Inﬂuence of partitioning parameters on the mechanical stability of austenite in a Q&P steel: A comparative in-situ study

The transformation-induced plasticity (TRIP)-eﬀect is an eﬃcient way to increase the formability in high perfor- mance steels. Hence, an optimal stability of the retained austenite is crucial to beneﬁt the most from this eﬀect. In the present work, in-situ high energy X-ray diﬀraction was used to study the austenite to martensite transfor- mation upon uniaxial tensile loading of a TRIP-assisted steel produced by the quenching and partitioning (Q&P) process. A detailed analysis of the diﬀraction patterns recorded during deformation allowed to study the austenite stability with respect to the applied partitioning conditions. The austenite stability was found to strongly depend on the applied heat treatment, and could be mainly attributed to the carbon content and to the tempering degree of the surrounding martensitic matrix. Partitioning at 260 °C resulted in a poor austenite stability, while the austenite was almost too stable after partitioning at 360 °C. The optimal combination of strength and ductility was found for partitioning at 400 °C. The micromechanical behavior was analyzed by the evolution of individual lattice strains and the change of full width at half maximum (FWHM). Yielding of austenite could be clearly identiﬁed by an increase of FWHM. Martensite showed an unexpected peak narrowing upon yielding. In the case of 2-step Q&P, austenite started to yield after martensite, while yielding occurred almost simultaneously in the case of 1-step Q&P.


Introduction
Current trends in the automotive industry lean towards an increased use of advanced high strength steels (AHSS) to meet the demands for higher passenger safety and increased energy efficiency. Among these, transformation-induced plasticity (TRIP)-aided sheet steels have received much attention due to their favorable combination of high strength and formability [ 1 , 2 ]. The high formability is achieved by the mechanically-induced austenite to martensite transformation that enhances work hardening and delays the onset of necking [ 3 , 4 ]. The beneficial effect of certain amounts of retained austenite is therefore exploited in all steel concepts of the 3 rd generation AHSS, including medium-Mn, TRIP-aided bainitic ferrite (TBF) and quenching and partitioning (Q&P) steels [5] .
In the case of the Q&P concept, which was first proposed by Speer et al. [6] , retained austenite is embedded in a martensitic matrix. After full or partial austenitization, the martensite forms during rapid quenching to a quenching temperature T q between the martensite start temperature M s and the martensite finish temperature M f (quenching step). During the subsequent partitioning step, which is either carried To benefit the most from the TRIP-effect, not only the quantity of austenite but also its stability is essential. If the stability is too low, austenite rapidly transforms upon loading, but if the stability is too high, no transformation occurs. In both cases, there is no advantageous contribution to enhanced formability. It is well known that there are several factors that affect austenite stability, i.e. carbon content [16][17][18][19][20] , grain size [19][20][21][22] and morphology [ 16 , 21-23 ] of the retained austenite, surrounding phases and their constraining effect [ 24 , 25 ], and the orientation of the austenite with respect to the external loading direction [ 18 , 19 , 21 ]. Advanced neutron and X-ray diffraction techniques combined with tensile testing have already proved to be powerful methods for an in-situ study of both the thermal [ 17 , 20 ] and mechanical stability [ 18 , 19 , 25 , 26 ] of retained austenite. Van Dijk et al. [17] studied the thermal stability of retained austenite in TRIP-steels with different Al and P contents during cooling from room temperature to 100 K. The results showed a lower thermal stability for low carbon austenite. Moreover, compressive stresses induced in the austenite upon final quenching from the bainitic transition temperature seemed to enhance the austenite stability during subsequent cooling. Further studies by Jimenez-Melero et al. [20] confirmed the influence of the austenite carbon content, and additionally revealed a higher thermal stability with decreasing austenite grain volume. Regarding the mechanical stability, Blondé et al. [18] studied a 0.2C-1.5Mn-0.3Si-1.8Al (wt%) TRIPsteel containing a microstructure of ferrite, bainite and austenite. The mechanical austenite stability was found to be mainly influenced by the carbon content in the retained austenite, grain orientation, temperature and load partitioning. Lower testing temperatures enhanced the austenite transformation. Furthermore, austenite grains with a lower carbon content preferentially transformed at the early stages of deformation. These findings are in accordance with results obtained by Yan et al. [26] , who studied the mechanical stability of austenite in a thermomechanically processed TRIP-steel. They concluded that the austenite carbon content is initially important, but other factors, like the influence of neighboring phases and the austenite grain size and morphology, become more relevant with proceeding deformation.
