Journal of Crystal Growth 457 (2017) 52–59

Contents lists available at ScienceDirect

Journal of Crystal Growth

journal homepage: www.elsevier.com/locate/jcrysgro

Surface rippling during solidification of binary polycrystalline alloy:

Insights from 3-D phase-field simulations
Kumar Ankit a,n, Hui Xing b, Michael Selzer a,c, Britta Nestler a,c, Martin E. Glicksman d
a
Institute of Applied Materials, Karlsruhe Institute of Technology, Haid-und-Neu Str. 7, 76131 Karlsruhe, Germany
b
Key Laboratory of Space Applied Physics and Chemistry, Shaanxi Key Laboratory for Condensed Matter Structure and Properties, Northwestern
Polytechnical University, Xi'an, China
c
Institute of Materials and Processes, Karlsruhe University of Applied Science, Moltkestr. 30, 76133 Karlsruhe, Germany
d
College of Engineering, Florida Institute of Technology, 150 West University Blvd., Melbourne, FL 32901-6975, United States

art ic l e i nf o a b s t r a c t

Communicated by Falini Giuseppe The mechanisms by which crystalline imperfections initiate breakdown of a planar front during direc-
Available online 20 May 2016 tional solidification remain a topic of longstanding interest. Previous experimental findings show that the
Keywords: solid-liquid interface adjacent to a grain boundary provides a potential site where morphological in-
Phase-field modeling stabilities initiate. However, interpretation of experimental data is difficult for complex 3-D diffusion
Solidification fields that develop around grain multi-junctions and boundary ridges. We apply a phase-field approach
Morphological instabilities to investigate factors that induce interfacial instabilities during directional solidification of a binary
Growth models polycrystalline alloy. Using 2-D simulations, we establish the influence of solid-liquid interfacial energies
Diffusion-limited patterns on the spatial localization of initial interfacial perturbations. Based on parametric studies, we predict that
grain misorientation and supersaturation in the melt provide major crystal growth factors determining
solute segregation responsible for surface rippling. Subsequent breakdown of boundary ridges into
periodic rows of hills, as simulated in 3-D, conform well with experiments. Finally, the significance of
crystal misorientation relationships is elucidated in inducing spatial alignment of surface ripples.
& 2016 Elsevier B.V. All rights reserved.

1. Introduction respectively. Several works discussing theoretical aspects of

The mechanism by which interfacial free energy is minimized
by grain boundary grooving at triple junctions is well known. By
definition, thermal grooves pertain to solid-vapor, whereas liquid
grooves correspond to solid-melt interfaces. In the present article,
we focus on the influence of liquid grooves in instigating surface
instability during solidification.
The dihedral angle, θd, at the triple point of a grain boundary
groove develops at equilibrium as a result of solute diffusion. The
equilibrated groove exhibits a pair of convex ridges that are
symmetrically disposed about the line of intersection of the solid-
liquid interface and the grain boundary, provided that the inter-
facial energy of the adjacent grains does not depend on the local
grain orientation to the melt. When constitutional supercooling
overcomes the interface-stabilizing thermal gradient, instability
arises and protruding ridges progressively amplify from the
groove. Mullins [1,2] established the kinetic law for grain bound-
1 1
ary grooving with corresponding power-law exponents of 4 and 3
for surface and volume diffusion-controlled kinetic regimes,
n
Corresponding author.
E-mail address: kumar.ankit@kit.edu (K. Ankit).
boundary grooving have been reported since the 1960s [3–6].
Bolling and Tiller [3] described stationary groove profiles of pure
materials analytically with solid-liquid interfacial energy, thermal
gradient, and dihedral angle. It is reported that the preferred
growth directions depend upon the anisotropy in surface energy.
Coriell and Sekerka [4,5] elucidated grain boundary-mediated
morphological instabilities by applying linear stability theory for
unidirectional solidification of pure materials and binary alloys.
The spreading of instabilities from initial perturbations can be
described by a Green's function [7]. Bokstein et al. [6] propose a
model of liquid grooving to investigate the kinetic (dissolution)
and diffusion regimes at grain boundaries with zero (or nearly
zero) dihedral angles.
Grain boundary grooving has been observed and reported in
numerous experiments on alloy solidification. By experimentally
analyzing the isothermal grain boundary grooving in the Sn–Pb
alloy, Hardy et al. [8] establish synergies with Mullins' power-law.
Independent experimental investigations on solidification of dilute
Pb–Sb [9] and Al–Cu [10] alloys also reveal that grain boundary
grooves provide potent sites for initial perturbations; all these
findings agree with theoretical predictions. Further, it is suggested
that crystal orientation has a strong influence on morphological
amplification of ridges and patterns evolving at later stages [9].

