A further study of the kinetics of recrystallization and grain growth of cold rolled TWIP steel

Fernando de las Cuevas^a,b,*, Claudio Aguilar^c, Javier Gil Sevillano^d

^aUPNA (Universidad Pública de Navarra), Campus de Arrosadia, Dpto. de Ciencias, 31006 Pamplona, Spain

^bSIEMENS GAMESA Renewable Energy, S.A, Avda. Ciudad de la Innovación 2, 31621 Sarriguren, Navarra, Spain

^cUSM (Universidad Técnica Federico Santa María), Dpto. de Ingeniería Metalúrgica y Materiales, Av. España 1680, Valparaíso, Chile

^dCEIT y TECNUN (Universidad de Navarra), M. Lardizábal 15, 20018 San Sebastián, Spain

(*Corresponding Author: fernando.delascuevas@unavarra.es)

1. INTRODUCTIONTOP

Over last decades, the twinning - induced plasticity Fe-Mn-C (TWIP) steels have been the focus on huge amount of research works due to their prominent strength – ductility compounding which develops from the occurrence of extended mechanical twinning during plastic deformation under mechanical loads (Grässel and Frommeyer, 1998; Frommeyer et al., 2000; Cornette et al., 2005; Scott et al., 2006; Bouaziz et al., 2008; Hamada et al., 2010; Bouaziz et al., 2011; De Cooman et al., 2011; Galán et al., 2012; Gil Sevillano and De las Cuevas, 2012; Chen et al., 2013; De las Cuevas et al., 2014; Ghasri-Khouzani and McDermid, 2015; Pierce et al., 2015; De las Cuevas and Gil Sevillano, 2017).

In TWIP steels, the fully austenitic microstructure can be retained by means of high level alloying with elements such as Mn, Al and Si. Al and Si are mainly used to adjust the magnitude of the stacking fault energy, g_SFE, of austenite (Frommeyer et al., 2000). Furthermore, they also strengthen the steel by solid solution hardening and stabilize austenite owing to their ability of slowing down the precipitation of carbides, especially cementite, leaving more carbon available for the enrichment of austenite (Leslie and Rauch, 1978).

Additions of Al generally increase g_SFE, whilst Si is reported to lower it at concentrations approximately ≥ 4 wt%, Schramm and Reed (1975), but to increase it at lower concentrations (Dumay et al., 2008). Then, by adjusting the chemical composition and controlling the carbon concentration, the austenitic structure is maintaned; the emergence of a´- martensite (BCC) and e– martensite (HCP) are inhibited and can conduct to prominent mechanical properties by TWIP effect, (Grässel et al., 1997; Grässel et al., 2000; Allain et al., 2004), leading to a dynamic Hall-Petch effect as the deformation proceed (De las Cuevas et al., 2010a). It is widely recognized that the TWIP effect occurs in a stable austenite with g_SFE approximately of 25 mJ·m⁻² (Frommeyer and Grässel, 1998). More recent research works on this relevant point, Pierce et al. (2014) has thrown more detailed aspects of the deformation mechanism, phase stability and stacking fault energy in TWIP steels.

In a previous work developed by De las Cuevas et al. (2010b), the static recrystallization and grain growth kinetics in a 22% Mn, 0.6% C (mass %) TWIP steel were studied. The present research work aims to obtain further data of the isothermal annealing behavior in TWIP steel to formulate any industrial application of this material since the desired compounding of elastic limit and tensile strength required for a specific application involves an unchanging austenitic structure with a preset grain size. Therefore, a thorough characterization of the precipitation in the interval of temperature 600 °C ≤ T ≤ 900 °C and its arguable effect in the static recrystallization and grain growth kinetics were studied in detail. With all the data obtained, a refined characterisation of the kinetic equation of recrystallization and grain growth for 60% cold rolled TWIP steel was performed to obtain the activation energy for both recrystallization and grain growth. Finally, textural evolution via X-ray of isothermal treatments of cold rolled TWIP steel at different stages: recrystallized structure (without apparent grain growth), and recrystallized plus grain growth with different grain sizes structures was characterized.

