011> slip directions and 3 potential {111} slip planes, where the selection of specific slip systems is governed by the Schmid factor. The operation of other competing plasticity mechanisms in the austenite grains is closely connected to the dislocation activities. For instance, the separation of perfect dislocations into two partial dislocations are connected by a ribbon known as stacking faults (SF), which is the precursor for the initiation of deformation twins [63,64]. The distance of two partial dislocations corresponds to the width of stacking faults which is inversely proportional to the stacking fault energy (SFE). The operation of different plasticity modes in austenite grains depends on the value of SFE, which is governed by deformation temperature and chemical composition [65,66]. According to a thermal dynamic model established for a Fe-Mn-C alloying system [65], the martensitic transformation operates for SFE lower than 18 mJ/m 2 , the deformation twinning occurs for SFE between 12 and 35 mJ/m 2 , and dominant dislocation plasticity takes place for SFE higher than 40 mJ/m 2 [65]. Although the exact value of SFE is varied among different models [65,67,68], these studies can qualitatively describe the correlation between the SFE and the operation of the different plasticity modes. Interestingly, it has been shown that the value of SFE does not continuously increase with the Mn content [67]. There is a minimum value of SFE with respect to the Mn content, beyond which the value of SFE increases with the increase of Mn content. According to the estimation of plasticity modes based on the value of SFE, it is expected that both martensitic transformation and deformation twins can be operative for a SFE in a range of 12-18 mJ/m 2 [65]. Irrespective of the multiple plasticity modes in austenite grains with proper SFE, the dislocation slip is frequently the dominant mechanism for the plastic deformation of the bulk sample, which is partially due to the frequent interaction between the dislocations and boundaries. The nucleation of martensite has been modeled through the arrangement of dislocations and the growth of martensite is facilitated by the motion of glissile transformation dislocations [69]. Nevertheless, the nano/micro-pillars compression test of the austenitic single crystal is dominated by certain plasticity modes other than dislocation slip, such as deformation twins [70], which is affected by the presence of a large surface-to-volume ratio.

Dislocation Plasticity
The retained austenite grains in most of the 3rd generation of AHSS have ultrafine grain size. The mechanical behavior of these single austenite grains can be investigated by the nanoindentation measurement. In general, the nano-mechanical behavior of austenite in AHSS is reflected from the load-displacement (P-h) curve, based on which the different plasticity modes of austenite can be examined. Like the bulk mechanical behavior of austenite, the deformation of the individual austenite grain during the small-scale tests also incorporates the elastic and plastic deformation. The elastic deformation of the austenite grains is frequently evaluated by the Hertzian elastic contact solution as below [71]: where P is the indentation load, E r is the reduced elastic modulus, R i is the tip radius, h is the penetration depth. Note that the tip of the indenter may not be perfectly sharp, and the tip blunting is unavoidable owing to the wear during frequent usage. The tip radius can be estimated from the radius of contact curvature using the atomic force microscopy (AFM) [72] or by fitting to the well-known metallic materials with the obvious strain burst (or pop-in) event and the well-documented elastic properties. Since both specimen and indenter are subjected to the elastic deformation during the nanoindentation test, the generated modulus from the P-h curve is the reduced elastic modulus (E r ) which can be described by elastic properties (E: Young's modulus, v: Poisson's ratio) of the specimen (s) and indenter (i) as below [73]: The typical Hertzian elastic contact solution fitted to the P-h curve of the retained austenite grains in medium Mn steel is shown in Figure 2. The discontinuity (strain burst or pop-in) coincides with the deviation of the P-h curve from the Hertzian elastic contact solution in the load-controlled nanoindentation test ( Figure 2). Note that both strain burst and load drop are observed for the displacement-controlled mode owing to the machine compliance [74].
where P is the indentation load, Er is the reduced elastic modulus, Ri is the tip radius, h is the penetration depth. Note that the tip of the indenter may not be perfectly sharp, and the tip blunting is unavoidable owing to the wear during frequent usage. The tip radius can be estimated from the radius of contact curvature using the atomic force microscopy (AFM) [72] or by fitting to the well-known metallic materials with the obvious strain burst (or pop-in) event and the well-documented elastic properties. Since both specimen and indenter are subjected to the elastic deformation during the nanoindentation test, the generated modulus from the P-h curve is the reduced elastic modulus (Er) which can be described by elastic properties (E: Young's modulus, v: Poisson's ratio) of the specimen (s) and indenter (i) as below [73]: The typical Hertzian elastic contact solution fitted to the P-h curve of the retained austenite grains in medium Mn steel is shown in Figure 2. The discontinuity (strain burst or pop-in) coincides with the deviation of the P-h curve from the Hertzian elastic contact solution in the load-controlled nanoindentation test ( Figure 2). Note that both strain burst and load drop are observed for the displacement-controlled mode owing to the machine compliance [74]. The deviation of the P-h curve of austenite grains from the Hertzian elastic contact solution is in the range of 0.1-0.2 mN. The occurrence of such deviation indicates the initiation of plasticity, which is frequently ascribed to the dislocation nucleation (incipient plasticity). The above conclusion is drawn based on the estimation of the maximum shear stress (τmax) underneath the indenter as below [71]: With the input of the indentation parameters, the τmax can thus be estimated from Equation (3) and it frequently approaches the ideal strength of austenite (G/6~G/30) [75]. Therefore, the dislocations could be nucleated beneath the indenter under the τmax. In other words, the initial pop-ins in P-h curves of austenite may indicate incipient plasticity [76]. Note that not all P-h curves of the austenite grains demonstrate pop-in at the position where the curve deviated from the Hertzian elastic contact solution. The austenite grains in Q&P steel and CFB steel are distributed with the defects resulting from the accommodation process of the displacive shear strain (martensitic or bainitic transformation). The The deviation of the P-h curve of austenite grains from the Hertzian elastic contact solution is in the range of 0.1-0.2 mN. The occurrence of such deviation indicates the initiation of plasticity, which is frequently ascribed to the dislocation nucleation (incipient plasticity). The above conclusion is drawn based on the estimation of the maximum shear stress (τ max ) underneath the indenter as below [71]: With the input of the indentation parameters, the τ max can thus be estimated from Equation (3) and it frequently approaches the ideal strength of austenite (G/6~G/30) [75]. Therefore, the dislocations could be nucleated beneath the indenter under the τ max . In other words, the initial pop-ins in P-h curves of austenite may indicate incipient plasticity [76]. Note that not all P-h curves of the austenite grains demonstrate pop-in at the position where the curve deviated from the Hertzian elastic contact solution. The austenite grains in Q&P steel and CFB steel are distributed with the defects resulting from the accommodation process of the displacive shear strain (martensitic or bainitic transformation). The presence of dislocations in the austenite grains in Q&P steel and CFB steel makes the nucleation of dislocations beneath the indenter unnecessary and thus no obvious pop-in for the initiation of plastic deformation. Nevertheless, the austenite grains in the medium Mn steel could be subjected to recrystallization and thus have lower dislocation density, increasing the possibility of indented volume containing no dislocations and thus the required nucleation of dislocations to initiate the plasticity.
