Effect of microstructure on the impact toughness of high strength steels

One of the major challenges in the development of new steel grades is to get increasingly high strength combined with a low ductile brittle transition temperature and a high upper shelf energy. This requires the appropriate microstructural design. Toughness in steels is controlled by different microstructural constituents. Some of them, like inclusions, are intrinsic while others happening at different microstructural scales relate to processing conditions. A series of empirical equations express the transition temperature as a sum of contributions from substitutional solutes, free nitrogen, carbides, pearlite, grain size and eventually precipitation strengthening. Aimed at developing a methodology that could be applied to high strength steels, microstructures with a selected degree of complexity were produced at laboratory in a Nb-microalloyed steel. As a result a model has been developed that consistently predicts the Charpy curves for ferrite-pearlite, bainitic and quenched and tempered microstructures using as input data microstructural parameters. This model becomes a good tool for microstructural design.


INTRODUCTION
One of the major challenges in the development of new steel grades is to get increasingly high strength combined with a low ductile brittle transition temperature and high upper shelf energy, Kv US .This requires a deep knowledge about microstructure and mesotexture effect on material properties, but is not an easy task for microstructures with an increasing degree of complexity in which the strength is controlled by low and high angle boundaries, while toughness depends on the effective grain size or cleavage unit size (D).Cleavage happens usually on {001} ferrite planes and the crack deflects at boundaries that can be considered as misorientation dependent energy barriers for propagation.The value of D was traditionally expressed as an optical ferrite boundary mean linear intercept, but EBSD has revealed that, even for ferrite-pearlite obtained after thermomechanical treatments (Bengochea et al., 1998, Novillo et al., 2005, Bhattacharjee et al., 2004) a relatively high density of low angle boundaries develops.The situation becomes more complex for bainite or martensite.In this context, EBSD techniques become a must to properly quantify the microstructures due to their fineness and particular distribution of low and high angle boundaries (mesotexture).The cleavage unit size is usually computed for boundaries with a misorientation above a certain threshold typically in the 10-15° (Díaz-Fuentes et al., 2003).EBSD is a powerful tool for developing appropriate relations between microstructure and mechanical properties, but work still remains in this field.
The first challenge concerns the development of a systematic methodology for EBSD quantification.The angular resolution of the EBSD scans can be theoretically estimated at 1°, but the lowest threshold angle for consistent size quantification was estimated at 2° (Iza-Mendia and Gutiérrez, 2013).Then, the boundaries were divided into: low (θ < 2°), medium (2° ≤ θ < 15°) and high (θ ≥ 15°) angle.Low angle boundaries will contribute to the dislocation strengthening, medium and high angle constitute the Hall-Petch contribution.High angle boundaries (θ ≥ 15°) was considered as the threshold angle that defines the cleavage unit size (Díaz-Fuentes et al., 2003).
The second challenge concerns the adaptation of the available formulations (Iza-Mendia and Gutiérrez, 2013;Gutiérrez, 2013).This opens the way to new propositions like the misorientation dependent Hall-Petch coefficient (Iza-Mendia and Gutiérrez, 2013) expressed by the following equation: .5 in which f i and θ i are respectively, the relative fr equency and the mean misorientation angle (rad) in the interval i, determined by EBSD; for ferrite α = 0.3, M = 3, μ = 8 × 10 4 MPa and b = 2.5 × 10 −7 mm and d (mm) = MED 2° is the mean equivalent diameter determined for θ ≥ 2°.The optimization of the toughness is a challenge by itself.The microstructure design philosophy works on ductile brittle transition temperature and K Vus concepts.In complex steel microstructures, Kv US is not directly linked to the grain size, but is mostly influenced by the volume fraction of various phases (Sung et al., 2011).The tran sition temperature can be expressed in different ways, one of them being the 50% ductile fracture appearance transition temperature (FATT).For low carbon ferrite-pearlite steels, the FATT is usually related to the steel composition and to a series of microstructural contributions (Irvine et al., 1967, Mintz e t al., 1979, Mintz et al., 1994).
The typical formulations were recently extended to bainitic and quenched and tempered microstructures (Gutiérrez, 2013).The aim of this work is now to develop a microstructure based model of the Charpy curves for microstructures with a variable degree of complexity.This model is aimed at being a tool for a better understanding of toughness development in high strength steels.

