Materials Transactions, Vol. 45, No. 10 (2004) pp.

2945 to 2950
#2004 The Japan Institute of Metals

Low Temperature Grain Boundary Diusion of Chromium

in SUS316 and 316L Stainless Steels
Masaki Mizouchi1 , Yoshihiro Yamazaki1 * , Yoshiaki Iijima1 and Koji Arioka2
1
2

Department of Materials Science, Graduate School of Engineering, Tohoku University, Sendai 980-8579, Japan
Institute of Nuclear Safety System, Incorporated, Mihama-cho, Fukui 919-1205, Japan

Grain boundary diusivity of chromium in SUS316 and 316L stainless steels has been determined in the temperature range between 518
and 1173 K. The magnitudes of the grain boundary diusivities in four kinds of specimens are in the order of the cold-worked SUS316, the
solution-treated SUS316L, the solution-treated SUS316 and the sensitized SUS316. The grain boundary diusivities in these specimens are
remarkably higher than those of previous works. The activation energies for the former are 8591 kJmol
1 , whereas those for the latter are
151234 kJmol
1 .
(Received May 24, 2004; Accepted August 10, 2004)
Keywords: grain boundary diusion, SUS316 and SUS316L stainless steels, inter-granular stress corrosion cracking (IGSCC), chromium

1.

Introduction

Austenitic stainless steels have been widely used as

structural materials in nuclear power plants because of their
excellent strength at high temperature and good corrosion
resistance. However, inter-granular stress corrosion cracking
(IGSCC) of these stainless steels has been observed in service
in Boiling Water Reactors (BWR) after a very long
duration.1) Incidents of failure by IGSCC have led to various
investigations of chemical, metallurgical and mechanical
factors associated with the cracking of the austenitic stainless
steels.1) Arioka and coworkers2,3) have studied the temperature dependence of IGSCC under hydrogenated water and
observed a low activation energy of 107 kJmol
1 . They have
also noted that the grain boundary sliding (creep) might play
a signicant role on IGSCC.2,3) On the other hand, the
activation energy for steady state creep in SUS316 above
798 K is 219 kJmol
1 , while that below 798 K is 110
kJmol
1 .2) To understand the degradation and fracture
processes of the stainless steels, information on grain
boundary diusion is important. In particular, experimental
data on it at lower temperatures around 600 K, which is an
ordinary temperature region in service of light water reactors
(LWR), are essential. However, experiments on the grain
boundary diusion in austenitic stainless steels have been so
far limited to the high temperature range above 800 K,4)
because experiments of grain boundary diusion as well as
volume diusion at lower temperatures are not easy.
Perkins et al.5) have measured the grain boundary
diusivity of 51 Cr in Fe-17Cr-12Ni alloy (concentration of
solute is expressed in the unit of mass%) in the temperature
range between 849 and 1323 K. They have obtained the
preexponential factor 3:5  10
14 m3 s
1 and the activation
energy 151 kJmol
1 . Smith6) has measured the grain boundary diusivity of 51 Cr in Fe-16.8Cr-11.6Ni-2.55Mo1.64Mn-0.5Si-0.04C alloy in the temperature range from
923 to 1285 K. The preexponential factor 1:5  10
12 m3 s
1
and the activation energy 191 kJmol
1 have been obtained.
ermak7) has measured the diusivity of 51 Cr in Fe-18CrC
*Corresponding

author, E-mail address: yamazaki@material.tohoku.ac.jp

12Ni and Fe-21Cr-31Ni alloys in the temperature range from

973 to 1223 K and obtained the diusion parameters, 1:2 
ermak
10
10 m3 s
1 and 234 kJmol
1 , for both of the alloys. C
et al.8) have also studied the inuence of carbon addition on
the grain boundary diusivity of 51 Cr in Fe-20Ni-10Cr alloy.
The grain boundary diusivity decreases with increasing of
the carbon content of 0.002 to 0.175 mass%. In all these
experiments, grain boundary diusivity has been deduced
from the analysis of type B kinetics using the numerical value
of the volume diusion coecient measured in a relatively
high temperature region. Therefore, these data on grain
boundary diusion are also limited to the corresponding high
temperature region.
In view of these considerations, the present paper reports
the volume and grain boundary diusion coecients of
chromium using 51 Cr tracer in the solution-treated SUS316
and SUS316L, the sensitized SUS316 and the cold-worked
SUS316 alloys over a temperature range of 5181173 K.
2.

