2013 Volume 53 Issue 5 Pages 900-908
The dynamic transformation behavior of deformed austenite was studied in a 0.79%C high carbon steel over the temperature range 743–823°C. The experiments were carried out in torsion under an atmosphere of argon and 5% H2. All these temperatures are above the orthoequilibrium Ae3 temperature of the steel. Strains of 0.25–4 were applied at a strain rate of 4 s–1. The experimental parameters were varied in order to determine the effects of strain and temperature on the formation of strain-induced ferrite and cementite. The critical strain for dynamic transformation was about 0.20. The volume fractions of the transformed phases increased with strain and decreased with temperature. The observed ferrite structures are Widmanstätten in nature and appear to have formed displacively. The carbon diffusion times required for formation of the observed spheroidal cementite particles were calculated; these are consistent with the occurrence of interstitial diffusion during deformation. Similar calculations indicate that substitutional diffusion does not play a role during dynamic transformation.
Ever since the classic work of Yada and co-workers in the 1980’s,1,2) it has been known that superequilibrium ferrite can be formed at temperatures as much as 166°C above the Ae3 as long as the austenite is being deformed. They investigated the behavior of low C steels containing 0.11%C–1.00%Mn–0.02%Si and 0.14%C–1.06%Mn–0.33%Si. The occurrence of dynamic transformation was observed under both laboratory testing conditions (plane strain compression) and in pilot rolling mill trials. Ferrite grain sizes of 1 to 2 μm were produced as long as a critical strain of about 0.5 was exceeded. They also reported that the ferrite volume fraction increased with deformation from about 5% at the critical strain to about 70 to 80% at accumulated strains of 3 to 4. Another important feature they observed was the re-transformation of ferrite into austenite when it was held above the Ae3 for about 1 s and 8 s, respectively. In order to produce direct evidence for dynamic transformation, Yada and coworkers deformed some steel samples in torsion in an x-ray diffraction apparatus.3) In this way, they were able to confirm the occurrence of dynamic transformation (DT) in real time.
In similar experiments, Chen and Chen4) used laser dilatometry to study the reverse transformation in a 0.18%C–0.60%Mn–0.21%Si steel. In their case, the kinetics of the reverse transformation were much slower than those observed by Yada;1,2) it required almost 40 s after hot compression at 860°C, i.e. at about 30°C above the Ae3 temperature of their steel. Sun et al.5) investigated the reverse transformation in a 0.17%C–0.71%Mn–0.27%Si steel (using a similar kind of laser dilatometer). They deformed their samples in compression at 840°C (Ae3 + 5°C) and 860°C (Ae3 + 25°C) in a Gleeble thermomechanical simulator. They showed that the ferrite formed during austenite deformation re-transformed into austenite at both 840 and 860°C during the first 200 s of holding after deformation. Using the double differentiation method6) they computed the critical strain for DT and observed that it increased with temperature.
Several researchers have modeled the dynamic transformation and the reverse transformation to austenite. These phenomena were simulated using Monte Carlo methods by Tong et al.,7) an approach that was later extended by Xiao et al.8) The latter authors allowed for the occurrence of dynamic recrystallization (DRX) as well as dynamic transformation (DT) during straining. Their simulations indicated that the two mechanisms can take place concurrently. However, DT is favored at the lower temperatures and higher strain rates and DRX under the opposite conditions.
The driving force for this kind of transformation has been calculated by a number of authors. For example, Parker9) in her Ph.D. work used the stored dislocation content of the deformed austenite to evaluate the driving force for the transformation and obtained values of 7.2–72 J mol–1. Similarly, in 2001 Hanlon et al.10) estimated the effect of a ‘homogeneous’ distribution of dislocations as well as of a well developed substructure on the Ae3. The presence of either of these was predicted to lead to a 10°C increase in the effective Ae3, which is much less than was observed by Yada and co-workers. Sun et al.5) also calculated the effect of the stored energy on the driving force. They concluded that the inhomogeneous nature of the dislocation distribution was responsible for higher values of the driving force than those quoted above. More recently, the present authors have calculated the effect of the inhomogeneous distribution in more detail and shown that Gibbs energy increases of as much as 200 J mol–1 can be readily produced.11) The latter, in turn, correspond to Ae3 temperature increases of 100°C or more.