The elastic-plastic deformation of the austenite grains is initiated after the deviation of the P-h curve from the Hertzian elastic contact solution. The plastic part of the P-h curve of the austenite grains can be studied by the GND concept using the following equation [77,78]: where A c = 24.5h 2 is the area projected under the ideal Berkovich indenter, M = 3 is the Taylor factor [77], C = 3 is the constraint factor [77], β = 0.5 describes the dislocation structure [77], b = 0.252 nm is Burgers vector of austenite, ρ SSD and ρ GND are dislocation densities in terms of the statistically stored dislocations (SSD) and GND in the plastic zone, respectively. Considering the small austenite grain size, the maximum indentation depth is generally shallow to avoid the indentation impression across the boundary. Therefore, the ρ SSD can be neglected under the low indentation depth. The ρ GND can be described by the following equation according to the classic Nix-Gao model as [79]: where f describes the proportion of the plastic zone radius (a pz ) to the contact radius (a c ), θ represents the angle between the specimen surface and the indenter. The fitting of the GND concept to the P-h curve of austenite is shown in Figure 2. In general, the overall fitting of the GND concept to the P-h curve can give the value of f, which can be affected by other defects such as the austenite grain boundaries. Considering the ultrafine austenite grain size, the nanoindentation investigation on the austenite grain may involve the influence of grain boundaries. The expansion of the indentation plastic zone with the increase of indentation load may encounter the austenite grain boundary. The high angle grain boundaries can inhibit the dislocation transmission event while the low angle grain boundaries are vulnerable to the transmission of dislocations across the boundaries [80]. The confinement of the dislocations by high angle grain boundaries can limit the indentation plastic zone size, leading to the increased resistance of austenite to penetration of the indenter and thus higher hardness. The single dislocation plasticity in retained austenite grains in AHSS is difficult to be achieved during the nanoindentation test as there are some other competing deformation mechanisms such as martensitic transformation and deformation twins. Considering the formation mechanism of deformation twins, the single dislocation plasticity can be obtained by employing the crystal orientation of the austenite with the favored generation of perfect dislocation [74,81]. Nevertheless, the martensitic transformation is less affected by the orientation as there are 24 or 12 possible martensitic variants according to the Kurdjumov-Sachs (K-S) or Nishiyama-Wasserman (N-W) orientation relationship [82,83]. The suppression of the martensitic transformation during nanoindentation investigation can be only realized through the direct alloying during casting or by partitioning of elements (C, Mn) and dislocation stabilization during thermo-mechanical processing. The single dislocation plasticity in austenite grains in medium Mn steel during nanoindentation investigation allows the capture of the intrinsic nanohardness [84], which is useful to extract the required mechanical properties for the modeling of the AHSS containing the metastable austenite grains. The intrinsic nanohardness of individual austenite grains in other 3rd generations of AHSS has not been reported as the corresponding martensitic transformation during nanoindentation is difficult to evade.
The nano-mechanical behavior of a medium Mn steel (Fe-1.2C-7Mn in wt.%) with a SFE of 28 mJ/m 2 has been investigated by the nanoindentation test [81]. Thus, the formation of deformation twins is expected according to the correlation between the SFE and the deformation twins [85]. Nevertheless, the suppression of the deformation twins can be achieved by the selection of the austenite orientation. The single dislocation plasticity Metals 2021, 11, 1927 8 of 29 is observed for the austenite grains with [111] orientation in a medium Mn steel [81]. TEM observation on the cross-section beneath the indenter reveals the existence of single dislocation plasticity without the formation of deformation twins ( Figure 3). The straight lines as shown in Figure 3a are found to be the perfect dislocations rather than the stacking fault ( Figure 3b). The detailed observation along different directions (g vectors) confirms the absence of deformation twins beneath the indenter (Figure 3c-e). Only perfect dislocation slip is observed in the austenite grain with [111] orientation [81]. Since the deformation twins are nucleated from three-layer stacking faults resulting from partial dislocations, the difficulty of the disassociation of the perfect dislocations into the partial dislocations is the dominant reason for the single dislocation plasticity during nanoindentation test on the austenite grains with [111] orientation.
The nano-mechanical behavior of a medium Mn steel (Fe-1.2C-7Mn in wt.%) with a SFE of 28 mJ/m 2 has been investigated by the nanoindentation test [81]. Thus, the formation of deformation twins is expected according to the correlation between the SFE and the deformation twins [85]. Nevertheless, the suppression of the deformation twins can be achieved by the selection of the austenite orientation. The single dislocation plasticity is observed for the austenite grains with [111] orientation in a medium Mn steel [81]. TEM observation on the cross-section beneath the indenter reveals the existence of single dislocation plasticity without the formation of deformation twins ( Figure 3). The straight lines as shown in Figure 3a are found to be the perfect dislocations rather than the stacking fault ( Figure 3b). The detailed observation along different directions (g vectors) confirms the absence of deformation twins beneath the indenter (Figure 3c-e). Only perfect dislocation slip is observed in the austenite grain with [111] orientation [81]. Since the deformation twins are nucleated from three-layer stacking faults resulting from partial dislocations, the difficulty of the disassociation of the perfect dislocations into the partial dislocations is the dominant reason for the single dislocation plasticity during nanoindentation test on the austenite grains with [111] orientation. The schematic illustration as shown in (c-e) are the tetrahedron demonstrating the orientation of {111} slip plane, and the perfect dislocations on different slip planes [81]. Reprinted with permission from ref. [81]. Copyright 2017, Elsevier.
The dislocation plasticity in the single austenite grains can also be investigated by compression test on the micro-pillars. The beneficial aspect of the compression test is the quantitative analysis of the mechanical behavior with the easy extraction of engineering stress-strain curves. Moreover, the generation of the GNDs can be avoided owing to the compression by flat punch [43]. However, different from the nanoindentation measurement armed with a sharp indenter, the micro-pillar compression test requires relatively large austenite grains to fabricate the micron/nano austenitic pillars. Consequently, most  [81]. Reprinted with permission from ref. [81]. Copyright 2017, Elsevier.
The dislocation plasticity in the single austenite grains can also be investigated by compression test on the micro-pillars. The beneficial aspect of the compression test is the quantitative analysis of the mechanical behavior with the easy extraction of engineering stress-strain curves. Moreover, the generation of the GNDs can be avoided owing to the compression by flat punch [43]. However, different from the nanoindentation measurement armed with a sharp indenter, the micro-pillar compression test requires relatively large austenite grains to fabricate the micron/nano austenitic pillars. Consequently, most of the retained austenite grains in the 3rd generation of AHSS are too small to be suitable for pillar fabrication. Thus, the nano-mechanical behavior of the austenite in the 3rd generation of AHSS is seldom investigated through a pillar compression test [81,86,87]. The size of austenite grains needs to be increased to facilitate the pillar fabrication, which can be realized by high-temperature annealing and the sufficient alloying of austenite stabilizers (Mn, C) as well. The micro-pillar with different diameters can be fabricated by FIB. The avalanche of dislocations after the nucleation leads to jumps (strain bursts) in the stressstrain curves (Figure 4a) [88]. In general, the morphology of the micro-pillar before and after the compression test can be captured by SEM observation. The typical micro-pillar of the austenite along the <111> direction in medium Mn steel after the compression test is shown in Figure 4b. Note that the <111> direction of the austenitic pillar is favorable for the formation of perfect dislocations without the generation of deformation twins, which is consistent with the results observed under the nanoindentation test [81]. Note that this medium Mn steel has a sufficiently high SFE to resist the martensitic transformation (~28 mJ/m 2 ). The formation of perfect dislocations is favored for the austenite with the [111] orientation because a perfect dislocation has smaller critical resolved shear stress as compared to the partial dislocation [81]. The slip traces are evident on the pillar surface as marked with the dashed line in Figure 4c. The arrangement of dislocations in the micro-pillars is inhomogeneous, with intensive dislocations distributed in the upper and lower parts of the deformed pillar ( Figure 4c). The dislocations are confined at the top and bottom parts of the pillar owing to the instrumental constraints, leaving the relatively low dislocation density at the center [89]. The perfect dislocations with straight morphology along the (-1 -1 1) slip plane are observed while they disappeared when observed along the different viewing direction according to the g.b = 0 (Figure 4d-e). The single dislocation plasticity is also found in the micro-pillar compression test of a high Mn austenitic steel (Fe-22Mn-0.6C in wt.%) [90], which could be explained by their comparable SFE among these two steel grades. Therefore, the orientation-dependent plasticity of the austenitic pillar is well confirmed. The dislocation plasticity is dominated over the deformation twins and stacking faults for the [111] crystal orientation.

