ISIJ International, Vol. 48 (2008), No. 11, pp.

1533–1541

Aluminum Deoxidation Equilibrium of Molten Fe–Ni Alloy

Coexisting with Alumina or Hercynite

Atsutaka HAYASHI,1) Takayuki UENISHI,2) Hirotoshi KANDORI,3) Takahiro MIKI4) and Mitsutaka HINO4)

1) Formerly Graduate Student, Tohoku Univ. Now at Nippon Steel & Sumikin Stainless Steel Corporation.
2) Formerly Tohoku Univ. Now at Aichi Steel Corporation. 3) Formerly Tohoku Univ. Now at Ministry of Land,
Infrastructure, Transport and Tourism, Japan. 4) Department of Metallurgy, Graduate School of Engineering, Tohoku
University, Aoba-yama 6-6-02, Sendai 980-8579 Japan.
(Received on June 3, 2008; accepted on July 31, 2008 )

Aluminum deoxidation equilibrium of molten Fe–Ni alloy was determined by chemical equilibrium method
at temperature of 1 873 to 1 973 K. Also, critical point in molten Fe–40mass%Ni alloy that coexisted with
Al2O3 and FeO · Al2O3 was determined experimentally at 1 973 K. Solvent extraction method was applied to
improve the analytical accuracy of the low aluminum content in molten Fe–Ni alloy. The relation between
aluminum and oxygen contents in molten Fe–Ni alloy equilibrated with Al2O3 or FeO · Al2O3 was estimated
in the whole alloy composition range at 1 873 to 1 973 K by utilizing the metallic solution model based on
Darken’s quadratic formalism and Redlich–Kister type polynomial.
KEY WORDS: quadratic formalism; thermodynamics; activity; excess free Gibbs energy; Redlich–Kister
polynomial; aluminum deoxidation equilibrium; nickel; iron; iron–nickel alloy; alumina; hercynite.

loys was carried out by application of the above-mentioned

1. Introduction
Aluminum has been used as a deoxidizer due to its
strong affinity with oxygen. Reliable thermodynamic infor-
mation of Al deoxidation equilibrium is important for in-
clusion control in clean steel and its alloy production. Many
researchers1–13) have reported Al deoxidation equilibrium of
molten iron. However, measurements on Fe–Ni alloy are
limited.14–20) Ishii et al.14) have reported Al deoxidation of
molten Ni–Fe alloy at temperature of 1 823–1 973 K. The
composition range was between 0 and 50 mass% Fe. Cho
and Suito15) have reported Al deoxidation equilibrium for
Fe–30, 50, 70mass%Ni alloy at 1 873 K. Li et al.16) have re-
ported Al deoxidation equilibrium for Fe–15, 29mass%Ni
alloy at 1 873 K. Ohta and Suito17) have measured Al deoxi-
dation equilibrium of Fe–10, 20, 40, 60mass%Ni alloys at
1 873 K. Lee et al.,18) Katsuki and Yamauchi,19) and Fuji-
wara et al.20) have reported Al deoxidation equilibrium for
Fe–36mass%Ni alloy at 1 773, 1 873, and 1 973 K, respec-
tively. However, Al deoxidation equilibrium of whole Fe–Ni
alloy composition range has not been determined yet at var-
ious temperatures.
Recently, some of the present authors have developed a
thermodynamic treatment to express deoxidation equilib-
rium on not only pure metals and also alloys. This model is
based on Darken’s quadratic formalism21,22) and Redlich–
Kister type polynomial,23,24) and the details are explained in
the previous papers.25–29) Aluminum deoxidation equilib-
rium of molten Fe, Fe–20mass%Ni and Fe–40mass%Ni
alloy was determined by chemical equilibrium method at
temperature of 1 873 to 1 973 K in the present work. Also,
numerical analysis on Al deoxidation of molten Fe–Ni al-
solution model using the present experimental data and lit-
erature values.
Alumina (Al2O3) normally is the deoxidation product
when aluminum is used as a deoxidizer. However, hercynite
(FeO · Al2O3) may form when the oxygen content in iron is
high. The critical point of Al2O3 and FeO · Al2O3 coexis-
tence has been investigated by several researchers.30–35) The
oxygen content in iron, that alumina and hercynite co-
existed at 1 873 K, was reported by Kim and McLean30)
(0.074 mass% O), Pillay et al.31) (0.079 mass% O), McLean
and Ward32) (0.0623 mass% O), and Novokhatsky and
Belov33) (0.0574 mass% O), Wasai and Mukai34) (0.074
mass% O), and Wasai et al.35) (0.072 mass% O). The alumi-
num content at this condition was reported by Novokha-
tsky and Belov33) (0.0070 mass% Al), Wasai and Mukai34)
(0.0060 mass% Al), and Wasai et al.35) (0.0018 mass%
Al). The results of Novokhatsky and Belov,33) Wasai and
Mukai,34) and Wasai et al.35) are shown in Fig. 1 with the
reported1–6) aluminum deoxidation results and the present
results obtained later in molten iron at 1 873 K. The point
that Al2O3 and FeO · Al2O3 coexists should of course be on
the Al2O3 saturation curve. However, large discrepancy is
observed in Fig. 1.
The present authors suggest that aluminum content at
critical point is much lower than the reported results33–35) as
shown in Fig. 1. Pillay et al.31) has reported that aluminum
content of the coexisting point at 1 973 K was less than
0.0005 mass% Al, which was much lower than other re-
serchers.33–35) The reason of the discrepancy may be due
to the difficulty of analysis at low aluminum content.
Aluminum deoxidation product increases with nickel ad-

