Acta Materialia 187 (2020) 4150

Contents lists available at ScienceDirect

Acta Materialia
journal homepage: www.elsevier.com/locate/actamat

Full length article

Instabilities in the periodic arrangement of elastically interacting

precipitates in nickel-base superalloys
M. Degeitera,b, Y. Le Bouar*,b, B. Appolairec, M. Perruta, A. Finelb
a
DMAS, ONERA, Universite
Paris-Saclay, Cha ^tillon F-92322, France
b
LEM, ONERA, CNRS, Universite
Paris-Saclay, Cha ^tillon F-92322, France
c
IJL, CNRS, Universite
de Lorraine, Nancy F-54011, France

A R T I C L E I N F O A B S T R A C T

Article History: Nickel-base superalloys display cuboidal precipitates aligned along the cubic directions, which are the elastic
Received 12 September 2019 soft directions. At high precipitate volume fraction, the microstructure is often described as a regular array of
Revised 6 January 2020 precipitates organized on a simple cubic macro-lattice. In the present work, we use a stability analysis and
Accepted 10 January 2020
3D phase field simulations to show that such a regular array is in fact unstable whatever the volume fraction
Available online 22 January 2020
of precipitates. The two main instability modes are the longitudinal [100] mode and the transverse [110]
mode along the ½110 eigenvector. We argue that these instabilities lead to formation of configurational
Keywords:
defects closely related to experimentally observed branches and herringbone patterns. The ro ^ les of elastic
Single-crystal superalloy
Microstructure
anisotropy and elastic homogeneity are also discussed.
Elasticity © 2020 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Stability analyses
Phase field
Pattern formation

1. Introduction terminates. Interestingly, what might initially appear as a nearly peri-

odic arrangement of roughly aligned cuboids is thus highlighted as an
In multiphase alloys, the coherent coexistence of misfitting phases intricate network of interweaving defects in the precipitate alignments.
generates internal elastic fields. Given their long-range and usually Other kinds of patterns are observed at lower volume fractions (Fig. 1b).
anisotropic character, these fields alter the kinetics of diffusion-con- They may be identified as adjacent precipitate alignments along [100]
trolled phase transformations, and also control the shapes and and [010] directions, forming right-angled structures called chevrons or
mutual arrangement of the precipitates. During microstructure evo- herringbone patterns [1316]. All these alignments defects and pat-
lution, the formation of long-range spatially-correlated patterns is terns are observed both experimentally and numerically (see Ref.
therefore observed [1,2]. [1921] for forks and dislocation patterns at high volume fraction, and
The microstructure of single-crystal nickel-based superalloys, Ref. [8,22,23] for herringbone patterns at lower volume fractions).
obtained by precipitation of the L12-ordered g 0 phase in the FCC g matrix, Though it is well-established that the anisotropy of the elastic
is formed under the influence of elasticity. Besides modifying the inter- moduli is responsible for the formation of these specific arrange-
face thermochemical equilibrium [35], elasticity causes the shape of a ments, and despite a significant number of studies devoted to the
growing g 0 precipitate to gradually change from spheroid to cuboid [6]. g /g 0 microstructure, the origin and evolution of the pattern defects
In addition, interacting g 0 precipitates tend to align along the elastically remain to be clarified.
soft directions of the matrix during coarsening [79]. For high precipitate In this contribution, we address the question of the origin of the
volume fraction, this reconfiguration process leads to the formation of pattern defects in the g /g 0 microstructure, and show that their for-
peculiar arrangements of nearly periodically aligned cuboids [1012]. mation is driven by specific shape-dependent instabilities of the peri-
However, the resulting microstructures systematically display odic arrangement of interacting precipitates.
defects in the precipitate alignments, as highlighted in Fig. 1. At high This paper is organized as follows. The general formalism used to
volume fraction (Fig. 1a), two types of defects are observed on both hor- perform the static stability analysis of a periodic arrangement of mis-
izontal and vertical alignments: branches, when an alignment separates fitting inclusions is presented in Section 2. This formalism is then
into two modulations, and macro-dislocations, when a modulation applied in Section 3 to successively analyze the importance of the
precipitate shape, the volume fraction and the elastic anisotropy. A
close attention is paid to the instability modes of a cubic arrangement
* Corresponding author. of cubic precipitates. In Section 4, an elastic phase field model is used
E-mail addresses: [email protected], [email protected] (Y. Le Bouar).

https://doi.org/10.1016/j.actamat.2020.01.022
1359-6454/© 2020 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
42 M. Degeiter et al. / Acta Materialia 187 (2020) 4150

where n ¼ k=k is a unit vector of reciprocal space. The Fourier trans-

form of u(r) is defined by
Z
1
uðkÞ ¼ dV uðrÞeikr : ð3Þ
V V
The elastic kernel B(n) is:

BðnÞ ¼ e0ij λijkl e0kl ni s 0ij Vjk ðnÞs 0kl nl ð4Þ

where s 0ij ¼ λijkl e0kl is the transformation stress tensor, and Vjk(n) is
1
the normalized Green tensor defined by Vij ðnÞ ¼ λiklj nk nl .
Let a distribution of N identical precipitates be periodically
arranged along the x, y and z directions of a simple cubic macrolattice,
with a the macrolattice parameter. The aim of the present calculation
is to analyze the stability of this arrangement with respect to small
disturbances of the precipitate positions, that we will call displace-
ments (not to be confused with material displacements). The micro-
structure is completely described by the shape function:
X
uðrÞ ¼ u0 ðrRÞ ð5Þ
R

where u 0 is the shape function of a single precipitate sitting at the ori-

gin of the macrolattice and where the sum runs over all the precipi-
tate positions R. In reciprocal space, u becomes:
X
uðkÞ ¼ u0 ðkÞ eikR : ð6Þ
R

The strain energy of the microstructure becomes

VX X 0
Eel ¼ BðnÞju0 ðkÞj2 eikðRR Þ ð7Þ
2 k R;R 0

where we have set Bð0Þ ¼ 0.

