3.1. Transformation Kinetic
Figure 2a presents the diagrammatic sketch of the thermal cycle, and
Figure 2b exhibits the dilatometry response obtained during cooling for these specimens. Based on the analysis of the dilatometry curves of the CC-10 steel, the experimental start temperature of martensitic transformation (
MS) was defined as 397 °C, at which a martensite volume fraction of 0.02 was formed. For the ISO-400 steel, the bainite transformation is practically complete during the isothermal holding at 400 °C, and there is no apparent volume expansion below
MS. The PAGs in the ISO-450 steel gradually transformed to bainite during isothermal holding at 450 °C, then slowly transformed to bainite during further cooling, and rapidly transformed to martensite when the temperature dropped below the
MS. The volume fraction of bainite or martensite was obtained by applying the lever rule to the average dilatometry curves and assuming that only martensitic transformation occurred below the
MS. The microstructure of the ISO-400 specimen is fully composed of isothermal bainite, whereas that of the ISO-45 specimen consists of approximately 20% isothermal bainite and 80% martensite. The Koistinen–Marburger (KM) model can be used to study the influence of the isothermal bainitic transformation on the kinetics of martensite formation:
where
fα is the martensite volume fraction,
TKM is the Koistinen–Marburger martensite start temperature, and α
m is the overall rate parameter. Equation (1) was fitted to the experimental curves. Since the KM model is not applicable for the early stages of the transformation and the bainitic transformation occurred before the martensitic transformation of the ISO-450 steel, data above a martensite fraction of 0.4 were adopted in the fitting. Here, the overall rate parameters of the CC-2 steel and ISO-450 steel are 0.0630 K
−1 and 0.0650 K
−1, respectively, indicating that the martensite transformation rate increased due to the pre-transformed bainite. The martensitic transformation rate is affected by factors such as transformation temperature and composition. As carbon is ejected to the untransformed austenite during the bainite transformation, the
MS of the untransformed austenite should be improved. Nevertheless, previous studies have shown that stress can accelerate both bainite and martensite transformation rates, suggesting that pre-formed bainite causes austenite deformation and promotes martensitic transformation [
24,
25]. It has been reported that the pre-transformed martensite will provide more nucleation sites for bainite [
11]. However, whether pre-transformed bainite can provide nucleation sites for bainite is still unknown.
Figure 2d shows how the bainite volume fraction climbed as the holding time increased. When held at 400 °C, bainite underwent an explosive transformation, and at 40 s, 80 percent of it had formed. Compared to -the ISO-400 steel, the bainite transformation at 450 °C proceeds slowly through the isothermal process, growing almost linearly up to 20%. The decrease in bainite transformation temperature will increase the activation energy for bainite transformation, thereby increasing the bainite transformation rate, which is consistent with previous research [
26].
3.2. Microstructural Feature
Figure 3 shows the SEM maps of the isothermal and non-isothermal transformed steels. CC-10 consists entirely of martensite, while CC-2 consists of martensite and a small fraction of bainite. The transformation point of this bainite was not found in the dilatometry curve of CC-2 steel, which is probably acceptable as the dilatometer sensitivity is about 10 percent volume fraction transformed. Compared to full martensite, the mixture of martensite and lath bainite (M/LB) normally exhibits better toughness along with excellent strength and is extensively applied in offshore steels. Further, large amounts of island-like structures were observed in the isothermally transformed steels, which were deemed to be the M-A constituents transformed from the C-rich residual austenite. The ISO-400 steel is entirely composed of lath bainite ferrite as the lath boundaries were decorated by dot/slender M-A constituents (indicated with arrows in
Figure 3d). In contrast to the ISO-400 steel, the ISO-450 steel is composed of two different morphologies—one is the mixture of lath bainite and martensite, and the other is the granular bainite. According to the statistical results derived from the SEM images, the volume fraction of GB in the ISO-400 steel is estimated to be around 10%. In addition, the M-A constituents in the ISO-450 steel exhibit a larger size than those in the ISO-400 steel.