http://dx.doi.org/10.1016/j.jcrysgro.2016.05.033
0022-0248/& 2016 Elsevier B.V. All rights reserved.
 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59
In order to expose and analyze the mechanisms by which in-
terfacial morphological instability initiates adjacent to grain
boundary grooves, optically transparent analogues, such as succi-
nonitrile-based alloys (SCN) and CBr2–C2Cl6, were frequently em-
ployed for such directional solidification experiments. The pio-
neering study on morphological instabilities adjacent to crystalline
imperfections by Schaefer and Glicksman [11] shows that grain
boundaries play a key role by initiating morphological instability
of primary ridges. As the initial ridges amplify, parallel secondary
ridges appear beside them. The primary ridges progressively break
down into periodic rows of hillocks and knobs, some of which
eventually evolve to ramified dendritic patterns. A detailed ela-
boration of these concomitant events is provided in Section 5.
Noël et al. [12,13] investigated the influence of grain bound-
aries and natural convection on microstructure formation in cel-
lular directional solidification of dilute SCN alloys. Their findings
confirmed that the dynamic grooving of grains occurs by the
propagation of ridges and depressions along grain boundaries.
Experimental results by Xing et al. [35] accentuate the potency of
grain boundaries in initiating morphological instabilities. The in-
itial perturbations observed adjacent to the grain boundary are
reported to be larger than others. Further, it is suggested that
preferred grain orientations dominate during growth competition
as the complex patterns evolve. Fig. 1(a–h) shows an exemplary
evolution of the solid-liquid interface near the grain boundary of
succinonitrile-0.9 wt% p-dichlorobenzene (SCN-DCB) alloy for an
applied thermal gradient of 6.2 K/mm and a pulling velocity of
2.80 m/s.
The phase-field simulation method, long established in the
material science community, provides a superlative numerical
approach to describe microstructural evolution during solidifica-
tion, solid-state and electrochemical transformations [14–18].
Two-dimensional phase-field simulations of morphological in-
stability adjacent to liquid grooves were reported earlier by Yeh
et al. [19] and by Miller et al. [20]. Both of these works capture the
critical onset of morphological instability near the intersecting
grain boundary that typically is characterized by “hump amplifi-
cation.” However, the influence of lateral flux-assisted solute seg-
regation results in a progressive breakdown of the solid-liquid
interface—a process termed “rippling.” For further aspects of this
process, which, heretofore, have not been addressed in detail, see
Fig. 19 in Morris and Winegard [9]. Indeed, further investigation is
warranted to achieve a more precise understanding of defect-in-
duced interfacial breakdown during directional solidification.
The remainder of this study is organized as follows: we first
Fig. 1. Temporal evolution of solid-liquid interface near the grain boundary of SCN-0.9 wt.%DCB (succinonitrile-p-dichlorobenzene) alloy. The time elapsed during evolution
(a–h) is 7 s.
53
describe the methodology, by using constructed free energy den-
sities, to approximate the variation of the grand-potentials of the
solid and liquid phases as functions of their chemical potentials. In
Section 3, a validation is presented of phase-field grain boundary
groove simulations checked against classical analytic solutions.
The importance of surface energy and grain misorientation in
initiating morphological instability is elaborated using 2-D simu-
lations in Section 4. Section 5 is devoted to 3-D analysis of the
mechanisms by which primary ridges amplify, along with a dis-
cussion of the significance of crystal misorientation relationships
inducing spatial alignment of surface ripples. Section 6 contains
conclusions and the outlook from this study.
2. Thermodynamics
The grand-chemical potential phase-field model used for our
numerical investigation is outlined in Appendix A. The free energy
of a phase α, in a 2-phase model binary alloy (A-B), are described
by the following form,
( ) ( ) ( ) ( ) ( )
fα cA, cB, T = Aα T c A2 + B α T c B2 + D α T cA + E α T cB + K α T , ( ) (1)
where cA and cB are the concentrations of components A and B,
respectively, and Aα , B α , D α , Eα and Kα are the temperature-de-
pendent coefficients for phase α. On imposing the binary alloy
constraint cA + cB = 1, one may rewrite the Equation (1) as follows:
( ) ( ) ( )
fα cA, T = O α T c 2A + P α T cA + Q α T , ( ) (2)
where O α = Aα + B α , P α = D α − 2Aα − Eα and Q α = B α + Eα + Kα . The
corresponding grand-potential densities are calculated algor-
ithmically as outlined by Choudhury et al. [21]. We point out that
the present generalization of free energy density by a polynomial
facilitates straightforward coupling with CALPHAD databases, as
exemplified by previous numerical studies on binary alloys [22–
27] and ternary alloys [28,29,21]. As the present investigation is
not limited to a specific alloy, we construct the parabolic free
energy densities such that the equilibrium mole fractions of B in
the solid grains and liquid phase are conveniently calculated to be
0.8 ( cBs, eq ) and 0.2 ( cBl, eq ), respectively. The relevant simulation
parameters are listed in their dimensional form in Table 1. Since
we are primarily interested in generic features of ripple formation,
the reported times are normalized by l02/Dl , where l0 = σ /( RT /Vm ) is
the capillary length, T denotes the isothermal transformation
temperature, s, interfacial energy, and Vm is the molar volume.
54 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59