The double sum on the right-hand side of Eq. (7) can be divided into a
sum over R ¼ R0 corresponding to the self-energy of the precipitates,
Fig. 1. Defects in the cuboidal microstructure: (a) Branch and macro-dislocation pat- and a sum over R ¼ 6 R0 which carries the dependence of the total strain
terns in the AM1 superalloy [17]; (b) Chevron patterns in a binary model nickel-based energy to the spatial arrangement of precipitates. As a conclusion, the sta-
superalloy [18].
bility of the arrangement is related to the configuration energy defined by
VX X 0
to analyze the dynamics resulting from of the two main instability Eelconf ¼ BðnÞju0 ðkÞj2 eikðRR Þ ð8Þ
2 k
modes predicted by the static stability analysis. The specific patterns R; R0
generated by these modes are presented and compared to the micro- 6 R0
R¼
structures observed in Ni-base superalloys. We finally investigate the
consequences of an inhomogeneity on the C0 shear modulus, which We now consider that the position R of each precipitate is close to
has been shown to significantly influence the microstructure evolu- some position R0 in a perfect cubic macrolattice and we define the
tion in these alloys [19]. displacements vectors uðR0 Þ ¼ RR0 .
Assuming small displacements, the configuration energy change
between the perfect and the distorted arrangements is, to the second
2. Stability analysis of periodic arrangements
order,
The subsequent analytical development is based on Khacha- VX
DEelconf ¼ BðnÞju0 ðkÞj2
turyan’s theory of elasticity [1], as well as on the work of Khacha- 2 k
turyan and Airapetyan [24]. Considering a coherent two-phase X  1    0
 ki kj ui ðR0 Þui ðR00 Þ uj ðR0 Þuj ðR00 Þ eikðR0R0 Þ : ð9Þ
mixture in stress-free boundary condition, and assuming small 2
0
strains and elastic homogeneity, the strain energy of the microstruc- R0 ; R 0
ture reads R0 6¼ R00
Z   
1 As usual with periodic structures, reciprocal lattice vectors are
Eel ¼ dVλijkl eij ðrÞ e0ij ðrÞ ekl ðrÞ e0kl ðrÞ ð1Þ
2 V decomposed as k ¼ H þ t where H is a vector of the reciprocal lattice
and t is a vector of the first Brillouin zone of the macrolattice.
where λijkl is elastic tensor, eij(r) is the total strain, and e0ij ðrÞ ¼ e0ij uðrÞ
Using this decomposition we have
is the stress-free strain tensor associated to the transformation eigen-
X X
strain e0ij . u (r) is the shape function equal to unity if r lies inside a pre- eikR0 ¼ eitR0 ¼ Ndt ¼0 : ð10Þ
cipitate, and zero otherwise. To simplify the notation, we assume R0 R0
that the volume V is orthorhombic with periodic boundary condi-
In addition, since u(R0) is defined at the lattice sites, we define its for-
tions.
ward and backward Fourier transforms as:
At mechanical equilibrium, the strain energy takes the form [1]:
1X
VX vi ðt Þ ¼ u ðR0 ÞeitR0 ð11Þ
Eel ¼ BðnÞjuðkÞj2 ð2Þ N R0 i
2 k6¼0
 M. Degeiter et al. / Acta Materialia 187 (2020) 4150 43

Fig. 2. (a) Eigenvalue spectrum of the curvature tensor for spherical precipitates with fv ¼ 37%. Longitudinal (L) and transverse (T) branches are indicated, and the dotted lines high-
light wave vectors with z ¼ 2p=ð12aÞ. (b) Sketches of the corresponding deformed arrangements, in which spherical precipitates are represented as blue spheres. (For interpretation
of the references to color in this figure legend, the reader is referred to the web version of this article.)

X
ui ðR0 Þ ¼ vi ðt ÞeitR0 : ð12Þ 3. Stability of a cubic array of precipitates
t

Introducing Eqs. (12) and (10) in Eq. (9) leads to the following In this section, stability analyses of the periodic distribution of
closed form for the configuration energy change: elastically interacting precipitates are carried out. The importance of
the precipitate shape, of the volume fraction and of the elastic anisot-
NX
DEelconf ¼ k ðt Þvi ðt Þvj ðt Þ ð13Þ ropy are successively analyzed.
2 t ij
3.1. Shape-dependent instabilities in the arrangement
where the curvature tensor for a unit cell of the periodic arrangement
is defined by
We first consider the case of spherical inclusions of radius R for
X H þ t 
kij ðt Þ ¼ VN B ðHi þ t i ÞðHj þ t j Þju0 ðH þ t Þj2 which the shape function in Fourier space is
H
Hþt " #
  ð14Þ 4 pR 3 kR cosðkRÞ sinðkRÞ
H u0 ðkÞ ¼ 3 : ð15Þ
B Hi Hj ju 0 ðHÞj2 : 3 V ðkRÞ3
H
It follows from (13) that the configuration energy change is a We choose the macrolattice parameter a ¼ 350 nm; and the volume
quadratic form of independent Fourier components of the displace- fraction fv ¼ 37%. The transformation eigenstrain is e0 ¼ 0:48% and
ments. Therefore, the configuration of identical precipitates on a the elastic constants are C11 ¼ 250 GPa; C12 ¼ 160 GPa and C44 ¼
macrolattice is stable with respect to the precipitate displacements 118:5 GPa. All the above parameters have been selected to allow a
if the tensor kij(t ) is positive for each Fourier mode t . As usual, the direct comparison with a previous work devoted to the stability of
positiveness of the tensor is analyzed by computing its eigenvalues. cubic arrangements of spherical precipitates [24]. The computation
If all eigenvalues are positive, the configuration is stable. If any of the curvature tensor at each Brillouin zone wave vector was car-
eigenvalue is negative, the configuration is unstable when the ried out with nRL ¼ 500; to ensure sufficient precision.
arrangement is perturbed by a displacement of mode t along the The three eigenvalues k0a of kij(t ) are plotted in Fig. 2a against the
corresponding eigenvector. Brillouin zone wavevector t of the cubic macrolattice, along the high-
For the cubic macrolattice considered here, the first Brillouin zone symmetry directions. The longitudinal k0L (i.e. eigenvector parallel to t )
is the set of vectors t such that p=a < t i  p=a. For the numerical and transverse transverse k0T1 and k0T2 (i.e. eigenvector normal to t )
computation of Eq. (14), the summation over reciprocal lattice sites H modes are indicated. Finally, Fig. 2b displays, for each branch, a sketch
has to be truncated. In our calculations, the summation is carried of the deformation of a cubic array of spheres for wave vectors with
over Hi ¼ 2pai =a; where ai is an integer such thatnRL  ai  nRL . z ¼ 2p=ð12aÞ. The branches of the curvature spectrum in Fig. 2a are
The number of replicated Brillouin zones nRL is chosen depending positive for all Brillouin zone wave vectors along the high-symmetry
on the precipitate shape, to ensure sufficient precision. Given the directions, implying that all the perturbations considered lead to an
Brillouin zone symmetry, the eigenvalues of kij(t ) are computed for increase in the configuration energy. We have also verified this property
the wave vectors t along the high symmetry directions. These direc- for all vectors of the first Brillouin zone (not shown). So the periodic dis-
tions are defined between the symmetry points G(0, 0, 0), X(p/a, 0, tribution of spherical inclusions is stable with respect to displacements
0), M(p/a, p/a, 0) and R(p/a, p/a, p/a). Since jt j ¼ 2p=L; where L is from their ideal positions. This conclusion is similar to that of Khacha-
the wavelength, G corresponds to an infinite-wavelength perturba- turyan and Airapetyan [24].1
tion (i.e. a translation), which induces no change in configuration
energy (kij ðGÞ ¼ 0). The perturbations at the other symmetry points 1
Note that the branches between G and X presented in Fig. 3a of Ref. [24] are ques-
have wave vector components equal to 0 or p/a and span over two tionable, especially for the behavior close to the limits of the first Brillouin zone and
precipitates. for the relative position of the longitudinal and transverse branches.
44 M. Degeiter et al. / Acta Materialia 187 (2020) 4150