2. MATERIAL AND EXPERIMENTAL PROCEDURESTOP

Hot rolled samples (5.4 mm thick) of TWIP steel with composition depicted in Table 1 were reduced by cold rolling at laboratory mill to different thicknesses, R: 3.24 mm (R = 40%), 2.70 mm (R = 50%), 2.16 mm (R = 60%) and 1.60 (R = 70%) and subsequently isothermally annealed. The annealing treatments were performed in a salt bath furnace in the interval of temperature 450 °C ≤ T ≤ 900 °C and they were interrupted by quenching water after different annealing times ranging 1 s ≤ t ≤ 43740 s. In a similar manner annealing treatments at 1000 ºC and 1100 ºC were carried out in a resistance furnace under Ar protective atmosphere in order to avoid decarburation of the steel, De las Cuevas and Gil Sevillano (2017), followed by water quenching too. The heating time was controlled by treating dummy samples with inserted thermocouples. The evolution of recrystallisation and grain growth was followed by control of the softening kinetics complemented by metallographic and EBSP-OIM observations. The samples were analysed using a Philips XL30 SEM microscope equipped with a TSL module for automatic EBSP acquisition. The grain sizes of the specimens were measured as the mean linear intercept using OIM-EBSP images. Further experimental points are described (De las Cuevas et al., 2010b). These annealing treatments provided equiaxed grain sizes after complete recrystallization and grain growth in the interval of 1.50 μm ≤ D < 50 μm.

The development of macro-texture was performed by X-ray diffraction in a Philips Xpert diffractometer. The selected specimens to study the evolution of macro-texture were: cold rolled 60%, recrystallized structure (neglected grain growth) with 1.50 ± 0.02 μm of grain size, and recrystallized plus grain growth structures with 12.2 ± 0.3 μm and 35 ± 1 μm.

In order to detect fine precipitates in the annealed samples, single stage carbon extraction replicas were prepared using the conventional method starting from surfaces prepared as for optical microscopy. Afterwards a selective electrolytic dissolution method to remove the matrix around the carbides was used (Marder, 1989). The extracted carbon replicas were examined using a Jeol JEM 200CX (STEM) electron microscope operated at 200 kV.

3. RESULTSTOP

3.1. Softening kineticsTOP

Owing to the industrial importance of 60% reduction, thorough characterization of the isothermal annealing behaviour after such reduction was carried out. Figure 1 shows the mean value of hardness (HV) versus annealing time (t) for reduction R = 60% TWIP cold rolled specimens. The dashed black line corresponds to the Vickers Hardness at 60% cold rolled structure, HV_CR = 506 ± 3 kgf•mm⁻². The softening kinetics was very similar for the four reductions.

First of all, the effect of annealing at 450 ºC was very weak. There was a small initial hardening after the shortest annealing. The shortest annealing corresponded to 1 s, but in fact it took about 8 s to reach the 450 ºC temperature. Thereafter an almost negligible softening took place. A weaker hardening at the start of the annealing was only perceptible at 600 ºC or 650 ºC. This initial hardening did not appear anymore for higher treatment temperatures. This stage of static strain ageing (SSA), is well known attributable to solid solution segregation at pre-existing dislocation lines (Vidoz et al., 1963). Figure 2 illustrates the transient static strain ageing stage of initial hardening (∆HV_i) at 450 ºC and 600 ºC as a function of level reduction.

At 600 ºC and 650 ºC the hardness evolution with time pointed typicallly the sigmoidal shape of recrystallisation softening. At high times some contribution of grain growth will also take place, but microstructural characterisation shows that grain growth is very slow at such temperatures. For 600 ºC, recrystallization is not completed for the maximum time verified (12.5 hours), the fraction recrystallized being known from the metallographic observations, [X_sof(t = 12.5 h)]_600ºC ≈ 87% as it is depictied in Fig. 3.

