Professional Documents
Culture Documents
Materials Science & Engineering A: B.L. Ennis, C. Bos, M.P. Aarnts, P.D. Lee, E. Jimenez-Melero
Materials Science & Engineering A: B.L. Ennis, C. Bos, M.P. Aarnts, P.D. Lee, E. Jimenez-Melero
A R T I C L E I N F O A B S T R A C T
Keywords: In this work, we show how the presence of microstructural banding and segregation affects the work-hardening
Mechanical characterization behaviour of a dual phase steel with improved formability. This steel contains chemical segregation inherited
Hardening from the casting process. Our previously developed 3D cellular automaton model allowed us to design thermo-
Steel mechanical processes to either promote or suppress banding. The bands are properly described as in-plane sheets
Austenite
of martensite grains. Mechanical testing data revealed a significant reduction in tensile strength in banded
Martensite
structures for a similar level of ductility. The work-hardening behaviour in the pre-yield regime, including the
Thermo-mechanical processing
yield strength itself, is not correlated to the incidence of segregation and/or microstructural banding. The re-
duction in ultimate tensile strength in banded structures stems from a reduced work-hardening capacity in the
post-yield regime. This is due to increased austenite stability in the banded steels, coupled to the anisotropic
strain localisation in the ferritic matrix between martensite bands.
⁎
Corresponding author at: Tata Steel Research and Development, 1970 CA IJmuiden, The Netherlands.
E-mail address: bernard.ennis@tatasteel.com (B.L. Ennis).
https://doi.org/10.1016/j.msea.2017.12.078
Received 2 October 2017; Received in revised form 19 December 2017; Accepted 20 December 2017
Available online 21 December 2017
0921-5093/ © 2017 Elsevier B.V. All rights reserved.
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
these discoveries, the influence of banding on the work hardening and transformation temperature [16]. The resulting function for the con-
fracture behaviour of austenite-containing steels remains less well un- centration profile used in the computer simulations is shown in Fig. 1a.
derstood due to the difficulties of resolving this phase using light mi- This profile is only a function of the y position of a cell, which means
croscopy, scanning electron microscopy (SEM) and electron backscatter that the element concentration is assumed to be constant in the x and z
diffraction (EBSD) [9]. Finite element analysis carried out by Shi et al. directions. This concentration profile was input into the CA model [12],
[10] indicated that a random distribution of the austenite phase leads to see Fig. 1b, and the selected cooling trajectories from the temperature
a higher tensile strength than when located predominantly within the of 850 °C used in the final thickness reduction step during hot rolling
bands. They coupled the strength increase to a lack of localised slip are shown in Fig. 1c.
bands in the ferrite matrix in the uniformly dispersed sample. In steels The simulated annealing cycle after hot and cold rolling consisted of
with austenite-enriched bands, confinement of ferrite between con- heating the deformed microstructure at 5 °C/s to 800 °C, holding for
nected austenite grains led to increased formation of slip bands within 60 s at this temperature and then cooling down to 650 °C at 1 °C/s, as
the ferrite matrix, effectively reducing the overall strength of the ma- shown in Fig. 1d. In the industrial annealing cycle, the material is
terial. Recently we demonstrated how the mechanically induced aus- subsequently quenched to room temperature. During quenching mar-
tenite transformation contributes to the work-hardening in non-banded tensite formation takes place, which is not accounted for in the model.
DH steels. We proposed a physically-based model based on an extended Since a low heating rate up to the annealing temperature can effectively
Kocks-Mecking analysis, coupled with in situ high-energy X-ray dif- remove the banding suppression achieved before in the hot-strip mill
fraction data collected upon straining, to describe the effect of the [17], experimental annealing cycles were adapted to mirror the cooling
mechanically-induced austenite transformation on the work-hardening rate in the hot-rolling process.
behaviour in a non-banded DH steel [11]. In this work, we assessed the One block of each type (NS, S) was used for each of the processing
effect of banding on the work-hardening behaviour in DH steels, by conditions (NB, B), yielding four basic microstructural variants. The
applying an equivalent physically-based model analysis of the work- main annealing parameters for each variant are given in Table 2,
hardening behaviour on banded and non-banded microstructures of the whereas the annealing cycles for both experiment and simulation are
same DH steel grade. Those microstructures were produced industrially shown graphically in Fig. 1d.
following predictions from a 3D cellular automaton (CA) model The finishing rolling temperature (FRT) was chosen to ensure de-
[12–14]. formation in the austenite region. The samples were then cooled at
different cooling rates (φ) through the austenite to ferrite transforma-
tion to the intermediate temperature at which the industrial strip steel
2. Experimental is coiled, namely coiling temperature (CT). After cold-rolling to 1 mm,
the samples were heated at different heating rates (HR) to 820 °C,
2.1. Material processing subsequently quenched using helium to the isothermal bainite holding
temperature of 425 °C, and held for 60 s prior, and finally quenched to
The steel under investigation is a high-strength dual phase steel with room temperature. The two non-banded steels, namely S-NB and NS-
improved formability, termed DH800, produced via continuous casting NB, were generated using a high heating rate; the other two (banded)
followed by hot-rolling (HR), cold-rolling (CR) and continuous an- samples S-B and NS-B corresponded to the lower heating rate used in
nealing. The chemical composition of this steel is given in Table 1, the CA model.
together with the segregation coefficient of the relevant elements [5].
