Materials Transactions, Vol. 49, No. 11 (2008) pp.

2625 to 2631
#2008 The Japan Institute of Metals

Free Energy Problem for the Simulations of the Growth

of Fe2 B Phase Using Phase-Field Method
Raden Dadan Ramdan1; * , Tomohiro Takaki2 and Yoshihiro Tomita1
1
Graduate School of Engineering, Kobe University, Kobe 657-8501, Japan
2
Graduate School of Science and Technology, Kyoto Institute of Technology, Kyoto 606-8585, Japan

The present simulation works develop phase-field method to replicate the growth of Fe2 B boride from austenite phase. The present works
also give a major concern on the free energy driving force problem for this transformation. In order to enhance the reliability of this study, free
energy driving force is left as the only variable parameter to be optimized while other parameters were taken from experimental data. Several
sets of free energy have been established based on the thermo-chemical database and the assumption that there is a parabolic relationship
between free energy of boride/austenite and boron concentration. On the other hand, in order to allow the growth of single Fe2 B phase,
temperature of 1223 K has been chosen as the evaluated temperature. Based on the comparison of the present numerical results and available
experimental data, one set of free energy has been considered to give the appropriate condition that approaching the experimental condition for
the growth of Fe2 B from austenite phase. [doi:10.2320/matertrans.MRA2008158]

(Received May 9, 2008; Accepted August 18, 2008; Published October 25, 2008)
Keywords: phase-field simulation, boronizing, free energy driving force, line compound

1. Introduction input data for the simulation of boronizing process by

Keddam.3) Different with the previous simulations work on
Boronizing is a kind of thermo-chemical treatment, where boronizing process3,8,9) that use the mass balance equation as
boron atom is diffused into the metal substrate to form a hard the basic model, phase-field model for the growth of boride
boride structure on the surface of metal. There are two major phase is derived from Ginzburg-Landau free energy func-
types of boride structure, FeB and Fe2 B, where the existences tional that consider the free energy value from thermo-
of both phases are depended on the condition of the process chemical database and related physical parameters of
such as temperature. The former phase exist when temper- materials such as interface energy and interface thickness
ature of the process is relatively low (below 1223 K), whereas of the corresponding material. This method has been well
the later phase might exist at higher temperature condition proved to simulate many kinds of phase transformations,
(1223–1373 K) and longer boronizing time.1,2) In addition, such as austenite-ferrite transformation in steel.10) One
FeB phase is also considered to have a higher hardness com- important point from this method is the unnecessary of
pared to Fe2 B phase, however this structure is more brittle calculating the complex condition of the interface, which
than Fe2 B phase.1,3) Therefore, in many cases, it is preferable might bring difficult problem for the simulation works.
to have a single phase of Fe2 B rather than having FeB phase However there is one remaining problem in developing the
in the boride structure, since the single phase of Fe2 B is phase-field simulations for the growth of Fe2 B phase, which
considered to offer a higher toughness compared to the is defining its free energy driving force. Since Fe2 B is a line
structure with the presence of FeB. Overall, boride structure compound that its free energy is defined to be independent
offers a higher hardness compared to other thermo-chemical of concentration (in the thermodynamic database), therefore
products, such as boride structure on AISI 5115 steel may there will be no derivative of its free energy on the
produce a hardness of around 1900 HV compared to around concentration which is required in constructing the diffusion
800 HV for the case of carburised layer on the same steel.4) profile model that couple the phase-field evolution model.
Even-though boronizing process offers a significant im- One of the common approach to solve this problem is by
provement on the quality of material, very limited references making an assumption that the free energy of this compound
can be found on this process. Among them, study on the pack has a parabolic relationship with concentrations.11–13) The
boronizing process1,3,5–7) took the most interest since this present study also adopted the similar approach, and in order
process offers a simpler way to produce boride layer. One of to evaluate the reliability of this approach, free energy
the progressive researches on the pack boronizing process variable was left as the only variable to be optimized,
have been done by Keddam3,8) who works on the simulation whereas other important parameters such as interface
of the growth kinetic of Fe2 B, taking into account the mobility, diffusion flux and diffusivity coefficient were taken
thermodynamic properties of the Fe-B phase diagram and from experimental data and the simulations results was also
adopting diffusivity of boron by Campos et al.9) In the present compared with the related experimental data. The derivation
study, phase-field method was applied to simulate the growth of phase-field model as well as its free energy driving force
of single Fe2 B boride from the austenite phase and taking will be explained in the following subsection.
several important experimental data for the pack boronizing
of ARMCO iron by Campos et al.,9) which is also used as the 2. Phase-Field Model

