enabling iCMe Research Summary

advances in Multi-scale
Modeling of solidification
and Casting processes
Baicheng Liu, Qingyan Xu, Tao Jing, Houfa Shen, and Zhiqiang Han

The development of the aviation, Recently, multi-scale modeling in- field method has become increasingly
energy and automobile industries re- cluding macro- and micro-modeling popular because of its advantages in
quires an advanced integrated prod- has become a hot topic in computa- morphology description—the den-
uct/process R&D systems which could tional materials engineering. Hence, dritic morphology plays a key role in
optimize the product and the process macro- and micro-modeling of so- the final performance of cast Mg alloy
design as well. Integrated computa- lidification and casting processes have components. The phase field method
tional materials engineering (ICME) been extensively studied in our re- is based on Ginzburg-Landau theory,
is a promising approach to fulfill this search group, and some advances in and reflects the influence of diffusion,
requirement and make the product and macro- and micro-simulation of mag- ordering potential and thermodynamic
process development efficient, eco- nesium and aluminum alloy castings, driving force by using the differential
nomic, and environmentally friendly. macrosegregation of large steel ingots, equations. However, most previous
Advances in multi-scale modeling of and superalloy directionally solidified research focused on metals with face-
solidification and casting processes, turbine blade casting are reported in centered cubic (f.c.c.) structure, such
including mathematical models as the paper. as Al, Cu and Ni, while metals with
well as engineering applications are hexagonal-close-packed (hcp) struc-
Modeling of dendrite
presented in the paper. Dendrite mor- ture, such as Mg and Zn, were rarely
Morphology of Cast
phology of magnesium and aluminum referred to. Experimental results in-
Mg alloy by the phase
alloy of solidification process by us- dicate that the dendritic morphology
field Method
ing phase field and cellular automa- of metals with hcp structure is totally
ton methods, mathematical models of Recently, numerical simulation for different from that with f.c.c. structure
segregation of large steel ingot, and microstructural evolution by phase because of the different crystal lat-
microstructure models of unidirec- tice.1–7
tionally solidified turbine blade cast- How would you… Mathematical Models
ing are studied and discussed. In ad- …describe the overall significance
dition, some engineering case studies, of this paper? Phase-Field Equations
including microstructure simulation Dendrite growth and microstructure
of aluminum casting for automobile during solidification process of The total free energy of the two-
industry, segregation of large steel super-alloy turbine blade casting, phase system is described by a phe-
aluminum and magnesium alloy nomenological Ginsburg-Landau
ingot for energy industry, and micro- casting, and large steel ingot are
structure simulation of unidirection- extensively by experiments and model as Equation 1, where F is the to-
ally solidified turbine blade castings Cellular Automaton Method. tal free energy of the system, f(ϕ,u,c)
for aviation industry are discussed. is the free energy density function, φ
…describe this work to a
materials science and engineering is the phase field variable parameter
introduCtion professional with no experience in ranging from negative one in the liq-
your technical specialty? uid to positive one in the solid, u and
Materials processing technologies
play an important role for the manu- This work is a general view of multi- c are the temperature and concentra-
facturing industry, for example, alu- scale modeling and simulation of tion, respectively, and γ(n) and δ(n)
solidification and casting processes, are the gradient energy and concentra-
minum and magnesium alloy castings including single crystal turbine blade
for the automobile industry, superal- casting, aluminum and magnesium tion field coefficient, respectively. (All
loy unidirectionally solidified turbine alloy castings, and large/heavy steel equations can be found in the table on
ingot. page 22.) The governing equations
blade castings for the aviation indus-
try, etc. On the other hand, modeling …describe this work to a layperson? based on thermodynamic theory for
and simulation can play a significant the phase-field coupled with tempera-
The work is to predict process-
role in guaranteeing the quality of microstructure-property of important ture field and solute field can then be
these castings, to shorten the R&D and critical castings, and hence to expressed as Equation 2. In this equa-
save R&D time and cost as well. tion, μ(n) is the kinetic coefficient, MC
time and to decrease the R&D cost.

