Ultramicroscopy 137 (2014) 40–47
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
An automated method of quantifying ferrite microstructures using electron backscatter diffraction (EBSD) data Sachin L. Shrestha a,n, Andrew J. Breen b, Patrick Trimby b, Gwénaëlle Proust c, Simon P. Ringer a,b, Julie M. Cairney a,b a
School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006, Australia Australian Centre for Microscopy and Microanalysis, The University of Sydney, NSW 2006, Australia c School of Civil Engineering, The University of Sydney, NSW 2006, Australia b
art ic l e i nf o
a b s t r a c t
Article history: Received 4 October 2013 Received in revised form 8 November 2013 Accepted 12 November 2013 Available online 21 November 2013
The identification and quantification of the different ferrite microconstituents in steels has long been a major challenge for metallurgists. Manual point counting from images obtained by optical and scanning electron microscopy (SEM) is commonly used for this purpose. While classification systems exist, the complexity of steel microstructures means that identifying and quantifying these phases is still a great challenge. Moreover, point counting is extremely tedious, time consuming, and subject to operator bias. This paper presents a new automated identification and quantification technique for the characterisation of complex ferrite microstructures by electron backscatter diffraction (EBSD). This technique takes advantage of the fact that different classes of ferrite exhibit preferential grain boundary misorientations, aspect ratios and mean misorientation, all of which can be detected using current EBSD software. These characteristics are set as criteria for identification and linked to grain size to determine the area fractions. The results of this method were evaluated by comparing the new automated technique with point counting results. The technique could easily be applied to a range of other steel microstructures. & 2013 Elsevier B.V. All rights reserved.
Keywords: Electron backscatter diffraction Data analysis HSLA steel Niobium
1. Introduction The characterisation and quantification of the different types of ferrite in steels is of great importance in materials engineering as the proportion and distribution of these ‘microconstituents’ strongly influence the steel's mechanical properties. Steel microstructures are extremely complex, and often consist of mixtures of different types of ferrite, such as acicular ferrite, polygonal ferrite and bainite. These structures have long been investigated using optical microscopy and scanning electron microscopy (SEM) and these studies have led to classification systems for the differentiation of steel microstructures, such as Dubé classification [1–3], but these systems do not allow for quantification. To truly understand the microstructure-property relationships for the complex microstructures of steels, quantitative measurements of the proportions of phases in the microstructures is required. The standard test method for determining volume fraction by systematic manual point counting (ASTM E562) [4,5] can be used to determine ferrite volume fractions, but this procedure is tedious, time consuming, and subject to operator bias and errors such as improper grid selection, sample quality, and incorrect measurements of fine ferrite microconstituents that are difficult to detect [6,7]. For these
n
Corresponding author. Tel.: þ 61 432990441; fax: þ 61 293517682. E-mail address: [email protected] (S.L. Shrestha).
0304-3991/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ultramic.2013.11.003
reasons, the identification and quantification of ferrite microconstituents in steels still remain a major challenge for researchers studying ferrous alloys. Electron backscatter diffraction (EBSD) has been proven to be beneficial to the microstructural analysis of steels [8–10]. This technique uses crystallographic orientation mapping and crystallographic/lattice structural differences to provide information such as grain size measurements and distribution, boundary characteristics, crystallographic orientations and their distributions (or texture), and phase identification. EBSD is often used to obtain crystallographic information that is combined with optical microscopy or X-ray diffraction to better identify constituent phases [11]. However the shortfall of EBSD for the quantification of steel microstructures is that many morphologies of ferrite have the same crystallographic structure. Indexing alone cannot then differentiate between different microconstituents of ferrite and thus their accurate identification and quantification remain difficult. In order to overcome this problem, previous studies have used diffraction pattern quality or index confidence to distinguish between different microconstituents [12–14]. The use of diffraction pattern quality for this purpose is based on the fact that different types of ferrite produced at different transformation temperatures have varying amounts of lattice defects. The diffraction pattern quality index is used to distinguish the degree of lattice imperfections and thus characterise the microstructure of selected high strength low alloy (HSLA) steels. Wu et al. [14] used diffraction pattern quality (referred as image
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
quality (IQ)) to separate different microconstituents and quantify them through a multi-peak model with reasonable results. Zaefferer et al. [15] used index confidence, which is similar to the image quality method but uses the fit between the diffraction pattern and indexing pattern to differentiate between ferrite microstructures. These techniques have been used on steels with fairly simple microstructures. However, a limitation is that more complex processing parameters will influence the amount of dislocations, and can distort the results obtained [13]. Another issue is that diffraction pattern quality is dependent on sample contamination and preparation and grain orientation [16,17], making comparisons between samples a challenge. Fractions of the microconstituents present obtained via these methods will also be different if obtained from different systems due to differences between the EBSD detectors' phosphorus screens, cameras, SEMs and software, on top of all the sample and operator variables. The technique presented in this paper employs the fact that different microstructures of ferrite exhibit preferential grain boundary misorientations, aspect ratios and mean misorientation, all of which can be detected using current EBSD systems. This combination of characteristics allows for automatic, quantitative measurement of the area fraction of each microconstituent. In this study the new method is applied to the ferrite microstructures of ultra thin strip cast (USC) steels produced by the CASTRIPs process [18–21]. These steels have an intricate microstructure consisting of acicular, polygonal, and bainitic ferrite [22,23]. Each of these ferrite microconstituents has distinctive characteristics in terms of grain size, different degrees of lattice imperfections
41
(i.e. dislocation or subgrain boundary density), grain boundary misorientation, and grain morphology. The proportion and distribution of each microconstituent influence the steel's mechanical properties; it is therefore important to obtain accurate and reliable data regarding the microstructure to be able to predict the mechanical properties. Polygonal ferrite is characterised by mainly equi-axed grains containing few low angle grain boundaries [2,24,25]. The presence of polygonal ferrite increases the ductility of steels, but if the grain size is unrefined it can result in lower strengths. Acicular ferrite consists of thin, lenticular plates that nucleate heterogeneously from non-metallic inclusions [2,26,27]. Their small grain size and high angle grain boundaries inhibit cleavage propagation and can increase toughness and strength. Bainite is in the form of very fine lenticular plates or laths that manifest as sheaves of ferrite plates that are usually separated by low-angle grain boundaries [2,26,28,29]. Bainitic microstructures contain a high density of dislocations that result in higher strength and lower ductility. With this new method, three distinctive characteristics (grain boundary misorientation, aspect ratio, and mean misorientation) are exploited to develop criteria for the identification of each ferrite grain. The grain size is then used to determine the area fraction of each ferrite microconstituent. The underlying Matlab code for this technique is freely available from the authors of this article under the working title ‘Ferrite Microstructure Quantifier’.
Table 1 Chemical composition (wt%) and processing parameters of the CASTRIP steels investigated. Specimen
Hot rolling reduction (%)
Hot rolling temperature (1C)
Coiling temperature (1C)
Nb
C
Mn
Si
N
Nb-free steel 04 Nb steel 08 Nb steel
12 12 38
879 895 897
544 547 567
o 0.001 0.041 0.084
0.034 0.036 0.031
0.93 0.87 0.83
0.2 0.21 0.2
0.008 0.007 0.006
Fig. 1. EBSD inverse pole figure maps of the microstructures of (a) Nb-free, (b) 04 Nb, and (c) 08 Nb steel hot rolled specimen along the rolling direction. Note these images represent a small portion of the actual dataset used.
42
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
2. Materials and methods 2.1. The materials and their processing The materials investigated in this study were as-hot-rolled low-carbon microalloyed ultra-thin cast strip (UCS) steels manufactured using the CASTRIPs process [19] containing o0.001, 0.041 and 0.084 wt% Nb (referred herein as the Nb-free, 04 Nb, and 08 Nb steels respectively). The chemical composition and processing parameters for these steels are listed in Table 1. Fig. 1 shows the microstructure of the three steels through EBSD orientation maps displaying grain boundaries with inverse pole figure colouring representing the orientation of grains. Nb is of particular interest in this steel as it decreases the transformation start temperature, thus favouring the formation of lowtemperature transformation products such as acicular and bainitic ferrite at the expense of polygonal ferrite [19,30–33]. 2.2. EBSD sample preparation and acquisition parameters Each specimen was sectioned so that a plane between the rolling direction and normal direction was analysed. The specimens were then mechanically ground by using 220–4000 grade sandpaper, polished with a 1 μm diamond suspension and then a 60 nm colloidal silica suspension. The samples were then electro-polished by using 4% perchloric acid in methanol as the electrolyte, which was cooled to approximately 35 1C, at a voltage of 28 V. The EBSD maps were collected using a Zeiss Ultra Plus field emission gun (FEG) SEM, equipped with an Oxford Instruments CHANNEL5 EBSD system with a Nordlys-S EBSD detector. The SEM was operating at 20 kV with an aperture of 60 μm and with the specimen tilted by 701. The parameters used for indexing were 7 bands of detection with 40 reflectors, 1 μm step size, using 1–4 frames of averaging and 8 8 binning for the large area scans.
