ISSN (1897-3310)

ARCHIVES Volume 9
Issue 4/2009
of
13 – 16
FOUNDRY ENGINEERING
3/4
Published quarterly as the organ of the Foundry Commission of the Polish Academy of Sciences

Modeling of the dendrite arms behavior

during solidification
A.A. Burbelko*, W. Kapturkiewicz, D. Gurgul, P. Stręk
Faculty of Foundry Engineering, AGH University of Science and Technology,
Reymonta 23, 30-059 Krakow, Poland
* Corresponding author. E-mail address: abur@agh.edu.pl

Received 25.06.2009; accepted in revised form 03.07.2009

Abstract
Modelling was carried out to investigate the internal dendrite grains structure formation from a liquid two-component solution. For the
simulation, our own model and computer program based on CAFD (Cellular Automata – Finite Differences) were used. In modelling, the
effect of process conditions and material-related parameters on the nature of the dendritic grain growth was examined. It was demonstrated
that increase of the secondary dendrite arm space may by a result of interruption of the arms growth as well as of overgrowing of concave
regions. A local melting down of the grains of a solid phase due to the segregation of admixtures reducing the alloy point of liquidus is also
possible.

Keywords: Cellular Automaton, Secondary Dendrite Arm Space, Remelting

modelling to be extended even more. The models developed

1. Introduction recently are capable of also including the effects of the diffusion
mass transport and liquid flow [3]. The effect of alloying
The structure of alloys formed during solidification affects the
constituents segregation is investigated using the potentials
utilisation properties of materials. This is the reason why in
offered by computers in the calculation of phase equilibrium
modern metallurgy the possibility of modelling the structure
diagrams [4].
formation process is so important, especially if different aspects
This study uses a mathematical model presented earlier in
of this complex phenomenon can be taken into consideration.
[5, 6, 7]to describe an interaction between the dendritic growth
The first mathematical model based on the method of CAFD
and its consequences, e.g. local remelting.
(Cellular Automata – Finite Differences) was presented by
Umantsev at al. [1, 2]. The development of computer modelling
of the solidification process using CA witnessed in the 90’s of the
past century has been discussed in [3]. 2. Model Description and Solution
A model described in [1] enabled computation of the
temperature field, allowing for the effect of the heat of phase
Method
transformations, interphase energy and surface curvature on the The transient temperature field in the presented simulation is
thermodynamic equilibrium temperature and non-equilibrium calculated from a numerical solution of the Fourier Equation:
character of the growth process (a velocity/undercooling
relationship). Progress in computer technique and numerical
methods enabled the range of the physical phenomena included in