*Graduate Student, Kobe University As previously mentioned, phase-field model for the pres-
Corresponding author, E-mail: [email protected] ent work is derived from Ginzburg-Landau free energy
2626 R. D. Ramdan, T. Takaki and Y. Tomita

functional (eq. (1)),10) where this functional is defined as the Table 1 Kinetic constant from experimental data by Keddam.3Þ
functional of free energy density (gð
; XB ; TÞ) and the Temperature (K) k (mm s
1=2 )
gradient energy density (").
is the field variable which is 1223 0.4584
set as 1 for boride and 0 for austenite phase, whereas XB and
1253 0.5943
T is the molar fraction of boron and absolute temperature
respectively. On the other hand the constant gradient 1273 0.6720
Z
 1323 0.8160
"2
G¼ gð
; XB ; TÞ þ jr
j2 dV ð1Þ
V 2
energy density, ", is related with the interface energy and
pffiffiffiffiffiffiffiffiffiffiffiffi value of normal velocity of interface is equal with the value
interface thickness as " ¼ 3=b.14) Assuming the surface of interface mobility for the present study.
region is  <
< ð1
Þ, we obtain b ¼ 2 tanh
1 ð1
2Þ, The normal velocity of interface is derived from
where  is the value of
at the onset of surface region definition of interface position,3) where it is defined that
( ¼ 0:1). the interface position is a function of kinetic constant, k,
Free energy density in the eq. (1) is constructed by the free and time, resulting an expression of normal velocitypof ffi
energy of boride (GFe 2B 
m ) and austenite (Gm ) as described in interface as well as interface mobility equal to M ¼ k=2 t,
the eq. (2). From this equation, pð
Þ is an interpolating where k is constant which is taken from experimental data
function which is expressed as, pð
Þ ¼
3 ð10{15
þ 6
2 Þ, by Keddam3) (Table 1) and t is time. For the present work,
whereas qð
Þ and W are double well potential and the the interface mobility is kept constant from the initial
potential height respectively. Potential height is defined as, growth of boride phase, since current understanding of
qð
Þ ¼
2 ð1

Þ2 , whereas double well potential is related interface mobility consider the interface mobility to have a
with the interface energy () and interface thickness () as, fixed value and it might increases exponentially with
W ¼ 6b=. temperature.15–17)
As mentioned in the introduction, defining free energy of
gð
; XB ; TÞ ¼ pð
ÞGFe
m
2B 
þ ð1
pð
ÞÞGm þ Wqð
Þ: ð2Þ
boride as a line compound will cause a problem since there
From the above Ginzburg-Landau free energy functional, is no derivative of its free energy on the concentration.
the evolution of phase field is derived following the Allen- Therefore in order to solve this problem and to find the
Cahn equation with the assumption that the total free energy optimized value of these free energies, both free energy of
decreases monotonically with time. For the case of one austenite and boride is approached with the assumption that
dimensional problem, this derivations produces the following free energy has a parabolic relationship with concentration
model of phase field evolutions, and their minimum free energy is similar with the free
@
G energy provided by the thermo-chemical database (SGTE
¼
M
database).18) The free energy of austenite is defined as
@t 
follow

¼ M
"2 r2
þ 4W
ð1

Þ 
Gm ¼ GAV þ 3:2  106 ðb
XB Þ2 ; ð5Þ
  where b is the concentration of boron for the minimum free
15 1


ð1

Þ
g þ

: ð3Þ energy of austenite and GAV is the free energy of austenite
2W 2
which is taken from the CALPHAD method for two sub-
where
g is the chemical driving force which is defined lattice model as follows
as
gðXB ; TÞ ¼ GFe 2B 
m ðXB ; TÞ
Gm ðXB ; TÞ. This equation