The multi-phase microstructure together with the austenite transformation upon straining lead to a complex macroscopic mechanical behavior of TRIP-assisted steels that is influenced both by the individual phase properties and the interplay between these phases. Understanding all these interactions is a challenging task that can be fruitfully supported by in-situ tensile testing [26][27][28][29][30] . The thus obtained results can serve as a basis for developing micromechanical models and material design studies [31][32][33] . Regarding Q&P steels, Song et al. [27] investigated a 0.2C-1.4Si-2.0Mn (wt%) steel subjected to a Q&P heat treatment with preceding intercritical annealing. Here, also austenite grains containing lower carbon content transform first. Furthermore, load distribution was found to occur from ferrite towards martensite and retained austenite, which facilitates the austenite to martensite transformation. Similar results were, for example, also obtained by Muránsky et al. [30] and Jimenez-Melero et al. [28] . Yielding first starts in the mechanically softer ferritic-bainitic matrix, while retained austenite, that represents the harder phase, can bear a higher load. Recently, Hidalgo et al. [25] investigated the deformation behavior of martensite/austenite microstructures, in which martensite strength was varied by different degrees of tempering. The work hardening behavior of the steel was found to be strongly affected by the strength ratio between austenite and martensite. Stronger martensite tempering lowered the mechanical stability of austenite and altered the austenite yield strength, which resulted in stress partitioning from martensite to austenite.
Although several studies used in-situ tensile testing for the characterization of AHSS, a comparative study concerning the influence of different partitioning conditions has not been exploited so far. Thus, aim of the present work is to provide a comparison of the mechanical stability of austenite in a low carbon steel subjected to four different Q&P heat treatments. The austenite to martensite transformation is investigated by in-situ tensile testing and continuous HEXRD pattern record-  HT260  260  3  260  300  HT360  360  HT400 400 HT400- 30 30 ing. The effect of partitioning temperature and time on the austenite transformation rate is studied and the main influences on the austenite stability are discussed. Furthermore, the micromechanical behavior is investigated by determining the evolution of individual lattice strains. Following up on a recently published work that focused on an in-situ investigation of the phase transformations occurring during Q&P heat treatments [10] , the present study provides insights into the processes taking place during mechanical deformation.

Investigated steel and heat treatments
The investigated material is a 0.2C/1-1.5Si/2.2-2.7Mn (wt%) coldrolled steel with a thickness of 1.6 mm. Steel strips were subjected to different Q&P heat treatments on an annealing simulator MULTIPAS, providing electrical resistance heating and gas jet cooling. All heat treatments comprised full austenitization at 850 °C, followed by an isothermal holding step at 750 °C for 10 s and rapid quenching to a T q of 260 °C for 3 s. In order to study the influence of different partitioning temperatures T p , subsequent partitioning was conducted at 260 °C, 360 °C and 400 °C for 300 s. The influence of partitioning time t p was investigated by an additional heat treatment with partitioning for 30 s at 400 °C. An overview of the four different heat treatments executed in this work can be seen in Table 1 .
The microstructure of the heat treated samples was investigated on a scanning electron microscope (SEM) ZEISS EVO 50 after sample preparation by conventional metallographic methods and Nital etching.