We now consider the case of cubic precipitates. The shape func- also be at the origin of the formation of topological defects, in particular
tion in Fourier space becomes through the pinching of precipitate rows, in the same way as the pinch-
ing observed in RayleighBe
nard convection patterns [28].
c3 sinðkx c=2Þ sinðky c=2Þ sinðkz c=2Þ
u0 ðkÞ ¼ ; ð16Þ The second instability is the long wavelength [zz0] transverse
V kx c=2 ky c=2 kz c=2
branch k0T2 ; illustrated at the bottom of Fig. 3b. The direction of the
where c is the edge length of the cube. We have considered edge displacements of the precipitate positions, given by the eigenvector
lengths which are multiple of a=ð2nRL þ 1Þ and reciprocal lattice vec- of the curvature tensor, is ½110. This instability preserves the (001)
tors t with components which are multiple of 2p/(a np), where np is planes of the macrolattice. Inside these planes, the macrolattice is
an integer. This choice corresponds to a simulation box of length a np alternatively rotated clockwise and counter-clockwise. The rotation
discretized in np ð2nRL þ 1Þ grid points, with periodic conditions. is supplemented by an elongation of the macrolattice (at constant
Because the edge lengths are a multiple of the grid spacing, this volume) either in the [100] or in the [010] direction. The resulting
choice ensures a rapid convergence such that the number of repli- configuration at the bottom of Fig. 3b is similar to a herringbone pat-
cated Brillouin zones can be reduced to nRL ¼ 50. All other parame- tern and could initiate the formation of chevron patterns. This point
ters were kept the same as for the periodic distribution of spheres, will be further addressed in Section 4.
and the corresponding stability landscape is presented in Fig. 3a. This instability mode slightly disrupts the favorable alignments
As clearly visible on the spectrum in Fig. 3a, three branches of the between the first neighbor precipitates in the elastically soft direc-
stability landscape become negative when considering cubes rather tions. On the other hand, this mode has the advantage of displacing
than spheres. Along the [100] direction, the longitudinal branch L is the second neighbor precipitates, initially aligned along the elasti-
negative for all wavelengths. For the [zz0] wave vectors, the trans- cally unfavorable ⟨110⟩ directions, in a transverse direction, which
verse branch T2 is positive for short wavelengths (i.e. close to M), increases their distance to the unfavorable alignment.
changes its sign for z around 2p/(3.4a) and remains negative for The third instability is due to the transverse branches along the
greater wavelengths. Finally, along the [111] direction, the transverse [zzz] direction. Because this instability is significantly weaker than
branches are slightly negative for large wavelengths. These calcula- the two others, it will not be further studied in this work.
tions demonstrate that the periodic arrangement of cubic precipitates To sum up, the calculations with 37% of precipitates have
on a cubic macrolattice is unstable for the volume fraction fv ¼ 37%. highlighted two important points:
The instability, which displays the most negative value of k0a ; sits
along the longitudinal [z00] branch, illustrated at the top of Fig. 3b.  When considering cubic anisotropy and elastic soft directions
Two features can be brought about to explain why. First, it has the along the cubic directions, the perfect arrangement of cubic pre-
property to maintain the alignments of precipitates along the cubic cipitates is not necessarily stable.
directions, which are the elastic soft directions. This instability mode  The approximation of spherical precipitates, used in a previous
also has the characteristics of avoiding the alignments of precipitates work, is not accurate enough to analyze the stability of the pre-
along the ⟨110⟩ directions. These directions are elastically unfavorable, cipitate arrangement in Ni-base superalloys.
as can be deduced from the configurational force maps [7,25]. Note
that the instability of the g /g 0 microstructure with respect to this
mode has already been reported in two-dimensional simulations [26]. ^le of the volume fraction on the stability
3.2. Ro
In pattern formation, the long-wavelength longitudinal instability of
lamellar patterns is called Eckhaus instability [27]. In this context, this The effect of the volume fraction of g 0 precipitates on the stability
instability has been shown to induce the formation of topological defects landscape is now analyzed. For that purpose, curvature spectra were
such as branches and macro-dislocations [2832]. During the formation computed for increasing value of fv, ranging from 5% to 95% within
and growth of cuboidal microstructures, the longitudinal instability may steps of 5%. The three eigenvalues were organized according to the

Fig. 3. (a) Eigenvalue spectrum of the curvature tensor for cubic precipitates with fv ¼ 37%. The transverse branch labelled T2 between M and G corresponds to the eigenvector along
½110 and the dotted lines highlight wave vectors with z ¼ 2p=ð12aÞ. (b) Two dimensional sketches of the corresponding deformed arrangements in a (001) plane, for which the dis-
placement vectors lie in the (001) plane.
 M. Degeiter et al. / Acta Materialia 187 (2020) 4150 45

Fig. 5. (a) Polar plots of the normalized elastic kernels B(n). (b) Eigenvalue spectra of
Fig. 4. Stability landscapes of the (a) longitudinal L and (b) transverse T2 (bottom)
the curvature tensor for different values of the elastic anisotropy ξ . Precipitates have a
branches of the periodic arrangement of cubic precipitates, for volume fractions vary-
cubic shape and the volume fraction is fv ¼ 37%. The spectrum for ξ ¼ 0 is superposed
ing from 5% to 95%. The curvatures are scaled using a bi-symmetric log transformation
    with the axis of abscissas. Also, to help the reading, labels are added to the longitudinal
[33] sgn k0a ðt ; fv Þ log10 1 þ jk0a ðt ; fv Þ=cj ; with c ¼ 1=250; to capture the wide range of
and transverse modes for ξ ¼ 1:5.
variations.

