ISIJ International, Vol. 41 (2001), No. 10, pp.

1076–1082
Review

Recent Advances in the Phase-field Model for Solidification

Machiko ODE, Seong Gyoon KIM1) and Toshio SUZUKI2)

Graduate student, The University of Tokyo, Tokyo 113-8656, Japan. 1) RASOM and Department of Materials Science
and Engineering, Kunsan National University, Kunsan 573-701, Korea. 2) Department of Metallurgy, The University of
Tokyo, Tokyo 113-8656, Japan.
(Received on March 16, 2001; accepted in final form on June 12, 2001)

The recent development of the phase-field models for solidification and their application examples are
briefly reviewed. The phase-field model is firstly proposed for pure material systems and then extended to
binary alloy, multi-phase and multi-component systems theoretically. Though the calculation conditions are
limited due to the sharp interface limit parameters in the early stage, it is widened in the thin interface limit
model. The development of the phase-field model is summarized from a viewpoint of the formulation of
phase-field equation and parameters. The important studies and the latest results such as application exam-
ples of free dendrite growth, directional solidification, Ostwald ripening, interface-particle interaction and
multi-phase simulation are mentioned. Finally future works of the phase-field model are prospected.
KEY WORDS: phase-field model; free dendrite growth; directional solidification; Ostwald ripening; particle–
interface interaction; peritectic reaction; eutectic growth.

is so unrealistic that no application examples are quantita-

1. Introduction
tively comparable to the analytical model or experimental
In the last decade, the capability to predict the solidifica- results. Since the computational efficiency is improved re-
tion microstructure is considerably advanced due to the im- markably and an arbitrary kinetic coefficient is available in
provement of the computer performance and the simulation the thin interface model, three-dimensional calculations and
models. Different techniques are used for the numerical the comparison between the phase-field calculation and that
simulation of microstructure depending on the system size. of solvability theory has become possible. The thin inter-
For example, molecular dynamics and Monte Carlo meth- face limit approach has been also applied to the binary al-
ods are favorable for nano-micro scale simulation, phase- loys19) and the phase-field model is established as a predic-
field models are for micro-meso scale, and cellular automa- tion tool for the solidification microstructure.20)
ta techniques are for meso-macro scale. Since the phase- So far there is a review paper on the recent theoretical
field model satisfies the local equilibrium condition at low development of the phase-field models21) but not on their
interface velocity and reproduced the solute trapping phe- application examples. In this paper, we briefly review the
nomena at high velocity regardless of the complexity of the phase-field models for solidification in section 2 and their
interface shape, it is expected as a powerful tool to simulate application examples such as free dendrite growth, direc-
the complex pattern evolution of the interface quantitative- tional solidification, Ostwald ripening, interface-particle in-
ly. The phase-field model for solidification has been for- teraction and multi-phase-field simulations in section 3.
mulated by Langer1) based on the model C proposed by Finally we mention future works of the phase-field model.
Halperin et al.2) and analyzed mathematically and thermo-
dynamically.3,4)
2. Governing Equations
Kobayashi5,6) showed the first example of the numerical
simulation using a phase-field model in which effects of 2.1. Phase-field Model for Pure Materials
noise and anisotropy on the dendrite shape were qualitative- In the phase-field model, the state of the phase is repre-
ly analyzed. The calculation result was so impressive that sented continuously by an order parameter, the phase-field,
many researchers rushed into the phase-field simulation. f . For example f 51, f 50 and a finite region in which
The phase-field model for binary alloys was soon pro- 0,f ,1 represent solid, liquid and the interface, respective-
posed7) and was extended to the multi-phase8) and multi- ly. The time change of the phase-field is assumed to be pro-
component alloys9) step by step, also the convection effect portional to the variation of the free energy functional.
was introduced to the phase-field model in both pure mater-
ial10–14) and binary alloy15–17) cases. ∂φ δF
5M ................................(1)
At the same time, the pure material model has been theo- ∂t δφ
retically improved to the thin interface limit model.18) In the
conventional sharp interface limit, the calculation condition where M is the phase-field mobility which is related to the