MATERIALS AND METHODS
A series of microstructures has been produced through different thermal and thermomechanical sequences (Table 1) on a Nb-microalloyed steel with the composition: 0.15 C−0.3 Si−1.4 Mn−0.012P−0.037 Al−0.033Nb−0.011V−0.007 N. The steel was reheated at 1200 °C for 30 minutes in order to dissolve Nb and subsequently cooled to the deformation temperature for S1 to S3 or water quenched for S4 and S5.S1 and S2 correspond respectively to 1-pass and 2-pass plane strain compression (PSC) tests.For S3, a 4-pass horizontal-vertical compression test was applied.After deformation, the specimen was fast cooled (5 °C s −1 and 15 °C s −1 respectively for low and high coiling temperatures).Coiling was simulated by 1 hour holding at the coiling temperature (650 or 300 °C), followed by slow cooling to room temperature.The deformations were always performed at a strain rate of 1 s −1 and the pass-temperatures and applied strains are shown in Table 1.
For S4 and S5 thermal treatments, after water quenching from 1200 °C, the steel was re-austenitised for 15 minutes at 900 °C.For S4, this was followed by immersion into a salt bath at temperatures of 650 or 300 °C, holding for 30 minutes and slow cooling to room temperature.For S5, the steel was water quenched from 900 °C and tempered for 30 minutes at 650 or 300 °C.
The microstructural characterisation was performed by FEG-SEM and EBSD.A detailed description of the results can be found in Iza-Mendia and Gutiérrez (2013).Tensile tests were carried out at room temperature with a strain rate of 10 −3 s −1 .Charpy tests were performed at 5.4 m s −1 on a Tinius Olsen 84 pendulum impact tester with a maximum capacity of 406 J.For more details see in Gutiérrez (2013).

RESULTS AND DISCUSSION
A relatively broad range of microstructures can be produced in Nb-microalloyed steels by changing the processing conditions (Fig. 1).Ferrite-pearlite or bainite-pearlite microstructures were produced depending on the coiling temperature.Quenching and tempering produces pearlite-free tempered martensite.Tempering at 650 °C preserves the morphology of the martensite, but induces the precipitation of carbides.
Considering all these microstructures as a whole a series of microstructural variables can be defined that affect in different ways strength and toughness.

Pearlite volume fraction and grain boundary carbides
Fractography showed that each time the brittle fracture origin was identified, it was associated to carbides.Grain boundary carbides cannot be avoided in most of the thermomechanically processed Nb-microalloyed steels with F+P microstructures, but their volume fraction and size depend on the carbon content and processing conditions.The carbide thickness, t, was found to be of around 0.08 for bainitic microstructures and between 0.12 and 0.27 μm for ferrite-pearlite.According to these values, their detrimental contribution to the FATT can be estimated to vary between 29 and 58 °C in this steel when using the following expression proposed by Mintz et al., (1979): ΔFATT c (°C) = 112t 0.5  (2) Pearlite also impairs toughness and its contribution can be expressed according to the following equation (Mintz et al., 1979): (3) Quenched and tempered microstructures (S5) are free of pearlite.For the rest, pearlite volume fraction varies between 14 and 26% and its detrimental contribution to the FATT can be estimated (Eq. 3) to range between 36 and 44 °C.

Boundary misorientation distributions and Grain size
EBSD maps reveal different boundary misorientation distributions, depending on the microstructure (Fig. 2a) most of them exhibiting a relatively high density of low angle boundaries.For polygonal ferrite these low angle boundaries result from the combination of the K-S orientation relationship and a variant selection during austenite transformation (Bengochea et al., 1998, Novillo et al., 2005).Consequently, even for simple ferrite microstructures, optically measured grain sizes can be significantly different from those obtained by EBSD (Bhattacharjee et al., 2004) unless the appropriate threshold angle and correlations are applied (Iza-Mendia and Gutiérrez, 2013).
The grain boundary misorientation distribution of a bainitic microstructure is the result of the displacive type of transformation.Low angle boundaries are mainly due to sub-units but also to some packet boundaries (Bhadeshia, 2001, Díaz-Fuentes et al., 2003).The boundary misorientation distribution has a characteristic shape, with a high density of low angle boundaries, almost no boundaries at intermediate misorientation angles and two peaks at Φ > 50°.The relative intensity of these peaks depends on the type of transformation: upper bainite, lower bainite or martensite (Zajac et al., 2005, Altuna andGutiérrez, 2005).
Low and high angle boundaries contribute to the strength, but groups of closely oriented grains make the crack to propagate on a coplanar manner defining the size of the cleavage unit (Gourgues et al., 2000).
A transverse section of the fracture surface analysed by EBSD in a sample with a bainitic microstructure (S1-300) is shown in Figure 2b.The traces of the {001} ferrite lattice planes are indicated, showing that, as expected, the cleavage crack propagates mainly on these planes.The crack deflects at each high angle boundary within the bainitic microstructure.The relation between the FATT and the cleavage unit size determined for the Nb-microalloyed steel by D = MED 15°, the EBSD mean equivalent diameter for a threshold angle (Φ) of 15° has been plot in Figure 2c for the different microstructures in Table 1.When comparing these results to those coming from C-Mn steels (Mintz et al., 1979, Zubialde et al., 2013) a systematic shift to higher transition temperatures results, excepting for S4-650 sample.

