Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S.

Qin)

25. Phase field models

25.1 Introduction In classical free boundary solidification model (known as Stefan problem), interface is a zero thickness mathematic discontinuity location. A grain is a domain isolated by interface from surrounding. Heat and mass equations are solved inside each domain. Interface migration is determined by gradients of temperature and solute around the interface. For a solidifying system that contains 5 grains, six sets of mass and heat equations are required to solve. Therefore, interface is requested to chase explicitly. The accurately chasing of interface location makes the classical method difficult in computing solidification with large number of grains or grains with complicated geometries. The classical free boundary solidification is illustrated schematically in figure 1(a).

(a)

(b) Figure 1. Solidification models (a) classical free boundary model (b) phase field model.

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

To avoid the chasing of the solid-liquid phase boundary, an order parameter called phase-field order parameter is introduced to represent the state of phase in space. The order parameter varies smoothly from one value in liquid to another value in the solid across a spatially diffuse interface region of thickness O. The whole system, no matter how many grains are contained and how many phases are involved, is described by only one set of governing equation. There is no need to chase interface location. The method is called phase-field model and is schematically illustrated in figure 1(b).

25.2 Free energy The free energy of system is given by the integration of free energy density across the whole volume

I , c, T dv
V

(25.1)

where is the free energy density which depends on the local values of phasefield I, temperature T and composition c. V is the volume of system. According to Ginzburg-Landau theory, free energy contains gradient and bulk contributions and can be expressed as following for the simplest case

I , c,T 

1 2 H I 2

 f (I, c,T )

(25.2)

where the first term in the right hand side of equation (25.2) is the contribution from the gradient of phase and is called Landau-Ginzberg gradient energy. H is the gradient coefficient and is related to interfacial property. f(I,c,T) is the bulk free energy which in phase transition is a double-well potential. In many case f(I,c,T) is a phenomenological free energy of biased double-well shape. For example people utilizes following equation to represent bulk free energy [1]

f I , T 

2I 3 I 5 I4 I2 1     T I 4 2 Z 3 5

(25.3)

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

Equation (25.3) is plotted in figure 2 with parameter Z=1 and T=1. The phases are represented as I=1 for liquid, I=-1 for solid and 1<I<1 for soid-liquid interface.

Figure 2. Phenomenological double well shaped free energy. Generally liquid metal solidifies into several solid phases. The generic phase-field model is to introduce a set of phase-field order parameters {I , I , I , } to represent those phases involved in solidifying system. Those phase-field order parameters obey the constraint of

N 1

I  x, y , z , t  1

(25.4)

The free energy density is represented as

I , c, T 

hI  1 g c , T    4Z I I  hI 
1 2 H I I  I I 2 2
N 2 N 2 2 N 2 1 N 1

(25.5)

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

where the three terms on the right hand side of equation (25.5) come from Landau-Ginzberg gradient energy, interfacial excess free energy, and bulk free energy, respectively. H can be expressed as following to include the effect of anisotropy [2].

H
H

section. The multiplier factor of g in third terms has a meaning of volume percent of E-phase in the point and is convenient to chose h(I ) as
hI



and Z

 H   KT    

(25.6)

are interface related parameters and will be determined in next

I 3 6I 2  15I  10

(25.7)

By noticed the meaning of the multiplier, the solute concentration at any point can be expressed into

g is the Gibbs free energy density of E-phase and can be obtained by alloy thermodynamics and Redlich-Kister equation. For binary alloys, it takes the expression of

  c    hI 
1 N 1

hI

(25.8)

g c , T 

1  c g T   c
A 0

where g A and g B are the Gibbs free energy density of phase E with pure element A and B, respectively. R is the gas constant and L is the Q-th model parameter in the Redlich-Kister equation and is available in handbook.

"

 c 1  c  ! L 1  2c  "

g B T  

RT c ln c  1  c ln 1  c vm

>

(25.9)

25.3 Governing equation

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

The evolution of system is always following a routine of reducing system free energy, i. e.

wG w t d 0
which takes the expansion form of

(25.10)

GG GI c ,T

wI GG wt  c ,T G c

,T

wc wt

,T

GG  GT

wT wt ,c

d0
,c

(25.11)

The subscripts outside the right bracket mean that the parameters are treated as constant. Due to the independence of I, c and T, equation (25.11) can be ensured when and only when
wI wt c ,T G G M GI c ,T

(25.12)

wc wt

&

,T

GG M c Gc

&

(25.13)
,T

wT wt

where M , Mc and MT are positive coefficients. Equation (25.11) can be further expressed as

'

,c

GG M T GT

'

(25.14)
,c

wI wt

c ,T

N  M

GG wI 1 2 1 ) 0 4 G I1 2 5 ) 2 4 2 3 wI0 2 5 ) 2 4 2 3
c ,T , , c ,T , ,

(25.15)

The constraint of equation (25.4) has been used during the derivation of equation (25.15). The equations (25.13)-(25.15) can be further expressed into