Table 1 3.2. Grooving kinetics: Mullins's theory

Numerical parameters used for the phase-field simulations.
A grain boundary groove interface in 3-D may be divided into
Parameter Values (Jm
3) Parameter Values Units
three local regions:
Oα −4.4 × 103 Δx = Δy = Δz 1 × 10−7 m
Pα 1.7 × 103 ⎛ c l, eq − c l ⎞ 0.5, 0.6 – 1. Junction of more than two grains, where a groove pit develops.
u ⎜⎜ = A A ⎟
2. The solid–liquid interface far away from the grain boundaries
l, eq s, eq ⎟
⎝ cA − cA ⎠
that may be approximated by an infinite (periodic) surface.
Qα 1.7 × 102 σs, l 0.5, 0.6, 0.7 Jm
2
3. The solid-liquid interface around a grain boundary but away
Oβ −8.9 × 103 sgb 1.0 Jm
2
⎛ σs, l ⎞ 0.5, 0.6, 0.7 –
from the junctions of more than two grains. The grooving ki-
Pβ 1.4 × 104
r ⎜⎜ = ⎟⎟ netics of the third region can be studied efficiently with 2-D
⎝ σgb ⎠
phase-field simulations reported earlier by [31–33].
Qβ 5.7 × 103 DAl = DBl 1 × 10−9 m2s
1
δαβ (Eq. (A.4)) 0.1 Vm 2.24 × 10−4 m3 mol
1
We know from Mullins's theory [2] that d3g increases linearly
with time, t, i.e.
3. Validation with theory dg = d 0 t α , (4)

where d0 is a prefactor and the power-law exponent, α, equals 3

3.1. Dihedral angle: Young's law
The initial condition chosen for validation with Young's law
consists of two grains, each of numerical grid width and thickness
of 2000Δx and 100Δx respectively. Periodic boundary conditions
are applied on the left and right-hand edges of the numerical
domain. The dihedral angle pertains to the local equilibrium state
resulting from the force balance at the groove's triple-junction,
given by
⎛ σgb ⎞
θd = 2 cos−1⎜ ⎟,
⎝ 2σs, l ⎠
where the grain boundary energy, sgb, and the solid-liquid inter-
facial energy, σs, l , can be analytically derived. See [30,19] for de-
tails. We assume for simplicity that the interfacial energies are
isotropic, and the solid-liquid surface energies for both grains are
identical. In accordance with Eq. (3) the simulated dihedral angle
decreases with an increase of grain boundary energy. cf. Fig. 2a–c.
1
for volume diffusion-controlled grooving. In order to validate our
3-D phase-field model, we choose an initial arrangement of
4 equiaxed grains in equilibrium with melt surrounding both sides.
A representative snapshot corresponding to a dihedral angle of
120° is shown in left inset of Fig. 2 d. The temporal variation of
groove depths, dg, corresponding to root and pit (illustrated by
yellow and white arrows in Fig. 2d) are also plotted. The power
law exponent α obtained by fitting the simulated dg is found to be
0.3334 in both the cases, which is a result in good agreement with
Mullins's kinetic law, Eq. (4). The pre-factor d0 corresponding to
(3) the pit is larger compared to that for the root, which qualitatively
complies with the previous finding of Zhang and Gladwell [34].
4. Numerical investigation of grain rippling
A polycrystalline solidification front is never free of imperfec-
tions, especially grain boundary grooves that potentially act as