Vol. 63 No. 4 • JOM www.tms.org/jom.html 19

 is a concentration mobility parameter
and is set as shown in Equation 2a. In
Slice
this equation, h(ϕ) is the solid fraction
Mg-matrix given by φ2 (3–2φ), DT is the thermal
diffusion coefficient, L is the latent
heat.
Thermodynamics Descriptions of Alloy
c System
a
b The free energy density function of
50 mm
the system consists of the free energy
Figure 1. Simulated resulst of dendrite morphology of magnesium alloy microstructures of the bulk phases and an imposed par-
(Mg-9wt.%Al). (a) Three-dimensional morphology; (b) synchrotron-radiation-based x-ray
tomography slice extracted showing hcp phase patter; (c) related sectional patter results abolic potential which is given and can
from 3-D simulations. be written as Equation 3, where g(φ)
– f2(1–φ)2， and WA and WB are the
Al wt.%
height of the parabolic potential which
16
will be determined as Equation 4.
13
11 The Anisotropic Functions of Interface
8 Energy and Mobility
6 An anisotropic function of interfa-
3 cial free energy and mobility, which
reflects underlying crystalline charac-
teristics of the hexagonal close packed
lattice, is proposed, as shown in Equa-
a b tions 5 and 6, where ni is unit vector
and is given as shown in Equation 5a
and i represents x, y, z, respectively, g0
and m0 are , respectively, the mean val-
ues of interface free energy and kinetic,
e1, e2, and x1, x2 are, respectively, the
first- and second-order anisotropic pa-
rameters of crystal-melt interface free
energy and mobility.
Numerical Simulation Results
and Discussion
c d 0.1 mm To yield well-developed three-
Figure 2. (a,b,c) Simulated and (d) experimental results of Mg alloy dendrites. (a) Simu- dimensional (3-D) dendritic mor-
lated 2-D equiaxed dendrite; (b) simulated 3-D equiaxed dendrite at early stage; (c) fully phologies, phase-field simulations
developed 3-D equiaxed dendrite; (d) metallographic results showing texture of Mg alloy of equiaxed dendritic solidification
dendrites.
growth in undercooled melts were
carried out. A nucleus was located in
Al wt.%
undercooled melts and an equiaxed so-
13 lidification process, governed by both
Al wt.%
11 13
phase-field equation accompanied with
12 associated temperature and solute field
9 10
9 equations, was carried out.
7 8
6
Based upon two anisotropic param-
5 5 eters, phase-field calculations yield the
3
3 entire dendritic solid-liquid interfacial
morphologies as shown in Figure 1a,
corresponding to the pattern of hexago-
nal dipyramid. X-ray tomography im-
ages based on synchrotron radiation of
a b c Figure 1b and the related two-dimen-
Figure 3. Simulated columnar dendrite growth of Mg alloy in 2-D and 3-D. (a) Columnar sional (2-D) sectional pattern of Figure
dendrites in 2-D; (b) columnar dendrites in 3-D; (c) section view showing the solute field at
X-Y, Y-Z, and X-Z planes.
1c also supply the proof of the possibil-
ity of three-dimensional results.