Note that the microstructural evolution and the carbon redistribution process of the heat treatments partitioned for 300 s were investigated by in-situ HEXRD in a recent study [10] . The macroscopic mechanical properties have also been determined therein by conventional tensile testing on a tensile testing machine BETA 250 from Messphysik. The tensile tests were carried out in accordance with DIN EN ISO 6892-1 [34] , using flat tensile test samples with an initial gauge length of 25 mm and the loading direction parallel to the rolling direction. Material with the same chemical composition was used, albeit with a different steel thickness of 1.19 mm.

In-situ tensile testing
The in-situ HEXRD experiments were conducted at the P07 beamline run by the Helmholtz-Zentrum Geesthacht at PETRA III, DESY [35] . Tensile test specimens with the longitudinal axis parallel to the rolling direction were fabricated from the heat treated steel strips. The used sample geometry can be seen in Fig. 1 a. The samples were mounted on an universal testing machine with a maximum load capacity of 20 kN. Uniaxial tensile load was applied at room temperature with a constant crosshead speed of 8.75 × 10 − 3 mm/s until sample fracture. During tensile testing, the applied load and crosshead displacement were recorded. While the applied stress was directly calculated from the applied load, a stiffness correction similar to that described in [36] was carried out for the strain to take into account the machine compliance. The thus obtained stress-strain curves are shown in Fig. 1 b. Note that the applied strain does not coincide with the macroscopic strain of the sample, since it cannot be accurately determined from the crosshead displacement. This also impedes the accurate determination of yield strength (YS) as will be addressed in Section 3.2 .
Complete diffraction patterns were continuously collected upon loading using a Perkin Elmer XRD 1621 fast area detector working in transmission mode. A photon energy of 103 keV ( = 0.1204 Å) was selected. The distance between sample and detector (1342 mm) was calibrated with a LaB 6 standard sample.

Data processing
Each dataset was analyzed from the beginning of tensile testing until reaching ultimate tensile strength (UTS), since subsequent necking occurred outside the area illuminated by the X -ray beam. An example of the obtained detector data showing only the fully recorded Debye Scherer rings is given in Fig. 2 a. Conversion of the 2D diffraction data into 1D diffraction data was carried out using the Data Analysis Work-beNch (DAWN) [ 37 , 38 ]. For all heat treatments, peaks corresponding to austenitic ( ) or ferritic ( ) phase can be clearly identified, as seen exemplary for 230/360 in Fig. 2 b. Peak positions, integrated intensities and full width at half maximum (FWHM) were determined by fitting the individual diffraction peaks with pseudo-Voigt profile functions using the commercial software TOPAS from Bruker AXS. For all analysis, only non-overlapping peaks were considered: (200), (220), (311) -peaks and (200), (211) -peaks. Note that -peaks include the diffraction signal of all bcc phases, i.e. ferrite, bainite and martensite. As will be discussed later in Section 3.1 , small amounts of ferrite and bainite might exist in the initial condition. Furthermore, the diffraction peaks corresponding to martensite formed during deformation overlap with those from preexisting -phase. Separating these contributions is not possible, and thus only a single peak function was considered for -peak fitting.
During tensile testing, the intensity of the austenite reflections gradually reduces (see Fig. 2 b), which reflects the transformation of the metastable austenite into martensite. The amount of austenite was determined from the ratio of the integrated intensities of the -and -peaks as described in [39] . For this calculation, 1D diffraction data integrated over the full azimuthal range was used.
Additionally, the evolution of the lattice strains parallel ( ∥) and perpendicular ( ⊥) to the loading direction (LD) was obtained from 10°segments of the 2D diffraction patterns as indicated in Fig. 2 a. To enhance grain statistics, the results obtained from opposite and thus symmetrical segments were averaged. The lattice spacings d hkl of the individual hkl-planes were calculated from the diffraction angle 2 hkl using Bragg's law where denotes the wavelength of the used X-ray beam. The lattice strains hkl can then be calculated via As can be seen in Fig. 1 b, unstable loading occurred at the beginning of tensile testing, which was probably caused by an insufficient preload for sample holding [40] . Hence, the reference condition d 0 was determined by extrapolation of the d hkl values of the steady-state elastic deformation to an applied load of 0 MPa and the values in the instable regime ( < 150 MPa) were corrected.
