Solidification Review

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/222541258
Solidification microstructures and solid-state parallels: Recent

developments, future directions
Article in Acta Materialia · February 2009

DOI: 10.1016/j.actamat.2008.10.020 · Source: OAI
CITATIONS READS
678 994
9 authors, including:
Mark Asta Alain Karma

University of California, Berkeley Northeastern University
342 PUBLICATIONS 23,977 CITATIONS 310 PUBLICATIONS 26,702 CITATIONS
SEE PROFILE SEE PROFILE
W. Kurz Mathis Plapp

École Polytechnique Fédérale de Lausanne French National Centre for Scientific Research
202 PUBLICATIONS 17,892 CITATIONS 110 PUBLICATIONS 5,420 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Oxide on Energy Application View project
Columnar-to-Equiaxed Transition in Solidification Processing (CETSOL) View project
All content following this page was uploaded by Mathis Plapp on 24 April 2018.
The user has requested enhancement of the downloaded file.

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Available online at www.sciencedirect.com
Acta Materialia 57 (2009) 941–971

www.elsevier.com/locate/actamat
Overview No. 146
Solidification microstructures and solid-state

parallels: Recent developments, future directions
M. Asta a, C. Beckermann b, A. Karma c, W. Kurz d,*, R. Napolitano e,
M. Plapp f, G. Purdy g, M. Rappaz d, R. Trivedi e
a
Department of Chemical Engineering and Materials Science, University of California at Davis, Davis, CA 95616, USA
b
Department of Mechanical and Industrial Engineering, The University of Iowa, Iowa City, IA 52242, USA
c
Department of Physics and Center for Interdisciplinary Research on Complex Systems, Northeastern University, Boston, MA 02115, USA
d
Institute of Materials, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland
e
Department of Materials Science and Engineering, Iowa State University, and Ames Laboratory USDOE, Ames, IA 50011, USA
f
Physique de la Matière Condensée, Ecole Polytechnique, CNRS, 91128 Palaiseau, France
g
Department of Materials Science and Engineering, McMaster University, Hamilton, Ont., Canada L8S 4L7
Received 11 January 2008; received in revised form 2 October 2008; accepted 2 October 2008
Available online 26 December 2008
Abstract
Rapid advances in atomistic and phase-field modeling techniques as well as new experiments have led to major progress in solidifi-
cation science during the first years of this century. Here we review the most important findings in this technologically important area
that impact our quantitative understanding of: (i) key anisotropic properties of the solid–liquid interface that govern solidification pat-
tern evolution, including the solid–liquid interface free energy and the kinetic coefficient; (ii) dendritic solidification at small and large
growth rates, with particular emphasis on orientation selection; (iii) regular and irregular eutectic and peritectic microstructures; (iv)
effects of convection on microstructure formation; (v) solidification at a high volume fraction of solid and the related formation of pores
and hot cracks; and (vi) solid-state transformations as far as they relate to solidification models and techniques. In light of this progress,
critical issues that point to directions for future research in both solidification and solid-state transformations are identified.
Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Solidification microstructures; Solid-state phase transformations; Solid-liquid interface; Phase-field modeling; Atomistic modeling
1. Introduction fication science advanced considerably through pioneering

studies of segregation, morphological stability and micro-
The basic problem of understanding and controlling structure selection, and, more recently, through rapid
microstructures during solidification continues to present developments in both experimental and multiscale compu-
numerous scientific and technological challenges. Reviews tational tools that have yielded a wealth of new informa-
from two workshops, held at Zermatt in 1988 and 1998, tion and quantitative insights into dendritic and two-
have been published in this journal reflecting the advances phase microstructures in a variety of alloys. The 2006 Ames
in solidification science at the end of the 20th century [1,2]. workshop distinguished itself from the previous Zermatt
The last workshop of this kind took place in Ames, IA, in workshops by the inclusion of one session devoted to
2006. Its main sessions focused on areas of solidification exploring parallels between solid-state phase transforma-
that have been extensively studied by one of the coauthors tions and solidification. This session highlighted the poten-
(R.T.) during the past 40 years. During this period, solidi- tial for cross-fertilization of ideas and techniques between
these two fields.
*
Corresponding author. Tel.: +41 21 693 2982; fax: +41 21 693 5890. The dual goal of this review, which is based on the pre-
E-mail address: wilfried.kurz@epfl.ch (W. Kurz). sentations of the Ames workshop and on research of the
1359-6454/$34.00 Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2008.10.020
942 M. Asta et al. / Acta Materialia 57 (2009) 941–971
following two years, is to describe recent developments in and of deformation, including nucleation and growth of
solidification science, as they relate to the formation and the voids, as well as percolation of the solid network.
control of solidification microstructure in metallic materi- Once the primary process of solidification is complete,
als that point to interesting future directions, and to discuss the solid generally transforms into new phases and second-
certain solid-state parallels. ary microstructures. As discussed in Section 7, solidifica-
Section 2 is devoted to reviewing recent progress made tion models and related experimental techniques such as
in using atomic-scale simulations to determine fundamen- directional solidification provide a valuable basis to exam-
tal equilibrium and non-equilibrium properties of the ine and understand various transformations in the solid
solid–liquid interface. Those include the excess free energy state. On the other hand, solid-state transformations pro-
of the solid–liquid interface and the interface attachment vide an interesting setting to test these models in new
kinetic coefficient, as well as their anisotropies, and other regimes that are not usually accessible in a solidification
parameters that characterize the departure from chemical context, such as the limit of absolute morphological stabil-
equilibrium at the interface. Even though those parameters ity at high growth rates.
form the basis of all microstructure modeling, they have We conclude each of the above sections with an outlook
been notoriously difficult to measure experimentally. statement that is intended to identify the most critical out-
Insight into the difficult topic of nucleation can also be standing questions, along with promising areas of future
obtained by the use of atomistic modeling. From this research. It is hoped that this review will stimulate future
standpoint, these techniques have provided an unprece- research to complete our picture of these important and
dented opportunity to determine the behavior of metallic interesting phenomena.
systems and have emerged as indispensible new tools in
solidification science. 2. Atomistics of the solid–liquid interface
Section 3 discusses insights in pattern selection obtained
using the phase-field method that has emerged as a method In phase transformations the evolution of microstruc-
of choice for quantitative simulations of interface dynam- ture is influenced both by transport phenomena and the
ics. Dendrite growth at small growth rates is discussed, formation of heterophase interfaces. In this section we con-
with particular emphasis on the question of orientation sider the second class of processes, as the related interfacial
selection. Anomalous dendrite growth directions with off properties, which are not well known, influence all solidifi-
crystal axes are explained in the setting of an enlarged cation phenomena. In solidification, atomistic processes at
two-dimensional parameter space characterizing the the solid–liquid interface are typically described with refer-
anisotropy of the interfacial free energy derived from atom- ence to three parameters: (i) the interfacial free energy, c,
istic simulations. Rapid dendrite growth and associated representing the reversible work to form an interface
transitions in the solidification front morphology are also between solid and liquid phases; (ii) the kinetic coefficient,
discussed. l, reflecting attachment kinetics; and (iii) the velocity-
In Section 4 results of experiments and phase-field com- dependent partition coefficient, k(V) = xS/xL, with xS and
putations of coupled growth of eutectic and peritectic alloys xL denoting solute mole fractions on solid and liquid sides
are presented. Recent theoretical and experimental results of the interface (see the list of symbols in the appendix).
of morphological stability and spacing selection mecha- Sharp-interface theories and phase-field simulations have
nisms of the two-phase interfaces are shown. Industrially demonstrated pronounced sensitivities of solidification
important irregular eutectic growth, which is controlled by rates and growth morphologies to the magnitudes and crys-
anisotropic interface properties, is also considered. talline anisotropies of these interfacial properties (see e.g.
Section 5 concerns the effect of convection on solidifica- [3–5] and Section 3).
tion microstructure. Coupling of macroscopic fluid flow While the availability of accurate values for these inter-
with microscopic temperature and solute fields is instru- facial parameters is thus critical for quantitative modeling
mental for a better understanding of practically relevant of solidification microstructures, significant challenges are
situations where convection always plays an important associated with their direct measurement. To date, limited
role. The influence of convection on columnar and equi- measurements have been reported for c anisotropy in alloys
axed dendritic growth, on dendrite trunk/arm develop- [6–9]. The magnitude of l in metals has been measured
ment, on the corresponding spacings, and on eutectics directly for only two elemental systems [10,11], and exper-
and peritectics is presented. imental data for the anisotropy of l or k(V) in such systems
Solidification under industrial conditions often involves is lacking. This situation has provided strong motivation
large thermal gradients and thus induces thermal strains for the application of atomic-scale simulations as a frame-
and residual stresses in the final product. This has many work for the direct calculation of both equilibrium and
practical consequences that engineers must account for. non-equilibrium properties of crystal–melt interfaces in
In this context, Section 6 is concerned with solidification metals, alloys and related model systems [12–14]. This sec-
microstructures at high solid fractions and related defects, tion reviews these results, derived within the framework of
such as pores and hot cracks. Understanding such defects molecular dynamics (MD) methods (an introduction to
requires a combined treatment of heat and mass transport MD methods can be found in Ref. [15]).
M. Asta et al. / Acta Materialia 57 (2009) 941–971 943
Owing to the required system sizes and time scales, the nitude of c is assumed to originate from the (negative)
MD simulations discussed here are based on the use of excess entropy associated with the ordering of the liquid
classical models for interatomic forces. Studies of crystal– against the solid. In this theory the Turnbull relation holds
melt interface properties have made use of both idealized exactly and, under the assumption that the liquid structure
pair-potential models, such as Lennard–Jones (LJ), and is characterized by tetrahedral packing, a number of
the more realistic many-body embedded atom method a = 0.86 is derived for fcc crystals. This value has been
(EAM) potentials for metals [16,17]. Comparisons of widely applied in modeling of experimental nucleation data
results obtained with different potential models has proven [28]. The SMT theory leads to an important prediction that
useful in determining the extent to which specific interface a varies significantly with crystal structure. Specifically, the
properties are sensitive to the details of the interatomic SMT theory predicts that the Turnbull coefficient for body-
interactions, or are dictated more generally by factors such centered cubic (bcc) crystals has a value 18% lower than
as the crystal structure. that for fcc crystals [25–27]. The lower bcc value is qualita-
In the following the topic is presented in three parts: tively consistent with experiments in Fe- and Ni-based sys-
solid–liquid interface energies and their anisotropies, inter- tems, where primary nucleation of metastable bcc phases
face kinetics and recent applications of atomistic simula- has been observed in rapidly quenched alloys with stable
tions in the investigation of crystal nucleation. fcc structures [29–36]. The SMT theory was recently
employed in calculations of a for a wider variety of com-
2.1. Crystal–melt interfacial free energies pounds, possessing complex ‘‘polytetrahedral” order [37–
39]. These calculations yield results pointing to an even
In the formation of solidification microstructures, both stronger dependence of a on crystal structure. Calculations
the magnitude of c and its crystalline anisotropy play for the Frank–Kasper (r) phase, the k and l phases in Al–
important roles. The former represents a critical term in Fe, and a quasicrystalline icosahedral (I) structure yield:
the reversible work to form a solid nucleus, and thus has aI/afcc 0.40 < ar/afcc 0.42 < ak/afcc 0.46 < al/afcc
a strong influence of nucleation kinetics. Further, signifi- 0.51 < abcc/afcc 0.82. These results, when combined with
cant structural variations in c can lead to the primary classical nucleation theory, are consistent with observa-
nucleation of metastable phases in deeply undercooled tions of primary nucleation in maximum undercooling
melts. The anisotropy of c accounts for a sizeable contribu- measurements for Al–(Cu)Fe, Al–(Cu–)Co, Al–(Ni–)Co
tion to the interface stiffness, which underlies Gibbs– and Al–Pd–Mn [28,40,41].
Thomson effects for anisotropic interfaces, and can thus Since the pioneering work of Broughton and Gilmer
be an important factor governing growth morphologies [42], atomistic simulations have been applied extensively
(see Section 3 below). We review here the current state of in calculations of c for a variety of systems [12–14].
understanding of both the magnitude and anisotropy of c Fig. 1 summarizes the results to date from atomistic calcu-
in metals in light of experimental observations and atomis- lations for metals modeled with EAM potentials [43]. The
tic-simulation results. results are plotted in a manner inspired by Eq. (1), as
cq2/3 vs. L, where the slopes of the best-fit lines corre-
2.1.1. Magnitudes of crystal–melt interfacial free energies spond to the average Turnbull coefficients for fcc and bcc
In the seminal work of Turnbull [18], the first measure-
ments of c were derived over 50 years ago from nucleation
0.12
studies in undercooled melts. Turnbull demonstrated a FCC: α =0.55 ACKFe
PAIRFe
strong correlation between c and the ratio of the latent heat SMFNi
RJMo
of melting (L, per atom) to the average interfacial area per FBDNi FAV
γρ-2/3 (eV/atom)
atom (expressed as q2/3, where q is the solid atomic 0.08

