Tensile Ductility of Nanostructured Bainitic Steels: Influence of Retained Austenite Stability

High silicon (>1.5%) steels with different compositions were isothermally transformed to bainite at 220 and 250 • C to produce what is often referred to as nanostructured bainite. Interrupted tensile tests were carried out and the retained austenite was measured as a function of strain. Results were correlated with tensile ductility. The role of retained austenite stability is remarkably underlined as strongly affecting the propensity to brittle failure, but also the tensile ductility. A simple quantitative relationship is proposed that clearly delimitates the different behaviours (brittle/ductile) and correlates well with the measured ductility. Conclusions are proposed as to the role of retained austenite fraction and the existence of a threshold value associated with tensile rupture.

From a microstructural point of view, the yield and tensile strength of these materials have been shown to be reasonably well correlated to the parameter V β /t β where V β is the volume fraction of bainitic ferrite (the rest normally being retained austenite) and t β the average lath thickness [10,11,17].However, tensile ductility has recently been shown to exhibit largely different values for microstructures exhibiting reasonably similar retained austenite fraction and bainitic ferrite lath thickness, as shown in Table 1, after [12].
Earlier work on the factors controlling the ductility of nanostructured bainite has insisted on the role of retained austenite fraction [18,19] and suggested the existence of an optimum value to achieve the maximum ductility.More recent work, based on measurements of retained austenite content before tensile tests and calculated evolutions, proposed that the stability of retained austenite would influence the material ductility, and indicated that there could be an optimum stability [20].Data from this same publication were later re-interpreted to propose the existence of a percolation mechanism, whereby ductility was imparted by a percolating network of retained austenite in the matrix of bainitic ferrite, and fracture occurred at an approximately constant volume fraction of 10% retained austenite [17].Also recently, a model for stress-assisted martensite formation was proposed by the present authors [21] and was found to provide a reasonable agreement with the experiment, though this approach does not provide indication as to the causes of early tensile failures.
The present work is concerned with further investigating this hypothesis through the use of interrupted tensile tests.This allows actual measurement of the retained austenite content as a function rather than estimated values to be used.

Materials
Three materials were used for the present investigation, with compositions as indicated in Table 2. References indicate both the carbon and silicon content.Both 06C-1.5Siand 1C-2.5Si were produced industrially as ingots, then hot-rolled to 120 mm (0.6C-1.5Si) or 35 mm (1C-2.5Si)bars.The 1C-1.5Si steel was manufactured using a vacuum induction furnace to obtain an approximately 35 kg ingot.After cooling to room temperature, the ingot was re-heated to 1150 • C and forged to a 40 mm bar.Prior to machining, all steels were annealed for 2 h at 700 • C. Chemical composition was determined on the hot-rolled or forged bars using optical emission spectrometry and combustion analysis (LECO).Tensile specimens, 6 or 8 mm in diameter, were manufactured from the hot-rolled bars, using material taken at mid-radius of the latter.They were initially machined with 0.3-0.5 mm additional thickness then heat-treated.The heat-treatments consisted of austenitising in a first salt bath or in a conventional furnace, followed by rapid cooling in a salt bath, to the isothermal transformation temperature.Both austenitising and isothermal transformation parameters varied.Austenitising was carried out for 1 h at temperatures between 860 and 1050 • C, while the temperature for isothermal transformation varied between 220 and 250 • C. The duration for isothermal holding was determined Metals 2017, 7, 31 3 of 7 from measurements in a Baehr dilatometer and varied depending on the material and austenitising.For convenience, relevant heat-treatment parameters are included in the specimen reference.As an example, 1C-2.5Si-1050-250 (16 h) refers to material 1C-2.5Siaustenitised at 1050 • C for one hour and isothermally transformed at 250 • C for 16 h; austenitising duration is not indicated as it was kept constant (1 h).Following heat-treatment, specimens were hard-machined to their final dimensions.
Conventional tensile tests were carried out using three to five specimens.Once yield strength, universal tensile strength and tensile elongation were known, interrupted tensile tests were carried out at selected values of plastic strain in the uniform elongation domain (so as to ensure absence of necking and non-uniform strain distribution in the specimens).Retained austenite measurements were carried out on both the tensile specimens' grip (reference value) and on transverse sections from the gauge length (value after destabilization by plastic strain).
For these experiments, samples were machined, ground and polished with 1 µm diamond paste, and then subjected to several cycles of etching and polishing to obtain an undeformed surface; finally, the samples were polished with colloidal silica.X-ray diffraction measurements were performed by means of a Bruker AXS D8 diffractometer equipped with a Co X-ray tube and Goebel mirror optics to obtain a parallel and monochromatic X-ray beam.Operational parameters and the procedure for obtaining the austenite content and composition are described elsewhere [22,23].