The evolution of hkl is not only affected by mechanical loading, but also by the transformation of austenite. During tensile testing, austenite with lower carbon content preferentially transforms into martensite, which results in higher lattice strains due to the ever-increasing carbon content of the remaining austenite. In addition, the volume expansion caused by martensite formation compresses austenite grains oriented in < 220 > direction [41] . To separate these contributions, the strain induced by austenite transformation can be determined by [18] where ⊥ and ∥ are the lattice strains perpendicular and parallel to LD, respectively, and is the Poisson's ratio calculated via The elastic moduli E hkl were obtained from a linear fit of the correspondinghkl curves within the elastic regime.
The (311) planes are reported to closely reflect the macroscopic behavior of austenite [42] . They were therefore chosen for the calculation of the average carbon content in austenite based on the known relationship between austenite lattice parameter a and the chemical composition (in wt%) [17] = 3 . 556 + 0 . 0453 + 0 . 00095 + 0 . 0056 , with Fig. 3 shows SEM images of the heat treated samples. The microstructure primarily consists of martensite, showing a higher degree of tempering for the 2-step cycles, and retained austenite in both blocky and lath form. The occurrence of bainite cannot be excluded, albeit this is not obvious in the SEM images due to the similar etching attack of bainitic ferrite and tempered martensite. Based on the results of a previous study [10] , the amount of bainite formed during the partitioning step is expected to be similar for all heat treatments (5-7%). Thus, disregarding this phase in the further analysis is not supposed to affect the final conclusions. In addition, the lack of additional diffraction peaks indicates that no carbide precipitation occurred during the heat treatments. Even though traces of carbides that cannot be detected by HEXRD might exist in the microstructure, carbides were also excluded in this study.

Table 2
Austenite phase fraction of the initial condition (f 0 ) and at UTS (f UTS ). The initial carbon content (C 0 ) is also given. Errors for the austenite phase fraction and carbon content were estimated as ± 1% and ± 0.01 wt%, respectively. The initial austenite phase fractions were determined as 12% (HT260), 11% (HT360), 14% (HT400) and 13% (HT400-30). The carbon enrichment in the austenite was estimated from the austenite lattice parameter before the onset of tensile testing. The thus obtained values are 0.72 wt% (HT260), 1.09 wt% (HT360), 1.11 wt% (HT400) and 0.99 wt% (HT400-30). These values are also summarized in Table 2 in Section 3.2 . The low carbon content of the HT260 condition can be accounted to negligible carbon diffusion during the partitioning step. Hence, the chemical stabilization of the remaining austenite is low and room temperature stability is assumed to be mainly accomplished by constraining effects of the surrounding martensite. Small amounts of fresh martensite might therefore be present in the microstructure. In contrast, the 2-step Q&P cycles result in a fast and clear carbon enrichment and no formation of fresh martensite upon final quenching is expected. Both the occurrence of fresh martensite for HT260 and the absence of fresh martensite for HT360 and HT400 were confirmed in [10] . Note that the carbon content estimated in this work is higher than that in our preceding study [10] , which can be attributed to the fact that the austenite lattice parameter at room temperature was used for the present calculation. It was recently demonstrated by Allain et al. [9] that this might lead to an overestimation of the austenite carbon content due to thermal eigenstrains arising during final cooling. Compared to our preceding study, the differences in carbon content are 0.44 wt% (HT260), 0.24 wt% (HT360) and 0.21 wt% (HT400) (note that HT400-30 has not been performed within this preceding study). Apart from the fact that the samples in the two studies were heat treated using different devices, which might affect the final microstructure, the higher carbon content measured at room temperature strongly indicates that hydrostatic tensile stresses arose upon final cooling. These stresses were caused by the different coefficients of thermal expansion of martensite and austenite, leading to constrained thermal contraction of austenite. Together with other mechanical contributions, e.g. compressive stresses introduced by the martensite formation, this defines the final stress state in austenite, which is exemplarily calculated for HT360 and HT400 in Section 3.4 .