Experimental Procedure

Plates 5 mm in thickness of SUS316 and 316L austenitic

stainless steels were supplied by Institute of Nuclear Safety
System, Incorporated. The chemical compositions of the
SUS316 and 316L stainless steels are shown in Table 1. Four
kinds of specimens were prepared to examine the eect of
carbon, carbide precipitation and cold work on grain
boundary diusion. At rst, for the solution heat treatment
the plates of SUS316 and 316L stainless steels were annealed
at 1423 K for 600 s and then quenched into water. Then, the
sensitized SUS316 specimen was made by heating at 923 K
for 172.8 ks in air. The cold worked SUS316 specimen was
made by cold-rolling the plate 5 mm down to 4 mm in
Table 1 Chemical compositions of SUS316 and 316L (N in mass ppm,
others in mass%).
Specimen
SUS316

Cr

Ni

Mo Mn

Si

Al

N
(ppm)

17.05 12.01 2.09 1.46 0.049 0.45 0.027 0.005 0.002

103

SUS316L 16.95 11.92 2.09 1.51 0.017 0.44 0.026 0.003 0.006

114

2946

M. Mizouchi, Y. Yamazaki, Y. Iijima and K. Arioka

thickness reduction of 20%. After these treatments, four

kinds of the plates were machined into the discs 12 mm in
diameter and 1 mm in thickness. The surfaces of discs were
electropolished at 273 K in an aqueous solution containing
4 vol% acetic acid and 1 vol% perchloric acid. Some specimens of the solution-treated SUS316 were used for measurements of volume diusion of chromium, because the volume
diusion coecient is required to deduce the grain boundary
diusivity in the analysis of type B kinetics. To cause grain
growth, they were annealed at 1623 K for 86.4 ks in a stream
of hydrogen gas puried by permeation through a palladium
alloy tube. The resultant grain size was 36 mm. On the other
hand, the grain sizes of the specimens of the solution-treated
SUS316 and 316L and the cold-worked SUS316 were 40
80 mm, while that of the sensitized SUS316 was 60100 mm.
All the specimens were ground on abrasive papers and
polished with ne alumina paste. The specimens for the
experiment of volume diusion were further annealed at
1273 K for 7.2 ks in an evacuated quartz tube to obtain a
stress-free surface. For the annealing, the specimen was
wrapped in a tantalum foil to prevent volatilization of
chromium from the surface of the specimen. Tantalum foil is
eective in keeping away the specimen from the quartz tube
to avoid the reaction between them. The distribution of grain
boundary character was measured using an orientation
imaging microscope. In particular, by compression of 20%,
the ratio of the length of random grain boundary to the total
length of grain boundary increases from 43 to 66%.
The radioisotope 51 Cr (
-ray, 0.320 MeV; half-life, 27.7 d)
was procured from ICN Biomedicals, Inc., U.S.A. in the form
of CrCl3 in 0.5 kmolm
3 HCl solution. It was electroplated
onto the polished surface of the specimen with the current
density 14 mA cm
2 for 30180 s. A quartz tube containing
the specimen wrapped with a tantalum foil was evacuated
under 1  10
4 Pa, and then it was put into an electric
furnace controlled within 1 K during diusion annealing at
a temperature in the range from 518 to 1173 K for 2.4 to
331.2 ks.
After diusion anneal, the cylindrical surface of the
specimen was reduced by a depth of 0.30.4 mm to remove
the eect of surface diusion. To measure the penetration
proles of the radioisotope into the specimen, two types of
the serial sectioning methods were employed; the ion-beam
sputter-microsectioning method for short penetration proles
of volume diusion and the serial sectioning method by
grinding on abrasive papers for long penetration proles of
grain boundary diusion. The intensity of
-ray emitted from
each section was measured with the help of a well-type Tlactivated NaI detector attached to a 1024 channel pulse
height analyzer.
3.