In the present study, the dynamic transformation behavior of deformed austenite was investigated in a 0.79%C high carbon steel by means of torsion testing. This composition was selected to extend the C concentration range over which the DT phenomenon was investigated from 0.06%C,12) 0.09%C,13) 0.21%C14) and 0.45%C15) to the eutectoid level. The tests were carried out to increasing strains at temperatures above both the para-and orthoequilibrium Ae3’s. The use of torsion testing has the advantage that it enables samples to be deformed to large strains. Furthermore, the local strain depends on the radius; multiple observations can therefore be made on a single sample, providing more data for analysis. Finally, the influence of temperature and strain on the dynamic transformation was investigated using scanning electron microscopy (SEM), transmission electron microscopy (TEM), and atom probe tomography (APT).
The chemical composition and Ae3 temperatures (orthoequilibrium and paraequilibrium) of the present steel are given in Table 1. These temperatures were determined using the ThermoCalc16) thermodynamic software with the FactSage FSStel database.17) The steel was supplied in the form of hot rolled plates.
| C | Mn | Si | Ae3 (°C) Orthoequilibrium | Ae3 (°C) Paraequilibrium |
|---|---|---|---|---|
| 0.79 | 0.65 | 0.24 | 733 | 709 |
Cylindrical torsion specimens 3.15 mm in radius and 22.4 mm in length were machined from the steel plates with the cylinder axes parallel to the rolling direction. The detailed thermomechanical schedule used in the current study is depicted in Fig. 1. The steel was heated at 1°C/s to 1150°C, austenitized for 20 min, and then cooled at 1°C/s to 1050°C. After holding for 20 s at this temperature, it was strained to ε = 1.0 at a strain rate of 2.0 s–1 to simulate roughing. After the first deformation, it was held at this temperature for 100 s to permit recrystallization and then cooled at 1°C/s to temperatures in the range 743–823°C. Here the range was selected so as to explore the phenomenon of dynamic transformation at temperatures above both the para-and orthoequilibrium Ae3’s.
Thermomechanical schedule for the torsion tests. In the first deformation, the specimens were strained to 1.0 at a strain rate of 2.0 s–1. In the second deformation, the specimens were strained to 0.25–4.0 at strain rate of 4 s–1.
The torsion samples were held for 1 min at the test temperature before the second deformation. Here they were deformed to strains in the range ε = 0.25 – 4.0, at a strain rate of 4.0 s–1. In order to arrest the progress of static and post-dynamic recrystallization, the specimens were water quenched in about 1.5 s immediately after the second deformation. This also prevented the static decomposition of austenite into pearlite during cooling (see below). A tubular horizontal radiation furnace containing an array of four quartz lamps with an atmosphere composed of argon and 5% H2 was used during heating and torsion testing to minimize decarburization and oxidation of the specimens. The temperature measurements were carried out using an Omega Super XL type K thermocouple that was placed in contact with the mid-length of the torsion specimen.
In order to determine whether some conventional transformation products were produced during quenching or not, the quench rates were measured, both at the surface, as well as along the axis of the sample. The cooling rates were then superimposed on the deformed CCT diagram of the steel. The results of this experiment are displayed in Fig. 2, where the cooling curves are compared with the CCT (undeformed) and CCT (deformed) diagrams for this steel. The latter were obtained as follows. First, the TTT diagram was calculated for the present composition from Ref. 18). Then, the CCT (undeformed) diagram was deduced from the TTT diagram by the Scheil-Avrami method as described by Grange and Kieffer19) and Kirkaldy and Sharma.20) Finally the CCT (deformed) diagram was estimated from the undeformed diagram using information from the Atlas of time-temperature diagrams for irons and steels compiled by Vander Voort.21)
Cooling rates at the surface and the interior during quenching in the present experiments. These are compared with the start and finish CCT diagrams for the steel investigated, including those for both the undeformed and deformed conditions.
The CCT diagrams of this figure are consistent with those of Refs. 22,23,24), determined on similar eutectoid steels. They indicate that the pearlite transformation is completely avoided during quenching, in agreement with the observed microstructures. They also account for the entirely martensitic microstructures along the specimen axes, where no strain had been applied. The presence of both ferrite and martensite at the surface (and intermediate locations) can thus be attributed to the effect of straining during the experiments. These microstructures will be presented in more detail below.
