Predicting Elastic Anisotropy of Dual-Phase Steels Based On Crystal Mechanics and Microstructure
Predicting Elastic Anisotropy of Dual-Phase Steels Based On Crystal Mechanics and Microstructure
a r t i c l e i n f o a b s t r a c t
Keywords: This work utilizes mean-field self-consistent and full-field fast Fourier transform-based homogenizations to study
Microstructures the effective elastic behavior of several steels: three dual-phase (DP), DP 590, DP 980, and DP 1180, and one
Elastic material martensitic (MS), MS 1700. Crystallographic textures and phase fractions of these steels are characterized using
Anisotropic material
electron microscopy along with electron-backscattered diffraction to initialize the models. A comprehensive set of
Numerical algorithms
Young’s modulus and Poisson’s ratio data, measured at the ambient temperature as a function of orientation with
Dual-phase steel
respect to the rolling direction for each steel sheet, is used to calibrate and validate the models. The calibration
of the models enabled us to estimate the single crystal elastic constants for both the martensitic phase and ferrite,
while calculating the orientation dependent effective properties. Half of the data was used in the calibration.
Subsequent predictions of the orientation dependent effective elastic properties for the remaining data verified
that the estimated single crystal properties are reliable. As the steels exhibit a different level of anisotropy in their
effective behavior, good predictions allowed us to discuss the role of texture, grain structure, phase fraction and
distribution on the effective properties. The results of this work represent a significant incentive to introduce elas-
tic anisotropy in numerical tools for simulating metal forming processes of dual-phase steels, in particular those
processes involving springback, using the texture informed crystal mechanics-based models to more accurately
estimate the effective elastic properties required by such simulations.
1. Introduction The crystal elastic properties play the central role in governing the
extent of backstress fields under load, or residual stress fields upon load
Simulations of manufacturing processes and mechanical designs in- removal [4]. A tensorial measure describing the resistance to stretch
volving polycrystalline metals are performed more accurately if the ma- a crystal elastically in an anisotropic manner is referred to as elastic
terial constitutive behavior is modeled as anisotropic. The basic build- stiffness. Capturing the distribution of elastic backstress fields is espe-
ing block of a polycrystalline material, the single crystal referred to as cially important for modeling the material behavior during strain-path-
a grain having a certain crystal lattice orientation, is known to exhibit changes [5]. Upon a strain-path-change, the material exhibits elastic
anisotropic behavior [1,2]. Thermo-mechanical processing of polycrys- unloading followed by a change in the yield stress from the one reached
talline aggregates generates a distribution of crystal lattice orientations, at the end of pre-straining, while the newly applied stress combines
which is referred to as crystallographic texture [3]. Texture is not ran- with the existing elastic backstress field [6,7]. After a metallic part is
dom but actually preferred depending on the directionality of active removed from a forming die, the field further evolves to a new equi-
crystallographic deformation mechanisms, which accommodate shape librium, which is accompanied with geometrical changes of the part
change during shaping operations. A preferred distribution of crystal referred to as springback. For example, upon completion of sheet metal
orientations can either enhance or suppress the anisotropy of effective forming, deep-drawn or stretch-drawn parts “spring back”, affecting the
properties. Interactions between individual grains of different crystal dimensional accuracy of a finished part. The dimensional changes are
orientations exhibiting anisotropy in their properties under load creates directly dependent on the effective elastic properties of the sheet. Thus,
inter-granular stress fields. The local mechanical fields are even more the constitutive model used in predicting springback in metal forming
heterogeneous in multi-phase materials such as dual-phase (DP) steels should consider elastic anisotropy, as highlighted in [8,9]. Accurate pre-
containing a distribution of martensite and ferrite phases with contrast- dictions of springback are accentuated with the development of mod-
ing mechanical characteristics. ern alloys for the transportation industry for vehicle light-weighting
∗
Corresponding author: Department of Mechanical Engineering, University of New Hampshire, 33 Academic Way, Kingsbury Hall, W119, Durham, NH 03824,
USA.
E-mail address: [email protected] (M. Knezevic).
https://doi.org/10.1016/j.ijmecsci.2018.12.021
Received 11 November 2018; Received in revised form 7 December 2018; Accepted 11 December 2018
Available online 12 December 2018
0020-7403/© 2018 Elsevier Ltd. All rights reserved.
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
purposes, such as the advanced-high strength steels (AHSS), which ex- components of an orthogonal coordinate transformation matrix relating
hibit high flow stress or aluminum and magnesium alloys, which exhibit a crystal to a sample frame as
low stiffness, both of which accentuate the magnitude of springback.
⎡cos 𝜙1 cos 𝜙2 − sin 𝜙1 cos Φ sin 𝜙2 − cos 𝜙1 sin 𝜙2 − sin 𝜙1 cos Φ sin 𝜙2 sin 𝜙1 sin Φ ⎤
Premier examples of AHSS are DP steels containing variable fractions 𝐠 = ⎢sin 𝜙1 cos 𝜙2 + cos 𝜙1 cos Φ sin 𝜙2 − sin 𝜙1 sin 𝜙2 + cos 𝜙1 cos Φ cos 𝜙2 − cos 𝜙1 sin Φ⎥
⎢ ⎥
of martensite vs. ferrite phases in their microstructure. This paper is ⎣ sin Φ sin 𝜙2 sin Φ cos 𝜙2 cos Φ ⎦
concerned with the effective elastic behavior of DP steels based on the (2)
single crystal mechanics and microstructure.
The first-order lower (Reuss) and upper (Voigt) bounds for the effec-
Estimates of the effective elastic behavior of DP steels depend on
tive elastic stiffness tensor have been well established in the literature
microstructure and the accuracy of available single crystal constants for
[15–22]. The bounds for the diagonal components are
martensite and ferrite along with the homogenization procedure linking ( ) ⟨ ⟩
the crystal to the overall effective elastic response of the polycrystal. ⟨𝐒𝑐 ⟩−1 ≤ 𝐶𝑖𝑗𝑖𝑗
∗ 𝑐
≤ 𝐶𝑖𝑗𝑖𝑗 (3)
Published experimental measurements and theoretical estimates of sin- 𝑖𝑗𝑖𝑗
gle crystal elastic constants for martensitic and ferrite phases are rare, while those for the off-diagonal components are
as they are difficult to measure and also difficult to theoretically cal- (⟨ ⟩ ) √
𝑐
culate. Furthermore, the reported values vary substantially. Tables that max 𝐶𝑖𝑗𝑘𝑙 , (⟨𝐒𝑐 ⟩−1 )𝑖𝑗𝑘𝑙 − Δ𝑖𝑗𝑖𝑗 Δ𝑘𝑙𝑘𝑙
summarize values based on a literature review are given in Appendix A. (⟨ ⟩ ) √
𝑐
As is evident, while the elastic crystal stiffness coefficients of iron (Fe) ≤ 𝐶𝑖𝑗𝑘𝑙
∗
≤ min 𝐶𝑖𝑗𝑘𝑙 , (⟨𝐒𝑐 ⟩−1 )𝑖𝑗𝑘𝑙 + Δ𝑖𝑗𝑖𝑗 Δ𝑘𝑙𝑘𝑙 , (4a)
have been reliably determined by many independent sources, the stiff-
⟨ ⟩
ness coefficients of ferrite and martensite have yet to be established with 𝑐
Δ𝑖𝑗𝑘𝑙 = 𝐶𝑖𝑗𝑘𝑙 − (⟨𝐒𝑐 ⟩−1 )𝑖𝑗𝑘𝑙 . (4b)
confidence.