Results and Discussion

3.1 Volume diusion of 51 Cr in SUS316

The boundary conditions used in the present experiments
are an instantaneous source and a semi-innite medium. The
appropriate solution of Ficks law is given by
 p
CX; t M= Dv t exp
X 2 =4Dv t;
1

X 2 / 10-14 m2 for A and B

0

10

20

30

40

50

60

SUS316

A (1173K, 7.2ks)

B (1121K, 3.6ks)

C (888K, 253.8ks)

X 2 / 10-14 m2 for C
Fig. 1 Typical penetration proles of ln CX; t versus X 2 for diusion of
51
Cr in SUS316.

where CX; t is the concentration of the radioactive tracer at

a distance X from the original surface, Dv the volume or
lattice diusion coecient, t the time of diusion anneal and
M the total amount of deposited tracer at X 0 and t 0.
This equation represents the material transport within the
lattice through volume diusion. Figure 1 shows the typical
penetration proles of ln CX; t versus X 2 for the diusion of
51
Cr in SUS316. The linearity observed in Fig. 1 proves that
eq. (1) will hold, and thus the volume diusion has been
observed. Figure 2 shows the Arrhenius plot of the volume
diusion coecient of 51 Cr in SUS316 by the present work
in comparison with those by Perkins et al.5) and Smith.6)
From the linear Arrhenius plot observed in the temperature
range from 888 to 1173 K the preexponential factor and the
activation energy are obtained to be 1:13  10
7 m2 s
1 and
234 kJmol
1 , respectively. The compositions of the alloys
used by Perkins et al.5) and Simth6) are similar to that used by
the present work. Although the diusion coecients obtained by Perkins et al. at lower temperatures are nearly equal
to those by the present work, at higher temperatures the
values of the diusion coecients by Perkins et al.5) and
Simth6) are larger than those by the present work. However,
strict comparison between them is unfruitful, because
stainless steels contain usually many elements besides
residual impurities, as shown in Table 1. The Arrhenius
lines for grain boundary diusivities in stainless steels
obtained by many authors form a wide band.4)
Grain boundary diusion of 51 Cr in SUS316 and
316L
To determine the diusion coecient within the lattice as
well as the grain boundary simultaneously, we need an
analytical procedure to separate the contributions of the

3.2

Low Temperature Grain Boundary Diusion of Chromium in SUS316 and 316L Stainless Steels

2947

T/K
-12 1700

1300

10

-13
-14

800

solution-treated
SUS316

present work
1 Perkins et al. (1973)
2 Smith (1975)

10

10

1000

-15

10

-16

10

1173K, 3.6ks

-17

10

1123K, 10.8ks

-18

10

-19

10

1073K, 79.2ks

-20

10

-21

10

993K, 141ks

-22

10

0.6

0.8

1.0

1.2

10

T-1 / 10-3K-1

30

40

50

X 6/5 / 10-6 m6/5

Fig. 2 Arrhenius plot of volume diusion coecient of 51 Cr in SUS316 by

present authors in comparison with those by Perkins et al. and Smith.

Fig. 3 Penetration proles of ln C
versus X 6=5 for diusion of

solution-treated SUS316.

lattice and the grain boundary to the overall depth prole of

the diusing species. When the value of mean diusion
distance Dv t1=2 is smaller than d=20 (d diameter of the
grains), the grain boundary diusion behavior can be
described by the type B kinetics referred to Harrison.9) For
the case of diusion from an instantaneous source in the type
B kinetics, Suzuokas solution is given by10,11)
sDgb 1:308Dv t1=2
@ ln C
=@X 6=5
5=3 ;

20

51

Cr in

51

Cr in

solution-treated
SUS316L

when the parameter  sDgb =Dv
1=2Dv t1=2 is larger

than 104 as is the case of the present experiments. s is the
segregation factor,  is the width of grain boundary and C
is
the mean concentration of the tracer at a distance X from the
original surface after a diusion time t. Dgb is the grain
boundary diusion coecients of the tracer. The product
sDgb represents the characteristic of grain boundary diusion and is usually called the grain boundary diusivity.
On the other hand, when lattice diusion is negligible and
a signicant transport of matter occurs only within the grain
boundary, the average concentration of indiused material,
as determined by a sectioning experiment, will vary with the
penetration distance in the same way as lattice diusion. This
situation refers to the Type C kinetics by Harrison.9) Under
the condition that Dv t1=2 is smaller than =20, the tail region
in a logarithmic plot of C
will vary linearly with the
penetration distance squared and the slope will yield Dgb
itself.
Figures 3 and 4 show the penetration proles of ln C

versus X 6=5 for the diusion of 51 Cr in the solution-treated
SUS316 and 316L, respectively. Each penetration prole
exhibits two regions: one is the near-surface region and the
other the deeper region of the prole. The former mainly
corresponds to the volume diusion process, while the latter
is due to a high-diusivity path: the diusion through this
path is faster than that in the grain. The slopes of the highdiusivity contribution in Figs. 3 and 4 are used for the