© 2001 ISIJ 1076

 ISIJ International, Vol. 41 (2001), No. 10

driving force for the interface. The Helmholtz free energy thin interface limit to relax the restriction of the interface
functional F has the form: width18) and expressed as follows:
−1
1 2 
∫
2 ε2L  1 δL 
F5  ε ∇φ 1 f (φ , T ) dV .................(2) M −15  1 K ,
V
2  Tm σ  µ k 4 D ⋅ cP 

∫
h(φ )(12 h(φ ))
0
where e is the gradient term coefficient which is related to
the interface energy. It is recognized that the free energy K5 dφ ......................(9)
1 φ (φ 21)
functional has two different contributions. The first term in
the integral has a positive value only in the interface region
and related to the interface energy. The second term repre- The interface width, 1.631025 m is possible at thin inter-
sents the free energy density which is the sum of the free face limit of alloy20) to get experimentally comparable re-
energies of solid, liquid and a double-well potential in the sults. Their thin interface limit model not only improves
interface region. The free energy density is expressed as: computer efficiency but also removes the limitation for the
kinetic coefficient. They showed that the numerical calcula-
f (f ,T)5h(f ) f S1(12h(f )) f L1Wg (f ) ...........(3) tion of the dendrite shape using the phase-field model
agreed well with the solvability theory in both 2-D and
h(f )5f 3(10215f 16f 2) or h(f )5f 2(322f ) ....(4) 3-D.25–27)
g(f )5f 2(12f )2 ..............................(5) Though Karma and Rappel assumed that the thermal
conductivities in solid and liquid were equal, the thin inter-
where W is the height of the double-well potential, f S and face limit model with the unequal conductivity has been
f L are the free energies of the solid and liquid phase, re- also proposed by Almgren.28)
spectively. Then the detailed equation of the phase-field is On the other hand, there is another approach to derive the
given as: phase-field equation. Since the temperature is assumed to
be constant in the free-energy functional based phase-field
∂φ model, it has been pointed out that there is a thermodynam-
5 M [ε 2 ∇ 2φ1h9(φ ){ f L 2 f S }2Wg9(φ )] ........(6)
∂t ic contradiction in the model and the phase-field equation is
derived assuming the positive entropy production.4) In the
In case of pure materials, a thermal diffusion equation is model, the total entropy of the system has been expressed
solved simultaneously. The latent heat generation term is as:
added to the conventional one,
 1 2 
∫
2
∂T L ∂φ
S5 2 ε 0 ∇φ 1 s(φ , e ) dV ..............(10)
5 D∇ 2T 1 h9(φ ) ....................(7) V
 2 
∂t cP ∂ t
where s(f , e) is the entropy density depending on the inter-
where D is the thermal diffusivity, L is the latant heat and cp nal energy density e and f . The phase-field parameters are
is the specific heat. derived similar way in the free energy functional model.22)
In the early phase-field model, the three parameters e , W
2.2. Phase-field Model for Binary Alloys
and M are selected arbitrarily.3–5) By assuming the equilibri-
um condition within the interface T5Tm the phase-field pa- In the first phase-field model for alloys, the interface has
rameters are connected to physical properties.22) The e and been assumed to be a region where the phase state change
W are related to the interface energy and interface width, M from one phase to another gradually. Hence the free energy
is related to the interface kinetic coefficient, m and they are in solid and liquid in Eq. (3) are the forms of composition-
expressed as Ref. 24). weighted mixture of the free energy of solute and solvent as
follows (WBM model).7)
3σ 6 2σδ T2µ f S5cf BS(T)1(12c) f AS(T),
W5 , ε 25 , M5 m .......(8)
2 Tmδ Tm 6 2 Lδ f L5cf BL(T)1(12c) f AL(T) ....................(11)
where s is the interface energy, Tm is the melting point and Due to the sharp interface limit, the first WBM model did
d is the interface width, respectively. Since the parameter M not exhibit the solute trapping. Then the solute gradient
is derived assuming the temperature is constant within the term was introduced into the free energy functional to re-
interface, small calculation mesh size or large undercooling produce the solute trapping phenomena.29) However, it can
is required to neglect the temperature change though the in- be taken into account under the thin interface limit condi-
terface region in the numerical simulation.23) Hence, the in- tion as a relaxation of the solute flux though the interface
terface width should be negligible small comparing to cap- without the solute gradient term.19)
illary length theoretically at the sharp interface limit. For There is another definition for the interface composition.
the computational efficiency, however, the value of the in- Steinbach et al. considered the interface as a fraction-
terface width is sometimes selected as large as possible be- weighted mixture of the solid and liquid with different
yond the restriction. For example, the order of interface compositions and free energy. The compositions within the
width is 1028 m in pure material23) and alloy,24) respectively. interface have been determined to keep the equilibrium par-
Karma and Rappel derived the phase-field mobility in the tition coefficient cS5kecL (0,f ,1).30) Kim et al. improved