Martensitic Transformation
The martensitic transformation during plastic deformation is particularly important for the mechanical properties of 3rd generation of AHSS containing metastable austenite grains. The enhancement of strain hardening behavior with the occurrence of martensitic transformation is typically known as the TRIP effect. The performance of martensitic transformation in elevating the work hardening behavior relies on the timing for its occurrence in a single austenite grain, which is usually described by the mechanical stability [23,91]. The understanding of the mechanical stability of single austenite grain helps to design the proper austenite stability for realizing the strength-ductility synergy of AHSS.
Most of the austenite grains in the 3rd generation of AHSS demonstrate the combined plasticity of dislocations and martensitic transformation during the nanoindentation investigation. There are two different pathways for the martensitic transformation, namely γ→α' transformation and γ→ε→α' transformation [69,92]. The γ→α' transformation accompanies a pop-in in the P-h curve ( Figure 5). Such pop-in is also observed for ε→α' transformation [93]. However, no evident pop-in is detected for the γ→ε transformation [93].
Here we mainly focus on the direct γ→α' transformation during the small-scale test.
along the (-1 -1 1) slip plane are observed while they disappeared when observed along the different viewing direction according to the g.b = 0 (Figure 4d-e). The single dislocation plasticity is also found in the micro-pillar compression test of a high Mn austenitic steel (Fe-22Mn-0.6C in wt.%) [90], which could be explained by their comparable SFE among these two steel grades. Therefore, the orientation-dependent plasticity of the austenitic pillar is well confirmed. The dislocation plasticity is dominated over the deformation twins and stacking faults for the [111] crystal orientation.   [81]. Reprinted with permission from ref. [81]. Copyright 2017, Elsevier.

Martensitic Transformation
The martensitic transformation during plastic deformation is particularly important for the mechanical properties of 3rd generation of AHSS containing metastable austenite grains. The enhancement of strain hardening behavior with the occurrence of martensitic transformation is typically known as the TRIP effect. The performance of martensitic transformation in elevating the work hardening behavior relies on the timing for its occurrence in a single austenite grain, which is usually described by the mechanical stability [23,91]. The understanding of the mechanical stability of single austenite grain helps to design the proper austenite stability for realizing the strength-ductility synergy of AHSS.
Most of the austenite grains in the 3rd generation of AHSS demonstrate the combined plasticity of dislocations and martensitic transformation during the nanoindentation investigation. There are two different pathways for the martensitic transformation, namely γ→α' transformation and γ→ε→α'transformation [69,92]. The γ→α' transformation accompanies a pop-in in the P-h curve ( Figure 5). Such pop-in is also observed for ε→α' transformation [93]. However, no evident pop-in is detected for the γ→ε transformation [93]. Here we mainly focus on the direct γ→α' transformation during the small-scale test. In addition to the occurrence of pop-in, the formation of martensite during nanoindentation measurement can be identified from the microstructural change associated with the indents as well. The identification of martensitic transformation during the nanoindentation test can be realized through the comparison of the phase change before and after the indentation test using the electron backscatter diffraction (EBSD) (Figure 6a) [95][96][97]. Considering that the EBSD observation is only sensitive to the transformed martensite close to the free surface and may resultantly miss the martensite underneath the indenter due to the shallow electron penetration depth, the martensite underneath the indenter can be captured by observation on the thin slide perpendicular to the surface of  [94]. Reprinted with permission from ref. [94]. Copyright 2015, Elsevier. (b) The P-h curve of individual austenite grain in Q&P steel (Fe-0.41%C-4.0%Mn-1.6%Si-1.0%Cr in wt.%) [28]. The arrows in (a) and (b) mark the pop-in. The insets in both (a) and (b) show the position of the indents. γ: austenite; α: ferrite; α': martensite. Reprinted with permission from ref. [28]. Copyright 2019, Elsevier.
In addition to the occurrence of pop-in, the formation of martensite during nanoindentation measurement can be identified from the microstructural change associated with the indents as well. The identification of martensitic transformation during the nanoindentation test can be realized through the comparison of the phase change before and after the indentation test using the electron backscatter diffraction (EBSD) (Figure 6a) [95][96][97]. Considering that the EBSD observation is only sensitive to the transformed martensite close to the free surface and may resultantly miss the martensite underneath the indenter due to the shallow electron penetration depth, the martensite underneath the indenter can be captured by observation on the thin slide perpendicular to the surface of indented austenite grain prepared by FIB thinning and lift-out strategy [94,98,99]. However, the preparation of the TEM sample made by using the FIB technique may involve the phase transformation induced by radiation [100,101] or by the creation of a free surface [102]. It is reported that magnetic force microscopy (MFM) can verify the boundaries between different magnetic domains (α is ferromagnetic and γ is paramagnetic) [103,104]. Therefore, the change of localized ferromagnetism beneath the indenter can be detected with the generation of possible domain patterns by using MFM imaging (Figure 6b) [105]. Like the EBSD measurement, the MFM can be carried out in a relatively easy and fast way to verify many indents, which is beneficial for statistical analysis [103]. In general, most of the transformed martensite is close to the indenter. Nevertheless, the expansion of the plastic zone may encompass the grain boundaries where the martensite embryo may exist. Consequently, the region close to the grain boundaries could also transform to martensite (Figure 6c).
Metals 2021, 11, x FOR PEER REVIEW 11 of 29 [102]. It is reported that magnetic force microscopy (MFM) can verify the boundaries between different magnetic domains (α is ferromagnetic and γ is paramagnetic) [103,104]. Therefore, the change of localized ferromagnetism beneath the indenter can be detected with the generation of possible domain patterns by using MFM imaging (Figure 6b) [105]. Like the EBSD measurement, the MFM can be carried out in a relatively easy and fast way to verify many indents, which is beneficial for statistical analysis [103]. In general, most of the transformed martensite is close to the indenter. Nevertheless, the expansion of the plastic zone may encompass the grain boundaries where the martensite embryo may exist. Consequently, the region close to the grain boundaries could also transform to martensite ( Figure 6c). Figure 6. The detection of the martensitic transformation during the nanoindentation test using different techniques, including the (a) EBSD measurement [106], Reprinted with permission from ref. [106]. Copyright 2020, Elsevier; (b) MFM [105], Reprinted with permission from ref. [105]. Copyright 2019, Elsevier; and (c) FIB-TEM/EBSD [94], Reprinted with permission from ref. [94]. Copyright 2015, Elsevier. (a-1) EBSD phase maps before and (a-2) after indentation test. Red color, yellow color and blue color in (a) represents the martensite, austenite, and cementite respectively. The region beneath the indenter contains the compressive hydrostatic core which is surrounded by the hemispherical plastic zone according to the theoretical consideration [107]. Since the martensitic transformation is dominated by the shear deformation (~0.2) with a minor dilatational exponent (~0.04) [108], the compressive hydrostatic pressure does not interact with the shear strain and shall suppress the volume expansion of Figure 6. The detection of the martensitic transformation during the nanoindentation test using different techniques, including the (a) EBSD measurement [106], Reprinted with permission from ref. [106]. Copyright 2020, Elsevier; (b) MFM [105], Reprinted with permission from ref. [105]. Copyright 2019, Elsevier; and (c) FIB-TEM/EBSD [94], Reprinted with permission from ref. [94]. Copyright 2015, Elsevier. (a-1) EBSD phase maps before and (a-2) after indentation test. Red color, yellow color and blue color in (a) represents the martensite, austenite, and cementite respectively. The region beneath the indenter contains the compressive hydrostatic core which is surrounded by the hemispherical plastic zone according to the theoretical consideration [107]. Since the martensitic transformation is dominated by the shear deformation (~0.2) with a minor dilatational exponent (~0.04) [108], the compressive hydrostatic pressure does not interact with the shear strain and shall suppress the volume expansion of martensitic transformation. Therefore, the initiation of martensite shall not be from the hydrostatic core but rather from the plastic zone. In addition, the strain burst represents the geometrical softening, which is inconsistent with the hardening effect of martensite. Therefore, the discontinuity in the P-h curve of indented austenite grains with martensitic transformation is unexpected and should be well explained. The occurrence of the pop-in is ascribed to the martensitic transformation by considering the variant selection during the nanoindentation loading process [94]. There are 24 or 12 possible martensite variants during the nanoindentation test for classic K-S or N-W relation, respectively. To simplify the discussion, the Bain relation is employed to qualitatively explain the strain burst in the P-h curve of metastable austenite grain [99]. The Bain model of martensitic transformation indicates that the martensitic variants accompany one contracted axis and two other expanded axes (Figure 7). The mechanical interaction energy (U), which is a product of the applied external stress (σ ij D ) and the stress-free transformation strain (ε ij T ) [109], assist the nucleation of martensite, and can be rewritten as [110]: where the invariant-plane strain ε ij T can be decomposed into normal ε n T and shear ε sh T components. It can be obtained from the phenomenological theory of martensite crystallography (PTMC) [111]. σ n D and σ sh D are the normal and shear stresses perpendicular and parallel to a habit plane (invariant plane) of a martensitic variant, respectively.