20 www.tms.org/jom.html JOM • April 2011

 Modeling of dendrite
Morphology of Cast
Mg alloy by Cellular
autoMaton Method
Mathematical models based on the
cellular automaton (CA) method are
being used in micro-scale modeling of
the dendrite growth, which can quanti-
tatively reproduce most of the dendritic a P1 1.74 mm P2 1.50 mm P3 1.25 mm
1 mm
features with affordable computational
costs. Böttger and Eiken et al.8 simu-
lated the dendrite morphology of Mg
alloy using phase field (PF) method
in two and three dimensions (2-D and
3-D). Huo et al.9 simulated the 2-D and
3-D as-cast microstructure of AZ91D
by using a modified cellular automa-
ton (CA) method, where a prescribed
b P1 1.72 mm P2 1.55 mm P3 1.28 mm
dendrite profile was used instead of 1 mm
physics based simulation considering Figure 4. Comparison of simulated and experimental results of step-shaped casting (grain
the solid-liquid interface anisotropy of size). (a) Simulated results (b) experimental results.
Mg. Yin et al.10 simulated the 2-D den-
drite morphology of AZ91 Mg alloy
using CA - finite element (FE) model
under hexagonal mesh.
In this study 2-D and 3-D CA mod- P6
els have been developed to simulate the
P7 P5
dendrite morphology evolution of cast
Mg alloy and the experiments were
carried out for validating the models. P1 P2 P3 P4

Mathematical Models a b

The present CA model employs a Figure 5. Schematic of an aircraft turbine wheel casting. (a) Aircraft turbine wheel casting;
scheme based on two sets of mesh. (b) specimen positions.
In the 2-D model, a hexagonal mesh
is used to perform CA capture proce-
dure to reflect the texture of Mg den-
drites, and an orthogonal mesh is used
to solve diffusion equations. In the 3-D
model, the CA calculation is performed
using a mesh defined by the hcp crystal
lattice, and the diffusion is solved us-
ing a cubic mesh. The growth kinetics
of solid-liquid interface is determined a 100 mm 100 mm 100 mm
based on the difference between the P1 0.118 mm P2 0.169 mm P3 0.270 mm
local equilibrium liquid composition
and the local actual composition ob-
tained by solving the solute diffusion
equation.11 The solid-liquid interface
curvature and growth anisotropy are
also considered in the present model.
Details of the model can be found in
the literature.12,13
b
Simulation Results 100 mm 100 mm 100 mm
P1 0.124 mm P2 0.177 mm P3 0.281 mm
The simulated morphology of
Figure 6. (a) Simulated and (b) experimental microstructure of an aircraft turbine wheel
equiaxed and columnar dendrites of casting (grain size).
AZ91D Mg alloy is shown in Figures

Vol. 63 No. 4 • JOM www.tms.org/jom.html 21

 Equations could be used to study the directional
solidification process and optimize the
γ 2 (n) δ2 (n) process parameters.
F = ∫V [f(φ,u,c) − | ∇φ |2 − | ∇c |2 ]dV (1)
2 2
Physical and Mathematical
∂φ δF ∂c δF  ∂T ∂h (φ ) Models

m (n ) =− , = ∇ M c (1 − c ) ∇  , = D ∇ 2 T + L (2)
∂t δφ ∂t  c δc  ∂t T ∂t The schematic of the directional
solidification process of turbine blade
 S L (
D + h (φ ) D − D  RT
S  V ) (2a) casting is shown in Figure 7. The metal
pouring and solidification process take
M
place in the directional solidification
hcp
f (φ,T,c ) = h (φ ) f
sub − reg
L
(T,c ) + 1 − h (φ ) fsub − reg
(T,c ) + (1 − c ) W A + cWB  g (φ ) (3) furnace under vacuum environment.
Macro-heat Transfer
3γ A 3 γB
WA = ,WB = (4) Complex heat transfer exists during
2T A δA 2T BδB the directional solidification process.
M M
The macro-temperature distribution
 
γ (n ) = γ 1 + ε  n 6 − 15n 4n 2 + 15n 2n 4 − n 6  + ε n 6  (5) within the casting and shell was cal-
0 1 x x y x y y  2 z  culated according to the transient non-
linear heat transport equation as shown
2 2 2
∂φ  ∂φ   ∂φ   ∂φ  in Equation 7, where T is the tempera-
n =  ∂x  +  ∂y  +  ∂z  (5a)
i ∂i       ture, t is the time, ρ is the density, c is
the specific heat, L is the latent heat, λ
 