Samples from industrial production material were taken after 2.2. Structural characterization
casting (AC) and further processed in the laboratory for this work. Two
slab samples, with dimensions of 135 × 75 × 225 mm3, were cut from Samples were taken adjacent to the top and middle blocks of the
the middle of consecutive slabs produced in an industrial continuous original as-cast feedstock to determine the degree of chemical segre-
casting line. These samples were further sectioned into three equal sized gation after casting. Optical microscopy was carried out on samples
blocks of 75 mm in height. The blocks from the middle of the original from each process variant, in the as hot-rolled (HR), cold-rolled (CR),
feedstock were designated S (segregated), since this is the most and cold-rolled & annealed condition (CRA). In the HR microstructure
common section for the appearance of both chemical macro- and micro- Nital was used to etch the pearlite phase, which consequently appears
segregation due to (a) the increased dendrite arm spacing as a result of dark in the micrographs. In the CRA microstructure 10% sodium me-
slower cooling rates and (b) the solute enrichment of the final solidi- tabisulphite was used to etch the combined martensite (α’) + austenite
fication zone [15]. The blocks taken from the top of the feedstock were (γ) phase, leaving ferrite lightly etched. The ferrite grain size could thus
designated NS (non-segregated), although the ‘S-NS’ designation refers be measured according to ASTM E112 [18] using Leica QWIN image
to the centre-line (macro) segregation. analysis software with a pixel size of 0.11 µm. For the banding analysis,
The blocks were subsequently hot- and cold-rolled, followed by a the CRA microstructure was etched with LePera reagent, as this pro-
continuous annealing designed specifically to either promote or sup- vides a better contrast between the ferrite matrix and the combined
press the occurrence of banding in the microstructure, generating B α’+γ second phase. In that case, the matrix is etched (blue) and the
(banded) and NB (non-banded) structures respectively [5]. The an- second phase is left unetched. Image analysis of the unetched phase
nealing process was simulated using a 3D cellular automaton model therefore results in the combined second phase fraction, fα’+γ. In the CA
[12]. The chemical segregation resulting from casting was modelled by model structures the matrix phase (ferrite) is coloured blue and the
combining a sine wave function for the spatial distribution and an second, carbon-rich, phase is coloured orange. This means that the
analytical function of the effect of chemical concentration on the phase orange phase in the simulated structures represents pearlite in the HR
condition, and martensite + austenite in the CRA condition. The vo-
Table 1 lume fraction of retained austenite fγ was measured by X-ray diffraction
Chemical composition and average segregation coefficients of the investigated DH800
(XRD) using a fully automated Panalytical Xpert PRO diffractometer.
steel.
The XRD patterns were recorded in the 2θ-range from 38 to 165° in
Element Fe C Mn Si Al Cr S P reflection mode using Co-Kα radiation (λ = 1.79 Å). The Rietveld
analysis of the XRD patterns was carried out using the Bruker TOPAS
Composition, wt% Bal. 0.15 2.05 0.42 0.55 0.40 < 0.01 < 0.01
software package.
Segregation – – 1.36 1.28 0.78 1.15 – –
coefficient, (1/k) The ASTM E1268-99 standard [19] describes the anisotropy index
(AI) method to determine the degree of banding in metallic
279
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
Fig. 1. Settings for CA simulations: (a) Mn concentration profile, (b) Voronoi 3D microstructure elements, (c) cooling trajectories from austenitising to coiling temperatures in the hot-
strip mill and (d) thermal cycles used in the annealing simulation after rolling (green dashed line) and in the experiments (solid red and blue lines) [12,13]. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)
microstructures. This index is determined by the ratio of the average completely decayed to white. In a banded structure, the percentage
number of interceptions per unit length of test lines perpendicular NL ⊥ decay is much more rapid when eroding perpendicular to the rolling
and parallel NL// to the deformation direction with the investigated direction (V) than parallel to it (H). By defining a characteristic decay
(banded) phase: distance, de[H,V], as the one at which the measured fraction of ferrite is
equal to 1/e of its initial value, it is then possible to define an “erosion-
NL ⊥
AI = based” anisotropy index as being the ratio of the decay distances in both
NL// (1) directions:
Although this metric can distinguish banded and non-banded de H
structures, it cannot adequately discriminate between strongly and AI =
deV (2)
weakly banded structures [20]. An alternative approach is based on the
measurement of the bi-directional (horizontal and vertical) evolution of The first step in this method is the detection of the phase of interest.
the fraction of matrix phase (in this case ferrite) when it is progressively This means that for the hot-rolled material the pearlite bands (Nital
eroded in the two perpendicular directions by a linear increment [9]. etching) and for the cold rolled material the martensite-containing
The image is converted into a black and white pixel image, and the dark bands (10% sodium metabisulphite etching) are detected as the dark
portion is progressive eroded by the linear increment until it has phase. No image processing steps were performed on the grey scale
Table 2
Processing conditions and resultant microstructures of each experimental block.