FBDCu ACKFe
FDNiAl
density): EAAl MENFe PAIRFe
2=3
VCAu
BCC: =0.3
BCC:α
c ¼ aLq ð1Þ VCAg
0.04
In Eq. (1) a is a constant that has come to be referred to as LOEPb
the Turnbull coefficient; it was originally reported to have a HCP Mg: α=0.5
=0.5
value of approximately 0.45 for metals, and this value de-
0
scribes well the values of c compiled from more recent
0 0.1 0.2 0.3 0.4
nucleation data in Ref. [19]. Slightly higher estimates are
L (eV/atom)
derived from data based on measurements other than
nucleation: a = 0.6 was quoted [20] for face-centered cubic Fig. 1. Calculated interface free energies for metals as a function of latent
(fcc) metals based on data [21] from the dihedral-angle heat of melting. The superscripts in the labels of the data points
technique [22,23], and a = 0.55(8) is obtained from analysis correspond to the authors who developed the different interatomic
potentials, and are explained in Ref. [12]. Note that for some of the
considering a combination of measurements in Ref. [24]. elements (e.g. Fe and Ni) more than one potential was used and, since
In the theory of crystal–melt interfacial free energies due these potentials generally gave different melting points and latent heats,
to Spaepen, Meyer and Thompson (SMT) [25–27] the mag- they led to different estimates for c.
metals (a result is also plotted for hcp Mg). The average fcc and sixfold anisotropy cubic-harmonic functions K1 and K2
value of a = 0.55 derived from the simulations is within the [47] (in terms of the components ni interface normal ^ n,
range of the experimental estimates quoted above. The sim- these P functions can P be written as
3 3
ulation results also confirm the trend towards lower aver- K 1 ¼ i¼1 n4i 3=5 and K 2 ¼ 3 i¼1 n4i þ 66n21 n22 n23 17=7Þ.
age values of a for bcc relative to fcc structures. In two The need for a two-term anisotropy expansion for c(^ n)
recent simulation studies [44,52] the crystal-structure in fcc metals was first demonstrated in atomistic simula-
dependence of a was considered in detail, employing inter- tions for EAM Ni [48] and Al [49], employing an MD tech-
atomic potentials that gave rise to both stable and metasta- nique that was specially designed to probe c anisotropies
ble fcc and bcc phases, allowing c to be calculated directly through analyses of equilibrium capillary fluctuations. Cal-
for both crystal structures, with the same underlying poten- culated results for e1 and e2, derived from this capillary
tial. For a system with repulsive power-law potentials (of fluctuation method (CFM) for a number of other fcc met-
the form 1/rn, with n = 6–8), a for bcc was found to be als and alloys are plotted in Fig. 2, adapted from Ref. [50]
on average about 15–20% lower than for fcc [52]; simula- (see references therein and Ref. [51] for citations to the ori-
tions for an EAM model of Fe gave values of a for bcc ginal work). In all cases it is found that e1 is positive, with
30–35% lower than fcc [44]. These results provide direct magnitudes ranging between 3% and 10%, while e2 is neg-
evidence of a sizeable effect of crystal structure on c, and ative, with values roughly an order of magnitude smaller.
confirm the trend towards a significant reduction in a for Spherical plots of the inverse of the interfacial stiffness
bcc structures relative to fcc. Eq. (6), shown in Fig. 2a, illustrate that positive e1 and neg-
In the plot of calculated interfacial free energies for ative e2 values favor dendrite growth directions along
EAM metals shown in Fig. 1, scatter in the data about h1 0 0i and h1 1 0i, respectively, as discussed in more detail
the best-fit lines is apparent, particularly for bcc metals. in Section 3.
The results highlight the fact that, although the Turnbull It is important to emphasize that the nature of c aniso-
scaling relation, with constant a for a given crystal struc- tropies calculated for fcc EAM metals have also been
ture, provides a good estimate of the value of c, it is not found in a wide variety of model fcc systems, including
exact. Indeed, atomistic simulations have shown clear vari- hard spheres [45], repulsive power-law [52] and LJ poten-
ations in a within a given crystal structure, depending on tials [53,54]. Additionally, these findings have been
the detailed nature of the interatomic potentials [52,46], obtained both by the CFM and an independent MD
and they also have provided some hints into the nature method involving interface cleaving [42,54,55]; the results
of these dependencies. For repulsive 1/rn potentials one are thus robust with respect to the details of the calcula-
obtains increasing values of a (from 0.47 to 0.60) as the tions. All of the calculated results for EAM and model sys-
range of the interaction decreases (from n = 6 to n = 1) tems are consistent with the ordering c100 > c110 > c111 for
[52]. Similarly, Morris et al. [46] calculated c for fcc Al the high-symmetry {1 0 0}, {1 1 0} and {1 1 1} planes; differ-
employing three different EAM potential models, each giv- ences between c for these orientations are generally found
ing similar values for the melting temperature, L and q; the to be on the order of a few percent. This relative ordering
potentials gave rise to values of a ranging between 0.41 and of the magnitudes of c values, as well as the signs of the cal-
0.60, and show a trend towards higher interfacial free ener- culated anisotropy parameters in Fig. 2, are consistent with
gies for potentials yielding more ordered liquid structures. experimental equilibrium-shape measurements for Al alloys
[6–8]. The fact that all fcc systems modeled to date have been
2.1.2. Crystalline anisotropies found to yield positive values for e1 and negative (or statis-
The importance of c anisotropy for the evolution of tically insignificant) values of e2, respectively, for a wide
solidification microstructures has long been appreciated, variety of interaction potentials suggests that this feature
particularly in the context of dendritic solidification. In of fcc c anisotropy, and the inherent competition between
metals and alloys, crystal–melt interfaces are typically h1 0 0i and h1 1 0i dendrite growth it implies, are intrinsic fea-
rough at the atomic scale, with interfacial free energies that tures dictated primarily by the symmetry of the crystal.
vary smoothly with the orientation of the interface normal Solvability theory and phase-field simulations [50] (see
(^
n). Historically, the orientation dependence of c in cubic below) have established that, in the space of e1 vs. e2 shown
metals has been described with a single anisotropy term in Fig. 2, dendrite growth along the commonly observed
parameterizing the strength of the fourfold anisotropy h1 0 0i crystallographic directions is stable for large values
(see Section 3). However, atomistic calculations have estab- of e1, while large and negative values of e2 tend to favor
lished that accurate parameterizations of c(^n) for fcc metals h1 1 0i growth, with a region of ‘‘hyperbranched” structures
generally require two anisotropy parameters associated in between (see Section 3). Additions of Zn to Al have been
with both four- and sixfold anisotropy terms: shown to lead to a transition from h1 0 0i to h1 1 0i growth
with increasing Zn concentration, with hyperbranched
nÞ ¼ co ½1 þ e1 K 1 ðh1 uÞ þ e2 K 2 ðh1 uÞ
cð^ ð2Þ
morphologies at intermediate compositions [56], and these
where h and / are the spherical angles of the interface nor- results have been interpreted to originate from concentra-
mal, c0 represents the orientation-averaged interfacial free tion dependence of c anisotropies. Recent MD calculations
energy, and e1 and e2 parametrize the strength of the four- [51] for a model LJ alloy system support the assumption
a b
A B [001]
view
<100>
Degenerate 1/ S
1/ Smaxima
0.15 maxima
ε1 A
B
<100> C
0.10
0.05
<110>
0 C <110>
0 0.01 0.02 1/ Smaxima
−ε 2
Fig. 2. Orientation selection maps (a) according to the minimum stiffness criterion in the plane (e1, e2) of anisotropy parameters and illustrative spherical
plots of inverse stiffness for different values of these parameters. (b) Dendrite growth directions as a function of e1 and e2. The symbols correspond to
atomistic calculations of c-anisotropy parameters e1 and e2 obtained for fcc metals and alloys. The region denoted by thin solid lines in the lower left is
shown in more detail in the inset on the right. Adapted from Haxhimali et al. [50].
that solute effects can induce significant changes in c aniso- In the first application of atomistic simulations to the
tropies. The calculated LJ results at six temperatures are study of crystal–melt interface kinetics, Broughton, Gilmer
shown in the inset to Fig. 2; anisotropies at T1 and T6 cor- and Jackson (BGJ) [63] also concluded that the high inter-
respond to the constituent pure materials (with the same face velocities obtained in MD at low undercoolings could
anisotropies in a LJ model), while the most concentrated not be explained within a diffusion limited growth model.
alloy compositions correspond to T3 and T4. The calcu- The authors proposed a growth law that fit well their
lated concentration dependencies for e1 and e2 are seen to MD data, and implies a form for the kinetic coefficient
be significant on the scale required to induce changes in given as:
dendrite growth orientations. V TL
l¼C ð4Þ
k B T 2M
2.2. Interface kinetic properties pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where V T ¼ K B T =M is the thermal velocity, M is the
In the solidification of a pure material, the interface atomic mass and C is a constant with a value of 1.2 derived
kinetic coefficient (l = V/DT) relates the degree of interface from a fit to the MD data for the (1 0 0)-oriented LJ inter-
undercooling (DT) to the interface velocity (V). In materi- face considered by BGJ. Subsequent to the work of BGJ,
als where significant structural rearrangement is required experimental measurements of l were performed by Rod-
for crystallization from the melt, growth is generally way and Hunt [11] using the Seeback effect to measure
described within a ‘‘diffusion limited” model (e.g. [57– the temperature of solidifying/melting interfaces; values
60]), where it is assumed that the activation energy for of l = 28(8) cm s1 K1 and 33(6) cm s1 K1 were re-
atom attachment to the crystal face is comparable to that ported from experiments for melting and growth, respec-
for diffusion in the liquid phase. This model leads to an tively. These values were found to agree (within
expression for l proportional to the liquid diffusivity (D). experimental error) with Eq. (3) using the BGJ value for
Turnbull and Bagley [61] noted that solidification in pure C. Importantly, Eq. (3) gives an estimate for Pb that is
metals can occur at very high undercoolings, where D is roughly 6 times larger than the measured value. The fact
very small, and thus growth cannot be diffusion limited. that Eq. (3) can lead to a significant overestimation of l
Rather, the frequency of liquid-atom ‘‘collisions” with in metals was pointed out originally by Hoyt et al. [64]
the crystal face was suggested to be the rate-limiting factor. who reported MD calculated kinetic coefficients for Cu
An upper bound for l within this collision-limited growth and Ni that are 4–5 times smaller than this upper bound.
model is given as [62]: The BGJ simulations and Pb measurements motivated
V sL the development of a classical density-functional theory
l< ð3Þ (DFT) for l by Mikeev and Chernov (MC) [65,66]. The
k B T 2M
MC theory yields an expression for l of the form of Eq.
where Vs denotes the sound velocity and TM is the melting (4), with a value of C = 0.72 for {1 0 0} interfaces, in rea-
temperature. For elemental fcc systems results of atomistic sonable agreement with the BGJ value. An important pre-
simulations have led to refinements in the estimated magni- diction of the MC theory is that C shows sizeable
tudes of l, and have led to the identification of significant crystalline anisotropy, with l100 > l110 l111, and l100/
crystalline anisotropies in this kinetic parameter. l110 = 1.4. Mikheev and Chernov noted that their pre-
dicted anisotropy between {1 0 0} and {1 1 1} interfaces was barriers [77–80] and nucleation kinetics [80–82], including
consistent with previous LJ MD simulation results by non-steady-state effects [83,84]. Central issues addressed
Burke et al. [67]. As discussed in Section 3, kinetic anisot- in these studies have included the diffuse nature and non-
ropy of this magnitude has important implications for den- spherical geometries of the crystal nuclei, particularly at
drite growth at high undercoolings. high undercoolings. Within the framework of classical
Over the past decade MD simulations have been applied nucleation theory quantitative predictions of nucleation
extensively to simulations of kinetic coefficients for LJ sys- rates derived from MD require accounting for the temper-
tems [68–70], hard-sphere [71] and EAM metals [43]. The ature and curvature dependence of the interfacial free
most recent LJ simulations have been based on simulation energy (c) (e.g. [85,86]). The theoretical considerations dis-
approaches distinct from those employed by BGJ (for a cussed in Section 2.1 suggest that the magnitude of c
review of such methods see [72]), yet they lead to consistent should generally increase with increasing temperature (i.e.
results. In recent simulations for hard spheres [71] l for the excess entropy is negative). This dependence has been
{1 0 0}, {1 1 0} and {1 1 1} interfaces were calculated to be demonstrated explicitly in recent MD simulations and their
1.44(7), 1.10(5) and 0.64(4), respectively, in the hard-sphere
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi analyses [83,87], although different detailed forms for this
units of K B =MT M ; given that the entropy of melting for dependence have been proposed.
hard spheres is L/TM = 1.17 kB [73], these results imply a The results of atomistic simulations, and nucleation
value of C = 1.23(6) for {1 0 0} interfaces, in excellent studies in model colloidal systems [88], have motivated
agreement with the BGJ result. In Ref. [71], available efforts aimed at the development of refined classical theo-
MD results for EAM systems were scaled into hard-sphere ries for homogenous nucleation (see Ref. [85], and refer-
units, yielding values ranging between 0.84 and 1.47 for ences therein). Additionally, diffuse-interface (phase-field)
{1 0 0} interfaces; when these values are divided by the models of crystal nucleation (see Refs. [85,89–92] and refer-
entropy of melting, they imply values for C in Eq. (3) rang- ences therein) have been extensively developed over the
ing between 0.8 and 1.4. Thus, the EAM values vary by past decade and have been shown to predict the magnitude
nearly a factor of 2, depending on the details of the of nucleation barriers in good agreement with atomistic-
EAM potential and MD method used. However, the aver- simulation results for hard-sphere and LJ model systems.
aged values obtained in all MD studies to date are gener- These theories automatically account for the diffuse nature
ally consistent with the predictions of Eq. (3), with the of the solid–liquid interface and give rise to temperature-
BGJ and hard-sphere value of C 1.2 providing a good dependent interfacial free energies, in good overall agree-
‘‘rule of thumb” estimate for {1 0 0} interfaces. ment with atomistic-simulation results.
Concerning the magnitudes of the kinetic anisotropies, Recently, MD simulations have also been applied in
LJ simulations by Huitema et al. [70] yield values of l100/ studies of phase selection in undercooled melts. Primary
l110 = 1.53 and l100/l111 = 2.00. For hard spheres the cor- nucleation of metastable bcc crystals in systems with stable
responding values are 1.31(9) and 2.25(18), respectively fcc phases have been demonstrated in simulation studies of
[71]. Comparable anisotropies have been found for all of charged-colloid and LJ systems [82,93–95]; interestingly, no
the fcc EAM systems studied to date by MD [43,71], with such observations have been reported for systems with
l100/l110 ranging between 1.25 and 1.71, and l100/l111 either hard-sphere or EAM-type potentials for metals
ranging between 1.2 and 2.7. For all fcc systems studied [78,88,96]. In the past few years the first MD simulations
by atomistic simulations to date it has been found that of heterogeneous nucleation have also been reported. Tech-
l100 > l110 > l111. niques have been demonstrated for calculating the nucle-
ation barrier and contact angles related to crystal
2.3. Atomistic studies of crystal nucleation nucleation on external ‘‘seed” particles [95,97]. In these
simulations the seeds are typically introduced as small crys-
Nucleation processes play a critical role in the formation tallites with variable structure and lattice spacing; it has
and evolution of solidification microstructures, particularly been demonstrated that the strongest reduction in the
in the context of phase selection and grain refinement nucleation barrier is obtained when the seeds are isostruc-
[28,74–76]. Over the past decade atomistic simulations have tural with the nucleating solid. Further, it has also been
been applied extensively in studies of both homogeneous shown that when the seeds have a different structure than
and heterogeneous crystal nucleation from the melt. In this the stable crystal they can enhance tendencies for forming
section we very briefly review these recent applications, as metastable bcc phases.
well as parallel theoretical developments. The intent is
not to provide a comprehensive review of the topic of crys- 2.4. Outlook: atomistics of the solid–liquid interface
tal nucleation, but rather to highlight the expanding role
that atomistic simulations are having as a tool for gaining It is anticipated that atomistic simulations will continue
insights into phenomena relevant for solidification to lead to new insights into the properties of crystal–melt
microstructures. interfaces relevant to solidification microstructures. Atom-
Recent atomistic simulations of homogeneous nucle- istic simulations provide a framework for performing
ation have involved explicit calculations of the nucleation ‘‘numerical experiments” to gain insights into atomistic
mechanisms and the relative importance of ‘‘chemical” Finally, it is expected that atomistic and first-principles
interactions, atomic size mismatch and crystal structure methods will play an increasingly important role for calcu-
in governing interfacial properties. Ultimately, these lating properties of solid–solid heterophase interfaces in the
insights should impact development of refined theories context of solid-state transformations. Calculations of
and scaling relations for both equilibrium and non-equilib- anisotropic interfacial energies and free energies for coher-
rium interfacial parameters; suggested areas for future such ent and semi-coherent solid–solid interfaces have been
applications are summarized below. Applications of atom- demonstrated [107–109] within the first-principles frame-
istic methods for direct calculations of interfacial parame- work of electronic density-functional theory (an introduc-
ters for larger-scale models of specific systems can also be tion to these methods can be found in Ref. [110]). The
expected to continue, and such applications for alloys will integration of these techniques with phase-field methods
require significant efforts at the development of more quan- (e.g. [109]) offers exciting possibilities to address outstand-
titative interatomic potential models. ing issues in solid-state phase transformations, as described
Further work is required to investigate composition in Section 7 below.
effects on c. To date only one atomistic study has
addressed the concentration dependence of the magnitude 3. Pattern selection in dendritic growth
of c [51] in a model system, featuring nearly ideal solution
thermodynamic properties and zero size mismatch. Stud- During solidification of a melt, the crystal appears with
ies of systems with more complex chemical interactions different morphologies leading, at the end of the transfor-
and size effects are needed to understand the extent to mation, to a certain microstructure. The latter is at the ori-
which such factors affect c. In terms of interface kinetics, gin of the properties of the final product, hence its
further MD simulations are required to further elucidate importance in the processing chain. Morphologies or pat-
the large range of scaled values obtained for EAM sys- terns, e.g. cells [111,112] and dendrites, are shaped by the
tems. Also, while the thermal-velocity temperature depen- complex dynamical evolution of the solid–liquid interface
dence of l implied by the BGJ model and hard-sphere and are both intricate and varied. Fundamentally under-
scaling arguments has been reported to be consistent with standing which pattern is selected under given growth con-
many MD studies, a recent calculation for EAM Ni and ditions has been a challenge in solidification science for well
Fe suggested the importance of interstitial-like diffusion over half a century. During the last decade computation-
mechanisms at high undercoolings [98]; such findings ally efficient phase-field models [5,113–115] and algorithms
should be further investigated. Kinetic simulations for [116–118], in turn, have reached a mature stage where they
alloys will also be useful as a framework for assessing can be used to simulate interface dynamics on experimen-
the accuracy of alternative theories of solute trapping tally relevant length and time scales [114,115,119,120].
[59,60,99–105]. Such studies should also include investiga- Advances in the atomistic simulation methods reviewed
tions of the nature of the anisotropies in alloy kinetic in the last section have made it uniquely possible to predict
coefficients. Recent work along these lines for the congru- key parameters that control the solid–liquid interface
ent solidification can be found in Ref. [106]; further, such dynamics at both low and high growth rate.
simulations would provide useful benchmarks for future This section concentrates on the major progress made
developments of sharp-interface kinetic theories, and with respect to dendrite theory, i.e. on the selection of tip
extensions of classical DFT models to include solute velocity, dendrite shape and orientation.
effects.
The ability to observe the formation of metastable 3.1. Dendrite velocity and shape selection
phases and heterogenous nucleation in atomistic simula-
tions raises many interesting possibilities for future work. 3.1.1. Low velocity
To make contact with experimental data derived in levi- The advent of solvability theory during the last two dec-
tated samples (see reviews in Refs. [28,41,76]), studies with ades of the last century [3,4,121,122] led to the understand-
more realistic potential models for metals would be of ing that the stability parameter
interest; the investigation of the correlation between the 2Dd 0
nature of the structural order in deeply undercooled melts r¼ ð5Þ
R2 V
and tendencies related to phase selection in nucleation
would be particularly interesting. In the context of hetero- which controls the operating state of the dendrite tip is
geneous nucleation, the simulation tools that have been determined by the anisotropy of the interfacial free energy
recently demonstrated provide a framework to make con- c in the low-velocity regime where the solid–liquid interface
tact with the free-growth theory that has been developed is in quasi-thermodynamic equilibrium. Here R is the den-
in the context of grain refinement in Al alloys [75]. Central drite tip radius, V is the dendrite growth rate, D is either
questions that could be addressed with simulations sur- the thermal or solutal diffusivity, depending on whether
round the nature of the chemical and geometrical features the solidification of a pure undercooled melt or the isother-
of seed particles that can induce potent catalysis of crystal mal solidification of an alloy is considered, and, concomi-
nucleation. tantly, d0 is either the thermal or chemical capillary
length. The Ivantsov transport relation determines the Pec- ments exploit the fact that this shape has the same fourfold
let number P = RV/(2D) as a function of the tip undercool- variation as c for small e4. This remains true for the more
ing or supersaturation, while solvability theory predicts r general form of cð^nÞ, defined by Eq. (2), but the amplitude
in terms of the c anisotropy [121]. With rare exceptions of the fourfold variation of c (or the equilibrium crystal
[123], analytical and computational studies of dendrite shape) is given by cð^nÞ. This implies that the experimental
growth have traditionally assumed that this anisotropy measurements of anisotropy to date do not determine e1
can be characterized by only one parameter, traditionally and e2 independently, as required to predict r, but rather
denoted by e4, which measures the magnitude of the four- determine their combination.
fold variation of c, c = c0(1 + e4cos4h), for materials with
an underlying cubic symmetry. 3.1.2. High velocity
By now, predictions of this theory have been largely val- For large growth rates, in contrast, the departure from
idated quantitatively by phase-field simulations in two and thermodynamic equilibrium at the solid–liquid interface
three dimensions [5,116]. Comparisons with experiments, strongly influences the selection of the dendrite tip operat-
however, have been few. One reason is that anisotropy is ing state. Phase-field simulations [119,126] of rapid solidifi-
difficult to measure experimentally in metallic systems, cation in pure undercooled Ni have shown that the
although progress in this direction has been made recently magnitude and anisotropy of l computed by atomistic sim-
[6–8]. Liu et al. [124] examined an Al–4%Cu alloy under ulations yield predictions of dendrite growth rates consis-
conditions of diffusion-controlled solidification. Their mea- tent with experimental data. The magnitude of the c
surements show that the dendrite tip selection follows anisotropy has been found to have a negligible effect on
microsolvability theory. the tip operating state at these high growth rates and the
The results of atomistic simulations in metallic systems extension of solvability theory to a kinetically dominated
embodied in Eq. (2) shed light on another reason. Namely, regime [127] has yielded a good agreement with phase-field
at least two anisotropy parameters, e1 and e2, which are the simulations and experimental data [119].
amplitudes of the first two terms in a cubic harmonic Two-dimensional phase-field simulations of rapid solidi-
expansion of cð^nÞ in three dimensions, are needed to char- fication in highly undercooled pure melts with isotropic
acterize this anisotropy. Results of solvability theory calcu- interface kinetics have shown the existence of a disordered
lations and phase-field simulations for equiaxed growth in growth morphology formed by repeated tip splitting at large
a pure undercooled melt with the form of cð^nÞ given by Eq. growth rates [128]. Three-dimensional phase-field simula-
(2) (Fig. 3) show that r depends sensitively on e1 and e2 tions that vary the magnitude ek = (l100 l110)/
[125]. Thus, a precise determination of both of these (l100 + l110) of the kinetic anisotropy exhibit a similar mor-
parameters is necessary to predict the dendrite tip operat- phology when ek is decreased below a critical value that
ing state and to compare theoretical predictions to experi- depends on undercooling [119]. This growth morphology
ments. Previously, such comparisons have been made by yields a circular envelope of the solidification front, as
only considering the anisotropy parameter e4 = (c100 c110) opposed to an angular envelope for stable dendrite growth
/(c100 + c110) measured experimentally by examination of without tip splitting. This transition from an angular to a cir-
equilibrium crystal shapes in {1 0 0} planes. These measure- cular envelope with increasing growth rate is distinct from
the one associated with absolute stability in rapid alloy
solidification [118,129,130] because the interface is still
highly unstable in the regime where the envelope is circular.
This high-velocity tip-splitting morphology with a circular
envelope is also fundamentally different from the seaweed
morphology forming in a capillary-dominated growth
regime, which has been associated with an underlying
steady-state doublon [131] or triplet [132] structure with a
split tip in two and three dimensions, respectively. In partic-
ular, solvability theory shows that this high-velocity disor-
dered morphology originates from the termination of the
main dendrite branch of steady-state growth solutions for
undercoolings larger than some threshold that increases
Fig. 3. Dendrite tip selection parameter r as a function of the two with ek [119,127]. For ek 0.13 predicted from MD for pure
crystalline anisotropy parameters e1 and e2 for h1 0 0i dendrites calculated Ni, this threshold is large enough that stable dendrite growth
using solvability theory (solid and dashed lines) and phase-field simula- persists to very large undercoolings for pure Ni or Cu.
tions (symbols) [125]. Solid (dashed) lines correspond to stable (unstable)
branches of steady-state dendrite growth solutions. These branches meet
3.2. Dendrite orientation selection
at a minimum value e1 , which shifts further to the right with more negative
e2. The function e1 ðe2 Þ defines the lower limit of existence of h1 0 0i dendrite
growth solutions in Fig. 2 and is in good agreement with the h1 0 0i- Metal dendrites typically grow along directions that cor-
hyperbranched boundary determined by phase-field simulations. respond to low index crystal axes, e.g. h1 0 0i directions in
fcc systems. Perhaps for this reason, the question of how a 3.2.2. Phase-field simulations and experiments
dendrite selects its growth direction has historically Phase-field simulations of dendrite growth in a pure
received little attention. undercooled melt that use the anisotropy form of c defined
Dendrites, however, sometimes select other directions, by Eq. (2) have been carried out to explore the selection of
‘‘off crystal axes”, ranging from early observations of dendrite growth directions [50]. The results shown in Fig. 2
h2 2 4 5i directions off the basal plane and the c-axis in Mg of the atomistic section reveal that h1 0 0i and h1 1 0i den-
alloys with hexagonal crystal symmetry [133,134] to h1 1 0i drites form near the positive e1 and negative e2 axis, respec-
and h1 1 1i directions for ammonium chloride in aqueous tively, as expected from the minimum stiffness criterion.
solutions [123], and to more recent detailed observations Not predicted by this criterion, however, is a large region
of h1 1 0i, h3 2 0i, h2 1 1i and unsteady curvilinear dendrite sandwiched between the h1 0 0i and h1 1 0i regions in
paths in fcc Al-based alloys [135–140]; anomalous growth Fig. 2 where the dendrite growth directions vary continu-
directions can be of practical interest as in the example of ously from h1 0 0i to h1 1 0i, as illustrated in the top panel
‘‘feathery crystals” in Al alloys that lead to highly aniso- of Fig. 4. Cubic symmetry dictates that these misoriented
tropic microstructures with undesirable mechanical proper- dendrites have 24 primary branches off crystal axes. These
ties. These observations have generated renewed interest in ‘‘hyperbranched” structures occupy a large region of
the question of how a dendrite selects its growth direction.
In turn, the expanded anisotropy parameter space of Eq.
(2), derived from atomistic simulations, has set the stage
for exploring this question theoretically in the experimen-
tally relevant low-velocity regime where dendrite growth is
dominated by capillary anisotropy.
3.2.1. Extremum stiffness criterion