3. Data analysis 3.1. Software parameters The data was analysed with a HKL Channel 5 system developed by Oxford Instruments and using MATLAB (Mathwork Inc.). The HKL Channel 5 system software consists of a data management module, Project Manager, and orientation map display software, Tango. A noise reduction function in Tango was employed to remove unindexed points (zero solutions) and spikes (misindexing). Noise reduction was used cautiously as overestimation can occur, increasing the average grain size estimation [34]. To minimise the use of
this function only scans with 483% indexing were used. Further noise and artefact reduction was conducted manually as described in Section 3.2. The grain area determination function in Tango was used to detect and analyse the characteristics of grains within the dataset. This function determines the position of all graindelimiting boundaries and calculates several of the grain's characteristics including area, equivalent circle diameter, aspect ratio, number of neighbours and the mean misorientation angle. The grain delimiting boundaries in this study are determined to have a minimum upper critical misorientation of 51 and a lower critical misorientation of 11, in order to detect low angle bainite lath grain boundaries. The upper critical misorientation is the set minimum angle required to wholly enclose a grain, whereas the lower critical misorientation is the minimum angle of a boundary that can be completed. These parameters were used with care, as reducing the minimal critical misorientation can incorrectly introduce sub-grain boundaries and noise in the results. 3.2. Analysis preparation via removal of artifacts Artefact removal is essential to improve data quality and steps need to be taken to eliminate these artifacts from the analysis. Standard noise reduction, described above, is conducted to remove zero solutions and isolated points that have been incorrectly indexed. Despite this, there are still artifacts or noise in the form of tiny grains/pixels detected by the grain area determination component. In this data, the pixel size is 1 μm2. The presence of these tiny grains/pixels is revealed by inspecting the aspect ratio distributions of detected grains within the dataset. For example in the Nb-free dataset, an analysis of grains of area 42 μm2 (42 pixels) revealed a spike in grains with an aspect ratio of 1.7 compared to the rest of the distribution as seen in Fig. 2a. Most of these grains are random pixels/artifacts produced by the EBSD scan. Using only grains of area 45 μm2 (4 5 pixels) results in a more expected aspect ratio distribution as shown in Fig. 2b. Very similar trends were shown for the 04 Nb and 08 Nb steel so grains of area r5 μm2 (which make up 7–9% of the overall area fraction) were removed in Matlab for all datasets.
4. Criteria for the classification of microconstituents Three criteria were used to distinguish the different microconstituents; grain boundary misorientation, aspect ratio and mean misorientation. The set of criteria used to classify each ferrite grain are provided in Table 2. Details of how each criterion was developed are provided in the following sections.
Fig. 2. Aspect ratio distribution curves for Nb-free steel microstructures with (a) grains of area 42 μm2, (b) grains of area 45 μm2. The red line represents the line of best fit for the data points represented by the blue marks. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
43
Table 2 Criteria for the identification of acicular, polygonal and bainitic ferrite. Microconstituents
Aspect ratio
Grain boundary angles
Acicular ferrite Polygonal ferrite
4 2.3 o 1.5 4 1.5
50–651
Mean misorientation
Grain size
Comments
o3
4150 μm
Excluding grains identified as acicular ferrite By difference
Bainitic
Fig. 3. EBSD inverse pole figure maps (a, c, and e) and band contrast maps (b, d, and f) of (a and b) bainite, (c and d) acicular ferrite, and (e and f) polygonal ferrite. Black, white and blue lines in the band contrast maps represents 5–201, 21–491, 50–651 grain boundaries respectively. Arrows indicate intragranular polygonal ferrite within the acicular ferrite microstructure. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4.1. Criterion 1: grain boundary misorientation Fig. 3 shows three different regions from the steels investigated in this study. The images across the top are EBSD inverse pole figure maps from regions that contain the three different types of ferrite that are commonly observed. Fig. 3a is bainite, Fig. 3c is acicular ferrite and Fig. 3e is polygonal ferrite. The images at the bottom are corresponding EBSD band contrast maps in which the grain boundaries have been coloured according to the misorientation (black¼5–201, white¼21–491, and blue ¼50–651) Fig. 4 shows the misorientation profile for regions containing predominantly one of the different microconstituents. These profiles were obtained using different regions of the Nb free steel. Very similar misorientation profiles were produced from bainite, acicular ferrite and polygonal ferrite regions in the 08 and 04 Nb steels. The noise in the profile for the polygonal ferrite is due to the limited number of grains scanned due to the large grain size (the size of the regions analysed was similar to those shown in Fig. 3). Bainite, acicular ferrite and polygonal ferrite all display distinct misorientation distributions. Similar grain boundary misorientation distributions were produced in a EBSD study by Park et al. [35] on steel microstructures containing predominately intragranular acicular and polygonal ferrite. The grain boundary misorientation distribution of both acicular ferrite and bainite are influenced by the Kurdjumov-Sachs orientation relationships with the parent austenite grain from which the acicular ferrite grains have nucleated [26,36,37]. The Kurdjumov-Sachs orientation relationship leads to a characteristic misorientation angle distribution usually described by having two peaks between 501 and 601, minimal misorientation between 351 and 451, and a large amount of misorientations below 251 [9].
Fig. 4. Grain boundary misorientation profiles of regions that contain predominantly bainite, acicular ferrite or polygonal ferrite. These profiles were obtained using the entire large scans realised on each type of steels.
The bainite regions have a very high proportion of low-angle grain boundaries with a strong peak in the misorientation profile below 101, along with some high angle grain boundaries, where two smaller peaks are present. The high proportion of low-angle grain boundaries is due to the morphology of bainite, which consists of packets of parallel ferrite plates or sub-grains connected by low-angle grain boundaries [2]. Bainite packets are usually separated by higher angle grain boundaries [36,38]. These features are shown in Fig. 3b where the sub-grains can be seen to contain grain boundaries in the 5–201 range as indicated by the black lines, and the grain boundaries between the packets mainly lie in the 50–651 range, indicated by the blue lines. Consistent with the low angle boundaries observed, the subgrains within these packets usually have similar crystallographic orientations [2], which is reflected in the gentle grading of colouring in Fig. 3a.
44
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
Acicular ferrite displays a higher proportion of high angle grain boundaries compared to bainite and polygonal ferrite, with two large peaks above 501, a significant number of low angle grain boundaries (with a significantly smaller peak in the 5–91 range and a obtuse hump between 91 and 141) and very few grain boundary misorientations between 201 and 501. The grain boundary misorientation distribution is reflected by the mainly blue (50–601) and black (5–201) boundaries in Fig. 3d. This high density of high angle boundaries contributes to the superior toughness of acicular ferrite [26,36], and is a consequence of acicular ferrite nucleating heterogeneously from non-metallic inclusions, creating a higher proportion of high angle grain boundaries, while maintaining an orientation relationship with the prior austenite grain through which it grew [26]. The distinct distribution of low angle grain boundaries is due to the idiomorphic intragranular polygonal ferrite mixed with the acicular ferrite in the sampled area of the microstructure as pointed out by the arrows in Fig. 3c. The intragranular polygonal ferrite nucleates from within the austenite grains on non-metallic inclusions and consists of several subgrains nucleated at the same inclusions, separated by low angle grain boundaries whist keeping high angle grain boundaries with respect to ferrite nucleated at other inclusions [39]. The intragranular polygonal ferrite present in Fig. 3c consists of low angle subgrain boundaries from 71 to 141 surrounded by high angle grain boundaries. The grain boundary misorientation distribution for polygonal ferrite shows more randomly oriented grains. This is also indicated by the random distribution of the black (5–201), white (21–491), and blue (50–641) colouring of the boundaries in Fig. 3f. The small peak at 5–71 can be attributed to a small number of low angle boundaries that were observed within some of the polygonal ferrite grains, in addition to the low angle grain boundaries that form randomly. The increase in the number of boundaries observed in the 50–651 range is thought to be due to sampling of neighbouring acicular ferrite grains, as few regions could be found with a reasonable number of grains to achieve a good sample that did not overlap slightly with neighbouring microconstituents. Fig. 5 compares the grain boundary misorientation distributions for the microstructures of the Nb-free, 04 and 08 Nb steels (as seen in Fig. 1). All three samples have clear peaks between 51 and 201, another two peaks between 501 and 601, and relatively few misorientations between 201 and 501. This is due to the high amounts of acicular ferrite and bainite within the microstructure that have Kurdjumov-Sachs orientation relationships with the parent austenite grain, which also explains the lack of misorientations between 201 and 501 [2,9,15,37]. These grain boundary misorientation distributions are similar to those obtained by Gourgues et al. [36], Dáz-Fuentes et al. [37] and Wang et al. [40], where their steels also contain significant amounts of acicular ferrite and bainite.
