1 Introduction

Lath martensitic steels are widely employed in diverse industrial applications, namely transportation, automotive, and pipeline due to their high strength that originates from the mutual interaction of dislocations, carbon in solid solution, and microstructure refinement.[1,2] These applications largely require an optimum balance between strength and toughness, which is achieved by the introduction of a specific overall texture and/or the microstructure constituent characteristics (i.e., retained austenite, block/packet size, and/or boundary network). These modifications are achieved through the application of specific thermomechanical processing route/s[3,4,5,6,7,8] and/or alloying addition.[9,10]

Cleavage fracture is, for example, known to be promoted by the presence of the rotated Cube orientation in the transformed microstructure, which deteriorates the overall toughness.[3] This is generally alleviated in microalloyed steels by the application of hot deformation at temperatures where there is no austenite recrystallization leading to a significant decrease in the development of the rotated Cube component upon transformation.[3] This process also introduces specific dislocation substructures in the parent austenite due to preferred slip activities, promoting specific variants (i.e., variant selection) upon transformation (e.g., to martensite).[11]

Refinement of the parent austenite grain prior to the transformation also leads to an improved toughness without sacrificing the strength.[8,12,13] Interestingly, this process does not alter the texture, rather there is a relatively weak transformation texture.[8] Nonetheless, this process appears to significantly change the grain boundary network of martensite. Interestingly, the trend is similar to that observed for the deformed austenite condition, where the variant selection leads to an increase in 60 deg/[111] boundary population at the expense of 60 deg/[110] boundaries.[7,8]

The grain boundary is known to have a marked effect on the material properties.[14] However, there are specific boundaries formed during martensitic (shear) transformation, where the phase transformation crystallographic constraints, rather boundary energy, define their character (i.e., boundary plane orientation, which is largely (110) planes with twist, tilt, or mixed characterFootnote 1).[15] However, it is not specifically clear how the boundary type/character generated through the shear transformation controls a given property of interest (e.g., toughness).

Hence, the aim of the current study was to explore the influence of the martensite boundary network characteristics formed from different parent austenite grain sizes on toughness. Here, the steel was subjected to different cyclic thermal treatments to obtain two parent austenite grain sizes prior to martensitic transformation, leading to two distinct boundary networks. The propensity for crack propagation in different boundaries was studied using molecular dynamics (MD) simulation to interpret the change observed in toughness of the different martensitic microstructures.

2 Experimental Procedure

The steel had a chemical composition of 0.043 C, 1.68 Mn, 0.2 Si, 0.017 Ti, 0.021 Nb, 0.002 B (in wt pct) and was supplied in the form of hot-rolled plate with a thickness of 12 mm. To obtain two distinct parent austenite grain size prior to the martensitic transformation, the steel was subjected to different cyclic heat treatments.[8] Rectangular samples with a dimension of 10 mm × 10 mm × 55 mm were machined along the hot rolling direction, having a 3-mm-diameter hole with a depth of 15 mm from both ends of the rectangular sample to position the thermocouples and monitor the temperature throughout the test. The heat treatment was conducted using an induction furnace embedded in the hot torsion rig, which was equipped with a quartz tube having a positive pressure of Argon gas atmosphere to prevent oxidation and decarburization along with a water-quenching system to immediately quench the sample after each reheating cycle.[6] The temperature difference between the two ends of the sample was kept within 10 °C during the test.

The samples were initially reheated to 1000 °C at 20 °C/s and held isothermally for 180 seconds to obtain a relatively coarse parent austenite grain size of 84 ± 0.5 µm (Figures 1(a) and (b)). They were then water-quenched immediately to achieve a fully lath martensitic microstructure (hereafter called coarse-grained martensite). Some of the samples were subjected to five further similar heat treatment cycles, consisting of reheating to 925 °C (just above Ac3 temperature of 900 °C) at 20 °C/s and holding isothermally for 5 seconds followed by water-quenching. This led to an average parent austenite grain size of 3 ± 0.1 μm after the five cycles prior to the martensitic transformation [Figures 1(c) and (d)], hereafter called fine-grained martensite.