GAV ¼ yFe yB o GFe:B 
þ yFe y o GFe:
allows the evolution of field variable from
¼ 0 (austenite)
to
¼ 1 (Fe2 B), where M
from this equation is the kinetic þ RTyB ln yB þ RTy ln y : ð6Þ
parameter related to the mobility of austenite-boride inter- where yFe and yB is the site fraction for iron and boron,
face M as respectively. For the present works, in stead of using site
pffiffiffiffiffiffiffi fraction, molar fraction values (XFe and XB , molar fraction
2W
M
¼ M: ð4Þ for iron and boron respectively) are used for the calculation.
6" The site fraction is defined by molar fraction as, yB ¼
Interface mobility is a parameter that relates the interface aXB =cð1
XB Þ, where a is equal to c since we consider the
normal velocity, v, to the driving force. Recent study on the fcc austenite of Fe. On the other hand yFe is set as 1 since

interface mobility by Z.T. Trautt et al.15) consider several there is no other alloying element in the matrix. o GFe:B is
o 
types of driving force that causing the movement of interface the reference energy term for boron austenite and GFe: is

such as deformation type of driving force, curvature and the reference energy term for vacancy in austenite. o GFe:B is
random walk. Considering the present condition of interface obtained from the optimization of SGTE database by Van
is a flat interface with negligible curvature and there is no Rompaey et al.19) as follow
contribution of deformation force during the process, there- o 
GFe:B ¼ HBser þ HFe
ser
þ 45496
77:5T: ð7Þ
fore we consider the movement of interface for the present
ser
studies is in the type of random walk movement. Correspond HFe and HBser is the enthalpy at reference state for iron and
to the Z. T. Trautt et al.,15) the interface mobility of random boron respectively and are obtained from SGTE database
walk type is independent of driving force, therefore the for pure elements.18) On the other hand, the reference
 Free Energy Problem for the Simulations of the Growth of Fe2 B Phase Using Phase-Field Method 2627

Table 2 11 sets of free energy for the austenite- Fe2 B transformation.

y(maximum solubility of
Case b F
boron in austenite)
1 0.1 3:  106 0.01
6
2 0.1 3:2  10 0.012
3 0.1 3:4  106 0.0125
4 0.1 3:6  106 0.0125
6
5 0.1 3:8  10 0.0125

Fe2B 6 0.125 3:8  106 0.025

6
7 0.15 3:8  10 0.04
8 0.175 3:8  106 0.045
6
9 0.2 3:8  10 0.047
10 0.225 3:8  106 0.045
6
11 0.25 3:8  10 0.03

Fig. 1 Common tangent curve for the first 5 sets of free energies with free
energy of austenite is kept constant for the b value of 0.1 and the free
energy of Fe2 B is varied with the F value as the variable parameter.


energy state for vacancy in austenite, o GFe: , in eq. (7) is
20)
obtained from Per Gustafson and defined as
o
GFe: ¼ HFe
ser

237:57 þ 132:4T
24:7T ln T

0:0038T 2
5:9E
8 T 3 þ 77358:5=T: ð8Þ
In order to allow a solubility range in the Fe2 B phase, the
chemical free energy of Fe2 B is considered to have a
parabolic function of composition, XB , following the similar
approach by Huh et al.21)
Fe2 B
Gm ¼ FðXB
0:33Þ2 þ o GFe2 B ; ð9Þ
where F is constant and o GFe2 B is the reference energy term
for the stoichimetric composition of Fe2 B and is obtained
from the optimization of SGTE database value by Van Fig. 2 Common tangent curve for the sets of free energies with the free
Rompaey et al.19) as follow energy of Fe2 B is kept constant for the F value at 3:4  106 and the free
energy of austenite is varied with the b value as the variable parameter.
o
GFe2 B ¼ HBser þ 2HFe
ser