It is evident that the addition of Nb has an effect on the misorientation distribution, and this is closely linked to the variations in the proportion of different ferrite microconstituents in these samples. The addition of Nb correlates to an increase in the frequency of low angle grain boundaries between 51 and 121 and a reduction in high angle grain boundaries above 501 compared to the Nb-free sample. This is consistent with the Nb additions promoting the formation of bainite thus leading to less high-angle grain boundaries and more subgrains formed by low-angle grain boundaries. The Nbfree sample has a much smaller peak at the low-angle range and larger peaks above 501 than the 04 and 08 Nb steels, suggesting it has more acicular ferrite and less bainite (a conclusion that can be confirmed by a visual inspection of Fig. 1a). The grain boundary misorientation distribution of the Nb-free sample is very similar to the acicular ferrite distribution shown in Fig. 4. This is due to larger areas of the Nb-free sample consisting of acicular and intragranular polygonal ferrite (Figs. 1 and 3c and d) compared to the Nb steels. A slightly higher frequency of grain boundaries with misorientation in the 20– 501 range is attributed to the higher hot rolling reduction in the 08 Nb steel. Hot rolling reduces the static recrystallisation of the austenite grains, subsequently reducing their size, providing more randomly oriented boundaries at the vicinity of prior austenite boundaries. Repeated recrystallisation during hot rolling also reduces the orientation relationship with the prior austenite grains. It is clear that the misorientation distribution can provide some information about the proportions of the different types of ferrite. However, to quantify the area fraction of the microconstituents, grain sizes must be accounted for in the data analysis. A high portion of grain boundary misorientations that relate to a certain ferrite microconstituent does not necessarily imply a high area fraction of that microconstituent within the steel. For example, a large number of high angle grain boundaries may relate to a high number of acicular ferrite grains, but due to the small grain size, the area fraction of acicular ferrite may not be large. The grains for each specimen were reconstructed from the EBSD maps using the application Tango. A minimum upper critical misorientation angle of 51 with a lower critical misorientation angle of 11 (to connect incomplete boundaries) were used to define the grains. Data for each of the grains was collected, including the grain area. To determine the misorientation angles associated with each grain, subsets of the grains were created by increasing the critical misorientation from 71 to 651 in increments of 21. The data was then uploaded into Matlabs and a script was used to automatically identify and calculate the area fraction of the grains associated with each grain boundary misorientation interval for each steel dataset. However, as some of the microconstituents have shared grain boundary misorientation characteristics, misorientation cannot be used as the sole criteria to define a microconstituent. For example, polygonal ferrite does not have a distinct grain boundary misorientation distribution and therefore cannot be differentiated solely by grain boundary misorientations. Bainite packets are enclosed by high angle grain boundaries and thus many of these grains can be mistaken for acicular ferrite if only high angles boundaries were used to identify acicular ferrite. Additional criteria are required to further differentiate these grains. 4.2. Criterion 2: aspect ratio
Fig. 5. Grain boundary misorientation profiles of Nb-free, 04 Nb, and 08 Nb CASTRIP steel microstructures.
Fig. 3 shows that bainite, acicular ferrite and polygonal ferrite have distinct grain morphologies. Acicular ferrite grains have a lath-like or lenticular plate shape based on 2-D interpretations [26,27] with aspect ratios typically greater than 3 [41,42]. A study of the 3-D morphology of acicular ferrite by Wu [43] revealed that acicular ferrite grains vary from lath to plate shape with a wide range of aspect ratios from 1 to 11, although 95% of the grains
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
measured had an aspect ratio greater than 2. Polygonal ferrite grains are usually identified as coarse, blocky and somewhat equiaxed [24,25], therefore having low aspect ratios. Bainite consists of sheaves of parallel ferrite plates or laths with relatively high aspect ratios [28], although this is dependent on the transformation temperature [29]. Using the grain area determination component in the application Tango, the aspect ratio of the reconstructed grains can be calculated. Hence it is possible to utilise this characteristic in combination with grain boundary misorientation profiling to distinguish between the microconstituents and more accurately determine their relative area fraction. For example, the combination of high aspect ratio and high angle grain boundaries is an effective criterion to distinguish the acicular ferrite from the other microconstituents (see Table 2). The aspect ratio criteria helps to prevent blocky polygonal ferrite grains with high angle grain boundaries from being classified as acicular ferrite, which can occur if grain boundary misorientation distributions are used as the sole criteria. Because the bainitic laths are connected by low angle grain boundaries, these sub-grain boundaries were not always well-detected, meaning that bainite displayed a range of aspect ratios and thus could not be distinguished using this criterion. In Fig. 2b, a distribution of aspect ratios was produced from the Nb-free steel dataset (for grains of area 45 μm2). An asymmetrical curve indicates that more than one microconstituent is present with different mean aspect ratios. Acicular and polygonal ferrite grains are assumed to contribute significantly to the shape of the distribution as they have distinct aspect ratios. Whilst bainite will be detected, it is not thought to lead to peaks in the data as bainite displayed a range of aspect ratios for the reasons stated above. Two Gaussian distributions were constructed with their sum matching the overall aspect ratio distributions with peaks around
45
1.5 and 2.3. The distributions around the peaks at 1.5 and 2.3 are assumed to be linked with polygonal and acicular ferrite grains respectively due to these ferrites’ particular aspect ratio characteristics. Similar bimodal distributions with the same peaks were found in all three steels, as shown in Fig. 6a–c, indicating that the same microconstituents are contributing to the shape of the distribution. For acicular ferrite, we are therefore able to combine aspect ratio with grain boundary misorientation information. We deduce that the grains with boundary misorientations of 50–651 and aspect ratios greater than 2.3 are very likely to be acicular ferrite. With these two criteria, it is possible to detect a high percentage of the acicular ferrite. Some acicular ferrite grains may have aspect ratios less than 2.3 (for example when a needle is oriented perpendicular to the surface of the specimen), but this will be a small percentage of the grains detected and will have a negligible effect on the area fraction results. Polygonal ferrite grains are likely to have aspect ratios less than 1.5. However, some polygonal ferrite grains will exceed the 1.5 aspect ratio. In this case, the implication is serious as these grains can have large grain sizes and the incorrect detection of these grains will lead to a significant underestimation of the area fraction of polygonal ferrite. A mean misorientation criteria, described in the next section, can assist in differentiating these grains and thus improve the accuracy of the measurements. 4.3. Criterion 3: mean misorientation The mean misorientation angle is defined as the average misorientation within a given grain. It is determined by calculating the mean misorientation values between random pairs of pixels
Fig. 6. Aspect ratio distribution curves for (a) Nb-free, (b) 04 Nb, and (c) 08 Nb CASTRIP steel microstructures. The red curves represent the overall distribution of aspect ratios of each steel respectively whilst the green and black curves are Gaussian distributions with their sum matching the overall aspect ratio distributions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
46
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
Table 3 Comparison of calculated area fraction of acicular, polygonal and bainitic ferrite in the CASTRIP steels investigated by using point counting and the new technique. Specimen
Nb-free steel 04 Nb steel 08 Nb steel
New automated technique
Point counting
Acicular ferrite (%)
Polygonal ferrite (%)
Bainite (%)
Acicular ferrite (%)
Polygonal ferrite (%)
Bainite (%)
31 29 23
33 30 26
36 41 51
33 27 24
32 26 28
36 47 48
within that grain. The value is affected by two things: the extent of deformation (including low angle grain boundaries) within the grain, and the size of the grain. Increased deformation will increase the mean misorientation of the grain. However, the mean misorientation of small grains can be inaccurate as a small amount of deformation, such as dislocations and sub-grain boundaries, can significantly increase the grain's mean misorientation compared to larger grains. Therefore this criterion is not useful for very small grains. Large bainitic sub-grain clusters have a higher mean misorientation due to the sub-grain boundaries within the detected “grain” compared to large polygonal ferrite grains, which typically have low dislocation densities. Thus there is a prospect to utilise this characteristic to distinguish large polygonal ferrite grains with an aspect ratio greater than 1.5 from clusters of sheaths of bainite. An inspection of the large polygonal grains within the steel data sets reveals that most have a mean misorientation smaller than 31, while bainite subgrain clusters tend to have a mean misorientation greater than 31. Therefore a second criterion to define polygonal ferrite is a mean misorientation smaller than 31 and an aspect ratio greater than 1.5. In order to avoid the detection of small bainite subgrains with high aspect ratios and low mean misorientation; this criterion must be only applied to larger grains. An inspection of the bainite subgrains reveals that these grains generally do not exceed 150 μm. 4.4. Summary of criteria The selection of criteria for the identification and quantification of acicular, polygonal and bainitic ferrite was based on grain boundary misorientation, aspect ratio, mean misorientation, and grain size. Note that these criteria may need to be sample specific due to different material processing and alloying, which can significantly alter microstructural features and morphology. A preliminary investigation of the microstructural features for these samples is necessary to determine the best values for these criteria. In this study, the same parameters were chosen for all three samples for consistency and accuracy. Acicular ferrite can be distinguished through its high aspect ratio ( 42.3) and grain boundary misorientation (50–651). Polygonal ferrite can be distinguished by its low aspect ratio (o1.5) whilst large polygonal ferrite grains (4150 μm) with higher aspect ratios (4 1.5) are differentiated by their relatively low mean misorientation (o 31). With the accurate determination of the acicular and polygonal ferrite, the rest of the microstructure is assumed to be bainite. This approach provides an area fraction of each of the ferrite microconstituents in the steel's microstructure.