1073K, 79.2ks

993K, 141ks

10

20
30
40
X 6/5 / 10-6 m6/5

50

Fig. 4 Penetration proles of ln C
versus X 6=5 for diusion of

solution-treated SUS316L.

calculation of the grain boundary diusivity sDgb using

eq. (2). The calculated values of sDgb are listed in Table 2 as
well as the grain boundary diusion parameters, where the
conditions of type B kinetics are satised. The Arrhenius
plots for the grain boundary diusivity of 51 Cr in the SUS316
and 316L are shown as solid circles and squares, respectively,
in Fig. 5 in comparison with those by the previous
authors.57) The grain boundary diusivity in SUS316L is
higher than that in SUS316. Furthermore, it should be noted
that the diusivities in SUS316 and 316L are about two-three
orders of magnitude higher than those observed in austenitic

2948

M. Mizouchi, Y. Yamazaki, Y. Iijima and K. Arioka

Table 2 Grain boundary diusivities and diusion parameters for

solution-treated SUS316 and 316L.

51

Cr in
solution-treated SUS316

Specimen

T/K

t/s

sDgb /m3 s
1

SUS316

1173

3600

2:08  10
18

2:12  106

SUS316

1123

10800

4:31  10
19

4:65  105

79200

19

1:77  106

20

9:90  106

19

5:56  10

3:45  106

1:23  10
19

1:78  107

SUS316

1073

SUS316

993

2:85  10

141000

SUS316L

1073

79200

SUS316L

993

141000

6:84  10

solution-treated SUS316L

1173K, 3.6ks
1123K, 10.8ks
1073K, 79.2ks
993K, 141ks

T/K
-15

10

1300

1000

800

1073K, 79.2ks

present work, SUS316

present work, SUS316L
1 Perkins et al. (1973)
2 Smith (1975)

(1990)
3 Cermk

-16

10

-17

10

993K, 141ks

-18

10

-19

10
15
X 6/5 / 10-6 m6/5

20

10

Fig. 6 Penetration proles of ln C
versus X 6=5 , where the short depth region
in Figs. 3 and 4 is enlarged.

-20

10

-21

10

-22

10

-23

10

-24

10

0.7

0.8

0.9
T-1 /

1.0

1.1

1.2

1.3

10-3K-1

Fig. 5 Arrhenius plots for grain boundary diusivities of 51 Cr in SUS316

and 316L.

stainless steels by the previous authors. It is well known that

the magnitude of grain boundary diusivity depends greatly
on the grain boundary character,12) even if the diusivity is
calculated from a linear tail in the plot of ln C
versus X 6=5
under the conditions of type B kinetics. As mentioned earlier,
the mean diusion distance Dv t1=2 is an important parameter in the analysis of penetration proles. In the paper of
Perkins et al.,5) the penetration depth was not shown. On the
other hand, Smith6) and Cermak7) have determined sDgb
from the tail slope where the ratio of the penetration depth to
Dv t1=2 is in the range from 10 to 606) and from 80 to 200,7)
respectively, whereas in the present experiments the ratio is
in the range from 1200 to 2700. Figure 6 shows the same
penetration proles as Figs. 3 and 4. However, the short
depth regions are enlarged, where the linear slopes are also
observed. Then, the value of sDgb is recalculated and plotted
in Fig. 5 as open circles and squares. The recalculated values
of sDgb are in a similar range that observed by the previous
authors.57) This means that the present experiment has
observed the grain boundary diusion via faster paths than
those observed by Smith and Cermak.
Figures 7, 8, 9 and 10 show the penetration proles of ln C

versus X 2 for the diusion of 51 Cr in the solution-treated
SUS316 and 316L, the sensitized SUS316 and the coldworked SUS316, respectively. In all proles, two regions are

observed as in Figs. 3 and 4. From the linear tail in these

proles, the grain boundary diusion coecients Dgb of 51 Cr
in these alloys are calculated and listed in Table 3. The
condition of Type C kinetics: Dv t1=2 is smaller than =20, is
fully satised if 5  10
10 m is put into , as in the usual case
of metals.12) The Arrhenius plots for the grain boundary
diusion coecients Dgb of 51 Cr in the solution-treated
SUS316 and 316L, the sensitized SUS316 and the coldworked SUS316 are shown in Fig. 11 in comparison with
those converted from the grain boundary diusivity sDgb by
putting s 5  10
10 m. The grain boundary diusion
coecients obtained by the present work are remarkably
higher than those of the previous authors. The linear
Arrhenius plots of the grain boundary diusion coecients
obtained by the type C kinetics can be expressed as follows:
Solution-treated SUS316
Dgb 1:021:08