1533 © 2008 ISIJ

 ISIJ International, Vol. 48 (2008), No. 11
Fig. 1. Aluminum deoxidation of liquid iron at 1 873 K.
dition to iron and shows the maximum at around Fe–
40mass%Ni.36) Also, aluminum deoxidation product in-
creases with temperature. Therefore, the confirmation of
the critical point will be easier if the Ni content and tem-
perature are high. Hence, critical point in molten Fe–
40mass%Ni alloy that coexisted with Al2O3 and FeO ·
Al2O3 was investigated at 1 973 K in the present work. Also,
the relation between aluminum and oxygen content in
molten Fe–Ni alloy equilibrated with Al2O3 or FeO · Al2O3
was estimated for the whole alloy composition range at
1 873 to 1 973 K by utilizing the present results.
2. Experimental Procedures
Measurement was performed with an induction furnace
10 kV A described elsewhere.37,38) The apparatus consisted
of a working quartz tube, gas purification trains and gas
flow meters. The melt temperature was measured with an
optical pyrometer calibrated with the melting points of iron
and nickel. Electrolytic metals of nickel and iron were uti-
lized in the experiment.
Iron and/or nickel weighing approximately 100 g were
placed in an alumina crucible, 25 mm in ID and 50 mm in
depth, which was put in an outer protective alumina cru-
cible, and the space between the two was filled with alu-
mina sand. Iron oxide (Fe2O3) was added to the crucible to
increase the initial oxygen content in the alloy for some ex-
periments. After completion of the working tube assembly,
the metal was melted in a purified 67vol%hydrogen–argon
gas mixture atmosphere for 2–3 h in order to reduce the dis-
solved oxygen in the melt. The atmosphere was then
changed to purified argon maintained at 100 mL/min. Alu-
minum or Fe–40mass%Ni–0.1mass%Al alloy, prepared
prior to the experiment, was added from the top of the
working tube after the melt was kept at the desired experi-
mental temperature. After the sample was held for 60 min,
which was predetermined period for equilibrium attain-
ment, the furnace power was switched off and the sample
was quenched.
The samples were cut, polished by SiC abrasive paper
and then electrolytically-polished twice for 3 min under the
condition of 5 V–0.5 A. Mixed acid that contains 80 mass%
acetic acid and 20 mass% perchloric acid was used for elec-
trolyte. The samples were ultrasonically cleaned in acetone
and immediately subjected to oxygen analysis by inert gas
© 2008 ISIJ
Fig. 2. Effect of electropolish on oxygen analysis.
fusion infrared absorptionmetry to avoid effect of oxygen
adsorped on sample surface. The effect of electropolish on
oxygen analysis is shown in Fig. 2. Some of the samples
were analyzed before and after electropolish. Oxygen con-
tent became lower after electropolish, and the average de-
crease of oxygen content was 0.0016 mass%. Therefore,
electropolish treatment is crucial to obtain accurate oxygen
content of alloys, especially in low oxygen content region.
Aluminum content was analyzed by an induction coupled
plasma emission spectrometry (ICP-AES). Solvent extrac-
tion method was applied to improve the analytical accuracy
of the low aluminum content in alloy. Wavelength of
396.152 nm was used for analyzing aluminum by ICP-
AES for most of the samples. However, wavelength of
167.079 nm was used for samples with low aluminum con-
tent due to its high intensity. As iron in the solution affected
the intensity of aluminum at this wavelength, iron in the so-
lution was eliminated by use of methyl isobutyl ketone
(MIBK) as shown in Fig. 3. Iron was completely removed
into the MIBK phase (upper phase in Fig. 3) by repeating
solvent extraction 4 times and the analysis became possible
to low Al content as 0.02 mg Al/L. It has been confirmed
that change of Ni content in the alloy was negligible during
the experiments.
The interface between metal and alumina crucible was
observed by SEM. Also, oxide composition at the vicinity
of the interface was analyzed by EDS.
1534
 ISIJ International, Vol. 48 (2008), No. 11

Fig. 3. Removal of iron by solvent extraction.