favorable alignments along the cubic directions as well as the unfa-

longitudinal L and the transverses T1 and T2 branches. This classifica-
vorable alignments along the ⟨110⟩ and ⟨111⟩ directions.
tion, in principle not valid between X and M, can easily be extended
The stability spectra are presented in Fig. 5b. At ξ ¼ 0; the curva-
in this region using the continuity of the branches. Eigenvalues for
ture spectrum is flat and equal to 0, which means that elastic energy
the longitudinal L and transverse T2 modes corresponding to the
is insensitive to perturbations. This is consistent with the Bitter-
main instabilities are plotted with respect to fv in Fig. 4a and Fig. 4b,
Crum theorem [34,35]: the strain energy of an elastically homoge-
respectively.
neous and isotropic coherent mixture depends neither on the shapes
For volume fractions greater than 0.2, the stability landscape is
nor on the spatial distribution of the precipitates, but only on their
not very sensitive to the volume fraction: indeed, instability regions
volume fraction.
similar to that obtained for fv ¼ 0:37 (Fig. 3a) are observed. In particu-
For ξ < 0, the spectra all display the same three instability ranges
lar, the longitudinal [z00] mode is always unstable as indicated by the
observed in Fig. 3a, with the same hierarchy from the most destabi-
negative eigenvalues for all wavelengths between G and X whatever
lizing mode to the weakest: [z00] longitudinal, [zz0] transverse T2
the volume fraction (Fig. 4a). On the contrary, the wavelengths range
and [zzz] transverse. In Fig. 5b, it also appears that a decrease of ξ
of negative eigenvalues for the transverse [zz0] T2 mode (between M
leads to an extension of instability regions as well as a decrease of
and G) shrinks by about a factor of 2 when fv increases from 0.3 to
the related eigenvalue. More precisely, whatever the negative value
0.9. Thus, according to this stability analysis, chevron and herring-
of the anisotropy, the longitudinal mode along [z00] is unstable for
bone lengths are expected to increase with the volume fraction.
all wavelength, but a decrease of the anisotropy progressively
The existence of numerous instability modes at low volume fraction
reduces the instability range of the other mode to larger wavelengths.
is consistent with the observation, in low volume fraction g /g 0 alloys
Increasing the anisotropy |ξ | increases the absolute value of the nega-
[13,22], of isolated precipitate alignments that cannot be described as a
tive eigenvalues as well as the range of the unstable wavelengths.
perturbation of a perfect cubic macrolattice of precipitates.
Hence, the longitudinal [z00] modes are still all unstable.
We therefore come to the conclusion that, whereas the anisotropy
^le of elastic tensor anisotropy
3.3. Ro of the elastic tensor is at the origin of the alignments of precipitates
along the cubic directions, it is also responsible for the instability of
The ro^le of the elastic tensor anisotropy on the stability of the
the simple cubic macro-lattice.
periodic arrangement is now investigated. We have used the elastic
constants C11 ¼ 250 GPa; C12 ¼ 160 GPa; and we have computed C44
so that the anisotropy parameter ξ ¼ ðC11 C12 2C44 Þ=C44 takes the 4. Instability-induced microstructure evolutions
targeted values of: 0;0:5;1;1:5. The transformation eigenstrain
was kept at e0 ¼ 0:48%; and the volume fraction at fv ¼ 37%. 4.1. Phase field model
The elastic anisotropy appears clearly in the polar plots of the
elastic kernel B(n) shown in Fig. 5a. Minima are located along the The stability analysis of the previous section relies on the energy
⟨100⟩ directions and maxima along the ⟨111⟩ directions. ⟨110⟩ direc- variation when the reference configuration is disturbed by displacing
tions are also high energy directions. Because the elastic interaction precipitates. This calculation has revealed the instability of the refer-
energy of two coherent precipitates aligned along a direction n is, at ence microstructure. However, as it is purely energetic and does not
long distance, proportional to B(n), these polar plots illustrates the include any kinetic processes, this calculation gives no information on
46 M. Degeiter et al. / Acta Materialia 187 (2020) 4150