Metals 2021, 11, x FOR PEER REVIEW 12 of 29 martensitic transformation. Therefore, the initiation of martensite shall not be from the hydrostatic core but rather from the plastic zone. In addition, the strain burst represents the geometrical softening, which is inconsistent with the hardening effect of martensite. Therefore, the discontinuity in the P-h curve of indented austenite grains with martensitic transformation is unexpected and should be well explained. The occurrence of the pop-in is ascribed to the martensitic transformation by considering the variant selection during the nanoindentation loading process [94]. There are 24 or 12 possible martensite variants during the nanoindentation test for classic K-S or N-W relation, respectively. To simplify the discussion, the Bain relation is employed to qualitatively explain the strain burst in the P-h curve of metastable austenite grain [99]. The Bain model of martensitic transformation indicates that the martensitic variants accompany one contracted axis and two other expanded axes (Figure 7). The mechanical interaction energy (U), which is a product of the applied external stress (σij D ) and the stress-free transformation strain (εij T ) [109], assist the nucleation of martensite, and can be rewritten as [110]: where the invariant-plane strain εij T can be decomposed into normal εn T and shear εsh T components. It can be obtained from the phenomenological theory of martensite crystallography (PTMC) [111]. σn D and σsh D are the normal and shear stresses perpendicular and parallel to a habit plane (invariant plane) of a martensitic variant, respectively.  [99]. σex: externally applied stress. Black circle and red dot represent the austenite and martensite atoms, respectively. Reprinted with permission from ref. [99]. Copyright 2010, Elsevier.
The energy barrier for the nucleation of martensite is dominated by the strain energy which should be overcome by the external stress in terms of the mechanical interaction energy. The mechanical interaction energy is varied for different martensitic variants. The martensitic variant with the highest mechanical interaction energy will be selected beneath the indenter [94]. Despite the complex stress field underneath the indenter, the  [99]. σ ex : externally applied stress. Black circle and red dot represent the austenite and martensite atoms, respectively. Reprinted with permission from ref. [99]. Copyright 2010, Elsevier.
The energy barrier for the nucleation of martensite is dominated by the strain energy which should be overcome by the external stress in terms of the mechanical interaction energy. The mechanical interaction energy is varied for different martensitic variants. The martensitic variant with the highest mechanical interaction energy will be selected beneath the indenter [94]. Despite the complex stress field underneath the indenter, the dominant stress state can be considered as compressive [99]. Thus, the martensitic variant with a contraction axis along the indentation direction is selected during the nanoindentation test to maximize the mechanical interaction energy [99]. The velocity for the propagation of martensite is estimated to be around 160-200 m/s in a grain size of 250 µm of a 304 austenitic stainless steel based on the acoustic-emission-frequency spectrum analysis [112]. Note that the propagation velocity is proportional to the austenite grain size. Therefore, it is expected that the propagation velocity is around 0.64-0.8 m/s for austenite with a grain size of 1 µm, which is still too large to be detected by the indenter. Under such circumstances, the indenter tip must have to move further to ensure a fixed loading rate in load-controlled mode, generating a strain burst [99] or reducing the applied load to guarantee a constant strain rate in displacement-controlled mode, inducing either load drop or strain burst for different loading modes [113]. Therefore, the magnitude of pop-in may correspond to the size/volume of the transformed martensite. However, it is difficult to rationalize the magnitude of pop-in with the amount of the transformed martensite as it shall also relate to the orientation of transformed martensite. Considering the phenomenological model of martensitic transformation, the martensitic variant that effectively accommodates the external stress is selected during the nanoindentation test, which is confirmed from the finite element calculation [94].
Among the multiple discontinuities (pop-ins) in the P-h curve, the first pop-in that deviates from the Hertzian contact solution may be induced by the dislocation nucleation [99]. The formation of martensite also initiates a pop-in and mostly takes place after the first one [98]. Therefore, the load at which the second pop-in takes place may represent the critical applied load for initiation of the martensitic transformation [98]. Based on this understanding, the second pop-in load has a positive correlation with the Mn content of austenite grains, which is consistent with the fact that Mn is an austenite stabilizer [98]. The mechanical stability of retained austenite is enhanced by carbon enrichment owing to the carbon partitioning during the tempering treatment [114]. The effect of carbon content on the mechanical stability of austenite is larger than that of the fraction and morphology of austenite in bearing steel [115]. Based on the nanoindentation investigation on the constituting phases, it is reported that the large lenticular martensite is not effective in C partitioning into the adjacent austenitic matrix in a medium Mn steel [116]. The grain size-dependent austenite stability is investigated by the nanoindentation test. The P-h curve of the coarse-grained sample (~12 µm) shows pop-in while no observable pop-in is found in that of the ultrafine-grained sample (~0.65 µm) in 304 L stainless steel [117,118]. However, the conclusions could be questionable as the pop-in event could also be related to the size of the martensite transformed beneath the indenter. In general, the small austenite grain size should also lead to the small martensite, which may not be discernible in the P-h curve. Therefore, a detailed observation of the microstructure beneath the indenter for the UFG specimen is necessary. The stress-induced martensite is found to be ahead of the dislocation plasticity of the transformed martensite in the austenite grains of a high carbon quenched and tempered steel [119]. The decrease of austenite grain size increases the required stress for both martensitic transformation and plastic deformation of transformed martensite, suggesting the small austenite grain has higher mechanical stability than the coarse counterpart [119]. The grain size-dependent nano-mechanical behavior of austenite is investigated with dominated mechanical twinning in nano/ultrafine-grained specimen while the prevalent strain-induced martensite nucleated at the shear bands in the coarse-grained sample during nanoindentation test [120]. The stability of the austenite grains relies on the grain orientation as has been demonstrated in 18CrNiMo7-6 steel [106]. The retained austenite grain with [001] orientation is much less stable as compared to the one with [111] orientation, which is verified by comparing the transformed martensite fraction beneath the indenter [106].