m (n ) = m 1 + ζ  n 6 − 15n 4n 2 + 15n 2n 4 − n 6  + ζ n 6  (6) is the heat conductivity; x, y and z are
0 1 x x y x y y  2 z  the coordinates; fs is the mass fraction
of solid phase; QR is the heat radiation
∂T  ∂2T ∂2T ∂2T  ∂f
ρc = λ  2 + 2 + 2  + ρL s + QR (7) exchange between the shell surface and
∂t  ∂x ∂y ∂z  ∂t
the furnace wall.
The huge number of memories re-
quired to calculate the view factors for
2 and 3. It can be seen that the pres- simulate the microstructure evolution each surface cell against others makes
ent model reproduces the Mg dendrites of an aircraft turbine wheel casting of it extremely complicated and difficult
pretty well and provides details about Al-7wt.%Si alloy. The casting shape for the heat radiation calculation in di-
texture, solute segregation, and growth and the position of the specimens are rectional solidification. In this paper,
anisotropy and competition. schematically shown in Figure 5, and a numerical method derived from the
the thinnest part of the casting is only 2 Monte Carlo method was proposed to
MiCrostruCture
mm in thickness. Specimens were tak- compute the heat radiation in direction-
Modeling of aluMinuM
en from the longitudinal cross-section al solidification, with much less memo-
alloy Casting by
of the wheel casting indexed as P1 to ry required and a higher accuracy. The
the Modified Cellular
P7 in sequence. The comparison be- details can be seen in Reference 15.
autoMaton Method
tween the simulated and experimental The microstructure simulation was
Besides dendrite morphology, microstructure of different specimen is based on the modified CA method.16 A
the microstructure evolution of Al- shown in Figure 6. continuous nucleation mode17 was em-
7wt.%Si alloy was also studied by ployed to calculate the nucleus number
MiCrostruCture
MCA method.14 A step-shaped sample in the undercooled. The growth speed
Modeling of single
casting to verify the model and then of the dendrite tip could be calculated
Crystal turbine blade
an engineering case study were carried based on the KGT model.18
Casting by MCa
out. The comparison between the simu-
Layer by Layer Calculation Method
lated and experimental microstructure Unidirectionally solidified and sin-
for Microstructure Simulation of
from specimens at different position gle crystal Ni-based superalloy turbine
Whole Turbine Blade
s(20, 40 and 60 mm in thickness) of the blade castings produced by Bridgman
step-shaped sample casting is shown in directional solidification technology Considering the characteristic of di-
Figure 4. As shown in Figure 4, grain are currently used in both the aero- rectional solidification, a layer by layer
size and secondary dendrite arm space nautic and energy industries as key method was proposed to calculate the
become smaller when the step thick- parts of the gas turbine engines. The microstructure formation of a single
ness decreases. final microstructure of the blade cast- crystal turbine blade.19 The mushy zone
For an engineering case study, a ing directly determines the casting included several layers in longitudinal
modified CA model coupled with fi- mechanical properties. As a powerful direction in macro-scale. We divided
nite difference method was used to tool, numerical simulation technology every layer into many cells further at