Dendrite growth directions have traditionally been
assumed to correspond to maxima of c. In the simplest
two-dimensional situation, where c = c0(1 + e4cos4h), these
maxima also correspond to minima of the interface stiff-
ness, S = c + d2c/dh2, which appears in the Gibbs–Thom-
son condition. This assumption is physically motivated
by the fact that capillary forces are least effective at
smoothing out protrusions of the solid–liquid interface in
directions where the stiffness is minimum. In three dimen-
sions, the interface stiffness becomes a tensor quantity,
but the trace of this tensor for a spherical crystal seed
@2c 1 @2c @c
S ¼ 2c þ 2
þ 2 þ coth ð6Þ
@h sin h @u2 @h
can be used similarly to formulate an extremum criterion
for dendrite growth directions [50,125]. As in the previous
section, h and u are the spherical angles of the interface
normal. An analytical calculation of the stiffness S defined
by Eq. (6) for the form of c defined by Eq. (2) reveals that
minima of S (maxima of 1/S in Fig. 2(a)) for cubic crystals
correspond either to h1 0 0i directions or h1 1 0i directions.
In the plane of anisotropy parameters of Fig. 2(a), the re-
gion of this plane where h1 0 0i dendrites are predicted to
form lies close to the positive e1 axis, while the h1 1 0i region
lies close to the negative e2 axis. These two regions are sep-
arated by a boundary (e1 = 20e2/3) where S is minimum
for all directions contained in any of the {1 0 0} planes. Fig. 4. Phase-field simulations of equiaxed growth (a) and experimental
Therefore, using this physically motivated but ad hoc min- results of dendrite orientation in directionally solidified Al–Zn alloys (b),
imum stiffness criterion, one would conclude that dendrites showing a continuous change of dendrite growth misorientation / as a
select either h1 0 0i or h1 1 0i directions and, exactly on the function of anisotropy parameter, e1, and Zn composition, respectively
[50]. The curve in (b) is an interpolation through the experimental
boundary between the h1 0 0i and h1 1 0i regions, some measurements (dots and squares). DS, directionally solidified by chilling,
unspecified directions contained in the set of {1 0 0} planes, 3 > V > 0.5 mm s1, 3 > G > 0 K mm1; BS, Bridgman solidification,
as shown in Fig. 2(a). V = 67 lm s1 and G = 10 K mm1.
parameter space, while the h1 1 0i dendrites are constrained relationship, the change from an angular to a spherical
to a narrow wedge above the negative e2 axis. envelope of the solidification front below the absolute sta-
The MD data for several pure fcc metals fall strikingly bility limit and grain refinement at large undercooling.
close to the h1 0 0i-hyperbranched boundary in Fig. 2. This The growth of partially faceted dendrite growth struc-
suggests that changes of anisotropy parameters resulting tures remains comparatively less understood than the
from solute addition are the underlying mechanism for growth of non-faceted dendrites reviewed in this section,
atypical dendrite growth directions observed in fcc metallic even though phase-field simulations of faceted growth have
alloys [56]. Results of directional experiments in Al–Zn started to be carried out [144,145]. In addition, recent pro-
alloys that show a continuous change of dendrite misorien- gress has been made to model the columnar growth of
tation as a function of Zn composition (bottom panel of more complex cellular/dendritic array structures during
Fig. 4) and recent MD simulations in a binary alloy that directional solidification using two-dimensional phase-field
show that alloying shifts the anisotropy parameters simulations [112,146]. This work identified scaling laws
towards the h1 0 0i-hyperbranched boundary both support that relate different microstructural length scales, including
this conjecture. the primary dendrite array spacing. However, extension of
A new experimental technique should also be men- these simulations to three dimensions remains a major
tioned, as it allows the use of instant pressure changes to challenge even on today’s computers. It is also a challenge
study transient and non-steady-state effects in dendritic to model the slow coarsening evolution of dendritic struc-
growth, particularly in the tip selection process [141]. The- tures in the mushy zone. Recent progress in this direction
oretical and experimental studies of non-steady-state by a combination of phase-field simulations and experi-
growth transients have also been carried out that provide ments by Voorhees and co-workers is reviewed in
new insights into transitions between globular and den- Section 6.
dritic microstructures in both equiaxed alloy solidification The study of pattern selection in multicomponent alloys
[142] and semi-solid processing [143]. Related progress in requires further work. Alloy phase-field models can in prin-
understanding transitions between columnar and equiaxed ciple be formulated for an arbitrary number of compo-
microstructures is reviewed in Section 5.2. nents, and such models have been used to simulate both
eutectic [147] and dendrite growth [148] in ternary alloys.
3.3. Outlook: pattern selection in dendritic growth However, formulating models with computationally tracta-
ble thin interface limits to achieve quantitative results
An accurate quantitative determination of c anisotropy remains a major challenge for more than two components
in Al–Zn and other alloys using either MD simulations and/or nonlinear phase coexistence curves. Recent progress
or equilibrium-shape measurements in multiple planes is to formulate computationally efficient phase-field formula-
critically needed to quantify the parameters e1 and e2, as tions for binary alloys with nonlinear phase coexistence
well as to investigate whether higher-order cubic harmonics curves shows promise in this direction [149].
are needed to represent the c anisotropy. In particular, the Finally, recent progress has been made to model com-
next higher-order term in the cubic harmonic expansion of plex dendritic microstructures in polycrystalline materials
c in Eq. (2) generates an eightfold cos8h variation in {1 0 0} using a classical phase-field approach with a scalar order
planes that can influence orientation selection because the parameter that represents the local grain orientation
combination of four- and eightfold variations in these [150,151]. The phase-field crystal (PFC) approach has also
planes produces stiffness minima in directions off crystal recently emerged as a powerful computational approach
axes [56,125]. In addition to the characterization of crystal- for multiscale modeling of polycrystalline microstructural
line anisotropy, these results bring new challenges for den- evolution [152]. By construct, this method resolves the
dritic growth theory. Fig. 3 shows that this theory can atomic-scale density wave structure of the crystalline mate-
predict the limit of existence of h1 0 0i dendrites in rial and can describe its evolution on time scales orders of
Fig. 2(b). This theory, however, is presently only developed magnitude longer than molecular dynamics simulations.
to predict independently the tip operating state parameter This method has recently been shown to describe well the
r and the growth misorientation in two dimensions. How equilibrium properties of the crystal–melt interface for
to extend this prediction to three dimensions is yet to be the case of pure Fe [153]. In a broader solidification con-
elucidated and appears to be a particularly challenging the- text, it offers a new avenue to tackle complex multiscale
oretical problem. problems where crystalline defects play a key role, from
Both MD simulations that characterize the departure the formation of complex structures such as twinned den-
from chemical equilibrium at the interface and its anisot- drites to grain coalescence during the late stages of
ropy, and three-dimensional phase-field simulations that solidification.
incorporate this knowledge and simultaneously resolve
the thermal and solutal diffusion fields, are still needed. 4. Pattern selection in coupled growth
This would lead to the understanding of the fundamental
relationships between different rapid solidification phe- The most widespread solidification microstructures
nomena, including the break in the velocity–undercooling besides dendrites are the composites formed in eutectic
alloys. In purely metallic (regular) systems the eutectic tional to the local gradient of spacing [155]. This effect
phases generally form in a coupled manner at the solidifica- always smoothes out inhomogeneities along the front, such
tion front, which is flat on a scale that is large, compared to that a lamellar front (in eutectics or in peritectics; see
the microstructural spacing. In irregular eutectics the below) can be stable over a limited range of spacing even
growth front is non-isothermal, but a certain coupling does when the front temperature vs. spacing curve has a positive
exist. In the last decade, major progress on the understand- slope (i.e. dT/dk > 0; Fig. 5(b)), as long as this slope
ing of pattern selection in coupled growth has been made remains below a critical value that increases with the tem-
thanks to the development of new experimental techniques perature gradient. As a consequence, the critical spacing
and efficient phase-field models for multi-phase growth. for the onset of the lamella elimination instability can be
Here, we concentrate on the influence of tri-junction substantially smaller than the minimum undercooling spac-
motion and interfacial properties on pattern selection, ing, depending on the temperature gradient and the growth
and the complex three-dimensional structure of eutectic speed. Even though the lateral tri-junction sliding is diffi-
composites and the associated instabilities and spacing cult to observe experimentally, the excellent agreement of
adjustment mechanisms. Microstructures in multicompo- the stability thresholds between simulations and experi-
nent alloys and peritectic coupled growth will also be ments strongly supports its existence.
briefly discussed. This section is divided into regular and
irregular growth. 4.1.2. Binary eutectic growth
Concerning the instabilities at large spacings, major
4.1. Regular eutectic and peritectic growth recent advances have come from the study of morpholog-
ical stability and spacing adjustment mechanisms in
4.1.1. Stability of coupled growth ‘‘thick” samples which contain a sufficient number of
A central question for the understanding of pattern lamellae for the dynamics to be three-dimensional but
selection in coupled growth is to identify and characterize remain small enough to exhibit only weak convection.
the dynamic instabilities that limit the range of stable For example, directional solidification experiments on
(and hence observable) lamellae or rod spacings [154]. the Al–Cu eutectic have been carried out in capillary tubes
For lamellar growth, new results concerning the instabili- [158], and the evolution of the eutectic spacing was exam-
ties both at small and at large spacings have been estab- ined by micro-milling and by the reconstruction of the 3D
lished. In a combined experimental and numerical study microstructure [159]. Furthermore, an experimental setup
[155], the lamella elimination instability that occurs at using a long-distance microscope in oblique view was
small spacings was investigated in detail. It was found in developed to observe the dynamics of growth fronts in
phase-field simulations that tri-junction points do not bulk samples of transparent alloys in situ [160–162]. In
move only normal to the growth front, as traditionally the Al–Cu eutectic, it was found that the ratio of the max-
assumed [156,157]: they can also slide parallel to the front, imum to the minimum spacing was approximately 1.2. In
as illustrated in Fig. 5(a), with a velocity that is propor- this alloy, a new instability was also discovered: after a
sudden change of the growth velocity by a factor of 4,
an initial bump on the lateral surface of a lamella can
a b develop and then grow parallel to the network of lamellae.
eutectic
z peritectic This mechanism, comparable to dendrite branching (i.e.
TE x secondary dendrite arms emitting tertiaries which can then
ζ(x) Liquid T(λ)
become new primary trunks), reduces the spacing (Fig. 6
[159]). In the transparent alloy, a zigzag instability occurs
β β TE, TP stable
α α λ above a critical spacing [160,161], which leads from
β straight to wavy lamellar arrays (Fig. 7). Three-dimen-
stable
sional phase-field simulations [163] have confirmed the lat-
Fig. 5. Schematic illustrations of (a) the sliding motion of three-phase ter result. In addition, these simulations have shown that
junctions (tri-junctions) and (b) the crucial influence of this motion on the the analogues of the oscillatory instabilities known in thin
stability of both eutectic and peritectic coupled growth. In (a), the samples do exist, but that they are not observable since
envelope of the composite interface is shown as the dashed line that passes they always occur for larger spacings than the zigzag
smoothly through tri-junctions. The displacement of this line along the z-
instability.
axis measured from the eutectic temperature (at z = 0) is f(x). The blue
arrows depict the motion of the tri-junctions of the central b lamella For rods, only preliminary results on stability are avail-
normal to this envelope. Consideration of this motion alone (Jackson– able [162,164,165], and these indicate that the main spacing
Hunt–Cahn assumption) leads to the prediction that coupled growth is adjustment mechanisms are rod elimination for small spac-
unstable for dT/dk > 0. The red arrows indicate the lateral motion of the ings and rod splitting for large spacings, and that the acti-
junctions in the direction of increasing spacing. This additional motion
vation of these mechanisms may strongly depend on the
can stabilize both eutectic and peritectic coupled growth for a finite range
of spacing where dT/dk > 0, as shown in (b) [154]. (For interpretation of geometry of the experiment and the resulting boundary
color mentioned in this figure the reader is referred to the web version of conditions. In both experiments and simulations it was
the article.) observed that the lamella-to-rod transition is not sharp
behavior [147,169]. A clear influence of crystallographic

effects on the shape of well-developed colonies has also
been reported [169–171]. Ternary-coupled eutectic growth
(involving three distinct solid phases) was investigated both
in metallic [172] and organic alloys [173–175]. A bewilder-
ing variety of microstructures was observed, which opens a
completely new field of study. The variety was even greater
for quarternary eutectics [176]. There is little recent theo-
retical work in this area, except a generalization of the
Jackson–Hunt calculation to ternary composites [177].
For a more detailed review on solidification in multicom-
ponent alloys, see Ref. [178].
4.1.4. Peritectic coupled growth

The question whether peritectic alloys can exhibit eutec-
tic-like coupled growth has been discussed for a long time,
but a definite answer has remained elusive. Observations of
growth in peritectic systems have remained rather sparse
until recently [179,180]; rather, mostly other microstruc-
tures, such as discrete bands, oscillatory structures and dis-
Fig. 6. Successive transverse sections of a directionnally solidified eutectic ordered composites, have been found [181–185]. The
Al–33.2 wt.% Cu alloy, showing a mechanism for lamella creation in three situation has now been considerably clarified. First, it
dimensions (average eutectic spacing = 4.8 lm). After an increase of the was established that convection has a strong influence on
growth speed from V = 1.25 to V = 5 lm s1, a perturbation normal to
the lamella (marked ‘‘A”) grows and develops into a new lamella which
microstructure formation in peritectic growth (see also Sec-
propagates along the pre-existing ones [159]. tion 5). Secondly, it was recognized that the lateral tri-junc-
tion motion discussed above for eutectics is also present in
peritectics and can lead to stable coupled growth even
but occurs gradually over a range of compositions (or vol- when the slope of the front temperature vs. spacing curve
ume fractions) [166]. is positive, which is often the case in peritectics. Further-
Finally, an investigation of the stability of eutectic more, a detailed experimental and numerical study of
growth at high velocities was recently carried out for Al– two-phase microstructure selection in the Fe–Ni system
Sm alloys [167]. This study showed that, when the temper- (Fig. 8 [186,187]) was carried out. The main results are
ature dependence of the solute diffusivity in the liquid is (Fig. 9): (i) stable coupled growth in Fe–Ni can occur only
taken into account, there exists a maximum undercooling above a critical temperature gradient where both phases
above which a low solubility eutectic will not be stable are morphologically stable and (ii) the range of stable spac-
[168], and this undercooling does not vary with composi- ing strongly depends on composition and is limited for
tion. These results are relevant for the eutectic-to-glass both large and small spacings by oscillatory instabilities.
transition. Recently, peritectic coupled growth was also observed in
the Cu–Sn alloy [188].
4.1.3. Multicomponent alloys
Recently, a whole series of studies has investigated 4.1.5. Initiation of coupled growth
eutectic growth in multicomponent systems. For eutectics A question of great interest is how coupled growth
with a small amount of ternary impurities, eutectic colonies develops from a single solid primary phase. Recent experi-
form. The instability leading to colony formation was mental and phase-field modeling studies have shed light on
found to be similar to the Mullins–Sekerka instability, two qualitatively different mechanisms of initiation
but with the difference that it can exhibit an oscillatory [186,188,189]. The first, which applies to both eutectic
Fig. 7. ‘‘Zig-zag” structures in a near-eutectic transparent eutectic alloy CBr4–C2Cl6. Left: top view of an in situ experiment using an optical long-distance
microscope in oblique view and subsequent image processing, G = 100 K cm1, V = 0.5 lm s1 (the growth direction is normal to the plane of the page,
the average spacing = 20 lm) [161]. Right: snapshot of a 3D phase-field simulation of a similar model system [163]. Growth direction is upward.
growth of both phases [188]. The second mechanism, which

has only been observed so far in the peritectic Fe–Ni sys-
tem, involves nucleation of the secondary phase at multiple
sites on the primary phase. The latter can initiate coupled
growth if the spacing between nuclei falls inside the stable
range of coupled growth spacing, or an island banding
microstructure often seen as a transient leading to coupled
growth [186].
4.2. Irregular eutectic growth
Irregular eutectic structures are developed when a non-