Results
Tensile tests exhibited two different behaviors which are illustrated in Figure 1.For specimens breaking in a brittle manner, both ultimate tensile strength (UTS) and elongation varied significantly (up to 800 MPa for the maximum stress and 4%-5% for the maximum elongation) in the three to five tests carried out on each identical condition, the maximum values were taken.For specimens breaking after necking, the reproducibility was typically within ±15 MPa for the UTS and ±0.7% for the elongation.
Metals 2017, 7, 31 3 of 7 isothermal transformation varied between 220 and 250 °C.The duration for isothermal holding was determined from measurements in a Baehr dilatometer and varied depending on the material and austenitising.For convenience, relevant heat-treatment parameters are included in the specimen reference.As an example, 1C-2.5Si-1050-250 (16 h) refers to material 1C-2.5Siaustenitised at 1050 °C for one hour and isothermally transformed at 250 °C for 16 h; austenitising duration is not indicated as it was kept constant (1 h).Following heat-treatment, specimens were hard-machined to their final dimensions.Conventional tensile tests were carried out using three to five specimens.Once yield strength, universal tensile strength and tensile elongation were known, interrupted tensile tests were carried out at selected values of plastic strain in the uniform elongation domain (so as to ensure absence of necking and non-uniform strain distribution in the specimens).Retained austenite measurements were carried out on both the tensile specimens' grip (reference value) and on transverse sections from the gauge length (value after destabilization by plastic strain).
For these experiments, samples were machined, ground and polished with 1 μm diamond paste, and then subjected to several cycles of etching and polishing to obtain an undeformed surface; finally, the samples were polished with colloidal silica.X-ray diffraction measurements were performed by means of a Bruker AXS D8 diffractometer equipped with a Co X-ray tube and Goebel mirror optics to obtain a parallel and monochromatic X-ray beam.Operational parameters and the procedure for obtaining the austenite content and composition are described elsewhere [22,23].

Results
Tensile tests exhibited two different behaviors which are illustrated in Figure 1.For specimens breaking in a brittle manner, both ultimate tensile strength (UTS) and elongation varied significantly (up to 800 MPa for the maximum stress and 4%-5% for the maximum elongation) in the three to five tests carried out on each identical condition, the maximum values were taken.For specimens breaking after necking, the reproducibility was typically within ±15 MPa for the UTS and ±0.7% for the elongation.Tensile data for all investigated conditions are summarized in Table 3, together with retained austenite content as measured in the grip.As shown, UTS varied between 2.0 and 2.2 GPa, with elongation as high as 17%.As already reported [10,11], such results are exceptional in the combination of strength and ductility that is achieved.It is also clear that retained austenite content Tensile data for all investigated conditions are summarized in Table 3, together with retained austenite content as measured in the grip.As shown, UTS varied between 2.0 and 2.2 GPa, with elongation as high as 17%.As already reported [10,11], such results are exceptional in the combination of strength and ductility that is achieved.It is also clear that retained austenite content alone does not Metals 2017, 7, 31 4 of 7 correlate directly with tensile elongation, as elongations around 5% and above 10% can be found for both ~20% or ~40% retained austenite contents.1), for a plastic strain expressed in %.
In an attempt to quantify the relationship between retained austenite stability and tensile ductility, the results were represented as per the following relationship [24]: where V γ,0 is the initial retained austenite content as measured in the specimen grip, and V γ the retained austenite in the gauge length after application of a true plastic strain of ε p .Some results are illustrated as an example in Figure 2.For all conditions investigated, the value of k was estimated using linear regression (throughout, this value is given for ε p in percent).
alone does not correlate directly with tensile elongation, as elongations around 5% and above 10% can be found for both ~20% or ~40% retained austenite contents.1), for a plastic strain expressed in %.
In an attempt to quantify the relationship between retained austenite stability and tensile ductility, the results were represented as per the following relationship [24]: where Vγ,0 is the initial retained austenite content as measured in the specimen grip, and Vγ the retained austenite in the gauge length after application of a true plastic strain of εp.Some results are illustrated as an example in Figure 2.For all conditions investigated, the value of k was estimated using linear regression (throughout, this value is given for εp in percent).Figure 3 shows the tensile elongation as a function of k for all conditions investigated.It is worth underlining that the corresponding dataset is for three different materials with a variety of Metals 2017, 7, 31 5 of 7 heat-treatments.A first and clear correlation is that relating high values of k (rapid mechanical destabilisation of retained austenite with increasing strain) with brittle behaviour.Indeed, all conditions for which k values of more than ~0.2 were measured, led to brittle fracture during full tensile tests.Inversely, below that threshold, there appears to be a direct correlation between improved mechanical stability (as measured through k) and tensile ductility.
Figure 3 shows the tensile elongation as a function of k for all conditions investigated.It is worth underlining that the corresponding dataset is for three different materials with a variety of heat-treatments.A first and clear correlation is that relating high values of k (rapid mechanical destabilisation of retained austenite with increasing strain) with brittle behaviour.Indeed, all conditions for which k values of more than ~0.2 were measured, led to brittle fracture during full tensile tests.Inversely, below that threshold, there appears to be a direct correlation between improved mechanical stability (as measured through k) and tensile ductility.In addition to measurements carried out on specimens tested within the uniform elongation domain, retained austenite content was sometimes estimated on surfaces directly underneath the rupture surface using the first specimens having undergone full tensile tests.These measurements were associated with a very large texture uncertainty and are therefore to be taken with caution; they were nevertheless frequently below 10% (e.g., 4% for 1C-2.5Si-950-220-22h,5% for 1C-2.5Si-950-250-40h).More reliably, a number of measurements within the uniform elongation domain yielded retained austenite content below 10% (0.6C-1.5Si-890-250-16h, γres 9% for 8% deformation; 0.6C-1.5Si-890-220-22h,γres 5% for 7% deformation).