At 700 ºC the hardness evolution with time showed less clearly the sigmoidal shape of recrystallisation softening. Judging by the final hardness values obtained after 12.15 h at 600 ºC and 700 ºC, the influence of the cold rolling reduction (from 40% to 70%) in the recrystallised grain size was small. Figure 4 illustrates such small influence at both temperatures. In addition, it was confirmed by metallographic examination, that the grain sizes of this steel after static recrystallisation are very small, in agreement with the results reported (Scott et al., 2006; De las Cuevas et al., 2010b).

Finally, from 800 ºC to 1100 ºC, the softening curves in all their time range correspond to the lower arm of their sigmoidal shape and, in fact, to grain growth. Hardness shows an approximately exponential decrease with time. This meant that in less than one second, recrystallisation was complete at T ≥ 800 ºC. These results were complemented with microstructural observations that confirm this recrystallization and grain growth behaviour.

3.2. Precipitation during isothermal annealing treatments of cold rolled TWIP steelTOP

The effect of annealing treatments on the precipitation behavior of 60% cold rolled TWIP steel was study in the interval of temperatures 600 °C ≤ T ≤ 900 °C at long soaking time, approaching equilibrium conditions. In Table 2 are reported the annealing treatments.

3.2.1. Annealing in the range 800 °C ≤ T ≤ 900 °CTOP

In Fig. 5 are shown the typical microstructures found for cold rolled 60% TWIP steel after soaking at 800 °C and 900 ºC. As it can be noted no precipitation was observed by SEM analysis. Only oxide particles and MnS were detected. In addition, in the thermodynamic study, the isothermal section of Fe-Mn-C system performed by Thermo-Calc (TCFe6 database) under the defined conditions of temperature at equilibrium did not reveal the precipitation of carbides.

3.2.2. Annealing in the range 600 °C ≤ T ≤ 700 °CTOP

In Fig. 6a is shown a SEM image of carbides precipitated on the austenitic grain boundaries of cold rolled 60% TWIP steel after a soaking at 700 °C. The semi-quantitative analysis SEM - EDS gave a C content of about 7%wt consistent with carbide of cementite type. In a similar manner, in Fig. 6b is shown carbide precipitated on austenitic grain boundaries at 600 °C.

- Extraction replica analysis by TEM

Owing to the detection of carbide precipitation via SEM–EDS analyses, in the cold rolled samples subjected to annealing treatments at 600 °C and 700 °C further investigation was performed by using TEM observations. The replica extractions were carried out on both TWIP steel samples soaked at 600 °C and 700 °C. The precipitation features in terms of type of precipitates and chemical composition were the same. The remarkable difference was in terms of amount of precipitates since at 700 °C the number of precipitates was significantly lower than at 600 °C. The types of precipitates were: (Fe, Mn)₃C – cementite and Vanadium carbo-nitrides. The presence of Vanadium carbo-nitrides is due to unwished presence of V in ferroalloy during ingot casting. In Fig. 7 is depicted a Vanadium carbo-nitride image characterized by fine size (< 50 nm) and its relevant diffraction pattern having a FCC structure.

In a similar manner, in Fig. 8 is reported a (Fe, Mn)₃C – cementite with the relevant diffraction pattern having an orthorhombic crystal structure.

Finally, the ternary system Fe-Mn-C was assessed by means of Thermo-calc (TCFe6 database) under the defined conditions: at temperature below 700 ºC and for Mn content in the range 16% ≤ Mn %wt ≤ 22% and C content 0.4% ≤ C %wt ≤ 0.6% the most stable carbide is (Fe, Mn)₃C cementite type with C content about 7%.

3.3. Textural evolution of isothermal treatments of cold rolled TWIP steelTOP

In order to assess the textural evolution of cold rolled 60% TWIP steel in both recrystallized and recrystallized plus grain growth stages, X-ray macro-texture measurements were performed. Figure 9 illustrates the ODFs φ₂sections (45º, 65º and 90º) of 60% cold rolled structure and recrystallized structure (without apparent grain growth). Figure 10 shows the same in recrystallized plus grain growth structure of TWIP steel for two grain sizes of structure. Recrystallized structure corresponds to 1.50 ± 0.02 μm of grain size. Recrystallized plus grain growth structures correspond to 12.2 ± 0.3 μm and 35 ± 1 μm.