5. Area fraction results compared to point counting The grain boundary misorientation, aspect ratio, mean misorientation, and grain size criteria, as listed in Table 2 were used to identify grains of each of the microconstituents and the areas of these grains were added up accordingly. The area fractions obtained from the new automated technique are shown in
Table 3. Manual point counting, as described in [4,5] was also used to estimate the area fractions of these microconstituents and the results are compared in Table 3. The area fractions determined by the new automated technique and the manual point counting of the ferrite microconstituents correlate well, with a maximum of 6% difference. The Nb-bearing steels display a higher area fraction of bainite compared to the Nbfree steels, with the 08 Nb steel having the highest area fraction of bainite. The calculated reduction in polygonal ferrite in the Nbbearing steels compared to the Nb-free steels correlates to observations made by Xie et al. [33] using light optical microscopy and visual inspection of the EBSD maps. The 08 Nb steel displays the lowest fraction of acicular ferrite. This is expected as the higher hot rolling reduction in the processing of the 08 Nb steel results in the decrease of the amount of acicular ferrite in these steels [18]. The Nb-free steel ferrite area fraction results showed that almost even amounts of acicular, bainitic, and polygonal ferrite are present. Of the three samples used in this study, the 04 Nb steel shows the largest difference between the manual point counting and the new automated technique results for bainite and polygonal ferrite, with a discrepancy of 6% in the amount of bainite observed. The criteria selected for the identification of each microconstituent (Table 2) is critical to the minimisation of error. Possible errors associated with the new technique depend on selection of criteria as well as quality of data from the EBSD scan and sample, EBSD analysis area sizes, and the specific samples. The criteria selected in this study are suitable for the selected steels, but different steels may require specific criteria values. A simple investigation into the ferrite characteristics will enable selection of the best criteria. This technique gives a good estimation of the relative microconstituent area fraction for strip cast steels with different microalloying and processing parameters. The results can be utilised to establish microstructure–property relationships and provide insight on the effects of microalloying and processing on the steel microstructures.
6. Conclusion A new automated method using a combination of EBSD and Matlab has been developed to effectively identify and quantify ferrite microconstituents in the complex microstructures of various steel grades produced by the CASTRIPs process. The unique characteristics of the ferrite microconstituents were investigated and exploited to identify the type of ferrite for each grain, and this is linked to their associated grain's size for area fraction calculations. Acicular and bainitic ferrite have distinct grain boundary misorientation profiles with acicular ferrite displaying predominately high angle grain boundary misorientation at 50–651. Bainite displays predominately low angle grain boundary misorientations at 5–201. Acicular and polygonal ferrite grains display preferred aspect ratios. The criteria determined to identify acicular ferrite involves a grain boundary misorientation of 50–651 and aspect ratio Z2.3. Most polygonal ferrite is detected using the criteria that the aspect ratio is r 1.5 whilst large grains (Z150 μm) with
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
higher aspect ratios are deemed to be polygonal ferrite grains if the mean misorientation is r31. With the accurate determination of the acicular and polygonal ferrite the rest of microstructure is assumed to be bainite, since polygonal, acicular, and bainitic ferrite are the dominant microconstituents in the microstructures of USC steels. The technique provides good results confirmed by corresponding manual point counting results and c observations from previous studies. With a careful assessment of the choice of values used in the criteria, this approach could easily be applied to other steel microstructures, even where other non-ferritic phases are present, that can be distinguished by using EBSD for phase identification. Acknowledgements The work was funded by the Australian Research Council and BlueScope Steel Pty. Ltd. The authors acknowledge the facilities, and the scientific and technical assistance, of the Australian Microscopy and Microanalysis Research Facility at Sydney Microscopy and Microanalysis, at the University of Sydney. The authors thank Emily Wilkinson and Bonnie Simeonov for the sample preparation of the steel specimens that were analysed. References [1] C.A. Dube, PhD thesis, Metallurgical Engineering, Carnegie Institute of Technology, Pittsburgh, PA, 1948. [2] H.K.D.H. Bhadeshia, R.W.K. Honeycombe, Steels: Microstructure And Properties, Butterworth-Heinemann, Oxford, 2006. [3] G. Thewlis, Mater. Sci. Technol. 20 (2004) 143–160. [4] A. International, J.J. Friel, Practical Guide to Image Analysis, ASM International, OH, USA, 2000. [5] G.F. Vander Voort, Metallography: Principles and Practices, ASM International, OH, USA, 1999. [6] S.W. Russell, C.D. Lundin, Final Report, Volume 2, The Development of Qualification Standards for Cast Duplex Stainless Steel (No. DOE/ID/13975-2). Materials Joining Group, University of Tennessee, Knoxville, TN, 2005. [7] A. International, Practical Guide to Image Analysis, ASM International, OH, USA, 2000. [8] M. Díaz-Fuentes, A. Iza-Mendia, I. Gutiérrez, Metall. Mater. Trans. A 34 (2003) 2505–2516. [9] E. Novillo, D. Hernández, I. Gutiérrez, B. López, Mater. Sci. Eng. A 385 (2004) 83–90. [10] L. Ryde, Mater. Sci. Technol. 22 (2006) 1297–1306. [11] K. Davut, S. Zaefferer, Metall. Mater. Trans. A 41 (2010) 2187–2196.
47
[12] J. Wu, P.J. Wray, C.I. Garcia, M. Hua, A.J. DeArdo, Mater. Manuf. Process. 22 (2007) 281–285. [13] A.J. DeArdo, C.I. Garcia, K. Cho, M. Hua, Mater. Manuf. Process. 25 (2010) 33–40. [14] J. Wu, P.J. Wray, C.I. Garcia, M. Hua, A.J. Deardo, ISIJ Int. 45 (2005) 254–262. [15] S. Zaefferer, J. Ohlert, W. Bleck, Acta Mater. 52 (2004) 2765–2778. [16] V. Randle, Mater. Charact. 60 (2009) 913–922. [17] S. Zaefferer, P. Romano, F. Friedel, J. Microsc. 230 (2008) 499–508. [18] C.R. Killmore, A. Phillips, H.R. Kaul, P.C. Campbell, M. Schueren, J.R. Williams, W.N. Blejde, Mater. Forum 32 (2008) 13–28. [19] C.R. Killmore, H.R. Kaul, J.R. Burg, K.R. Carpenter, J.G. Williams, D.G. Edelman, P.C. Campbell, W. Blejde, in: 3rd International Conference on ThermoMechanical Processing of Steels, AIM, September 10–12, Padova, Italy, 2008. [20] C.R. Killmore, D.G. Edelman, K.R. Carpenter, H.R. Kaul, J.G. Williams, P.C. Campbell, W.N. Blejde, Mater. Sci. Forum 654 (2010) 198–201. [21] C. Killmore, K. Carpenter, H. Kaul, J. Williams, D. Edelman, P. Campbell, W. Blejde, Development of a Family of High strength Low Carbon Microalloyed Ultra-thin Cast Strip Products Produced by the CASTRIP® Process, in: MS&T 2010 Proceedings, AIST, Houston, USA, 2010. [22] Xie, Kelvin Y. Lan Yao, Chen Zhu, Julie, M. Cairney, Chris, R. Killmore, Frank, J. Barbaro, James, G. Williams, and Simon, P. Ringer. Metallurgical and Materials Transactions A 42, no. 8 (2011): 2199–2206. [23] F.A. Burgmann, Y. Xie, J.M. Cairney, S.P. Ringer, C. Killmore, F. Barbaro, J. Williams, Mater. Forum 32 (2008) 9–12. [24] G. Krauss, S.W. Thompson, ISIJ Int. 35 (1995) 937–945. [25] B.L. Bramfitt, Metallographer's Guide: Practices and Procedures for Irons and Steels, ASM International, Materials Park, Ohio, 2002. [26] H.K.D.H. Bhadeshia, Bainite in Steels: Transformations, Microstructure and Properties, IOM Communications, London, 2001. [27] O. Grong, D.K. Matlock, Int. Mater. Rev. 31 (1986) 27–48. [28] J. Wang, S. van der Zwaag, Z. Yang, H.-S. Fang, Mater. Lett. 45 (2000) 228–234. [29] S.B. Singh, H.K.D.H. Bhadeshia, Mater. Sci. Eng. A 245 (1998) 72–79. [30] Q. Yu, Z. Wang, X. Liu, G. Wang, Mater. Sci. Eng. A 379 (2004) 384–390. [31] J. Strid, K.E. Easterling, Acta Metall. 33 (1985) 2057–2074. [32] Y.-w. Wang, C. Feng, F.-y. Xu, B.-z. Bai, H.-s. Fang, J. Iron Steel Res. Int. 17 (2010) 49–53. [33] K. Xie, L. Yao, C. Zhu, J. Cairney, C. Killmore, F. Barbaro, J. Williams, S. Ringer, Metall. Mater. Trans. A 42 (2011) 2199–2206. [34] K.P. Mingard, B. Roebuck, E.G. Bennett, M. Thomas, B.P. Wynne, E.J. Palmiere, J. Microsc. 227 (2007) 298–308. [35] K.-T. Park, S.W. Hwang, J.H. Ji, C.H. Lee, Met. Mater. Int. 17 (2011) 349–356. [36] A.F. Gourgues, H. Flower, T. Lindley, Mater. Sci. Technol. 16 (2000) 26–40. [37] M. Dáz-Fuentes, A. Iza-Mendia, I. Gutierrez, Metall. Mater. Trans. A 34 (2003) 2505–2516. [38] A. Lambert-Perlade, T. Sturel, A. Gourgues, J. Besson, A. Pineau, Metall. Mater. Trans. A 35 (2004) 1039–1053. [39] L. Cheng, K. Wu, Acta Mater. 57 (2009) 3754–3762. [40] W. Wang, Y. Shan, K. Yang, Mater. Sci. Eng. A 502 (2009) 38–44. [41] R.A. Farrar, P.L. Harrison, J. Mater. Sci. 22 (1987) 3812–3820. [42] G. Magudeeswaran, V. Balasubramanian, S. Sathyanarayanan, G.M. Reddy, A. Moitra, S. Venugopal, G. Sasikala, J. Iron Steel Res. Int. 17 (2010) 51–56. [43] K.M. Wu, Scr. Mater. 54 (2006) 569–574.