Figure 4 shows the band contrast morphology and the grain boundary (GB) distribution of the steels. The non-isothermal steels and the ISO-400 steel exhibit a lath-like morphology, which is in accordance with the SEM maps. The ISO-450 steel also exhibits a lath-like morphology, while massive bainite ferrite was observed in the local region of the ISO-450 steel, as shown in the enlarged map in
Figure 4d
1. In addition, island-like M-A constituents consist of unindexed pixels or pixels indexed as BCC, with the lower BC being observed in
Figure 4d
1. The M-A constituents normally consist of substructures like martensite variants and austenite grains. These martensite variants transformed from the C-rich austenite and maintained the K-S orientation relationship with their neighboring retained austenite [
27]. The grain boundary (GB) misorientation, as obtained from the EBSD analysis, is presented in
Figure 4e. The density of high-angle boundaries (HAGBs, >45°) was at its maximum in the ISO-450, followed by the CC-2, CC-10, and ISO-400. The density of low-angle boundaries (LAGBs, <15°) was at its maximum in CC-10. The higher density of the HAGBs in the CC-2 steel can be attributed to the refinement of microstructure caused by the geometrical partitioning of the prior austenite grain by pre-transformed bainitic ferrite. The strong variant selection (the V1&V4 variant pair) and large amounts of defects like dislocations induced by fast quenching rate generate a higher density of LAGBs in CC-10 steel. Granular bainite in the ISO-450 steel is typically characterized by coarse bainitic grains, however, it exhibits the maximum density of HAGBs. This phenomenon is explicable as the fraction of granular bainite only accounts for 10% and numerous M-A constituents also contribute to the increase in the density of HAGBs.
A systematic characterization of the M-A constituents is presented in
Figure 5. The size and structure of the M-A constituents transformed in the isothermal stage were dominated by the transformation temperature—
Figure 5a,b exhibits the morphology of the M-A constituents in the ISO-400 steel and ISO-450 steel, respectively. There are large amounts of massive M-A constituents in the ISO-450 steel, both the maximum size (L
max) and the minimum size (L
min) of these M-A constituents exceed 2 μm. Compared to those in the ISO-450 steel, the M-A constituents in the ISO-400 steel mainly exhibit the dot and fine–slender morphologies. Statistical information on the number and density of the various M-A constituent types in the ISO-450 and ISO-400 steel is displayed in
Figure 5c. It demonstrates that the number and density of massive M-A components in the ISO-450 steel is double that of the ISO-400 steel.
Figure 5b
1 is a BC map that overlaps the grain boundary map, showing that the massive M-A constituents consist of various martensite variants and that these martensite sub-grains are separated by high-angle grain boundaries.
Figure 5a
1,b
2 are overlaid with the fcc-phase map.
Figure 5b
2 shows that the massive M-A constituents in the ISO-450 steel contain block-like austenite grains, with a significantly larger size and higher fraction compared to the austenite grains in the M-A constituents of the ISO-400 steel. Some austenite grains in the ISO-450 steel reach sizes approaching 1 μm. In contrast, the M-A constituents in the ISO-400 steel exhibit a significant reduction in size, along with noticeably decreased martensite size and increased lattice distortion. The austenite content in the M-A constituents of the ISO-400 steel is very low, and austenite films were observed at the grain boundaries.
Figure 6 and
Figure 7 present the TEM images of M-A constituents in the isothermal transformed steels treated at 450 °C and 400 °C, respectively.
Figure 6d exhibits the bainitic morphology of the ISO-400 steel, confirming that the ISO-400 steel is primarily composed of lath-shaped bainitic ferrite. The M-A constituents in the ISO-400 steel are smaller in size compared to those in the ISO-450 steel. These constituents contain martensite sub-grains smaller than 100 nm and exhibit a high density of dislocations, which enhances the hardness of the M-A constituents in the ISO-400 steel. Additionally, fine bainitic laths are distributed around the M-A constituents in the ISO-400 steel. In contrast, the massive M-A constituents at the prior austenite grain boundaries in the ISO-450 steel consist of a random arrangement of martensite and austenite, which is also named the type-III M-A constituent according to its structure [
28]. The dark-field images reveal that the austenite is located at the edges of M-A constituents and martensite lath boundaries. And, the M-A constituents in the ISO-450 steel are surrounded by blocky bainitic ferrite and several lath-shaped bainitic ferrites, consistent with the EBSD results.