After completion of recrystallization, Fig. 9b, the main components of the rolling texture (Fig. 9a), Brass, Goss and S, remain but with much weaker intensity levels. However, grain growth induces progressive qualitative texture changes. The intensity of the texture is weak, with maxima of about 2.5 random but the Brass and Goss components disappear and, different components built a new, more complicated texture, as it is evident in Fig. 10. This behaviour was observed before, De las Cuevas et al. (2010b), for cold rolled TWIP steel 22%Mn-0.6%C of smaller grain size (until ~ 9 μm) and other TWIP steel compositions: (I-600-011 (22.27%Mn-0.19%Si-0.5%C-0.011%N) and L-500-081 (22.7%Mn-0.21%Si-0.5%C-0.081%N)), Bracke et al. (2009), and Fe-30Mn-3Si-3Al, Vercammen et al. (2004) with similar grain growth structure.

4. DISCUSSIONTOP

4.1. Modelling of recrystallization behavior of TWIP steelTOP

Due to the EBSP-OIM results together with the evolution of hardness with the annealing treatment (sigmoidal evolution of hardness with time), annealing treatments at 600 ºC, 650 ºC and 700 ºC were selected in order to study the kinetics of recrystallisation. The time evolution of the fraction of recrystallized, X_soft material for the four reductions and annealing temperatures were modelled by a Johnson-Mehl-Avrami-Kolmagorov (JMAK) type curve (Avrami, 1939; Johnson and Mehl, 1939) as it is shown in Eq. (1).

Where k_soft is the time exponent. For austenite recrystallisation this time exponent is only 3 or 4, as JMAK theory predicts, in some lightly deformed fine-grained texture-free Fe-C-Mn steel grades of a uniform grain size. Most often this exponent takes values of the order of 1 for the various Fe-C-Mn steel grades (Humphreys and Hatherly, 2004; Luo et al., 2004). In general, this exponent is related to the geometry of the transformation. However, according to bibliography, the geometry of transformation cannot always be deduced from the value of the exponent, k_soft (Doherty et al., 1997; Humphreys and Hatherly, 2004; Luo et al., 2004).

B, which accepts an Arrhenius expression described by Eq. (2) is a factor containing the activation energy for recystallisation, Q_soft. In general, this value contains all the temperature dependent terms because thermal activation affects the growth strongly through boundary/interface mobility, and therefore the nucleation density depends very strongly on driving force.

B₀ depends on the chemical composition of the cold rolled TWIP steel.

In order to obtain the parameters of JMAK equation (k_soft and B), linear regressions between against ln(t) have been made in the Eq. (3).

The calculated time exponents for TWIP steel range from 0.8 to 0.3, with some trend to decrease as the recrystallisation temperature increases (De las Cuevas et al., 2010b).

Figure 11a illustrates the softening fitting at 600 ºC for 40%, 50%, 60% and 70% reductions. The beginning of recrystallisation is practically the same for the four reductions. However, owing to the stored free energy of high cold rolled TWIP steel (60% and 70%) less time is taken for ending the recrystallisation process as it is demonstrated by the different gradients of green and pink (70% and 60% respectively) curves compared with blue and red slopes (40% and 50% respectively). Figure 11b shows the same evolution of static recrystallisation as a function of time at 700 ºC. As it can be observed at 700 ºC more clearly than at 600 ºC, the greater the deformed material (60% and 70%), the more free energy is stored in the material, therefore less time is taken to complete the recrystallisation process. Figure 11c depicts for 60% cold rolled TWIP the same representation at 600 ºC, 650 ºC and 700 ºC. Clearly, at higher temperatures (700 ºC), faster recrystallisation process takes place.