ARCHIVES of FOUNDRY ENGINEERING Volume 9, Issue 4/2009, 13-16 13

 ∂T ∂f
c = ∇(λ∇T ) + L S (1)
∂τ ∂τ
where: T is the temperature, τ is the time, λ is the thermal conductivity,
fS is the solid fraction, c is the volumetric specific heat, and L is the
volumetric latent heat of solidification.
The transient field of solute concentration has been computed from
a numerical solution of the diffusion equation:
∂C ∂f
= ∇ ⋅ (D∇C ) + C (1 − k ) S (2)
∂τ ∂τ
where: D is the solute diffusion coefficient, and k is the solute partition
coefficient. For the liquid and interface cells, C is the concentration of
solute in liquid (CL), and for solid cells C is equal to CS/k (CS is the solid
composition).
The anisotropy of the growth of dendritic crystals is caused by the
anisotropy of their physical properties, such as the surface tension and
kinetics of growth. In numerous models it has been described by the
anisotropy of the computational grid. The directions of the dendrite arm
growth obtained by modelling were parallel to the lines of the cellular
automaton grid, or to a bisector of the angle between them [2]. Various
techniques are known for dealing with this problem. Wang at. al. [8]
uses a modified decentred square growth technique. Beltran-Sanchez
and Stefanescu [9,10] propose a virtual interface tracking scheme.
In [6, 7, 11]the reasons for the artificial anisotropy in solidification
modelling by the cellular automata method were examined and our own
solution was proposed. Determination of the interface normal direction
is based on the vector connecting the centre of mass of the
transformation product contained in a circle, the centre of which lies in
the examined cell, with the centre of this cell [12]. As proved in [6] by
the growth rate circle test, the minimum level of the artificial anisotropy
is ensured by using the environment of a radius √5a, where a is the cell
size.
The solidification rate is calculated from a ratio between the
amplitude of the growth velocity vector and the longest projection of the
cell diagonal on this vector, according to equation [5]:
Δf S u wall is constant and equals the initial temperature of the liquid. In
= calculations, the value of the kinetic coefficient equal to μ=10-3
Δτ a( cos θ + sin θ )
, (3)
where: u is the growth velocity, and θ is the angle between the interface
normal direction and the grid axes.
The model allows for a non-equilibrium character of the
solidification process. The modulus of a vector of the interface
migration rate depends on the kinetic undercooling:
u = μΔTk , (4)
where: μ is the kinetic parameter, and ΔTk is the kinetic undercooling.
The kinetic undercooling is calculated as the difference between the
temperature of the thermodynamic equilibrium, determined for a local
chemical composition of liquid on the solidification front (Teq(C)), and
the local temperature (T) with capillary undercooling correction (ΔTc)
related to interface curvature:
ΔTk = Teq (C ) − T − ΔTc , (5)
14 ARCHIVES of FOUNDRY ENGINEERING Volume 9, Issue 4/2009, 13-16
where: capillary undercooling ΔTc = ΓσΚ , Γ is the Gibbs-Thomson
coefficient, σ is the surface tension, and Κ is the interface curvature.
The anisotropy of the surface tension and interface mobility were
taken into account. Other details pertinent to the applied diffusion-
kinetic model have been related in [5-7].
A set of the equations (1), (3)-(5) were solved using an implicit
scheme. Next, having substituted the solidification rate in interface cells
determined from equation (3) to equation (2), the field of concentration
was determined using an explicit scheme.
The point that describes the liquid alloy state in the interface during
solidification on the thermodynamic equilibrium diagram (the liquid
figurative point) is shifted in respect of the liquidus line towards a two-
phase area. Its distance from the liquidus line is a measure of the
thermodynamic driving force of the phase transformations, due to
which, from the liquid, the crystals of a solid phase are precipitating,
while the figurative point is moving towards a state of equilibrium (the
liquidus line) (Fig. 2).
Using the above mentioned scheme in a solution of a set of model
equations enables a very accurate approach to the non-equilibrium
character of the process of phase transformations with precise
determination of the coordinates of the figurative points with respect to
the line of an equilibrium diagram.
3. The conditions and results of
modelling
The results presented here have been obtained on a grid of
501x801 cells for a material with the properties corresponding to
the properties of a Fe-0.6 wt. % C alloy. On the right and left
boundaries of the modelling area (Fig. 2) a periodic boundary
condition has been applied; for the lower boundary, an adiabatic
condition holds good. The upper boundary is a mould wall of
constant temperature. On this boundary, at a half width distance
of its area, two nuclei were deposited. The initial undercooling of
the liquid phase below the alloys’ equilibrium liquidus
temperature was 0.04·L/c = 13.5 K. The temperature of the mould
m/(s·K) was adopted. The contact angle of mould wetting by the
solid phase was assumed to be equal to 90º.
The preset angles between the axis of the coordinates and the
directions of the highest growth velocity of the main arms of these
two grains are 22.5º and 2º.
Diagram in Fig. 1 indicates that symmetry of dendrite growth
in the presented model doesn’t depend on the anisotropy of
calculation mesh. Dots on this diagram show divergence of
dendrite tip position vector from the mesh axis direction. Given
direction 2º is shown by dotted line. Limits of deviation
(calculated for tip distance from desired direction equal to one CA
cell size) are shown in this diagram as solid lines. Order of
divergence magnitude obtained in the simulation is equal to one
cell size.
The evolution of the grains obtained by simulation is shown
in Fig. 2. From Fig. 3 it follows that the main arms growing at an
angle of 2º (arm no. 1 in Fig. 2), 22.5º (arm no. 2) and 67.5º (arm
no. 3) to the mould wall normal direction are, at the initial stage
of solidification, growing with the same velocity equal to
1.23·10-3 m/s. As a result of the, so called, “soft collision,” the follows that the decrease in spacing between the secondary arms
rate of the dendrite arm growth starts decreasing sooner, the more may be caused by both the decay of the secondary arms as well as
their growth direction approaches the direction parallel to the by their fusion (the disappearance of a concave space between
mould wall. For arms 4 and 5 – this happens practically at the them).
very beginning of the solidification process; for arm 3 after a 40
lapse of about 5 ms, and for arm 2 – after 10 ms.
1
Line 1 (Fig. 3) is slightly convex on its whole length with the

Lenght of arm [mm]

convexity pointing upwards. This effect is due to the conditions of 30
heat transfer from the solidification front changing with the
growing distance from the mould wall.
5 20 2
Calculated from tip position
3
Desired value 10
4 4
Angle [deg]