Fig. 1
figure 1

(a, c) The band contract image for (a) coarse-grained martensite and (c) fine-grained martensite and (b, d) their corresponding reconstructed parent austenite grain structures using TSL OIM V8 software. Austenite grain boundaries are identified by black lines. The triangle inset in (d) represents color codes referring to the normal direction (Color figure online)

Full size image

The microstructure was characterized by a FEI Quanta 3D scanning electron microscope equipped with an electron backscatter diffraction (EBSD) detector. The samples for EBSD were prepared using standard mechanical grinding and polishing followed by final polishing with a colloidal silica slurry. Data acquisition was performed at an operating voltage of 20 kV and a beam current of 4 nA using a step size of 0.15 μm. The EBSD post processing was conducted using the TSL OIM analysis V8 software. Four EBSD maps were acquired, covering a total area of ~ 57,600 µm2 (i.e., 4 × 120 µm × 120 µm) for each heat-treated condition. The microstructure constituents (i.e., parent austenite grain, block, and packet sizes) were measured using the linear intercept method. The statistical error was computed using the \(\mathrm{standard error}=\frac{S}{\sqrt{N}},\) where S and N are standard deviation and number of measurements, respectively.

To evaluate the toughness, a U-notch grove was machined on the heat-treated rectangular samples (10 mm × 10 mm × 55 mm) using a wire-cut machine in the rolling-transverse direction plane (RD-TD) towards the normal direction (ND) with respect to the original hot-rolled plate, according to ASTM E23 Charpy impact testing standard (Figure 2(a)). The Charpy impact test was performed using an Instron MPX pendulum impact machine over the temperature range of − 196 °C to 25 °C for the fine-grained martensitic microstructure and − 80 °C to 25 °C for the coarse-grained martensite. Both fine- and coarse-grained martensitic microstructures were subjected to similar temperatures up to − 80 °C with ~ 20 °C temperature interval (i.e., ambient, 0 °C, − 20 °C, − 40 °C, − 60 °C, and − 80 °C). The Novec liquid was used to cool down the sample to − 80 °C, while the temperature of − 196 °C was reached using liquid nitrogen. Three Charpy samples were tested for each condition and, the average Charpy energy value was subsequently calculated for a given condition. The fracture surface was evaluated using a SEM Zeiss Supra 55VP with a secondary electron detector operating at 20 kV voltage, and 8 nA beam current.

Fig. 2
figure 2

(a) The schematic representation of Charpy testing sample. RD, TD, and ND represent rolling direction, transverse direction, and normal direction, respectively. (b) The Charpy energy as a function of test temperature for martensite produced from different parent austenite grain sizes (i.e., ~ 84 and ~ 3 μm). Asymmetric hyperbolic tangent fitting was performed for Charpy testing data for both conditions using NIST software package.[18] The fracture surface of coarse-grained (c–f) and fine-grained (g–j) martensite microstructures at different Charpy testing temperatures. (c and g) were adopted from Ref. [8], which is an open access article, permitted to use under the terms of the Creative Commons CC-By

Full size image

3 Results and Discussion

The cycling heat treatment resulted in a significant refinement of parent austenite grain size from 84.0 ± 0.5 μm in the first cycle to 3.0 ± 0.1 μm after 6 thermal cycles (Figure 1). During the first cycle, the initial microstructure initially transforms to austenite upon heating above the Ar3 temperature and then transformed to martensite during the subsequent water quench. The martensite formation significantly refines the microstructure, increasing the density of the nucleation sites for austenite formation during the subsequent reheating cycle. This restricts the austenite growth, leading to a smaller austenite grain size and subsequently finer martensite upon quenching.[16,17] Subsequent thermal cycling routines progressively lead to the parent austenite grain refinement, reaching 3.0 ± 0.1 μm after 6 cycles (Figures 1(c) and (d)).

Due to the low carbon content in the steel composition (in wt pct C), it can be assumed that the austenite to martensite phase transformation follows the Kurdjumov–Sachs (K–S) orientation relationship.[15] Therefore, 24 variants can theoretically be formed within a given austenite grain during the martensitic transformation. With the absence of variant bias, this leads to 16 distinct lattice misorientations (i.e., intervariant boundaries) through the impingement of these variants due to crystal symmetry (Table I). In turn, each individual parent austenite grain is subdivided by specific intervariant boundaries, that can be employed to define martensite microstructure constituents (i.e., packet and block). The block boundaries are formed through the impingement of two distinct variants having an identical habit plane (i.e., V1–V2 to V1–V6, Table I), though the interaction of the two distinct variants from different families leads to the formation of packet boundaries (e.g., V1–V7 to V1–V24, Table I). Hence, the blocks were outlined considering all boundaries having greater than 5 misorientation angles, i.e., prior austenite grain boundaries along with all intervariant boundaries presented in Table I. The packets delineated all of the boundary network, excluding the lattice misorientations associated with the block boundaries. It appeared that the change in the parent austenite grain size significantly altered the martensite characteristics (i.e., block and packet). The size of the block and packet reduced from 5 ± 0.18 and 18 ± 0.40 μm, respectively, for the coarse-grained martensite, to 2 ± 0.04 and 3 ± 0.05 μm, respectively, for the fine-grained martensitic microstructure. This resulted in an increase in the overall hardness from 314 ± 0.6 HV10 kgf to 343 ± 0.5 HV10 kgf of the martensite transformed from coarse and fine parent grain size, respectively (Table II).