In the torsion tests, the equivalent stresses and strains at the maximum radius were calculated using the expressions:25)
| (1) |
| (2) |
In order to reveal the microstructure, cross-sections perpendicular to the longitudinal axis were cut from the deformed specimens. These were mounted for scanning electron microscopy on a FEGSEM, operated at 15 kV. The mounted samples were polished to a 1200 grit surface finish using SiC papers. Both 3 and 1 μm diamond pastes were used in the final polishing stage. The polished samples were preheated slightly using a hot air blower and then etched with a 2% nital solution for times that varied from 5 to 10 s.
Thin foils for TEM were prepared by twin jet electropolishing using a solution of 5% perchloric acid in methanol at 253 K (–20°C) and an operating voltage of 50 V. Bright-field and dark-field images and selected area electron diffraction patterns were obtained on a Philips CM 20, operated at 200 kV.
A standard two-stage electropolishing procedure30) was used to prepare the needle-shaped atom probe samples using 33% nitric acid in methanol for the first stage, followed by 2% perchloric acid in butoxyethanol at 16 V. APT of the specimen was conducted using the Oxford nanoScience 3DAP. The pulse repetition rate was 20 kHz, the pulse fraction was 0.2 and the sample temperature was maintained at 60 K (–213°C). The compositions of the various phases were obtained from volumes of interest, free from any visible element segregation (e.g., clusters, precipitates, Cottrell atmospheres and boundaries) based on the number of atoms present after background noise was subtracted.
Phase identification was carried out by means of microhardness measurements (using a weight of 1 kg), and scanning electron microscopy (SEM). In addition, an image analyzer was used to determine the percentages of deformation-induced ferrite and cementite by measuring their respective areas and subtracting them from the total area.
The critical strain for the initiation of dynamic transformation was found to be about 0.20.31) A scanning electron micrograph of the steel at 753°C (i.e. Ae3 + 20°C) is illustrated in Fig. 3 after straining to ε = 0.25 at a strain rate of
Scanning electron micrographs showing the presence of Widmanstätten ferrite plates and carbide precipitates. The specimen was deformed at 753°C (Ae3 + 20°C) to a strain of 0.25 in 62 ms and then quenched. The austenite present after straining was converted into martensite. Some characteristic straight interfaces are highlighted with white arrows.
Given the relatively high carbon content, appreciable amounts of cementite (Fe3C) were formed, in addition to the ferrite. The finely dispersed cementite spheroids can be seen to better effect at higher magnification in the lower right hand side figure. Due to the inhomogeneous strain distribution in the parent austenite, some of the areas have not yet attained the critical strain for DT. These regions eventually transform into martensite on quenching. The three phases in the microstructure are readily identified by their morphological characteristics.
In a few places, at higher applied strains, the initial plate-like structures were transformed into equiaxed grains decorating the prior austenite grain boundaries. An example is provided in Fig. 4, where a scanning electron micrograph of a specimen deformed to a strain of 0.5 at 4 s–1 at 753°C (i.e. Ae3 + 20°C) is shown. Both ferrite plates and ferrite grains can be seen. Spheroids are present in some of the ferrite plates as well as in some of the ferrite grains. Conversely, some of the plates and grains are devoid of spheroids. Some cementite plates are located next to the spheroid-free ferrite plates. Thus the microstructure of the deformed and quenched material is quite complex in nature.
Scanning electron micrograph of a specimen deformed to a strain of 0.5 at 4 s–1 at 753°C (Ae3 + 20°C). The deformed and quenched specimen contains ferrite, cementite and martensite.