4
dg

2
Groove pit (numerical)
Groove root (numerical)
100 1000
0 1 2 Time (t)
Fig. 2. Simulations of isotropic grain boundary grooves: (a) σs, l = 0.5σgb , (b) σs, l = 0.6σgb and (c) σs, l = 0.7σgb . The illustrated snapshots of the groove region correspond to a
numerical domain size of 2000Δx × 200Δx , with periodic boundary conditions on the left and right edges. (d) Comparison of Mullins's theory with numerically simulated
grain-boundary grooving kinetics for volume-diffusion controlled regime. Data points represent the simulated values (log-log scale) whereas the lines are power law fits
(growth exponent ¼ 0.3334). The inset picture (on the left) shows an example phase-field simulation with periodic boundary conditions. The grains are plotted with offset to
reveal inner curvatures (see legend) in the vicinity of the triple-junction (right inlet), groove pits (white arrow) and roots (yellow arrow). (For interpretation of the references
to color in this figure legend, the reader is referred to the web version of this article).
 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59
sites for initiating morphological instabilities during directional
growth. As outlined in Section 1, such observations have been
reported in independent experimental works [11,13,35] on the
directional solidification of SCN-based alloys. In particular, it is
suggested that pulling velocity and alloy concentration also
influence the growth of morphological instabilities. As our step
towards a precise understanding of factors that lead to morpho-
logical instabilities of the solid-melt interface near grain bound-
aries, we numerically investigate the influence of surface energy
and grain misorientation on the spatial localization of incipient
perturbations. In this section, the reported phase-field simulations
correspond to the case of an isothermal solidification with a pre-
scribed melt supersaturation, u, proceeding from a preexisting
symmetric arrangement of 2 grains, as shown in Fig. 3a.
4.1. Symmetric grooves: influence of surface energy
In Fig. 3, we illustrate the influence of the ratios, r, between the
interfacial energy, σs, l , and the grain boundary energy, sgb, both of
which are assumed to be isotropic. We observe a spatial shift in
the location of incipient perturbations depending upon this ratio.
The temporal evolution shown in Fig. 3b, where the ratio r ¼0.5,
suggests that the initial perturbations develop along the curved
portion of the solid-liquid interface, and then continue to amplify.
At a larger ratio, where r ¼0.7, the ability of the grain boundary
groove to initiate a perturbation large enough to amplify appears
to be hampered, as is shown in Fig. 3c. Such a regime is char-
acterized by the concurrent amplification of perturbations loca-
lized along the curved surface away from the groove. It is note-
worthy that the flatter portion of solid-liquid interface, i.e., the
region around the grain center, is unable to initiate perturbations
irrespective of the value of r. The interfacial energy-dependent
localization of initial perturbations surrounding grain boundary
grooves reported in these simulations is corroborated by the ex-
perimental findings reported by Noël et al. [13], who observed the
onset of similar instability behavior which depended on the
pulling velocity for their SCN-based alloy. Apparently, such a ca-
pillary-mediated localization is implicitly related to the ratio of
driving force (which for the present case is supersaturation in the
melt) and the interfacial energy. In the present simulations where
the magnitude of melt supersaturation is held constant, a smaller
0.82
t = 12.0
r = 0.5
0.17
0.81
t = 8.0
0.17
0.76
(b) t = 2.4
0.16
0.4
(a) t = 0.0
0.2
0.80
(c) t = 15.2
0.16
0.82
t = 55.2
r = 0.7
0.15
0.81
t = 153.6
0.16
Fig. 3. Temporal evolution of the phase contours plotted over the corresponding
composition profile for (a) r ¼0.5 and (b) r ¼0.7, where r = σs, l/σgb .
55
interfacial energy allows the initiation of instabilities close to the
groove cusp.
The initial profile of the two grains chosen for the present phase-
field simulations is elliptical, which corresponds to a groove dihedral
angle of zero. From a numerical perspective, such a choice optimizes
the simulation run time, as the onset of morphological instabilities
relies upon the local gradient of interfacial curvature. Apparently, the
lateral flow of solute along the solid-liquid surface for an initially
planar solidification front is zero, corresponding to a groove angle of
180° as was the case in the works of Yeh et al. [19] and Miller et al.
[20]. As grain boundary grooving develops, the interfacial curvature
gradients promote the necessary flow of solute along the interface.
Obviously, such an initial condition is not favored from a numerical
perspective, as it prolongs simulation run-times.
4.2. Influence of anisotropy in surface energy
The correlation between grain boundaries and the formation of
ordered cellular arrays is reported in the experimental work of
Noël et al. [13]. It is suggested that undulations and cell alignment
along the grain boundary, typically observed during the initial
stages, gives way to flat, regular polygonal arrays of cells. In the
following section, we incorporate a smooth form of anisotropy in
solid-liquid energy (see Eq. A.4) to study the influence of grain
misorientation and melt supersaturation on the morphological
stability of the solid-liquid interface. We will establish the influ-
ence of these parameters on 3-D ordering based on insights de-
rived from these numerical simulations and reported in Section 5.
Here, we numerically simulate the evolution of grains at dif-
ferent supersaturations in the melt, u, proceeding from an initial
configuration illustrated in Fig. 3a. In Fig. 4, we compare grain
contours for grain misorientations of 30° and 45°, respectively. At a
supersaturation of u¼ 0.6, Fig. 4a, with solidification proceeding,
the primary ridge instability that corresponds to the more favor-