faceted (nf) phase is coupled with a faceted (f) phase, as
observed most notably in the Fe–C(graphite) and Al–Si
systems. In such eutectics, local morphological adjustment
of interphase spacing is severely encumbered by the limited
branching ability of the highly anisotropic faceted phase
containing planar defects. Thus, the spatially non-uniform
Fig. 8. Peritectic coupled growth of lamellar and fibrous d and c in a or irregular structure that evolves during f–nf eutectic
directionally solidified specimen of Fe–4.1 at.% Ni (G = 18 K mm1,
solidification is inherently three-dimensional, where the
V = 10 lm s1). Longitudinal (a) and transverse (b) micrographs. The
microstructure is correlated with the local Ni composition, which increases relationship between the growth mechanisms of the faceted
in radial direction due to macrosegregation [187]. phase and the complex non-isothermal interface structure
gives rise to a more diverse range of solidification micro-
structures than that exhibited by regular eutectics and to
a number of growth mode transitions involving both intrin-
sic and extrinsic mechanisms of nucleation and growth.
From a theoretical standpoint, only limited progress
has been made over the last decade toward understanding
the dynamics of irregular eutectic growth, and new devel-
opments remain constrained to two-dimensional (2D)
descriptions which are inherently limited. Most recently,
a fourth-order shape function was used [190] to extend
the analytical MK model [191] (which employs a cubic
function) to allow for the deformation of the phases in
a non-isothermal coupled growth configuration. Accord-
ingly, both the protrusion of the leading (faceted) phase
and the depression of the non-faceted phase are parame-
terized with the assumed shape function and related to
the contact angles for the respective phases at the triple
junction. While this modification to the MK theory
addresses an important feature of the non-isothermal
Fig. 9. Microstructure selection map for steady-state directionally solid-
interface, it appears that significant advancement can only
ified peritectic Fe–Ni alloys with plane front coupled growth of fibrous
and lamellar structures in the center, oscillatory structures on the left and arise from more descriptive 3D parameterization of the
cellular composites at low G/V values [187]. irregular growth front coupled with models that account
for the variation in local structure, the role of crystallog-
raphy and crystal defects, the simultaneous operation of
and peritectic alloys, involves a growth instability occur- multiple nucleation and growth mechanisms, and the
ring during spreading of the second phase on the primary competition that leads to various observed growth mode
phase. This mechanism was studied experimentally in a transitions.
transparent eutectic alloy in thin samples [189]. After Focused ion beam tomography has recently been used
nucleation of the second phase, a ‘‘finger” develops which by Lasagni et al. [192] for high-resolution investigation of
propagates laterally along the solid–liquid interface. Above the 3D structure of Sr-modified and unmodified Al–Si
a certain velocity, it develops an oscillatory instability, alloys (Fig. 10), and volume-based structural parameters
which leads to the emergence of very fine lamellae. This have been used to identify the flake/fiber transition, which
is followed by numerous lamella elimination events until is not indicated through conventional spacing measure-
a stable state is reached. Similarly, peritectic coupled ments [193]. These efforts to understand the 3D nature of
growth in Cu–Sn alloys was initiated by repeated sideways irregular eutectic growth are limited in number, however,
Fig. 10. Microstructure of an unmodified Al–12 wt.% Si alloy. (a) Metallographic image and (b) 3D reconstruction of the deep-etched Si structure [192].
and several intrinsic and extrinsic growth mode transitions Al9Si2Fe2 in the solute boundary layer just ahead of the
in irregular eutectic systems remain poorly understood. primary aluminum phase stimulates nucleation of eutectic
We focus here on Al–Si as the prototypical irregular silicon and serves to initiate the eutectic reaction. They
eutectic alloy and examine recent investigations into intrin- show further that, for Fe levels low enough (0.0032
sic growth morphologies and impurity effects leading to wt.%) to preclude Al9Si2Fe2 nucleation, a refined Si struc-
microstructural modification. Recent investigations of the ture is observed [202], presumably due to the increased
intrinsic high-velocity flake/fiber (sometimes called quench undercooling associated with more sluggish nucleation.
modification) transition in the binary Al–Si eutectic have These assertions have generated some controversy [202–
indicated that this transition occurs over a rather wide tem- 205] and, while the role of the Al9Si2Fe2 phase (compared
perature range and in several stages. Observation of in- to the role of other phases, such as AlP) remains unre-
plane fingering of flakes, flake-like packets of rods and solved, trace levels of Fe appear to have a distinct effect
out-of-plane rod growth suggest that anisotropic interfacial on the eutectic structure.
properties are critical here [193], emphasizing the need for
experimental and atomistic determination (see Section 2) of 4.3. Outlook: pattern selection in coupled growth
these properties. Low-velocity transitions to angular script-
like and other morphologies have also been associated with Clearly, the further study of the dynamics of 3D pat-
this anisotropy and related to a twin-enhanced growth terns of lamellar and fibrous eutectics is critical to under-
mechanism [194]. At low velocities, the script-like structure stand the mechanisms leading to spacing selection. In
gives way to the strong selection of a h0 0 1i textured, fac- particular, the role of topological defects (lamella termina-
eted and twinned silicon morphology, which can be attrib- tions, lamellar fault lines, grain boundaries) and the effects
uted to the effective kinetic operation of a complex of external forces (convection, lateral temperature gradi-
structure of coherent {3 1 0} and {2 1 0} twins in the silicon ents), crystalline anisotropy and thermal stresses develop-
phase [195]. ing at the ab interfaces need to be further clarified. The
Because the competition between the available mecha- effect of a third component on the stability of eutectics
nisms of nucleation and growth are central to the selection needs to be investigated in off-eutectic compositions in
of f–nf morphologies, solute elements, even in trace which one of the phases is a compound, and a theoretical
amounts, can play a critical role in altering the observed model that includes contributions of interface energy and
dynamics. In the case of Al–Si, recent results bring into interface kinetics for the compound phase needs to be
question whether the influence of impurities on growth developed. One open question in this area pertains to the
mechanisms can ever be completely avoided in practice. mechanism that sets the eutectic colony size. The stability
It has been reported [196] that, in addition to the well- of eutectic microstructures at high growth rate has also
known modification effect, Sr additions may alter the been examined for systems in which the solute-partitioning
mechanism by which eutectic grains nucleate. Indeed, three coefficient is small. This treatment needs to be extended to
distinct mechanisms for silicon nucleation have been the case in which non-equilibrium effects are present at the
reported, each at different Sr levels [197]. Subsequent find- interface.
ings [198–200] suggest that Sr may poison the catalytic Microstructure and phase competition (e.g. from eutec-
effects of other impurities, thereby suppressing Si nucle- tic to dendrites, from eutectic to glass) are other subjects of
ation and increasing the eutectic grain size in commercial great practical and scientific interest. For multi-phase
alloys. It has also been proposed by Shankar et al. [201] eutectic coupled growth, a classification of the observed
that trace amounts of Fe (>30 ppm, which is commonly microstructures needs to be established and microstructure
observed even in high purity Al) may have a strong influ- selection maps need to be developed. For irregular eutec-
ence on eutectic growth morphology. Examining this issue tics, a better understanding of the 3D morphology and
in hypoeutectic Al–Si alloys, they claim that nucleation of associated chemical and thermal fields is required. In addi-
tion, further investigation into the faceted phase growth mation. The last decade has seen a flurry of activities in
mechanisms in f–nf (irregular) eutectics and the role of the area of convection–microstructure interaction. The
crystal defects, such as twin boundaries, is needed. The role research has benefited not only from novel experiments,
of anisotropy on the maximum stable spacing needs to be but even more so from the recently emerged capability
established theoretically and experimentally. to perform direct numerical simulations of solidification
Since the selection of peritectic structures strongly with convection on a microstructural scale. The following
depends on the dynamics of nucleation and growth, theo- subsections provide an overview of recent research on
retical models are required to precisely predict the selection convection effects on various solidification morphologies,
of microstructures. These studies, however, have been such as plane front, dendrites, eutectics, peritectics and
undertaken in systems in which some convection effects monotectics.
were present. Experimental studies under purely diffusive
growth conditions are required to unambiguously validate 5.1. Coupled morphological and convective instabilities
the model that is based on diffusive growth only.
Progress toward these goals can be made by combining Fluid flow along a solid–liquid interface induces mor-
new experimental and simulation techniques. In particular, phological instabilities. Such instabilities have been ana-
phase-field modeling is expected to be instrumental in con- lyzed theoretically for the case of directional solidification
necting interfacial mechanisms with local 3D microstruc- [211–213]. It has been found that the solidification front
tural evolution. However, progress on phase-field exhibits slow, spatio-temporal dynamics when the wave-
treatments of f–nf interfaces is needed: such models are length of the cellular convection is far longer than that of
available for single-phase solidification only [206–208], the morphological cells. When the flow is strong, the mor-
and have remained restricted to linear interface kinetics. phological instability becomes spatially localized and con-
Atomistic simulations such as those described in Section fined at inward flow-stagnation regions on the front.
2 are certainly needed to obtain quantitative information Localized microstructures induced by fluid flow in direc-
about interface kinetics of faceted interfaces, such that ulti- tional solidification were also observed experimentally in
mately faceted interfaces with multiple growth mechanisms both transparent and metallic alloys [214–216]. The real-
and nonlinear kinetics can be incorporated into phase-field time observations revealed two types of localized patterns
models of f–nf eutectic growth. of a generic nature: focus type and outward hexagonal flow
These fundamental issues should be addressed if our type [214]. A delay of morphological instability, due to sol-
understanding of chemical modification, mechanical vibra- utal convection, was observed downstream of well-oriented
tions and applied electric or magnetic fields are to be effec- solid ridges [216].
tively utilized to develop new alloys and/or growth Coupled convective and morphological instabilities dur-
morphologies offering enhanced properties. For example, ing directional solidification of an alloy have also been sim-
it would be highly interesting to find a way for the produc- ulated using the phase-field method [217–219]. The
tion of a truly spherulitic Si morphology in Al–Si castings, simulations revealed the importance of lateral solute segre-
in analogy with ductile (nodular) cast iron. In this context, gation, resulting from the gravity-induced flow, on the sta-
a paper on detailed phase-field modeling of spherulites is bility of the interface. For the cases slightly above the onset
relevant [210]. of instability, the morphologies of the shallow cells are
Finally, an interesting topic for further studies is the affected by the convection as well. Cells with different
relation between eutectic growth (mostly of f–nf type) wavelengths and depths coexist due to flow-induced segre-
and metallic glass formation. When the partition coeffi- gation. The unique feature of these phase-field simulations
cients of both eutectic phases are very small, when the is that they show the full coupling of long (convective) and
eutectic temperature is low and the glass transition temper- short (morphological) wavelength instability modes
ature is high, a eutectic-to-glass transition is often observed beyond the linear stability limit. The planar to cellular
at high solidification rates. While considerable effort has transition during upward directional solidification was
been directed toward understanding and predicting glass- investigated experimentally using a transparent succinonit-
forming ability in metallic alloys [209], this topic has not rile–acetone alloy [220]. Buoyancy-driven melt convection
generally been treated within the context of growth. can significantly stabilize the planar morphology. At the
Clearly, a better understanding of nucleation and growth plane-to-cell transition, quasi-hexagonal and elongated
of eutectics competing with glass formation would be help- cells exist simultaneously, both at grain boundaries as well
ful in the search for optimal glass-forming alloys. as inside the grains. With increasing solidification length, a
quasi-hexagonal pattern of randomly rearranging cells cov-
5. Microstructure and convection ers the interface. The cell size varies over a wide range and
no wavelength selection was found. These and other com-
The work presented in the earlier sections was cen- plexities still do not allow for a complete and unified under-
tered on diffusional phase transformations. In real situa- standing of coupled morphological and convective
tions, however, convection is omnipresent and often instabilities. From a numerical simulation point of view,
plays a dominant role in solidification microstructure for- the simultaneous presence of short and long wavelength
instabilities represents a major challenge still to be [248]. The effects of convection on dendritic growth in the
conquered. high tip velocity regime and on dendrite orientation selec-
tion, as reviewed in Section 3, also still need to be investi-
5.2. Convection effects on dendritic growth gated in more detail.
Dendrite growth in a convective environment during
5.2.1. Dendrite tip growth directional solidification of binary alloys has been modeled
While tip kinetics during free dendritic growth into a phenomenologically by taking into account the local
quiescent undercooled melt are reasonably well under- growth conditions (V, G and local composition) and an
stood, the elucidation of the well-documented effects of effective diffusion coefficient that depends on the confine-
melt convection [221,222] has remained an important chal- ment (e.g. specimen diameter) [124,251]. The stability
lenge. Phase-field models have been developed [223–225] parameter r was found to be the same for solidification
and widely used to simulate solidification in the presence with or without convection only when the effective diffu-
of fluid flow [226–236]. Other methods, such as the sion coefficient was used to compute r in the case with con-
sharp-interface tracking technique [237–239], lattice Boltz- vection. It is possible that the dendrite tip operating state is
mann models [240–243], cellular automaton techniques influenced by the overlap of diffusion fields from nearby
[244–246] and Monte Carlo simulations [247], have also dendrites in directional growth, and thus is quantitatively
become available. different from that for free growth. However, this differ-
Quantitative phase-field simulations of free dendritic ence, if present, has not been systematically investigated.
growth in a pure undercooled melt have been carried out Dendrite tip growth during settling of equiaxed grains in
in three dimensions [229,233]. Fig. 11 shows a simulated an undercooled melt was investigated in several experimen-
3D morphology at the limit of current computational capa- tal studies involving transparent alloys [252–254]. Measure-
bilities. These studies have shown that fluid flow modifies ments were performed to investigate the effect of the flow
the convective heat transport away from the dendrite tip (or settling) velocity and of the angle of the flow with
in good agreement with an analytical solution [233,248] respect to the dendrite growth direction [254]. While the
that replaces the Ivantsov diffusive transport relation men- dendrite tips pointing into the flow generally grow at a
tioned in Section 3; modifications of the Ivantsov transport velocity that is greater than the value predicted by the stan-
theory have also been developed based on boundary layer dard diffusion theory, the dendrite tips in the wake of the
or stagnant film type models of the effects of natural con- crystals grow at a velocity that is substantially less than
vection on the heat and solute transport at the dendrite in the absence of flow. When the flow is at an angle normal
tip [249,250]. For the range of flow velocities and melt to the dendrite axis (90°), the tip growth velocity is unaf-
undercoolings investigated, and in the limit of vanishing fected by the flow. Interestingly, the average growth veloc-
interface kinetic effects, the dendrite tip selection parameter ity of all six primary dendrite arms of a settling equiaxed
r was found to be unaffected by the flow. This finding is crystal was found to be in almost perfect agreement with
consistent with solvability theory [248], which predicts that the value predicted by the standard diffusion theory,
r is only altered when the flow velocity is much larger than despite the presence of considerable settling speed changes
the dendrite growth rate. Additional advances in numerical and crystal rotation.
techniques are needed to explore higher flow velocities
where r is predicted to deviate from the diffusion value 5.2.2. Dendrite arm development and spacings
The effect of convection on the columnar (primary) den-
drite trunk or cell spacing in directional solidification has
also emerged as a subject of strong research interest.
Numerous experimental studies have been performed
where the flow is induced by thermal and/or solutal buoy-
ancy forces [255–263]. While these experimental investiga-
tions reveal several interesting phenomena, few general
statements regarding the convection effect can be made.
The primary spacing either increased or decreased com-
pared to purely diffusive conditions (as in a zero gravity
environment). This is related to the complex nature of
the natural convection: in some alloys, the rejected solute
causes an increase in the melt density, while in others it
causes a decrease. This results in different microscopic con-
vection patterns between the dendrite arms near the tips.
Fig. 11. Three-dimensional phase-field simulation of free dendritic growth
The direction of the gravity vector relative to the solidifica-
of a pure substance into an undercooled melt flowing around the crystal
(from left to right); in the simulation, the crystal was held fixed in order to tion direction and the existence of lateral thermal gradients
study the effect of the flow relative to the crystal (figure courtesy of C. also play important roles. Macroscopic convection on the
Beckermann, University of Iowa [233]). scale of the crucible induces long-range variations in the
solute concentration that affect the primary spacing as well. [235,266,267] and the selection of columnar grains in the
Often, convection is unsteady or even oscillatory, and presence of a shear flow [236]. While these numerical stud-
steady-state growth conditions and stable arm spacings ies represent key contributions to the understanding of
are never achieved. convection effects in directional solidification, additional
The complex nature of the effects of thermo-solutal nat- work remains to be performed. In particular, these studies
ural convection on the primary dendrite (trunk) spacing is need to be extended to three dimensions and more realistic
well illustrated by the phase-field simulations of Steinbach growth conditions.
and co-workers [264,265] for directional solidification of an Extensive experimental studies on the effect of convec-
Al–4% Cu alloy. As shown in Fig. 12(a), when the dendrite tion on the dendrite morphology during directional solidi-
array is solidified upward, segregated Cu-rich melt of a fication of alloys have been performed by using magnetic
higher density flows downward along the dendrite trunks fields to induce a flow of known strength [268–271]. As
and accumulates in the space between the primary arms. shown in Fig. 13(a), the application of magnetic fields of
This solute accumulation by the flow causes an increase different strengths allowed for the validation of a relation
in the primary arm spacing to 600 lm, compared to [272,273] that predicts the primary dendrite arm spacing
330 lm under purely diffusive conditions. The flow is con- as a function of the flow velocity. Here, the flow induced
fined to the interdendritic region and is relatively weak, by the rotating magnetic field causes the primary dendrite
since the thermal gradient ahead of the growth front arm spacing to decrease. In addition, these experiments
induces stable density stratification. On the other hand, confirmed previous phase-field simulations [274] of the
during downward solidification (Fig. 12(b)), strong and effect of convection on the secondary dendrite arm spacing.
highly unsteady convection exists in front of the growth The interdendritic flow induced by the magnetic field
front, since this configuration is both thermally and solu- increases the secondary spacing, and the regular coarsening
tally unstable. This flow washes away the rejected solute exponent of 1/3 in the absence of flow increases to 1/2
between the dendrite trunks and causes a decrease in the (Fig. 13(b)). More complex microstructural effects are pres-
primary spacing to 270 lm. Other phase-field simulation ent at very high magnetic field strengths (up to 10 T)
studies have investigated the inclination of dendrite arms [275,276]. Depending on the growth rate, the magnetic field
Fig. 12. Two-dimensional phase-field simulations of the effect of thermo-solutal convection on the primary dendrite arm spacing in an Al–4% Cu alloy
directionally solidifying with a thermal gradient of 10 K mm1 and a velocity of 40 mm s1; (a) upward solidification and (b) downward solidification. The
grey shades indicate that the solute concentration in the melt varied between 4% Cu (dark) and 7% Cu (light) (figure courtesy of I. Steinbach, ACCESS,
Aachen [265]).
Fig. 13. Effect of a rotating magnetic field on (a) the primary and (b) the secondary dendrite arm spacings in a directionally solidified A357 aluminum
alloy (figure courtesy of L. Ratke and S. Steinbach, DLR, Cologne [270,271]).
causes tilting, rotation and general disorder in the dendrite [287–289]. They generally predict an increase in the lamel-
pattern, an increase in the primary dendrite arm spacing lar spacing with convection intensity. However, the effect
along with an enhanced side-branch structure and even a of buoyancy-driven convection, or its absence in micro-
transition to an equiaxed grain structure. These phenom- gravity, on eutectic microstructure has largely remained
ena were attributed to the thermo-electromagnetic convec- unexplained [290]. Thus, it was hypothesized that freezing
tion caused by the high magnetic field. Overall, the use of rate fluctuations due to irregular convection might be
magnetic fields in experiments designed to understand the responsible for spacing adjustments in a fibrous micro-
role of convection during solidification has resulted in sig- structure [290]. An increase in freezing rate must cause
nificant progress during the past 10 years, and additional new fibers to form, either by branching or by nucleation,
experimental work along these lines would be highly and a decrease in freezing rate would cause fibers to termi-
desirable. nate by overgrowth of the matrix phase. However, this
In addition to the arm spacings during directional solid- mechanism was found to result in a finer microstructure,
ification, gravity-driven convection can also influence the instead of the increase in the spacing that is generally
side-branch characteristics of freely grown equiaxed den- observed with increasing convection. Thus, a different
drites, and significant differences to corresponding micro- mechanism was proposed whereby the freezing rate oscilla-
gravity experiments have been observed [277–279]. tions cause the average interfacial melt composition to
Forced convection, as in DC casting of aluminum alloys, deviate from the eutectic because the system does not solid-
can affect the formation of feathery grains made of twinned ify at the extremum. This results in the formation of a com-
dendrites [280]. Finally, the dendritic microstructure can position boundary layer of sufficient thickness that the
also be altered by convection that is induced by vibrations, system would become sufficiently sensitive to convection
and dendrite fragmentation can result [260,281]. [290]. Direct numerical simulation of eutectic solidification
with melt convection should become possible in the near
5.2.3. Columnar-to-equiaxed transition future and could contribute much to the understanding
Direct numerical simulations of the columnar-to-equi- of the above and other effects.
axed transition (CET), using the phase-field method, are The effect of gravity-induced fluid flow on eutectic
now possible for purely diffusive conditions and in two microstructure has also been investigated experimentally
dimensions [282]. However, corresponding numerical simu- for Al–Cu alloys [291]. Different convection modes were
lations taking into account the presence of melt flow, den- achieved by varying the composition from hypo- to hyper-
drite fragmentation and movement of small equiaxed eutectic. Single-phase-to-eutectic, lamellar-to-rod-eutectic
grains are still awaiting the availability of increased compu- and rod-to-lamellar-eutectic transitions were observed.
tational capabilities. Experimental investigations of den- The fluid flow effect on eutectic spacing in eutectic or
drite fragmentation in metallic alloys solidifying under near-eutectic alloys was found to be very small, whereas
natural (non-forced) convection conditions have recently it increases the eutectic spacing in hypoeutectic alloys due
been performed using synchrotron radiography [283,284]. to the increase in the effective diffusion coefficient value.
This very powerful experimental technique allows for The spacing can increase or decrease in hypereutectic alloys
real-time, direct observation of solidification and convec- since the bulk composition changes continuously due to the
tive transport in metals on a microstructural length scale. rejection of lighter Al, so that the spacing value depends on
It was found that fragmentation of higher-order dendrite the solidification fraction. Models were developed to pre-
branches could result from solute-rich liquid flowing from dict the variations in eutectic microstructure caused by
the tips into open regions of the dendritic network [284]. fluid flow.
The effect of the transport and settling of fragments on Density-change-driven flow was found to have a poten-
CET was observed in detail [283]. tially strong effect on eutectic growth [292]. When the den-
In industry, dendrite fragmentation is often accom- sities of the two solid phases are different, there are flow
plished by electromagnetic stirring of the melt, and consid- components both perpendicular and transverse to the
erable grain refinement can be achieved. For such an solid–liquid interface. The direction of the transverse flow
application, a dendrite fragmentation criterion has been is always from the less dense solid phase to the denser solid
developed [285]. However, the fragmentation of dendrites phase. Assuming equal densities of the two solid phases can
in the presence of flow certainly deserves much additional lead to large errors in the phase volume fractions. For some
research attention. Also, the subsequent evolution of the alloys, the minimum undercooling and the lamellar spacing
fragments under convective conditions [286] has not been differ substantially from the values given by the Jackson–
studied sufficiently. Hunt theory for equal densities. For Fe–C alloys, where
the b phase has a lower density than the a phase, the lamel-
5.3. Convection effects on multi-phase solidification lar spacing decreases by a factor of 1.5 due to density-
change-driven flow. For most other alloys, the density of
5.3.1. Eutectic growth the b phase is higher than that of the a phase and the lamel-
Several models have been developed for the lamellar lar spacing increases, as one would expect in the presence
spacing during eutectic growth in the presence of melt flow of flow.
5.3.2. Peritectic and monotectic growth need for a more quantitative model of equiaxed growth
Convection effects in peritectic systems have continued in a constitutionally undercooled boundary layer of a
to attract research interest [182,293–295]. Convection was columnar dendritic growth front that includes dendrite
found to reduce the spacing in a banded peritectic micro- fragmentation, transport, dissolution and growth of frag-
structure. A variety of novel pattern-forming processes ments in the presence of fluid flow.
due to the coupling of convection and phase change were The limited range of processing parameters for which
discovered. The coupling of flow oscillations near the inter- aligned monotectic composites can be grown is controlled
face with solidification results in ordered layered structures. by fluid flow, but the precise mechanisms remain to be
Strong oscillatory convection produces a large treelike pri- determined. There has been progress in treating the effect
mary phase that is surrounded by the peritectic phase of simple flows on linear morphological stability, but
matrix. Nucleation and convection effects in peritectic sys- extension into the nonlinear regime remains to be done.
tems have also been examined in recent phase-field simula- Three-dimensional time-dependent simulations of fluid
tions [296]. flow are now possible, but become difficult for complex
The important role of convection during monotectic flows that exhibit flow transitions. These confounding
growth has been known for some time. Phase-field simula- issues need to be considered more carefully to make any
tions of monotectic solidification with convection in both real progress.
liquid phases were performed to better understand the full
complexities of the transport processes involved [297]. For 6. Microstructures and defects at high solid fractions
fibrous monotectic growth, a new relation for the fiber
spacing in the presence of Marangoni convection was The kinetics of dendrite tips and the dynamics of eutec-
derived [298,299]. A static magnetic field was used to elim- tic fronts have been the focus of much of the attention of
inate convection during directional solidification of an Al– physicists and materials scientists (see Sections 3 and 4).
Bi monotectic alloy, and the fiber spacing was observed to Although these are important topics for the understanding
decrease [300]. and modeling of solidification microstructures, phenomena
occurring deep in the mushy zones are just as important in
5.4. Outlook: microstructure and convection the formation of two major solidification defects, namely
microporosity and hot tearing. This is particularly the case
One of the main present challenges is to develop a quan- for dilute alloys with a large solidification interval, DT0. In
titative understanding of convection effects that strongly such alloys, the fraction of solid gs at which a non-equilib-
influence dendrites in actual castings. While significant pro- rium eutectic might form from the last liquid is close to
gress has been made during the last decade in unraveling unity. Solidification shrinkage associated with the forma-
the complex effects of convection on solidification micro- tion of this eutectic might not be compensated by liquid
structure development, much of the understanding remains flow, thus leading to the formation of shrinkage porosity.
incomplete or fragmented. Most of the existing theories or The space available for the growth of such pores is dictated
experimental findings are only valid in very specific situa- by the morphology of the primary phase, and thus coarsen-
tions or narrow regimes. For example, it is still not clear ing is an important parameter. Further, when gs is typically
how to calculate the primary and secondary dendrite arm larger than 0.9, the primary phase undergoes a morpholog-
spacings in a mushy zone for any value and direction of ical change from isolated solid elements (dendrites or
the local flow velocity. This is important not only for pre- grains) surrounded by continuous liquid films, to a contin-
dicting the microstructure in solidified materials, but also uous solid network with isolated liquid droplets. This coa-
for determining the permeability of the mushy zone during lescence phenomenon leads to the percolation of the solid,
solidification and for modeling phenomena such as macro- i.e. to the formation of a cluster of grains of increasing size,
segregation, porosity formation and hot tearing. Experi- finally equal to that of the considered domain. This transi-
ments and numerical simulations where the flow is tion is very critical in the formation of hot tears or hot
externally imposed and well defined promise to yield the cracks.
most insight, and additional progress in both experimental This section briefly reviews recent progress made in the
(e.g. using magnetic fields for flow control or by in situ characterization and understanding of microstructure evo-
observations with synchrotron radiation) and numerical lution at high solid fractions and of formation of micropo-
techniques can be expected. However, in most solidification rosity and hot tearing.
processes convection is caused by solidification itself and
the flow is heavily influenced by the developing microstruc- 6.1. Microstructure evolution at high solid fractions
ture. Furthermore, convection is usually unsteady and of a
long-range nature, and can thus change the environment in On the experimental side, serial sectioning [301–303], X-
which the microstructure grows. ray tomography [304–306] and in situ X-ray radiography
Studies of the possibilities for improving processing by [284,307,308] have emerged over the past 10 years as pow-
modifying the interaction between flow fields and interface erful techniques for the observation of microstructure evo-
morphologies are scarce. For example, there is a strong lution in metallic alloys at low and high solid fractions.
While the first technique can be applied to solidified speci- been developed recently [316–318]. In such models, grains
mens only, the advent of high-intensity synchrotron X-ray assumed to be globular are randomly distributed in space
sources has allowed in situ observations of microstructure with random crystallographic orientations. A simplified
formation in metallic systems, mainly low-absorbing ele- 1D microsegregation-coalescence model can then be used
ments such as aluminum. Using serial sectioning, Voorhees to predict the solidification of each individual grain and
and co-workers [309] observed the evolution of microstruc- of the liquid channels in between them. As clusters of
tures, in particular the particle size distribution and the grains of increasing size form, the gradual transition from
number of particle contacts during isothermal holding of continuous liquid films to a continuous solid network can
Sn–Pb alloys. They compared their results with phase-field be modeled (Fig. 14).
predictions [310]. Similar observations have been made by
Suéry and co-workers [305] at higher temperatures for Al– 6.2. Mechanical properties of the mushy zone
Cu alloys, but using an in situ tomography technique. As a
main conclusion of these studies, the evolution of the As clustering occurs during solidification, the rheology
microstructure at a high solid fraction is dictated not only of the mushy zone gradually evolves from that of a viscous
by Ostwald ripening, but also by coalescence (or bridging) fluid with grains in suspension at low gs, to that of a fully
of, and growth of the necks in between, individual parti- solid material near gs = 1. At intermediate values of gs
cles. Besides the observation of individual coarsening or when liquid films are still interconnected, the drastically
coalescence events [311], new statistical approaches are different rheological behaviors of the solid skeleton and
required for the representation of the solid–liquid (s‘) inter- of the intergranular liquid make the mushy solid to appear
face evolution. For example, after determining the mean compressible, even though each individual phase is not.
curvature H and the Gaussian curvature K at each point For such complex materials, two-phase average mass and
of the solid–liquid interface, Voorhees and co-workers momentum equations have been derived [319–321], but
[312] use a H–K space representation to follow the evolu- they must be complemented by constitutive equations.
tion of the surface density of these points. While that of the liquid can be approximated by Darcy’s
While coarsening phenomena have been investigated law, the viscoplastic compressible behavior of the mushy
fairly thoroughly from an experimental and a theoretical solid is described by a model in which the pressure in the
point of view [312], coalescence or bridging of two particles solid appears, in addition to the equivalent von Mises
or dendrite arms has not been studied to any great extent.
Being the opposite of grain boundary wetting, this last phe-
nomenon is primarily dictated by the Gibbs–Smith condi-
tion, which compares the grain boundary energy cgb and
twice the solid–liquid interfacial energy 2cs‘ [313]. Based
on simple thermodynamic considerations [314], it can be
shown that, for cgb < 2cs‘, a grain boundary is ‘‘attractive”,
i.e. solid bridges are established as soon as the two solid–
liquid interfaces of the grains get within an interaction dis-
tance of the order of the thickness d of the diffuse solid–
liquid interface. However, when cgb > 2cs‘, the grain
boundary is ‘‘repulsive” and bridging requires an addi-
tional coalescence undercooling, DTb = (cgb 2cs‘).
(DSfd)1, where DSf is the volumetric entropy of fusion.
For an alloy, it has been shown that liquid films remain
at a grain boundary until a coalescence line (or surface),
parallel to the liquidus but located DTb below, is reached
by a combination of cooling and back-diffusion. This
approach was substantiated by phase-field simulations
[314] and more recently by MD calculations [315].
Coalescence in alloys can be modeled in detail only for
very small volume elements since it involves three widely
different length scales, associated with diffusion, capillarity
and (real) thickness of the diffuse interface. However, its
influence on the gradual formation of a coherent solid Fig. 14. Computed grain structure evolution in two dimensions in a
manifests itself on the scale of a fairly large population of directionally solidified Al–1 wt.% Cu alloy. The four images (a–d)
represent various transverse sections of increasing volume fraction of solid
grains since it also includes statistical aspects related to
of a directionally solidified specimen. They clearly show the gradual
the crystallographic orientation and the respective position percolation, i.e. the formation of increasingly large size clusters of grains.
of the grains. Therefore, in order to approach such prob- The connectivity of the grains or clusters appear with various grey levels
lems, granular or discrete element methods (DEM) have while the liquid is black. (After Vernède et al. [316,317].)
stress. Further details of the rheological behavior of a