Most of the nanoindentation investigations on the austenite stability involve applying a load beyond the elastic deformation regime [94,98,99]. The formation of martensite in these studies is termed the strain-induced martensitic transformation [99]. However, the stress-induced martensitic transformation is less investigated. The difference between the stress and strain-induced martensite is in their different martensitic nucleation processes [121]. The stress-induced martensite has the same nucleation site as compared to the martensite formed during the quenching process. In contrast, strain-induced martensite refers to the nucleation of martensite at the site created during the plastic deformation process (i.e., intersection of slip bands) [122]. The stress-induced martensite is investigated by applying an ultralow load within the elastic deformation regime. The martensitic transformation takes place adjacent to the annealing twin boundaries. The applied elastic stress facilitates the martensitic transformation through the mechanical interaction energy to overcome the nucleation barrier [95]. The nano-mechanical behavior of individual austenite grain is affected by the geometry of the indenter. The nanoindentation with the cube cone indenter favors mechanical twinning while the Berkovich indenter generates a single martensitic transformation, which is rationalized through the sharper indenter (smaller centerline-to-face-angle) induces a higher shear stress for the generation of deformation twins [97]. The above experimental observation can also be explained by the fact that the sharper indenter corresponds to the higher representative strain and thus this may induce the different mechanical responses of austenite during the nanoindentation test.
The blocky retained austenite exhibits intermediate nanohardness values, which is lower than high-carbon martensite but is larger than bainite in bainitic steels [123]. Nevertheless, it is found that carbon enriched austenite demonstrates a nanohardness comparable to the primary low carbon martensite in Q&P steel [28]. The high nanohardness of the austenite grains is contributed by the martensitic transformation beneath the indenter through either the plastic deformation of the martensite or the dispersion hardening [98], depending on the distance of the transformed martensite to the indenter. It is found that the nanohardness of deformation-induced martensite in CrMnNi steel with a high alloying element is only 24% higher than the parent austenitic matrix [124]. Furthermore, it is concluded that the strain hardening from the martensitic transformation is induced by the dynamic Hall-Petch effect owing to the presence of intensive boundaries from transformed martensite rather than by the "intrinsic" hardness from the deformation of martensite. Nevertheless, the above conclusion is debatable in that the parent austenitic matrix could also be strengthened by the underneath martensite (dispersion hardening), making it slightly softer than the transformed martensite.
The nano-mechanical behavior of the austenite at different temperatures is investigated by the molecular dynamics (MD) simulation (Figure 8a,b) [113]. It is found that the martensitic transformation dominates over the whole austenite grain during nanoindentation investigation at a deformation temperature of 100 K (Figure 8c). The increase of deformation temperature reduces the amount of transformed martensite which is mainly confined to the area surrounding the indenter (Figure 8b,c). The remaining austenite is deformed by dislocation plasticity. No deformation twinning is observed for the warmtemperature deformation of austenite (500 K) (Figure 8c). Although the molecular dynamics investigation is effective to reveal the deformation behavior at different temperatures, the MD simulation involves a very high strain rate (20 m/s) [113], which cannot be achieved by the conventional loading rate of the nanoindentation test. Moreover, the austenite grain boundary is fixed, which limits transformation at the area close to the fixed boundary (Figure 8c).  The nano-mechanical response of austenite during the micro-pillar compression test is different from that of the bulk counterpart. The martensite transformed in the micronsized austenitic pillars in medium Mn steel (Fe-9Mn-0.6C in wt.%) has a lower dislocation density as compared to the martensite in bulk samples, which can be explained through the large surface-to-volume ratio of micro-pillar [126]. The shape strain of martensite transformed in a micro-pillar is well accommodated by the free surface without deforming the surrounding austenitic matrix and the martensite itself. The nucleation of martensite in the micron-pillar is at the intersected site of two stacking faults bundles [126]. It has been estimated that only a single nucleation site is pre-existed in bulk austenite with a grain size of 100 μm [127]. Therefore, the micron-pillar may lack the nucleation site of martensite owing to its small pillar size. Thus, the formation of the martensite nucleation site through the plastic deformation is a prerequisite for the commencement of martensitic transformation, i.e., intersected stacking faults [126].
The martensitic transformation is found to be orientation-dependent during the compression test of the austenitic micro-pillars pillar in medium Mn steel (Fe-9Mn-0.6C in wt.%) [86,87]. The martensitic transformation can be identified in the austenitic micropillars with the orientation of [100] and [110]. Nevertheless, the [100] pillars demonstrate a far larger magnitude of strain burst than the [110] ones (Figure 9a), which can be ascribed to the much higher volume fraction of transformed martensite in [100] pillars considering the general correlation between the martensitic transformation and strain burst [86]. The [100] micro-pillars (~483 MPa) demonstrate a lower average critical stress as compared to that in [100] micro-pillars (~700 MPa) [87]. The deformation twins are observed in the deformed [110] micro-pillars but not in the deformed [100] micro-pillars (Figure 9b,c). The critical stress to initiate the deformation twins in the [100] micro-pillars is estimated to be around ~465 MPa, which is approaching the stress for generating martensite. Considering the similar theoretical stress for initiating the martensite in both micro-pillars (650-750 MPa), the reduced critical stress for triggering the martensitic transformation in [100] micro-pillars as compared to [100] micro-pillars could be ascribed to the twin-assisted nucleation of martensite at the intersections of two twin bundles (Figure 9d) [87]. Consequently, the orientation-dependent martensitic transformation observed during the micro-pillar The micro-pillar compression experiments can also be used to investigate the martensitic transformation in metastable retained austenite grains in AHSS. It is found that the martensitic transformation does not closely relate to the pop-ins during compression tests of nanoplate observed by in situ TEM [125]. The pop-in is explained instead through the generation of slip band and the geometrical softening of the nanoplate [125]. In contrast, most of the micro-pillar compression investigations suggest that the obvious strain burst is triggered by the formation of martensite [86,87,126]. The magnitude of strain burst is comparable to the calculated strain based on the estimation of the martensite area fraction normalized along the compression direction [126]. The discrepancy between these studies should be verified by further well-controlled experiments. For instance, the austenitic pillars with large diameters are desirable to avoid artifact such as buckling during the in situ compression test.
The nano-mechanical response of austenite during the micro-pillar compression test is different from that of the bulk counterpart. The martensite transformed in the micronsized austenitic pillars in medium Mn steel (Fe-9Mn-0.6C in wt.%) has a lower dislocation density as compared to the martensite in bulk samples, which can be explained through the large surface-to-volume ratio of micro-pillar [126]. The shape strain of martensite transformed in a micro-pillar is well accommodated by the free surface without deforming the surrounding austenitic matrix and the martensite itself. The nucleation of martensite in the micron-pillar is at the intersected site of two stacking faults bundles [126]. It has been estimated that only a single nucleation site is pre-existed in bulk austenite with a grain size of 100 µm [127]. Therefore, the micron-pillar may lack the nucleation site of martensite owing to its small pillar size. Thus, the formation of the martensite nucleation site through the plastic deformation is a prerequisite for the commencement of martensitic transformation, i.e., intersected stacking faults [126].
The martensitic transformation is found to be orientation-dependent during the compression test of the austenitic micro-pillars pillar in medium Mn steel (Fe-9Mn-0.6C in wt.%) [86,87]. The martensitic transformation can be identified in the austenitic micropillars with the orientation of [100] and [110]. Nevertheless, the [100] pillars demonstrate a far larger magnitude of strain burst than the [110] ones (Figure 9a), which can be ascribed to the much higher volume fraction of transformed martensite in [100] pillars considering the general correlation between the martensitic transformation and strain burst [86]. The [100] micro-pillars (~483 MPa) demonstrate a lower average critical stress as compared to that in [100] micro-pillars (~700 MPa) [87]. The deformation twins are observed in the deformed [110] micro-pillars but not in the deformed [100] micro-pillars (Figure 9b,c). The critical stress to initiate the deformation twins in the [100] micro-pillars is estimated to be around~465 MPa, which is approaching the stress for generating martensite. Considering the similar theoretical stress for initiating the martensite in both micro-pillars (650-750 MPa), the reduced critical stress for triggering the martensitic transformation in [100] micro-pillars as compared to [100] micro-pillars could be ascribed to the twinassisted nucleation of martensite at the intersections of two twin bundles (Figure 9d) [87]. Consequently, the orientation-dependent martensitic transformation observed during the micro-pillar compression test is ascribed to the different transformation pathways, including the twin-assisted and free-standing martensitic transformation (Figure 9e) [87].  . TEM images show the microstructure in deformed austenitic pillars with the orientation of (b) [100] and (c) [110]. The nucleation process of the martensite at the intersection of the stacking faults/twins in the micron-pillars with the orientation of [100]. (d) The calculated critical stress for triggering the deformation twins and martensite. (e) Schematic illustration of the different transformation pathways in the micro-pillars with varied orientations [86,87]. Reprinted with permission from ref. [87]. Copyright 2019, Elsevier.