ably-oriented grain is found to amplify at a faster rate compared to
its counter-part, which is overgrown at a later stage. Drawing si-
milarities with the case-study reported in Section 4.1, the ob-
served rippling of solid-liquid surface in the present case is also
found to be dependent on grain misorientation. At a lower su-
persaturation in the melt, u ¼0.5, the overall rate of rippling is
reduced (Fig. 4b). Interestingly, the left ridge and the perturbation
located away from the groove, corresponding to the more favor-
ably-oriented grain, are coarser as compared to Fig. 4a. The 2-D
studies reported in this section signify an intricate correlation
between the grain misorientation and the observed evolution of
incipient ripples.
200
Δθ=30ο
(a)
Y grid-points
ο
Δθ=45
100
0
500 600 700 800 900 1000 1100 1200 1300 1400 1500
300
ο (b)
Δθ=30
Δθ=45 ο
Y grid-points
200
100
0
500 600 700 800 900 1000 1100 1200 1300 1400 1500
X grid-points
Fig. 4. Comparison of grain instability contours for assigned relative misorienta-
tions of 30° and 45°, respectively. The non-dimensional supersaturation, u, equals
(a) 0.6 and (b) 0.5. A representative time-step is selected for comparison In both
cases, when temporally amplifying the initial ridge on the right-hand side of the
groove cusp that overgrows the one on its left.
56 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59
5. Grain rippling in 3-D
The factors which determine whether or not a planar crystal
surface growing in the melt will break down into a cellular or
dendritic structure remain topics of interest. However, the influ-
ence of interface imperfections, such as grain boundary grooves,
on the stability of planar surfaces was reported in several experi-
mental works, but has never been numerically investigated in 3-D.
We use the 3-D phase-field modeling in the following section,
outlined in Appendix A, to investigate the influence of grain mis-
orientation as one of factors that trigger instability of an advancing
solid-liquid interface.
5.1. Earlier experiments
Before discussing our 3-D phase-field simulation results, the
seminal work of Schaefer and Glicksman (See Fig. 3 in [11]) that
underpins the concept of crystalline defect-mediated surface rip-
pling is well worth revisiting. Specifically, Schaefer and Glicksman
[11] showed experimentally that grain boundary and sub-bound-
ary grooves intersecting a nominally planar solid-liquid interface
play essential roles in initiating morphological instability. An in-
tersecting straight grain boundary forms a persistent groove-like
perturbation on the interface, the lateral dimensions and depth of
which depend inversely on the square-root of the local thermal
gradient acting normal to the interface. Morphological instability
initiates by the localized growth of a pair of ridges that form
parallel to the grain boundary groove, and which approach each
other and merge along the triple line formed by the grain
boundary and the planar solid-liquid interface. Ridge growth along
the grain boundary groove is immediately followed by the for-
mation of weaker, parallel, secondary ridges, with a shallow in-
tervening valley connecting the primary and secondary ridges.
Primary ridges then progressively break down into a chain of
(a)
t=0
(c)
t = 12.8
Fig. 5. 3-D phase-field simulation showing the morphological evolution of an advancing crystal-melt interface in the vicinity of a grain boundary groove. (a) Initial grain
boundary groove appears on the interface. (b) Ridges form and bulge out towards the melt. The primary ridges formed near the groove are taller in comparison to the weak
secondary ridges adjacent to them. (c) Modulations are observed along the ridges which continue to amplify. Secondary ridges die-out. (d) Hillocks form as periodic ripples
along the ridges as the solid-liquid interface advances to form more complex patterns.
periodic hillocks in the third dimension that follow the peak of the
primary ridges. Hillock formation is initiated where a grain
boundary groove terminates at an intersection with another
groove, or where the grain touches the container. The secondary
ridges frequently disappear as the primary ridges extend. Where
two grain boundary grooves happen to intersect, a deep pit de-
velops, which, in turn, often forms several “pre-dendritic disks.”
These pre-dendritic disks extend rapidly as protuberances into the
undercooled or supersaturated melt to form a trio of highly-ra-
mified dendrites that grow in preferred crystallographic directions
set by the orientation of each of the three participating grains.
5.2. Rippling simulations
In order to validate our numerical technique with the in-
stability mechanism suggested above, we started with a sym-
metric arrangement of 4 equiaxed grains, similar to Fig. 2d, from
which we simulated the early-stage evolution of the solid-liquid
interface. The supersaturation in the melt contacting the 4 grains
is u ¼0.5, with misorientations in range of 10° to 20°. The temporal
evolution of the solid-liquid interface adjacent to the boundary
groove is illustrated in Fig. 5, shows initial formation of parallel
ridges and then periodic modulation in the third dimension. As the
parallel ridges amplify, secondary ridges form beside them. Once
the primary ridges develop more height, they quickly modulate
into hillocks that continue to amplify and form a ramified pattern.
It is worth noting that the simulated progression in 3-D is in close
agreement with experiment.
Next, we investigated the influence of the grain quadruple
junction and its spatial arrangement of secondary ridges and ripples
in 3-D. The initial setting comprises a supersaturated melt, with
u¼0.5, in contact with a polycrystalline seed consisting of 4 grains
symmetrically arranged in an array. The grains are misoriented with
respect to the most-preferred-orientation in the order Grain 1
(b)
t = 8.0
(d)
t = 19.2
 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59 57