280
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
Fig. 2. (a) Illustration of the banding parameters measured by progressive erosion, (b) relationship between anisotropy index and peak maximum position of the different variants after
hot-rolling (HR) and after cold-rolling and annealing (CRA). CA indicates microstructures generated by the cellular automaton model simulation; L and T indicate the microstructure in
the longitudinal and transverse directions respectively. The ultimate tensile strength in the CRA condition is shown by the red circles. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)
input images, these were converted as-is to black and white. The ob- at room temperature using an Instron 100 kN tensile tester, with the
tained binary image of the detected phase serves as input for the sub- rolling direction parallel to the tensile direction. The strain rate in the
sequent quantitative image analysis steps. In all cases, a horizontal gauge length was kept constant at 5 × 10−3 s−1 throughout the test;
closure step was applied to the initial binary image to suppress the strain and strain rate were measured and controlled using a clip-on
influence of discontinuous bands. The band continuity could potentially strain gauge.
be quantified by varying the number of closures steps [21], but that
falls out of the scope of this study. 3. Results
After phase identification, the area fractions are determined of the
binary image after progressive erosion in the vertical (deV) and hor- 3.1. Microstructure evolution through processing
izontal (deH) directions resulting in two area fraction decay curves. This
is illustrated in Fig. 2a by the curves with closed (V) and open (H) In our previous work, we showed the presence of significant
circles. By subtracting the vertical decay curve from the horizontal banding which was inherited from the cast structure. The counter
decay curve, a difference curve is obtained (solid blue line in Fig. 2a) segregation of Al on one hand and Mn and Si on the other tends to
from which several characteristic values can be extracted. The peak promote banding, especially at the centre-line where the macroscopic
maximum of the difference curve is of itself a measure for the thickness segregation is the most exaggerated. Segregation factors for the major
of the bands: the thicker the bands, the larger shift of the maximum to alloying elements in this steel were calculated as the ratio of the
higher erosion values. The position of the peak is also used in defining composition of each element at the centre-line to the bulk composition,
an objective “erosion-based” anisotropy index, which corresponds to see Table 1. Banding in this steel appears as continuous strings of
the ratio of horizontal to vertical decay at the peak position. The area martensite grains aligned parallel to the rolling direction. Since the
under the difference curve (red solid line in Fig. 2a) is a measure for the segregation in bands has been shown to be directly inherited from the
amount of bands: for strongly banded structures the integrated area as cast structure [5], the segregation factors derived are the inverse of
curve follows the area detected prior to the erosion steps, however for the Scheil partitioning coefficient: the values in the table are, thus, in
non-banded structures there will be a big difference between the area good agreement with other published values of the partitioning coef-
detected prior to the erosion steps and the area under the difference ficient [23,24].
curve. To compensate for the grain refinement resulting from cold- The microstructures in the as-hot rolled condition are shown in
rolling, when going from the hot rolled to cold rolled end micro- Fig. 3, along with the simulated microstructure from the CA model,
structure, the image of the cold rolled material was taken at 500× which shows reasonable agreement with the actual microstructures
whereas the hot rolled product was taken at 200×. Given that this is achieved. This is also reflected in the anisotropy index analysis shown
not exactly equivalent to the actual cold-rolling reduction applied in Fig. 2b, whereby the non-banded microstructures have a low aniso-
(70%), the peak maximum is normalised into µm for ease of compar- tropy and a relatively lower peak position than the banded samples for
ison. both the real and simulated microstructures. Whilst it was not possible
to suppress banding completely in the segregated sample S-NB, the
2.3. Mechanical testing average grain size and inter-band spacing is significantly reduced when
compared to the other segregated sample S-B. The difference between
Samples were taken from the mid-section of each structural variant the S-B and NS-B samples is the appearance of more pronounced bands
and machined into identical tensile specimens with a 50 mm gauge at the centre-line in the S-B sample, as reflected in the higher anisotropy
length per EN ISO 6892-1 [19]. The specimens were strained to failure index and peak position in both the longitudinal and transverse
281
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
Fig. 3. (a, d) The microstructures predicted by the cellular automaton model in the as hot-rolled (HR) condition for process conditions to suppress (NB) or promote (B) the formation of
banding are compared to (b, c, d and e) micrographs of the observed HR structure as etched in 5% Nital.
directions, see Fig. 2b. The bands in the NS-B sample are more nu- contains the real CRA microstructure of the SB and S-NB variants in
merous and more closely spaced than in the S-B sample. both the longitudinal and transverse directions. The bands in the S-B
The simulated S-SB and S-NB microstructures in the as cold-rolled sample extend not only parallel to the rolling direction but also per-
and annealed (CRA) condition are shown in Fig. 4a and d respectively. pendicular in the rolling plane, which is consistent with the CA model.
The CA model simulations show that it is possible to suppress the These bands extend across the transverse direction of the steel (in plane
centre-line band with this process, but that some banding is still present and perpendicular to the rolling direction) and as such they are better
due to micro-segregation. With the notable exception of the NS-NB described as continuous 2D bands. Although continuous bands are
microstructure, the anisotropy index decreases after cold-rolling and present in the longitudinal direction of the S-NB sample, they are not
annealing which is not completely captured by the model, see Fig. 2b. present in the transverse direction.
Since the CA segregation model is based upon the extrusion into the z
direction of the sine function, the CA model may not be completely
valid in describing the 3D grain and alloy distribution. Fig. 4 also
Fig. 4. (a, d) The microstructures predicted by the cellular automaton model in the as cold-rolled and annealed condition (CRA) for process conditions to (a) promote (B) or (d) suppress
(NB) the formation of banding are compared to the observed microstructures in (b, e) longitudinal and (c, f) transverse directions of (b, c) the segregated and banded variant (S-B) and (e,
f) the segregated, non-banded variant (S-NB) as etched with 10% sodium metabisulphite.