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
An automated method of quantifying ferrite microstructures using electron backscatter diffraction (EBSD) data Sachin L. Shrestha a,n, Andrew J. Breen b, Patrick Trimby b, Gwénaëlle Proust c, Simon P. Ringer a,b, Julie M. Cairney a,b a
School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006, Australia Australian Centre for Microscopy and Microanalysis, The University of Sydney, NSW 2006, Australia c School of Civil Engineering, The University of Sydney, NSW 2006, Australia b
art ic l e i nf o
a b s t r a c t
Article history: Received 4 October 2013 Received in revised form 8 November 2013 Accepted 12 November 2013 Available online 21 November 2013
The identification and quantification of the different ferrite microconstituents in steels has long been a major challenge for metallurgists. Manual point counting from images obtained by optical and scanning electron microscopy (SEM) is commonly used for this purpose. While classification systems exist, the complexity of steel microstructures means that identifying and quantifying these phases is still a great challenge. Moreover, point counting is extremely tedious, time consuming, and subject to operator bias. This paper presents a new automated identification and quantification technique for the characterisation of complex ferrite microstructures by electron backscatter diffraction (EBSD). This technique takes advantage of the fact that different classes of ferrite exhibit preferential grain boundary misorientations, aspect ratios and mean misorientation, all of which can be detected using current EBSD software. These characteristics are set as criteria for identification and linked to grain size to determine the area fractions. The results of this method were evaluated by comparing the new automated technique with point counting results. The technique could easily be applied to a range of other steel microstructures. & 2013 Elsevier B.V. All rights reserved.
Keywords: Electron backscatter diffraction Data analysis HSLA steel Niobium
1. Introduction The characterisation and quantification of the different types of ferrite in steels is of great importance in materials engineering as the proportion and distribution of these ‘microconstituents’ strongly influence the steel's mechanical properties. Steel microstructures are extremely complex, and often consist of mixtures of different types of ferrite, such as acicular ferrite, polygonal ferrite and bainite. These structures have long been investigated using optical microscopy and scanning electron microscopy (SEM) and these studies have led to classification systems for the differentiation of steel microstructures, such as Dubé classification [1–3], but these systems do not allow for quantification. To truly understand the microstructure-property relationships for the complex microstructures of steels, quantitative measurements of the proportions of phases in the microstructures is required. The standard test method for determining volume fraction by systematic manual point counting (ASTM E562) [4,5] can be used to determine ferrite volume fractions, but this procedure is tedious, time consuming, and subject to operator bias and errors such as improper grid selection, sample quality, and incorrect measurements of fine ferrite microconstituents that are difficult to detect [6,7]. For these
n
Corresponding author. Tel.: þ 61 432990441; fax: þ 61 293517682. E-mail address: [email protected] (S.L. Shrestha).
0304-3991/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ultramic.2013.11.003
reasons, the identification and quantification of ferrite microconstituents in steels still remain a major challenge for researchers studying ferrous alloys. Electron backscatter diffraction (EBSD) has been proven to be beneficial to the microstructural analysis of steels [8–10]. This technique uses crystallographic orientation mapping and crystallographic/lattice structural differences to provide information such as grain size measurements and distribution, boundary characteristics, crystallographic orientations and their distributions (or texture), and phase identification. EBSD is often used to obtain crystallographic information that is combined with optical microscopy or X-ray diffraction to better identify constituent phases [11]. However the shortfall of EBSD for the quantification of steel microstructures is that many morphologies of ferrite have the same crystallographic structure. Indexing alone cannot then differentiate between different microconstituents of ferrite and thus their accurate identification and quantification remain difficult. In order to overcome this problem, previous studies have used diffraction pattern quality or index confidence to distinguish between different microconstituents [12–14]. The use of diffraction pattern quality for this purpose is based on the fact that different types of ferrite produced at different transformation temperatures have varying amounts of lattice defects. The diffraction pattern quality index is used to distinguish the degree of lattice imperfections and thus characterise the microstructure of selected high strength low alloy (HSLA) steels. Wu et al. [14] used diffraction pattern quality (referred as image
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
quality (IQ)) to separate different microconstituents and quantify them through a multi-peak model with reasonable results. Zaefferer et al. [15] used index confidence, which is similar to the image quality method but uses the fit between the diffraction pattern and indexing pattern to differentiate between ferrite microstructures. These techniques have been used on steels with fairly simple microstructures. However, a limitation is that more complex processing parameters will influence the amount of dislocations, and can distort the results obtained [13]. Another issue is that diffraction pattern quality is dependent on sample contamination and preparation and grain orientation [16,17], making comparisons between samples a challenge. Fractions of the microconstituents present obtained via these methods will also be different if obtained from different systems due to differences between the EBSD detectors' phosphorus screens, cameras, SEMs and software, on top of all the sample and operator variables. The technique presented in this paper employs the fact that different microstructures of ferrite exhibit preferential grain boundary misorientations, aspect ratios and mean misorientation, all of which can be detected using current EBSD systems. This combination of characteristics allows for automatic, quantitative measurement of the area fraction of each microconstituent. In this study the new method is applied to the ferrite microstructures of ultra thin strip cast (USC) steels produced by the CASTRIPs process [18–21]. These steels have an intricate microstructure consisting of acicular, polygonal, and bainitic ferrite [22,23]. Each of these ferrite microconstituents has distinctive characteristics in terms of grain size, different degrees of lattice imperfections
41
(i.e. dislocation or subgrain boundary density), grain boundary misorientation, and grain morphology. The proportion and distribution of each microconstituent influence the steel's mechanical properties; it is therefore important to obtain accurate and reliable data regarding the microstructure to be able to predict the mechanical properties. Polygonal ferrite is characterised by mainly equi-axed grains containing few low angle grain boundaries [2,24,25]. The presence of polygonal ferrite increases the ductility of steels, but if the grain size is unrefined it can result in lower strengths. Acicular ferrite consists of thin, lenticular plates that nucleate heterogeneously from non-metallic inclusions [2,26,27]. Their small grain size and high angle grain boundaries inhibit cleavage propagation and can increase toughness and strength. Bainite is in the form of very fine lenticular plates or laths that manifest as sheaves of ferrite plates that are usually separated by low-angle grain boundaries [2,26,28,29]. Bainitic microstructures contain a high density of dislocations that result in higher strength and lower ductility. With this new method, three distinctive characteristics (grain boundary misorientation, aspect ratio, and mean misorientation) are exploited to develop criteria for the identification of each ferrite grain. The grain size is then used to determine the area fraction of each ferrite microconstituent. The underlying Matlab code for this technique is freely available from the authors of this article under the working title ‘Ferrite Microstructure Quantifier’.
Table 1 Chemical composition (wt%) and processing parameters of the CASTRIP steels investigated. Specimen
Hot rolling reduction (%)
Hot rolling temperature (1C)
Coiling temperature (1C)
Nb
C
Mn
Si
N
Nb-free steel 04 Nb steel 08 Nb steel
12 12 38
879 895 897
544 547 567
o 0.001 0.041 0.084
0.034 0.036 0.031
0.93 0.87 0.83
0.2 0.21 0.2
0.008 0.007 0.006
Fig. 1. EBSD inverse pole figure maps of the microstructures of (a) Nb-free, (b) 04 Nb, and (c) 08 Nb steel hot rolled specimen along the rolling direction. Note these images represent a small portion of the actual dataset used.
42
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
2. Materials and methods 2.1. The materials and their processing The materials investigated in this study were as-hot-rolled low-carbon microalloyed ultra-thin cast strip (UCS) steels manufactured using the CASTRIPs process [19] containing o0.001, 0.041 and 0.084 wt% Nb (referred herein as the Nb-free, 04 Nb, and 08 Nb steels respectively). The chemical composition and processing parameters for these steels are listed in Table 1. Fig. 1 shows the microstructure of the three steels through EBSD orientation maps displaying grain boundaries with inverse pole figure colouring representing the orientation of grains. Nb is of particular interest in this steel as it decreases the transformation start temperature, thus favouring the formation of lowtemperature transformation products such as acicular and bainitic ferrite at the expense of polygonal ferrite [19,30–33]. 2.2. EBSD sample preparation and acquisition parameters Each specimen was sectioned so that a plane between the rolling direction and normal direction was analysed. The specimens were then mechanically ground by using 220–4000 grade sandpaper, polished with a 1 μm diamond suspension and then a 60 nm colloidal silica suspension. The samples were then electro-polished by using 4% perchloric acid in methanol as the electrolyte, which was cooled to approximately 35 1C, at a voltage of 28 V. The EBSD maps were collected using a Zeiss Ultra Plus field emission gun (FEG) SEM, equipped with an Oxford Instruments CHANNEL5 EBSD system with a Nordlys-S EBSD detector. The SEM was operating at 20 kV with an aperture of 60 μm and with the specimen tilted by 701. The parameters used for indexing were 7 bands of detection with 40 reflectors, 1 μm step size, using 1–4 frames of averaging and 8 8 binning for the large area scans.