the growth rate of the destabilizing modes. In fact, this approach fea-
tures three limitations: (i) it gives no information on the growth rate
of the destabilizing modes; (ii) it is limited to infinitesimal precipitate
displacements; (iii) it relies on prescribed shapes for the precipitates,
while elastic interactions are known to induce morphological evolu-
tions [7]. To overcome these limitations, we have undertaken a kinetic
study of the microstructure evolution using a phase field model.
Phase field models have been extensively used to analyze the
microstructure evolution in nickel-base superalloys [22,23,3642].
We have used the classical phase field model detailed in Refs. [26,37].
In this model, the two phases alloy is described by a single concentra-
tion field and a homogeneous free energy density approximated by a
standard double-well potential. This model is not able to describe the
coalescence of ordered precipitates but is sufficient to analyze the
microstructure evolutions considered in the present study. Also we
mention that the model is constructed in such a way that, under
stress, interfaces perpendicular to the elastically soft directions adopt
a predefined profile (here ð1 þ tanhð2x=wÞÞ=2; where w is the inter- Fig. 6. Evolution of a microstructure initially perturbed along a k0L ðz; 0; 0Þ destabilizing
face width) with a predefined surface tension [37]. The parameters of mode with z ¼ 2p=8a. The volume fraction is fv ¼ 37%; and the snapshots correspond
to reduced times (a) t  ¼ 0; (b) t  ¼ 9000; and (c) t  ¼ 18000.
the phase field model are taken from Degeiter et al. [26] and corre-
spond to a simplified Ni-base superalloy at 950BC. In this section, the
elastic constants are C11 ¼ 197 GPa; C12 ¼ 144 GPa and C44 ¼ 90 GPa; (ii) The precipitates that have been made farther apart from each
and the precipitate volume fraction is fv ¼ 37% as in Section 3.1. The other along [100] become thinner along this direction and larger
eigenstrain is e0 ¼ 0:059 and the interface energy, assumed isotropic, along [010] and [001] to give platelet precipitates. With their
is s ¼ 10 mJ/m2. Times are given in reduced units using the charac- periodic replicas, they form rafts perpendicular to the [100] elas-
teristic time t0 ¼ 8d2 =D where d is the discretization step and D is the tic soft direction.
diffusion coefficient. The interface width is w ¼ 3:2d; corresponding (iii) The other precipitates adopt irregular shapes.
to about 7 points in the interface defined as the interval between 1%
and 99% of the concentration jump.
Note that the coexistence of all three kinds of shapes has been
4.2. Elastically homogeneous case reported in the 3D characterization of a Ni-base superalloy [43].
Just before the first coalescence event (t  ¼ 18000), the shapes of
Two kind of calculations have been performed to analyze the two precipitates that are about to merge exhibit concave sides, as a result
main instabilities identified previously. In both cases, the initial con- of the configurational forces that are attractive near the corners and
figuration is set as an arrangement of cubic precipitates, whose posi- repulsive near the center of the faces [44].
tions have been shifted with respect to the periodic reference It is worth stressing that the simulation in Fig. 6 is a simplified
according to a destabilizing mode. The amplitude and wavelength of case with a 1D [100] longitudinal instability with periodic images
the modes were chosen so that the displacements are multiple of the along the [010] and [001] directions. As mentioned earlier, this kind
grid spacing d. To smooth the initial interfaces, the concentration of instability could be associated with the formation of branches and
field is convolved with a gaussian function. macro-dislocations because it is similar to the pinching mechanism
We first consider a microstructure with 8 precipitates inside a observed in the Eckhaus instability in Rayleigh-Be
nard convection
3.7 £ 0.5 £ 0.5 mm3 box discretized with d ¼ 3:6 nm, and we impose cells pattern [28]. Although the full description of such a process
periodic boundary conditions along the three cubic directions. This would require much larger simulation boxes, we believe that the
microstructure is initially perturbed with a [z00] longitudinal mode. existence of the [100] longitudinal instability could be responsible for
If a denotes the initial periodicity of the macrolattice, the wavelength the formation of the defects highlighted in Fig. 1a.
of the initial mode is L ¼ 8a; which corresponds to z ¼ 2p=8a; and The second mode investigated is the T2 transverse along the [zz0]
the initial amplitude is 28.8 nm (8 grid spacings). The evolution is direction, imposed on a reference arrangement of 4 £ 4 £ 1 precipi-
presented in Fig. 6 at three different times, the last one preceding tates, periodically distributed in a simulation box of volume
coagulation of neighboring precipitates. 1.8 £ 1.8 £ 0.5 mm3 discretized with d ¼ 3:6 nm. The initial perturba-
At t  ¼ 0 in Fig. 6a, the precipitates are cubes with diffuse interfa- tion mode spans over 4 precipitates along [100] and [010]
ces, and the imposed slight variation in their longitudinal spacings is (z ¼ 2p=4a) with an amplitude of 36 nm (10 grid spacings).
difficult to see: the spacings between the precipitates on the left- and The evolution of this perturbed configuration during coarsening is
right-hand sides of the simulation box are smaller and larger than in presented in Fig. 7.
reference arrangement, respectively. As expected from the linear stability analysis, the precipitate cen-
As shown in Fig. 6b and c, and in agreement with the stability ters do not go back to the sites of the reference cubic macrolattice. In
analysis, the dynamics does not bring the microstructure back to the fact, the phase field calculation reveals that the instability induces a
reference alignment of equidistant cubic precipitates. More precisely, significant shape evolution of the precipitates while their center of
the center of mass of the precipitates does not significantly move and mass remains at their initial perturbed position during the simula-
rather we observe a morphological evolution of the precipitates: tion. Due to the small wavelength considered in this simulation, only
two types of morphologies are observed: plate-like and irregular. The
formation of platelet-like precipitates creates large matrix channels
(i) The precipitates that have been made closer to each other by the alternatively in the vertical and horizontal directions (see 2D sections
perturbation along [100] elongate progressively in this direction in Fig. 7b and c). During the microstructure evolution, we also
to form rectangular cuboids. This evolution is similar to the one observe that the precipitates that are close to each other align along
observed in 2D simulations by Su and Voorhees [7] during the cubic directions, which are the elastic soft directions (e.g. [7]). At the
interaction of two precipitates. end of the simulation, a herringbone pattern is formed, which is
 M. Degeiter et al. / Acta Materialia 187 (2020) 4150 47

Fig. 7. Evolution of a periodic microstructure initially perturbed along a k0T2 ðz; z; 0Þ Fig. 8. Eigenvalue spectra of the curvature tensor for periodic arrangements of cubic
destabilizing mode with z ¼ 2p=4a. The volume fraction is fv ¼ 37%; and the phase precipitates at (a) fv ¼ 37% and (b) fv ¼ 60%. The continuous lines are obtained from Eq.
field simulation results correspond to the reduced times (a) t  ¼ 0; (b) t  ¼ 12000; and (14). Blue circles and red squares correspond to curvatures computed using the FFT
(c) t  ¼ 24000. solver with DC 0 ¼ 0% and DC 0 ¼ 50%; respectively. (For interpretation of the references
to color in this figure legend, the reader is referred to the web version of this article.)

similar to chevron patterns observed in Fig. 1b. Note that chevron

patterns have also been reported in two-dimensional phase field sim- instabilities identified in the elastically homogeneous case. Note
ulations [22,23]. finally, that the FFT solver that we have used is best suited to diffuse
As a conclusion, beyond the stability analysis, the phase field cal- interfaces. Therefore, as for the initial configuration of the phase field
culations have shown that the development of the instabilities does calculations (see Sec. 4), the interfaces have been smoothed by apply-
not proceed by the movement of the precipitates but by a change of ing a small gaussian filter. Considering sharp interfaces would require
their morphologies. The instabilities lead to the coexistence of precip- more sophisticated solvers [45,46] that are beyond the scope of the
itates with different shapes and to the formation of specific micro- present work.
structures that can be observed in electron micrographs. The spectra are presented in Fig. 8a and b for the two volume frac-
tions fv ¼ 37% and fv ¼ 60% and compared to the homogeneous spec-
trum obtained with Eq. (14).
4.3. Inhomogeneity on C0 elastic constant
First, we have checked that smoothing the interfaces does not
change drastically the spectra for the homogeneous elastic constants.
4.3.1. Linear stability analysis This is the case as demonstrated by the blue dots obtained with dif-
The influence of an elastic inhomogeneity DC 0 ¼ ðCg0 0 Cg0 Þ=Cg0 is fuse interfaces very close to the black lines. The differences are only
now investigated by comparing calculations using two sets of elastic visible near the edges of the Brillouin zone.
constants: homogeneous constants used in Section 4 and inhomoge- Then, the influence of the elastic inhomogeneity DC0 can be
g0 g0 g0
neous constants considering C11 ¼ 193; C12 ¼ 113 and C44 ¼ 132 deduced by comparing the spectra for DC 0 ¼ 0 (blue circles) and for D
GPa. These values, proposed in [19], correspond to DC 0 ¼ 50%. C 0 ¼ 50% (red squares). For the [z00] modes, our calculations show
When considering an elastic inhomogeneity, the curvature tensor that increasing DC0 induces a decrease of all the branches. In particu-
cannot be computed with Eq. (14). We have therefore used the FFT lar, the longitudinal mode becomes more unstable. For the [z z 0]
mechanical solver of the phase field code to compute the energy of modes, the consequences of the elastic inhomogeneity DC0 are more
configurations perturbed by sinusoidal waves of increasing ampli- complex and depend both on the volume fraction and on the consid-
tudes. The energy increases with respect to the amplitudes are fitted ered branch. When focusing on the destabilizing transverse T2 mode,
with quadratic functions so as to retrieve the components of the cur- we observe that an increase of DC0 leads to a very slight increase of
vature tensor. In the following, we have only analyzed the two main the destabilizing wavelength range for fv ¼ 37% whereas a decrease
48 M. Degeiter et al. / Acta Materialia 187 (2020) 4150