We present estimates of the single crystal elastic coefficients for Here, no implicit summation on repeated indices is used.1 The angle
martensite and ferrite, based on the calibration and validation of mean- brackets ⟨ ⟩ are used to denote the volume averaged value over con-
field self-consistent (SC) and full-field fast Fourier transform (FFT)- stituent crystals in a polycrystal. Sc = (Cc ) − 1 is the single crystal com-
based homogenizations towards a comprehensive set of Young’s mod- pliance tensor.
ulus and Poisson’s ratio data measured at the ambient temperature as The constitutive relation for linear elasticity that holds for
a function of orientation with respect to the rolling direction (RD) for anisotropic materials expressed in terms of matrices in the Voigt form
three DP steels, DP 590, DP 980, and DP 1180, and one martensitic (MS) [23] is
steel, MS 1700. The effective elastic properties have been measured at
⎛Σ11 ⎞ ⎛𝐶1111 𝐶1122 𝐶1133 𝐶1123 𝐶1131 𝐶1112 ⎞ ⎛𝐸11 ⎞
∗ ∗ ∗ ∗ ∗ ∗
15° increments from the RD of the sheets. The specific models consid-
⎜Σ22 ⎟ ⎜𝐶2211
∗
𝐶2222
∗
𝐶2233
∗
𝐶2223
∗
𝐶2231
∗ ∗ ⎟⎜
𝐶2221 𝐸22 ⎟
ered in the present work are the elasto-plastic SC (EPSC) [10–12] and the ⎜ ⎟ ⎜ ∗ ∗ ⎟⎜ ⎟
elasto-viscoplastic FFT (EVPFFT) [13,14] in which the strain is restricted Σ 𝐶
⎜ ⎟ = ⎜ 3311
33 𝐶3322
∗
𝐶3333
∗
𝐶3323
∗
𝐶3331
∗
𝐶3312 ⎟ ⎜𝐸33 ⎟ (5)
to be elastic. While the former is much more computationally efficient, ⎜Σ23 ⎟ ⎜𝐶3211
∗
𝐶3222
∗
𝐶2333
∗
𝐶2323
∗
𝐶2331
∗
𝐶2312 ⎟ ⎜2𝐸23 ⎟
∗
⎜Σ ⎟ ⎜𝐶 ∗ 𝐶3122 𝐶3133 𝐶3132 𝐶3131 ∗ ⎟⎜
𝐶3112 ⎟
⎟ ⎜2𝐸31 ⎟
∗ ∗ ∗ ∗
the latter is regarded as more accurate and can be used to validate the ⎜ 31 ⎟ ⎜ 3111
⎝Σ12 ⎠ ⎝𝐶1211
∗
𝐶1222
∗
𝐶1233
∗
𝐶1223
∗
𝐶1231
∗
𝐶1212 2𝐸12 ⎠
∗ ⎠⎝
former. Texture and phase fractions of these steels are characterized
using scanning electron microscopy (SEM) and electron-backscattered In this reducing order notation, the macroscopic Cauchy stress ten-
diffraction (EBSD) to initialize the models. Taking the characterized mi- sor, 𝚺, and macroscopic strain tensor, E, are expressed as column vectors
crostructure and the single crystal elastic constants as inputs, the mod- and the effective stiffness tensor is expressed as a symmetric matrix. The
els calculate the orientation dependent effective elastic behavior of the inverse relation is
steels. The constants for martensite and ferrite are varied to fit the ori-
⎛𝐸11 ⎞ ⎛𝑆1111 𝑆1122 𝑆1133 2𝑆1123 2𝑆1131 2𝑆1112
∗ ∗ ∗ ∗ ∗ ∗
entation dependent and microstructure sensitive effective response of ⎞ ⎛Σ11 ⎞
DP 590 and MS 1700 using an optimization scheme and then verified ⎜𝐸22 ⎟ ⎜𝑆2211∗
𝑆2222
∗
𝑆2233
∗
2𝑆2223
∗
2𝑆2231
∗ ∗ ⎟⎜
2𝑆2221 Σ22 ⎟
⎜ ⎟ ⎜ ∗ ∗ ⎟⎜ ⎟
by predicting the response of DP 980 and DP 1180. Thus, the effec- ⎜𝐸 33 𝑆
⎟ = ⎜ 3311 𝑆3322
∗
𝑆3333
∗
2𝑆3323
∗
2𝑆3331
∗
2𝑆3312 ⎟ ⎜Σ33 ⎟ (6)
tive property measurements along with the microstructural measure- ⎜2𝐸23 ⎟ ⎜2𝑆3211∗
2𝑆3222
∗
2𝑆2333
∗
4𝑆2323
∗
4𝑆2331
∗
4𝑆2312 ⎟ ⎜Σ23 ⎟
∗
⎜2𝐸 ⎟ ⎜2𝑆 ∗ 2𝑆3122
∗
2𝑆3133
∗
4𝑆3132
∗
4𝑆3131
∗ ∗ ⎟⎜
4𝑆3112 ⎟
ments are computationally linked to establish single crystal elastic con- ⎜ 31 ⎟ ⎜ 3111 ⎟ ⎜Σ13 ⎟
stants for ferrite and martensite. As the steels exhibit a different level ⎝2𝐸12 ⎠ ⎝2𝑆1211∗
2𝑆1222
∗
2𝑆1233
∗
4𝑆1223
∗
4𝑆1231
∗
4𝑆1212 Σ12 ⎠
∗ ⎠⎝
where 𝛿 ij represents the Kronecker delta symbol, C12 , C12 , and C44 are 1
The Einstein indicial notation of implicit summation on repeated indices is
the single crystal elastic stiffness constants for a cubic crystal, and gij are employed in this paper, except when explicitly noted otherwise.