1077 © 2001 ISIJ

 ISIJ International, Vol. 41 (2001), No. 10

this idea and the compositions of solid and liquid at the in- 2.3. Multi-phase-field Model and Multi-component
terface are determined to have the same chemical potential Model
(KKS model).19) One of the advantages of the phase-field model is that it
f S5cS f BS(T)1(12cS) f AS(T), is extended to the multi-phase and multi-component sys-
tems straightforwardly. In the multi-phase-field model, each
f L5cL f BL(T)1(12cL) f AL(T), phase is distinguished by each order parameter f 1, f 2,…, f n
c5h(f )cS1(12h(f ))cL, m S(cS(x, t))5m L(cL(x, t)) ......(12) and the order parameter can be considered as a volume
fraction of the phase of the reference number.
Since the free energy density in KKS model corresponds
to the common tangent rule, no excess chemical free energy n

which prevents adopting the large interface width appears. ∑ φ 51 .................................(16)

i =1
i
They have also derived the phase-field mobility in the thin
interface limit for alloys. In the n-phase-field system, the free energy functional is
In case of alloys, a solute diffusion equation is solved si- written as:8)
multaneously.
1
∫ ∑ ∑  2 ε
n i
∂c  D (φ )  F5 2
φi ∇φk 2 φk ∇φi 1
2
5∇ ⋅  c ∇f c  ......................(13) ik
∂t  f cc  V i51 k51

where Dc is the solute diffusivety and fc and fcc are the first 
1 h(φi ) f i (φ , T ) 1Wik p (φi , φk ) dV ............(17)
and the second derivatives of the free energy density. The 
solute diffusion equation assures that the time change of the
solute is proportional to the gradient of the chemical poten- The phase-field parameters are determined by the same way
tial fc, fcc is added to guarantee a constant diffusivities in as in a single phase model.
both the bulk solid and liquid phases. In the multi-component systems, the free energy of solid
Phase-field parameters, e and W, are the same as the ones and liquid are written as:
in the pure material case in the sharp interface limit. The
phase-field mobility in the thin interface limit is also de- n
rived by Kim et al.19) in the same manner as the pure mater-
ial case.
∑ c 51 .................................(18)
i =1
i

n n
ε 2  RT 12k e 1 
M 5 −1

σ  Vm m µ DL 2W
1
ε
ς (cSe , cLe ) , f S5 ∑ ciS f i S (T ), f L 5 ∑c L L
i f i (T )

e i 51 i 51

( µiS5 µiL with in the interface) ...............(19)

ς (cSe , cLe )5 f ccS (cSe ) f ccL (cLe )(cLe 2cSe )2
The cross terms in the solute diffusion equation are still

∫
h(φ0 )[12h(φ0 )] dφ 0
1
under discussion and only a dilute multi-component alloy
3 ⋅ ...(14)
0 [12h(φ0 )] f cc (cS )1h(φ0 ) f cc (cL ) φ0 (12φ0 )
S e L e model is proposed9) and the phase-field mobility in the thin
interface limit is also available with the infinite kinetic co-
Since the mobility is the first-order correction in that of the efficient.
sharp interface limit, the accuracy increase with the larger
ε3 n