Table I The 24 Possible Variants Generated Through Austenite–Martensite Phase Transformation Considering the Kurdjumov–Sachs orientation Relationship, Which is Reprinted with Permission from Ref. [15]
Full size table
Table II The Microstructure Characteristics of Martensite Produced Through Different Thermal Cycles
Full size table

The parent austenite grain size also significantly altered the Charpy impact energy of martensite. In general, the coarse-grained martensite displayed a lower impact toughness compared to the fine-grained martensitic microstructure at all test temperatures. The difference was relatively low (~ 11 J) at ambient temperature. This is evident from their fracture surfaces, displaying a ductile failure dominated by abundant dimples, though the coarse-grained martensite occasionally exhibited the presence of a few cleavage facets (Figure 2). The impact toughness difference between the microstructures progressively increased with a reduction in the test temperature, becoming ~ 40 J at − 80 °C (Figure 2(b)). The ductile–brittle transition temperature (DBTT) for the coarse-grained martensite was estimated to be − 40 °C, where the fracture surface consisted of both dimples and cleavage facets (Figures 2(b) and (d)). The DBTT was estimated to be ~ − 110 °C for the fine-grained martensite, as the fracture surface of − 80 °C was still dominated by dimples (Figure 2(i)). At the lower shelf energy region, both microstructures displayed a brittle fracture surface dominated by cleavage facets (Figures 2(f) and (j)). Note that the DBTT at a specific energy level, such as 40 J, is used to provide a reference point to compare the toughness of different materials under similar conditions. Considering the 40 J criteria, the DBTT significantly reduced from ~ − 50 °C for the coarse-grained martensite to ~ − 100 °C for the fine-grained martensite.

In general, the material toughness can be influenced by multiple factors, such as chemical composition (impurities),[19] microstructure constituents,[20] grain morphology,[21] texture,[22] hardness,[23] grain size,[13] and the grain boundary network.[14] The steel composition is identical for both thermal cycle conditions. Therefore, the effect of impurities (e.g., S and P) and/or inclusions (e.g., MnS) on toughness would be similar for both microstructures. In addition, they are fully martensitic, containing aggregates of elongated laths with different alignments for both conditions. In other words, no metastable second phase (i.e., retained austenite) was identified in the martensitic microstructures.

The texture is known to influence the toughness of steel, specifically the cleavage fracture is delayed by a decrease in the rotated Cube orientation (i.e., {001}<110>).[3,4,5] However, the overall texture is expected to be qualitatively similar for both martensitic microstructures due to the texture memory effect.[24] There is an inverse relationship between hardness and toughness.[23] However, the parent austenite grain refinement leads to the enhancement of both hardness and toughness of martensite in the current study (Figure 2 and Table II). The austenite grain refinement significantly reduces the martensitic constituents such as block and packet sizes (Table II), increasing the length of intervariant boundaries. Therefore, the increase in hardness can be attributed to dislocations introduced through the martensitic transformation and a higher fraction of boundary area in the fine-grained martensite compared to the coarse one.[8,17] However, it is demonstrated that not all boundaries have the same influence on toughness, as the boundary propensity to crack initiation and/or propagation (i.e., toughness) largely depends on the grain boundary plane orientation.

Although the {001} boundaries are generally known to promote intergranular fracture in steels,[25] some reports suggest that boundaries with {110} orientation also favor the crack initiation and propagation.[26] Interestingly, the intervariant boundaries formed through martensitic transformation are terminated on {110} or close to {110} crystallographic planes due to the phase transformation constraints.[7,15,27] Recently, it was shown that the enhancement of intervariant boundary population in bainitic steels does not necessarily increase the material toughness, rather it may have a detrimental effect on the toughness on some occasions.[28] This was explained due to the identical orientation of {110} preferred slip planes for dislocation glide in body-centered cubic (bcc) materials with the intervariant boundary orientations, which make the dislocations slip in a direction parallel to the intervariant boundary or being restricted by the intersected lath boundaries. Therefore, the intervariant boundary spacing limits the slip length, impeding general yielding/plasticity at the stress concentration, which favors fracture (reducing the toughness).[28]

The current result, to some extent, contradicts the recent findings, as the martensite microstructure refinement due to a decrease in the austenite grain size enhances the toughness, despite reducing the boundary spacing (i.e., packet and block size refinement, Table II). However, it is important to emphasize that the martensitic transformation results in 16 distinct intervariant boundaries based on the Kurdjumov–Sachs orientation relationship (K–S OR), displaying distinct characters (twist, tilt, and mixed, Table I).[15] However, it is not clear how the character of the intervariant boundary impacts the general yielding (crack initiation and/or propagation).