640
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
where C(x) is the local elastic stiffness at x in a grain c. The local elastic ∫ 1 [𝐺𝑘𝑖,𝑗𝑙 (𝐱 − 𝐱′ ) + 𝐺𝑙𝑖,𝑗𝑘 (𝐱 − 𝐱′ )]𝑑 𝐱′ is uniform in the volume
stiffness can be expressed in terms of an elastic stiffness of a fictitious 𝑉𝑐 2
homogenous reference medium as of the inclusion, meaning that the strain inside the inclusion
is also uniform. The symmetric Eshelby tensor is defined as
𝐂(𝐱) = 𝐂0 + 𝛿𝐂(𝐱), (9) 𝑐
𝑆klpq = ∫ 12 [𝐺ki,jl (𝐱 − 𝐱′ ) + 𝐺li,jk (𝐱 − 𝐱′ )]𝑑𝐱′ 𝐶ijpq
∗ [29]. Then, the strain in
𝑉𝑐
where C0 is the elastic stiffness of the homogenous reference medium
the inclusion is
and 𝛿C(x) is the deviation of the local elastic stiffness at x from [ ( )−1 ( )]−1
elastic stiffness of the homogenous reference medium. The equilib- 𝜀𝑐𝑘𝑙 = 𝐼𝑘𝑙𝑚𝑛 + 𝑆𝑘𝑙𝑝𝑞
𝑐
𝐶𝑝𝑞𝑖𝑗
∗ 𝑐
𝐶𝑖𝑗𝑚𝑛 − 𝐶𝑖𝑗𝑚𝑛
∗
𝐸𝑚𝑛 = 𝐴𝑐𝑘𝑙𝑚𝑛 𝐸𝑚𝑛 (14)
rium statement in the absence of a body force can be written using
Eqs. (8) and (9) where Iijkl is the fourth rank identity tensor, 𝐴𝑐𝑖𝑗𝑘𝑙 is a localization (or
( ) concentration) tensor. Using Eqs. (5), 7a), (7b) and (14), the effective
𝐶𝑖𝑗𝑘𝑙
0
𝑢𝑘,𝑙𝑗 (𝐱) + 𝛿𝐶𝑖𝑗𝑘𝑙 (𝐱)𝑢𝑘,𝑙 (𝐱) ,𝑗 = 0, (10)
stiffness is
where uk (x) is the local displacement vector and the product of 𝛿Cijkl (x) 𝐶ijkl
∗ 𝑐
=< 𝐶ijpq 𝐴𝑐pqmn > < 𝐴𝑐mnkl >−1 . (15)
and uk,l (x) is known as the polarization field, 𝜙ij . Eq. (10) can be viewed
as the equilibrium statement for the homogenous reference medium Since the localization tensor, 𝐴𝑐𝑝𝑞𝑚𝑛 , is a function of the effective stiff-
with a prescribed distribution of the fictitious body force, 𝜙ij,j . Using ness, 𝐶𝑖𝑗𝑘𝑙
∗ , Eq. (15) is solved numerically using the fixed point iterations
the Green function method allows transformation of Eq. (10) into an [10].
integral equation
2.1.2. Full-field FFT homogenization for estimating effective elastic
( ) ( ) properties
𝑢̃ 𝑘,𝑙 (𝐱) = 𝐺𝑘𝑖,𝑗𝑙 𝐱 − 𝐱 𝜙𝑖𝑗 𝐱′ 𝑑 𝐱′ ,
′
(11)
∫ Effective properties of a microstructural cell embedding crystal prop-
𝑅3 erties can be solved using full-field approaches like finite elements with
where 𝑢̃ 𝑘,𝑙 (𝐱) is the local displacement gradient fluctuation, Gki (x − x′) sub-grain mesh resolution [30–36] or a Green’s function-based method,
is Green’s function for the infinite homogenous reference medium, and which relies on the efficient FFT algorithm to solve the convolution in-
the integral is generically over the entire three-dimensional space, R3 . tegral representing stress equilibrium under strain compatibility con-
Eqs. (5), (7a), and (7b) together with the solution of the integral in straint over a voxel-based microstructural cell. The later approach is
Eq. (11) facilitate solving for the effective properties of the polycrystal utilized here as a more accurate but less computationally efficient alter-
[26]. In this work, we will use two approaches to solve for the integral native to the SC homogenization.
in Eq. (11) in order to estimate homogenized effective stiffness for poly- In this formulation, Eq. (11) is solved in Fourier space, which is then
crystals: (1) the mean-field SC approach and (2) the full-field FFT-based followed by the inverse transform to obtain the strain as
( ( ) )
approach. ̂0 (𝐤) 𝜙̂ kl (𝐤) ,
𝜀ij (𝐱) = 𝐸ij + 𝐹 𝑇 −1 sym Γ ijkl
(16)
2.1.1. Mean-field self-consistent homogenization for estimating effective where the symbols "ˆ" and FT−1 indicate direct and inverse Fourier trans-
elastic properties forms, respectively, while sym indicates the symmetric portion and k is
This section summarizes the solution to Eq. (11) using the SC ap- a point (frequency) in Fourier Space. The fourth order tensor Γ̂ 0𝑖𝑗𝑘𝑙 (𝐤) is
proach, which is valid for granular media i.e. polycrystals. The SC ap-
proach facilitates the evaluation of the effective elastic stiffness in a Γ̂ 0𝑖𝑗𝑘𝑙 (𝐤) = −𝑘𝑗 𝑘𝑙 𝐺̂ 𝑖𝑘 (𝐤); 𝐺̂ 𝑖𝑘 (𝐤) = [𝐶𝑘𝑗𝑖𝑙 𝑘𝑙 𝑘𝑗 ]−1 . (17)
computationally efficient manner. The underlying assumptions can be
Guessing the stress, 𝜆ij and corresponding elastic strain, eij
(i) , (i) , at an
found elsewhere e.g. [26,27], while the actual SC scheme used here is
iteration (i), the polarization field is
from [10,28].