22 www.tms.org/jom.html JOM • April 2011

microscale. At a macro-time step, the experiment with varying withdrawal bon segregation in a 360-t steel ingot,
microstructure evolution in the micro- rates is shown in Figure 11. No stray despite the MP process as well as other
computational domain was calculated. grains were found at the edge of the modern technologies of metallurgical
Finally modeling results of all layers platform both in the simulated and ex- production was used.25
were combined together to form that of perimental results. This paper presents some simulated
the whole turbine blade. results based on recently developed
Grain Competition and Evolution
Each grain’s orientation was ran- macrosegregation model.27 The model
in the Grain Selector
domly determined in this model. was first validated by using a bench-
Therefore, the grains which were not The grain evolution process within mark test, and then used to simulate
well-aligned with respect to the maxi- the seed selector during solidification the macrosegregation in a 300-ton MP
mum gradient of the temperature field at different times is shown in Figure 12. steel ingot.
would grow at a much slower speed The liquid is not displayed so that the The model is based on the con-
than those which were best aligned, solid-liquid interface could be clearly tinuum theory, and it involves a fully
which made them grow behind the shown. It is indicated that a great num- coupled numerical solution of mass,
growth front. This made it possible for ber of tiny equiaxed grains emerged at momentum, energy, and species con-
those well-aligned grains to have their the bottom surface of the starter block, servation equations in the liquid, solid,
secondary and tertiary side arms grow and transferred into a much smaller and mushy zones. The thermosolutal
out and occupy the space just in front number of columnar grains when convection and the induced macrose-
of the less well-aligned grains. The growing upwards, which is well known gregation in large steel ingots can be
above process can be used to describe as the grain competition process deter- numerically simulated by the model. It
the competitive grain growth in the mined by the heat flux direction and the should be mentioned that multiphase
starter block of grain selector.20 grain’s fastest growth direction. The solidification models are available in
grain number decreased quickly in the the literature that take into account
Simulation Results of Single
starter block, leaving only less than ten melt convection and grain sedimenta-
Crystal Turbine Blade Castings
columnar grains growing into the spi- tion (for example, References 28 and
In order to investigate the influence ral. The grains in the spiral continued 29). The application of these sophisti-
of varying withdrawal rates on the to decrease when growing upwards, cated models to large industrial ingots,
casting microstructure, both experi- and only one of them survived from however, is limited due to the large
ments and simulation of two groups this grain selection, which finally grew computational resources required to
of Ni-based superalloy blade castings into a whole single crystal casting. The resolve the variety of the phenomena
with different processing parameters measuring experiments from S1 to S7 over the process scale.
(Group A and B) were carried out. The sections were compared quantitatively The validation of the model was per-
simulated microstructure evolution of with the simulation, which shows a rea- formed on the well-known Hebditch–
blade casting at different time in group sonable correspondence.20 Hunt (HH) benchmark experiment.30
A, with a constant withdrawal rate of In this experiment, macrosegregation
Modeling and siMulation
7.0 mm/min. is shown in Figure 8. It resulting from the solidification of a
of MaCrosegregation
could be seen that stray grains existed Pb-48wt.%Sn alloy in a rectangular
of a large steel ingot
at the edge of the platform of the blade. cavity was measured. The thermo-
The simulated microstructure evolu- The production of large steel ingots physical data and parameters used in
tion of blade casting at different time in with improved chemical homogeneity the calculation were identical to the
group B, with varying withdrawal rates is of great concern for steelmakers to ones used in Reference 31 an exist-
is shown in Figure 9. It could be seen meet stringent requirements for high ing benchmark simulation. Figure 13
that no stray grains were found at the quality and large integrated forgings shows the segregation ratio of tin in
platform of the blade casting, and the typically used for pressure vessels of the solidifying cavity. It can be ob-
produced casting was a whole single nuclear power plants. Macrosegrega- served that the present predictions
crystal super lloy blade. tion, as one of the main defects in large compare well with the corresponding
The microstructures in different sec- steel ingot, refers to chemical inhomo- results in Reference 31. Furthermore,
tions of the whole blade of group A by geneity in cast metals at the scale of the the predicted segregation maps at the
experiment with a constant withdrawal product. It cannot be mitigated through end of solidification are generally in
rate is shown in Figure 10. It could the subsequent processing, and hence good agreement with the experimental
be seen both in the simulated and ex- negatively impacts the properties of fi- results of HH (detailed comparison re-
perimental results that stray grains ap- nal products.21,22 The multi-ladle pour- fers to Reference 27).
peared at the edge of the platform and ing (MP) process, i.e., sequential pour-
Simulated Results and Discussion
grew up to the top of the blade. That ing of liquid steel with different carbon
explains why a high withdrawal rate contents,23–26 has been widely used to The simulation was carried out for a
would not be acceptable to get a com- suppress macrosegregation in large in- large 300-t steel ingot (Fe-0.36wt.%C)
plete single crystal blade casting. dustrial steel ingots, especially several 4.5 m in height and 3.4 m in mean
The microstructures in different sec- tons in weight. However, it is very dif- width. Teeming was ignored, and no
tions of the whole blade of group B by ficult to completely avoid certain car- superheat was considered for the mol-