Fig. 4. SEM-EDS results for sample 56.

details were explained in the previous papers.25–29) Pure

3. Results and Discussion
3.1. Deoxidation of Fe–Ni Alloy with Al
Experimental results are shown in Table 1. The alu-
minum content of sample 51 and 52 could not be deter-
mined due to analytical limit. Result of SEM-EDS for sam-
ple 56 is shown in Fig. 4. Crucible is on the right-hand side
and resin is on the left-hand side. Resin filled the space be-
tween metal and crucible generated during sample quench-
ing. New oxide phase formation can be seen for sample 56.
Molar ratio between iron and aluminum of the new phase
was 2. Hence, it was confirmed that formed hercynite
(FeO · Al2O3) was in equilibrium with molten alloy in this
sample. Formation of hercynite was also observed for sam-
ples 51 to 55. Dissolution of nickel oxide in alumina and
hercynite was found to be negligible by SEM-EDS analysis.
3.2. Numerical Analysis on Al Deoxidation of Fe–Ni
Alloy
Quadratic formalism and Redlich–Kister type polyno-
mial were used in the present numerical analysis and the
substance is chosen as a standard state (Raoultian standard
state) for condensed phases in the present work. Dissolved
oxygen in the melt equilibrating with 101 325 Pa (1 atm)
oxygen is selected as a standard state for oxygen. The rela-
tion between the oxygen activity and oxygen partial pres-
sure in this standard state can be expressed as following
equation.
ao
(PO2/P –o)1/2 ...............................(1)
(P –o: standard pressure in Pa, PO2: partial pressure of
O2 gas in Pa)
Therefore, the oxygen activity or equilibrium oxygen par-
tial pressure is independent of the kind of metal solvent.
Al deoxidation reaction can be expressed by Eq. (2).
2Al(l)3O
Al2O3(s) .........................(2)
This equation can be separated into following equations.
3
2 Al(l)  O2 (g)
Al 2O3 (s) ...................(3)
2

1535 © 2008 ISIJ

 ISIJ International, Vol. 48 (2008), No. 11

Table 1. Experimental results on Al deoxidation of molten

Fe–Ni alloys. ∆G°f,Al2O3
ln K Al

RT

ln aAl2O3  2 ln γ Al  3 ln γ O  2 ln X Al  3 ln X O

 2 ln γ Al  3 ln γ O  2 ln X Al  3 ln X O
(
aAl2O3
1) ......................................(5)

Where Xi, ai and g i denotes mole fraction, activity and the

activity coefficient of component i, respectively. The activ-
ity of Al2O3 is taken as unity, since deoxidation product is
pure Al2O3 except samples 51 to 56.
The excess free energy change of mixing in Fe–Ni–
Al–O ternary system can be expressed as Eq. (6) by
Redlich–Kister type polynomial.

∆GMex
X Fe X Ni {0 Ω Fe–Ni  1Ω Fe–Ni ( X Fe  X Ni )}

 X Fe X Al {0 Ω Fe–Al  1Ω Fe–Al ( X Fe  X Al )}
 X Fe X O {0 Ω Fe–O  1Ω Fe–O ( X Fe  X O )}
 X Ni X Al {0 Ω Ni–Al  1Ω Ni–Al ( X Ni  X Al )}
 X Ni X O {0 Ω Ni–O  1Ω Ni –O ( X Ni  X O )}
 X Al X O 0 Ω Al–O ...........................................(6)