∑D
value of the partition coefficient. The above mobility is de- 1
M −15 ζ j (cLe j , cSe j ) ............(20)
rived in the 1-D situation but its definition in 2-D and 3-D σ 2W j 51 ji
is still under active discussion.28, 31)
Strictly speaking, the phase-field equation for alloys There is another attempt to simulate multi-component
should also be derived to assure the positive entropy pro- systems with the phase-field model.33) Grafe et al. introduce
duction with the entropy functional of the form:32) the thermodynamics database such as Thermo-Calc into the
phase-field equation and Dictra software is adapted to de-
 1 2

∫
2 1 2 1 termine the diffusion matrix in the solute field equation.
S5 2 ε 02 ∇φ 1 ε 02c ∇c 1 ε 02e ∇e  1
 2
V 2 2 

 3. Application Examples
1s (φ , e, c ) dV ............................................(15)
 3.1. Free Dendrite Growth
Dendritic crystal growth analysis is popular in the phase-
The thermodynamically consistent model is ideal to adopt field studies and there are lots of application examples for
the open or non-isothermal system theoretically. The free pure materials,18,23,25–27,34–39) pure materials with convec-
energy functional based phase-field equation is, however, tion,10–13) binary alloys24,40,41) and ternary alloys.9)
also applicable in case of metric alloys because the thermal The first numerical simulation using a phase-field model
diffusivity is much larger than the solute one in metal sys- was also the study of the free dendrite growth in a pure
tems and the temperature can be regard as constant locally. material.5,6) Figure 1 shows the first numerical example of

© 2001 ISIJ 1078

 ISIJ International, Vol. 41 (2001), No. 10

Fig. 2. The secondary arm spacing vs. local solidification time in

Al–4.5wt%Cu alloy.41) The calculated arm spacing and
tangent shows good agreement with that of experiment.64)

Fig. 1. The first numerical phase-field simulation. The an-

isotropy is introduced in the driving force for the inter-
face. The amplitude parameter d is (a) 0.2 (b) 0.3 (c) 0.5
respectively.63) The effect of the anisotropy and the com-
petitive growth of the side branches are nicely repro-
duced. (By courtesy of R. Kobayashi)
Fig. 3. The directional solidification of the CBr4–8 mol%C2Cl6
the free dendrite growth using a phase-field model by alloy.20) The {111} plane of the crystal is parallel to the
Kobayashi. plane of the thin film and the k110l direction is parallel to
the pulling direction; (a) V510 m m/s, (b) V515 m m/s and
The anisotropy introduced in gradient term coefficient as
(c) V520 m m/s and G50.83104 K/m. The calculation re-
follows. sults show good agreement with the experimental
data.45–47)
∂φ / ∂ y
ε 5 ε (11ν cos kθ ), tan θ 5
∂φ / ∂ x
3.2. Directional Solidification
where n is the amplitude of the anisotropy and k represents The cellular pattern formation in the directional solidifi-
the k-hold symmetry. Noise is introduced to reproduce cation is also studied using a phase-field model. Losert et
well-developed side branches and added into the phase-field al. studied the range of the imposed perturbation wave-
equation. By varying the value of the phase-field parame- length from singlet to doublet pattern formation in a pure
ters arbitrary, he showed the effect of the anisotropy and material using the thin interface limit model.42) Conti
noise in the shape of the dendrite qualitatively. Wheeler et showed so-called banded structure in rapidly solidified bi-
al. adopted the phase-field parameters which were deter- nary alloys43) and Boettinger et al. showed that the segrega-
mined from physical properties and studied the dendrite tip tion pattern and the breakdown of planar interfaces into
shape more quantitatively. They also pointed out the calcu- cellular structures near the absolute stability interface ve-
lation result changed depending on the interface thickness locity.44) They showed the dynamics of the transition of the
because of the sharp interface limit.23) Then Karma and interface pattern that were observed in the experiment qual-
Rappel showed that phase-field simulation agreed with the itatively. However, there is few quantitative information
prediction of the solvability theory in 2-D and 3-D by about the pattern formation. In other word, though the cal-
proposing the thin interface limit.25–27) culated microstructure looks like the experimental one, the
Thermal noise has been introduced for the quantitative solidification conditions such as temperature gradient and
analysis of side brunch.36) By solving Navier-Stokes equa- interface velocity are restricted. On the other hand, Kim et
tion simultaneously, the convection effect has been also al. showed the calculated cellular pattern which agreed
studied using the thin interface limit model.10,11,13) To re- quantitatively with the experimental data in a binary sys-
duce the computational time, adaptive mesh techniques tem.20) Figure 3 shows that the calculated cellular pat-
have been proposed.34,37) tern using the same condition as the one used in the experi-
Though there are numerical examples of the free dendrite ment.45–47) As pointed out before, their model is not perfect
growth in alloys,24,40) almost all the results are qualitative. thermodynamically but computationally efficient, and gives
There is one example of the prediction of the secondary us quantitative predictions of the microstructure.
arm spacing, which is comparable to the experimental In the directional solidification of alloys, solute-trapping
data41) Figure 2 shows that the calculated arm spacing phenomena is also examined. Since the solute trapping phe-
agrees well with that of the experiment. nomena did not emerged in the first alloy phase-field model