Although the time to reach the practical end of recrystallisation is very small above 650 ºC and the recrystallised grain size is so small as well (D₀ = 1.50 ± 0.02 μm in all cases), a further intent at a refined characterisation of the kinetic equation of recrystallisation for 60% cold rolled TWIP steel was performed. Taking into account the multiple linear Eq. (5.4), a plane has been fitted to experimental points (softened fraction at 600 ºC, 650 ºC and 700 ºC) optimising the square of the correlation index, R² as it is shown in Fig. 12a. The coefficients of polynomial regression are k_soft and Q_soft, and lnB₀ corresponds to an independent term as it is depicted in Fig. 12b. Reasonable values of k_soft and Q_soft have been obtained respectively, 0.54 ± 0.15 and 281 ± 70 kJ·mol⁻¹ where the activation energy for recrystallisation corresponds practically with the activation energy for bulk self-diffusion in austenite, 270 kJ·mol⁻¹, Humphreys and Hatherly (2004), and for Mn diffusion in the austenite lattice, 265 kJ·mol⁻¹ (Sun and Pugh, 2000). This is in accordance with Scott et al. (2006) who found Q_soft ~ 300 kJ·mol⁻¹ for the same TWIP steel but for a cold rolled reduction of 50%.

4.2. Modelling of grain growth kinetics behavior of TWIP steelTOP

Once recrystallization is complete the new grain structure starts to grow. The result is a new strain-free polycrystalline structure with a Gibbs free energy much lower than the energy of the deformed state (Humphreys and Hatherly, 2004).

The kinetics of grain growth was characterized via empirical equation Eq. (4) (Burke and Turnbull, 1952):

Where D is the mean grain size after a growth time t, starting the growth from a mean grain size D₀. The grain growth exponent n_GG theoretically takes the value of 2 (Burke and Turnbull, 1952): The factor K_GG follows the Arrhenius form (Eq. (5)):

Where Q_GG is the activation energy for grain growth, and (K_GG)₀ a pre-exponential constant. Owing to the small recrystallised grain size as demonstrated in the EBSP-OIM measurements (D₀ = 1.50 ± 0.02 μm), Eq. (4) leads to the following approximation (Eq. (6)):

Combining all the data from annealing treatments in the temperature range of 800 ºC ≤ T ≤ 1100 ºC and annealing times ranging from 12 s ≤ t ≤ 43740 s, which correspond to grain growth stage for TWIP steel, a good linear fitting is obtained using Eq. (7) where the slope of the linear fit is the time exponent n_GG.

The grain growth equation has been optimised using numerical methods that maximize the square of the correlation index, R². The values Q_GG and n_GG that optimise R² in the linear regression are respectively 384 ± 60 kJ·mol⁻¹ and 4.05 ± 0.80. By using Student’s t-distribution, the dashed black line in Fig. 13 corresponds to the confidence interval of the slope n_GG of the linear regression, whilst the dashed blue lines represent the confidence interval of the population of experimental data (blue points).

Values of n_GGlarger than two indicate a grain boundary drag process associated with inclusions, carbides or chemical segregation (solute drag) at the boundaries (German, 1978; Humphreys and Hatherly, 2004). An exponent other than two means a nonlinear dependence of grain boundary velocity on driving pressure (Burke and Turnbull, 1952). Values of n_GG equals to 2 are rarely found experimentally even in very pure materials and average values are close to 2.4. The larger measured exponents, n_GG, are a consequence of the materials used not being ideal (high purity metals) i.e. not consistent with the basic assumptions about the material which are incorporated in the models (Humphreys and Hatherly, 2004).