The partial molar excess free energy change of Fe, Al and

O can be derived as follows.
∆GFeex
RT ln γ Fe
∂∆G ex ∂∆G ex ∂∆G ex

∆G ex  X Ni  X Al  XO
∂X Ni ∂X Al ∂X O

X Al (1 X Fe ) 0 Ω Fe–Al
 X Al ( X Al  2 X Fe  2 X Fe X Al  2 X Fe
2 1
) Ω Fe–Al
 X Ni X Al 0 Ω Ni–Al  2 X Ni X Al ( X Ni  X Al ) 1Ω Ni–Al
 X O (1 X Fe ) 0 Ω Fe–O
 X O ( X O  2 X Fe  2 X Fe X O  2 X Fe
2 1
) Ω Fe–O
 X Ni X O0 Ω Ni–O  2 X Ni X O ( X Ni  X O ) 1Ω Ni–O
 X Al X O0 Ω Al–O  X Ni (1 X Fe ) 0 Ω Fe–Ni
 X Ni ( X Ni  2 X Fe  2 X Fe X Ni  2 X Fe
2 1
) Ω Fe– Ni ....(7)

∆GAlex
RT ln γ Al
∂∆G ex ∂∆G ex ∂∆G ex

∆G ex X Ni (1 X Al ) X O
∂X Ni ∂X Al ∂X O

X Fe (1 X Al ) 0 Ω Fe–Al
 X Fe ( X Fe2 X Al 2 X Fe X Al 2 X Al
2 1
) Ω Fe–Al
3
3O
O2 (g) ...............................(4)  X Ni (1 X Al ) 0 Ω Ni–Al
2
 X Ni ( X Ni 2 X Al 2 X Ni X Al 2 X Al
2 1
) Ω Ni–Al
The Gibbs free energy change of Eq. (3) is identical to the  X Fe X O0 Ω Fe–O2 X Fe X O ( X Fe X O ) 1Ω Fe–O
Gibbs free energy of Al2O3 formation (DG°f,Al2O3 ) and that of  X Ni X O0 Ω Ni–O2 X Ni X O ( X Ni  X O ) 1Ω Ni–O
Eq. (4) is zero due to the relation of Eq. (1). Therefore, the
Gibbs free energy change of Eq. (2) will be identical to the  X O (1 X Al ) 0 Ω Al–O X Fe X Ni0 Ω Fe–Ni
Gibbs free energy change of Al2O3 formation (DG°f,Al2O3 ). 2 X Fe X Ni ( X Fe X Ni ) 1Ω Fe–Ni ..............................(8)
Equation (2) and the Gibbs free energy change of Al2O3
formation can be utilized for Al deoxidation of any metal or
alloy, and this is a great advantage. The equilibrium con-
stant of Eq. (2), KAl, can be expressed as follows.

© 2008 ISIJ 1536

 ISIJ International, Vol. 48 (2008), No. 11

∆GOex
RT ln γ O in the present work are shown below.

∂∆G ex
∂∆G ex
∂∆G ex 0
W Fe–O
415 400142.4T/J .................(11)

∆G ex X Ni  X Al (1 X O )
∂X Ni ∂X Al ∂X O
W Fe–O
298 300117.8T/J ...................(12)
1

 X Fe X Al 0 Ω Fe–Al  X Fe X Al ( X Fe X Al ) 1Ω Fe–Al
 X Ni X Al 0 Ω Ni–Al  X Ni X Al ( X Ni  X Al ) 1Ω Ni–Al 0
W Fe–Al
275 700106.9T/J .................(13)
 X Fe (1 X O ) Ω Fe–O
0