1079 © 2001 ISIJ

 ISIJ International, Vol. 41 (2001), No. 10

Fig. 5. The pushing/engulfment of an alumina prticle by the in-

terface in Fe–0.5 mol%C alloy. The forces acting on the
particle are estimated by the interface shape at each time-
step.66) (a) pushing, (b) engulfment.

Fig. 4. The interaction of the solute fields of each particle during

Ostwald ripening process in Al–4.0wt%Cu alloy.65) The
melting and disappearance of small particles, and the
growth and agglomeration of larger particles are qualita-
tively reproduced. (a) 1.0 s, (b) 20 s, (c) 40 s, (d) 60 s.

due to the sharp interface limit, the solute gradient term

was reconsidered29,48) However, since the chemical potential
gradient exists across the interface under the thin interface
limit condition, the solute trapping is reproduced naturally
without the term49,19) It is shown that the calculation results
agree with the prediction by the Aziz’s equation.
3.3. Ostwald Ripening
Fig. 6. Peritectic transformation during directional solidification
Since the phase-field model can take the interaction be- in Fe–0.36wt%C alloy (Ṫ522 K/s, =T52250 K/cm).
tween the solute fields around each particle into account, it The austenite nucleation is imposed at the dendrite sur-
is very useful to analyze the Ostwald ripening problem. face when the temperature is below the peritectic one.
The austenite growth by consuming ferrite behind the
Figure 4 shows that the solute fields around the particles
dendrite tip is nicely reproduced.67,68) (By courtesy of I.
during the Ostwald ripening. The first Ostwald ripening Steinbach)
analysis adopted very small calculation mesh size and the
volume fraction changed during the coarsening process by
3.5. Application Example Using Multi-phase-field
the curvature effect.50) Diepers et al. studied the convection
Model
effect on the Ostwald ripening process qualitatively.
Vaithyanathan and Chen studied the difference in the simu- So far only the multi-phase-field model can simulate the
lation result between 2-D and 3-D simulation using the complex microstructure of the eutectic,8,54–56) peritec-
phase-field model for solid solution.51) tic8,54,55,57,58) and monotectic alloys.17) The idea of the non-
single phase-field model has been proposed by Wheeler et
3.4. Interface-particle Problem al. for eutectic growth56) and Steinbach et al. for multi-
It is difficult to simulate the interface shape around an phase-field model8) at almost the same time. The various
insoluble particle with the conventional shape interface eutectic lamella structure is reproduced54,55) and the peritec-
model. Even if local equilibrium is assumed, the interface tic phase growth behind the dendrite tip region is nicely
shape is still remained to be undetermined. The first analy- reproduced.58) Figure 6 shows that the peritectic phase
sis for the interface shape around the insoluble particle in growth during the free dendrite growth and Fig. 7 shows
alloy is studied by Ode et al. Then they showed the time that the rod structure growth. In the multi- phase field
history of the interface shape during the particle pushing/ model the quantitative comparsion with the experimantal
engulfent process and estimate the critical velocity for par- data also becomes possible. Figure 8 shows eutectic lamel-
ticle pushing/engulfment.52,53) Figure 5 shows that time his- la structure changes depending on the composition.
tory of the particle pushing/engulfment by the interface of Multi-phase-field models are also applied to the grain
Fe–C alloy. growth analysis8,59) Each phase-field represents the each
grain with a specific orientation. However, a single phase-
field model with an orientation contribution term in the
free energy functional, so-called vector-valued phase-field