Vol. 63 No. 4 • JOM www.tms.org/jom.html 23

ten steel. The schematic of MP process tern in a large steel ingot is predicted, cess should have some positive effect
is shown in the left side of Figure 14, including an intense positive segre- on reducing macrosegregation from
showing that the first ladle contains gation in the hot top and a conically industrial experience and the simula-
150 tons molten steel, the second 90 shaped negative segregation zone at tion demonstrates the capability of the
tons and the third 60 tons, while the the bottom of the ingot. It is indicated model to predict the macrosegregation
final macrosegregation pattern of car- that certain carbon macrosegregation for large steel ingot in industry. None-
bon in the right side of Figure 14. is not completely avoided in the simu- theless, the flow and mixture behavior
The classical macrosegregation pat- lated MP ingot. However, the MP pro- of the molten steel from multiple ladles

Heating
zone

Blade
casting

Baffle
Cooling
zone

Chill
Withdrawal
dicretion
Withdrawal
equipment
t = 15 min. t = 20 min. t = 25 min. t = 30 min.

Figure 7. Schematic of Bridgman directional Figure 8. Simulated microstructure evolution in group A at different
solidification process. time.

a b
0.1 mm
t = 15 min. t = 20 min. t = 25 min. t = 30 min.

Figure 9. Simulated microstructure evolution in group B at dif- Figure 10. Experimental microstructures of blade casting in group
ferent time. A with stray grains.

S7
S6
S5
S4

S3

S2
S1
a b t = 4.5 min. t = 14.5 min. t = 22.0 min.
0.1 mm
Figure 11. Experimental microstructures of blade casting in Figure 12. Simulated grain evolution in the grain selector.
group B without stray grains.

24 www.tms.org/jom.html JOM • April 2011

 a b
Figure 13. Macrosegregation map in a solidifying cavity at t = 400 s. (a) Existing simulation;
(b) present simulation.