 X Fe ( X Fe2 X O2 X Fe X O2 X O2 ) 1Ω Fe–O

 X Ni (1 X O ) 0 Ω Ni–O
 X Ni ( X Ni 2 X O2 X Ni X O2 X O2 ) 1Ω Ni–O
 X Al (1 X O ) 0 Ω Al–O X Fe X Ni 0 Ω Fe–Ni
2 X Fe X Ni ( X Fe X Ni ) 1Ω Fe–Ni...............................(9)
Equation (10) can be deduced by substituting Eqs. (8) and
(9) into Eq. (6).
X Fe (25 X Al ) 0 Ω Fe– Al
2 X Fe ( X Fe 2 X Al 5 X Fe X Al 5 X Al
2 1
) Ω Fe– Al
 X Ni (25 X Al ) 0 Ω Ni–Al
2 X Ni ( X Ni 2 X Al 5 X Ni X Al 5 X Al
2 1
) Ω Ni–Al
 X Fe (35 X O ) 0 Ω Fe– O
 X Fe (3 X Fe 6 X O10 X Fe X O10 X O2 ) 1Ω Fe– O
 X Ni (35 X O ) 0 Ω Ni–O
 X Ni (3 X Ni 6 X O10 X Ni X O10 X O2 ) 1Ω Ni–O
(3 X Al 2 X O5 X Al X O ) 0 Ω Al–O
5 X Fe X Ni 0 Ω Fe– Ni 10 X Fe X Ni ( X Fe  X Ni ) 1Ω Fe– Ni
2 RT ln X Al 3RT ln X O∆G °f,Al2O3
0 ...............(10)
Equation (10) is the fundamental equation for numerical
analysis on Al deoxidation of molten Fe–Ni alloy.
First, numerical analysis was conducted for Al deoxida-
tion of molten Fe. Hence, XNi in Eq. (10) was set zero. The
binary interaction parameters 0W Fe–O and 1W Fe–O were deter-
mined in the previous work.26) The binary interaction pa-
rameters 0W Fe–Al and 1W Fe–Al were obtained from 0W Fe–O and
1
W Fe–O26) and the assessed equilibrium constant of Al deoxi-
dation in molten iron reported by Itoh et al.39) The Gibbs
free energy of Al2O3 formation was taken from NIST-
JANAF Thermochemical Tables.40) The parameters utilized
1
W Fe–Al
79 94035.85T/J....................(14)
DG°f,Al2O3
1 682 300324.15T/J .............(15)
The following relation can be derived by substituting zero
for XNi and rearranging Eq. (10).
YFe(Al –O)
X Fe (35 X O ) 0 Ω Fe– O
 X Fe (3 X Fe 6 X O10 X Fe X O10 X O2 ) 1Ω Fe– O
 X Fe (25 X Al ) 0 Ω Fe– Al
2 X Fe ( X Fe 2 X Al 5 X Fe X Al 5 X Al
2 1
) Ω Fe– Al
2 RT ln X Al 3RT ln X O∆G°f,Al2O3

(2 X O3 X Al 5 X Al X O ) 0 Ω Al–O ................(16)
Literature values1–13) were also utilized for analysis on Al
deoxidization equilibrium in molten Fe. Interaction parame-
ter 0W Al–O can be obtained from the slope of the regressed
line that runs through the origin by taking the left-hand side
of Eq. (16) as a vertical axis and (2XO3XAl5XAlXO) as
a horizontal axis at each temperature. The relation between
left-hand side of Eq. (16) as a vertical axis and (2XO
3XAl5XAl XO) at 1 873 K and 1 973 K is shown in Figs. 5
and 6, respectively. Interaction parameter 0W Al–O was deter-
mined as a temperature function as follows.
0
W Al–O
856 3001 497T/J ...................(17)
The regressed curves determined in the present work are
shown with literature values in Fig. 7. The present result
agreed especially well with the results of Gokcen and Chip-
man9) and Kobayashi et al.13)
In case of numerical analyzing Al deoxidation of molten
Fe–Ni alloy, binary interaction parameters of Fe–Ni, Ni–Al,
and Ni–O systems must be additionally considered with the
binary interaction parameters used for the numerical analy-
sis of Al deoxidation in molten Fe. The previously re-
ported26) binary interaction parameters of Fe–Ni and Ni–O
were utilized in the present work. Experimental results for

Fig. 5. Determination of 0W Al–O at 1 873 K. Fig. 6. Determination of 0W Al–O at 1 973 K.

1537 © 2008 ISIJ

 ISIJ International, Vol. 48 (2008), No. 11

Fig. 7. Relation between aluminum and oxygen equilibrium concentration in Al deoxidation of Fe.

Al deoxidation of molten Fe–Ni alloys reported by Ishii and

Ban-ya,14) Cho and Suito,15) Li and Suito16) and Ohta and
Suito17) were utilized for the analysis. These binary interac-
tion parameters of Fe–Ni system26) and Ni–O system26) are
shown in the following.
0
W Fe–Ni
16 9115.1622T/J .................(18)
1
W Fe–Ni
10 1804.146656T/J.................(19)
0
W Ni–O
106 50044.80T/J .................(20)
1
W Ni–O
35 50015.92T/J ....................(21)
Fig. 8. Determination of 0W Ni–Al and 1W Ni–Al at 1 873 K.
Equations (18) and (19) are valid for any composition of
Fe–Ni binary, while Eqs. (20) and (21) are valid in the di-
lute region of O in molten Ni. Rearranging Eq. (10) derives
Eq. (22).

Fig. 9. Determination of 0W Ni–Al and 1W Ni–Al at 1 973 K.