One of the criteria to determine the propensity of a boundary to crack initiation and propagation is the grain boundary energy (i.e., high-energy boundaries are detrimental to the toughness and vice versa).[14] The current authors have calculated the energy of different intervariant boundaries formed during shear transformation in steel using molecular statics simulation.[29] These boundaries exhibited a broad range of energy from 306 to 1109 mJ/m2 (Figure 3). In general, the boundaries with a mixed character mostly have the highest energy, while the symmetric tilt boundaries reveal the lowest energy (e.g., the 60 deg/[111] symmetric tilt boundary). Interestingly, the latter has a relatively small fraction, which slightly increases with austenite grain refinement (Figure 3). However, the high-energy mixed boundaries are also present in a very small fraction, and the change in their populations with the austenite grain refinement is negligible. Therefore, these types of boundaries are not expected to have a significant effect on toughness. However, the 60 deg/[110] boundaries with a twist character exhibit a relatively high energy (765 mJ/m2), also having the maximum population among all intervariant boundaries. Additionally, a reduction in austenite grain size markedly reduces its population in the martensitic structure (Figure 3). This is due to a change in the variant selection mechanism from the 3-variant clustering (V1V3V5) towards the 4-variant clustering (V1V2V3V4) with a decrease in the parent austenite grain, which leads to a progressive promotion of 60 deg/[111] and 10.5 deg/[011] intervariant boundaries at the expense of 60 deg/[110] martensite intervariant boundaries.[8]

Fig. 3
figure 3

The fraction of intervariant boundaries associated with K–S OR in the lath martensite transformed from different parent austenite grain sizes, as listed in Table I, and their corresponding boundary energy calculated in Ref. [29] using molecular statics simulation. T, ST, and M subscripts refer to twist, symmetric tilt, and mixed boundaries

Full size image

Therefore, the low-energy 60 deg/[111] symmetric tilt boundary and high-energy 60 deg/[110] twist boundary were chosen to investigate the role of grain boundary energy on the propensity of crack propagation, and subsequently the overall toughness in martensite, using MD simulation implemented by LAMMPS. A boundary structure database of martensite intervariant boundaries representing the low-energy 60 deg/[111] and relatively high-energy 60 deg/[110] boundaries having a pre-existing crack along them was generated using in-house Python code. The total length of the boundary and initial crack were 278 and 140 Å, respectively, for both boundary types (Figure 4b,f). The interface model related to these intervariant boundaries was initially equilibrated by MD simulation with respect to the isobaric–isothermal ensemble at the pressure of 0 bar and temperature of 300 K. Then, each boundary was subjected to uniaxial tensile deformation (i.e., normal to the boundary planes) at a constant strain rate of 108 s−1 to monitor the crack propagation using MD simulation. Here, it was assumed that the martensitic lath has a bcc structure due to the low carbon content.[15] Therefore, the MD simulation was performed using pure iron. In this regard, MD simulation was implemented by the LAMMPS software[30] with embedded atom method potentials defining the crack propagation length for each intervariant boundary.[31] A non-linear conjugate algorithm was used to achieve the optimal grain boundary structure with minimum energy. Successive rigid body translations of one lattice with respect to the other were used to sample different initial configurations, which improves the probability of finding the correct equilibrium grain boundary structure for the energy minimization.[32] The Open Visualization Tool (OVITO) was employed to visualize the simulation output data generated by the simulation.[33]

Fig. 4
figure 4

(a) The crack propagation length over time for the low-energy 60 deg/[111] and high-energy 60 deg/[110] intervariant boundaries. Atomic configurations in α-Fe for (b–e) the low-energy 60 deg/[111] and (f–i) high-energy 60 deg/[110] intervariant boundaries with a planted crack along with the boundary plane under tensile loading after different loading times