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

wI wt

c ,T

7 6 8 9@A 8 7
M
N

wT wI wT wI 2 2  H K c  KK cc w x w y w y wx hccI    hcI ' 60Z c ,T

87

2 2 c  2 K K T I  K I

(25.16)

wc wt

B C

,T

wg M c wc

,T

M c c g cc

(25.17)
,T

wT wt

,c

w g M T wT

,c

M T T g TT

(25.18)
,c

where the prime means the differentiation respect to its parameter. g and ' are defined as

'

DF E E  hI F g E  g F @ G G G G > 1  hI @ w g H w g G w cH G cH  c   g H  g   cI  c h w
g
N

>g

(25.19)

wg wc  wc wI

HG

c ,T

(25.20)

The subscripts of g in equations (25.17) and (25.18) refer to the deferential of g respect to the subscript parameters. ' plays a role of driving force, and which sometimes be replaced with thermodynamic driving force in literature. To illustrate the meaning of ' and compare with the thermodynamic driving force, one specifies D phase as solid and E as liquid. c only depends on temperature according to solidification theory, and therefore, one has, F /I =0. |'| and the thermodynamic driving force are illustrated in figure 3 as |CcDc| and |CD|, respectively. It can be found from equation (25.20) or figure 3 that |'|=0 only when

I P

>w g R cR , T  w cR

 w g c , T  w c

Q Q

Q@

c ,T

(25.21)

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

Equation (25.21) is exactly the equilibrium common tangent law. The replacement of |CcDc| with |CD| is not an appropriate approximation due to the great difference between their values.

Figure 3. Kinetic and thermodynamic driving force for phase transition

25.4 Parameters determination The parameters H , Z and M in the phase-field equation are to be matched with the interface energy V , interface thickness 2O and interface propagation rate v . H and Z are determined at equilibrium condition and M is determined at dynamic steady condition.

ISP SI P TUISP ISP UT ISP ISP IWP I

IVP

TUISP

From the knowledge of alloy thermodynamics, an equilibrium state implies that the amount of each individual phase is not changed and the common tangent law is satisfied. For a planar interface at equilibrium condition, the equilibrium phasefield parameter I e obeys the following equation according to equation (25.16) after taking account the relationship of I =1-I

P I

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

X`Y X
2

d 2I e 1  I 1  I 2 dx 2Z

X`Y X X  X 
1  2I

(25.22)

where x is along the normal direction of the interface. The solution of equation (20) with the boundary conditions of I eo1 as xo-fand I eo0 as xo+f is

I e x 

1 x 1  tanh 2 2Z H 2

acb acb

(25.23)

Defined the interface area as the region between I e=0.9 to I e=0.1, one get the relationship of

dUe


2.2 2Z H

dUe dUe fcg f

dI e dx
2

(25.24)

Multiply equation (25.22) with dI e/dx and perform the integration, one get
2 1 I e 1I e 4Z

fcg f

1 2 H 2

(25.25)

From the definition of interfacial energy and Gibbs free energy, one has

hci V

1 2 H V 2

hci dIh  1ch i Ih dx 4Z

e 2

e2

1  Ih  dv
e 2

(25.26)

Bring equations (25.23) and (25.25) into (25.26) arrives

pcq

pcq 12 Z pcq
2H

(25.27)

When a planar interface is moving with a constant velocity v , equation (25.16) changes into an expression of following after replace I with 1-I

rSs

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

tcu t ct v u ct u t
2

v d 2I  2 dx M

1 dI  I 1  I dx 2 Z

tcu t

t  t   t 
1  2I

 hc I '

(25.28)

For any of the solid-liquid interface, the concentration of solid phase under diffusion controlled growth depends on the local temperature only (F /I =0). ' is, therefore, a function of liquid concentration and temperature. In order to maintain the thickness of interface during propagation, M must be a function liquid concentration. This assumption has not any contradiction with our earlier definition because the earlier constraint for M is only for M t0.

wUrxs

wUrSs

r s

wUrxs

For a certain value of ', the third term in equation (25.28) can be integrated with respecting to I . The result is a double well potential with the asymmetry ' and maximum value ]

1 I 1  I 2 Z

yc y

y  y   y  y
1  2I
v d 2I  2 dx M

 hc I ' dI

1 I 4Z

yc y

1  I 2  hI '

(25.29)

Such a double well potential can be well described by a quadratic polynomial in I . Equation (25.29) can eventually be expressed in a formation of [3]

c c c 
]
'

dI  9I I  W I  1 0 dx

(25.30)

9W 3 1 1  W 6 2
9 1 W  6 2

(25.31)

(25.32)

The only solution of equation (25.30) with the boundary conditions of I o1 as xo-fand I o0 as xo+f is

I x 

9 1  exp x 2 2H

(25.33)

Graduate Institute of Ferrous Technology, POSTECH Rongshan Qin (R. S. Qin)

and this solution only exists when M

6H '

c c

(25.34)

25.5 Numerical calculation Finite difference or finite element methods can be used to solve the three governing equations. Figure 4 is one of the simplest examples.

Figure 4. Phase-field simulation of interface migration References 1. A. Karma and W-J Rappel, Phys. Rev. E 57 (1998) 4323. 2. A.A. Wheeler, B.T. Murray, and R.J. Schaefer, Physica, 66(1993)243. 3. R.S.Qin and E. R. Wallach, Journal of Crystal Growth, Vol. 253, pp549556, (2003).