1
W Ni–Al
101 40089.47T/J...................(24)

Estimated Al deoxidation equilibrium curves of molten

.........................................(22)
Utilizing Eqs. (11)–(15), (17)–(21), present experimental
results and reported by various researchers and taking
YNi(Al) in Eq. (22) as a vertical axis and 2(XNi2XAl
5XNi XAl5X 2Al)/(25XAl) as a horizontal axis, 0W Ni–Al and
1
W Ni–Al can be determined from the intercept and the slope
of the regressed line, respectively. The results at 1 873 K
and 1 973 K are shown in Figs. 8 and 9, respectively, and
the following values were obtained from regression.
0
W Ni–Al
429 700188.2T/J .................(23)
Fe–Ni alloy at 1 873 K are shown in Fig. 10 with present
experimental data and literature values.14–17) The present es-
timated result agrees extremely well with all of experimen-
tal results, and it was confirmed that quadratic formalism
and Redlich–Kister type polynomial could be utilized to ex-
press Al deoxidation equilibrium in the whole region of the
Fe–Ni alloy.
The relation between equilibrium Al and O content in Al
deoxidation of Ni was calculated utilizing equation substi-
tuting XFe
0 in Eq. (10). Equations. (17), (20), (21), (23)
and (24) were utilized for the binary interaction parameters.
The relation is compared with the results reported by Ishii
and Ban-ya14) in Fig. 11. It was confirmed from the present

© 2008 ISIJ 1538

 ISIJ International, Vol. 48 (2008), No. 11
Fig. 10. Relation between aluminum and oxygen equilibrium
concentration in Al deoxidation of Fe–Ni alloy at
1 873 K.
Fig. 11. Relation between aluminum and oxygen equilibrium
concentration in Al deoxidation of Ni.
numerical analysis that the Al deoxidation equilibrium of
molten Fe, Ni and Fe–Ni alloys could be quantitatively ex-
pressed with high accuracy.
3.3. Equilibrium among Fe–Ni Alloy, Al2O3 and
FeO · Al2O3
Equilibrium between hercynite (FeO · Al2O3) and molten
iron alloy can be expressed by Eq. (25).
Fe(l)2Al(l)4O
FeO · Al2O3(s) .............(25)
The equilibrium constant of Eq. (25), KFeO · Al2O3, can be ex-
pressed as follows.
∆G°f,FeO ⋅ Al2O3
ln K FeO ⋅ Al2O3