282
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
Table 3
Volume fraction and lattice parameter of the second phase constituents in the CRA condition, primarily austenite (γ) and martensite (α′), together with the carbon content of both
constituents and the ferrite grain diameter (Dα).
Variant Dα (µm) fα′+γ (vol%) fγ (vol%) fα’(LOM) (vol%) fγ(UTS)/fγ0 aγ (Å) [C]α′ (wt%) [C]γ (wt%)
and for XMn, XSi and XCr we have used the bulk alloy content, since the
local alloy content varies from band to band. Given that the materials
are subjected to an isothermal bainitic holding treatment at 425 °C for
60 s prior to quenching to room temperature, austenite grains whose
carbon content resulted in an Ms at or above the holding temperature
will already have transformed to martensite upon quenching from
820 °C to that isothermal temperature (425 °C). Any remaining auste-
nite grains will then be further enriched in carbon through the bainite
reaction, thus remaining for the most part in a metastable state at room
Fig. 5. Equivalent stress of each of the microstructural variants of DH steel as a function temperature.
of the applied equivalent strain at room temperature. The resulting calculated carbon content in martensite is low enough
not to give a significant degree of crystal tetragonality [27,28] and is
consistent with the minimum carbon content in austenite of 0.444 wt%
3.2. Characterization of microstructural variants
required for it to remain metastable at room temperature [29]. It is
worth noting that the calculated carbon content of the martensite phase
The volume fraction of the combined second phase (including aus-
in the banded and non-banded variants in the CRA condition does not
tenite and bainite) in each variant was measured using light optical
differ significantly. However, the calculated carbon content of the
microscopy (LOM) combined with image analysis, see Table 3. It is not
austenite phase of the banded variants is considerably higher than in
possible to separate austenite and martensite using this technique due
the non-banded variants, which will increase the mechanical stability of
to the similarity in colour after etching. For this reason, the retained
this phase.
austenite fraction is determined using XRD, and both its volume frac-
The retained austenite fractions of the NS-NB and S-B variants were
tion and lattice parameter are also given in the table. The martensite
also measured at the UTS (fγ(UTS)) and the ratios thereof to the original
fraction is calculated by subtracting the amount of retained austenite,
phase fractions (fγ0) are also given in the table.
measured by XRD, from the total amount of second phase obtained
using light optical microscopy.
The carbon content in austenite (XCγ) was derived by subtracting the 3.3. Mechanical behaviour
amount of carbon in the martensite (XCα’) from the total carbon in the
material (XCtotal): The equivalent stress-equivalent strain curve for each variant at
room temperature is shown in Fig. 5, and the principal mechanical
XCγ = XCtotal − XCα (3) properties are given in Table 4. All variants present a similar mechan-
′
ical response up to the onset of plastic deformation. The post-yielding
Due to the presence of significant quantities of carbide-suppressing properties of the four variants are also relatively close, except for ul-
elements such as Al and Si in this steel [24,25], we can assume that the timate tensile strength (UTS), which was consistently higher in the non-
remaining carbon atoms can be found predominantly in the retained banded samples than the banded samples.
austenite. The carbon content of the martensite was calculated using From Table 3 we can see that the average grain size of the ferrite
the following polynomial equation for the martensite start temperature, phase, Dα, in the CRA condition does not vary significantly across the
Ms [26]: different microstructural variants, and hence the related Hall-Petch
grain boundary strengthening, σy(HP), according to [31]:
Table 4
Mechanical properties and initial dislocation structure of different variants in CRA condition. Rp0.2 is the 0.2% proof strength, UTS the ultimate tensile strength, Au and At are the uniform
and total elongation, respectively. The parameters derived from the extended Kocks-Mecking plot area also given in the table, where σy is the yield stress, E is the elastic modulus, Θa the
anelastic (pre-yield) work-hardening rate, L is the effective dislocation segment length and ρ is the dislocation density, see Section 4 in the text and ref. [30].
NS-NB 385 890 14.1 18.9 389 209 5131 159 3.3 × 1013
S-NB 374 875 16.1 20.4 366 202 4973 168 3.0 × 1013
NS-B 374 808 16.5 20.8 369 218 4549 166 3.6 × 1013
S-B 363 813 15.9 19.6 353 202 5161 174 2.7 × 1013
283
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
1
σy (HP ) ∝
Dα (6)
The sample with highest strength (UTS) and lowest uniform elon-
gation (Au) is the NS-NB sample, which also contains the highest vo-
lume fraction and lowest average carbon content of retained austenite.
Based on these results, it seems that the volume fraction and overall
mechanical stability of the retained austenite are not the only para-
meters controlling the ductility in banded DH steel structures.