© 2001 ISIJ 1080

 ISIJ International, Vol. 41 (2001), No. 10

4. Concluding Remarks
The recent developments of the phase-field models and
their application examples are briefly reviewed. So far, no
other method can be compared with phase-field models for
simulating the complex interface evolution in not only pure
materials but also in multi-phase and multi-component sys-
tems. Since there are application trials for solid transforma-
tion, evaporation and electrodeposition processes, the
phase-field models will be applied to almost all the forma-
tion processes of microstructure in the near future.
The key point of the phase-field model for solidification
will be the quantitative prediction of the microstructure for
industrial materials. The multi-phase and multi-componet
models should be extended to simulate the microstructure
evolution from the crystallization of primary and secondary
phases to the solid state transformation continuously and
quantitatively. Since the basic idea has been already accom-
plished, treatment of nucleation, coupling with a thermody-
namic database and timesaving techniques should be inves-
tigated in detail. There are a few models to produce a new
phase seed such as coupling with the classic nucleation the-
ory, however, it needs more examination to obtain reliable
results. The trial to couple the thermodynamic database
with the phase-field calculation has started recently and it is
necessary to widen the scope of the application to industrial
Fig. 7. The rod like eutectic structure in 3-D calculation by a multi-phase and multi-component alloys. Now, the most se-
multi-phase-field model (353353100 m m). (a) 0 s, (b) rious problem in a practical use of the phase-field model is
0.5 s, (c) 1.5 s, (d) 2.5 s.69) (By courtesy of I. Steinbach)
the computational time. Timesaving algorithm and parallel
computing technique are expected for the further applica-
tion.
Acknowledgements
The authors are grateful to R. Kobayashi of Hokkaido
University and I. Stainbach of ACCESS e.V. for their fig-
ures.

REFERENCES
1) J. S. Langer: Directions in Condensed Matter Physics, Models of
Pattern Formation in First-order Phase Transition, (1986), 165.
2) B. I. Halperin, P. C. Hohenberg and S. Ma: Phys. Rev. B, 10 (1974),
139.
3) G. Caginalp: Phys. Rev. A, 39 (1989), 5887.
4) O. Penrose and P. C. Fife: Physica D, 43 (1990), 44.
5) R. Kobayashi: Bull. Jpn. Soc. Ind. Appl. Math., 1 (1991), 22.
6) R. Kobayashi: Physica D, 63 (1993), 410.
7) A. A. Wheeler, W. J. Boettinger and G. B. McFadden: Phys. Rev. A,
45 (1992), 7424.
8) I. Steinbach, F. Pezzolla, B. Nestler, M. Seeselberg, R. Prieler, G. J.
Schmitz and J. L. L. Rezende: Physica D, 94 (1996), 135.
9) M. Ode, J. S. Lee, S. G. Kim, W. T. Kim and T. Suzuki: ISIJ Int., 40
(2000), 870.
10) R. Tonhardt and G. Amberg: Phys. Rev. E, 62 (2000), 828.
11) R. Tonhardt and G. Amberg: J. Cryst. Growth, 213 (2000), 161.
Fig. 8. The lamellar eutectic growth in CBr4–C2Cl6 alloy, (a) 12) R. Tonhardt and G. Amberg: J. Cryst. Growth, 194 (1998), 406.
steady tilt pattern, (b) 2-l oscillatory pattern, (c) 1-l os- 13) X. Tong, C. Beckermann and A. Karma: Modeling of Casting,
cillatory tilted pattern, (d) 2-l oscillatory tilted pattern. Welding and Advanced Solidification Processes VIII, TMS,
The real phase-diagram data is used to determine the Warrendale, PA, (1998), 613.
phase-field parameters and the calculation conditions are 14) D. M. Anderson, G. B. McFadden and A. A.Wheeler: Physica D,
the same in the experiment.70) 135 (2000), 175.
15) H.-J. Diepers, C. Beckermann and I. Steinbach: Acta mater., 47
(1999), 3663.
model, is also proposed for grain boundary analysis. The
16) H.-J. Diepers, C. Beckermann and I. Steinbach: Proc. Modeling of
vector-valued phase-field model can reproduce grain orien- Casting, Welding and Advanced Solidification Processes VIII, TMS,
tation arbitrary and the grain coalescence.60–62) Warrendale, PA, (1998), 565.
17) B. Nestler, A. A. Wheeler, L. Ratke and C. Stocker: Physica D, 141