RT

ln aFeO ⋅ Al2O3  ln γ Fe 2 ln γ Al 4 ln γ O
 ln X Fe 2 ln X Al 4 ln X O ................(26)
Here, DG°f,FeO · Al2O3 is the Gibbs free energy change of her-
cynite (FeO · Al2O3) formation. The following relations can
be obtained by substituting Eqs. (8)–(10) into Eq. (26).
1539
Fig. 12. Determination of critical point that alumina and her-
cynite coexist with Fe–40mass%Ni alloy at 1 973 K.
...............................(27)
Dissolution of nickel oxide in hercynite was found to be
negligible from the SEM-EDS analysis. Hence, the activity
of hercynite in Eq. (27) was assumed as unity in the present
work.
The critical point that alumina and hercynite coexist with
molten Fe–40mass%Ni alloy at 1 973 K can be estimated
from the present work. Oxide in equilibrium with samples
55 and 56 in Table 1 was hercynite and that with sample 57
was alumina as shown in Fig. 12. Therefore, the critical
point should be in the range of 0.00013[mass%Al]
0.00075 and 0.019[mass%O]0.037 from the present ex-
perimental result. The Gibbs free energy change of her-
cynite formation DG°f,FeO · Al2O3 at 1 973 K should be between
1 268 kJ and 1 262 kJ to satisfy the range of critical
point from Eq. (27) and the parameters obtained in the
present work. The Gibbs free energy of FeO · Al2O3 forma-
tion can be obtained from combining the reported values of
Kubaschewski,41) Barin et al.42) and Richardson et al.43)
with NIST-JANAF Thermochemical Tables40) as shown in
Fig. 13. However, extrapolation of literature values to
1 973 K does not satisfy the required range of 1 268 to
1 262 kJ. Therefore, the Gibbs free energy change of her-
cynite formation was estimated as Eq. (29) using the aver-
age entropy value of these literature values,40–43) 406.0 J/K,
and the estimation in average value of DG°f,FeO · Al2O3 the pres-
ent work at 1 973 K.
Fe(l)2Al(l)2O2(g)
FeO · Al2O3(s)...........(28)
© 2008 ISIJ
 ISIJ International, Vol. 48 (2008), No. 11
Fig. 13. Determination of the Gibbs free energy change of her-
cynite formation.
Fig. 14. Relation between aluminum and oxygen contents in
molten Fe–Ni alloy saturated with alumina or hercynite
at 1 873 and 1 973 K.
DG°f,FeO · Al2O3/J
2 066 000406.0T............(29)
The relation between aluminum and oxygen content in
molten Fe–Ni alloy saturated with alumina or hercynite at
1 873 and 1 973 K estimated from Eqs. (10) and (27) is
shown in Fig. 14. The critical point shifts to higher Al con-
tent with increase of Ni content and has a maximum at ap-
proximately 90 mass% Ni. This can be explained qualita-
tively by the following reason. Hercynite tends to form on
the conditions that have high Fe and O activity from the fol-
lowing equation.
Fe(l)OAl2O3(s)
FeO · Al2O3(s).............(30)
The activity of iron decreases with increase of Ni content in
alloy. However, the oxygen activity coefficient at infinite di-
lution in molten Ni, g O(in
° Ni), is 0.425 and that in molten Fe,
g O(in
° Fe), is 0.0151 at 1 973 K from Eq. (9) on the condition
of XNi
1 or XFe
1. The activity coefficient of oxygen in
molten Ni is 28 times larger than that in molten Fe at
1 973 K. Nickel content dependence on aFe(l), g O° and
aFe(l) · g O° in Fe–Ni(–O) system at 1 973 K is shown in Fig.
15. The O activity coefficient increases rapidly in high Ni
content region while the Fe activity decreases gradually
with Ni addition to the alloy. The value of aFe(l) · g O° has the
maximum at approximately 80 mass% Ni. This composition
does not agree with 90 mass% Ni, which Al content at the
critical point had the maximum in Fig. 14. This difference
is due to the effect of Al on O activity coefficient in the
© 2008 ISIJ
Fig. 15. Nickel content dependence on aFe(l), g O° and aFe(l) · g O° for
Fe–Ni(–O) system in 1 973 K.
Fe–Ni–O system. However, the reason for tendency of her-
cynite formation at high Ni content can be explained quali-
tatively with this simple consideration.
The critical point that alumina and hercynite coexists
with molten Fe was estimated to be [mass ppm Al]
0.44
and [mass%O]
0.052 at 1 873 K and it is shown in Fig. 1.
The oxygen content at the critical point was close to
0.074 mass% O,30) 0.079 mass% O,31) 0.0623 mass% O,32)
0.0574 mass% O,33) 0.074 mass% O34) and 0.072 mass%
O35) in the literatures. However, reported aluminum content
does not agree with the literatures. This may be due to diffi-
culty of Al analysis in molten Fe.
4. Conclusion
Aluminum deoxidation equilibrium of molten Fe–Ni
alloy was determined by chemical equilibrium method at
temperature of 1 873 to 1 973 K. Al deoxidation equilibrium
in molten Fe, Ni and Fe–Ni alloys was quantitatively ex-
pressed by binary interaction parameters determined on
basis of the quadratic formalism and Redlich–Kister type
polynomial using the present results and literature values.
Critical point in molten Fe–40mass%Ni alloy that coex-
ists with Al2O3 and FeO · Al2O3 was determined experimen-
tally at 1 973 K. Solvent extraction method was applied to
improve the analytical accuracy of the low aluminum con-
tent in molten Fe–Ni. The Gibbs free energy change of
FeO · Al2O3 was estimated from the present result. The criti-
cal aluminum and oxygen contents in molten Fe–Ni alloy
equilibrated with Al2O3 or FeO · Al2O3 was estimated for
the whole alloy composition range at 1 873 to 1 973 K by
utilizing the present result and the previous numerical
analysis on Al deoxidation of molten Fe–Ni alloy. The criti-
cal point shifts to higher Al content with Ni addition to Fe.
The critical point that alumina and hercynite coexists with
molten Fe was estimated to be [mass ppm Al]
0.44 and
[mass%O]
0.052 at 1 873 K.
Acknowledgement
Authors would like to thank late Mr. M. Yonemoto, for-
mer undergraduate student, for helping experiments and
chemical analysis.
1540
 ISIJ International, Vol. 48 (2008), No. 11