4. Discussion
The intercept method for determining the so-called proof stress that
characterises yielding, as recommended in international standards [32],
can lead in this case to significant errors due to the very high work
hardening rate of these DH steels, especially in the early stages of
plastic deformation. However, the pre-yield behaviour appears to be Fig. 7. Work hardening rate of each microstructural variant as a function of the
similar for each of the microstructural variants. This pre-yield beha- equivalent (true) stress.
viour, which is due to reversible dislocation glide [30,33], is manifested
by a drop in the work-hardening rate below the theoretical maximum
related to the anelastic work-hardening rate (Θa) as follows [30,33]:
for a perfectly elastic material, as the applied stress approaches the
yield stress. In order to determine accurately the yield stress of these M 2Es 3 1 − s 2
steels, it is necessary to find the transition from the pre-yield, reversible Θa =
ρL2 (1 + ν )(s − arcsin(s ) 1 − s 2 ) (8)
glide regime into irreversible plastic deformation. This pre-yield be-
haviour is best illustrated using an extended Kocks-Mecking (eK-M) plot where is the Taylor factor and equals 3.06 [38], E is the elastic modulus
of the work-hardening rate as a function of the nominal equivalent whose value is collected in Table 4 for each of the variants, ν is the
stress [11,34]. Poisson ratio of ferrite and takes a value of 0.29 [39], ρ is the dis-
Fig. 6a shows the low strain portion of the tensile a tensile curve of location density, L is the effective dislocation segment length and s is
the NS-NB variant, and the dashed lines illustrate how the 0.2% proof the ratio of the yield stress to the critical stress, σc, for generation of
stress (Rp0.2) is determined. The enlargement of the low-strain data dislocations from a Frank-Read source given by:
shows that the yield stress, in its physical meaning of the stress at which
M Gb
Frank-Read sources are activated, cannot be determined un- σc =
L (9)
ambiguously. An alternative representation of the tensile curve is the
Kocks-Mecking (KM) plot, which is a graphical representation of the with G, the shear modulus = 81 GPa [40] and b, the modulus of the
work-hardening rate, Θ = dσf/dε vs. the flow stress σf, [35–37]. By Burgers’ vector of ferrite = ½a < 111 > = 0.248 nm. By combining Eqs.
definition, the flow stress is the sum of the yield stress σy and the work- (8) and (9) we can fit the work-hardening behaviour of the different
hardening contribution σw: variants from Fig. 7 to derive, and thereby compare the initial dis-
σf = σy + σw location structures. The fit parameters are the effective dislocation
(7)
segment length, L and dislocation density, ρ, which are given in Table 4
By extending the curve to include the pre-yield region (σ < σf) and for each variant. Since the dislocation segment lengths in a real material
plotting the data against the nominal equivalent stress, see Fig. 6b, we form a distribution, this will result in a range of applied stresses at
can also derive information regarding the initial dislocation structure. which dislocations will be activated as Frank-Read sources: longer
This curve also shows two branches of different slope, with a relatively segments at lower stress, shorter segments at higher stress. The exten-
sharp transition point between them. The approx. linear high-stress part sion of the yield-stress range is, thus, directly related to the width of the
of the KM-plot describes the plastic behaviour of the material; the low- segment-length distribution: a wider segment-length distribution will
stress branch on the other hand, provides information on the pre-yield cause a more gradual transition between the pre-yield and post-yield
behaviour of the material, and thus on the dislocation structure. This is ranges in the extended Kocks-Mecking plot. The parameter L should
Fig. 6. (a) Yield regime of a tensile test in this case of the NS-NB variant, showing the Rp0.2 construction and the yield stress determined from (b) extended Kocks-Mecking (KM) plot of the
data in (a), showing a well-defined yield point (σy). The solid line represents experimental data, the blue dashed line the model description of the pre-yield behaviour from Eq. (8), see
text.
284
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
therefore be regarded as an effective segment length related to the ef- compared to NS-NB.
fective average yield stress. This approach for deriving the parameters
was verified by Arechabaleta et al. by comparing the data from a tensile 5. Conclusion
test to X-ray diffraction measurements [30].
As σ approaches σc, dislocation segments are activated as Frank- In this work, we have shown how the application of segregation
Read sources and the pre-yield behaviour of Eq. (8) changes into the profiles inherited from the casting process in a 3D cellular automaton
plastic Stage-III behaviour. This transition, clearly present in the ex- model can be reliably used to predict the microstructure of a dual phase
tended KM-plot of Fig. 6b, is the physical definition of the yield stress steel with improved formability at different stages in the production.
σy. This value can be thus correctly interpreted as the critical stress for Despite the simple 1D description of the chemical segregation in the
generation of dislocations from a Frank-Read source i.e. σc = σy. If the model, it captures adequately the sheet nature of the martensite grains
pre-yield behaviour consisted solely of linear elastic deformation, then in the resultant microstructure. Mechanical testing was carried out on
given the generally accepted value of the elastic modulus of the un- samples treated to achieve different combinations of segregation and
distorted lattice in steel of 210 GPa [40], the maximum work-hardening banding, and the results show a significant reduction in tensile strength
rate in the pre-yield regime of a purely elastic material would achieve in banded steels for a similar level of ductility. The work-hardening
this value, with a sharp drop at the yield stress. Instead, the initial part behaviour in the pre-yield regime, including the yield strength itself, is
of the plot gradually decreases, with a slope that is significantly higher not correlated to the incidence of segregation and/or microstructural
than that of the plastic Stage-III hardening (Θpost) which according to banding. We demonstrate that the reduction in ultimate tensile strength
classical dislocation theory follows a linear relationship [41]: in banded microstructures is the result of a reduced work-hardening
capacity in the post-yield regime. This is due to an increased austenite
1
Θpost = (αM GbU − Ωσw ) stability in the banded steels, coupled to the anisotropic strain locali-
2 (10)
sation in the ferrite matrix between the martensite bands.