Each crystal is represented as an ellipsoidal heterogeneity within 𝜙𝑖𝑗 (𝑖) (𝐱) = 𝜆𝑖𝑗 (𝑖) (𝐱) − 𝐶 0 𝑖𝑗𝑘𝑙 𝑒𝑘𝑙 (𝑖) (𝐱). (18)
the infinite homogenous reference medium under applied displacement. The new guess for strain field at (i + 1) is then
( ( ) )
Elastic stiffness of the reference homogenous medium is set equal to the
𝑒𝑖𝑗 (𝑖+1) (𝐱) = 𝐸𝑖𝑗 + 𝐹 𝑇 −1 𝑠𝑦𝑚 Γ̂ 0𝑖𝑗𝑘𝑙 (𝐤) 𝜙̂ (𝑘𝑙𝑖) (𝐤) . (19)
unknown effective elastic stiffness i.e. 𝐶𝑖𝑗𝑘𝑙 0 = 𝐶 ∗ , while the elastic stiff-
𝑖𝑗𝑘𝑙
𝑐
ness of each crystal, 𝐶𝑖𝑗𝑘𝑙 , is assumed to be uniform within the crystal The stress field is iteratively solved using a suitably defined residual
volume c. The problem can be reduced to Eshelby’s inclusion problem 𝑅𝑘 (𝝈 (𝑖+1) ) = 𝜎𝑘 (𝑖+1) + 𝐶𝑘𝑙 0 𝜀𝑙 (𝑖+1) (𝝈 (𝑖+1) ) − 𝜆𝑘 (𝑖) − 𝐶𝑘𝑙 0 𝑒𝑙 (𝑖+1) = 0. (20)
[27] with an appropriate choice of eigenstrain, meaning that the strain
In Eq. (20), the Voigt form is employed as
and stress within the heterogeneity are constant. The polarization ten-
sor, 𝜙ij (x′), is zero outside the inclusion volume and constant within 𝜎ij → 𝜎𝑘 , 𝑘 = 1..6,
(21)
the inclusion i.e. fictitious body force is only present in the inclusion 𝐶ijkl → 𝐶mn 𝑚, 𝑛 = 1..6.
volume. Eq. (11) can be integrated over the volume of the inclusion to The solution is obtained using the Newton’s solution procedure
calculate the average displacement gradient deviation in the inclusion ( )−1
𝜕 𝑅𝑘 ||
𝜎𝑘 (𝑖+1,𝑗+1) = 𝜎𝑘 (𝑖+1,𝑗) − 𝑅𝑙 (𝝈 (𝑖+1,𝑗) ), (22)
𝜕 𝜎𝑙 ||𝝈
[26] ( 𝑖 +1 ,𝑗)
( ) ( )
1 where j enumerates the iterations for stress. The convergence is con-
𝑢̃ 𝑐𝑘,𝑙 = 𝐺 𝑐
𝐱 − 𝐱′ 𝑑 𝐱′ 𝑑𝐱 𝐶𝑖𝑗𝑚𝑛 − 𝐶𝑖𝑗𝑚𝑛
∗
𝑢𝑐𝑚,𝑛 , (12)
𝑉𝑐 ∫ ∫ 𝑘𝑖,𝑗𝑙 trolled by
𝑉𝑐 𝑉𝑐
𝜕𝑅𝑘 ||
= 𝛿kl + 𝐶kq 0 𝐶ql −1 (23)
where 𝑢𝑐𝑚,𝑛 is the average displacement gradient in the inclusion i.e. crys- 𝜕𝜎𝑙 ||𝝈 (𝑖+1,𝑗 )
tal. Defining the strain in the inclusion as the symmetric part of the dis-
The stress solution is used as a next guess in Eq. (18) until convergence
placement gradient, 𝑢𝑐𝑘,𝑙 , and invoking the minor symmetry of the elastic
is achieved in which 𝝈 approaches 𝝀 and 𝜺 approaches e.
stiffness, gives
Upon the convergence, a homogenization over voxels of the mi-
( )
1 1[ ( ) ( )] crostructural cell is performed as follows
𝜀𝑐kl = 𝐸kl + 𝑐
𝐺ki,jl 𝐱 − 𝐱′ + 𝐺li,jk 𝐱 − 𝐱′ 𝑑 𝐱′ 𝑑 𝐱 𝐶ijmn ∗
−𝐶ijmn 𝜀𝑐mn ∑ ( )
𝑉𝑐 ∫ ∫ 2
𝑉𝑐 𝑉𝑐 𝑋,𝑌 ,𝑍 𝜎ij (𝐱 )
Σij = ; 𝑁 = # of voxels in 𝑋, 𝑌 , 𝑍,
𝑁3
(13) ∑ ( )
𝑋,𝑌 ,𝑍 𝜀ij (𝐱 )
where Ekl is the macroscopically imposed strain tensor i.e. 𝐸ij = . (24)
𝑁3
the strain in the homogenous reference medium. The integral
641
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
Fig. 1. Stereographic pole figures showing initial texture of (a) DP 590, (b) DP 980, and (c) DP 1180 steel sheets.
Since an equivalent expression to Eq. (15) for the full-field FFT ho- Finally, shear modulus, 𝜇 12 , is
mogenization is not readily available, an alternative procedure is used 1
to obtain the effective macroscopic stiffness for a given microstructural Shear Modulus = (29)
4𝑆1212
∗
cell. The procedure involves several steps:
The stiffness tensor, C∗ , given with respect to the reference frame
i Set six simulations such that of the sheet, can be transformed using the coordinate transformation
⎡1.0 0.0 0.0⎤ ⎡0.0 0.0 0.0⎤ ⎡0.0 0.0 0.0⎤ law into another frame 𝐞′𝑖 . The direction cosines transformation matrix
𝐸𝑖𝑗(1) = ⎢0.0 0.0 0.0⎥, 𝐸𝑖𝑗(2) = ⎢0.0 1.0 0.0⎥, 𝐸𝑖𝑗(3) = ⎢0.0 0.0 0.0⎥, for change of basis is a dot product between vectors representing the
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ reference frame of the sheet and the transformed frame, 𝑄𝑖𝑗 = 𝐞𝑖 ⋅ 𝐞′𝑗 .
⎣0.0 0.0 0.0⎦ ⎣0.0 0.0 0.0⎦ ⎣0.0 0.0 1.0⎦
′
⎡0.0 0.0 0.0⎤ ⎡0.0 0.0 0.5⎤ ⎡0.0 0.5 0.0⎤ 𝐶𝑖𝑗𝑘𝑙
∗
= 𝑄𝑖𝑝 𝑄𝑗𝑞 𝑄𝑘𝑟 𝑄𝑙𝑠 𝐶𝑝𝑞𝑟𝑠
∗
(30)
𝐸𝑖𝑗(4) = ⎢0.0 0.0 0.5⎥, 𝐸𝑖𝑗(5) = ⎢0.0 0.0 0.0⎥, 𝐸𝑖𝑗(6) = ⎢0.5 0.0 0.0⎥.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ The transformation matrix, Qij , can be conveniently established using
⎣0.0 0.5 0.0⎦ ⎣0.5 0.0 0.0⎦ ⎣0.0 0.0 0.0⎦ ′
two angles, where one rotates about e3 (i.e. 𝜙1 in Eq. (2)) to obtain 𝐶𝑖𝑗𝑘𝑙
∗
(25)
along any in-plane sheet orientation and another rotates about e1 (i.e.