93363 þ 481:99T

79:05T ln T
0:007T 2 þ 731991T
1 : ð10Þ
For the present research, 11 sets of free energy will be Regarding the maximum solubility of boron in austenite
evaluated to find the optimum condition that is considered to phase, which is also the results of the construction of
approach the appropriate free energy condition for the common tangent curve of free energy, the present simulations
transformation of austenite to Fe2 B boride phase. The first works produce a higher value of this point than that is
5 sets of free energy use the fixed value of austenite free provided by the available data of this value from the phase
energy with b constant (eq. (7)) at 0.1 and F constant diagram. However since phase-field simulations works on the
(eq. (12)) will be used as the variable parameter to be calculation of the growth of one phase from another phase,
optimized. These variations results in the sharper curve of and for the present works the growth of Fe2 B from austenite
boride free energy from the case 1 up to case 5 as can be phase instead of nucleation of Fe2 B phase, alteration of this
seen from the common tangent figure of these free energies point will not effect the calculation. The precise determi-
(Fig. 1). In addition, these variations of free energy also nation of this value is very crucial for the case of nucleation
change slightly the boron solubility value as is shown by the y process, where for the case of Fe2 B nucleation, this phase
term in the Table 2. On the other hand, the last 7 sets of free will start to nucleate when the solubility of boron in austenite
energy use the optimum value of F constant from the first 5 is exceeded.
sets of free energy as the fix parameter, and the valued of b
constant as the variable parameter to be optimized. These 3. Boron Diffusion Model
variations also change the sharpness of the free energy curve
and the boron solubility as well, as can be seen in the From Ginzburg-Landau free energy functional in the
Fig. 2 and Table 2 respectively. All set of free energies are eq. (1), the nonlinear diffusion equation is derived following
tabulated in the Table 2. the Cahn-Hilliard equation as follow
2628 R. D. Ramdan, T. Takaki and Y. Tomita

(a) (b)

Fig. 3 Initial boron concentration profile condition (a) and initial phase-field condition (b).

  2  the one dimensional case as provided in the experimental

@XB @ g @2 g
¼ r MB ð
; XB ; TÞ rX B þ r
; ð11Þ data at references8) and we obtained a value of 2 mol/m. On
@t @XB2 @XB @

the other hand, zero Neumann boundary condition is applied
where MB is diffusion coefficient of boron, and is defined as a on the both side of phase-field profile and the right side of
function of diffusion coefficient of boron in austenite, MB , concentration profile. The interface thickness, , is set at 8
and diffusion coefficient of boron in boride, MBFe2 B , as in the times of grid size, which is equal to 400 nm, temperature of
following expression the process of 1223 K, whereas the interface energy, , is set
at 1.0 J/m2 .
MB ð
; XB ; TÞ ¼ ðMBFe2 B Þ pð
Þ ðMB Þð1
pð
ÞÞ : ð12Þ
Both MB and MBFe2 B has a temperature dependence function as 5. Simulation Results and Discussions
MBi ¼ Mo expð
E=RTÞ; ð13Þ
Figure 4 illustrates numerical results that replicate the
where MBi is the diffusion coefficient of boron in austenite or 2 h boronizing process of ARMCO iron for case 1–5 of
boride (i ¼  or Fe2 B), and Mo is the pre-exponential different set of Fe2 B free energy. Experimental data for this
parameter of diffusivity. Mo values for boron in austenite and condition3) provide a value of approaching 40 mm as the
boride were set at 4:4  10
8 m2 s
1 and 1:13  10
6 m2 s
1 case depth of Fe2 B layer which is almost similar with the
respectively.3) depth of thickness obtained by the case 4 and 5 of free
energy in the Fig. 4(a). In addition there is a decreasing of
4. Computational Model depth of thickness from the case 1 up to case 5, which
indicates the decreasing of kinetic of the process for these
One dimensional finite difference simulations of eqs. (3) various sets of free energy. This condition is believed due to
and (11) were performed with the solutions of these equations the lower free energy driving force provides by the higher
are approached by explicit time difference method. The index (1–5) case of free energy for the composition of boron
computational domain is divided into 10000 finite difference around 0.3 molar fraction (composition for the case of Fe2 B
grids, where the individual grid size is set at 0.05 mm phase), as also suggested from Fig. 1. It can be seen from
resulting the size of computational domain of 500 mm. Initial this figure that the lower difference of free energy between
boron concentration was set at 0.0035 molar fraction of boron Fe2 B and austenite phase occurs for the composition around
in the austenite phase and 0.33 molar fraction of boron in the 0.3 molar fraction of boron. On the other hand, Fig. 4(b)
Fe2 B phase (Fig. 3(a)). On the other hand, initial phase-field shows the boron concentration profile for the related boride
variable (Fig. 3(b)) was set as one dimensional equilibrium growth in the Fig. 4(a). It can be seen that case 5 approach
profile, where it was considered that initially Fe2 B phase the equilibrium condition of Fe2 B phase, which is consid-
(
¼ 1) exist up to grid size of 10 (the origin is at i ¼ 10). ered to be around 0.33 molar fraction of boron.1,19) One
Neumann boundary condition is applied at the left side of important point to be noted here is, the equilibrium
boron concentration profile, with the gradient value is composition of boron in boride phase as is resulted from
calculated as the flux (J) of boron entering the surface as the numerical simulations is different with the equilibrium
J ¼ D@C =@x, where D is diffusivity of boron in Fe2 B, composition as depicted in the common tangent curve
1:19  10
13 m2 s
1 ,8) and @C=@x is the average of the (Fig. 1 and 2), where from this curve the equilibrium
gradient of surface concentration on the paste thickness for composition of boron in Fe2 B occurs below 0.3 molar
 Free Energy Problem for the Simulations of the Growth of Fe2 B Phase Using Phase-Field Method 2629