Grain 2 Grain 3

Grain 1 Grain 4

(a) t = 0.0 (b) t = 40.0

(c) t = 28.8 (d) t = 32.0 (e) t = 35.2

Fig. 6. (a) Morphological evolution of an initially flat polycrystalline ‘seed’ in contact with supersaturated melt, illustrated at representative time-steps. The corresponding
axial tilts (θ) for grains 1, 2, 3 and 4 are 15.5,°, 25.5°, 31.6° and 18.1°, respectively. (b) Magnified view of grain 1 /melt surface close to quadruple junction.

(θ = 15.5°) < Grain 4 (θ = 18.1°) < Grain 2 (θ = 25.5°) < Grain 3 (θ = 31.6°), Towards grain center
where θ represents axial-tilt (See Fig. A8c.) Fig. 6 illustrates pro-
gressive amplification of primary and secondary ridges at re-
presentative time-steps.
As the melt supersaturation was chosen to be large enough to
initiate morphological instabilities near the groove, as well as at
the grain center, primary ridges develop adjacent to the groove as
do ripples located close to the grain centers. Another interesting
characteristic is that the primary ridge close to the grain quadruple
junction is found to amplify at a much larger rate as compared to
those along the triple line regions. The influence of grain mis-
orientation is also apparent; the fastest amplifying ridges corre-
spond to the most-favorably oriented grain, which is grain 1 in the (a)
present setting. At a given time, the amplitude of ripples closer to
Grain Boundary
multi-junctions is much larger as compared to those near grain
centers. Before the amplifying ripples evolve to a complex pattern,
the following order of amplification rate is observed: (1) Primary
ridges next to the quadruple junction; (2) primary ridges along the
triple junction; (3) secondary ridges near the quadruple junction;
(4) secondary ridges adjacent to the triple junction, (5) secondary
ridges closer to the grain centers.
Finally, we comment upon the influence of grain misorienta-
tion on the spatial ordering of primary ridges and ripples found in
the experiments reported by Noël et al. [13]. In Fig. 7a, we illus-
trate surface ripples adjacent to the grain boundary interface be-
tween grain 3 (θ ¼ 31.6°) and grain 2 (θ ¼ 25.5°). The primary rid- (b)
ges that are tilted along θ with respect to the most-preferred-or-
ientation (chosen to be the same as the viewing angle) are found Fig. 7. Top-view illustrating the spatial arrangement of primary ridges and ripples
across the grain boundary interfaces of grains (a) grain 3 (left) and grain 2 (right),
to be ordered along the grain boundary. However, the extent of and (b) grain 1 (left) and grain 4 (right).
ordering progressively declines on moving away from the grain
boundary interface. The ordering of primary and secondary ridges, 6. Conclusions and outlook
shown in Fig. 7b, in the form of oblong cells is observed to be more
pronounced across the interface between grain 1 (θ ¼15.5°) and We explored persistent perturbations and supersaturation—
grain 4 (θ ¼18.1°). factors that initiate morphological instability of a solidifying
58 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59

planar solid-liquid interface by using phase-field simulation. We N

focused primarily on the morphological evolution occurring near Ψ (T , μ, ϕ) = ∑ Ψα (T , μ) hα (ϕ),

α=1
grain boundaries at sites where perturbations develop. First, the
K− 1
required force balance at the equilibrium dihedral angle was es-
Ψα (T , μ) = fα (c α ( T , μ), T ) − ∑ μi ciα (T , μ),
tablished for different ratios of surface and grain boundary en- i=1 (A.2)
ergies, evidenced by agreement between simulated dihedral an-
gles and Young's law. Next, simulated grooving kinetics was found where hα ( ϕ) is an interpolation function of the form
to obey kinetics in close agreement with Mullins's groove growth ( )
hα ϕ = ϕα2 3 − 2ϕα . ( )
power-law. The validated model was then used to conduct a sys- The gradient energy density, a ( ϕ, ∇ϕ), is given by
tematic numerical investigation to decompose the influences of
N, N
grain misorientation and melt supersaturation on the critical onset
σαβ ⎡⎣ a c qαβ ⎤⎦ qαβ ,
2 2
of interfacial ripples. For the first time, we demonstrate that the
a (ϕ, ∇ϕ) = ∑ ( )
α, β = 1
ratio of solid-liquid interface and grain boundary energies is one of ( α < β) (A.3)
the factors which cause localization of incipient perturbations.
Using 3-D phase-field modeling, we established the role of ( )
where ac qαβ defines the form of the surface energy anisotropy of
grain misorientation on influencing the spatial alignment of in- the evolving phase boundary and σαβ is the interfacial free energy
terface ripples. We also showed that the evolution of primary per unit area of the α /β phase boundary, which may additionally
ridges along straight grain boundaries is similar to the instability depend on the relative orientation of the interface. The vector
phenomena observed in directional solidification experiments on quantity qαβ = ϕα ∇ϕβ − ϕβ ∇ϕα is a generalized gradient vector
SCN-based alloys [11,13]. The present work, whilst providing va- normal to the α /β interface.
luable insights, falls short of a fully detailed parametric study. The To assign an isotropic surface energy to the α /β phase bound-
phase-field model could be extended to investigate the influences ( )
ary, ac qαβ is chosen to be 1. Weakly anisotropic crystals with an
of the thermal gradient and melt convection to facilitate a more underlying cubic symmetry are modeled in three dimensions by
quantitative comparison of numerical results with experiments. the expression
The topic of complex pattern evolution, typically observed at later
stages as the initial perturbations progressively amplify, has not ⎛ 4 ⎞
⎜ qαβ ⎟
4
been addressed in the present article. Complementary numerical a c (qαβ ) = 1 − δαβ ⎜ 3 ∓ 4 4 ⎟,
studies addressed to these interesting topics are currently in ⎜ qαβ ⎟
⎝ ⎠ (A.4)
progress.
d ⎛
4 ⎞ ⎡ d ⎤2
with q
4 ⎝ ⎠ ( )
= ∑i = 1 ⎜ qi4 ⎟ and q 4 = ⎢⎣ ∑i = 1 qi2 ⎥⎦ . The anisotropic