is observed for fv ¼ 60%. This suggests that increasing DC0 would simulation, no significant evolution of the precipitate positions or
delay the formation of chevron and herringbone patterns in high vol- shapes can be measured in the inhomogeneous simulation. For much
ume fraction alloys. longer times, the precipitates move and their shape evolves, but the
amplitude of the initial mode neither grows nor decays during the
4.3.2. Microstructure evolution simulation, and appears to be kinetically frozen. The movement of
In this subsection, we carry out phase-field calculations to further the precipitates results from the development of other longitudinal
investigate the two main instabilities of the periodic arrangement of modes with shorter wavelength, as it is clearly visible on the phase
cubic precipitates with an elastic inhomogeneity on the C0 shear modu- profiles presented in the bottom of Fig. 9b. Note that a similar behav-
lus. This allows us to go beyond the static stability analysis performed in ior was reported in 2D simulations [26].
the previous section and, in particular, such calculations have the advan- The behavior of the elastically inhomogeneous system can be
tage to handle shape evolutions, a key ingredient as already shown in rationalized using the following results from the literature. First,
(Section 4). In this section, we use a high volume fraction (fv ¼ 60 %) cor- when the precipitates are harder than the matrix, as it is the case
responding to usual monocrystalline nickel-based superalloys. here for the C0 component, the elastic inhomogeneity has been shown
First, we handle the longitudinal mode along the [100] direction, to favor compact precipitate shapes [4750]. Hence, shape changes
with a wavelength L ¼ 16a and an amplitude of 36 nm (Fig. 9). This observed in the homogeneous simulations are less favorable in the
mode is applied on a microstructure consisting of 16 £ 1 £ 1 precipi- inhomogeneous case. Second, elastic inhomogeneity has been
tates periodically arranged in a simulation box of volume reported to slow down the evolution kinetics. This can explain why
7.4 £ 0.5 £ 0.5 mm3 discretized with d ¼ 3:6 nm. the evolution is slower when DC 0 ¼ 50% (Fig. 9b) than when DC 0 ¼ 0
The movement of the precipitates is analyzed using a new method (Fig. 9a).
proposed in Ref. [26]. It relies on the phase cx(r) (not to be confused We now analyze the evolution of a configuration initially perturbed
with phase field f(r)) which is a continuous equivalent of the discrete with the transverse T2 mode along the [zz0] direction. We consider a
phase shift cx ðR0 Þ ¼ kx ux ðR0 Þ induced by the movement of one pre- periodic microstructure composed of 4 £ 4 £ 1 precipitates in a
cipitate along [100]. This quantity is plotted at different time steps 1.8 £ 1.8 £ 0.5 mm3 box discretized with d ¼ 3:6 nm. The initial ampli-
(Fig. 9 bottom) to follow the precipitates movement along [100]. tude is 36 nm and the wave vector component z ¼ 2p=4a is chosen
To highlight the consequences of elastic inhomogeneity, the close to the limit of the instability region predicted by the linear analy-
microstructure evolution with DC 0 ¼ 50% (Fig. 9b) is compared to the sis (Fig. 8). More precisely, the wavelength is 6% and 16% lower than
microstructure obtained with homogeneous elasticity (Fig. 9a). the stability limit in the homogeneous and inhomogeneous cases,
In the homogeneous case (Fig. 9a), the evolution is similar to the respectively. Therefore, within the framework of the linear stability
one observed in Fig. 6 for a lower volume fraction: the precipitate analysis considering constant precipitate shapes, the configuration is
positions remain the same while their shapes are evolving up to a stable in both the homogeneous and inhomogenous cases.
point where a coagulation event occurs (t*  1.5 104). In the inhomo- The final configuration of the phase field simulations, presented in
geneous case (Fig. 9b), the microstructure evolution is much slower. Fig. 10a, demonstrates that, in the simulation with homogeneous
At the time when a coagulation event occurs in the homogeneous elasticity, the initial perturbation induces precipitate shape changes

Fig. 9. Phase field results for the longitudinal perturbation along [100] with (a) homogeneous elasticity and (b) inhomogeneous elasticity. Top: final microstructures at
(a) tf ¼ 15000; and at (b) tf ¼ 350000. Middle: Cross sections of the microstructure evolution. Bottom: horizontal profiles of the topological phase cx(r) at t  ¼ 0 (red) and at differ-
ent time steps (black). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
 M. Degeiter et al. / Acta Materialia 187 (2020) 4150 49

isothermal annealing, even if the microstructure approximates well

on average a cubic arrangement of precipitates, it must necessarily
contain defects. To characterize microstructure evolution in detail, it
is therefore important to study the dynamics of these defects.
Preliminary observations on 2D phase field calculations of the
g /g 0 microstructure in Ref. [19] have revealed that, during an isother-
mal heat treatment, branches and macro-dislocations migrate in the
microstructure according to mechanisms close to dislocation climb-
ing (see also Ref. [22,51]) and gliding mechanisms. In a future work,
we intend to extend this analysis by examining the dynamics of these
arrangement defects using the recently proposed S-PFM methods,
which allows the simulation of large-scale microstructures [52].

Declaration od competing interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.