Nb-rich precipitates of different sizes
The precipitation strengthening found in Nbmicroalloyed steels is usually well described by Ashby-Orowan's equation (Gladman, 1997, Gladman 1999, Kestenbach, 1997).The extra strengthening for F-P microstructures beyond that expected from solutes and grain size (Hall-Petch) can be estimated using the following equation (Iza-Mendia et al., 2012): with σ yexp the experimental yield strength and the second term the application of a Pickering´s type of equation (Pickering, 1978;Pickering, 1993, Gutiérrez andAltuna, 2008) for which only Mn and Si are considered here given that the concentration of other substitutional elements is negligible in this steel.In equation (4) d = MED 2° and k HP is estimated according to Equation (1).The obtained results are shown in Figure 3a.
For F+P microstructures, (S1-650 to S4-650), the lowest Δσ y (~30 MPa) is obtained for the sequence S4-650.Reheating after quenching followed by a long holding stage at 900 °C induces full precipitation of Nb and precipitate coarsening.Nb-rich precipitates formed in austenite produce a moderate contribution to yield or tensile strength (Altuna et al., 2012).This explains the position close to C-Mn steels of the point for S4-650 in Figure 3c.
Nb precipitation in austenite reduces the achievable strengthening contribution.The precipitation strengthening contribution in themomechanically processed samples (S1-650 to S3-650) depends on the sequence and more specifically on the Nb in solution before transformation to ferrite.The precipitates formed in ferrite are extremely fine (<3 nm) and produce a precipitation strengthening propor tional to the free Nb in austenite before phase transformation (Herman et al., 1992) that according to Altuna et al., (2012) can be expressed by the following equation: with Nb free in (wt%).Calculating the exact contribution of Nb, Δσ yNb, requires using N free -corrected yield stress (leading to: Δσ yNb = Δσ y -5544N free ).In such a case, the slope of the linear regression (Eq.5) reduces to a value of about 1900 MPa per %Nb free (Altuna et al., 2012).
The S3-650 four pass sequence (Δσ y ≅ 50 MPa) with relatively low finishing temperature was designed for a strain induced precipitation in austenite greater than for S1-650 and S2-650 (Δσ y in the range 80-100 MPa).A model for precipitation in austenite (López, 2006) indicates that reheating at 1200 °C is not able to dissolve Nb beyond 0.025%.Substituting this value into Equation ( 5) gives the maximum precipitation strengthening potential of Nb around 90-95 MPa.Not far from the values estimated for Δσ y after S1 and S2-650 sequences.
The detrimental effect of precipitation strengthening on the FATT for F-P microstructures can be calculated using the following equation (Mintz et al., 1979): ΔFATT precipitates (°C) = 0.5Δσ y (6) Substituting the Δσ y values in Figure 3a into this equation, ΔFATT precipitates will range from ~10 to 45 °C, depending on the 650 °C sequence.

Transformation dislocation density
The presence of Nb in solution before transformation increases the hardenability of the steel, and quite frequently quasi-polygonal and non-polygonal ferrite microstructures are obtained in Nb-microalloyed steels under cooling conditions that would produce equiaxed ferrite in C-Mn steels, but here a low coiling temperature was deliberately applied to get bainite.A displacive transformation produces dislocated substructures with an additional strength that depends on transformation temperature (Bhadeshia, 2001) or tempering temperature (Malik and Lund, 1972) (Fig. 3b).For the two quenched and tempered samples (S5-300 and S5-650), the Δσ y from Equation (4) has also been plot.Given that, the martensite start temperature M s ≅ 440 °C determined by dilatometry is higher than the tempering temperature for (S5-300) the Δσ y has been assigned the Ms. Tempering at 650 °C produces a significant decrease of Δσ y which, is consistent with the expected behavior.
It is usually assumed that the dislocations resulting from cold working do not contribute too much to the FATT, but there is no information concerning the eventual effect of transformation dislocations present in bainite.As a first approximation, the same coefficient than the precipitation strengthening in Equation (6) was considered, leading to the following expression (Gutiérrez, 2013): ΔFATT bainite (°C) = 0.5Δσ y (7)