∗′ along any orientation in three-dimensional
Φ in Eq. (2)) to obtain 𝐶𝑖𝑗𝑘𝑙
ii Run each simulation to obtain the individual components of elastic
′
stiffness (3D) space. The transformed stiffness matrix 𝐶𝑖𝑗𝑘𝑙
∗ is then inverted to get
𝑒𝑓 𝑓 ∗′
𝑆𝑖𝑗𝑘𝑙 , which is then used to determined Young’s modulus and Poisson’s
Σ(1) (1)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝐸𝑘𝑙 = 𝐶
∗ (1)
𝑖𝑗11 𝐸11 = 𝐶𝑖𝑗11
∗
′
1 𝑆2211
∗
Σ(2) (2) (2)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝐸𝑘𝑙 = 𝐶𝑖𝑗22 𝐸22 = 𝐶𝑖𝑗22
∗ ∗ ∗ ratio along selected in-plane sheet orientation using ∗′
and − ∗′
,
𝑆1111 𝑆1111
𝑒𝑓 𝑓 respectively.
Σ(3)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙
∗ (3)
𝐸𝑘𝑙 = 𝐶𝑖𝑗33
∗ (3)
𝐸33 =𝐶 𝑖𝑗33
( )
1 ∗
Σ(4) ∗ (4) ∗ (4) ∗ (4)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝐸𝑘𝑙 = 𝐶𝑖𝑗23 𝐸23 + 𝐶𝑖𝑗32 𝐸32 = 𝐶𝑖𝑗23 + 𝐶𝑖𝑗32
∗
= 𝐶𝑖𝑗23
∗
3. Results
2
( )
1 ∗
Σ(5)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙
∗ (5)
𝐸𝑘𝑙 = 𝐶𝑖𝑗13
∗ (5)
𝐸13 + 𝐶𝑖𝑗31
∗ (5)
𝐸31 = 𝐶 + 𝐶𝑖𝑗31
∗
= 𝐶𝑖𝑗13
∗
3.1. Microstructure
2 𝑖𝑗13
( )
1 ∗
Σ(6)
𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙
∗ (6)
𝐸𝑘𝑙 = 𝐶𝑖𝑗12
∗ (6)
𝐸12 + 𝐶𝑖𝑗21
∗ (6)
𝐸21 = 𝐶 + 𝐶 ∗𝑖𝑗21 = 𝐶𝑖𝑗12
∗
The materials investigated in this study are commercial steel
2 𝑖𝑗12
𝑖𝑗 = 11, 22, 33, 23, 13, 12. (26) sheets received from US Steel. The studied steels are typical of auto-
body/structure applications. We begin by presenting texture for the
iii Form the effective elastic stiffness tensor from the components ob- sheets as characterized using EBSD. Fig. 1 shows stereographic pole fig-
tained in Eq. (26). ures visualizing the texture for the ferrite phase in DP 590, DP 980, and
DP 1180. The data for DP 590 was taken from [37]. The EBSD scans
2.2. Obtaining Young’s modulus, Poisson’s ratio, and shear modulus from were run over a very large area for each steel to obtain statistically
an effective elastic tensor significant data. The texture evolution in the sheets resembles classi-
cally reported orthotropic rolled texture for body-centered cubic (BCC)
The effective elastic stiffness and compliance tensors as outputs from materials, where a majority of the grains are concentrated around the
the homogenization approaches are defined with respect to the rolling 𝛾-fiber and a portion of the 𝛼-fiber [3,38–45]. We have attempted to
direction (RD) as e1 , the transverse direction (TD) as e2 , and the nor- separate the texture of the martensitic phase in each steel and obtained
mal direction (ND) as e3 of the steel sheets: 𝐂∗ = 𝐶𝑖𝑗𝑘𝑙
∗ 𝐞 ⊗𝐞 ⊗𝐞 ⊗𝐞
𝑖 𝑗 𝑘 𝑙 these to be weak (i.e. approximately random). While the measured EBSD
and 𝐒∗ = 𝑆𝑖𝑗𝑘𝑙
∗ 𝐞 ⊗ 𝐞 ⊗ 𝐞 ⊗ 𝐞 , respectively. Young’s modulus along e
𝑖 𝑗 𝑘 𝑙 1 scans consisted of over a million index-able points with high confidence
can be determined by examining the case of a simple tension in e1 . index for ferrite for each sheet, textures used in the simulations were
Eq. (6) reduces to 𝐸11 = 𝑆1111
∗ Σ which can be rearranged into
11 compacted to 1,000 weighted orientations using the recently developed
∑ procedures [46,47]. Sum of the weights corresponded to the fraction of
1
Young’s Modulus = 11 = ∗ . (27) ferrite for each steel. Texture used in the simulation for the martensitic
𝐸11 𝑆1111
phase was also represented by 1000 orientations, which were approxi-
Similarly, Poisson’s ratio, 𝜈 12 can be determined using the simple tension mated as uniform random. Likewise, for ferrite, the sum of the weights
case as of orientations for the martensite phase corresponded to the measured
𝑆∗ fraction of martensite for each steel. The phase fractions were measured
𝐸22
Poisson’s Ratio = − = − 2211 (28) using a combination of EBSD and SEM imaging, where the latter method
𝐸11 𝑆1111
∗
642
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
Table 1 the crystal stiffness coefficients of ferrite and martensite while calculat-
Measured fraction of martensitic phase per steel. ing the orientation dependent effective data for Young’s modulus and
Steel DP 590 DP 980 DP 1180 MS 1700 Poisson’s ratio with respect to RD for the two sheets. The data is taken
from [49]. Additionally, the data for the shear modulus was available
Fraction 0.077 0.33 0.42 0.90
for MS 1700 [50]. We set a direct-search optimization algorithm to find
the optimal constants while matching the experimental data with the
was used to verify the former method. Since regions of martensite ap- output of the SC homogenization.