It is of interest that the principal constituents of the microstructure do not change as the temperature is increased above the Ae3. Three phases, namely, ferrite, cementite and martensite, are present over the entire temperature range from 743°C (i.e. Ae3 + 10°C) to 823°C (i.e. Ae3 + 90°C), the highest temperature investigated. This is clearly demonstrated in Figs. 5(a) to 5(d), where the samples were deformed to a strain of 4 at a strain rate of 4 s–1 at the temperatures shown (Fig. 5(a) –743°C (Ae3 + 10°C), Fig. 5(b) –763°C ((Ae3 + 30°C), Fig. 5(c) –793°C (Ae3 + 60°C) and Fig. 5(d) –823°C (Ae3 + 90°C)). The presence of martensite even after such large strains indicates that the local dislocation density is not high enough in these regions to provide a sufficient driving force for DT.11) That is the austenite in these locations did not attain the critical strain for dynamic transformation. As in the case of Fig. 4, the density of the cementite particles is not uniform across the microstructure.
Presence of ferrite and carbides in the microstructure after deformation to a strain of 4 at 4 s–1 at the following temperatures: (a) 743°C (Ae3 + 10°C), (b) 763°C (Ae3 + 30°C), (c) 793°C (Ae3 + 60°C) and (d) 823°C (Ae3 + 90°C). Here the Ae3 is the orthoequilibrium temperature; the paraequilibrium intervals are 24°C greater.
The concurrent presence of ferrite, cementite and martensite in the microstructure was also verified in the TEM micrographs. The presence of ferrite (F) and martensite (M) (i.e. austenite) in the microstructure of a sample deformed to a strain of 4 at 4 s–1 at 763°C (Ae3 + 30°C) is illustrated in Fig. 6(a). The morphology of a typical Widmanstätten ferrite plate is depicted in Fig. 6(b), where it can be seen to have a width of about 200 nm. The appearance of such plates at lower magnifications, as seen in IQ and IPF maps by EBSD, is demonstrated in Figs. 6(c) and 6(d), respectively. Here the Widmanstätten nature of the plates can be readily distinguished.
(a) Transmission electron micrograph illustrating the simultaneous presence of ferrite (F) and martensite (M), i.e. austenite, in the microstructure. The sample was deformed to a strain of 4 at 4 s–1 at 763°C (Ae3 + 30°C); (b) Micrograph illustrating the morphology of a Widmanstätten ferrite plate and the relative absence of internal dislocations; (c) Appearance of a colony of DT Widmanstätten plates in an IQ map of a 0.09C–0.036Nb steel;29) (d) IPF map of the Widmanstätten plates of Fig. 6(c) as revealed by EBSD, showing the small misorientations between the plates.32)
The ferrite diffraction pattern of a Widmanstätten plate in the present steel is displayed in Fig. 7(a). The cementite precipitates inside the ferrite grains and their diffraction pattern are shown in Fig. 7(b). In the latter figure three cementite precipitates are highlighted by white arrows. The martensite present in the microstructure had two different morphologies, namely, lath and plate shaped. These are illustrated in Figs. 8(a) and 8(b), respectively.
Transmission electron micrographs displaying (a) the ferrite (F) morphology (diffraction pattern inset); (b) three cementite precipitates indicated by arrows and their corresponding diffraction pattern (inset). The sample was deformed in torsion to a strain of 4 at 4 s–1 at 763°C (Ae3 + 30°C) and quenched immediately afterwards.
Transmission electron micrographs of the martensite showing the two different morphologies observed (a) lath-like and (b) plate-like. The sample was deformed in torsion to a strain of 4 at 4 s–1 at 763°C (Ae3 + 30°C).