Acknowledgements

The authors thank Didier Locq and Pierre Caron (ONERA) for fruitful
discussions and for providing electron micrographs of single-crystal
superalloys.

References

[1] A.G. Khachaturyan, Theory of Structural Transformations in Solids, Dover Publica-

Fig. 10. Snapshots (left: 3D image, right: 2D cross section) of cubic macro-lattices per- tions: New York, 1983. Chapter 7-11.
turbed by a transverse T2 mode along [zz0] (z ¼ 2p=4a) with (a) homogeneous elastic- [2] Y. Le Bouar, A. Loiseau, A.G. Khachaturyan, Origin of chessboard-like structures in
ity and (b) inhomogeneous elasticity. decomposing alloys. theoretical model and computer simulation, Acta. Mater. 8
(1998) 2777–2788.
[3] F. Larche
, J.W. Cahn, Thermochemical equilibrium of multiphase solids under
and leads to the formation of a herringbone pattern, as in Fig. 7. It stress, Acta Metall. 26 (1978) 1579–1589.
means that the energy gain associated with the morphological evolu- [4] M.E. Gurtin, P.W. Voorhees, The continuum mechanics of coherent two-phase
tion exceeds the slight energy increase that is predicted by the linear elastic solids with mass transport, Proc. R. Soc. Lond. A 440 (1993) 323–343.
[5] D. Schneider, O. Tschukin, A. Choudhury, M. Selzer, T. Bohlke, B. Nestler, Phase
analysis performed at constant precipitate shape. field elasticity model based on mechanical jump conditions, Comput. Mech. 55
The microstructure evolution is very different with inhomoge- (2015) 887–901.
neous elasticity (Fig. 10b). Indeed, the initial perturbation progres- [6] M.E. Thompson, C.S. Su, P.W. Voorhees, The equilibrium shape of a misfitting pre-
cipitate, Acta Metall. Mater. 42 (1994) 2107–2122.
sively vanishes and a perfect cubic macro-lattice is recovered. Two [7] C.H. Su, P.W. Voorhees, The dynamics of precipitate evolution in elastically
reasons can be given to explain the difference in behavior between stressed solids-II. particle alignment, Acta Mater. 44 (1996) 2001–2016.
homogeneous and inhomogeneous calculations. First the relevant [8] K. Thornton, N. Akaiwa, P.W. Voorhees, Large-scale simulations of Ostwald ripen-
ing in elastically stressed solids: I. development of microstructure, Acta Mater 52
eigenvalue of the curvature tensor is slightly higher for the inhomo-
(2004) 1353–1364.
geneous case (Fig. 8b). Second, as mentioned above, when the precip- [9] K. Thornton, N. Akaiwa, P.W. Voorhees, Large-scale simulations of Ostwald ripen-
itates are harder than the matrix, the elastic inhomogeneity favors ing in elastically stressed solids: II. Coarsening kinetics and particle size distribu-
compact precipitate shapes. tion, Acta Mater. 52 (2004) 1365–1378.
[10] D.J. Chellman, A.J. Ardell, The coarsening of g 0 precipitates at large volume frac-
As a conclusion, we have shown with two examples that the inho- tion, Acta. Metall. 22 (1974) 577–588.
mogeneity of the shear modulus C0 may qualitatively change the micro- [11] P. Caron, T. Khan, Improvement of creep strength in a nickel-base single crystal
structure evolution resulting from some instability. Therefore defects superalloy by heat treatment, Mater. Sci. Eng. 61 (1983) 173–184.
[12] P. Caron, High g 0 solvus new generation nickel-based superalloys for single-crys-
related to these instabilities (chevron patterns, macro-dislocations,...) tal turbine blade applications, in: T.M. Pollock (Ed.), Superalloys, 2000, 2000,
are expected to be sensitive to this inhomogeneity coefficient. pp. 737–746.
[13] A.J. Ardell, R.B. Nicholson, On the modulated structure of aged ni-al alloys, Acta
Metall. 14 (1966) 1295–1309.
[14] W.T. Loomis, J.W. Freeman, D.L. Sponseller, The influence of molybdenum on the g 0
5. Conclusions phase in experimental nickel-base superalloys, Metall. Trans. 3 (1972) 989–1000.
[15] V. Biss, D.L. Sponseller, The effet of molybdenum on g 0 coarsening and on ele-
In this contribution, the origin of the defects in the precipitate vated-temperature hardness in some experimental nickel-base superalloys, Met-
all. Trans. 4 (1973) 1953–1960.
alignments of the g /g 0 microstructure in nickel-base superalloys has [16] T. Grosdidier, A. Hazotte, A. Simon, Precipitation and dissolution processes in g /g 0
been addressed, by means of semi-analytical stability analysis and single-crystal nickel-based superalloys, Mater. Sci. Eng. A 256 (1998) 183–196.
phase field calculations. [17] F. Diologent, Comportement en fluage et en traction de superalliages monocristal-
lins a base de nickel, Universite
d’Orsay - Paris XI, p. 47, 2002 Phd thesis.
In the context of microstructure formation in nickel-based super- [18] X. Li, X.N. Zhang, C.P. Liu, C.Y. Wang, T. Yu, Z. Zhang, Regular g /g 0 phase interface
alloys, our calculations and phase field simulations have revealed instability in a binary model nickel-based single-crystal alloy, J. Alloys Compd.
that the perfectly periodic arrangement of cuboidal precipitates is 633 (2015) 366–369.
[19] M. Cottura, Y. Le Bouar, B. Appolaire, A. Finel, Role of elastic inhomogeneity in the
unstable whatever the volume fraction. Specifically, the stability
development of cuboidal microstructures in ni-based superalloys, Acta Mater. 94
analysis together with the phase field simulations have shown that (2015) 15–25.
configurational instabilities lead to formation of defects closely [20] S.Y. Hu, L.Q. Chen, A phase-field method for evolving microstructures with strong
related to experimentally observed branches and herringbone pat- elastic inhomogeneity, Acta Mater. 49 (2001) 1879–1890.
[21] J. Zhu, L.-Q. Chen, J. Shen, Morphological evolution during phase separation and
terns. Also, our phase field simulations have shown that precipitate coarsening with strong inhomogeneous elasticity, Model. Simul. Mater. Sci. Eng. 9
shape evolutions are correlated to these defects. Thus, during an (2001) 499–511.
50 M. Degeiter et al. / Acta Materialia 187 (2020) 4150