Deformation Twins
The coherent twin boundaries are highly symmetric grain boundaries that can largely enhance the strength of metallic materials without sacrificing ductility [128,129]. However, the possibility for the operation of deformation twins in the austenite grains of CFB steel and the Q&P steel is generally low. This is because the formation of deformation twins requires low SFE (20-40 mJ/m −2 ) for the dissociation of perfect dislocations [85] and most of the austenite grains in CFB steel and Q&P steel does not have sufficient Mn and C content to satisfy the required SFE. However, the deformation twins can be occasionally found in the medium Mn steel with elevated Mn/C content [17,18,130,131]. In particular, the deformation twins in the high Mn steel such as TWIP steel are prevalent and are believed to be responsible for the high work hardening rate, which is generally known as the TWIP effect [132,133]. The formation of deformation twins enhances the work-hardening behavior by altering the dislocation structures [134]. The glissile dislocations in the matrix become sessile dislocations within the twin, accompanying with change of Burgers vector. For polycrystalline materials, the twin boundaries can serve as barriers to the dislocation slip, decreasing the dislocation mean free path and leading to the intensive interaction between dense dislocations and deformation twins [132,135,136]. Nevertheless, it is reported that the individual contribution of the deformation twins to the flow stress of TWIP steel is small as compared to the forest dislocation hardening, indicating the secondary role of deformation twins in TWIP steel [62]. Furthermore, it is argued that the  [100] and (c) [110]. The nucleation process of the martensite at the intersection of the stacking faults/twins in the micron-pillars with the orientation of [100]. (d) The calculated critical stress for triggering the deformation twins and martensite. (e) Schematic illustration of the different transformation pathways in the micro-pillars with varied orientations [86,87]. Reprinted with permission from ref. [87]. Copyright 2019, Elsevier.

Deformation Twins
The coherent twin boundaries are highly symmetric grain boundaries that can largely enhance the strength of metallic materials without sacrificing ductility [128,129]. However, the possibility for the operation of deformation twins in the austenite grains of CFB steel and the Q&P steel is generally low. This is because the formation of deformation twins requires low SFE (20-40 mJ/m −2 ) for the dissociation of perfect dislocations [85] and most of the austenite grains in CFB steel and Q&P steel does not have sufficient Mn and C content to satisfy the required SFE. However, the deformation twins can be occasionally found in the medium Mn steel with elevated Mn/C content [17,18,130,131]. In particular, the deformation twins in the high Mn steel such as TWIP steel are prevalent and are believed to be responsible for the high work hardening rate, which is generally known as the TWIP effect [132,133]. The formation of deformation twins enhances the workhardening behavior by altering the dislocation structures [134]. The glissile dislocations in the matrix become sessile dislocations within the twin, accompanying with change of Burgers vector. For polycrystalline materials, the twin boundaries can serve as barriers to the dislocation slip, decreasing the dislocation mean free path and leading to the intensive interaction between dense dislocations and deformation twins [132,135,136]. Nevertheless, it is reported that the individual contribution of the deformation twins to the flow stress of TWIP steel is small as compared to the forest dislocation hardening, indicating the secondary role of deformation twins in TWIP steel [62]. Furthermore, it is argued that the interstitial atoms such as C atoms rather than the deformation twins play a crucial role in improving the work hardening of TWIP steel [137,138]. Note that the origin of enhanced work hardening behavior in TWIP steel with carbon atoms has already been discussed around 30 years ago [139]. The key to answering whether the TWIP effect is important for the TWIP steel is to reveal the exact resistance of single twin boundary to dislocation motions and the role of twinning arrangements on the strength.
The perfect dislocations in austenite grains with proper SFE (20-40 mJ/m −2 ) will be energetically favorable for dissociation into two partial dislocations, including leading partial dislocations and trailing partial dislocations. These two partial dislocations glide to different distances, connected by a ribbon known as the stacking fault, which is considered as a precursor for the generation of deformation twins [63,64]. Nevertheless, only the proper SFE is insufficient for deformation twins as they also depend on the crystal orientation. It is found that the deformation twins are favored in a single austenitic pillar with [001] orientation in a medium Mn steel (Fe-1.2C-7.0Mn in wt.%) under both nanoindentation and compression tests [81]. In contrast, only perfect dislocations are observed in the austenite grains with [111] orientation [81]. Therefore, the formation of mechanical twins is governed by the orientations which are closely related to the Schmid factor for twinning in fcc crystals [140,141].
The generation of deformation twins beneath the indenter is believed to be responsible for the observed large strain bursts during the nanoindentation test [81]. However, such belief has not been unambiguously verified. Considering the scenario that the simultaneous occurrence of the dislocation and deformation twins in a single indent while only one pop-in is detected in the P-h curve, it is a challenge to determine whether the yielding of single austenitic grain with a pop-in is induced by deformation twinning or dislocation glide [140]. Another possibility is that the plastic yielding of the [001] grain could be induced by both slip and twinning owing to their similar critical resolved shear stresses. However, this possibility has not been confirmed in a previous study [140].
In addition to the nanoindentation test, the formation of deformation twins can also be investigated by a micro-pillar compression test with a typical flow curve as shown in Figure 10(a-1) [142]. The deformed micro-pillar is sheared along a single direction as depicted by SEM observation (Figure 10(a-2)). To reveal the underlying mechanism for such deformed morphology, the cross-section of the micro-pillar is fabricated by FIB and observed under TEM (Figure 10(a-3)). Interestingly, the deformed morphology coincides with the arrangement of the large deformation twins (Figure 10(a-2,a-3)), suggesting that the pillar morphology is induced by the generation of deformation twins. In addition to the large deformation twins with thickness of around 1 µm, profuse nanometer-sized deformation twins can also be observed in the micro-pillar compression test ( Figure  10b) [142]. Nevertheless, a small number of twins with large thickness has also been observed in the micro-pillar compression test of TWIP steel in another study ( Figure  10c) [70]. The formation of deformation twins with varied thickness may be due to the competition between the nucleation and growth processes of deformation twins. surface in terms of vacancy Frank loops induced by ion radiation. The dissociation of such loops beneath the surface may assist the nucleation of twins. The dissociation of dislocation into partial dislocations at the free surface provides the nucleation of twins. It is found that the pre-existing perfect dislocations within the pillar are also essential for the generation of deformation twins [144]. The growth of deformation twins in pillars can be ascribed to the glide of Shockley partial dislocations (SPDs) on successive consecutive {111} planes from the pillar surface [142]. . Reprinted with permission from ref. [74]. Copyright 2017, Elsevier. (c) The engineering stress-strain curve and deformation twins in another study [70]. Reprinted with permission from ref. [70]. Copyright 2015, Elsevier. . Reprinted with permission from ref. [74]. Copyright 2017, Elsevier. (c) The engineering stress-strain curve and deformation twins in another study [70]. Reprinted with permission from ref. [70]. Copyright 2015, Elsevier.