Macroscopic mechanical behavior and change of austenite phase fraction
The mechanical properties have been determined by conventional tensile testing using different sample geometry [10] . Although the macroscopic strain cannot be exactly derived from crosshead displacement, the principal macroscopic mechanical behavior observed before [10] is well captured by in-situ tensile testing. As can be seen in Fig. 1 b, UTS decreases with increasing T p /t p , which can be attributed to stronger martensite tempering. The respective values are 1493 MPa (HT260), 1219 MPa (HT360), 1163 MPa (HT400) and 1248 MPa (HT400-30). Furthermore, higher T p and prolonged holding lead to enhanced elongation. YS was calculated using the 0.2% offset-method and the obtained values for in-situ and ex-situ tensile testing (given as in-situ/ex-situ) are 977/879 MPa (HT260), 1007/1101 MPa (HT360), 918/1113 MPa (HT400) and 964/1055 MPa (HT400-30), respectively. The discrepancy between these values, ranging from 90 MPa (HT400-30) up to 194 MPa (HT400), emphasizes that YS is not measured in a reliable way during in-situ tensile testing. This affects the further discussion in two ways: First, only a qualitative assertion of the austenite transformed before macroscopic yielding is possible. Second, it is not possible to definitely correlate macroscopic yielding to changes in the lattice strains of the individual phases shown in the next section. Fig. 4 shows the fraction of transformed austenite until reaching UTS for the different heat treatments with the respective YS indicated by dashed lines. It can be clearly seen that the austenite phase fraction decreases with the applied stress for all investigated conditions. For HT260 and HT400-30, austenite transformation starts already before reaching the onset of macroscopic yielding. Subsequently, the transformation continuously proceeds for HT260, while the increase initially occurs less rapid for HT400-30, but accelerates above 1000 MPa. In the case of HT360 and HT400, large-scale austenite to martensite transformation starts above 1150 MPa and 950 MPa, respectively. When reaching UTS, the transformed austenite fraction is the highest for HT260 (70%), followed by HT400 (56%) and HT400-30 (47%). The HT360 condition shows the lowest austenite transformation (31%) and a high quantity of initial austenite (8%) remains untransformed. A summary of the values determined for the initial condition (f 0 ) and at UTS (f UTS ) is given in Table 2 . These results demonstrate the influence of the partitioning parameters on the austenite stability and will be further discussed together with the evolution of the austenite carbon content in Section 3.4 .

Microscopic mechanical behavior
The strain evolution of individual grain families in dependence of the external loading direction is shown in Fig. 5 for all investigated conditions. The change of the lattice strain originates from mechanical loading and the mechanically-induced martensite formation. The latter can be separated from the mechanical contribution using Eq. (3) , and the thus obtained c is shown with open symbols in Fig. 5 . Regarding -phase ( Fig. 5 a-d), c starts to evolve with the onset of transformation. (200) and (311) planes are strained in tension, which reflects the increasing austenite carbon content, while (220) planes tend to increase in compression. As mentioned earlier, this is caused by the volume expansion of the freshly formed martensite parallel to < 220 > [41] . The formation of martensite from higher carbon containing austenite also affects c of martensite ( Fig. 5 e-h). A slight increase in tension is observed for (200) planes, while (211) planes are hardly affected. The increasing lattice strain of (200) indicates that martensite with this orientation preferentially forms from the high carbon austenite. Similar findings were also obtained in [25] .