Limits, calculated for tip divergency

+/- one cell from desired direction 5
3 0
0 10 20 30 40
Time [ms]
2
Fig. 3. History of primary dendrite arm length in time
for the dendrite grains shown in Fig. 2.
1
Fig. 4 shows the distribution of kinetic undercooling on the
0 10 20 30
solidification front within the entire space of a 2D dendrite during
Time [ms]
its growth. The distribution of capillary undercooling is shown in
Fig. 1. Precision of generation of a dendrite primary arm direction Fig. 5. As follows from these drawings, for the assumed value of
in the simulation. the kinetic coefficient, both these values are similar. This fact
prompts the need for analysing once again the fact of disregarding
the value of kinetic undercooling in some CAFD models.

5 ms 10 ms 15 ms

20 ms 25 ms 30 ms
Fig. 2. Simulated evolution of the two grains. Fig. 4. Distribution of the kinetic undercooling [K]
in the space of the 2D dendrites.
The black lines in Fig. 5 indicate the position of zero capillary
undercooling, and hence of zero curvature. From this drawing it

ARCHIVES of FOUNDRY ENGINEERING Volume 9, Issue 4/2009, 13-16 15

Fig. 5. Distribution of the capillary undercooling [K] in the space of
the 2D dendrites (lines – positions of 0 capillary undercooling).
The model presented here enables a forecasting of the process
of partial remelting of the grains during solidification. The areas
of remelting marked in white are shown in Fig. 6.
The results presented here indicate that the effect of artificial
anisotropy on the results of modelling has been reduced to an
acceptable level.
4. Conclusions
In simulation based on the model developed previously, a
distribution of the values of the kinetic and capillary undercooling
within the whole space of a 2D dendrite during its growth was
obtained.
It has been stated that during the dendritic solidification the
values of the kinetic and capillary undercooling are of the same
order of magnitude which prompts the need for analysing once
again the fact that the kinetic undercooling values have been
disregarded in some CAFD models.
It has been proved that the change in the secondary dendrite
arm spacing may be caused by either a decay of their growth or
by their fusion (the concave spaces disappear).
In modelling, the effect of the dendrite remelting was
obtained by both an interaction of the secondary arms and a soft
collision of the neighbouring grains.
Acknowledgements
This work was supported by the Polish Ministry of Science and
Education as research project N507 023 32/0839.
16 ARCHIVES of FOUNDRY ENGINEERING Volume 9, Issue 4/2009, 13-16
Fig. 6. Places of the remelting of dendrite grains (white).
References
[1] A.R. Umantsev, V.V. Vinogradov, and V.T. Borisov:
Kristallografia Vol. 30 (3) (1985), p. 455 (In Russian)
[2] A.R. Umantsev, V.V. Vinogradov, and V.T. Borisov:
Kristallografia Vol. 31 (5) (1986), p. 1002 (In Russian)
[3] M.F. Zhu, C.P. Hong, D.M. Stefanescu, Y.A. Chang: Metall.
Mater. Trans. B Vol. 38B (2007), p. 517
[4] M.-F. Zhu, W. Cao, S.-L. Chen, C.-P. Hong, Y. A. Chang:
Journal of Phase Equilibria and Diffusion Vol. 28 (1) (2007),
p. 130
[5] A.Burbelko: Mezomodeling of Solidification Using a
Cellular Automaton (UWND, Poland, Krakow 2004) (In
Polish)
[6] A.Burbelko, W. Kapturkiewicz, D. Gurgul, in: Solidification
Processing 2007: Proceedings of the 5th Decennial
International Conference on Solidification Processing. The
University of Sheffield, UK, (2007) p. 31
[7] A.Burbelko, E. Fraś, W. Kapturkiewicz and E. Olejnik: Mater.
Sci. Forum, Vol. 508 (2006), p. 405
[8] W.Wang, A. Kermanpur, P.D. Lee and McLean: Int. J. Cast
Metals Res. Vol. 15 (2002) p. 269
[9] L.Beltran-Sanchez and D.M. Stefanescu: Metall. Mater.
Trans. A Vol. 34A (2003) p. 367
[10] L.Beltran-Sanchez and D.M. Stefanescu: Metall. Mater.
Trans. A Vol. 35A (2004) p. 2471
[11] A.Burbelko, W. Kapturkiewicz, D. Gurgul: Computer
Methods in Materials Science Vol. 7 (1) (2007) p. 182
[12] U.Dithley, V. Pavlik, in: Modeling of Casting, Welding and
Advanced Solidification Processes VIII, edited by B.G.
Thomas and C. Beckermann, TMS, Warrendale (1998) p.
589