mushy solid can be found in Refs. [322–324].
6.3. Microporosity formation
Shrinkage microporosity in alloys is a result of a lack of

feeding of the mushy zone, i.e. the density increase associ-
ated with solidification cannot be fully compensated by an
interdendritic fluid flow opposite to the displacement of the
isotherms [325]. As this flow induces a pressure drop in the
interdendritic liquid when the temperature decreases, the
solubility limit of gases such as hydrogen or nitrogen
decreases. Simultaneously, their actual composition in the
liquid increases due to segregation and might exceed at
some point their solubility limit, thus allowing pore nucle-
ation. However, the contributions of gas diffusion and cap-
illarity must be accounted for the nucleation and the
growth of pores. Fairly comprehensive microporosity mod-
els [326–331] have been developed since the first work of
Piwonka and Flemings [332], but they will not be detailed Fig. 15. Three-dimensional high-resolution X-ray tomography of a
here. However, because it is directly linked with the mor- shrinkage porosity formed in an Al–4.5 wt.% Cu alloy [403].
phology of the primary phase, the capillarity contribution
is briefly discussed here. Again, X-ray tomography has
been instrumental to reveal the very complex and intercon- tearing criteria based on the stress, strain or strain rate
nected morphology of shrinkage porosity in 3D (see experienced by the mushy zone [337,338], sometimes with
Fig. 15). Two-dimensional sections make them appear as a normalization with the cooling rate, have been devel-
small, disconnected holes. The highest curvature of such oped, but the mechanisms of straining and feeding of the
pores corresponds to the portion of its surface in contact mushy zone have been approached only recently using
with the interdendritic liquid, since the gas–liquid (or two-phase models.
gas–solid) interfacial energy c‘g is typically one order of The first, yet simple, two-phase approach for the predic-
magnitude larger than cs‘. In Al–Cu alloys, radius of curva- tion of hot tearing was derived in 1999 by Rappaz, Drezet
ture as small as 5 lm have been measured thus inducing a and Gremaud (RDG) [339]. Similar to other criteria for
Laplace overpressure of about 400 kPa with respect to the porosity formation [340], this approach considers steady-
surrounding liquid. Such contributions must absolutely be state directional solidification in a given thermal gradient
accounted for in porosity models if realistic simulations are and a strain rate e_ s? in the perpendicular direction. The
to be produced. pressure drop Dp‘ from the liquidus Tliq to a coherency
temperature Tcoh can be calculated, Tcoh being defined as
6.4. Hot tearing the temperature at which continuous liquid films disappear.
This RDG criterion can also be used under non-steady-
If microporosity is the major defect in shape casting, hot state conditions by first calculating the strain rate in the
tearing is certainly the major defect of continuous or semi- fully solid part and then feeding in the mushy zone consid-
continuous casting processes and of welding [325]. ering the strain rate component perpendicular to the ther-
Although similar to shrinkage-induced microporosity since mal gradient [341]. Although this criterion compares
it occurs in large solidification interval alloys, hot tearing fairly well with experiments [336], it has two main weak-
involves tensile stresses in the partially coherent solid. nesses: (i) the strain rate in the mushy zone is not well
Compared to microporosity models, hot tearing was defined, as it is taken from the extrapolation of a deforma-
approached with very simple criteria related, for example, tion calculation performed for a fully solid material; and
to the solidification interval [325] or the feeding ability (ii) the model is very sensitive upon Tcoh.
[333] of the alloy. Clyne and Davies [334] recognized the The first limitation of the RDG criterion was removed
importance of the microstructure in the formation of this using a real two-phase formulation in which the mechani-
defect, namely the persistence of continuous liquid films cal behavior of the compressible mushy solid is accounted
at grain boundaries in a region of the mushy zone where for [324,342] (see Section 6.2). In such models, the calcula-
feeding is nearly impossible. They defined a criterion based tion of the velocity vs of the mushy solid is coupled with
on the time spent by the mushy zone in this critical region that of the pressure drop Dp‘ in the liquid, the latter now
of the mushy zone, typically for 0.9 < gs < 0.99. However, including the contributions of solidification shrinkage, as
when applied to various solidification processes, this crite- for microporosity (see Section 6.3), and of the compres-
rion fails to predict the right trend [335,336]. Other hot sion/expansion of the mushy solid. The second limitation
is strongly related to the evolution of the microstructure at couple such methods with the phase-field technique if one
high gs value. desires to model simultaneously diffusion in the solid phase
Because they are based on continuous one- or two-phase and preferential solute segregation at grain boundaries.
media, hot tearing approaches neglect the importance of Based on such micromodels, the percolation of an ensem-
localization of strains at grain boundaries. Yet the effect ble of grains will have to be studied in three dimensions
of grain boundaries on hot tearing was clearly demon- in order to capture the gradual transition of the mushy
strated by welding two single crystals of nickel-based zone under more realistic conditions. For example, in
superalloy under identical conditions but with an increas- two dimensions, either the solid or the liquid phase is con-
ing tilt boundary misorientation (Fig. 16) [343]. These tinuous, whereas in three dimensions both can be continu-
observations showed that hot cracks formed only when ous at the same time. This will require the development of
the misorientation exceeded a given value, which was then dedicated 3D granular or discrete element models that
interpreted based on a grain boundary energy and coales- encompass solidification and coalescence. We have almost
cence model. Such observations clearly indicate the neces- all the tools necessary for the study and modeling of the
sity to account for the microstructure evolution at a high growth of microporosity in a well-developed dendritic net-
solid fraction, and thus granular or DEM approaches work. Fast X-ray tomography combined with multi-phase-
really should be developed. Although limited yet to 2D field simulations will definitely give a quantitative assess-
geometries, such developments have already been under- ment of the important contribution of capillarity during
taken and proved their usefulness [316–318,344]. pore formation. Hot tearing modeling will also go along
the same direction of granular or discrete element models.
6.5. Outlook: microstructures and defects at high solid Such methods have already been employed successfully for
fraction a whole range of problems (e.g. sintering, fragmentation,
soil mechanics, granular flows). A better knowledge of
Over the past 40 years, theories and models have the mechanical properties of the semi-solid material and
focused a lot on the early stages of solidification (formation the characterization of the microstructure near a solid frac-
of dendrites, eutectics, etc.). Yet the evolution of the micro- tion of unity are necessary. A 3D model that encompasses
structure at high solid fraction is a key parameter for the solidification, coalescence, feeding and deformation in the
prediction of defects such as microporosity and hot tearing. mushy zone has to be developed.
Besides the improvement of commercial solidification
codes for the prediction of such defects, our understanding 7. Solid-state transformation parallels
of microstructure evolution at high solid fraction must be
improved. Along this line, a detailed modeling of coales- In the 1950s and 1960s, the distinction between the
cence based on molecular dynamics or phase-field crystal fields of solidification and solid–solid transformations
methods, correlated with in situ X-ray tomography obser- was not as clearly defined as it is today; for example, back
vations, is necessary. For alloys, it might be necessary to then the same theories were used to study both the growth
a
[001] [001]
c 1000
[010]
[010] 800
Crack length [µm]
600
[100]
b 400
= 8°
Hot crack
200
No crack
0
= 24°
0 2 4 6 8 10 12 14 16
θ/2
Fig. 16. Hot tearing weld experiment. (a) Experimental setup with laser beam (in grey) used for welding of two nickel-base superalloy single crystals with
symmetric orientation with respect to the weld plane (angle a between [100] and weld direction, 2a = h; (b) laser welds of two bi-crystals exhibiting tilt
grain boundary angles h of 8 (above) and 24°. (c) Crack length as a function of grain boundary half angle a. Under identical welding conditions, only low
angle grain boundary welds with h 6 12° did not exhibit hot cracking [343].
of dendrites and Widmanstaetten plates. In the following recent review, the work presented is not limited to the last
years, the fields of solidification and solid-state transfor- decade.
mations have deepened and diverged. In recent years,
however, there have been a few tentative attempts to close 7.1. Directional transformation
this gap.
The theory of solid-state phase transformations really 7.1.1. Single-phase growth
started with the landmark paper by Zener some 60 years Directional transformation (Bridgman type growth) is
ago [345], which laid the foundation of more quantitative characterized by an imposed growth rate in a positive tem-
models by Hillert [346], Trivedi [347] and Jackson and perature gradient. The conditions are chosen such as to
Hunt [156]. These models have since been used extensively avoid nucleation of the product phase in the parent phase
in solidification research. This section is not a review of and to obtain a macroscopically flat transformation front.
solid-state transformations but an attempt to highlight This technique allows direct examination of the interface
links between the two fields, stressing that there can be response, i.e. the development of different interface mor-
fruitful cross-fertilization of ideas and techniques between phologies of an alloy (plane front, cells, dendrites) and
solidification and solid-state transformation. related interface/tip temperatures as a function of velocity
There are certain similarities in the evolution of both in a selected temperature gradient [352,353]. Directional
fields, in that solute transport, interfacial energy and inter- transformation of the alloy can be carried out from liquid
facial kinetics [348] have been considered significant in to room temperature so that the consecutive liquid–solid
both cases. Similar phenomena include the growth of nee- and solid–solid transformation fronts in a given sample
dles and plates (e.g. dendrites vs. Widmanstaetten struc- can be studied under well-defined steady-state conditions
tures), coupled growth of eutectics and eutectoids, and [354]. In the high-temperature solid-state transformation
massive transformations. There are, however, several from d-ferrite to c-austenite in Fe–Ni or Fe–Cr alloys it
major factors that are different: convection effects are cru- has been shown that solidification models apply well to
cial in solidification only, whereas strain energy effects the observed solid-state transformation structures
often play an important role in solid-state precipitation. [354,355] as the interface seems to be mobile and the parent
In addition, and related to the fact that at least two crystals d has higher diffusion coefficients than c. Further, it could
must meet at the solid–solid transformation front, aniso- be unambiguously shown that in a Fe–17.5 at.% Co alloy
tropies of interfacial free energy and interfacial mobility [357] absolute stability [356] of the solid–solid transforma-
are generally more pronounced in solid–solid transformation front was reached. In Fig. 17 the growth fronts for two
tions: interfaces are often faceted, requiring the formation different velocities (1 and 10 lm s1) can be seen: The L/d
and lateral propagation of ledges for their migration fronts (top) of both quenched specimens correspond to
[349]. Consequently, the strongest parallels between solidi- plane front growth as they form below constitutional und-
fication and solid–solid transformations will be found in ercooling for solidification. The d/c solid-state transforma-
faceted growth (for both) on the one hand, and in non-fac- tion front (centre) is cellular at V = 1 lm s1 (<Va) and
eted growth (e.g. comparing the motion of ‘‘incoherent” planar at 10 lm s1 (>Va) [357]. The calculated limit of
solid-state interfaces with that of typical metallic solid–
liquid interfaces) on the other.
Phase-field modeling techniques that have proved so
valuable in solidification science have also been applied
to solid-state transformations. The review on the subject
by Chen [350] is helpful for work before 2002. Multiscale
models are highly promising approaches with the potential
to yield quantitative insight into microstructure evolution
[109]. A more tractable technique, which can cope with
orders of magnitude larger space and time frames, has been
developed recently by Elder et al. [351] and is called the
phase-field crystal (PFC) method (a review of this method
can be found in Ref. [152]).
Most of the experimental studies in solid–solid transfor-
mations are carried out isothermally, similar to the solidi- Fig. 17. Consecutive phase changes from top to bottom (L/d and d/c) in
fication of alloys in an undercooled melt. In contrast, directionally transformed hypoperitectic Fe–17.5 at.% Co alloy;
directional transformation is used widely in solidification, G = 18 K mm1. (a, top) Plane front d ferrite growing at a smaller rate
but has not been very common in the study of solid-state than the critical velocity for constitutional undercooling for solidification
and (a, bottom) cellular c transformed in solid state at V = 1 lm s1. (b,
transformations, despite its potential.
top) Plane front d below the critical rate of constitutional undercooling
In the following the subject is therefore subdivided into and (b, bottom) plane front c above absolute stability for solid-state
directional and isothermal transformations of single- or transformation at V = 10 lm s1 [357]. In (b) the low temperature
two-phase alloys. As the topic has not been subject to a austenite–ferrite transformation produced a ferrite grain (light grey).
absolute stability, Va, for this solid-state transformation is