To further clarify the austenite fraction in the steels, XRD patterns were utilized for analysis, as shown in
Figure 8. Although the austenite fraction below 2% cannot be accurately measured by XRD, a significantly increased intensity of the {111}
γ peak indicates a significant increase in the austenite fraction. In fact, considering that the ISO-450 steel contains only 20% bainite, the austenite fraction within the M-A constituents of the ISO-450 steel is much higher than that of the ISO-400 steel. During the bainitic transformation, the carbon will be excluded into the surrounding austenite grains from the bainitic ferrite. A higher isothermal temperature enhances the carbon diffusion rate, further stabilizing the retained austenite. Additionally, based on the Williams equation, the dislocation densities of CC-10 steel, CC-2 steel, ISO-450 steel, and ISO-400 steel were calculated as 6.1253 × 1011, 5.9233 × 1011, 4.1972 × 1011, and 4.9567 × 1011 cm
−2, respectively.
Using the high-temperature LSCM observation, the real-time features of bainite at the early stage of the transformation were recorded, as shown in
Figure 9. The phase transformation does not occur when the temperature is above 500 °C. The minor fraction of bainitic ferrite transformed above 450 °C as signified in
Figure 9. However, this temperature is higher than that detected using a dilatometer (
Figure 2a). This discrepancy is caused by the sensitivity of the dilatometer, and is similar to the discovery of a small fraction of bainite in CC-2 steel. With a further holding time at 450 °C, bainite gradually transformed. Nevertheless, the bainite formation is inhomogeneous at the grain scale since the remaining grains are still isolated from the transformation region. The remaining austenite grains changed instantly, indicating the start of martensitic transformation, when the temperature dropped below 400 °C. This result is in good agreement with the SEM findings shown in
Figure 3e. The pre-transformed granular bainite was sandwiched between the lath structures. Meanwhile, it is very interesting to find that a newly formed island-like M-A constituent was observed among the pre-transformed bainitic ferrite, as marked with a cycled yellow square.
According to previous studies, the formation of bainitic ferrite is thought to be displacive, and the carbon partitioning into the remaining austenite happens right after each bainite platelet grows. The bainite transformation at a high temperature is slow, and no apparent explosive bainite transformation occurs. As undercooling increases and carbon diffusion proceeds, the compositional heterogeneity of austenite grains increases, forming C-rich and C-poor regions, which in turn promotes bainite transformation. The increased cooling rate shortens the holding time at high temperatures, suppressing bainite transformation and restricting it within a small amount of unstable austenite. When held at 450 °C, bainitic ferrite preferentially forms in regions with lower concentrations of alloying elements. The carbon partitioning during the isothermal process is the primary cause of the formation of island-like structures in the bainitic microstructure. The bainitic ferrite formed at 450 °C expels carbon into the surrounding austenite as it grows. These C-enriched austenite grains wrapped by pre-transformed bainite ferrite partially transform into C-enriched martensite and residual austenite (ISO-450 specimen), or completely transform into C-enriched martensite (ISO-400 specimen) according to the different MS points of the residual austenite.
The difference in the morphology of bainitic ferrite results from the variant selection of bainite. Previous research has indicated that low-temperature bainite transformation requires a greater number of V1&V2 variant pairs to accommodate the strain induced during the transformation process [
29]. At a high transformation temperature, however, bainite transformation primarily relies on the self-accommodation of austenite grains, leading to the elevated fraction of V1&V4 or V1&V1 variant pairs [
30]. The V1&V2 variants are associated with high-angle grain boundaries, approximately 60°, whereas the V1&V4 variants correspond to small-angle grain boundaries, around 10°. The variant selection is responsible for the differences in the misorientation distribution.