1081 © 2001 ISIJ

 ISIJ International, Vol. 41 (2001), No. 10

(2000), 133. 48) W. J. Boettinger, A. A. Wheeler, B. T. Murray and G. B. McFadden:

18) A. Karma and W.-J. Rappel: Phys. Rev. E, 53 (1996), R3017. Mater. Sci. Eng., A178 (1994), 217.
19) S. G. Kim, W. T. Kim and T. Suzuki: Phys. Rev. E, 60 (1999), 7186. 49) N. A. Ahmad, A. A. Wheeler, W. J. Boettinger and G. B. McFadden:
20) S. G. Kim, W. T. Kim and T. Suzuki: Proc. the Science of Casting Phys. Rev. E, 58 (1998), 3436.
and Solidification Conf., Hobbystar S.R.L., Brasov, (2001), 29. 50) J. A. Warren and B. T. Murray: Modelling Simul. Mater. Sci. Eng., 4
21) A. A. Wheeler, N. A. Ahmad, W. J. Boettinger, R. J. Braun, G. B. (1996), 215.
McFadden and B. T. Murray: Adv. Space Res., 16 (1995), 163. 51) V. Vaithyanathan and L. Q. Chen: Scripta mater., 42 (2000), 967.
22) S.-L. Wang, R. F. Sekerka, A. A. Wheeler, B. T. Murray, S. R. 52) M. Ode, J. S. Lee, S. G. Kim, W. T. Kim and T. Suzuki: ISIJ Int., 39
Coriell, R. J. Braun and G. B. McFadden: Physica D, 69 (1993), 189. (1999), 149.
23) A. A. Wheeler, B. T. Murray and R. J. Schaefer: Physica D, 66 53) M. Ode, J. S. Lee, S. G. Kim, W. T. Kim and T. Suzuki: ISIJ Int., 40
(1993), 243. (2000), 153.
24) J. A. Warren and W. J. Boettinger: Acta metall. mater., 43 (1995), 54) B. Nestler and A. A. Wheeler: Physica D, 138 (2000), 114.
689. 55) M. Seesselberg, J. Tiaden, G. J. Schmitz and I. Steinbach: Proc. the
25) A. Karma and W.-J. Rappel: Phys. Rev. Lett., 77 (1996), 4050. 4th Decennial Int. Conf. on Solidification Processing, Univ. of
26) A. Karma and W.-J. Rappel: J. Cryst. Growth, 174 (1997), 54. Sheffield, Sheffield, (1997), 440.
27) A. Karma and W.-J. Rappel: Phys. Rev. E, 57 (1998), 4323. 56) A. A. Wheeler, G. B. McFadden and W. J. Boettinger: Proc. R. Soc.
28) R. F. Almgren: SIAM J. Appl. Math., 59 (1999), 2086. Lond., A452 (1996), 495.
29) A. A. Wheeler, W. J. Boettinger and G. B. McFadden: Phys. Rev. E, 57) G. J. Schmitz and B. Nestler: Mater. Sci. Eng., B53 (1998), 23.
47 (1993), 1893. 58) J. Tiaden: J. Cryst. Growth, 198–199 (1999), 1275.
30) J. Tiaden, B. Nestler, H. J. Diepers and I. Steinbach: Physica D, 115 59) M. K. Venkitachalam, L.-Q. Chen, A. G. Khachaturyan and G. L.
(1998), 73. Messing: Mater. Sci. Eng., A238 (1997), 94.