Discussion and Conclusions
As discussed earlier, it has been proposed that the tensile ductility of bainite formed at low temperatures is correlated with the amount of retained austenite initially available, and limited to a percolation threshold, below which further plastic deformation is no longer possible [17].
The present data provide two important results.First, and as can be seen in Table 3, there is no correlation between initial retained austenite content and tensile ductility.In fact, retained austenite contents of 35%-40% can be associated with "brittle" behavior (1C-2.5Si-950-220-22h,1C-2.5Si-1050-220-40h,1C-1.5Si-950-220-22h)but also with very good tensile ductility (1C-2.5Si-950-250-16h).On the contrary, the approximate quantification of retained austenite stability via k exhibits an excellent correlation with tensile ductility.Furthermore, results suggest the existence of a critical value of k, beyond which brittle behavior cannot be avoided (low retained austenite stability).Below this value, the results suggest a continuous benefit in increasing retained austenite stability to enhance tensile ductility.In particular, the present results do not provide evidence that these microstructures may exhibit "excessive" austenite stability as suggested in earlier publications [22].
Second, the above results do not confirm the suggested existence of a threshold retained austenite content [17], below which ductile deformation is no longer possible.

Discussion and Conclusions
As discussed earlier, it has been proposed that the tensile ductility of bainite formed at low temperatures is correlated with the amount of retained austenite initially available, and limited to a percolation threshold, below which further plastic deformation is no longer possible [17].
The present data provide two important results.First, and as can be seen in Table 3, there is no correlation between initial retained austenite content and tensile ductility.In fact, retained austenite contents of 35%-40% can be associated with "brittle" behavior (1C-2.5Si-950-220-22h,1C-2.5Si-1050-220-40h,1C-1.5Si-950-220-22h)but also with very good tensile ductility (1C-2.5Si-950-250-16h).On the contrary, the approximate quantification of retained austenite stability via k exhibits an excellent correlation with tensile ductility.Furthermore, results suggest the existence of a critical value of k, beyond which brittle behavior cannot be avoided (low retained austenite stability).Below this value, the results suggest a continuous benefit in increasing retained austenite stability to enhance tensile ductility.In particular, the present results do not provide evidence that these microstructures may exhibit "excessive" austenite stability as suggested in earlier publications [22].
Second, the above results do not confirm the suggested existence of a threshold retained austenite content [17], below which ductile deformation is no longer possible.
Interestingly, poor ductility (high k values) was largely associated with transformation at 220 • C, whereas transformation at 250 • C tended to systematically provide ductile behavior (Table 3).
A detailed investigation (SEM, TEM, 3D-APT) of the potential origins of this difference will be published separately.

Figure 1 .
Figure 1.Example of the two different behaviours (brittle and ductile) identified on engineering stress-strain curves.Brittle behaviour may lead to higher maximum stress though the reproducibility is poor.

Figure 1 .
Figure 1.Example of the two different behaviours (brittle and ductile) identified on engineering stress-strain curves.Brittle behaviour may lead to higher maximum stress though the reproducibility is poor.

Figure 2 .
Figure 2. Destabilisation of retained austenite as a function of true plastic strain for two selected conditions.

Figure 2 .
Figure 2. Destabilisation of retained austenite as a function of true plastic strain for two selected conditions.

Figure 3 .
Figure 3. Tensile elongation as a function of the value of k for all conditions investigated.Hollow symbols are for brittle ruptures; full symbols are for ductile ruptures.

Figure 3 .
Figure 3. Tensile elongation as a function of the value of k for all conditions investigated.Hollow symbols are for brittle ruptures; full symbols are for ductile ruptures.

Table 2 .
Chemical composition (wt %) for the three steels used in the present investigation, as determined using optical emission spectrometry and combustion (LECO) analysis.

Table 3 .
Initial retained austenite content and tensile properties for all conditions investigated.The notation for the reference is explained in the text.* indicate brittle behaviour with no true UTS value and variable maximum elongation (the highest value of all tests is then given).
V γ,0 is the retained austenite content as measured in the grip; UTS is Ultimate tensile strength; TE is Total elongation; k is the constant in Equation (

Table 3 .
Initial retained austenite content and tensile properties for all conditions investigated.The notation for the reference is explained in the text.* indicate brittle behaviour with no true UTS value and variable maximum elongation (the highest value of all tests is then given).
Vγ,0 is the retained austenite content as measured in the grip; UTS is Ultimate tensile strength; TE is Total elongation; k is the constant in Equation (