where U and Ω are the dislocation storage rate and dynamic recovery
rate, respectively, G is the elastic shear modulus, b is the Burger's Acknowledgements
vector, σw is the work-hardening contribution to the flow stress, is the
Taylor factor and α is a material factor related to the dislocation density This work was made possible by the facilities and support of Tata
and distribution. Steel, the Diamond-Manchester Collaboration and the Research
The extended Kocks-Mecking plot for each microstructural variant is Complex at Harwell, funded in part by EPSRC (EP/I02249X/1).
shown in Fig. 7. The data was smoothed using a 5-point quadratic
polynomial filter [42]. The data in both Table 4 and Fig. 7 show that Data statement
there is no clear correlation between microstructural banding and the
values for the dislocation density, effective segment length, elastic A representative sample of research data from the experiments
modulus and yield strength. The observed difference in mechanical along with the plot data for the graphs in this manuscript is provided in
behaviour is not significantly affected by the dislocation structure and supplementary material: [doi: 10.17632/4pdrfnn3kn.1].
pre-yield work-hardening behaviour of the original ferritic matrix. The
work-hardening rate of the different variants lies close to one another at References
very low stresses. After the maximum value of the work-hardening rate
has been reached, it then drops below the theoretical maximum for a [1] German Association of the Automotive Industry (VDA), VDA 239-100 Sheet Steel
perfectly elastic material (Young's modulus, E), due to reversible dis- for Cold Forming, 2016. 〈https://www.vda.de/en/services/Publications/vda-239-
100.html〉.
location glide [30,33] as the applied stress approaches the yield stress, [2] M. De Meyer, D. Vanderschueren, B.C. De Cooman, The influence of the substitution
which is thus back-calculated from the minimum value of the work- of Si by Al on the properties of cold-rolled C-Mn-Si TRIP steels, ISIJ Int. 39 (1999)
hardening rate. 813–822, http://dx.doi.org/10.2355/isijinternational.39.813.
[3] K. Eberle, P. Cantinieaux, P. Harlet, New thermomechanical strategies for the
The difference in mechanical properties of the different variants can production of high strength low alloyed multiphase steel showing a transformation
be explained by examination of the work-hardening rate in the post- induced plasticity (TRIP) effect, Steel Res. 70 (1999) 233–238.
yield (plastic) regime, see inset of Fig. 7. Firstly, the work-hardening [4] J. Maki, J. Mahieu, B.C. De Cooman, S. Claessens, Hot-dip galvanising of Si-free C-
Mn-Al TRIP steels, in: M. Lamberights, C. de R.M. (CRM) (Eds.), Galvatech 2001,
rate of the banded variants do not show clear minima and instead ex- Verlag Stahleisen, 2001, pp. 623–630.
hibit the gradual transition from pre- to post-yield work-hardening [5] B.L. Ennis, E. Jimenez-Melero, R. Mostert, M.B. Santillana, P.D. Lee, The role of
more typical of dual phase steels without retained austenite aluminium in chemical and phase segregation in a TRIP-assisted dual phase steel,
Acta Mater. 115 (2016) 132–142, http://dx.doi.org/10.1016/j.actamat.2016.05.
[11,43–45]. The gradual transition is the result of equalisation of
046.
stresses between the martensite and ferrite phases. As demonstrated by [6] C.C. Tasan, J.P.M. Hoefnagels, M.G.D. Geers, Microstructural banding effects clar-
Tasan et al. [6] in a banded DP steel, the effect of the martensite bands ified through micrographic digital image correlation, Scr. Mater. 62 (2010)
is to localise the strain into the ferrite phase between the bands. This 835–838, http://dx.doi.org/10.1016/j.scriptamat.2010.02.014.
[7] M. Mazinani, W.J. Poole, Effect of martensite plasticity on the deformation behavior
will induce deformation in the ferrite phase without necessarily acti- of a low-carbon dual-phase steel, Metall. Mater. Trans. A 38 (2007) 328–339,
vating the mechanical transformation of austenite. By contrast, the non- http://dx.doi.org/10.1007/s11661-006-9023-3.
banded variants show the distinctive rise in work-hardening rate at the [8] H. Ghadbeigi, C. Pinna, S. Celotto, J.R. Yates, Local plastic strain evolution in a high
strength dual-phase steel, Mater. Sci. Eng. A 527 (2010) 5026–5032, http://dx.doi.
yield point associated with the austenite transformation [11]. org/10.1016/j.msea.2010.04.052.
Moreover, the post-yield work-hardening rate is higher for the non- [9] M.P. Aarnts, M. Zuiderwijk, R.A. Rijkenberg, A. Arlazarov, D. Barbier, A. Hazotte,
banded steels than for the banded variants, although the slopes of the B. Krebs, L. Germain, M. Gouné, L. Barbé, J. Kommenda, O. Karlsson, J. Hagström,
M. Fischer, F. Friedel, V. Marx, E. Füreder, C. Commenda, Microstructural
work-hardening curves of all variants are parallel i.e. a higher con- Quantification of Multi-phase Steels (Micro-quant), European Commission,
tribution of work-hardening stress due to the transformation work. Brussels, Belgium, 2011, http://dx.doi.org/10.2777/83656.