The lattice strains caused only by mechanical loading ( hkl,corr ) were obtained by subtracting c from hkl and they are presented as closed symbols in Fig. 5 . Grains with their plane normal oriented parallel to LD are in tension and grains with their plane normal oriented perpendicular to LD are in compression. For all conditions, hkl,corr exhibits an initial linear, i.e. elastic, behavior, and the elastic parameters determined for the different (hkl) planes are summarized in Table 3 . Different slopes can be observed for the various (hkl) reflections. For both austenite and martensite, (200) planes are the most compliant. This coincides with the behavior predicted by the elastic anisotropy factor (A hkl ) for cubic structures: A 200, < A 311, < A 220, and A 200, < A 211, [43] . For textured materials, the elastic response of individual lattice planes is known to be sensitive to the preferred grain orientation. However, it can be seen in Fig. 2 a that there is no clear intensity variation along the various diffraction rings, which indicates weak texture and thus negligible effect on the lattice strain behavior in the elastic regime is expected.
In general, entering the plastic regime is expressed by deviation from linearity in the hkl -curves caused by the evolution of intergranular and interphase stresses. The onset of yielding typically involves a lower increase of lattice strain, visible as downward inflection in hkl -curves as they are shown in Fig. 5 . Planes that have not yet yielded retain their linear elastic behavior, but might exhibit an increase of lattice strain caused by load redistribution. This is expressed as upward inflection in hkl -curves. Interestingly, the austenite lattice strains for HT260 in Fig. 5 a continuously increase without showing any decline that would indicate yielding. Moreover, the lattice strains of (200) and (311) shown in Fig. 5 b-d for the 2-step cycles increase again as soon as large-scale austenite to martensite transformation occurs (see Fig. 4 ). Muránsky et al. [30] studied the temperature-dependent transformation behavior of a ferritic-bainitic TRIP-assisted steel. At room temperature, the beginning of the collaborative deformation and austenite transformation coincided with the end of load redistribution between -and -phase, which manifests itself in the austenite strain remaining constant. In the present case, the continuously increasing austenite lattice strains might indicate that load needs to be constantly transferred to the remaining austenite to sustain the austenite to martensite transformation.
Additional information about the yielding behavior can be obtained from the peak profiles. It is generally accepted that an increase of crystal defects associated with plastic deformation results in peak broad-   ening. Fig. 6 shows the change of FWHM obtained for the individual (hkl) reflections of austenite and martensite parallel to LD. For each heat treatment condition, (hkl) reflections of austenite show an increase starting at similar stress levels, which is marked by a line close to the x-axis. The increase initiates slightly below YS for HT260 (663 MPa, Fig. 6 a), and above YS for HT360 (975 MPa, Fig. 6 b), HT400 (911 MPa, Fig. 6 c) and HT400-30 (1000 MPa, Fig. 6 d). The onset of yielding is expected to depend on the austenite carbon content and constraining effects from the surrounding martensite. Hence, a higher carbon content leads to strengthening of the austenite in the 2-step cycles, while the low carbon austenite in HT260 yields at lower applied stress. Furthermore, the slightly lower onset of yielding for austenite in HT400 compared to HT360 and HT400-30 reflects the stronger martensite tempering. Note that peak broadening could be also influenced by the crystallite size, but this effect is assumed to only become important when the amount of transformed austenite considerably increases.