The micron-pillar compression test of the single austenitic crystal shows that there is an obvious strain burst in the flow curve ( Figure 10(a-3),c). The contribution of the deformation twins to the engineering strain can be estimated as below [70]: where ε t is the strain induced by deformation twins, T is the overall thickness of deformation twins, γ t is the strain sheared by deformation twinning which is 0.707 [143], and α is the deviation angle of active twinning direction from the pillar axis. H is the altitude of the micro-pillar before the compression test. The engineering strain estimated based on the above equation is similar to the observed strain burst in the engineering stress-strain curve, suggesting that the strain burst could be induced by the occurrence of deformation twins [70]. In particular, the deformation twins are carried out by the initiation of the layerby-layer shearing because all partial dislocations for a twinning glide on the paralleled planes with the same Burgers vector [81]. This may be the fundamental reason for the occurrence of large strain bursts during the generation of deformation twins. A physically based model is developed by Liang and Huang [70], which takes the different sources and activation criteria of partial dislocations into consideration. This model captures the key feature of the flow curve including the twinning burst and also the twin evolution such as twin fraction and morphology, quantifying the contribution of the formation of deformation twins to the engineering strain ( Figure 11) [70].   The simulated stress-strain curve at different engineering strains together with the corresponding morphology of deformation twins [70]. The red color in (b-d) represents the deformation twins. Reprinted with permission from ref. [70]. Copyright 2015, Elsevier.
The nucleation of deformation twinning in austenitic micro-pillars during compression test is different from the bulk counterparts. There are different mechanisms for nucleation of deformation twinning during the compression test of austenitic micro-pillar in TWIP steel. The partial dislocations emitted from the pillar surface are found to be responsible for the growth of deformation twins (Figure 12a) [142]. In contrast, it is found that the interaction between dislocations is a prerequisite for generation of twin with the partial dislocations emitted from dislocation intersections (Figure 12b) [90]. Interestingly, the nucleation of twins could also be assisted by the defects created during the preparation of pillars at the ion-milling step [144]. The FIB techniques can induce damage to the pillar surface in terms of vacancy Frank loops induced by ion radiation. The dissociation of such loops beneath the surface may assist the nucleation of twins. The dissociation of dislocation into partial dislocations at the free surface provides the nucleation of twins. It is found that the pre-existing perfect dislocations within the pillar are also essential for the generation of deformation twins [144]. The growth of deformation twins in pillars can be ascribed to the glide of Shockley partial dislocations (SPDs) on successive consecutive {111} planes from the pillar surface [142]. Figure 11. (a) The comparison between the simulated stress-strain curves and the one obtained by experiments. The inset shows the formation of twins in the austenitic pillar after the compression test. (b-d) The simulated stress-strain curve at different engineering strains together with the corresponding morphology of deformation twins [70]. The red color in (b-d) represents the deformation twins. Reprinted with permission from ref. [70]. Copyright 2015, Elsevier.   [74]. Reprinted with permission from ref. [74]. Copyright 2017, Elsevier.
The compression of micro-pillar in metallic materials with dominant dislocation plasticity demonstrates a size effect that is increased critical resolved shear stress with decreased diameter, which can be explained by the dislocation associated mechanisms such as dislocation starvation [145,146] and limited dislocation source size [147,148]. Nevertheless, the size effect can also be observed for the austenitic single crystal with the operation of the deformation twins, namely the stress for the twinning process increases with the decrease of pillar diameter ( Figure 13) [144]. In particular, the yield stress (initiation of first strain burst) substantially increases with decreased pillar diameter ranging from 3 µm down to 0.5 µm (Figure 13), following a power-law relation with a smaller exponent (0.43) as compared to the one induced by dislocation activities (0.6-0.7) [144]. This is consistent with the nucleation mechanism of twinning from the surface dislocation emission. The deformation twins are found to be the dominant plastic mode in the pillars, suggesting that the size effect should be controlled by deformation twins on the aspect of twin nucleation and growth rather than by dislocation activity [144]. The preparation of the micro-pillar by FIB technique involves the ion bombardment which will induce a damaged layer incorporating the dislocation loops and other defects on the sample surface [149,150]. The dislocation loops induced by ion-radiation demonstrate a radius of several tens of nanometers, which can be activated at smaller stress as compared to the theoretical limit, suggesting that the introduction of dislocations by ion-radiation can alternatively serve as dislocation sources to compensate the limited pre-existing dislocation sources within the small sample volume for initiating plasticity [151]. The ion radiation can generate plastic deformation through point defects such as vacancy and embedded ions, inducing the local volume expansion, and producing the compressive residual stress state due to the mismatch between the irradiated and non-irradiated parts, which could help to dissociate the vacancy Frank loops for initiation of deformation twins [144]. The compressive stress induced by ion-radiation in the subsurface layer is found to decrease with an increase of pillar diameter, leading to the observed size effect for deformation twins during the compression test [144].
limit, suggesting that the introduction of dislocations by ion-radiation can alternatively serve as dislocation sources to compensate the limited pre-existing dislocation sources within the small sample volume for initiating plasticity [151]. The ion radiation can generate plastic deformation through point defects such as vacancy and embedded ions, inducing the local volume expansion, and producing the compressive residual stress state due to the mismatch between the irradiated and non-irradiated parts, which could help to dissociate the vacancy Frank loops for initiation of deformation twins [144]. The compressive stress induced by ion-radiation in the subsurface layer is found to decrease with an increase of pillar diameter, leading to the observed size effect for deformation twins during the compression test [144]. The effect of the twin boundaries on the strength of steel can be investigated by employing the pillar compression test. The twin boundaries are introduced and incorporated in the pillar with a diameter of 4 μm [90]. The pillar with twin boundaries demonstrates slightly larger and more jerky flow stresses than the pillar without twin boundaries, which The effect of the twin boundaries on the strength of steel can be investigated by employing the pillar compression test. The twin boundaries are introduced and incorporated in the pillar with a diameter of 4 µm [90]. The pillar with twin boundaries demonstrates slightly larger and more jerky flow stresses than the pillar without twin boundaries, which is ascribed to the operation of secondary slip systems [90]. However, the comparison of the mechanical response between the small-volume single crystal and the polycrystalline bulk counterpart is inappropriate as the former plasticity is affected by the existence of free surfaces. The dislocations in the deformed micro-pillars can easily glide to the free surface with no storage of dense dislocations within the pillar, which is different from the mechanical behavior of the polycrystalline bulk material.
The plastic mode of the austenite including the pure dislocation plasticity and deformation twinning during the micro-pillar compression test depends on the crystal orientation [81]. It is found that the critical resolved shear stress of micro-pillars along different orientations favorable for either dislocation slip and deformation twins is similar ( Figure 13) [90]. Based on this observation, it is further proposed that the initiation of plasticity in micro-pillars of TWIP steel is governed by the dislocation activity rather than by deformation twins, which is controversial with Liang's work [144]. The formation of ε-martensite is observed in the deformed micro-pillars with deformation twins. Such ε-martensite is absent in bulk TWIP steel (Fe-22Mn-0.6C in wt.%) [90]. This observation may be explained if we consider the effect of ion damage with gallium implantation which is a ferrite stabilizer [100].
It is reported that the critical stress for initiation of deformation twins is higher than the one for perfect dislocation slip [74]. The plastic deformation of pillar oriented for deformation twins is initially carried out with generation and glide of leading partial dislocations. The partial dislocation transmits into twins and its emission from the intersected site of planar defects take place with the increase of plastic deformation, inducing the formation of double deformation twins [74]. The double-twin is the one that the stacking of the twin is rearranged into the original austenite matrix. The thin double-twins are observed within the thicker twins along with the primary slip system in the deformed micro-pillars. The dislocation interactions among the primary and conjugate slip systems can generate a sessile dislocation configuration known as the Lomer-Cottrell locks for inhibiting dislocation motion [152]. The nucleation of the deformation twin can be derived from the interaction between partial dislocations and Lomer-Cottrell locks [153]. In contrast, the planar dislocation glide is dominated in the micro-pillars with favorable orientation for dislocation plasticity. However, planar dislocation glide is mainly confined in the area close to the top part of along the primary slip plane, resulting in localized plastic deformation [74].