pear as dark regions in image quality (IQ) maps [37,48], the volume Since the elastic stiffness coefficients of ferrite are expected to be sim-
fraction of martensite was determined using a threshold procedure from ilar to those of Fe, we used them as the initial guess. The elastic crys-
IQ maps of high resolution. The obtained estimates were verified by an- tal stiffness coefficients of Fe have been reliably determined by many
alyzing many SEM images. Table 1 presents the measured fraction of independent sources (Appendix A). The two-tier direct-search optimiza-
martensite for each steel. tion algorithm was set to minimize the error between the calculated
and mean of the measured effective values for two steels in seven di-
3.2. Calibration rections (RD = 0°, 15°, 30°, 45°, 60°, 75°, and TD = 90°) to determine the
crystal elastic stiffness coefficients for ferrite and martensite. For every
To study elastic anisotropy of dual-phase steels as a function of their combination of the constants, the effective stiffness tensors are calcu-
microstructural features, we first employ the SC homogenization to esti- lated for the two steels using the SC homogenization model. The two
mate the effective elastic properties. The microstructural variables con- effective stiffness tensors are then passed to the program for perform-
sidered by the SC homogenization are restricted to texture and phase ing the change of basis and obtaining Young’s modulus, Poisson’s ratio,
fraction. The steels used in this study have similar texture but con- and shear modulus for two steels in every direction. The first stage of
tain an increasing volume fraction of martensite within a ferrite matrix the program considered only the experimental data for Poisson’s ratio
(Table 1). The content of martensite is likely the main microstructural and varied all six coefficients at once. The second stage of the program
feature governing the difference in their elastic behavior. To begin, tex- kept the anisotropy ratio constant and varied only the overall magni-
ture and phase fraction of DP 590 as a predominantly ferritic streel and tude of the constants to minimize the error for Young’s Modulus. Thus,
MS 1700 as a predominantly martensitic steel are used to appropriately both the level of anisotropy and the magnitude of the constants were
initialize the SC modeling framework put together to extract the single adjusted. The optimization scheme attempted to increase and decrease
crystal constants. As a consequence of cubic crystal symmetry, BCC for each crystal constant by a predetermined step size and the combination
ferrite and body-centered tetragonal for martensite require only three with the lowest error from the experimental data was used as the new
crystal stiffness coefficients, C11 , C12 , and C44 to define their single- initial value. This process continued until no change in the new initial
crystal stiffness tensors. These three crystal constants per phase are a constant values from the old values occurred.
user defined input in the SC model, in addition to the input in terms of Fig. 2 shows the comparison between measured and calculated vari-
the phase fractions and texture per phase. The objective is to determine ation of effective Young’s modulus and Poisson’s ratio with the in-plane
643
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
Fig. 3. Contour plots showing anisotropy of Young’s modulus with orientation for single crystalline (a) ferrite phase and (b) martensitic phase.
Fig. 4. Front, side, and top views of 3D voxel-based microstructural cells generated in DREAM.3D to synthetically represent grain structure of DP 590: (a) microstruc-
tural cell #, 1 (b) microstructural cell #2, (c) microstructural cell #3, (i) texture shuffle #1, (ii) texture shuffle #2, (iii) texture shuffle #3. The colors represent
different grains. These numerical setups are used in the EVPFFT simulations of the effective elastic properties for DP 590.
644
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
645
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
the full-field FFT predictions. Additionally, the property bounds are plot- Table 4
ted. First, it is observed that the effective values are within the bounds. Comparison of measured [50] and predicted in-
Next, the effective elastic stiffness estimated using the full-field FFT plane shear modulus (𝜇12 ).
solver and the SC approach are very similar, validating the single crystal DP 980 (GPa) DP 1180 (GPa)
elastic stiffness constants for both phases. Finally, the good agreement
Measured 78.1 75.0
between the SC and FFT-based estimates suggests that texture and phase SC Predicted 79.40 77.53
fractions play the major role in governing the effective elastic behavior,
while grain, phase, and misorientation distributions are secondary. Al-
though the grains of the martensite are much different than grains of
the ferrite, the morphology effects still seem secondary. since the same set of single crystal constants for ferrite and marten-
site (Table 3) are used. The initial textures shown in Fig. 1 were used
4. Discussion to initialize the ferrite phase per steel used in the SC model. Compar-
ison of measured [49] and predicted variation of the effective proper-
This work utilized the single crystal mechanics, homogenization the- ties with the in-plane sheet orientation is given in Fig. 6. Furthermore,
ories, sets of microstructural characterization data, and sets of orienta- Table 4 shows the predicted values of the shear modulus for both steels.
tion dependent Young’s modulus and Poisson’s ratio data to reliably Considering the predicted magnate of the effective properties and the
determine the single crystal elastic stiffness constants for ferrite and shape of the curves, we regard these predictions as reasonably good.
martensite. Two steels were suitably selected to facilitate the accurate Appendix B presents values of the elastic stiffness tension components
determination of the constants per phase, one containing over 90% of for the four studied steels.
ferrite (DP 590) and another containing up to 90% of martensite (MS Clearly, the four steels studied in this work exhibit a moderate level
1700). Crystallographic texture and phase fractions have been charac- of anisotropic effective elastic behavior, where both their Young’s mod-
terized using EBSD and SEM. Taking texture and phase fractions as input ulus and the Poisson’s ratios depend on orientation with respect to the
into the model, the effective elastic behavior of two steels was modelled sheet loading direction. As has been determined, the individual crystals
using the SC homogenization while calibrating the constants by match- exhibit the crystalline anisotropy and symmetry. The contour plots pre-
ing the measured anisotropy. To this end, an optimization algorithm sented in Fig. 3 reveal the level of orientation dependence of Young’s
was developed to search for the optimal constant. The optimal constants modulus for ferrite and martensite in 3D. Interestingly, soft vs. stiff
were then successfully used within the full-field FFT model to verify the directions are the opposite for the two phases. Moreover, the levels
elasticity response of DP 590. of elastic anisotropy are different. The estimated Zener anisotropy ra-
2𝐶
To further verify the established values, we simulate the orientation tio, 𝐴 = 𝐶 −44𝐶 , for ferrite is A = 2.0, which is slightly less than the
11 12
dependent Young’s modulus and Poisson’s ratio for DP 980 and DP 1180 value for Fe (A = 2.4 for Fe). While a value of 1.0 for A is for an
using the SC approach. These calculations are regarded as predictions isotropic material, that value for martensite is A = 0.95. The values of the
646
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
constants and the small level of anisotropy we obtain for martensite ing metal forming processes of dual-phase steels using the texture in-
are in good agreement with the experimental measurements presented formed crystal mechanics-based models to more accurately estimate the
in [52], which are given in table A4 of the Appendix A. The effective effective elastic properties required by such simulations. Furthermore,
behavior of polycrystalline aggregates of these steels is a direct con- texture evolution can be simulated during a forming process using crys-
sequence of texture and phase fractions, homogenizing the crystalline tal plasticity models such as EPSC and EVPFFT implying that the evolu-
effects. tion of the effective elastic behavior with plastic strain is also calculated
In closing, we reflect on possible sources influencing the accuracy during the forming process.