A somewhat surprising behavior was observed for martensite. In contrast to [25] , where only peak broadening was observed as the martensite started to yield, FWHM decreases in the present case, as can be also seen in Fig. 6 . Peak narrowing and the evolution of asymmetric peak profiles have been reported for martensite upon plastic straining [44][45][46][47] . This was associated to a rearrangement and annihilation of dislocations, which reduced the dislocation density [ 44 , 45 ]. On the contrary, Hutchinson et al. [ 46 , 47 ] accounted this phenomenon to a reduction of internal stresses (Type II), which had been introduced during phase transformation. Based on this assumption, they concluded that peak narrowing occurs asymmetrically, because crystallites that contribute to the peak intensity at the high angle side of the peak, i.e. crystallites that are in compression, preferentially deform elastically, and thus the right peak side shifts more than the other does. Assuming that the decrease of FWHM also defines the onset of martensite yielding in the current work, certain martensite regions start to plastically deform at 735 MPa (HT260, Fig. 6 a), 826 MPa (HT360, Fig. 6 b), 774 MPa (HT400,   Fig. 6 d). This roughly coincides with the macroscopic YS and is also marked by a line close to the x-axis in Fig. 5 and 6 . As the deformation proceeds, FWHM increases again, which might be caused by an increasing amount of lattice defects, especially in the freshly formed martensite.
However, it must be noted that the overall interpretation is not straightforward, because (1) (hkl) reflections of austenite represent the average of all -grains with different carbon content and morphology, and (2) (hkl) reflections of martensite represent both initial and freshly formed martensite. It must be further mentioned, that intensity changes along the different diffraction rings of both austenite and martensite have been observed during tensile testing for all heat treated conditions. Albeit not further discussed here, this indicates grain rotations due to austenite transformation and plastic deformation as reported e.g. by Jimenez-Melero et al. [28] .

Change of austenite carbon content and influence of the partitioning condition on austenite stability
Austenite with lower carbon content preferentially transforms into martensite [18] . Thus, the average carbon content in the remaining austenite is assumed to gradually increase, especially at the very beginning of the austenite to martensite transformation. The change of the austenite carbon content ΔC has been estimated from c by combining Eq. (5) and (6) . Fig. 7 shows the evolution of the austenite phase fraction f (upper graph) and the austenite carbon content C (lower graph) upon loading. For all heat treatments, C continuously increases with the austenite phase fraction being reduced.
In the case of HT260, the initial austenite fraction (12%) contains only 0.72 wt% carbon ( Table 2 ). Thus, a low austenite stability is expected, which is indeed reflected in the present in-situ experiments. As can be seen in Fig. 7 , austenite transformation starts already in the elastic regime and steadily proceeds upon loading. Simultaneously, the carbon content of the not yet transformed austenite strongly increases and reaches 1.16 wt% at UTS. The low austenite stability after 1-step Q&P processing was also predicted in previous studies conducted by the present authors [ 10 , 48 ].
A contrary result was found for the HT360 condition. The austenite phase fraction (11%) was slightly lower than for HT260, but the carbon content was remarkably higher (1.09 wt%). This enhanced chemical stabilization was reflected by a delayed onset of austenite transformation compared to HT260. However, the austenite turned out to be too stable and only one third of the initial austenite transformed before reaching UTS. The untransformed austenite (8%) contained a slightly increased carbon content of 1.21 wt%. Consequently, the contribution of the TRIPeffect to the work hardening behavior of the HT360 sample is limited, and therefore this heat treatment is assumed unsuitable for producing Q&P steels with high formability.
The 'optimal' austenite stability to benefit the most from the TRIPeffect and the related formability improvement can be expected if the vast amount of austenite transforms in the plastic region [1] . This requirement was found to be fulfilled the best by the samples partitioned at 400°C. The initial austenite fraction was the highest among all heat treatments investigated and its carbon content was remarkably high (see Table 2 ). About 56% (HT400) and 47% (HT400-30) of the austenite transformed, and similar to HT260 and HT360, an increase of carbon content of the remaining austenite was detected. At UTS, the untransformed austenite contained 1.23 wt% (HT400) and 1.14 wt% (HT400-30) carbon.