3. Data analysis 3.1. Software parameters The data was analysed with a HKL Channel 5 system developed by Oxford Instruments and using MATLAB (Mathwork Inc.). The HKL Channel 5 system software consists of a data management module, Project Manager, and orientation map display software, Tango. A noise reduction function in Tango was employed to remove unindexed points (zero solutions) and spikes (misindexing). Noise reduction was used cautiously as overestimation can occur, increasing the average grain size estimation [34]. To minimise the use of
this function only scans with 483% indexing were used. Further noise and artefact reduction was conducted manually as described in Section 3.2. The grain area determination function in Tango was used to detect and analyse the characteristics of grains within the dataset. This function determines the position of all graindelimiting boundaries and calculates several of the grain's characteristics including area, equivalent circle diameter, aspect ratio, number of neighbours and the mean misorientation angle. The grain delimiting boundaries in this study are determined to have a minimum upper critical misorientation of 51 and a lower critical misorientation of 11, in order to detect low angle bainite lath grain boundaries. The upper critical misorientation is the set minimum angle required to wholly enclose a grain, whereas the lower critical misorientation is the minimum angle of a boundary that can be completed. These parameters were used with care, as reducing the minimal critical misorientation can incorrectly introduce sub-grain boundaries and noise in the results. 3.2. Analysis preparation via removal of artifacts Artefact removal is essential to improve data quality and steps need to be taken to eliminate these artifacts from the analysis. Standard noise reduction, described above, is conducted to remove zero solutions and isolated points that have been incorrectly indexed. Despite this, there are still artifacts or noise in the form of tiny grains/pixels detected by the grain area determination component. In this data, the pixel size is 1 μm2. The presence of these tiny grains/pixels is revealed by inspecting the aspect ratio distributions of detected grains within the dataset. For example in the Nb-free dataset, an analysis of grains of area 42 μm2 (42 pixels) revealed a spike in grains with an aspect ratio of 1.7 compared to the rest of the distribution as seen in Fig. 2a. Most of these grains are random pixels/artifacts produced by the EBSD scan. Using only grains of area 45 μm2 (4 5 pixels) results in a more expected aspect ratio distribution as shown in Fig. 2b. Very similar trends were shown for the 04 Nb and 08 Nb steel so grains of area r5 μm2 (which make up 7–9% of the overall area fraction) were removed in Matlab for all datasets.
4. Criteria for the classification of microconstituents Three criteria were used to distinguish the different microconstituents; grain boundary misorientation, aspect ratio and mean misorientation. The set of criteria used to classify each ferrite grain are provided in Table 2. Details of how each criterion was developed are provided in the following sections.
Fig. 2. Aspect ratio distribution curves for Nb-free steel microstructures with (a) grains of area 42 μm2, (b) grains of area 45 μm2. The red line represents the line of best fit for the data points represented by the blue marks. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
43
Table 2 Criteria for the identification of acicular, polygonal and bainitic ferrite. Microconstituents
Aspect ratio
Grain boundary angles
Acicular ferrite Polygonal ferrite
4 2.3 o 1.5 4 1.5
50–651
Mean misorientation
Grain size
Comments
o3
4150 μm
Excluding grains identified as acicular ferrite By difference
Bainitic
Fig. 3. EBSD inverse pole figure maps (a, c, and e) and band contrast maps (b, d, and f) of (a and b) bainite, (c and d) acicular ferrite, and (e and f) polygonal ferrite. Black, white and blue lines in the band contrast maps represents 5–201, 21–491, 50–651 grain boundaries respectively. Arrows indicate intragranular polygonal ferrite within the acicular ferrite microstructure. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4.1. Criterion 1: grain boundary misorientation Fig. 3 shows three different regions from the steels investigated in this study. The images across the top are EBSD inverse pole figure maps from regions that contain the three different types of ferrite that are commonly observed. Fig. 3a is bainite, Fig. 3c is acicular ferrite and Fig. 3e is polygonal ferrite. The images at the bottom are corresponding EBSD band contrast maps in which the grain boundaries have been coloured according to the misorientation (black¼5–201, white¼21–491, and blue ¼50–651) Fig. 4 shows the misorientation profile for regions containing predominantly one of the different microconstituents. These profiles were obtained using different regions of the Nb free steel. Very similar misorientation profiles were produced from bainite, acicular ferrite and polygonal ferrite regions in the 08 and 04 Nb steels. The noise in the profile for the polygonal ferrite is due to the limited number of grains scanned due to the large grain size (the size of the regions analysed was similar to those shown in Fig. 3). Bainite, acicular ferrite and polygonal ferrite all display distinct misorientation distributions. Similar grain boundary misorientation distributions were produced in a EBSD study by Park et al. [35] on steel microstructures containing predominately intragranular acicular and polygonal ferrite. The grain boundary misorientation distribution of both acicular ferrite and bainite are influenced by the Kurdjumov-Sachs orientation relationships with the parent austenite grain from which the acicular ferrite grains have nucleated [26,36,37]. The Kurdjumov-Sachs orientation relationship leads to a characteristic misorientation angle distribution usually described by having two peaks between 501 and 601, minimal misorientation between 351 and 451, and a large amount of misorientations below 251 [9].
Fig. 4. Grain boundary misorientation profiles of regions that contain predominantly bainite, acicular ferrite or polygonal ferrite. These profiles were obtained using the entire large scans realised on each type of steels.
The bainite regions have a very high proportion of low-angle grain boundaries with a strong peak in the misorientation profile below 101, along with some high angle grain boundaries, where two smaller peaks are present. The high proportion of low-angle grain boundaries is due to the morphology of bainite, which consists of packets of parallel ferrite plates or sub-grains connected by low-angle grain boundaries [2]. Bainite packets are usually separated by higher angle grain boundaries [36,38]. These features are shown in Fig. 3b where the sub-grains can be seen to contain grain boundaries in the 5–201 range as indicated by the black lines, and the grain boundaries between the packets mainly lie in the 50–651 range, indicated by the blue lines. Consistent with the low angle boundaries observed, the subgrains within these packets usually have similar crystallographic orientations [2], which is reflected in the gentle grading of colouring in Fig. 3a.
44
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
Acicular ferrite displays a higher proportion of high angle grain boundaries compared to bainite and polygonal ferrite, with two large peaks above 501, a significant number of low angle grain boundaries (with a significantly smaller peak in the 5–91 range and a obtuse hump between 91 and 141) and very few grain boundary misorientations between 201 and 501. The grain boundary misorientation distribution is reflected by the mainly blue (50–601) and black (5–201) boundaries in Fig. 3d. This high density of high angle boundaries contributes to the superior toughness of acicular ferrite [26,36], and is a consequence of acicular ferrite nucleating heterogeneously from non-metallic inclusions, creating a higher proportion of high angle grain boundaries, while maintaining an orientation relationship with the prior austenite grain through which it grew [26]. The distinct distribution of low angle grain boundaries is due to the idiomorphic intragranular polygonal ferrite mixed with the acicular ferrite in the sampled area of the microstructure as pointed out by the arrows in Fig. 3c. The intragranular polygonal ferrite nucleates from within the austenite grains on non-metallic inclusions and consists of several subgrains nucleated at the same inclusions, separated by low angle grain boundaries whist keeping high angle grain boundaries with respect to ferrite nucleated at other inclusions [39]. The intragranular polygonal ferrite present in Fig. 3c consists of low angle subgrain boundaries from 71 to 141 surrounded by high angle grain boundaries. The grain boundary misorientation distribution for polygonal ferrite shows more randomly oriented grains. This is also indicated by the random distribution of the black (5–201), white (21–491), and blue (50–641) colouring of the boundaries in Fig. 3f. The small peak at 5–71 can be attributed to a small number of low angle boundaries that were observed within some of the polygonal ferrite grains, in addition to the low angle grain boundaries that form randomly. The increase in the number of boundaries observed in the 50–651 range is thought to be due to sampling of neighbouring acicular ferrite grains, as few regions could be found with a reasonable number of grains to achieve a good sample that did not overlap slightly with neighbouring microconstituents. Fig. 5 compares the grain boundary misorientation distributions for the microstructures of the Nb-free, 04 and 08 Nb steels (as seen in Fig. 1). All three samples have clear peaks between 51 and 201, another two peaks between 501 and 601, and relatively few misorientations between 201 and 501. This is due to the high amounts of acicular ferrite and bainite within the microstructure that have Kurdjumov-Sachs orientation relationships with the parent austenite grain, which also explains the lack of misorientations between 201 and 501 [2,9,15,37]. These grain boundary misorientation distributions are similar to those obtained by Gourgues et al. [36], Dáz-Fuentes et al. [37] and Wang et al. [40], where their steels also contain significant amounts of acicular ferrite and bainite.