The morphology of M-A constituents is strongly affected by the shape and crystallography of the surrounding bainite. N. Takayama points out that M-A constituents are elongated along the growth direction or habit plane of the surrounding bainitic ferrite when the neighboring variants of bainitic ferrite share them [
15]. Therefore, the transformation temperature of M-A constituents dominates the morphology of M-A constituents, although these M-A constituents were finally formed in the cooling to room temperature from the remaining austenite [
31]. M-A constituents are basically formed at the lath boundaries or the prior austenite grain boundaries. The explosive bainitic transformation and refined bainitic laths, which limit the geometric space for M-A constituents, restrict the growth of massive M-A constituents and increase the fraction of dot-like M-A constituents, as the isothermal temperature decreases.
3.3. Low-Temperature Toughness
Table 1 presents the hardness and Charpy impact properties at temperatures of −60 °C and −80 °C. The hardness was at its maximum in the CC-10 steel (376 HV
10), followed by CC-2, ISO-450, and ISO-400 steel. Compared to non-isothermally transformed steels, isothermally transformed steels exhibit a lower hardness. This is attributed to the relief of lattice distortion and the reduction in internal stresses during the isothermal process.
For low-temperature toughness, untempered martensite steels show significant internal distortion and a high density of defects, which always exhibit poor toughness. Compared to the full martensite steels (CC-10 steel), the martensite + lath bainite structure (CC-2) has a higher density of HAGBs, a significantly refined effective grain size, a reduced dislocation density, and an improved low-temperature toughness [
32]. Regarding isothermally transformed steels, the low-temperature toughness of the ISO-450 steel is significantly reduced. Although the ISO-450 steel predominantly consists of lath-like structures with the highest density of HAGBs, these boundaries are partially contributed by massive M-A constituents, which do not effectively hinder crack propagation. Moreover, the massive M-A constituents will cause stress concentration, promoting crack initiation and growth. In contrast, the ISO-400 steel primarily comprises a ferritic matrix with lath morphology and relatively small M-A constituents. Small-sized M-A constituents have little impact on low-temperature toughness.
Figure 10 exhibits the load-deflection curves of the experimental steels at temperatures of −60 °C and −80 °C. Supersaturated as-quenched martensitic steels typically exhibit a high ductile–brittle transition temperature (DBTT) due to lattice distortion and internal stresses, as they lack a tempering process. However, the CC-10 steel shows better toughness compared to the ISO-450 steel. Under a −60 °C impact, the load of the ISO-450 steel deeply dropped after the peak load, demonstrating that once a crack initiates, it immediately adopts an unstable propagation mode until the crack tip blunting. The other steels, in contrast to the ISO-450 steel, proceed with a stable crack propagation mode after the peak load. Under a −80 °C impact, compared to other fractured steels, the crack in the ISO-450 formed earlier, and once the crack forms, it adopts a completely unstable propagation mode. The ISO-400 steel consistently maintains a stable crack propagation mode until the extending cracks become excessively sharp, leading to instability. However, as the crack tip blunts, it reverts to a stable propagation mode. The CC-2 sample exhibits the best toughness, with the smallest unstable propagation region, indicating that the resistance to crack propagation is enhanced.
Figure 11 displays SEM fractographs of fractured Charpy specimens impact tested at −60 °C. The ISO-400 steel exhibits a completely ductile ‘fibrous-fracture appearance’, as characterized by the presence of multiple dimples on the fracture surfaces. In the ISO-450 steel, quasi-cleavage fracture accounted for approximately 45% of the fracture surface in addition to fibrous fracture. It is noteworthy that many secondary cracks were connected to the primary crack and extended laterally in the ISO-450 steel. Those secondary cracks possibly initiated around the brittle component located at grain boundaries, carbides, or M-A constituents. They nucleated in great quantities during loading, propagated rapidly, and either merged into the main crack or facilitated its propagation.