REFERENCES 25) T. Miki, F. Ishii and M. Hino: Mater. Trans., 44 (2003), 1817.
26) T. Miki and M. Hino: ISIJ Int., 44 (2004), 1800.
1) R. J. Fruehan: Metall. Trans., 1 (1970), 3403. 27) T. Miki and M. Hino: ISIJ Int., 45 (2005), 1848.
2) L. E. Rohde, A. Choudhury and M. Wahlster: Arch. Eisenhüttenwes., 28) W. Y. Cha, T. Nagasaka, T. Miki, Y. Sasaki and M. Hino: ISIJ Int., 46
42 (1971), 165. (2006), 996.
3) D. Janke and W. A. Fischer: Arch. Eisenhüttenwes., 47 (1976), 195. 29) T. Miki, K. Tsujita, S. Ban-ya and M. Hino: Calphad, 30 (2006),
4) H. Suito, H. Inoue and R. Inoue: ISIJ Int., 31 (1991), 1381. 449.
5) H. Suito, R. Inoue and A. Nagatani: Steel Res., 63 (1992), 419. 30) C. K. Kim and A. McLean: Proc. of Int. Symp. Met. Slag Gas Reac-
6) J. D. Seo, S. H. Kim and K. R. Lee: Steel Res., 69 (1998), 49. tions and Processes, Electrochemical Soc., Princeton, N.J., (1975),
7) G. K. Sigworth and J. F. Elliott: Met. Sci., 8 (1974), 298. 284.
8) S.-W. Cho and H. Suito: ISIJ. Int., 34 (1994), 177. 31) T. C. Pillay, J. d’Entremont and J. Chipman: J. Am. Ceram. Soc., 43
9) N. A. Gokcen and J. Chipman: J. Met., 5 (1953), 173. (1960), 583.
10) J. C. d’Entremont, D. L. Guernsey and J. Chipman: Trans. Metall. 32) A. McLean and R. G. Ward: J. Iron Steel. Inst., 204 (1966), 8.
Soc. AIME, 227 (1963), 14. 33) I. A. Novokhatskiy and B. F. Belov: Russ. Metall., (1966), Nr. 1, 12.
11) A. McLean and H. B. Bell: J. Iron Steel Inst., 203 (1965), 123. 34) K. Wasai and K. Mukai: J. Jpn. Inst. Met., 52 (1988), 1088.
12) J. H. Swisher: Trans. Metall. Soc. AIME, 239 (1967), 123. 35) K. Wasai, K. Mukai, H. Fuchiwaki and A. Yoshida: ISIJ Int., 39
13) K. Kobayashi, Y. Ohmori and M. Sanbongi: Tetsu-to-Hagané, 53 (1999), 760.
(1967), 323. 36) A. Hayashi, T. Miki, Z. Ren and M. Hino: CAMP-ISIJ, 19 (2006),
14) F. Ishii, S. Ban-ya and M. Hino: ISIJ Int., 36 (1996), 25. 79.
15) S.-W. Cho and H. Suito: Steel Res., 66 (1995), 237. 37) F. Ishii and S. Ban-ya: Tetsu-to-Hagané, 75 (1989), 2188.
16) G. Li, R. Inoue and H. Suito: Steel Res., 67 (1996), 528. 38) F. Ishii and S. Ban-ya: ISIJ Int., 32 (1992), 1091.
17) H. Ohta and H. Suito: ISIJ Int., 43 (2003), 1301. 39) H. Itoh, M. Hino and S. Ban-ya: Tetsu-to-Hagané, 83 (1997), 773.
18) S.-B. Lee, S.-M. Jung, H.-G. Lee and C.-H. Rhee: ISIJ Int., 42 40) NIST-JANAF Thermochemical Tables, 4th ed., ed. by M. W. Chase,
(2002), 679. Jr., J. Phys. Chem. Ref. Data, (1998).
19) J. Katsuki and T. Yamauchi: CAMP-ISIJ, 7 (1994), 1076. 41) O. Kubaschewski: High Temp.–High Press., 4 (1972), 1.
20) H. Fujiwara, A. Hattori and E. Ichise: Tetsu-to-Hagané, 85 (1999), 42) I. Barin, O. Knacke and O. Kubaschewski: Thermochemical Proper-
201. ties of Inorganic Substances, Berlin, New York, Springer-Verlag,
21) L. S. Darken: Trans. Met. Soc. AIME, 239 (1967), 80. (1977).
22) L. S. Darken: Trans. Met. Soc. AIME, 239 (1967), 90. 43) F. D. Richardson, J. H. E. Jeffes and G. Withers: J. Iron Steel Inst.,
23) M. Hillert and L.-I. Staffanson: Acta Chem. Scand., 24 (1970), 3618. 166 (1950), 213.
24) N. Saunders and A. P. Miodownik: Calphad (Calculation of Phase
Diagrams.), A Comprehensive Guide, Pergamon, Oxford, (1988), 91.

1541 © 2008 ISIJ