In the case of medium carbon TWIP steel (22%Mn-0.6%C), at temperature range of 600 ºC ≤ T ≤ 700 ºC the recrystallised grain size is very fine (D₀ = 1.50 ± 0.02 μm) and equiaxed, the textures of recrystallisation and grain growth are very weak and it was found carbide precipitation of (Fe, Mn)₃C–cementite and Vanadium carbonitrides on the austenite grain boundaries. Concerning, the activation energy of TWIP steel for recrystallisation (281 ± 77 kJ·mol⁻¹), it is pointed out that it copes with both the activation energy for self-diffusion in austenite, 270 kJ·mol⁻¹, (Humphreys and Hatherly, 2004), and for Mn diffusion in the austenite lattice, 265 kJ·mol⁻¹ (Sun and Pugh, 2000). It clearly implies that at 600 ºC and 700 ºC where carbide precipitation takes place, the effect of grain boundary drag by the very fine precipitates does not have an important effect. Therefore, one possibility is that the quantity of precipitates is not enough to have a relevant pinning effect. In addition, the distribution of precipitates at such temperatures could be very dispersed where the mean length between precipitates, L_prec, is larger than or equal to the recrystallised grain size, D₀ (neglected grain growth), L_prec ≥ D₀ as it is depicted in Fig. 14a. Nevertheless, the activation energy for grain growth is found to be much higher (384 ± 60 kJ·mol⁻¹) than for recrystallisation process. The reason for such difference is believed to be associated to grain boundary drag (pinning of migrating grain boundaries by precipitates) being only effective for grain sizes above some threshold value, D_threshold. For the vanadium carbides to dissolve, the temperature of heating for quenching should be at least 1150 ºC in austenitic steels with high Mn content (Kalashnikov et al., 2001). In our case, annealing treatments were performed up to 1100 ºC. Therefore, the most plausible explanation is that the quantity of precipitates is enough to have relevant grain boundary drag during grain growth because the mean length between precipitates, L_prec, is smaller than some threshold value of grain size, L_prec < D_threshold, being the recrystallised grain size smaller than the grain size after some grain growth, D_threshold >> D₀ as it is described in Fig. 14b.

4.3. Texture development during recrystallization and grain growth, and their mechanismsTOP

The results strongly induce to conclude that the factor controlling the recrystallisation behaviour and the final microstructure of TWIP steels is the nucleation. One of the main reasons for the very small size of the grains after completion of recrystallisation is most likely due to the availability of many nucleation sites because of the tremendous number of mechanical twin intersections after 60% cold rolling deformation. Furthermore, the apparent homogeneity of the deformation structure does not provide any preferred nucleation site. Hence, the nucleation events may occur in a randomly dispersed manner throughout the deformed austenitic matrix which is confirmed by the EBSP-OIM images illustrated in Fig. 3. It is noticed that very small recrystallised grain sizes of austenitic stainless steel (metastable austenite) and TWIP steels had previously been reported (Vercammen et al., 2004; Bracke et al., 2009).

In this research work, as the same components are found in the cold rolled texture and in the recrystallised texture, the nuclei formed should have the same orientation distribution than the deformation texture. This clearly points against oriented nucleation. Besides the grain size change, the only concomitant structural change that occurs in grain growth is the increasing presence of recrystallized twin boundaries with respect to the recrystallized structure. That means an increased volume fraction of recrystallized twin variants of the original components of the texture (De las Cuevas et al., 2010b). Only extensive grain growth promotes the emergence of new orientations of very low intensity (Vercammen et al., 2004; Bracke et al., 2009).

It is well known that recovery processes are difficult in low-SFE materials, and thus the driving force for recrystallisation is higher than in high-SFE materials such as Al (Hurley and Humphreys, 2003a; Hurley and Humphreys, 2003b). This condition may account for the near site-saturated character of the nucleation process in TWIP steels. Furthermore, the retained rolling texture in the recrystallisation process, as well as the homogeneous nature of deformation structure, which implies the nucleation without any preferred orientation mechanism and thus inhibits a sequential spread of nucleation events, also contributes to a very high nucleation rate at the onset of recrystallisation. The limited growth rate observed in our experiments before the end of complete recrystallisation can be attributed as a consequence of the profuse near site-saturated character of the nucleation and the ensuing impingement of growing grains soon after their nucleation.

5. CONCLUSIONSTOP

The main conclusions of this research work concerning the 22% Mn-0.6% (% in mass) TWIP steel are:

ACKNOWLEDGEMENTSTOP

Economical support from the European Union, Research Programme of the Research Fund for Coal and Steel (contract RFSR-CT-00030) and from the Spanish Ministry of Science and Innovation (action MAT2005-23927-E). We would like to thank D. Badiola for their scientific support in this research work. Finally, I would like to show my deep appreciation to Jerónimo, Martín, Claudia y Pachi.

REFERENCESTOP