during the teeming process needs to be references

taken into account for further study. 1. L. Gránásy, T. Pusztai, T. Börzsönyi, J.A. Warren, and
J.F. Douglas, Nature Mater., 3 (2004), pp. 645–650.
ConClusions 2. S.-L. Wang, R.F. Sekerka, A.A. Wheeler, B.T. Murray,
Initial Final
S.R. Coriell, R.J. Braun, and G.B. McFadden, Physica
Phase-field and cellular automaton D, 69 (1993), pp. 189–200.
3. G. Caginalp and W. Xie, Phys. Rev. E, 48 (1993), Figure 14. Schematic of MP (left) and
models for simulating the dendrite pp. 1897–1909. simulated results of final macrosegrega-
morphology of cast Mg alloys have 4. S.A. David and T. DebRoy, Science, 257 (1992), pp. tion pattern (right) of a 300-t steel ingot.
been developed, by incorporating the 497–500.
5. A. Karma, “Prediction of Dendrite Growth Directions”
anisotropic formulae, which provide (Presented at the TMS Annual Meeting, San Francis- 19. J. Yu et al., “Experimental Study and Numerical
the insight into both the solid-liquid co, CA, 2005). Simulation of Directionally Solidified Turbine Blade
interfaces patterns and solidification 6. K. Pattersen, O. Lohne, and N. Ryum, Metall. Trans. Casting” (Presented at the Asian Foundry Congress,
A, 21 (1990), pp. 221–230. Seoul, Korea, 2007).
morphologies evolution in hcp based 7. F. Czerwinski, Acta Mater., 50 (2002), pp. 3265– 20. D. Pan et al., JOM, 62 (5) (2010), pp. 30–34.
metals or alloys. An MCA model was 3281. 21. G. Lesoult, Mater. Sci. and Eng. A, 413-414 (2005),
used to study the microstructure evolu- 8. J. Eiken, B. Böttger, and I. Steinbach, Modeling of pp. 19–29.
Casting, Welding and Advanced Solidification Pro- 22. C. Beckermann, Int. Mater. Rev., 47 (5) (2002), pp.
tion of aluminum alloy casting for en- cesses-XI (Warrendale, PA: TMS, 2006), pp. 489–496. 243–261.
gineering application. A mathematical 9. Liang Huo et al., Materials Science Forum, 561-565 23. M. Tateno, Transactions ISIJ, 25 (1985), pp. 97–108.
model for macrosegregation pattern of (2007), pp. 1797–1800. 24. H. Ushiyama, G. Yuasa, and T. Yajima, Ironmaking
10. H.B. Yin and S.D. Felicelli, Modeling and Simulation and Steelmaking, 5 (1978), pp. 121–134.
large steel ingot was also studied. Fi- in Materials Science and Engineering, 17 (7) (2009), 25. V.I. Bogdanov, V.A. Durynin, and V.V. Tsukanov,
nally, a mathematical model based on 075011. Russian Metallurgy (Metally) (7) (2007), pp. 626–633.
CA-FD method was developed for the 11. M.F. Zhu and D.M. Stefanescu, Acta Materialia, 55 26. D.R. Liu et al., Int. J. Cast Metals Research, 23 (6)
(2007), pp. 1741–1755. (2010), pp. 354–363.
three dimensional simulation of grain 12. Liang Huo, Zhiqiang Han, and Baicheng Liu, “Two- 27. W.S. Li, H.F. Shen, and B.C. Liu, Steel Research
evolution in unidirectional solidifica- and Three-dimensional Cellular Automaton Models International, 81 (11) (2010), pp. 994–1000.
tion process of superalloy turbine blade for Simulating Dendrite Morphology Evolution of Cast 28. H. Combeau et al., Metall. Mater. Trans. B, 40
Magnesium Alloys” (Presented at the 139th TMS An- (2009), pp. 289–304.
casting and the experimental results nual Meeting and Exhibition, Washington State Con- 29. M. Wu and A. Ludwig, Acta Materialia, 57 (2009),
were studied and compared with the vention Center, Seattle, 14–18 February 2010). pp. 5621–5631.
simulated results. 13. Liang Huo, Zhiqiang Han, and Baicheng Liu, Mate- 30. D.J. Hebditch and J.D. Hunt, Metall. Trans., 5
rials Science Forum, 638-642 (2010), pp. 1562–1568. (1974), pp. 1557–1564.
aCknowledgeMents 14. B. Li, Q.Y. Xu, and B.C. Liu, Materials Science Fo- 31. N. Ahmad et al., Metall. Mater. Trans. A, 29 (1998),
rum, 561-565 (2007), pp. 1787–1792. pp. 617–630.
This work is financially supported 15. J. Yu et al., J. Materials Science and Technology,
23 (2007), p. 47. Baicheng Liu, professor, Tao Jing, professor,
by National Basic Research Program 16. M. Rappaz and C.A. Gandin, Acta Metallurgica et Qingyan Xu, associate professor, Houfa
of China (Nos. 2005CB724105 and Materialia, 41 (1993), p. 345. Shen, professor, and Zhiqiang Han, associate
2006CB605208), and National Impor- 17. P. Thevoz, J.L. Desbiolles, and M. Rappaz, Metall. professor, are with the Department of Mechanical
Trans. A, 20A (1989), p. 311. Engineering, Tsinghua University, Beijing 100084,
tant S & T Program No. 4 (2009 ZX04 18. W. Kurz, B. Giovanola, and R. Trivedi, Acta Metal- China. Prof. Liu can be reached at liubc@tsinghua.
014). lurgica, 34 (1986), p. 823. edu.cn.

Vol. 63 No. 4 • JOM www.tms.org/jom.html 25