Concluding Remarks
Continuous efforts in understanding the nano/micro-mechanical behavior of austenite are beneficial for the alloy design, service behavior of small-volume materials in the industry such as MEMS, and the thorough capture of deformation mechanism of bulk materials. The nanoindentation measurement is a useful technique for extracting the bulk mechanical properties without destroying the specimen. The engineering stress-strain curves of individual constituting phases in AHSS can be extracted from the single small-scale test based on the nanoindentation test combined with microhardness measurements using the inverse method or by numerical methods and finite element analyses, which properly compare with the mechanical response obtained from the tensile tests [154,155]. The cyclic nanoindentation test has been employed to investigate the mechanical behavior of the austenitic stainless steel [156][157][158], demonstrating an interesting reverse transformation that does not exist in the macroscopic tensile test. However, such cyclic nanoindentation test has seldom been reported in the retained austenite grains in the AHSS, which requires further investigation. The nanohardness is calculated from the P-h curve by using the Oliver-Pharr method which relies on the precise detection of the indentation depth [73]. However, the nanohardness of austenite is strongly affected by the martensitic transformation. As has been discussed in Section 3.2, the metastable austenite grains during the nanoindentation test are heavily subjected to the martensitic transformation with the occurrence of strain burst which represents the geometrical softening, while the subsequent deformation of martensite contributes to the hardening of austenite [98,99]. The occurrence of the pop-in or the martensitic transformation makes the hardness closely dependent on the maximum depth, austenite stability, martensite size, and variants. The nano-mechanical behavior of metastable austenite grains in quenched and tempered steel with high carbon content exhibits double plastic deformation stages, including the initial martensite transformation followed by deformation of transformed martensite [119]. Therefore, it is quite challenging to decouple the contribution of transformed martensite to the calculation of nanohardness of the metastable austenite grains. In addition, the elastic deformation of the austenite during the unloading process is also affected by the formation of martensite, making the calculation of the intrinsic elastic modulus of austenite complex. One solution to obtain the inherent properties of the austenite grains is to inhibit the formation of martensite during the nanoindentation test [84]. It has been shown by MD simulation that the austenite grain can be stabilized by the high temperature in which the transformed volume of martensite is reduced [113]. It is reported that the nanoindentation test can be carried out at high temperatures with the reliable characterization of deformation behavior [159]. Therefore, it may be possible to obtain the nanohardness of the austenite at the different high temperatures without martensitic transformation and use these data to extrapolate to the nanohardness at room temperature. However, such nano-mechanical behavior of the austenite grains at high temperatures is less investigated in the open literature. Since the SFE of austenite grains increases with deformation temperature, the martensitic transformation may change to the deformation twinning, and then to the dominated dislocation plasticity encompassing the planar slip, cross-slip, and climb of dislocations.
The intrinsic austenite stability can be precisely revealed by the nanoindentation test. The formation of martensite is frequently accompanied by a strain burst after the initiation of dislocation plasticity. The load at which the second strain burst takes place is considered as the critical load for the formation of martensite and thus can be used to measure the stability of austenite. However, it is noted that such pop-in load for martensitic transformation is difficult to be used for comparison among different experiments. The popin stress is more useful than the pop-in load for comparison purposes. Such pop-in stress is difficult to be obtained owing to the complex stress field underneath the tip. Moreover, the nucleation of the martensite beneath the indenter is influenced by the interaction between the dislocations and the embryo of martensite at the austenite grain boundary. Such interaction depends on the indentation depth and the thickness of the indented austenite grains. The intrinsic stability of austenite grains is governed by the defect (dislocation) density, and chemical compositions (C/Mn), all of which can be investigated by nanoindentation test. It is noted that the austenite grains with filmy morphology are more stable than the blocky counterparts in Q&P steel during a tensile test, which is ascribed to the presence of hydrostatic pressure and higher defect density. In addition, the chemical compositions in the austenite grains with different morphologies could be different. Since there are different governing factors in affecting the austenite stability simultaneously, it is thus unclear whether the morphology of austenite can intrinsically affect the austenite stability, rather than coupling with other factors. It is known that the morphology of austenite grains in the rolled samples along the different observation directions is different. In other words, the granular morphology of austenite grains along the rolling direction may demonstrate the lamellar morphology along the transverse direction. By employing the nano-mechanical behavior of the austenite grains with different orientations, the statistical sense of the austenite stability with different morphology can be verified.
The tailored austenite stability beneficially enhances the work hardening behavior of AHSS containing the metastable austenite grains. The austenite grains with different mechanical stability can avoid either the scenario for the exhaustion of the austenite grains at the early deformation process or the remaining of the substantial amount of austenite after the necking process, both of which are undesirable for the enhancement of mechanical properties. The different austenite mechanical stability can be realized through tuning the governing factors such as the chemical compositions, grain size/morphology, and defect density. The MD simulation on the transformation behavior of the austenite during the indentation test suggests that the transformed amount of martensite largely depends on the deformation temperature [84]. Since the deformation temperature affects the austenite stability in a manner that can be equivalent to the chemical compositions, it is hypothesized that the single austenite grains with the varied chemical partitioning can be useful for the controlled release of TRIP effect, continuously improving the work hardening behaviors of AHSS. The chemical inhomogeneity such as Mn content has been proved to be effective for the development of high-performance medium Mn steel processed by splash heat treatment [5]. The Mn banding, which is generated during the solidification process, can also be used to tailor the austenite stability across a larger scale than the splash heat treatment [160].
In addition to the dislocation and martensitic transformation, the deformation twins are an important plasticity mode for the austenite grains with proper SFE in the AHSS. However, the impact of deformation twins on the work hardening of the TWIP steel is still under debate. The classic viewpoint on the deformation twins indicates that they play a critical role in enabling the high strength and good ductility of TWIP steel, which is typically referred to as the dynamic Hall-Petch effect by separating the austenite into smaller grains [132,133,135,136]. Nevertheless, the calculation of the overall dislocation density and the estimation of the contribution of dislocations to the flow stress by the classical Taylor hardening law suggest that the dislocations mostly contribute to the flow stress of TWIP steel [62]. Consequently, the newly developed viewpoint suggests that the interstitial carbon atoms rather than the deformation twins are critically important for the enhanced work hardening response of TWIP steel [137,138]. However, the presence of the deformation twins can also increase the dislocation density as they act as barriers for the dislocation slip. Therefore, the respective contribution from the deformation twins to the flow stress can be incorporated in the part contributed by the increased overall dislocation density. In other words, the contribution made by deformation twins to the flow stress is indirect through the increased dislocation density and it should be decoupled from the total increased dislocation density. The quantitative estimation of the single twin boundaries to the flow stress is important to identify if the engineering of such coherent twin boundary to the flow stress is effective. The resistance of the twin boundary to the dislocation glide should be weaker than the grain boundary [38]. From the aspect of alloy design, the ultimate question is about how to increase the deformation twin density and the optimization of the twin arrangement as well [161]. The contribution of the single twin boundaries to the flow stress has been estimated by using the small-scale test of a micropillar with a diameter of 4 µm [162]. However, it is difficult to correlate the results obtained from the small-scale test to the bulk mechanical behavior. For instance, the mechanical behavior of micro-pillar and nano-pillar during the compression test is largely affected by the presence of a free surface. This dilemma may be resolved by fabricating the pillar with relatively large diameter to capture the bulk mechanical behavior of TWIP steel. It has been shown that the size-dependent plasticity can be disregarded for pillar diameter larger than 7.6 µm based on a theoretical consideration of TWIP steel [90]. The proper comparison of the mechanical behavior of the large pillars with and without the deformation twins may be useful to verify the importance of the deformation twins to the strength of TWIP steel. Data Availability Statement: No data, models, or code were generated or used during the study.

Conflicts of Interest:
On behalf of all authors, the corresponding author states that there is no conflict of interest.