(a) (b)

Fig. 4 Numerical results that replicate 2 h boronizing process on ARMCO iron, (a) phase-field profile for case 1–5; (b) boron
concentration profile for case 1–5.

(a) (b)

Fig. 5 Numerical results that replicate 2 h boronizing process on ARMCO iron, (a) phase-field profile for case 5–11; (b) boron
concentration profile for case 5–11.

fraction of boron. This situation is considered due to the By using the optimum F constant of Fe2 B free energy
presence of flux of boron, which is introduced into the equation from the case 5 as the fixed parameter, evaluation
system for representing the open system condition of on the variations of austenite free energy has been conducted.
boronizing process, that allow the additional of mass of Figure 5 shows the numerical results for the case 5–11 of
concentration from the outer of the concerned system. free energy which is also carried out to replicate the 2 h
Considering the boronizing system as an open system is boronizing process of ARMCO iron. It can be seen from
very important, since it has also been well-established from Fig. 5(a) that a significant decreasing of depth of thickness
the previous research8,22) that boron potential (which is occurs from the case 5 to 11, which also indicates a lower
represented by the flux value of boron in the present free energy driving force occurs in this direction. In addition,
research) plays an important role in determine the way of case 5 is considered to approach the real kinetic condition
boron element impregnates into the inner surface and to the since a value of approaching 40 mm is obtained for this case,
composition of boron as well. Overall, considering the while other cases of free energy show a significant lower
numerical results of Fig. 4(a) and 4(b), the set of free value of depth of thickness. On the other hand, Fig. 5(b)
energy for the case 5 is considered to give the appropriate shows the numerical results of boron concentration profile for
results from these variations of free energy. the case 5–11. Case 5,6,7 and 8 produce a considerable value
2630 R. D. Ramdan, T. Takaki and Y. Tomita

Fig. 6 The Highlighted of common tangent curve for the optimum case of
Fig. 7 Prediction of depth of thickness that replicate 2 h boronizing process
free energy.
on ARMCO iron for different temperature.