Free interstitial elements
The determination of the content of free interstitial elements as carbon and nitrogen in microalloyed steels is difficult.For ferrite-pearlite, the free N can be estimated by a combination of hot rolling and coiling models.For bainitic and Q&T microstructures, it would not be possible to dissociate the effect of interstitial elements from that of dislocations.Nevertheless, taking into account the coefficients usually assigned to this element (Gladman, 1971, Pickering, 1978;Pickering, 1993), a maximum contribution of 39 MPa to the yield strength and of 59 °C to the FATT results for all N dissolved in ferrite.

Application
Assuming that in the transition region the fracture energies in the ductile and brittle regions are both independent of the testing temperatu re and practically equal to the upper, Kv US , and lower, Kv LS , shelf energies (Todinov, 2001), a modified law of mixtures given by: can applied (Gutiérrez, 2013), with n = 0.7 and X d being the ductile fraction measured on the f racture surface, that can be expressed by the following empirical equation: with m = 2.5 and k that has been deduced for this steel to be a function of the cleavage unit size (in mm) of the form: The coefficients k and m define the slope of the Charpy curve in the transition and can vary depending on the steel.
The equation for the FATT (Gutiérrez, 2013) has now been modified by replacing the sum of the contribution from grain boundary carbides and from  has been deduced that expresses KV US as the sum of contributions to the FATT (Eq.11) other than grain size.
Equation (8) to Equation ( 12) together with Equation (4) are the basic equations of the proposed model that allows calculating the Charpy curves for the different microstructures in Table 1.The results shown in Figure 4 have been calculated using the following equation: FATT (°C) = (88 + ΔFATT-11Mn + 42Si + 15 (% pearlite) 1/3 + 0.5Δσ y -14(D) −0.5 (13) FIGURE 4. Experimental Charpy curves and predictions of the model (lines).Some FATT correction factor was required for some curves in order to obtain a better fit with experiment.This correction factor is indicated in Table 2.
which differs from Equation ( 11) in the term ΔFATT that is used here as an error term to express the deviation between the model predictions (Eq.11) and the experimental data.
The proposed model describes very well most of the experimental curves, as can be seen in Figure 4.A deviation happens for two F + P and one Q&T (S5-650) microstructures (Table 2).For these three cases, the predicted FATT requires a correction factor, ΔFATT, in an interval of ± 20 °C which is very reasonable taking into account the normal scatter for this parameter (Mintz et al., 1979).

CONCLUSIONS
The proposed model describes reasonably well the Charpy curves for a relatively broad range of microstructures (ferrite-pearlite, bainite and quenched and tempered) obtained in the same Nb-microalloyed steel.
This model is probably not of general application because some of the parameters defining the Charpy curve, for example the upper shelf impact energy that has been expressed as the sum of FATT terms, excepting grain size term, will certainly be affected by different correlations when changing the steel.Nevertheless, it constitutes an interesting tool for microstructural design and its formulation could open new perspectives.

FIGURE 1 .
FIGURE 1. Examples of the obtained microstructures.

FIGURE 2
FIGURE 2. a) EBSD boundary misorientation angle distribution; b) EBSD images showing IPF+IQ maps and detail of the indicated region showing IPF map, boundaries with misorientation >15° and {001} planes traces; c) relation between the experimental FATT and the cleavage unit size (Optical: MLI or EBSD: Q ≥15°) for C-Mn steels and the Nb-microalloyed steel in this work.
FIGURE 3. Extra yield strength (Δσ y ) beyond that expected from substutional solutes and grain size (Eq.5): a) as a function of the applied sequence and b) comparison for Q&T S5-300 and S5-650 between experimental and expected values.

TABLE 1 .
Summary of the applied thermal and thermomechanical methods to produce the different microstructures

TABLE 2 .
Evaluation of the deviation of the FATT from the model predictions (ΔFATT)