The compositions of the various phases were averaged over several volumes and are listed in Table 2. Here martensite (prior austenite) is considered to be the phase that contains more than 2 at%C, polygonal ferrite as having less than 0.08 at%C, and carbide-free plate ferrite as less than 0.5 at%C. Although the martensite C concentration is less than that of the nominal composition of the steel (3.6 at%), this is consistent with the presence of C in the numerous carbides that were observed, both as dispersed particles and as thin films between the ferrite plates. The polygonal ferrite concentration is in agreement with the paraequilibrium phase diagram and the carbide-free plate ferrite concentration with the results of previous APT studies of bainite formation.33,34) The latter result is thus consistent with the interpretation of the Widmanstätten ferrite being formed by a displacive mechanism. This conclusion is supported by the lack of partitioning of the substitutional elements Mn and Si (not illustrated here). In the plate ferrite that contained carbides, the C concentration was somewhat higher, Table 2. (The overestimation of the Si level in the present results can be attributed to certain deficiencies of the APT method.30))
| Steel composition | Martensite | Polygonal ferrite | Plate ferrite free of visible carbides | Plate ferrite containing carbides | ||
|---|---|---|---|---|---|---|
| wt% | at% | at% | at% | at% | at% | |
| C | 0.79 | 3.60 | 2.56±0.02 | 0.02±0.004 | 0.45±0.008 | 0.80±0.005 |
| Mn | 0.65 | 0.64 | 0.79±0.01 | 0.78±0.02 | 0.76±0.01 | 0.74±0.004 |
| Si | 0.24 | 0.49 | 0.84±0.01 | 0.77±0.02 | 0.76±0.01 | 0.75±0.005 |
A representative C atom map of the martensite referred to above is presented in Fig. 9(a) and of the Widmanstätten ferrite in Fig. 9(b). A C concentration profile across the Fig. 9(b) sub-boundary is depicted in Fig. 9(c), where the segregation to dislocations can be readily seen. Detailed examination of the torsion sample deformed at 763°C also revealed considerable C segregation to shear bands, see Fig. 10(a). The enrichment at these bands attained levels as high as 5–9 at%, Fig. 10(b). In the plate ferrite, carbides formed in these regions of C enrichment.
C atom maps from representative (a) martensitic; (b) ferritic plate regions. (c) C concentration profile across a sub-boundary in the plate ferrite. The specimen was deformed to a strain of ε = 4.0 at 4.0 s–1 at 763°C (Ae3 + 30°C).
(a) C atom map of plate ferrite; (b) concentration profile measured in the sample shown in (a). The specimen was deformed to a strain of ε = 4.0 at 4.0 s–1 at 763°C (Ae3 + 30°C).
Using Eqs. (1) and (2), equivalent stresses and strains were calculated, leading to stress-strain curves of the type illustrated in Fig. 11. Here, two softening mechanisms are operating concurrently at large strains.13) The first is dynamic transformation (DT). As ferrite is softer than austenite at a given temperature, the formation of strain-induced ferrite from work hardened austenite reduces the flow stress of the specimen. As mentioned earlier, the critical strain for dynamic transformation determined by the double differentiation method is about 0.20,31) which is almost always less than the critical strain for dynamic recrystallization (DRX), the second softening mechanism.35) The critical strains for the initiation of DRX were also determined in Ref. 31) and found to be in the range 0.35–0.45. The initiation of DT does not, however, lead to the immediate presence of a stress peak, i.e. to net softening, since the volume fraction of ferrite produced per unit strain is too low to produce this. The peak stresses are associated instead with strains of about 0.6–0.8, and involve both DT and DRX. These observations are consistent with the results of simulation experiments7,8) carried out using Monte Carlo models, which have shown that the initiation of DRX does not prevent DT from taking place, but retards its progress through the material.
Stress-strain curves of specimens deformed to a strain of 4.0 at 4 s–1 over the present temperature range. After the peak, the flow stress decreases due to the combined effect of dynamic transformation (DT) and dynamic recrystallization (DRX).
The volume fraction of DT ferrite that forms depends on the applied strain and temperature. This is depicted in Fig. 12 for specimens deformed at 753°C (i.e. Ae3 + 20°C) and 783°C (i.e. Ae3 + 50°C) at 4 s–1. From Ref. 31), it is clear that the critical strain for DT to occur in this particular steel is about 0.20, below which no appreciable phase transformation takes place (see inset). The rate of transformation is highest in the strain interval from 0.6 to 2.0 and decreases with temperature. Beyond this strain, the fraction transformed approaches saturation, as called for by the deformation-modified phase diagram.11)
Effect of strain on the percentage of austenite that transforms dynamically into ferrite and cementite at 753°C (Ae3 + 20°C) and 783°C (Ae3 + 50°C) at 4 s–1. The critical strain for dynamic transformation to occur in this particular steel is about 0.20, below which no appreciable phase transformation takes place (see inset). At zero strain, the amount of transformed phases is zero as indicated by the grey square (see inset).
It is important to note that the volume fractions of the transformed phases are highest at the sample surface, where the quench rate is also the highest, while the volume fractions of the martensite (i.e. austenite not subject to DT) are highest near the axis of the specimens, where conversely the quench rate is the slowest.