Acknowledgments strength of an α /β interface is given by the parameter δαβ . A polar

plot of function A.4 is shown In Fig. A8a. An alternative formula-
tion, corresponding to a strongly anisotropic crystal of faceted type
KA thanks GEOLAB for providing financial support and a mul-
tidisciplinary research platform. HX acknowledges his support was employed elsewhere [37–40]. The evolution equation for the
from National Natural Science Foundation of Shaanxi Province in N phase-field variables can be written as,
China (No. 2015JQ5125). BN acknowledges the financial support of ∂ϕα ⎛ ∂a ( ϕ, ∇ϕ) ∂a ( ϕ, ∇ϕ) ⎞ 1 ∂w ( ϕ) ∂Ψ ( T , μ, ϕ)
τϵ = ϵ ⎜⎜ ∇· − ⎟⎟ − − − Λ,
DFG in the framework of Sino-German research effort on “Phase ∂t ⎝ ∂∇ϕα ∂ϕα ⎠ ϵ ∂ϕα ∂ϕα (A.5)
transformations under extreme conditions” (No. NE 822/22-1).
MEG acknowledges his support from the Allen S. Henry Dis- where Λ is the Lagrange parameter that maintains the constraint
N
tinguished Professorship at Florida Institute of Technology. ∑α = 1 ϕα = 1. The equilibrium crystal shape simulated (corre-
sponding to function A.4) using the volume-preservation techni-
que [41] is illustrated in Fig. A8b.
Appendix A. Phase-field model The concentration fields are obtained by a mass conservation
equation for each of the K
1 independent concentration variables
In this appendix, for the sake of completeness, we briefly re- ci. The evolution equation for the concentration fields can be de-
count the multiphase-field equations used to study micro- rived as,
structural evolution. The reader is referred to our prior studies ⎛ K− 1 ⎞
∂ci
[36,22,24,28,25] for a more detailed description of the simulation = ∇·⎜⎜ ∑ Mij (ϕ)∇μj ⎟⎟,
methods employed. ∂t ⎝ j=1 ⎠ (A.6)
The evolution of phases is governed by the phenomenological
minimization of the grand potential functional Ω, and
⎡ ⎛ ⎞⎤ N
Ω (T , μ, ϕ) = ∫V ⎢⎣ Ψ (T , μ, ϕ) + ⎜⎝ ϵa (ϕ, ∇ϕ) + 1ϵ w (ϕ) ⎟⎠ ⎥⎦ dV . (A.1)
Mij (ϕ) = ∑ Mijα gα (ϕ),
α=1 (A.7)
Here T is the temperature, μ is a chemical potential vector con- where each represents the mobility matrix of the phase α
Mijα
sisting of K
1 independent chemical potentials, ϕ is the phase- (related to solute diffusivities, Dl and Dg in the melt and grains,
field vector containing the volume fractions of the N-phases and ϵ respectively).
is the length scale related to the interface. a ( ϕ, ∇ϕ) an w ( ϕ) re- The removal of artificial, thin-interface solute trapping is
present the gradient and obstacle potential type energy density, achieved by incorporating the form of anti-trapping current pro-
respectively, and finally V denotes the domain volume. posed by Choudhury and Nestler [36] based on a double-obstacle
The grand potential density Ψ ( T, μ, ϕ), which is the Legendre potential formulation. The model equations are discretized using a
transform of the free energy density of the system, f ( T, c , ϕ), is finite-difference method. The spatial derivatives are subdivided
written as an interpolation of individual grand potential densities, into left-sided, right-sided, and centered differences, such that
namely, grid anisotropy is minimized. The time derivative obeys an explicit
 K. Ankit et al. / Journal of Crystal Growth 457 (2017) 52–59
Small Curvature Large
(a) (b)
Fig. A8. (a) Interfacial energy plotted in polar coordinates and (b) the corresponding equilibrium shape simulated using the volume preservation algorithm of Nestler et al.
(c) The axial tilt with respect to the most-preferred-orientation aligned along the z-axis and rotation are used to define the crystal misorientation for the phase-field
simulations of directional solidification.
Euler scheme. The grand-chemical potential model is im-
plemented in the Pace3D software package version 2.1.1. As a test