[22] V. Vaithyanathan, L.Q. Chen, Coarsening of ordered intermetallic precipitates with [38] A. Gaubert, Y. Le Bouar, A. Finel, Coupling phase field and viscoplasticity to study
coherency stress, Acta Mater. 50 (2002) 4061–4073. rafting in ni-based superalloys, Philos. Mag. B 90 (2010) 375–404.
[23] Y. Wang, D. Banerjee, C.C. Su, A.G. Khachaturyan, Field kinetic model and com- [39] N. Zhou, C. Shen, M. Mills, Y. Wang, Large-scale three-dimensional phase field
puter simulation of precipitation of L12 ordered intermetallics from F.C.C. solid simulation of g 0 rafting and creep deformation, Philos. Mag. B 90 (2010) 405–436.
solution, Acta Mater. 46 (1998) 2983–3001. [40] L.T. Mushongera, M. Fleck, J. Kundin, Y. Wang, H. Emmerich, Effect of re on direc-
[24] A.G. Khachaturyan, V.M. Airapetyan, Spatially periodic distributions of new phase tional g 0 -coarsening in commercial single crystal ni-base superalloys: a phase
inclusions caused by elastic distorsions, Phys. Status Solidi 26 (1974) 61–70. field study, Acta Mater. 93 (2015) 60–72.
[25] M. Degeiter, Etude nume
rique de la dynamique des de
fauts d’alignement des [41] M. Cottura, B. Appolaire, A. Finel, Y. Le Bouar, Coupling the phase field method for
pre
cipite
s g 0 dans les superalliages monocristallins a  base de nickel, Universite
de diffusive transformations with dislocation density-based crystal plasticity: appli-
Lorraine, France, 2019 Phd thesis. cation to ni-based superalloys, J. Mech. Phys. Solids 94 (2016) 473–489.
[26] M. Degeiter, M. Perrut, B. Appolaire, Y. Le Bouar, A. Finel, A new analysis of the [42] J.V. Goerler, I. Lopez-Galilea, L.M. Roncery, O. Shchyglo, W. Theisen, I. Steinbach,
microstructure of ni-based single-crystal superalloys : relevant topological Topological phase inversion after long-term thermal exposure of nickel-base
parameters for efficient microstructural modeling, in: M. Hardy, E. Huron, superalloys: Experiment and phase-field simulation, Acta Mater. 124 (2017)
U. Glatzel, B. Griffin, B. Lewis, C. Rae, V. Seetharaman, S. Tin (Eds.), Superalloys 151–158.
2016: Proceedings of the Thirteenth International Symposium on Superalloys, [43] A.C. Lund, P.W. Voorhees, A quantitative assessment of the three- dimensional
Wiley Online Library, 2016, pp. 323–331. microstructure of a g  g 0 alloy, Philos. Mag. 83 (14) (2003) 1719–1733.
[27] M.C. Cross, P.C. Hohenberg, Pattern formation outside of equilibrium, Rev. Mod. [44] A.C. Lund, P.W. Voorhees, The effects of elastic stress on coarsening in the ni-al
Phys. 65 (3) (1993) 851–1112. system, Acta Mater. 50 (2002) 2585–2598.
[28] H. Sakaguchi, Defect creation by the Eckhaus instability, Prog. Theor. Phys. 855 (5) [45] A. Ruffini, Y. Le Bouar, A. Finel, Three-dimensional phase-field model of disloca-
(1991) 927–932. tions for a heterogeneous face-centered cubic crystals, J. Mech. Phys. Solids 105
[29] S. Rasenat, E. Braun, V. Steinberg, Eckhaus instability and defect nucleation in (2017) 95–115.
two-dimensional anisotropic systems, Phys. Rev. A 43 (10) (1991) 5728–5731. [46] P.L. Valdenaire, Crystal plasticity: Transport equations and dislocation density,
[30] A.C. Newell, T. Passot, Instabilities of dislocations in fluid patterns, Phys. Rev. Lett. Mines ParisTech, France, 2016. Phd thesis.
68 (12) (1992) 1846–1849. [47] D.M. Barnett, J.K. Lee, H.I. Aaronson, K.C. Russell, The strain energy of a coherent
[31] Y. Hu, R. Ecke, G. Ahlers, Convection under rotation for Prandtl numbers near 1: spherical precipitate, Scr. Metall. 8 (1974) 1447–1450.
linear stability, wave-number selection, and pattern dynamics, Phys. Rev. E 55 (6) [48] J.K. Lee, A study on coherency strain and particle morphology via discrete atom
(1997) 6928–6949. method, Metall. Mater. Trans. A 27 (1994) 1449.
~ a, C. Pe
[32] B. Pen
rez-García, A. Sanz-Anchelergues, D.G. Míguez, A.P. Mun ~ uzuri, Transverse [49] M. Kato, T. Fujii, S. Onaka, Elastic strain energy of sphere, plate and needle inclu-
instabilities in chemical turing patterns of stripes, Phys. Rev. E 68 (2003) 056206. sions, Mater. Sci. Eng. A 211 (1996) 95–103.
[33] J. Beau, W. Webber, A bi-symmetric log transformation for wide-range data, [50] X. Li, K. Thornton, Q. Nie, P.W. Voorhees, J.S. Lowengrub, Two- and three-dimen-
Meas. Sci. Technol. 24 (2013) 027001. sional equilibrium morphology of a misfitting particle and the gibbsthomson
[34] F. Bitter, On impurities in metals, Phys. Rev. 37 (1931) 1527–1547. effect, Acta Mater. 52 (2004) 5829–5843.
[35] J.W. Cahn, F. Larche
, A simple model for coherent equilibrium, Acta Metall. 32 [51] Y. Le Bouar, A. Loiseau, A.G. Khachaturyan, Coarsening of the chessboard-like
(1984) 1915–1923. structures: a TEM study and computer simulations, in: P.E.A. Turchi,
[36] J. Boisse, N. Lecocq, R. Patte, H. Zapolsky, Phase-field simulation of coarsening of g A. Gonis (Eds.), Phase Transformations and Evolution in Materials, TMS (The Min-
precipitates in an ordered g 0 matrix, Acta Mater. 55 (2007) 6151–6158. erals, Metals & Materials Society), 2000, pp. 55–71.
[37] G. Boussinot, A. Finel, Y. Le Bouar, Phase-field modeling of bimodal microstruc- [52] A. Finel, Y. Le Bouar, A. Dabas, B. Appolaire, Y. Yamada, T. Mohri, A sharp-interface
tures in nickel-based superalloys, Acta Mater. 57 (2009) 921–931. phase field method, Phys. Rev. Lett. 121 (2018) 025501.