To clarify the nucleation sites of the cracks, the EBSD micrographs showing the locations just beneath the fracture surfaces on the transverse cross-sections of the broken Charpy specimens are presented in
Figure 12. The ISO-450 steel shows a straighter crack propagation path in the quasi-cleavage fracture region, as depicted in
Figure 12b. The microstructure of the ISO-450 steel consists of M/LB and GB. This heterogeneity, resulting from the distribution of M/LB and GB, introduces microstructural plastic incompatibilities. GB exhibits lower strength and hardness compared to the M/LB. During impact loading, plastic deformation is predominantly localized in the GB, leading to stress concentration at the interface between GB and M/LB. Due to the heavy deformation of the bainitic ferrite matrix, high stress concentrates on the boundary of M-A and makes it crack or debond [
16,
33]. Crack initiation in the ISO-450 steel originates from broken M-A constituents, especially their massive morphology, as shown in
Figure 12c. According to the calculation of the size of the Griffith crack based on Griffith theory, the width of the MA constituents can represent the initiation size of the cleavage crack [
34].
Table 2 summarizes the numerical data of M-A constituents in the ISO-450 steel and ISO-400 steel. Although the ISO-400 sample exhibits a higher number density of M-A constituents, the majority are of the dot-type morphology, as illustrated in
Figure 5c. In contrast, the number fraction of massive-type M-A constituents in the ISO-450 steel is approximately twice that observed in the ISO-400 steel. The average size of the M-A constituents in the ISO-450 steel exceeds 1 μm, and this value is widely recognized as a critical threshold associated with reduced fracture toughness. The increased number of coarse M-A constituents intensifies local stress concentrations around brittle phases and decreases the interspacing between microcracks, thereby contributing to crack initiation and propagation. In addition, the blocky austenite grains in the M-A austenite of the undeformed specimens were not observed near the fracture surface, indicating that the martensite transformation occurred in this residual austenite. The presence of deformation-induced martensite in the M-A constituents increases the brittleness tendency of the M-A constituents by intensifying the internal stress concentration within the M-A constituents or at their boundary. Meanwhile, since cleavage cracks propagate along specific crystallographic planes, HAGBs between cleavage planes enhance the ability to deflect cracks. These HAGBs are present at a significantly higher density within the lath structure. The blocky ferrite in GB lacks high-angle grain boundaries that can effectively resist crack propagation, with resistance relying mainly on prior austenite grain boundaries, as illustrated in
Figure 12c. The brittle M-A constituents and coarse blocky ferrite greatly increase the risk of crack formation and propagation in granular bainite. In line with the weakest link theory, the primary cracks preferentially nucleate in granular bainite and propagate following the path of least resistance rapidly, as shown in
Figure 12b
1. The detrimental effects of massive M-A constituents on toughness can be summarized as follows: (i) massive M-A constituents fractured under the shear stress, leading to crack initiation; (ii) M-A constituents induce stress concentration, reducing the critical stress for cleavage crack propagation; (iii) microcracks, initially impeded by prior austenite grain boundaries, consume the resistance to the crack propagation, thereby facilitating the propagation of the primary crack; (iv) the unstable propagation of the fracture occurs through the coalescence of pre-formed micro-cracks within the GB matrix.
In contrast, the fracture behavior of the ISO-400 steel predominantly exhibits a plastic fracture mode, characterized by significant plastic deformation near the fracture surface and the great accumulation of geometrically necessary dislocations within the grains. Microcracks are effectively impeded by packet boundaries within the lath structure, as shown in
Figure 12a
1. The primary crack propagation path is narrow and frequently deflected by packet boundaries, block boundaries, and prior austenite grain boundaries, indicating that the fracture propagation energy is progressively dissipated by the high-angle grain boundaries. Small-sized M-A constituents have little influence on the toughness. The reduced interface between small-sized M-A constituents and ferrite matrix-relevant stress concentration. While the increase in the density of the M-A constituents of the ISO-400 steel may raise the stress concentration factor, the surrounding bainitic ferrite, characterized by its lath morphology, exhibits significantly higher shear resistance. The high-angle packet and block boundaries within the lath structure serve as effective barriers to crack propagation, thereby mitigating the negative effects of small-sized M-A constituents. In the quenching process, a reduction in the instantaneous cooling rate is inevitable. Since bainitic transformed at a high temperature will reduce low-temperature toughness, it is crucial to control the cooling rate around 450 °C, particularly in the central region of high-strength heavy plates. Slower cooling rates around 400 °C help promote the formation of lath bainite and improve the low-temperature toughness. This also explains the excellent properties observed at the 1/4 thickness position of heavy plates.