case, we used identical simulation parameters, as reported in Fig. 3
with a smaller grid size of 0.1Δx to observe the influence of in-
terfacial energies on rippling.
On comparing the evolution of phase contours (assuming iso-
tropic interfacial energies) corresponding to 1.0Δx with 0.1Δx , it
becomes apparent that the simulation results are independent of
grid size.
References
[1] W. Mullins, J. Appl. Phys. 28 (1957) 333–339.
[2] W. Mullins, Trans. Metall. Soc. AIME 218 (1960) 354–361.
[3] G. Bolling, W. Tiller, J. Appl. Phys. 31 (1960) 1345–1350.
[4] S. Coriell, R. Sekerka, J. Cryst. Growth 19 (1972) 90–104.
[5] S. Coriell, R. Sekerka, J. Cryst. Growth 19 (1973) 285–293.
[6] B. Bokstein, L. Klinger, I. Apikhtina, Mater. Sci. Eng. A 203 (1995) 373–376.
[7] R. Sekerka, J. Cryst. growth 10 (1971) 239–250.
[8] S. Hardy, G. McFadden, S. Coriell, J. Cryst. Growth 114 (1991) 467–480.
[9] L. Morris, W. Winegard, J. Cryst. Growth. 5 (1969) 361–375.
[10] R. Sharp, A. Hellawell, J. Cryst. Growth 6 (1970) 334–340.
[11] R. Schaefer, M. Glicksman, Metall. Trans. 1 (1970) 1973–1978.
[12] M. Noël, H. Jamgotchian, B. Billia, J. Cryst. Growth 181 (1997) 117–132.
[13] M. Noël, H. Jamgotchian, B. Billia, J. Cryst. Growth 187 (1998) 516–526.
[14] L.-Q. Chen, Ann. Rev. Mater. Res. 32 (2002) 113–140.
[15] W. Boettinger, J. Warren, C. Beckermann, A. Karma, Ann. Rev. Mater. Res. 32
(2002) 163–194.
[16] K. Thornton, J. Ågren, P. Voorhees, Acta Mater. 51 (2003) 5675–5710.
[17] I. Singer-Loginova, H. Singer, Rep. Prog. Phys. 71 (2008) 106501.
[18] B. Nestler, A. Choudhury, Curr. Opin. Solid State Mater. Sci. 15 (2011) 93–105.
59
0
Rotation axis
Z Ax
ial
tilt
45
Y
(c) X
[19] S. Yeh, C. Chen, C. Lan, J. Cryst. Growth 324 (2011) 296–303.
[20] W. Miller, A. Popescu, G. Cantù, J. Cryst. Growth 385 (2014) 127–133.
[21] A. Choudhury, M. Kellner, B. Nestler, Curr. Opin. Solid State Mater. Sci. 19
(2015) 287–300.
[22] D. Molnar, R. Mukherjee, A. Choudhury, A. Mora, P. Binkele, M. Selzer,
B. Nestler, S. Schmauder, Acta Mater. 60 (20) (2012) 6961–6971.
[23] E. Wesner, A. Choudhury, A. August, M. Berghoff, B. Nestler, J. Cryst. Growth
359 (2012) 107–121.
[24] K. Ankit, A. Choudhury, C. Qin, S. Schulz, M. McDaniel, B. Nestler, Acta Mater.
61 (11) (2013) 4245–4253.
[25] K. Ankit, R. Mukherjee, T. Mittnacht, B. Nestler, Acta Mater. 81 (2014) 204–210.
[26] K. Ankit, R. Mukherjee, B. Nestler, Acta Mater. 97 (2015) 316–324.
[27] K. Ankit, T. Milttnacht, R. Mukherjee, B. Nestler, Comput. Mater. Sci. 108B
(2015) 342–347.
[28] R. Mukherjee, A. Choudhury, B. Nestler, Model. Simul. Mater. Sci. 21 (7) (2013)
075012.
[29] J. Hötzer, M. Jainta, P. Steinmetz, B. Nestler, A. Dennstedt, A. Genau, M. Bauer,
H. Köstler, U. Rüde, Acta Mater. 93 (2015) 194–204.
[30] J. Warren, R. Kobayashi, A.E. Lobkovsky, W.C. Carter, Acta Mater. 51 (2003)
6035–6058.
[31] M. Bouville, S. Hu, L.-Q. Chen, D. Chi, D. Srolovitz, Model. Simul. Mater. Sc. 14
(2006) 433–443.
[32] M. Bouville, D. Chi, D. Srolovitz, Phys. Rev. Lett. 98 (2007) 1–4.
[33] M. Bouville, D. Chi, D. Srolovitz, Solid State Phenom. 129 (2007) 89–94.
[34] W. Zhang, I. Gladwell, J. Cryst. Growth 277 (2005) 608–622.
[35] H. Xing, C. Wang, J. Wang, C. Chen, Sci. China Ser. G Phys. Astron. 54 (2011)
2174–2180.
[36] A. Choudhury, B. Nestler, Phys. Rev. E 85 (2) (2012) 21602.
[37] B. Nestler, H. Garcke, B. Stinner, Phys. Rev. E 71 (2005) 041609.
[38] K. Ankit, B. Nestler, M. Selzer, M. Reichardt, Contrib. Miner. Pet. 166 (2013)
1709–1723.
[39] K. Ankit, J. Urai, B. Nestler, J. Geophys. Res. Solid Earth 120 (2015) 3096–3118.
[40] K. Ankit, M. Selzer, B. Hilgers, C. Nestler, J Petrol. Sci. Res. 4 (2015) 79–96.
[41] B. Nestler, F. Wendler, M. Selzer, B. Stinner, H. Garcke, Phys. Rev. E 78 (2008)
011604.