Furthermore, the carbon content of the austenite in the banded variants [10] J. Shi, S. Turteltaub, E. Van der Giessen, Analysis of banded morphology in mul-
tiphase steels based on a discrete dislocation–transformation model, Model. Simul.
is significantly higher than in the non-banded variants. This will result Mater. Sci. Eng. 19 (2011) 74006, http://dx.doi.org/10.1088/0965-0393/19/7/
in higher mechanical stability of this phase [46,47], and therefore to a 074006.
lower contribution to the work-hardening rate in the post-yield region. [11] B.L. Ennis, E. Jimenez-Melero, E.H. Atzema, M. Krugla, M.A. Azeem, D. Rowley,
D. Daisenberger, D.N. Hanlon, P.D. Lee, Metastable austenite driven work-hard-
This is supported by the significantly higher fraction of retained aus-
ening behaviour in a TRIP-assisted dual phase steel, Int. J. Plast. 88 (2017)
tenite remaining at the UTS of the banded variant S-B than in the non- 126–139, http://dx.doi.org/10.1016/j.ijplas.2016.10.005.
banded variant NS-NB which explains the significantly lower UTS in S-B [12] C. Bos, M.G. Mecozzi, D.N. Hanlon, M.P. Aarnts, J. Sietsma, Application of a three-
285
B.L. Ennis et al. Materials Science & Engineering A 713 (2018) 278–286
dimensional microstructure evolution model to identify key process settings for the (1949) 212–213.
production of dual-phase steels, Metall. Mater. Trans. A 42 (2011) 3602–3610, [30] Z. Arechabaleta, P. van Liempt, J. Sietsma, Quantification of dislocation structures
http://dx.doi.org/10.1007/s11661-011-0696-x. from anelastic deformation behaviour, Acta Mater. 115 (2016) 314–323, http://dx.
[13] C. Bos, M.G. Mecozzi, J. Sietsma, A microstructure model for recrystallisation and doi.org/10.1016/j.actamat.2016.05.040.
phase transformation during the dual-phase steel annealing cycle, Comput. Mater. [31] F.B. Pickering, Cahn, et al. (Ed.), Structure-Property Relationships in Steels, Wiley,
Sci. 48 (2010) 692–699, http://dx.doi.org/10.1016/j.commatsci.2010.03.010. 1993, pp. 45–79.
[14] M.G. Mecozzi, C. Bos, J. Sietsma, Microstructure modelling of solid-state transfor- [32] The International Organisation for Standardisation (ISO), ISO 6892-1: Metallic
mations in low-alloy steel production, Mater. Sci. Forum 706–709 (2012) Materials - Tensile Testing - Part 1: Method of Testing at Ambient Temperature,
2782–2787, http://dx.doi.org/10.4028/www.scientific.net/MSF.706-709.2782. Geneve, Switzerland, 2016.
[15] S. Karagadde, L. Yuan, N. Shevchenko, S. Eckert, P.D. Lee, 3-D microstructural [33] P. van Liempt, J. Sietsma, A physically based yield criterion I. determination of the
model of freckle formation validated using in situ experiments, Acta Mater. 79 yield stress based on analysis of pre-yield dislocation behaviour, Mater. Sci. Eng. A
(2014) 168–180, http://dx.doi.org/10.1016/j.actamat.2014.07.002. 662 (2016) 80–87, http://dx.doi.org/10.1016/j.msea.2016.03.013.
[16] B.L. Ennis, E. Jimenez-Melero, R. Mostert, B. Santillana, P.D. Lee, Dataset con- [34] H. Mecking, U.F. Kocks, Kinetics of flow and strain-hardening, Acta Metall. 29
cerning the analytical approximation of the Ae3 temperature, Data Brief 10 (2017) (1981) 1865–1875, http://dx.doi.org/10.1016/0001-6160(81)90112-7.
330–334, http://dx.doi.org/10.1016/j.dib.2016.11.073. [35] U.F. Kocks, Laws for work-hardening and low-temperature creep, J. Eng. Mater.
[17] F.G. Caballero, A. García-Junceda, C. Capdevila, C. García de Andrés, Evolution of Technol. 98 (1976) 76, http://dx.doi.org/10.1115/1.3443340.
microstructural banding during the manufacturing process of dual phase steels, [36] H. Mecking, Work hardening of single-phase polycrystals, in: Encycl. Mater. Sci.
Mater. Trans. 47 (2006) 2269–2276, http://dx.doi.org/10.2320/matertrans.47. Technol., Elsevier Science, London, UK, 2001, pp. 9785–9795.
2269. [37] U.F. Kocks, H. Mecking, Physics and phenomenology of strain hardening: the FCC
[18] ASTM, E112-10 standard test methods for determining average grain size, in: Annu. case, Prog. Mater. Sci. 48 (2003) 171–273, http://dx.doi.org/10.1016/S0079-
B. ASTM Stand. Vol. 03. 01, Met. –Mechanical Testing; Elev. Low-Temperature 6425(02)00003-8.
Tests Metallogr., ASTM International, West Conshohocken, PA, USA, 2011. [38] R.E. Stoller, S.J. Zinkle, On the relationship between uniaxial yield strength and
[19] ASTM, E1268-99 Standard practice for assessing the degree of banding or or- resolved shear stress in polycrystalline materials, J. Nucl. Mater. 283–287 (2000)
ientation of microstructures (reapproved 2001, 2007), in: Annu. B. ASTM Stand. 349–352, http://dx.doi.org/10.1016/S0022-3115(00)00378-0.