6 lm s1, in agreement with experiment. One has in
Fig. 17(b) the interesting case of two planar transformation
fronts: the liquid/d front, which grows at a speed below the
limit of constitutional undercooling, and the d/c front,
which grows above the limit of absolute morphological sta-
bility. The reason for this behavior is the huge difference in
the diffusion coefficients between the solid and liquid
phases. The solid-state plane front grows close to local
equilibrium and in steady state the product phase is com-
pletely supersaturated as a result of a narrow concentration Fig. 18. Comparison of measured and calculated growth temperatures as
pile up in ferrite. Such a transformation where the product a function of growth velocity of pearlite. Open and full circles are
phase has the same composition as the parent phase is experimental data; the other symbols are from theory [364].
called a ‘‘massive transformation”. In this sense, two con-
secutive massive transformation interfaces are observed
(Fig. 17(b)). The simultaneous observation of two or more important in a eutectoid can also become important in a
interfaces would provide critical information when strain eutectic at high growth rates or in polymer eutectics. For
energy, interface kinetics or rapid solidification effects example, the scaling laws, k3V = const. and DT6 / V, have
due to the larger Peclet number in the solid influence the been observed in a polymer system, and it has been shown
solid–solid interface structure. that accounting for interface diffusion gives the scaling law
between the velocity and spacing, while the scaling law for
7.1.2. Two-phase growth undercooling is controlled by the interface kinetics for sec-
For every three-phase reaction involving one liquid and ondary nucleation of the molecular attachment process
two solid phases, there exists a solid-state analogue, e.g. [365].
eutectic/eutectoid; peritectic/peritectoid; monotectic/
monotectoid. The eutectoid (pearlite) reaction in carbon 7.2. Isothermal transformation
steels has attracted perhaps the most attention, as these
steels are widely used. Eutectoid growth is also amendable 7.2.1. Single-phase growth
to study by directional transformation techniques [358– Certain solid-state precipitates can be said to be den-
362]; the analogy with eutectic growth is nearly complete. dritic, as, for example, in the case of the precipitation of
The models of eutectic growth at low rates have been c in b brass [366] and the formation of M23C6 carbides in
experimentally verified for both the lamellar [159] and alloy steels [367]. Well-developed dendritic growth has also
rod eutectic [363] in systems whose required parameters been observed in glass as it undergoes crystallization
have been accurately measured. For solidification, the scal- [368,369]. Sometimes grain boundary-initiated precipitates
ing laws between velocity and interphase spacing and possess a 2D dendritic morphology. Perhaps more typical
between undercooling and velocity were found to follow is the formation of a needles from b brass, a reaction that
the model based on diffusion in the parent phase including leads to Widmanstaetten precipitates, which show very lit-
capillarity, which are k2V = const. and DT2 / V [346,156]. tle tendency to side branch. However, these precipitates do
In contrast, in eutectoid systems the above equations show approximately paraboloidal fronts near the tip
often agree less well with the experimental data. For pearl- regions, and the side faces are locked into a low energy/
ite growth, Pearson and Verhoeven [358,359] showed that low mobility zone. The analogy with dendrite growth, i.e.
the scaling laws are obeyed, but that the values of the con- the capillarity-corrected Ivantsov model [370,371], has been
stants disagree. This discrepancy has been recently re-eval- used to predict lengthening kinetics, with some success
uated by Steinbach and Apel [364] with phase-field [372–374]. Recently, good agreement with measured tem-
calculations to fit experimental undercoolings of pearlite perature dependence of growth was found using current
with theory (Fig. 18). Further experiments are necessary phase-field models for 2D dendritic growth [375]. The basic
to confirm the assumptions of this model. concept of 2D dendrite tip stability was used to predict the
Carpay [362] has carried out directional transformations selection of tip radius of Widmanstaetten precipitates at
in several eutectoid systems, and significant deviations in low and high growth rates [376]. The tip selection originat-
the exponent in the scaling laws have been obtained. This ing from the constraint due to low mobility sides in this
can be attributed to the various physical phenomena that case was found to coincide with the maximum growth rate
become important in solid–solid transformation, which condition. Since the Peclet number in the solid can be large,
include interface diffusion, strain energy, interface kinetics rapid transformation effects were also considered, which
or diffusion in the two new phases. In fact, different models gave results [376] similar to the Ivantsov marginally stabil-
that include these effects have been developed and com- ity dendrite model [377].
pared with the experimental results to establish the mecha- Several recent studies have used simplified phase-field
nism of coupled growth. Some of the effects that are models to simulate the precipitates’ evolution from an ini-
tial fluctuation [378,379]. The formation of Widmanstaet- parent phase to a regular array of precipitates [391,392].
ten side plates from prior grain boundary precipitates has Brener and Temkin developed a model that took into
been modeled [378], using a highly anisotropic interfacial account the elastic effects that arise because of the sharp con-
free energy but neglecting anisotropy of mobility and strain centration change ahead of the transformation front [393].
energy. For further progress in this area, we look to a They showed that elastic effects are required to obtain a
newly developed and highly promising adaptation of the match with the experimentally observed result for realistic
phase-field approach by Greenwood and Provatas [375], values of the physical parameters. The final morphology is
which accounts for the interplay of anisotropies of interfa- very like that expected from a cooperative eutectoid, except
cial energy, interfacial mobility and self-strain energy, as that only two phases (a = a0 + b) participate in the overall
well as the strain energy associated with the gradient of a reaction. The question whether stability considerations offer
misfitting solute in the parent crystal. Phase-field modeling a criterion for the selection of the growth velocity and spac-
of various aspects of transformations in Fe–C steels, such ing [394] or whether a continuous family of stable steady-
as grain growth, massive transformation, nucleation and state solutions exists [395] is still open.
coarsening, has also been undertaken [380,381].
Multi-phase diffusion couples have been used widely in 7.3. Outlook: solid-state transformation parallels
the study of morphological stability in the solid state, par-
ticularly for the multicomponent case. Concepts of interfa- While the parallels between liquid–solid and solid–solid
cial stability drawn from solidification theory have found phase transformations are many and striking, we must
extensive application to evaluate the non-equilibrium return to, and even stress, the points of divergence, involv-
effects at the interface that are often critical in ternary steels ing the influence of strong anisotropy of interfacial proper-
of composition Fe-CX, where X is a substitutional ternary ties and the effects of strain and strain energy on the solid
element such as Ni, Mn or Cr [382,383], which may or may state. These factors restrict the types of transformation that
not partition at the interface. For example, the observed can profitably be taken as analogues. Nevertheless,
kinetics are often precisely parabolic and controlled by researchers in the two subdisciplines share many funda-
the diffusion of the interstitial solute. A much more slowly mental concerns, especially those dealing with the stability
diffusing substitutional solute (e.g. Ni) does not partition, of growth fronts and transformation products.
but its near-interface behavior is critical in determining For the future, we suppose that, by careful selection of
the interfacial boundary conditions for the diffusion of car- systems to study, one can take advantage of the ease of
bon. The method has proven ideal for the study of transi- quenching and retaining transformation structures in
tions between full local equilibrium [384] states, for which a solid-state transformation products and the potential for
pile up of the substitutional solute occurs in the parent high-resolution chemical analysis of the interfacial and
phase, and partial local equilibrium (‘‘paraequilibrium” near interfacial concentration fields. The absence of con-
[385]) interfacial states, as well as the investigation of the vection effects may also be considered a benefit.
morphological stability of the transformation interface There is considerable potential for the use of directional
[386]. growth experiments for the study of solid–solid reactions as
In in situ observations of solidification and solid-state they are superior to isothermal experiments with respect to
transformation of peritectic Fe–Ni alloys it was found that a control of velocity and temperature gradient and growth
the peritectic c phase precipitating at grain boundaries of under steady-state conditions. The interface composition
the supersaturated d-phase developed morphological insta- and interface temperature are quantitatively accessible.
bilities which grew perpendicularly to the grain boundaries At high temperatures, solidification models are sometimes
[387,388]. Solidification theory has been successfully applicable and allow interesting comparisons between
applied for the interpretation of these observations. solid–liquid and solid–solid growth fronts. In contrast,
In solid-state transformations anisotropy in interface the effects of strain energy, complex interface kinetics and
energy and in interface mobility play a dominant role in surface diffusion often play an important role in eutectoid
governing precipitate morphologies. For such cases transformations and give rise to different scaling laws
phase-field models developed for faceted solids may be use- between velocity, undercooling and spacing. Directional
ful for more quantitative insights. transformation experiments are therefore capable of pro-
viding critical information on the importance of these
7.2.2. Two-phase growth effects on transformation kinetics. We also emphasize the
In several studies of isothermal growth, the effect of useful analogy between a temperature gradient in direc-
alloying elements on eutectoid growth has been examined. tional solidification experiments and a gradient of fast-dif-
For example, the ideas of Hillert [389] on the effects of Mn fusing interstitial solute (e.g. C, H, O, N) in isothermal
additions on the growth of pearlite were tested quantita- directional experiments like studies of decarburization
tively [390]. and internal oxidation.
A solid-state reaction which has no liquid–solid ana- In future, we expect that experimental research on solid–
logue is discontinuous precipitation. In this case, a migrat- solid transformations will maintain a primary focus on the
ing grain boundary acts as a conduit for solute from the results of ultra-high-resolution techniques that yield both
structural and chemical information [396]. The simulation Modeling directions: The integration of molecular
of these transformations will increasingly involve advanced dynamics and phase-field techniques, the phase-field crystal
computationally intensive techniques, such as molecular method and micro–macro modeling techniques will lead to
dynamics [397], phase-field calculations [375], multiscale more realistic predictions of real processes. Enhanced accu-
models [109] and the recently developed phase-field-crystal racy in the atomistic simulations will require continued
methods [152]. This latter method resolves the atomic-scale efforts aimed at the development of more quantitative
density wave structure of a polycrystalline material and interatomic potential models. Three-dimensional modeling
describes its evolution in the presence of elastic interactions and incorporation of fluid flow in these models is neces-
on time scales that are orders of magnitude longer than sary. Faster computational algorithms and improved phys-
molecular dynamics simulations. Thus, it appears ideally ical models for the calculation of multicomponent phase
suited to study solid-state phase transformations and equilibria and improved thermodynamic data banks are
promises to become a major modeling technique for the prerequisites for any successful simulation [402]. Atomistic
next decades. modeling of heterogeneous nucleation and inoculation
Further, we suggest that it would be worthwhile to shows great potential for progress of our understanding
examine two of the little-explored solid-state transforma- of the first moments of phase transformations. Integrated
tions: peritectoid and monotectoid systems. Here the modeling of solidification with solid-state transformations
absence of convection will be a major advantage for and properties would help process optimization and alloy
obtaining relevant experimental results. development.
Finally, a promising approach for the modeling of the Open questions: What is the effect of concentration
properties of cast alloys has to be mentioned. Gandin dependent and off-diagonal diffusion terms on diffusional
et al. integrated the modeling of various aspects of phase transformation kinetics in multicomponent systems? What
transformations (solidification, homogenization and pre- are the mechanisms of cell to dendrite and dendrite to cell
cipitation), with structural hardening and mechanical transitions? How do 3D instabilities affect the operating
behavior. Their results indicate ways to improve the com- states of eutectics with anisotropic solid–solid interface
plete processing chain from the melt to the final heat trea- energies? Is it really impossible to grow large eutectic single
ted solid [398,399]. grains? Which microstructure selection mechanisms oper-
ate in three-phase eutectics? Which operating states exist
8. Conclusions in faceted–non-faceted eutectics? What are the mechanisms
of eutectic growth transitions from lamellar to fibrous to
This review shows the progress that has been made in spherulitic morphologies? Which mechanisms lead to cou-
solidification science over the recent years. To summarize, pled peritectic growth? Does coupled peritectoid growth
some important challenges and major directions for future exist? How does melt convection affect; dendritic growth
research are given. They are subdivided into experimental at high flow velocities, dendritic and eutectic spacings,
and computational techniques, and open questions. and dendrite fragmentation? How does stress and deforma-
Experimental challenges: Direct measurements of inter- tion affect initiation and growth of pores and cracks in the
face properties and their anisotropies in real systems are mushy zone with a high solid fraction? What are the effects
required to confirm the results of atomistic computations. of strain energy and anisotropic interface kinetics in solid-
Use of colloidal model systems could improve substantially state transformations? How can the scaling laws of eutec-
our understanding of nucleation. Extensive use of 3D imag- toid growth be improved? What are the analogies and dif-
ing techniques [400] and real-time in situ observations will ferences between rapid (high Peclet number) growth in the
produce new insight into the complex phenomena of growth solid state and rapid solidification?
and defect formation. The quantitative evaluation of micro- These are some of the more important questions that
structures and mechanical properties of the mushy zone, spring to mind when analyzing the present understanding
especially at high volume fractions of solid, are important of phase transformations. These questions clearly indicate
for the understanding of nucleation and growth of second that, despite the substantial progress that has been made
phases, pores and cracks. As fluid flow strongly influences in solidification science over the recent years, there is much
the microstructures, its control and absence in directional room left for future contributions to this scientifically and
solidification by the use of small-diameter tubes is a power- technologically important field.
ful technique, complementary to space experiments. More
work on reliable phase diagrams, which are the basis of Acknowledgements
any microstructure modeling, is necessary. Inverse methods
using microstructure selection maps for the quantitative M.A. and A.K. thank the US Department of Energy for
determination of heterogeneous phase equilibria can be use- support through Grants DE-FG02-07ER46400 (A.K.),
ful in this respect [401]. The highly important field of solid- DE-FG02-01ER45910 (M.A.) and DE-FG02-06ER46282
state transformations can profit from directional transfor- (M.A.), and funds from the Computational Materials Sci-
mation experiments, which permit excellent control of a ence Network. The work of C.B. was supported, in part,
wide range of transformation conditions. by NASA under Contract NNM04AA18G.
Appendix A References
[1] Kurz W, Trivedi R. Acta Metall Mater 1990;38:1.