Full size image

The MD simulation showed the application of tensile deformation perpendicular to the boundary plane led to the propagation of the crack along the interface of the boundary for both 60 deg/[111] symmetric tilt and 60 deg/[110] twist intervariant boundaries (Figure 4). It appeared that the character of intervariant boundary and its energy markedly affect the crack propagation characteristics. For the 60 deg/[111] tilt boundary, the crack propagation led to continuous nucleation of two different types of dislocations (i.e., ½\([\overline{1 }11]\) and ½\([111]\)), which were emitted in the matrix at either side of the boundary (Figure 4b-e). In other words, the generation, and movement of different types of dislocations (i.e., ½\([\overline{1 }11]\) and ½\([111]\)) is responsible for the crack propagation in low-energy 60 deg/[111] tilt boundary. However, the crack propagation along the high-energy 60 deg/[110] twist boundary was accompanied with the formation of mechanical twins with a {112} plane orientation along with the emission of ½\([1\overline{1 }1]\) and ½\([\overline{1 }\overline{1 }1]\) dislocations ahead of the twins as they propagate (Figures 4(f) and (i)).

In general, the crack propagation rate in the low-energy 60 deg/[111] symmetric tilt boundary was much slower than the high-energy 60 deg/[110] twist boundary (Figure 4(a)), taking 2300 and 1350 picosecond (ps) for the boundary to fail, respectively. This suggests that the low-energy 60 deg/[111] symmetric tilt boundary has a lower propensity to crack propagation compared with the high-energy 60 deg/[110] twist boundary. This could be due to a higher excess free volume at the interface of the high-energy 60 deg/[110] twist boundary, promoting more atomic vibrations during the tensile deformation, compared with low-energy 60 deg/[111] symmetric tilt boundary (Figure 4).

EBSD examination of the cross-section perpendicular to the fracture surface for coarse-grained martensitic microstructure subjected to impact testing at ambient revealed the presence of voids/cracks with a wide range of sizes and morphologies (Figure 5). They appeared as round or elongated morphologies, which may be initially nucleated at inclusions and/or intervariant boundaries. Some micro-cracks also appeared along the intervariant boundaries (shown by white arrows in Figure 5). Since this region is subjected to severe plastic deformation, it is difficult to identify the lattice misorientation of intervariant boundaries with micro-cracks. They most likely have high-energy (e.g., 60 deg/[110] twist intervariant boundaries), as they are more prone to the crack initiation than low-energy boundaries (e.g., 60 deg/[111] symmetric tilt boundaries).[34] The fine-grained martensitic microstructure contains fewer connected high-energy boundaries (e.g., 60 deg/[110]), which slows down the coalescence of intergranular micro-cracks and promotes dimple fracture, ultimately enhancing the toughness and reducing the DBTT temperature. In contrast, the coarse-grained martensite microstructure includes more connected high-energy 60 deg/[110] boundaries, which accelerates the formation of intergranular cracks, resulting in the partial observation of cleavage facets in the fracture surface, even at room temperature (Figure 2(c)). This becomes more pronounced for low temperature impact testing, where crack propagation of the main crack is accelerated when it encounters high-energy boundaries, as they have higher crack propagation rate than low-energy boundaries (e.g., 60 deg/[110] vs 60 deg/[111], Figure 4). This leads to a reduction in the toughness and an increase in the DBTT temperature compared with the fine-grained martensite.

Fig. 5
figure 5

Band contrast image of cross-section perpendicular to the fracture surface for the coarse-grained martensitic microstructure subjected to the impact testing at ambient

Full size image

4 Conclusion

The current study revealed that parent austenite grain refinement prior to martensitic transformation significantly altered the microstructure constituent characteristics (i.e., block and packet size, and intervariant boundary network), ultimately improving the mechanical properties (i.e., toughness and hardness) of martensite. MD simulation demonstrated that the propensity to crack initiation and/or propagation largely depends on the intervariant boundary character/energy. The crack propagation rate in the high-energy 60 deg/[110] twist boundary was much faster than the low-energy 60 deg/[111] with a symmetric tilt character, suggesting that the martensite intervariant boundaries with twist character (i.e., 60 deg/[110]) are more susceptible to crack initiation and propagation compared to the symmetrical tilt boundaries (i.e., 60 deg/[111]). This was consistent with the improved toughness in the fine-grained martensite, where the population of low-energy boundaries reduced at the expense of high-energy intervariant boundary, compared with the coarse-grained martensite. This demonstrates that the mechanical properties (i.e., toughness and hardness) of martensitic microstructure can be significantly enhanced through engineering of the intervariant boundary network.