It is evident that the addition of Nb has an effect on the misorientation distribution, and this is closely linked to the variations in the proportion of different ferrite microconstituents in these samples. The addition of Nb correlates to an increase in the frequency of low angle grain boundaries between 51 and 121 and a reduction in high angle grain boundaries above 501 compared to the Nb-free sample. This is consistent with the Nb additions promoting the formation of bainite thus leading to less high-angle grain boundaries and more subgrains formed by low-angle grain boundaries. The Nbfree sample has a much smaller peak at the low-angle range and larger peaks above 501 than the 04 and 08 Nb steels, suggesting it has more acicular ferrite and less bainite (a conclusion that can be confirmed by a visual inspection of Fig. 1a). The grain boundary misorientation distribution of the Nb-free sample is very similar to the acicular ferrite distribution shown in Fig. 4. This is due to larger areas of the Nb-free sample consisting of acicular and intragranular polygonal ferrite (Figs. 1 and 3c and d) compared to the Nb steels. A slightly higher frequency of grain boundaries with misorientation in the 20– 501 range is attributed to the higher hot rolling reduction in the 08 Nb steel. Hot rolling reduces the static recrystallisation of the austenite grains, subsequently reducing their size, providing more randomly oriented boundaries at the vicinity of prior austenite boundaries. Repeated recrystallisation during hot rolling also reduces the orientation relationship with the prior austenite grains. It is clear that the misorientation distribution can provide some information about the proportions of the different types of ferrite. However, to quantify the area fraction of the microconstituents, grain sizes must be accounted for in the data analysis. A high portion of grain boundary misorientations that relate to a certain ferrite microconstituent does not necessarily imply a high area fraction of that microconstituent within the steel. For example, a large number of high angle grain boundaries may relate to a high number of acicular ferrite grains, but due to the small grain size, the area fraction of acicular ferrite may not be large. The grains for each specimen were reconstructed from the EBSD maps using the application Tango. A minimum upper critical misorientation angle of 51 with a lower critical misorientation angle of 11 (to connect incomplete boundaries) were used to define the grains. Data for each of the grains was collected, including the grain area. To determine the misorientation angles associated with each grain, subsets of the grains were created by increasing the critical misorientation from 71 to 651 in increments of 21. The data was then uploaded into Matlabs and a script was used to automatically identify and calculate the area fraction of the grains associated with each grain boundary misorientation interval for each steel dataset. However, as some of the microconstituents have shared grain boundary misorientation characteristics, misorientation cannot be used as the sole criteria to define a microconstituent. For example, polygonal ferrite does not have a distinct grain boundary misorientation distribution and therefore cannot be differentiated solely by grain boundary misorientations. Bainite packets are enclosed by high angle grain boundaries and thus many of these grains can be mistaken for acicular ferrite if only high angles boundaries were used to identify acicular ferrite. Additional criteria are required to further differentiate these grains. 4.2. Criterion 2: aspect ratio
Fig. 5. Grain boundary misorientation profiles of Nb-free, 04 Nb, and 08 Nb CASTRIP steel microstructures.
Fig. 3 shows that bainite, acicular ferrite and polygonal ferrite have distinct grain morphologies. Acicular ferrite grains have a lath-like or lenticular plate shape based on 2-D interpretations [26,27] with aspect ratios typically greater than 3 [41,42]. A study of the 3-D morphology of acicular ferrite by Wu [43] revealed that acicular ferrite grains vary from lath to plate shape with a wide range of aspect ratios from 1 to 11, although 95% of the grains
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
measured had an aspect ratio greater than 2. Polygonal ferrite grains are usually identified as coarse, blocky and somewhat equiaxed [24,25], therefore having low aspect ratios. Bainite consists of sheaves of parallel ferrite plates or laths with relatively high aspect ratios [28], although this is dependent on the transformation temperature [29]. Using the grain area determination component in the application Tango, the aspect ratio of the reconstructed grains can be calculated. Hence it is possible to utilise this characteristic in combination with grain boundary misorientation profiling to distinguish between the microconstituents and more accurately determine their relative area fraction. For example, the combination of high aspect ratio and high angle grain boundaries is an effective criterion to distinguish the acicular ferrite from the other microconstituents (see Table 2). The aspect ratio criteria helps to prevent blocky polygonal ferrite grains with high angle grain boundaries from being classified as acicular ferrite, which can occur if grain boundary misorientation distributions are used as the sole criteria. Because the bainitic laths are connected by low angle grain boundaries, these sub-grain boundaries were not always well-detected, meaning that bainite displayed a range of aspect ratios and thus could not be distinguished using this criterion. In Fig. 2b, a distribution of aspect ratios was produced from the Nb-free steel dataset (for grains of area 45 μm2). An asymmetrical curve indicates that more than one microconstituent is present with different mean aspect ratios. Acicular and polygonal ferrite grains are assumed to contribute significantly to the shape of the distribution as they have distinct aspect ratios. Whilst bainite will be detected, it is not thought to lead to peaks in the data as bainite displayed a range of aspect ratios for the reasons stated above. Two Gaussian distributions were constructed with their sum matching the overall aspect ratio distributions with peaks around
45
1.5 and 2.3. The distributions around the peaks at 1.5 and 2.3 are assumed to be linked with polygonal and acicular ferrite grains respectively due to these ferrites’ particular aspect ratio characteristics. Similar bimodal distributions with the same peaks were found in all three steels, as shown in Fig. 6a–c, indicating that the same microconstituents are contributing to the shape of the distribution. For acicular ferrite, we are therefore able to combine aspect ratio with grain boundary misorientation information. We deduce that the grains with boundary misorientations of 50–651 and aspect ratios greater than 2.3 are very likely to be acicular ferrite. With these two criteria, it is possible to detect a high percentage of the acicular ferrite. Some acicular ferrite grains may have aspect ratios less than 2.3 (for example when a needle is oriented perpendicular to the surface of the specimen), but this will be a small percentage of the grains detected and will have a negligible effect on the area fraction results. Polygonal ferrite grains are likely to have aspect ratios less than 1.5. However, some polygonal ferrite grains will exceed the 1.5 aspect ratio. In this case, the implication is serious as these grains can have large grain sizes and the incorrect detection of these grains will lead to a significant underestimation of the area fraction of polygonal ferrite. A mean misorientation criteria, described in the next section, can assist in differentiating these grains and thus improve the accuracy of the measurements. 4.3. Criterion 3: mean misorientation The mean misorientation angle is defined as the average misorientation within a given grain. It is determined by calculating the mean misorientation values between random pairs of pixels
Fig. 6. Aspect ratio distribution curves for (a) Nb-free, (b) 04 Nb, and (c) 08 Nb CASTRIP steel microstructures. The red curves represent the overall distribution of aspect ratios of each steel respectively whilst the green and black curves are Gaussian distributions with their sum matching the overall aspect ratio distributions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
46
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
Table 3 Comparison of calculated area fraction of acicular, polygonal and bainitic ferrite in the CASTRIP steels investigated by using point counting and the new technique. Specimen
Nb-free steel 04 Nb steel 08 Nb steel
New automated technique
Point counting
Acicular ferrite (%)
Polygonal ferrite (%)
Bainite (%)
Acicular ferrite (%)
Polygonal ferrite (%)
Bainite (%)
31 29 23
33 30 26
36 41 51
33 27 24
32 26 28
36 47 48
within that grain. The value is affected by two things: the extent of deformation (including low angle grain boundaries) within the grain, and the size of the grain. Increased deformation will increase the mean misorientation of the grain. However, the mean misorientation of small grains can be inaccurate as a small amount of deformation, such as dislocations and sub-grain boundaries, can significantly increase the grain's mean misorientation compared to larger grains. Therefore this criterion is not useful for very small grains. Large bainitic sub-grain clusters have a higher mean misorientation due to the sub-grain boundaries within the detected “grain” compared to large polygonal ferrite grains, which typically have low dislocation densities. Thus there is a prospect to utilise this characteristic to distinguish large polygonal ferrite grains with an aspect ratio greater than 1.5 from clusters of sheaths of bainite. An inspection of the large polygonal grains within the steel data sets reveals that most have a mean misorientation smaller than 31, while bainite subgrain clusters tend to have a mean misorientation greater than 31. Therefore a second criterion to define polygonal ferrite is a mean misorientation smaller than 31 and an aspect ratio greater than 1.5. In order to avoid the detection of small bainite subgrains with high aspect ratios and low mean misorientation; this criterion must be only applied to larger grains. An inspection of the bainite subgrains reveals that these grains generally do not exceed 150 μm. 4.4. Summary of criteria The selection of criteria for the identification and quantification of acicular, polygonal and bainitic ferrite was based on grain boundary misorientation, aspect ratio, mean misorientation, and grain size. Note that these criteria may need to be sample specific due to different material processing and alloying, which can significantly alter microstructural features and morphology. A preliminary investigation of the microstructural features for these samples is necessary to determine the best values for these criteria. In this study, the same parameters were chosen for all three samples for consistency and accuracy. Acicular ferrite can be distinguished through its high aspect ratio ( 42.3) and grain boundary misorientation (50–651). Polygonal ferrite can be distinguished by its low aspect ratio (o1.5) whilst large polygonal ferrite grains (4150 μm) with higher aspect ratios (4 1.5) are differentiated by their relatively low mean misorientation (o 31). With the accurate determination of the acicular and polygonal ferrite, the rest of the microstructure is assumed to be bainite. This approach provides an area fraction of each of the ferrite microconstituents in the steel's microstructure.