List of symbols [2] Boettinger WJ, Coriell SR, Greer AL, Karma A, Kurz W, Rappaz
Co concentration M, et al. Acta Mater 2000;48:43.
D liquid-phase diffusivity [3] Langer JS. In: Souletie J, Vannimenus J, Stora R, editors. Chance
and matter, Lectures on the theory of pattern formation, Les
d0 capillary length
Houches, Session XLVI, North Holland; 1987. p. 629.
g volume fraction [4] Kessler D, Koplik J, Levine H. Adv Phys 1988;37:255.
K permeability [5] Karma A, Rappel W-J. Phys Rev E 1998;57:4323.
k(V) velocity-dependent partition coefficient [6] Liu S, Napolitano RE, Trivedi R. Acta Mater 2001;49:4271.
kB Boltzman constant [7] Napolitano RE, Liu S, Trivedi R. Interface Sci 2002;10:217.
[8] Napolitano RE, Liu S. Phys Rev B 2004;70:214103.
L latent heat of melting
[9] Passerone A, Eustathopoulos N. Acta Mater 1982;30:1349.
M atomic mass [10] Glicksman ME, Schaefer RJ. J Cryst Growth 1967;1:297.
^n interface normal [11] Rodway GH, Hunt JD. J Cryst Growth 1991;112:554.
P Peclet number (RV/2D) [12] Hoyt JJ, Asta M, Karma A. Mater Sci Eng R 2003;41:121.
p pressure [13] Hoyt JJ, Asta M, Haxhimali T, Karma A, Napolitano RE, Trivedi
R radius R, et al. MRS Bull 2004;29:935.
[14] Davidchack RL, Laird BB. J Phys Chem B 2005;109:17802.
S interface stiffness [15] Frenkel D, Smit B. Understanding molecular simulation: from
T temperature algorithms to applications. San Diego (CA): Academic Press; 2002.
Tcoh coherency temperature [16] Daw MS, Baskes MI. Phys Rev Lett 1983;50:1285.
Tliq liquidus temperature [17] Daw MS, Baskes MI. Phys Rev B 1984;29:6443.
[18] Turnbull D. J Appl Phys 1950;21:1022.
TM melting temperature
[19] Kelton KF. Solid State Phys 1991;45:75.
Tsol solidus temperature [20] Granasy L, Borzsonyi T, Pusztai T. Phys Rev Lett 2002;88:
t time 206105.
V interface velocity [21] Waseda Y, Miller WA. Trans Jpn Inst Metals 1978;19:546.
Va absolute stability [22] Glicksman ME, Vold CL. Acta Metall 1969;17:1.
[23] Glicksman ME, Vold CL. Scr Metall 1983;5:493.
Vs sound velocity
[24] Eustathopoulos N. Int Met Rev 1983;28:189.
VT thermal velocity [25] Spaepen F. Acta Metall Mater 1975;23:729.
xL solute mole fractions in liquid [26] Spaepen F, Mayer RB. Scr Metall Mater 1976;10:257.
xS solute mole fractions in solid [27] Thompson CV. Ph.D. thesis, Harvard University; 1979.
a Turnbull coefficient [28] Herlach DM, Galenko P, Holland-Moritz D. Metastable solids from
c crystal–melt interfacial free energy undercooled melts. Oxford: Pergamon; 2007.
[29] Cech RE. Trans AIME 1956;206:535.
c0 orientation-averaged interfacial free energy [30] Ghosh G. Mater Sci Eng A 1994;189:277.
DS entropy of melting [31] Volkmann T, Loeser W, Herlach DN. Metall Mater Trans A
DT interface undercooling 1997;28A:453.
DTb coalescence undercooling [32] Zambon A, Badan B, Eckler K, Gärtner F, Norman AF, Greer AL,
et al. Acta Mater 1998;46:4657.
DTo equilibrium solidification interval
[33] Li M, Lin Z, Song G, Yang G, Zhou Y. Mater Sci Eng A
d thickness of the solid–liquid interface 1999;268:90.
e1 amplitude of the first term in a cubic [34] Herlach DM. J Phys Condens Matter 2001;13:7737.
harmonic expansion of c in 3D [35] Greer AL, Walker IR. J Non-Cryst Solids 2003;317:78.
e2 amplitude of the second term in a cubic [36] Notthoff C, Feuerbacker V, Franz H, Herlach DM, Holland-Moritz
harmonic expansion of c in 3D D. Phys Rev Lett 2001;86:1038.
[37] Holland-Moritz D. Int J Non-Equilib Process 1998;11:169.
e4 amplitude of the fourfold variation of c [38] Holland-Moritz D. J Non-Cryst Solids 1999;252:839.
es strain [39] Holland-Moritz D. Mater Sci Eng A 2001;304–306:108.
ek kinetic anisotropy [40] Holland-Moritz D, Herlach DM, Grushko B, Urban K. Mater Sci
e_ s strain rate Eng A 1994;182:766.
g viscosity [41] Schroers J, Holland-Moritz D, Herlach DM, Urban K. Phys Rev B
2000;61:14500.
h angle [42] Broughton JQ, Gilmer GH. J Chem Phys 1986;84:5759.
l kinetic coefficient = V/DT [43] Hoyt JJ, Asta M, Sun DY. Philos Mag 2006;86:3651.
q atomic density [44] Sun DY, Asta M, Hoyt JJ, Mendelev MI, Srolovitz DJ. Phys. Rev. B
r stability parameter 2004;69:020102.
req equivalent von Mises stress [45] Davidchack RL, Morris JR, Laird BB. J Chem Phys
2006;125:094710.
rs stress tensor in the solid [46] Morris JR, Mendelev MI, Srolovitz DJ. J Non-Cryst Solids
/ angle 2007;353:3565.
u angle [47] Fehlner WR, Vosko SH. Can J Phys 1976;54:2159.
v velocity [48] Hoyt JJ, Asta M, Karma A. Phys Rev Lett 2001;86:5530.
[49] Morris JR. Phys Rev B 2002;66:144104.
[50] Haxhimali T, Karma A, Gonzales F, Rappaz M. Nat Mater [102] Galenko PK, Danilov DA. J Cryst Growth 1999;197:992.
2006;5:660. [103] Ahmad NA, Wheeler AA, Boettinger WJ, McFadden GB. Phys Rev
[51] Becker CA, Olmsted D, Asta M, Hoyt JJ, Foiles SM. Phys Rev Lett E 1998;58:3436.
2007;98:125701. [104] Cai Y, Zhang G, Fu H. J Appl Phys 2000;87:7735.
[52] Davidchack RL, Laird BB. Phys Rev Lett 2005;94:086102. [105] Danilov D, Nestler B. Acta Mater 1996;54:4659.
[53] Morris JR, Song X. J Chem Phys 2003;119:3920. [106] Kerrache A, Horbach J, Binder K. Europhys Lett 2008;81:58001.
[54] Davidchack RL, Laird BB. J Chem Phys 2003;118:7651. [107] Sluiter M, Kawazoe Y. Phys Rev B 1996;54:10381.
[55] Davidchack RL, Laird BB. Phys Rev Lett 2000;85:4751. [108] Asta M, Ozolins V, Woodward C. JOM 2001;53:16.
[56] Gonzales F, Rappaz M. Metall Trans A 2006;37:2797. [109] Vaithyanathan V, Wolverton C, Chen LQ. Phys Rev Lett
[57] Jackson KA. Interface Sci 2002;10:159. 2002;88:125503.
[58] Chernov AA. Modern crystallography, vol. 3. Berlin: Springer; 1984. [110] Hafner J, Wolverton G, Ceder G. MRS Bull 2006;31:659.
[59] Aziz MJ, Boettinger WJ. Acta Metall Mater 1994;42:527. [111] Bi ZQ, Sekerka RF. J Cryst Growth 2002;237:138.
[60] Boettinger WJ, Aziz MJ. Acta Metall 1989;37:3379. [112] Kauerauf B, Zimmermann G, Rex S, Billia B, Jamgotchian H, Hunt
[61] Turnbull D, Bagley BG. In: Hannay NG, editor. Treatise on solid JD. J Cryst Growth 2001;223:277.
state chemistry, vol. 5. New York: Plenum; 1975. p. 526. [113] Karma A. Phys Rev Lett 2001;87:115701.
[62] Coriell SR, Turnbull DH. Acta Metall 1982;30:2135. [114] Ramirez JC, Beckermann C, Karma A, Diepers H-J. Phys Rev E
[63] Broughton JQ, Gilmer GH, Jackson KA. Phys Rev Lett 2004;69:051607.
1982;49:1496. [115] Echebarria B, Folch R, Karma A, Plapp M. Phys Rev E
[64] Hoyt JJ, Sadigh B, Asta M, Foiles SM. Acta Mater 1999;47:3181. 2004;70:061604.
[65] Mikheev LV, Chernov AA. J Cryst Growth 1991;112:591. [116] Provatas N, Goldenfeld N, Dantzig J, LaCombe JC, Lupulescu A,
[66] Chernov AA. J Cryst Growth 2004;264:499. Koss MB, et al. Phys Rev Lett 1999;82:4496.
[67] Burke E, Broughton JQ, Gilmer GH. J Chem Phys 1988;89:1030. [117] Plapp M, Karma A. Phys Rev Lett 2000;84:1740.
[68] Tepper HL, Briels WJ. J Chem Phys 2002;116:5186. [118] Fan J, Greenwood JM, Haataja M, Provatas N. Phys Rev E
[69] Tepper HL, Briels WJ. J Chem Phys 2001;115:9434. 2006;74:031602.
[70] Huitema HEA, Vlot MJ, van der Eerden JP. J Chem Phys [119] Bragard J, Karma A, Lee YH, Plapp M. Interface Sci 2002;10:121.
1999;111:4714. [120] Greenwood M, Haataja M, Provatas N. Phys Rev Lett
[71] Amini M, Laird BB. Phys Rev Lett 2006;97:216102. 2004;93:246101.
[72] Hoyt JJ, Asta M, Karma A. Interface Sci 2002;10:181. [121] Barbieri A, Langer JS. Phys Rev A 1989;39:5314.
[73] Laird BB. J Chem Phys 2001;115:2887. [122] Ben Amar M, Brener EA. Phys Rev Lett 1993;71:589.
[74] Perepezko JH. In: Rappaz M, Beckermann C, Trivedi R, editors. [123] Chan S-K, Reimer H-H, Kahlweit M. J Cryst Growth 1976;32:
Solidification processes and microstructures. A symposium in honor 303.
of Wilfried Kurz. Warrendale (PA): TMS; 2004. p. 295. [124] Liu S, Li J, Lee J, Trivedi R. Philos Mag 2006;86:3717.
[75] Greer AL, Quested TE. Philos Mag 2006;86:3665. [125] Haxhimali T. Ph.D. thesis, Northeastern University; 2006.
[76] Kelton KF, Greer AL, Herlach DM, Holland-Moriz D. MRS Bull [126] Nestler B, Danilov D, Galenko R. J Comput Phys 2005;207:221.
2004;29:940. [127] Lee YH. Ph.D. thesis, Northeastern University; 1998.
[77] ten Wolde PR, Ruiz-Montero MJ, Frenkel D. J Chem Phys [128] Mullis AM, Cochrane RF. Acta Mater 2001;49:2205.
1999;110:1591. [129] Boettinger WJ, Warren JA. J Cryst Growth 1999;200:583.
[78] Auer S, Frenkel D. Nature 2001;409:1020. [130] Danilov D, Nestler B. Phys Rev Lett 2004;93:215501.
[79] Auer S, Frenkel D. Annu Rev Phys Chem 2004;55:333. [131] Brener EA, Muller-Krumbhaar H, Temkin DE. Phys Rev E
[80] Trudu F, Donadio D, Parrinello M. Phys Rev Lett 2006;97:105701. 1996;54:2714.
[81] Streitz FH, Glosli JN, Patel MV. Phys Rev Lett 2006;96:225701. [132] Abel T, Brener EA, Muller-Krumbhaar H. Phys Rev E
[82] Desgranges C, Delhommelle J. J Chem Phys 2007;126:054501. 1997;55:7789.
[83] Aga RS, Morris JR, Hoyt JJ, Mendelev M. Phys Rev Lett [133] Pettersen K, Ryum M. Metall Trans A 1989;20:847.
2006;96:245701. [134] Pettersen K, Lohne O, Ryum M. Metall Trans A 1990;21:221.
[84] Huitema HEA, ven der Eerden JP, Janssen JJM, Human H. Phys [135] Henry S, Jarry P, Jouneau P-H, Rappaz M. Metall Mater Trans A
Rev B 2000;62:14690. 1997;28:207.
[85] Wu DT, Granasy L, Spaepen F. MRS Bull 2004;29:945. [136] Henry S, Jarry P, Rappaz M. Metall Mater Trans A 1998;29:2807.
[86] Spaepen F. Solid State Phys 1994;47:1. [137] Henry S, Minghetti T, Rappaz M. Acta Mater 1998;46:6431.
[87] Bai XM, Li M. J Chem Phys 2006;124:124707. [138] Henry S. PhD thesis, EPFL Lausanne; 1999.
[88] Gasser U, Weeks ER, Schofield A, Pusey PN, Weitz DA. Science [139] Semoroz A, Durandet Y, Rappaz M. Acta Metall Mater
2001;292:258. 2001;49:529.
[89] Gránásy L, Börzsönyi T, Pusztai T. Phys Rev Lett 2002;88:206105. [140] Semoroz A. Ph.D. thesis, EPFL Lausanne; 1999.
[90] Gránásy L, Pusztai T. J Chem Phys 2002;117:10121. [141] Koss MB, LaCombe JC, Chait A, Pines V, Zlatkowski M,
[91] Tóth G, Gránásy L. J Chem Phys 2007;127:074709. Glicksman ME, et al. J Cryst Growth 2005;279:170.
[92] Gránásy L, Pusztai T, Saylor D, Warren JA. Phys Rev Lett [142] Diepers H-J, Karma A. In: Rappaz M, Beckermann C, Trivedi R,
2007;98:035703. editors, Solidification processes and microstructures. A symposium
[93] ten Wolde PR, Ruiz-Montero MJ, Frenkel D. J Chem Phys in honor of Wilfried Kurz. Warrendale (PA): TMS; 2004. p. 369–71.
1996;104:9932. [143] Martinez RA, Karma A, Flemings MC. Metall Mater Trans A
[94] Auer S, Frenkel D. J Phys Condens Matter 2002;14:7667. 2006;37:2807.
[95] Browning AR, Doherty MF, Fredrickson GH. Phys Rev E [144] Uehara T, Sekerka RF. J Cryst Growth 2003;254:251.
2008;77:041604. [145] Debierre J-M, Karma A, Celestini F, Guérin R. Phys Rev E
[96] Desgranges C, Delhommelle J. Phys Chem Phys 2007;127:144509. 2003;68:041604.
[97] Wang H, Gould H, Klein W. Phys Rev E 2007;76:031604. [146] Greenwood M, Haataja M, Provatas AN. Phys Rev Lett
[98] Ashkenazy Y, Averback RS. Europhys Lett 2007;79:26005. 2004;93:246101.
[99] Aziz MJ, Kaplan J. Acta Metall 1988;36:2335. [147] Plapp M, Karma A. Phys Rev E 2002;66:061608.
[100] Aziz MJ. Metall Mater Trans A 1996;27:671. [148] Zhang R, Jing T, Jie W, Baicheng L. Acta Mater 2006;54:2235.
[101] Jackson KA, Beatty KM, Gudgel KA. J Cryst Growth [149] Tong Chaohui, Greenwood M, Provatas N. Phys Rev B
2004;271:481. 2008;77:064112.
[150] Gránásy L, Pusztai T, Warren JA, Douglas JF, Börzsönyi T, [194] Shamsuzzoha M, Hogan LM. Cast Metals 1991;4:103.
Ferreiro V. Nat Mater 2003;2:92. [195] Napolitano RE, Meco H, Jung C. J Metals 2004;56:16.
[151] Gránásy L, Pusztai T, Börzsönyi T, Warren JA, Douglas JF. Nat [196] Nogita K, Dahle AK. Mater Characterization 2001;46:305.
Mater 2004;3:645. [197] Dahle AK, Nogita K, Zindel JW, McDonald SD, Hogan LM.
[152] Provatas N, Dantzig JA, Athreya B, Chan P, Stefanovic P, Metall Mater Trans A 2001;32:949.
Goldenfeld N, et al. JOM 2007;59:83. [198] McDonald SD, Dahle AK, Taylor JA, St. John DH. Metall Mater
[153] Wu K-A, Karma A. Phys Rev B 2007;76:184107. Trans A 2004;35:1829.
[154] Karma A, Plapp M. J Metals 2004;56:28. [199] McDonald SD, Nogita K, Dahle AK. Acta Mater 2004;52:4273.
[155] Akamatsu S, Plapp M, Faivre G, Karma A. Metall Mater Trans A [200] Dahle AK, Nogita K, McDonald SD, Dinnis C, Lu L. Mater Sci
2004;35:1815. Eng A 2005;413:243.
[156] Jackson KA, Hunt JD. Trans AIME 1966;236:1129. [201] Shankar S, Riddle YW, Makhlouf MM. Metall Mater Trans A
[157] Langer JS. Phys Rev Lett 1980;44:1023. 2004;35:3038.
[158] Trivedi R, Liu S, Mazumder P, Simsek E. Sci Technol Adv Mater [202] Shankar S, Riddle YW, Makhlouf MM. Metall Mater Trans A
2001;2:309. 2005;36:1613.
[159] Walker H, Liu S, Lee JH, Trivedi R. Metall Mater Trans A [203] Dahle AK, Hillert M. Metall Mater Trans A 2005;36:1612.
2007;38:1417. [204] Dahle AK, Hillert M. Metall Mater Trans A 2006;37:1353.
[160] Akamatsu S, Bottin-Rousseau S, Faivre G. Phys Rev Lett [205] Shankar S, Makhlouf MM. Metall Mater Trans A 2006;37:1353.
2004;93:175701. [206] Eggleston JJ, McFadden GB, Voorhees PW. Physica D 2001;150:
[161] Akamatsu S, Bottin-Rousseau S, Faivre G. Philos Mag 91.
2006;86:3703. [207] Debierre J-M, Karma A, Celestini F, Guerin R. Phys Rev E
[162] Akamatsu S, Bottin-Rousseau S, Perrut M, Faivre G, Witusiewicz 2003;68:041604.
VT, Sturz L. J Cryst Growth 2007;299:418. [208] Suzuki T, Kim SG, Kim WT. Mater Sci Eng A 2007;449–451:99.
[163] Parisi A, Plapp M. Acta Mater 2008;56:1348. [209] Perepezko JH, Hildal K. Philos Mag 2006;86:3681.
[164] Teng J, Liu S, Trivedi R. Acta Mater 2008;56:2819. [210] Gránásy L, Pusztai T, Börzsönyi T, Tóth GI, Tegze G, Warren JA,
[165] Serefoglu M, Napolitano RE. Acta Mater 2008;56:3862. et al. Philos Mag 2006;86:3757.
[166] Liu S, Lee JH, Enlow D, Trivedi R. In: Rappaz M, Beckermann C, [211] Buehler L, Davis SH. J Cryst Growth 1998;186:629.
Trivedi R, editors. Solidification processes and microstructures. A [212] Chen YJ, Davis SH. J Fluid Mech 2000;421:369.
symposium in honor of Wilfried Kurz. Warrendale (PA): TMS; [213] Chen YJ, Davis SH. J Fluid Mech 1999;395:253.
2004. p. 257. [214] Jamgotchian H, Bergeon N, Benielli D, Voge P, Billia B, Guerin R.
[167] Wang N, David S, Walker H, Trivedi R. Trans Indian Inst Metals Phys Rev Lett 2001;87:166105.
2007;60:69. [215] Jamgotchian H, Nguyen-Thi H, Bergeon N, Billia B. Int J Thermal
[168] Trivedi R, Magnin P, Kurz W. Acta Metall 1987;35:971. Sci 2004;43:769.
[169] Akamatsu S, Faivre G. Phys Rev E 2000;61:3757. [216] Noel N, Jamgotchian H, Billia B. J Cryst Growth 1998;187:516.
[170] Han SH, Trivedi R. Unpublished data, Iowa State University, Ames, [217] Lan CW, Tu CY. J Cryst Growth 2000;220:619.
IA; 2008. [218] Lan CW, Yang YW, Chen HZ. Metall Mater Trans A 2002;33:3011.
[171] Hecht U, Witusiewicz VT, Drevermann A, Bottger B, Rex S. Mater [219] Lan CW, Lee MH, Chuang MH, Shih CJ. J Cryst Growth
Sci Forum 2005;508:57. 2006;295:202.
[172] Rex S, Bottger B, Witusiewicz V, Hecht U. Mater Sci Eng A [220] Kauerauf B, Zimmermann G, Murmann L, Rex S. J Cryst Growth
2005;413:249. 1998;193:701.
[173] Witusiewicz VT, Hecht U, Sturz L, Rex S. J Cryst Growth [221] Glicksman ME, Huang S-C. In: Zierip J, Oertl Jr H, editors.
2006;286:137. Convective transport and instability phenomena. Karlsruhe: G.
[174] Witusiewicz VT, Hecht U, Sturz L, Rex S. J Cryst Growth Braun; 1982. p. 557.
2006;286:431. [222] Glicksman ME, Coriell SR, McFadden GB. Annu Rev Fluid Mech
[175] Witusiewicz VT, Hecht U, Sturz L, Rex S. J Cryst Growth 1986;18:307.
2006;297:117. [223] Beckermann C, Diepers HJ, Steinbach I, Karma A, Tong X. J
[176] Fisher DJ, Kurz W. Metall Trans 1974;5:1508. Comput Phys 1999;154:468.
[177] Himemiya T, Umeda T. JIM 1999;40:665. [224] Anderson DM, McFadden GB, Wheeler AA. Physica D
[178] Hecht U, Gránásy L, Pusztai T, Böttger B, Apel M, Witusiewicz V, 2001;151:305.
et al. Mater Sci Eng R-Rep 2004;46:1. [225] Anderson DM, McFadden GB, Wheeler AA. Physica D
[179] Lee JH, Verhoeven JD. J Cryst Growth 1994;144:353. 2000;135:175.
[180] Vandyoussefi M, Kerr HW, Kurz W. Acta Mater 2000;48:2297. [226] Ai X, Shu Y, Li BQ. Comput Math Appl 2006;52:721.
[181] Boettinger WJ. Metall Trans 1974;5:2023. [227] Galenko PK, Funke O, Wang J, Herlach DM. Mater Sci Eng A
[182] Mazumder P, Trivedi R. Phys Rev Lett 2002;88:235507. 2004;375–377:488.
[183] Trivedi R, Park JS. J Cryst Growth 2002;235:572. [228] Galenko PK, Herlach DM, Phanikumar G, Funke O. Mater Sci
[184] Lograsso TA, Fuh BC, Trivedi R. Metall Mater Trans A Forum 2006;508:431.
2005;36:1287. [229] Jeong JH, Dantzig JA, Goldenfeld N. Metall Mater Trans A
[185] Trivedi R, Shin JH. Mater Sci Eng A 2005;413/14:288. 2003;34:459.
[186] Lo TS, Dobler S, Plapp M, Karma A. Acta Mater 2003;51:599. [230] Lan CW, Hsu CM, Liu CC. J Cryst Growth 2002;241:379.
[187] Dobler S, Lo TS, Plapp M, Karma A. Acta Mater 2004;52:2795. [231] Lan CW, Shih CJ. Phys Rev E 2004;69:31601.
[188] Kohler F. Ph.D. thesis, EPFL Lausanne; 2008. [232] Lan CW, Shih CJ. J Cryst Growth 2004;264:472.
[189] Akamatsu S, Moulinet S, Faivre G. Metall Mater Trans A [233] Lu Y, Beckermann C, Ramirez JC. J Cryst Growth 2005;280:320.
2001;32:2039. [234] Narski J, Picasso M. Fluid Dyn Mater Proc 2007;3:49.
[190] Guzik E, Kopycinski D. Metall Mater Trans A 2006;37:3057. [235] Tong X, Beckermann C, Karma A, Li Q. Phys Rev E
[191] Magnin P, Kurz W. Acta Metall 1987;35:1119. 2001;63:061601.
[192] Lasagni F, Lasagni A, Holzapfel C, Muecklich F, Degischer HP. [236] Tonhardt R, Amberg G. J Cryst Growth 2000;213:161.
Adv Eng Mater 2006;8:719. [237] Al-Rawahi N, Tryggvason G. J Comput Phys 2002;180:471.
[193] Napolitano RE, England LG. In: Solidification of aluminum [238] Udaykumar HS, Marella S, Krishnan S. Int J Heat Mass Transfer
alloys. Charlotte (NC): TMS; 2004. 2003;46:2615.
[239] Zhao P, Heinrich JC, Poirier DR. Int J Numer Methods Fluids [282] Badillo A, Beckermann C. Acta Mater 2006;54:2015.
2005;49:233. [283] Nguyen-Thi H, Reinhart G, Mangelinck-Noel N, Jung H, Billia B,
[240] Chakraborty S, Chatterjee D. Phys Lett A 2006;351:359. Schenk T, et al. Metall Mater Trans A 2007;38:1458.
[241] Medvedev D, Kassner K. Phys Rev E 2005;72:56703. [284] Ruvalcaba D, Mathiesen RH, Eskin DG, Arnberg L, Katgerman L.
[242] Medvedev D, Kassner K. J Cryst Growth 2005;275:1495. Acta Mater 2007;55:4287.
[243] Miller W, Rasin I, Succi S. Physica A 2006;362:78. [285] Campanella T, Charbon C, Rappaz M. Metall Mater Trans A
[244] Li D, Li R, Zhang P. Appl Math Model 2007;31:971. 2004;35:3201.
[245] Shin YH, Hong CP. ISIJ Int 2002;42:359. [286] Li T, Lin X, Huang W. Acta Mater 2006;54:4815.
[246] Zhu MF, Lee SY, Hong CP. Phys Rev E 2004;69:061610. [287] Ma D, Jie WQ, Li Y, Ng SC. Acta Mater 1998;46:3203.
[247] Das A, Ji S, Fan Z. Acta Mater 2002;50:4571. [288] Wang WM, Liu JM, Chen XY, Liu ZG. J Cryst Growth
[248] Bouissou Ph, Pelce P. Phys Rev A 1989;40:6673. 2002;240:313.
[249] Schrage DS. J Cryst Growth 1999;205:410. [289] Weiqiang Z, Hua F, Yuansheng Y, Zhuangqi H. J Cryst Growth
[250] Li Q, Beckermann C. J Cryst Growth 2002;236:482. 1998;194:263.
[251] Lee J, Liu S, Miyahara H, Trivedi R. Metall Mater Trans B [290] Regel LL, Wilcox WR, Popov D, Li F. Acta Astronaut 2001;48:101.
2004;35:909. [291] Lee JH, Liu S, Trivedi R. Metall Mater Trans A 2005;36:3111.
[252] Appolaire B, Albert V, Combeau H, Lesoult G. Acta Mater [292] Coriell SR, McFadden GB, Mitchell WF, Murray BT, Andrews JB,
1998;46:5851. Arikawa Y. J Cryst Growth 2001;224:145.
[253] Appolaire B, Albert V, Combeau H, Lesoult G. ISIJ Int [293] Liu S, Trivedi R. Metall Mater Trans A 2006;37:3293.
1999;39:263. [294] Mazumder P, Trivedi R. Appl Math Model 2004;28:109.
[254] Badillo A, Ceynar DL, Beckermann C. J Cryst Growth [295] Mazumder P, Trivedi R, Karma A. Metall Mater Trans A
2007;309:197. 2000;31:1233.
[255] Chen H, Chen YS, Wu X, Tewari SN. J Cryst Growth 2003;253:413. [296] Siquieri R, Emmerich H. Philos Mag Lett 2007;87:829.
[256] Drevet B, Nguyen-Thi H, Camel D, Billia B, Dupouy MD. J Cryst [297] Nestler B, Wheeler AA, Ratke L, Stocker C. Physica D
Growth 2000;218:419. 2000;141:133.
[257] Hui J, Tiwari R, Wu X, Tewari SN, Trivedi R. Metall Mater Trans [298] Ratke L. Metall Mater Trans A 2003;34:449.
A 2002;33:3499. [299] Stocker C, Ratke L. J Cryst Growth 2000;212:324.
[258] Kauerauf B, Zimmermann G, Rex S. Mater Sci Forum 2000;329– [300] Yang S, Liu W, Jia J. J Mater Sci 2001;36:5351.
330:271. [301] Alkemper J, Voorhees PW. J Microsc 2001;201:388.
[259] Ming L, Mori T, Iwasaki H. Mater Sci Eng A 1999;265:217. [302] Alkemper J, Mendoza R, Voorhees PW. Adv Eng Mater 2002;4:
[260] Nguyen-Thi H, Dabo Y, Drevet B, Dupouy MD, Camel D, Billia B, 694.
et al. J Cryst Growth 2005;281:654. [303] Mendoza R, Alkemper J, Voorhees PW. Z Metallkd 2005;96:155.
[261] Spinelli JE, Rosa DM, Ferreira IL, Garcia A. Mater Sci Eng A [304] Ludwig O, DiMichiel M, Salvo L, Suéry M, Falus P. Metall Mater
2004;383:271. Trans A 2005;36:1515.
[262] Trivedi R, Mazumder P, Tewari SN. Metall Mater Trans A [305] Limodin N, Salvo L, Suéry M, DiMichiel M. Acta Mater
2002;33:3763. 2007;55:3177.
[263] Trivedi R, Miyahara H, Mazumder P, Simsek E, Tewari SN. J Cryst [306] Li B, Brody HD, Kazimirov A. Phys Rev E 2004;70:062602.
Growth 2001;222:365. [307] Mathiesen RH, Arnberg L, Ramsøskar K, Weitkamp T, Rau C,
[264] Apel M, Diepers HJ, Steinbach I. In: Gandin CA, Bellet M, editors. Snigirev A. Metall Mater Trans B 2002;33:613.
Modeling of casting, welding and adv. solidification processes – [308] Schenk T, Gastaldi J, Nguyen Thi H, Reinhart G, Mangelinck-Noel
XI. Warrendale (PA): TMS; 2006. p. 505–12. N, Klein H, et al. J Cryst Growth 2005;275:201.
[265] Diepers HJ, Steinbach I. Mater Sci Forum 2006;508:145. [309] Rowenhorst DJ, Kuang JP, Thornton K, Voorhees PW. Acta Mater
[266] Natsume Y, Ohsasa K, Narita T. Mater Trans 2002;43:2228. 2006;54:2027.
[267] Nomoto S, Izumi T, Shiohara Y, Umeda T. J Jpn Inst Metals [310] Krill III CE, Chen LQ. Acta Mater 2002;50:3057.
2001;65:310. [311] Mendoza R, Thornton K, Savin I, Voorhees PW. Acta Mater
[268] Hainke M, Steinbach S, Dagner J, Ratke L, Muller G. Mater Sci 2006;54:743.
Forum 2006;508:199. [312] Kammer D, Voorhees PW. Acta Mater 2006;54:1549.
[269] Hainke M, Steinbach S, Ratke L, Muller G. Trans Indian Inst [313] Smith CS. Trans Metall Soc AIME 1948;175:15.
Metals 2005;58:639. [314] Rappaz M, Jacot A, Boettinger WJ. Metall Mater Trans A
[270] Ratke L, Steinbach S, Muller G, Hainke M, Roosz A, Fautrelle Y, 2003;34:467.
et al. Mater Sci Forum 2006;508:131. [315] Karma A, private communication.
[271] Steinbach S, Ratke L. Mater Sci Eng A 2005;413–414:200. [316] Vernède S, Rappaz M. Philos Mag 2006;86:3779.
[272] Lehmann P, Moreau R, Camel D, Bolcato R. Acta Mater [317] Vernède S, Jarry P, Rappaz M. Acta Mater 2006;53:4023.
1998;46:4067. [318] Rappaz M, Drezet JM, Mathier V, Vernède S. Mater Sci Forum
[273] Lehmann P, Moreau R, Camel D, Bolcato R. J Cryst Growth 2006;519–521:1665.
1998;183:690. [319] Ni J, Beckermann C. Metall Mater Trans B 1991;22:349.
[274] Diepers HJ, Beckermann C, Steinbach I. Acta Mater 1999;47:3663. [320] Stangeland A, Mo A, Nielsen O, Eskin D, M’Hamdi M. Metall
[275] Li X, Ren ZM, Fautrelle Y. J Cryst Growth 2006;290:571. Mater Trans A 2004;35:2803.
[276] Li X, Fautrelle Y, Ren Z. Acta Mater 2007;55:1377. [321] Stangeland A, Mo A, Eskin D. Metall Mater Trans A 2006;37:2219.
[277] Corrigan DP, Koss MB, LaCombe JC, de Jager KD, Tennenhouse [322] Ludwig O, Drezet JM, Martin CL, Suéry M. Metall Mater Trans A
LA, Glicksman ME. Phys Rev E 1999;60:7217. 2005;36:1525.
[278] Giummarra C, LaCombe JC, Koss MB, Frei JE, Lupulescu A, [323] Ludwig O, Drezet JM, Ménésès P, Martin CL, Suéry M. Mater Sci
Glicksman ME. J Cryst Growth 2005;274:317. Eng A 2005;413–414:174.
[279] McCay MH, Hopkins JA, McCay TD. Metall Mater Trans A [324] M’Hamdi M, Mo A, Martin CL. Metall Mater Trans A
1998;29:1137. 2002;33:2081.
[280] Henry S, Gruen GU, Rappaz M. Metall Mater Trans A [325] Campbell J. Castings. Oxford: Elsevier; 2003.
2004;35:2495. [326] Lee PD, Chirazi A, See D. J Light Metals 2001;1:15.
[281] Nguyen-Thi H, Reinhart G, Zhou B, Billia B, Liu Q, Lyubimova [327] Sung PK, Poirier DR, Felicelli SD. Model Simul Mater Sci Eng
TP, et al. J Cryst Growth 2005;275:1579. 2002;10:551.
[328] Péquet C, Gremaud M, Rappaz M. Metall Mater Trans A [367] Habraken L, Brouwer JL. De ferri metallographia. Brussels: Presses
2002;33:2095. Academique Europeenes; 1966.
[329] Stefanescu DM. Int J Cast Metals Res 2005;18:129. [368] Trivedi RK, Laorchan V. Acta Metall 1988;36:1941.
[330] Carlson KD, Lin Z, Beckermann C. Metall Mater Trans B [369] Trivedi RK, Laorchan V. Acta Metall 1988;36:1951.
2007;38:541. [370] Trivedi RK. Scr Metall 1969;3:613.
[331] Wang JS, Lee PD. Int J Cast Metal Res 2007;20:151. [371] Trivedi RK. Acta Metall 1970;18:287.
[332] Piwonka TS, Flemings MC. Trans AIME 1966;236:1157. [372] Purdy GR. Met Sci J 1971;5:81.
[333] Feurer U. Giessereiforschung 1976;28:75. [373] Stevens DE, Purdy GR. Scr Metall 1974;8:323.
[334] Clyne TW, Davies G. J Brit Foundry 1981;74:65. [374] Simonen EP, Trivedi RK. Acta Metall 1977;25:945.
[335] Suyitno, Kool WH, Katgermann L. Metall Mater Trans A [375] Greenwood M, Provatas N, unpublished data.
2005;36:1537. [376] Trivedi RK. In: Aaronson et al., editors. Solid–solid phase trans-
[336] Eskin DG, Suyitno, Katgerman L. Prog Mater Sci 2004;49:629. formation. Warrendale (PA): TMS; 1982. p. 77.
[337] Lahaie D, Bouchard M. Metall Mater Trans B 2001;32:697. [377] Lipton J, Kurz W, Trivedi R. Acta Metall 1987;35:957. 965.
[338] Magnin B, Maenner L, Katgermann L, Engler S. Mater Sci Forum [378] Loginova I, Agren J, Arnberg G. Acta Mater 2004;52:4055.
1996;217–222:1209. [379] Wang Y, Ma N, Chen Q, Zhang F, Chen S-L, Chang YA. JOM
[339] Rappaz M, Drezet JM, Gremaud M. Metall Mater Trans A 2005;57:32.
1999;30:449. [380] Huang CJ, Browne DJ. Metall Mater Trans A 2006;37:589.
[340] Niyama E, Uchida T, Morikawa M, Saito S. AFS Int Cast Metals J [381] Huang CJ, Browne DJ, McFadden S. Acta Mater
1982:52. 2006;54:11.
[341] Drezet JM, Rappaz M. In Light Metals. Warrendale (PA): TMS; [382] Phillion A, Zurob HS, Hutchinson CR, Guo H, Malakhov DV,
2001. p. 887. Nakano J, Purdy GR. Metall Mater Trans A 2004;35:1237.
[342] Mathier V, Drezet JM, Rappaz M. Model Simul Mater Sci Eng [383] Zurob HS, Hutchinson CR, Béché A, Purdy GR, Bréchet YJM.
2007;15:121. Acta Mater 2008;56:2203.
[343] Wang N, Mokadem S, Rappaz M, Kurz W. Acta Mater [384] Kirkaldy JS. Can J Phys 1958;36:907.
2004;52:3173. [385] Hillert M, ‘‘Paraequilibium” Internal report of the Swedish Institute
[344] Phillion AB, Cockcroft SL, Lee PD. Acta Mater 2008;4:55. for Metals Research; 1953.
[345] Zener C. Trans AIME 1947;167:550. [386] Kirkaldy JS, Young DJ. Diffusion in the condensed state. Lon-
[346] Hillert M. Jernkont Annal 1957;141:757. don: Institute of Metals; 1987.
[347] Trivedi RK. Metall Trans 1970;1:921. [387] Shibate H, Aray Y, Suzuki M, Emi T. Metall Mater Trans B
[348] Hillert M. Acta Mater 1999;47:4481. 2000;31:981.
[349] Aaronson HI, Furuhara T, Hall MG, Hirth JP, Nie J-F, Purdy GR, [388] Arai Y, Emi T, Fredriksson H, Shibate H. Metall Mater Trans A
et al. Acta Mater 2006;54:1227. 2005;36:3065.
[350] Chen LQ. Annu Rev Mater Res 2002;32:113. [389] Hillert M. Solid state phase transformations. Warrendale
[351] Elder KR, Katakowski M, Haataja M, Grant M. Phys Rev Lett (PA): TMS-AIME; 1982.
2002;88:245701. [390] Hutchinson CR, Shiflet G. Scr Mater 2004;50:1.
[352] Trivedi R, Kurz W. Int Mater Rev 1994;39:49. [391] Turnbull D. Acta Metall 1955;3:55.
[353] Kurz W. Adv Mater Eng 2001;3:443. [392] Hillert M. Metall Trans 1972;3:2729.
[354] Vandyoussefi M, Kerr HW, Kurz W. Acta Mater 1997;45:4093. [393] Brener EA, Temkin DE. Acta Mater 2003;51:797.
[355] Lima M, Kurz W. Metall Mater Trans A 2002;33:2337. [394] Klinger L, Brechet Y, Purdy GR. Acta Mater 1997;45:4667.
[356] Mullins W, Sekerka RF. J Appl Phys 1964;35:444. [395] Brener EA, Temkin DE. Acta Mater 1999;47:3759.
[357] Sumida M, Kurz W. EPFL, unpublished; 2001. [396] Howe JM, Reynolds WT, Vasudevan VK. Metall Mater Trans A
[358] Pearson DD, Verhoeven JD. Metall Trans A 1984;15:1037. 2002;33:2391.
[359] Verhoeven JD, Pearson DD. Metall Trans A 1984;15:1947. [397] Foiles SM, Hoyt JJ. Acta Mater 2006;54:3351.
[360] Whiting MJ. Scr Mater 2000;43:969. [398] Gandin CA, Bréchet Y, Rappaz M, Canova G, Ashby M, Shercliff
[361] Trivedi RK. Unpublished work. H. Acta Mater 2002;50:901–27.
[362] Carpay FMA. Int Met Rev 1978;23:1. [399] Gandin CA, Jacot A. Acta Mater 2007;55:2539.
[363] Teng J, Liu S, Trivedi R. Acta Mater 2008;56:2819. [400] Thornton K, Poulsen HF, editors. MRS Bull 2008;33:587–629.
[364] Steinbach I, Apel M. Acta Mater 2007;55:4817. [401] Hunziker O, Kurz W. Metall Mater Trans A 1999;30:3167.
[365] Teng J, Ph.D. Thesis, Ames, IA: Iowa State University; 2007. [402] Eiken J, Böttger B, Steinbach I. Phys Rev E 2006;73:066122.
[366] Malcolm JA, Purdy GR. TMS AIME 1967;239:1391. [403] Felberbaum M, Suéry M, Salvo L, Rappaz M, to be published.
View publication stats