that approaching the equilibrium condition for Fe2 B phase, 6. Conclusions

however since only case 5 approaching the experimental
kinetic condition of the growth of Fe2 B phase (Fig. 5(a)), it Development of phase-field simulations that replicate the
can be concluded that for these variations of free energy, pack boronizing of ARMCO iron has been conducted with
case 5 gives the appropriate free energy that is considered to the stressing on the evaluation of free energy driving force for
approach the real free energy driving force for the Fe2 B the Fe2 B transformation from austenite phase. Based on the
transformation from austenite phase. Furthermore, since the available free energy data from SGTE data base and using the
present variations of free energy that involves the variations assumption that there is a parabolic relationship between free
of F value (constant that determine the shape of Fe2 B free energy of boride and austenite with the boron concentration,
energy curve) and b constant (constant that determine the 11 set of free energy driving force has been established. By
concentration for the minimum free energy of austenite) comparing the experimental data of the depth of thickness of
produce a significant influences on the calculations, therefore Fe2 B phase for the similar process and evaluating the boron
consideration on these constants is very important for concentration profile for each case of free energy, case 5
determination of the free energy system of Fe2 B trans- shows the appropriate numerical results condition which
formation. The highlighted common tangent curve of the considerably approaching the real condition for the Fe2 B
optimum free energy (case 5), which has also been depicted transformation from austenite phase. In addition, the opti-
in the Fig. 1 is shown in the Fig. 6. This construction of free mum model for the present simulations also shows a good
energy as has been previously mentioned produces a agreement for the prediction of depth of thickness of boride
maximum solubility of boron in austenite phase at higher for different temperature. However since the present simu-
value, 0.0125 molar fraction of boron, compared to the lations process works on the ARMCO iron, which neglecting
0.00043 molar fraction of boron,19) which is provided in the the presence of alloying element in the steel, therefore it is
equilibrium phase diagram. However as has been explained considered if the present model might works well on the plain
before, since phase-field simulation works on the growth of carbon steel, where the presence of alloying element is
structure which is not started from the nucleation stage, this restricted.
deviation will not affect the calculation. In addition, instead
of having a line compound of Fe2 B, the production of phase
REFERENCES
diagram from the present construction of free energy will
produce Fe2 B phase line which is not independent on the 1) L. G. Yu, X. J. Chen, K. A. Khor and G. Sundararajan: Acta Mater. 53
concentration since the present construction of free energy (2005) 2361–2368.
assumes free energy to be a function of concentration and 2) D. S. Kukharev, S. P. Fisenko and S. I. Shabunya: J. Eng. Phys. Therm.
temperature. 69 (1996) 145–150.
3) M. Keddam: Appl. Surf. Sci. 236 (2004) 451–45.
Furthermore, the optimum model obtained in the present 4) B. Selçuk, R. Ipek and M. B. Karam
s: J. Mater. Proc. Tech. 141 (2003)
simulation works has been used to predict the depth of 189–196.
thickness for different temperature of 1223, 1253 and 1273 K 5) V. Jain and G. Sundararajan: Surf. Coat. Tech. 149 (2002) 21–26.
for 2 h of the process. The simulations of these situation is in 6) S. Y. Lee, G. S. Kim and B. S. Kim: Surf. Coat. Tech. 177–178 (2004)
178–184.
a good agreement with each condition of experimental
7) C. Li, M. S. Li and Y. C. Zhou: Surf. Coat. Tech. 201 (2007) 6005–
works,9) which is around 40 mm for 1223 K, between 50– 6011.
60 mm for temperature of 1253 K and around 70 mm for 8) M. Keddam: Appl. Surf. Sci. 253 (2006) 757–761.
temperature of 1273 K as shown in the Fig. 7. 9) I. Campos, J. Oseguera, U. Figueroa, J. A. Garcia, O. Bautista and
 Free Energy Problem for the Simulations of the Growth of Fe2 B Phase Using Phase-Field Method 2631

G. Keleminis: Mater. Sci. Eng. A 352 (2003) 261–265. 17) S. I. Vooijs, Y. V. Leeuwen, J. Sietsma and S. V. Zwaag: Metall. Mater.
10) A. Yamanaka, T. Takaki and Y. Tomita: Mater. Trans. 47 (2006) Trans. A 31A (2000) 379–385.
2725–2731. 18) A. T. Dinsdale: CALPHAD 15 (1991) 317–425.
11) S. Y. Hu: Comp. Coupl. Phase Diag. Thermochem. 31 (2007) 303–312. 19) T. V. Rompaey, K. C. Hari Kumar and P. Wollants: J. Alloy. Compd.
12) D. Y. Li and L. Q. Chen: Acta Mater. 46 (1998) 2573. 334 (2002) 173–181.
13) I. Steinbach: Phys. D 94 (1996) 135. 20) P. Gustafson: Scandinavian J. Metall. 14 (1985) 259–267.
14) T. Takaki, T. Hasebe and Y. Tomita: J. Cryst. Growth 287 (2006) 21) J. Y. Huh, K. K. Hong, Y. B. Kim and K. T. Kim: J. Elect. Mater. 33
495–499. (2004) 1161–1170.
15) Z. T. Trautt, M. Upmanyu and A. Kurma: Science 314 (2006) 632–635. 22) I. Campos, M. Islas, E. González, P. Ponce and G. Ramı́rez: Surf. Coat.
16) M. G. Mecozzi, J. Sietsma and S. Van der Zwaag: Comput. Mater. Sci. Tech. 201 (2006) 2717–2723.
34 (2005) 290–297.