Vol. 03. 01, Met. –Mechanical Testing; Elev. Low-Temperature Tests Metallogr., [39] J.A. Benito, J. Jorba, J.M. Manero, A. Roca, Change of Young's modulus of cold-
ASTM International, West Conshohocken, PA, USA, 2000: pp. 820–848. deformed pure iron in a tensile test, Metall. Mater. Trans. A 36 (2005) 3317–3324,
[20] B. Krebs, A. Hazotte, L. Germain, M. Gouné, Quantitative analysis of banded http://dx.doi.org/10.1007/s11661-005-0006-6.
structures in dual-phase steels, Image Anal. Stereol. 29 (2010) 85–90, http://dx.doi. [40] G. Ghosh, G.B. Olson, The isotropic shear modulus of multicomponent Fe-base solid
org/10.5566/ias.v29.p85-90. solutions, Acta Mater. 50 (2002) 2655–2675, http://dx.doi.org/10.1016/S1359-
[21] K.S. McGarrity, J. Sietsma, G. Jongbloed, Characterisation and quantification of 6454(02)00096-4.
microstructural banding in dual phase steels Part 1 – General 2D study, Mater. Sci. [41] H. Mecking, U.F. Kocks, Kinetics of flow and strain-hardening, Acta Metall. 29
Technol. 28 (2012) 895–902, http://dx.doi.org/10.1179/1743284712Y. (1981) 1865–1875, http://dx.doi.org/10.1016/0001-6160(81)90112-7.
0000000004. [42] A. Savitzky, M.J.E. Golay, Smoothing and differentiation of data by simplified least
[22] R.M. Fisher, G.R. Speich, L.J. Cuddy, H. Hu, Phase transformations during steel squares procedures, Anal. Chem. 36 (1964) 1627–1639, http://dx.doi.org/10.
production, in: Proceedings Darken Conference Physical Chem. Metall., US Steel, 1021/ac60214a047.
Monroeville, PA, USA, 1976, pp. 463–488. [43] P. van Liempt, Workhardening and substructural geometry of metals, J. Mater.
[23] D. Chakrabarti, C. Davis, M. Strangwood, Characterisation of bimodal grain struc- Process. Technol. 45 (1994) 459–464, http://dx.doi.org/10.1016/0924-0136(94)
tures in HSLA steels, Mater. Charact. 58 (2007) 423–438, http://dx.doi.org/10. 90382-4.
1016/j.matchar.2006.06.014. [44] H. Vegter, P. van Liempt, Proceedings of the 13th International Conference on
[24] M. Gomez, C.I. Garcia, D.M. Haezebrouck, A.J. Deardo, Design of composition in Metal Forming, Steel Res. Int., vol. 81, 2010, n/a-n/a. 〈http://doi.org/10.1002/
(Al/Si)-alloyed TRIP steels, ISIJ Int. 49 (2009) 302–311, http://dx.doi.org/10. srin.201190002〉.
2355/isijinternational.49.302. [45] Y. Bergström, A dislocation model for the stress-strain behaviour of polycrystalline
[25] D. Simonovic, C.K. Ande, A.I. Duff, F. Syahputra, M.H.F. Sluiter, Diffusion of carbon α-Fe with special emphasis on the variation of the densities of mobile and immobile
in bcc Fe in the presence of Si, Phys. Rev. B 81 (2010) 54116, http://dx.doi.org/10. dislocations, Mater. Sci. Eng. 5 (1970) 193–200, http://dx.doi.org/10.1016/0025-
1103/PhysRevB.81.054116. 5416(70)90081-9.
[26] S.M.C. van Bohemen, Bainite and martensite start temperature calculated with [46] E. Jimenez-Melero, N.H. van Dijk, L. Zhao, J. Sietsma, S.E. Offerman, J.P. Wright,
exponential carbon dependence, Mater. Sci. Technol. 28 (2012) 487–495, http:// S. van der Zwaag, Martensitic transformation of individual grains in low-alloyed
dx.doi.org/10.1179/1743284711Y.0000000097. TRIP steels, Scr. Mater. 56 (2007) 421–424, http://dx.doi.org/10.1016/j.
[27] G. Kurdjumow, E. Kaminsky, Eine röntgenographische Untersuchung der Struktur scriptamat.2006.10.041.
des gehärteten Kohlenstoffstahls, Z. Phys. 53 (1929) 696–707, http://dx.doi.org/ [47] R. Blondé, E. Jimenez-Melero, L. Zhao, J.P. Wright, E. Brück, S. van der Zwaag,
10.1007/BF01330828. N.H. van Dijk, High-energy X-ray diffraction study on the temperature-dependent
[28] J. Mazur, Lattice parameters of martensite and austenite, Nature 166 (1950) 828. mechanical stability of retained austenite in low-alloyed TRIP steels, Acta Mater. 60
[29] W.J. Wrazej, Lattice spacing of retained austenite in iron-carbon alloys, Nature 163 (2012) 565–577, http://dx.doi.org/10.1016/j.actamat.2011.10.019.
286