Solidification Review

Uploaded by

Copyright:

Available Formats

Solidification Review

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Solidification Review

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Solidiﬁcation microstructures and solid-state parallels: Recent

Article in Acta Materialia · February 2009

Mark Asta Alain Karma

SEE PROFILE SEE PROFILE

W. Kurz Mathis Plapp

SEE PROFILE SEE PROFILE

Oxide on Energy Application View project

Columnar-to-Equiaxed Transition in Solidification Processing (CETSOL) View project

The user has requested enhancement of the downloaded file.

Available online at www.sciencedirect.com

Acta Materialia 57 (2009) 941–971

Overview No. 146

Solidiﬁcation microstructures and solid-state

1. Introduction ﬁcation science advanced considerably through pioneering

942 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 943

atom (expressed as q2/3, where q is the solid atomic 0.08

density): EAAl MENFe PAIRFe

944 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 945

946 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 947

948 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 949

3.2.1. Extremum stiﬀness criterion

950 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 951

952 M. Asta et al. / Acta Materialia 57 (2009) 941–971

behavior [147,169]. A clear inﬂuence of crystallographic

4.1.4. Peritectic coupled growth

M. Asta et al. / Acta Materialia 57 (2009) 941–971 953

growth of both phases [188]. The second mechanism, which

4.2. Irregular eutectic growth

Irregular eutectic structures are developed when a non-

954 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 955

956 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 957

958 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 959

960 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 961

stress. Further details of the rheological behavior of a

6.3. Microporosity formation

Shrinkage microporosity in alloys is a result of a lack of

962 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 963

964 M. Asta et al. / Acta Materialia 57 (2009) 941–971

absolute stability, Va, for this solid-state transformation is

M. Asta et al. / Acta Materialia 57 (2009) 941–971 965

966 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 967

[1] Kurz W, Trivedi R. Acta Metall Mater 1990;38:1.

968 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 969

970 M. Asta et al. / Acta Materialia 57 (2009) 941–971

M. Asta et al. / Acta Materialia 57 (2009) 941–971 971

View publication stats