31) G. B. McFadden, A. A. Wheeler and D. M. Anderson: Physica D, 60) R. Kobayashi, J. A. Warren and W. C. Carter: Physica D, 119 (1998),
144 (2000), 154. 415.
32) Z. Bi and R. F. Sekerka: Physica A, 261 (1998), 95. 61) J. A. Warren, R. Kobayashi and W. C. Carter: J. Cryst. Growth, 211
33) U. Grafe, B. Boettger, J. Tiaden and S. G. Fries: Scripta mater., 42 (2000), 18.
(2000), 1179. 62) R. Kobayashi, J. A. Warren and W. C. Carter: Physica D, 140 (2000),
34) N. Provatas, N. Goldenfeld and J. Dantzig: Proc. Solidification, 141.
TMS, Warrendale, PA, (1999), 151. 63) R. Kobayashi: J. Jpn. Assoc. Cryst. Growth, 18 (1991), 209.
35) Y.-T. Kim, N. Provatas, N. Goldenfeld and J. Dantzig: Phys. Rev. E, 64) T. F. Bower, H. D. Brody and M. C. Flemings: Trans. TMS-AIME,
59 (1999), R2546. 236 (1966), 624.
36) A. Karma and W.-J. Rappel: Phys. Rev. E, 60 (1999), 3614. 65) M. Ode, S. G. Kim, W. T. Kim and T. Suzuki: Proc. Processing
37) R. J. Braun and B. T. Murray: J. Cryst. Growth, 174 (1997), 41. Materials for Properties 2, TMS, Warrendale, PA, (2000), 1065.
38) S.-L. Wang and R. F. Sekerka: Phys. Rev. E, 53 (1996), 3760. 66) M. Ode, J. S. Lee, T. Suzuki, S. G. Kim and W. T. Kim: Proc.
39) B. T. Murray, A. A. Wheeler and M. E. Glicksman: J. Cryst. Growth, Modeling of Casting & Solidification Processes 1999, CAMP, Seoul,
154 (1995), 386. (1999), 125.
40) J. F. McCarthy: Acta mater., 45 (1997), 4077. 67) J. Tiaden and U. Grafe: Proc. Int. Conf. on Solid–Solid Phase
41) M. Ode, S. G. Kim, W. T. Kim and T. Suzuki: ISIJ Int., 41 (2001), Transformation ’99 (JIMIC-3), JIM, Sendai, (1999), 737.
345. 68) J. Tiaden, U. Grafe and I. Steinbach: Proc. Cutting Edge of
42) W. Losert, D. A. Stillman, H. Z. Cummins, P. Kopezynski, W.-J. Computer Simulation of Solidification and Casting, ISIJ, Tokyo,
Rappel and A. Karma: Phys. Rev. E, 58 (1998), 7492. (1999), 13.
43) M. Conti: J. Cryst. Growth, 198–199 (1999), 1251. 69) M. Apel, B. Boettger, H.-J. Diepers and I. Steinbach: Abs. ICCG 13,
44) W. J. Boettinger and J. A. Warren: J. Cryst. Growth, 200 (1999), 583. IOCG Kyoto, (2001), 173.
45) S. Akamatsu, G. Faiyre and T. Ihle: Phys. Rev. E, 51 (1995), 4751. 70) M. Ginibre, S. Akamatsu and G. Faivre: Phys. Rev. E, 56 (1997),
46) S. Akamatsu and T. Ihle: Phys. Rev. E, 56 (1997), 4479. 780.
47) S. Akamatsu and G. Faiyre: Phys. Rev. E, 58 (1998), 3302.

© 2001 ISIJ 1082