3.4. Effect of Temperature on Phase Proportions and Microhardness ValuesThe effect of temperature on dynamic transformation is illustrated in Fig. 13, where the specimens were deformed to a strain of 4 at 4 s–1. Here it is evident that the amounts of deformation-induced ferrite and cementite decrease as the temperature is increased. This is because the driving force for the γ-to-α transformation decreases as the temperature is increased, resulting in lower amounts of ferrite and cementite at the higher temperatures. This trend was observed up to 823°C (i.e. Ae3 + 90°C), the highest experimental temperature employed in the present investigation. Extrapolation of the current data to higher temperatures suggests that DT ferrite will continue to form up to temperatures 120–130°C above the orthoequilibrium Ae3. As these quantities are inversely proportional to the amount of martensite, the hardness increases with increase in deformation temperature.
Dependences of the ferrite plus cementite volume fraction and microhardness on ΔT (Experimental Temperature – Ae3). The specimen was deformed to a strain of ε = 4.0 at 4.0 s–1.
In order to understand how strain-induced ferrite and cementite form during DT, it is important to estimate how far C as well as Mn and Si can diffuse during deformation. The following relationship can be used to calculate the diffusion distances of C and Mn in ferrite.
| (3) |
| (4) |
The estimated diffusion distances are shown in Fig. 14, from which it can be seen that, at these temperatures, carbon takes about 100 μs to diffuse a distance of 100 nm in ferrite, Fig. 14(a). This is the approximate time required for carbon to diffuse out of the Widmanstätten ferrite plates during their formation. The carbide spheroids present in the microstructure can form in even shorter times. However, as shown in Fig. 14(b), Mn can only diffuse a small fraction of a nanometer, i.e. less than one atomic diameter, within this time frame. The formation of displacive ferrite, which involves a lattice transition from FCC to BCC, is expected to take place at velocities of about one-third the speed of sound.38) These velocities correspond to elapsed times of the order of picoseconds.11) Thus the carbon partitioning associated with the formation of cementite plates and particles can take place during straining, but there is insufficient time for Mn (or Si) redistribution. This leads to the conclusion that the time scales involved in DT are sufficient for the attainment of a kind of local paraequilibrium but not for that of orthoequilibrium.
Estimated mean diffusion distances of (a) C and (b) Mn in ferrite over the experimental temperature range.
The present results as well as those reported earlier5,12,13,14,15,39) indicate that DT ferrite probably forms during the finishing stages of strip rolling, when pancaking accompanied by strain accumulation takes place at temperatures below the Tnr but above the Ae3. This view is supported by the drops in mean flow stress (MFS) that take place above the Ae3 in rolling simulations.40) It thus has applications in the control of strip rolling mills, where the interpass times are short, leading to strain accumulation in the austenite. The formation of DT ferrite (a softer phase) from austenite under these conditions will reduce the roll separation force, thus influencing the rolling load and the control of shape. Nevertheless, since the parameters that control this mechanism are not well understood at the moment, it has not yet been possible to devise controlled rolling schedules that exploit this type of transformation. Perhaps it could be employed in planetary hot rolling, where large reductions can be applied in a single pass just above the Ae3.
The following conclusions can be drawn from the present study.
(1) The critical strain for the formation of strain-induced ferrite and cementite above the Ae3 in a 0.79%C steel is about 0.20.
(2) The volume fractions of the Widmanstätten ferrite and cementite increase with applied strain and decrease as the temperature is increased above the Ae3.
(3) DT ferrite plates are relatively thin (about 200 nm thick) and appear to have formed by a displacive mechanism.
(4) As the applied strain is increased, some of the plates are gradually converted into approximately equiaxed grains.
(5) The carbon diffusion distances associated with the formation of ferrite plates and spheroidal cementite particles are consistent with elapsed times of less than a microsecond over the experimental temperature range. Thus there is sufficient time for interstitial diffusion to take place during DT. However, the times available do not allow for any substitutional diffusion.
The authors are grateful to the Natural Sciences and Engineering Research Council of Canada for funding this work. The authors also thank Professor Elena V Pereloma of Wollongong University, Australia and Dr. Ilana Timokhina of Monash University, Australia for the provision of their TEM and APT facilities.