Differences between the two conditions partitioned at 400°C arise from the different partitioning times. The more pronounced increase of carbon content and earlier start of martensite formation ( Fig. 7 ) for HT400-30 indicates that the carbon partitioning process is not entirely completed. Worth mentioning is that although partitioning is less proceeded after 30 s and some of the austenite might be quite unstable, the higher amount of transformed austenite is found after partitioning for 300 s. This might be caused by stronger martensite tempering with prolonged holding. Hidalgo et al. [ 24 , 25 ] found that increasing softening of martensite resulting from higher degree of tempering lowered the mechanical stability of the austenite. This was attributed to the increasing capability of the martensite to accommodate the volume expansion occurring during austenite transformation. The high austenite stability for HT360 might also be the result of the less tempered martensite.
Another factor influencing the austenite stability is the internal stress state. Especially compressive stresses are known to be beneficial in stabilizing retained austenite [ 49 , 50 ]. Based on preceding dilatometric studies [10] , final stresses in austenite could be calculated for HT360 and HT420. Applying the procedure described in [51] , compressive stresses of -19 MPa for HT360 and -300 MPa for HT400 were determined. Contrary to expectations, HT360 revealed a lower compressive stress although the austenite stability is the highest of all samples investigated. This suggests that other stabilizing factors might be more important. Note that a calculation of the internal stress state was not possible for HT260 and HT400-30 due to lacking dilatometric data at room temperature.
In total, the evolution of the austenite transformation and the austenite stability observed for the investigated Q&P conditions could be ascribed to the effects of austenite carbon content and the tempering degree of the surrounding martensite. Other effects that might affect the stability of the austenite, i.e. grain size, morphology and distribution of austenite, have not been considered in this work. However, these factors are mainly determined by the quenching step. Since the same quenching conditions were used for the studied heat treatments, these characteristics are not expected to differ greatly, and are thus assumed to contribute only to a small extent to the observed differences in austenite stability. In contrast, carbon content and tempering of martensite are highly influenced by the partitioning conditions, and it is therefore considered reasonable to focus on these factors. Overall, the results undoubtedly confirm the positive contribution of an effective austenite transformation to enhanced work hardening behavior via the TRIP-effect. The most promising combination of strength and ductility was found for the samples partitioned at 400 °C.

Summary
The austenite to martensite transformation and the evolution of the individual lattice strains in a low C steel subjected to four different Q&P heat treatments with varying partitioning conditions have been investigated using in-situ tensile testing under synchrotron radiation. The transformed austenite fraction and the amount of austenite that remained untransformed was found to strongly depend on the selected partitioning temperature. The sample subjected to 1-step Q&P with a high initial austenite fraction (12%) but low carbon content (0.72 wt%) showed the lowest stability. Rapid transformation starting at low applied stresses was observed. In contrast, HT360 condition showed slightly lower austenite content (11%), but remarkably higher carbon content (1.09 wt%). This resulted in a very stable austenite and nearly no austenite transformation. The ideal austenite stability was found for the samples partitioned at 400°C. Initial transformation was low, but successively increased upon straining. Together with the martensite softening due to stronger tempering, this resulted in high elongation.
The change of FHWM was studied in order to determine the onset of yielding for both austenite and martensite. Yielding of austenite was correlated to an increase of FWHM, as it is typically observed for metals due to an increase of crystal defects. In contrast, peak narrowing upon plastic deformation was observed for martensite. This phenomenon has been observed before and was correlated either to a reduction of dislocation density or an elimination of residual internal stresses. The analysis in this study revealed similar behavior for all 2-step cycles. Plastic deformation first started in the martensite followed by austenite. In case of the 1-step cycle, yielding started nearly simultaneously and the lower yield stress of austenite was ascribed to the higher amount of low carbon austenite. With the onset of large-scale austenite to martensite transformation, the lattice strains of (200) and (311) increased again for the 2-step cycles, while a constant increase was observed for the austenite lattice strains of the 1-step cycle. This increase could be explained by the higher load that was transferred to the austenite in order to sustain austenite transformation.

Declaration of Competing Interest
None.