of the predictions, in addition to the intrinsic assumptions involved in
the homogenization approached we employed to estimate the effective
properties. We have used EBSD and SEM to initialize texture and phase Acknowledgments
fraction input to the models. While we have attempted to provide as
statistically significant as possible data based on these techniques, it This research was sponsored by the U.S. National Science Foundation
is possible that spatial variation in the microstructure of studied steels and was accomplished under the CAREER grant no. CMMI-1650641. The
can cause these to be inaccurate. In particular, texture of martensite authors also acknowledge discussions with Professor Yannis P. Korkolis
was assumed as uniform random. To relax this concern, future research of the University of New Hampshire and Dr. Ricardo A. Lebensohn of
will involve generating the input to the models based on the data col- Los Alamos National Laboratory.
lected using neutron diffraction. Neutron diffraction allows for the mea-
surement of microstructural features averaged over large volumes due
to the deep penetration of thermal neutrons into most materials com- Appendix A
bined with beam spot sizes of ∼1 cm2 . The constants determined in
this work asummed single chemical composition for ferrite and single This appendix presents a summary of available single crystal elastic
chemical composition for martensite. The four studied steels have very coefficients for 𝛼-Iron (Fe), ferrite, and martensite, including the source
similar chemical composition but not identical. Solute atoms could in- of the values. (Tables A1–A5)
duce lattice strains, which could change lattice parameters. Such dis-
tortions could affect the single crystal elastic constants, and as a re-
Table A1
sult, the effective behavior. For example, the studied steels have a slight
Single crystal elastic stiffness coefficients in GPa for 𝛼-
difference in the content of carbon (C). C atoms within the interstitial iron (Fe) measured using either the ultrasonic pulse-echo
sites increase the interatomic constants and should increase the stiff- technique or resonant ultrasonic spectroscopy at room
ness on one hand. On the other hand, these atoms increase the vol- temperature (∼300 K) and atmospheric pressure.
ume of the lattice, reducing the stiffness. Thus, there are some com-
Source C11 C12 C44
peting effects [53] but considering the similarity in the composition of
the studied steels, the effect on solute on the determined crystal con- [55] 236.88 140.63 116.01
stants is likely to be secondary. In addition, any variation in the density [56] 241.50 146.63 111.73
[57] 228.09 133.48 110.86
of dislocations, especially in martensite, can have some effect on the
[58] 209.36 113.66 111.36
constants [54]. [59] 242.00 146.50 112.00
[59] 237.00 141.00 116.00
5. Conclusions [60] 233.10 135.44 117.83
[61] 228.00 132.00 116.50
[61] 223.00 127.00 115.00
This work has shown that it is possible to reliably estimate sin- [62] 231.40 134.70 116.40
gle crystal stiffness coefficients for ferrite and martensite using an ap- [63] 230.10 134.60 116.60
proach involving single crystal mechanics, homogenization theories, mi- [64] 226.00 140.00 116.00
crostructural characterization data, and orientation dependent Young’s [65] 232.20 135.60 117.00
[52] 231.50 135.00 116.00
modulus and Poisson’s ratio data. The microstructure data of DP 590 [66] 230.37 134.07 115.87
and MS 1700 steels was used to initialize the homogenization models
in terms of texture and phase fraction. The models were run within an
optimization scheme to extract the single crystal stiffness constants for Table A2
ferrite and martensite, while reproducing the effective property data. Single crystal elastic stiffness coefficients in GPa for 𝛼-
Subsequently, the constants were verified by predicting the effective iron estimated theoretically using first-principles calcula-
property data for DP 980 and DP 1180. It is shown that the model tak- tions based on density functional theory (DFT) methods.
ing the calibrated constants and microstructure as input can predict the Source C11 C12 C44
effective elastic properties for all studied steels containing various levels
[67] 276.6 145.8 97.58
of martensitic volume fraction. The effective behavior of the predomi- [68] 289 118 115
nantly martensitic steel, MS 1700, was predicted as nearly isotropic due [69] 279 140 99
to nearly isotropic behavior of a martensite single crystal. In contrast, [70] 271 145 101
the behavior of the predominantly ferrite steel, DP 590, was predicted [71] 303 150 126
[72] 297.8 141.9 106.7
as the most anisotropic amongst the studied steels, as a consequence of
[73] 279.2 148.8 93.0
anisotropic behavior of a ferrite single crystal and texture. Thus, the ef-
fective behavior of steels exhibiting a different level of anisotropy was
successfully predicted. Since the effective elastic stiffness estimated us- Table A3
ing the full-field FFT solver of several microstructural realizations, for Single crystal elastic stiffness coefficients in GPa for fer-
constant texture and phase fractions and the SC approach for the same rite, estimated theoretically using a micromechanical
texture and phase fraction, were nearly identical, it is inferred that tex- approach involving a crystallography-based anisotropic
thermomechanical continuum model.
ture and phase fractions play a primary role in governing the effective
elastic behaviors of steels, while grain, phase, and misorientation distri- Source C11 C12 C44
butions are secondary. The results of this work represent a significant
[74] 233.3 135.5 118.0
incentive to introduce elastic anisotropy in numerical tools for simulat-
647
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
Table A4
Elastic stiffness tensor values in GPa for martensite measured using resonant ultrasonic
spectroscopy at room temperature (∼300 K) and atmospheric pressure.
Source C1111 C2222 C3333 C1122 C1133 C2233 C2323 C3131 C1212
[52] 268.1 268.4 267.2 111.2 110.2 111.0 79.06 78.72 78.85
648
A.M. Cantara, M. Zecevic and A. Eghtesad et al. International Journal of Mechanical Sciences 151 (2019) 639–649
[37] Zecevic M, Korkolis YP, Kuwabara T, Knezevic M. Dual-phase steel sheets under [57] Kimura RI. On the elastic moduli of ferromagnetic materials. Part I. Dynamical mea-
cyclic tension–compression to large strains: experiments and crystal plasticity mod- surements of the elastic morduli of iron crystals. Nippon Sugaku-Buturigakkwai Kizi
eling. J Mech Phys Solids 2016;96:65–87. Dai 3 Ki 1939;21:686–706.
[38] Bunge H-J. Texture analysis in materials science. Mathematical methods. Göttingen: [58] Kimura RI. On the elastic moduli of ferromagnetic materials. Part II. The change
Cuvillier Verlag; 1993. in Young’s modulus, due to magnetisation and temperature. Proc Physico-Math Soc
[39] Ray R, Jonas JJ, Hook R. Cold rolling and annealing textures in low carbon and Jpn 3rd Ser 1939;21:786–99.
extra low carbon steels. Int Mater Rev 1994;39:129–72. [59] Alexandrov KS, Ryzhova TV. The elastic properties of crystals. Soviet Phys-Crystal-
[40] Randle V, Engler O. Introduction to texture analysis. Macrotexture, microstructure lograp 1961;6:228–52.