5. Area fraction results compared to point counting The grain boundary misorientation, aspect ratio, mean misorientation, and grain size criteria, as listed in Table 2 were used to identify grains of each of the microconstituents and the areas of these grains were added up accordingly. The area fractions obtained from the new automated technique are shown in
Table 3. Manual point counting, as described in [4,5] was also used to estimate the area fractions of these microconstituents and the results are compared in Table 3. The area fractions determined by the new automated technique and the manual point counting of the ferrite microconstituents correlate well, with a maximum of 6% difference. The Nb-bearing steels display a higher area fraction of bainite compared to the Nbfree steels, with the 08 Nb steel having the highest area fraction of bainite. The calculated reduction in polygonal ferrite in the Nbbearing steels compared to the Nb-free steels correlates to observations made by Xie et al. [33] using light optical microscopy and visual inspection of the EBSD maps. The 08 Nb steel displays the lowest fraction of acicular ferrite. This is expected as the higher hot rolling reduction in the processing of the 08 Nb steel results in the decrease of the amount of acicular ferrite in these steels [18]. The Nb-free steel ferrite area fraction results showed that almost even amounts of acicular, bainitic, and polygonal ferrite are present. Of the three samples used in this study, the 04 Nb steel shows the largest difference between the manual point counting and the new automated technique results for bainite and polygonal ferrite, with a discrepancy of 6% in the amount of bainite observed. The criteria selected for the identification of each microconstituent (Table 2) is critical to the minimisation of error. Possible errors associated with the new technique depend on selection of criteria as well as quality of data from the EBSD scan and sample, EBSD analysis area sizes, and the specific samples. The criteria selected in this study are suitable for the selected steels, but different steels may require specific criteria values. A simple investigation into the ferrite characteristics will enable selection of the best criteria. This technique gives a good estimation of the relative microconstituent area fraction for strip cast steels with different microalloying and processing parameters. The results can be utilised to establish microstructure–property relationships and provide insight on the effects of microalloying and processing on the steel microstructures.
6. Conclusion A new automated method using a combination of EBSD and Matlab has been developed to effectively identify and quantify ferrite microconstituents in the complex microstructures of various steel grades produced by the CASTRIPs process. The unique characteristics of the ferrite microconstituents were investigated and exploited to identify the type of ferrite for each grain, and this is linked to their associated grain's size for area fraction calculations. Acicular and bainitic ferrite have distinct grain boundary misorientation profiles with acicular ferrite displaying predominately high angle grain boundary misorientation at 50–651. Bainite displays predominately low angle grain boundary misorientations at 5–201. Acicular and polygonal ferrite grains display preferred aspect ratios. The criteria determined to identify acicular ferrite involves a grain boundary misorientation of 50–651 and aspect ratio Z2.3. Most polygonal ferrite is detected using the criteria that the aspect ratio is r 1.5 whilst large grains (Z150 μm) with
S.L. Shrestha et al. / Ultramicroscopy 137 (2014) 40–47
higher aspect ratios are deemed to be polygonal ferrite grains if the mean misorientation is r31. With the accurate determination of the acicular and polygonal ferrite the rest of microstructure is assumed to be bainite, since polygonal, acicular, and bainitic ferrite are the dominant microconstituents in the microstructures of USC steels. The technique provides good results confirmed by corresponding manual point counting results and c observations from previous studies. With a careful assessment of the choice of values used in the criteria, this approach could easily be applied to other steel microstructures, even where other non-ferritic phases are present, that can be distinguished by using EBSD for phase identification. Acknowledgements The work was funded by the Australian Research Council and BlueScope Steel Pty. Ltd. The authors acknowledge the facilities, and the scientific and technical assistance, of the Australian Microscopy and Microanalysis Research Facility at Sydney Microscopy and Microanalysis, at the University of Sydney. The authors thank Emily Wilkinson and Bonnie Simeonov for the sample preparation of the steel specimens that were analysed. References [1] C.A. Dube, PhD thesis, Metallurgical Engineering, Carnegie Institute of Technology, Pittsburgh, PA, 1948. [2] H.K.D.H. Bhadeshia, R.W.K. Honeycombe, Steels: Microstructure And Properties, Butterworth-Heinemann, Oxford, 2006. [3] G. Thewlis, Mater. Sci. Technol. 20 (2004) 143–160. [4] A. International, J.J. Friel, Practical Guide to Image Analysis, ASM International, OH, USA, 2000. [5] G.F. Vander Voort, Metallography: Principles and Practices, ASM International, OH, USA, 1999. [6] S.W. Russell, C.D. Lundin, Final Report, Volume 2, The Development of Qualification Standards for Cast Duplex Stainless Steel (No. DOE/ID/13975-2). Materials Joining Group, University of Tennessee, Knoxville, TN, 2005. [7] A. International, Practical Guide to Image Analysis, ASM International, OH, USA, 2000. [8] M. Díaz-Fuentes, A. Iza-Mendia, I. Gutiérrez, Metall. Mater. Trans. A 34 (2003) 2505–2516. [9] E. Novillo, D. Hernández, I. Gutiérrez, B. López, Mater. Sci. Eng. A 385 (2004) 83–90. [10] L. Ryde, Mater. Sci. Technol. 22 (2006) 1297–1306. [11] K. Davut, S. Zaefferer, Metall. Mater. Trans. A 41 (2010) 2187–2196.
47
[12] J. Wu, P.J. Wray, C.I. Garcia, M. Hua, A.J. DeArdo, Mater. Manuf. Process. 22 (2007) 281–285. [13] A.J. DeArdo, C.I. Garcia, K. Cho, M. Hua, Mater. Manuf. Process. 25 (2010) 33–40. [14] J. Wu, P.J. Wray, C.I. Garcia, M. Hua, A.J. Deardo, ISIJ Int. 45 (2005) 254–262. [15] S. Zaefferer, J. Ohlert, W. Bleck, Acta Mater. 52 (2004) 2765–2778. [16] V. Randle, Mater. Charact. 60 (2009) 913–922. [17] S. Zaefferer, P. Romano, F. Friedel, J. Microsc. 230 (2008) 499–508. [18] C.R. Killmore, A. Phillips, H.R. Kaul, P.C. Campbell, M. Schueren, J.R. Williams, W.N. Blejde, Mater. Forum 32 (2008) 13–28. [19] C.R. Killmore, H.R. Kaul, J.R. Burg, K.R. Carpenter, J.G. Williams, D.G. Edelman, P.C. Campbell, W. Blejde, in: 3rd International Conference on ThermoMechanical Processing of Steels, AIM, September 10–12, Padova, Italy, 2008. [20] C.R. Killmore, D.G. Edelman, K.R. Carpenter, H.R. Kaul, J.G. Williams, P.C. Campbell, W.N. Blejde, Mater. Sci. Forum 654 (2010) 198–201. [21] C. Killmore, K. Carpenter, H. Kaul, J. Williams, D. Edelman, P. Campbell, W. Blejde, Development of a Family of High strength Low Carbon Microalloyed Ultra-thin Cast Strip Products Produced by the CASTRIP® Process, in: MS&T 2010 Proceedings, AIST, Houston, USA, 2010. [22] Xie, Kelvin Y. Lan Yao, Chen Zhu, Julie, M. Cairney, Chris, R. Killmore, Frank, J. Barbaro, James, G. Williams, and Simon, P. Ringer. Metallurgical and Materials Transactions A 42, no. 8 (2011): 2199–2206. [23] F.A. Burgmann, Y. Xie, J.M. Cairney, S.P. Ringer, C. Killmore, F. Barbaro, J. Williams, Mater. Forum 32 (2008) 9–12. [24] G. Krauss, S.W. Thompson, ISIJ Int. 35 (1995) 937–945. [25] B.L. Bramfitt, Metallographer's Guide: Practices and Procedures for Irons and Steels, ASM International, Materials Park, Ohio, 2002. [26] H.K.D.H. Bhadeshia, Bainite in Steels: Transformations, Microstructure and Properties, IOM Communications, London, 2001. [27] O. Grong, D.K. Matlock, Int. Mater. Rev. 31 (1986) 27–48. [28] J. Wang, S. van der Zwaag, Z. Yang, H.-S. Fang, Mater. Lett. 45 (2000) 228–234. [29] S.B. Singh, H.K.D.H. Bhadeshia, Mater. Sci. Eng. A 245 (1998) 72–79. [30] Q. Yu, Z. Wang, X. Liu, G. Wang, Mater. Sci. Eng. A 379 (2004) 384–390. [31] J. Strid, K.E. Easterling, Acta Metall. 33 (1985) 2057–2074. [32] Y.-w. Wang, C. Feng, F.-y. Xu, B.-z. Bai, H.-s. Fang, J. Iron Steel Res. Int. 17 (2010) 49–53. [33] K. Xie, L. Yao, C. Zhu, J. Cairney, C. Killmore, F. Barbaro, J. Williams, S. Ringer, Metall. Mater. Trans. A 42 (2011) 2199–2206. [34] K.P. Mingard, B. Roebuck, E.G. Bennett, M. Thomas, B.P. Wynne, E.J. Palmiere, J. Microsc. 227 (2007) 298–308. [35] K.-T. Park, S.W. Hwang, J.H. Ji, C.H. Lee, Met. Mater. Int. 17 (2011) 349–356. [36] A.F. Gourgues, H. Flower, T. Lindley, Mater. Sci. Technol. 16 (2000) 26–40. [37] M. Dáz-Fuentes, A. Iza-Mendia, I. Gutierrez, Metall. Mater. Trans. A 34 (2003) 2505–2516. [38] A. Lambert-Perlade, T. Sturel, A. Gourgues, J. Besson, A. Pineau, Metall. Mater. Trans. A 35 (2004) 1039–1053. [39] L. Cheng, K. Wu, Acta Mater. 57 (2009) 3754–3762. [40] W. Wang, Y. Shan, K. Yang, Mater. Sci. Eng. A 502 (2009) 38–44. [41] R.A. Farrar, P.L. Harrison, J. Mater. Sci. 22 (1987) 3812–3820. [42] G. Magudeeswaran, V. Balasubramanian, S. Sathyanarayanan, G.M. Reddy, A. Moitra, S. Venugopal, G. Sasikala, J. Iron Steel Res. Int. 17 (2010) 51–56. [43] K.M. Wu, Scr. Mater. 54 (2006) 569–574.