& orientation mapping. Gordon and Breach Science Publishers; 2000. [60] Rayne JA, Chandrasekhar BS. Elastic constants of Iron from 4.2 to 300°K. Phys Rev
[41] Von Schlippenbach U, Emren F, Lücke K. Investigation of the development of the 1961;122:1714–16.
cold rolling texture in deep drawing steels by ODF-analysis. Acta Metallurgica [61] Lord AEJ, Beshers DN. Elastic stiffness coefficients of Iron from 77° to 673°K. J Appl
1986;34:1289–301. Phys 1965;36:1620–3.
[42] Lücke K, Hölscher M. Rolling and recrystallization textures of BCC steels. Texture, [62] Rotter CA, Smith CS. Ultrasonic equation of state of iron: I. low pressure, room
Stress, Microstruct 1991;14:585–96. temperature. J Phys Chem Solids 1966;27:267–76.
[43] Knezevic M, Nizolek T, Ardeljan M, Beyerlein IJ, Mara NA, Pollock TM. Texture [63] Guinan MW, Beshers DN. Pressure derivatives of the elastic constants of 𝛼-iron to
evolution in two-phase Zr/Nb lamellar composites during accumulative roll bonding. 10 kbs. J Phys Chem Solids 1968;29:541–9.
Int J Plast 2014;57:16–28. [64] Leese J, Lord A Jr. Elastic stiffness coefficients of single‐crystal iron from room tem-
[44] Jahedi M, Paydar MH, Zheng S, Beyerlein IJ, Knezevic M. Texture evolution and perature to 500 °C. J Appl Phys 1968;39:3986–8.
enhanced grain refinement under high-pressure-double-torsion. Mater Sci Eng A [65] Dever D. Temperature dependence of the elastic constants in 𝛼‐iron single crystals:
2014;611:29–36. relationship to spin order and diffusion anomalies. J Appl Phys 1972;43:3293–301.
[45] Smith DH, Bicknell J, Jorgensen L, Patterson BM, Cordes NL, Tsukrov I, Knezevic M. [66] Adams JJ, Agosta D, Leisure R, Ledbetter H. Elastic constants of monocrystal iron
Microstructure and mechanical behavior of direct metal laser sintered Inconel alloy from 3–500 K. J Appl Phys 2006;100:113530.
718. Mater Charact 2016;113:1–9. [67] Souissi M, Numakura H. Elastic Properties of Fe–C and Fe–N Martensites. ISIJ Int
[46] Eghtesad A, Barrett TJ, Knezevic M. Compact reconstruction of orientation distribu- 2015;55:1512–21.
tions using generalized spherical harmonics to advance large-scale crystal plasticity [68] Vočadlo L, de Wijs GA, Kresse G, Gillan M, Price GD. First principles calcula-
modeling: Verification using cubic, hexagonal, and orthorhombic polycrystals. Acta tions on crystalline and liquid iron at Earth’s core conditions. Faraday Discuss
Mater 2018;155:418–32. 1997;106:205–18.
[47] Knezevic M, Landry NW. Procedures for reducing large datasets of crystal orienta- [69] Guo G, Wang H. Gradient-corrected density functional calculation of elastic con-
tions using generalized spherical harmonics. Mech Mater 2015;88:73–86. stants of Fe, Co and Ni in bcc, fcc and hcp structures. Chin J Phys 2000;38:949–61.
[48] Pinard PT, Schwedt A, Ramazani A, Prahl U, Richter S. Characterization of du- [70] Caspersen KJ, Lew A, Ortiz M, Carter EA. Importance of Shear in the bcc-to-hcp
al-phase steel microstructure by combined submicrometer EBSD and EPMA carbon transformation in Iron. Phys Rev Lett 2004;93:115501.
measurements. Microsc Microanal 2013;19:996–1006. [71] Sha X, Cohen RE. First-principles thermoelasticity of bcc iron under pressure. Phys
[49] Deng N, Korkolis YP. Elastic anisotropy of dual-phase steels with varying martensite Rev B 2006;74:214111.
content. Int J Solids Struct 2018;141-142:264–78. [72] Zhang H, Johansson B, Vitos L. Ab initio calculations of elastic properties of bcc
[50] Deng N, Korkolis YP. Determination of the shear modulus of orthotropic Fe-Mg and Fe-Cr random alloys. Phys Rev B 2009;79:224201.
thin sheets with the anticlastic-plate-bending experiment. J Eng Mater Technol [73] Shang SL, Saengdeejing A, Mei ZG, Kim DE, Zhang H, Ganeshan S, Wang Y, Liu ZK.
2018;140:041011–041011-041017. First-principles calculations of pure elements: equations of state and elastic stiffness
[51] Groeber MA, Jackson MA. DREAM. 3D: a digital representation environment for the constants. Comput Mater Sci 2010;48:813–26.
analysis of microstructure in 3D. Integr Mater Manufact Innovat 2014;3:5. [74] Tjahjanto DD, Turteltaub S, Suiker ASJ. Crystallographically based model for trans-
[52] Kim SA, Johnson WL. Elastic constants and internal friction of martensitic steel, formation-induced plasticity in multiphase carbon steels. Continuum Mech Thermo-
ferritic-pearlitic steel, and 𝛼-iron. Mater Sci Eng A 2007;452:633–9. dyn 2008;19:399–422.
[53] Ledbetter H, Austin M. Effects of carbon and nitrogen on the elastic constants of AISI [75] Tasan CC, Diehl M, Yan D, Zambaldi C, Shanthraj P, Roters F, Raabe D. Integrated
type 304 stainless steel. Mater Sci Eng 1985;70:143–9. experimental–simulation analysis of stress and strain partitioning in multiphase al-
[54] Granato AV, Lücke K. Theory of mechanical damping due to dislocations. J Appl loys. Acta Mater 2014;81:386–400.
Phys 1956;27:583–93. [76] Fellinger MR, Hector LG, Trinkle DR. Effect of solutes on the lattice parameters
[55] Goens E, Schmid E. Über die elastische Anisotropie des Eisens. Naturwissenschaften and elastic stiffness coefficients of body-centered tetragonal Fe. Comput Mater Sci
1931;19:520–4. 2018;152:308–23.
[56] Kimura R, Ohno K. On the elastic constants of single crystals of iron. Sci Rept Tohoku
Univ 1934;23:359–64.
649