0:52
 10
5 exp
87:5  3:6 kJmol
1 =RT m2 s
1

Solution-treated SUS316L
Dgb 2:842:50

1:33
 10
5 exp
90:6  3:0 kJmol
1 =RT m2 s
1

Sensitized SUS316
Dgb 4:8719:29

3:89
 10
6 exp
85:1  7:8 kJmol
1 =RT m2 s
1

Cold-worked SUS316
Dgb 3:062:07

1:24
 10
5 exp
90:5  2:5 kJmol
1 =RT m2 s
1

As seen in Fig. 11, the values of Dgb deduced from the

values of sDgb are on the lines extrapolated from the

Low Temperature Grain Boundary Diusion of Chromium in SUS316 and 316L Stainless Steels

X 2 / 10-8 m2 for A

X 2 / 10-8 m2 for D
5

2949

sensitized SUS316

solution-treated
SUS316

A (698K, 2.4ks)

A (652K, 7.2ks)

B (668K, 3.9ks)

B (598K, 84.6ks)

C (652K, 7.2ks)

C (548K, 316.8ks)

D (548K, 316.8ks)

1
X2 /

10-8 m2

for A, B and C

Fig. 7 Penetration proles of ln C versus X 2 for diusion of

solution-treated SUS316.

51

Cr in

5
X 2 / 10-8 m2 for B and C

Fig. 9 Penetration proles of ln C versus X 2 for diusion of

sensitized SUS316.

51

Cr in

X 2 / 10-8 m2 for B
0.0

solution-treated
SUS316L

0.5
cold-worked
SUS316

568K, 148.32ks

A (652K, 7.2ks)

548K, 316.8ks
B (568K, 148.32ks)

533K, 331.2ks

C (548K, 316.8ks)

D (533K, 331.2ks)

X 2 / 10-8 m2
Fig. 8 Penetration proles of ln C versus X 2 for diusion of
solution-treated SUS316L.

0
51

Cr in

5
X 2 / 10-8 m2 for A, C and D

Fig. 10 Penetration proles of ln C versus X 2 for diusion of 51 Cr in coldworked SUS316.

Arrhenius relations of eqs. (3) and (4), suggesting that the

assumption of s 5  10
10 m for these alloys is reasonable. The magnitudes of Dgb among the solution-treated
SUS316 and 316L and the sensitized SUS316 can be easily
understood, because the value of Dgb decreases with
increasing the carbon content8) and carbide13) in the grain
boundary. The value of Dgb in the cold-worked SUS316 is
largest among the four kinds of specimens as shown in

Fig. 11. According to the measurements of grain boundary

character, the ratio of the length of random grain boundary to
the total length of grain boundary increases from 43 to 66%
by thickness reduction of 20%. Substantially, the cold-work
increases faster diusion paths, because the diusion through
random grain boundaries is faster than the diusion through

2950

M. Mizouchi, Y. Yamazaki, Y. Iijima and K. Arioka

Table 3 Grain boundary diusion coecients for 51 Cr in solution-treated

SUS316 and 316L, sensitized SUS316 and cold-worked SUS316.
Specimen

T/K

t/s

10

12

10

698

2400

2:71  10

solution-treated SUS316

668

3900

1:62  10
12

solution-treated SUS316

652

7200

1:11  10
12

solution-treated SUS316

617

86400

2:81  10
13

solution-treated SUS316

598

84600

2:71  10
13

solution-treated SUS316

568

148320

1:08  10
13

solution-treated SUS316

548

316800

3:90  10
14

10

solution-treated SUS316
solution-treated SUS316L

518
698

321660
2400

1:58  10
14
4:01  10
12

10

solution-treated SUS316L

652

7200

1:76  10
12

solution-treated SUS316L

598

84600

4:01  10
13

solution-treated SUS316L

568

148320

1:18  10
13

solution-treated SUS316L

548

316800

6:27  10
14

14

solution-treated SUS316L

533

331200

4:46  10

solution-treated SUS316L

518

321660

1:75  10
14

sensitized SUS316
sensitized SUS316

698
652

2400
7200

2:15  10
12
4:42  10
13

sensitized SUS316

617

86400

1:78  10
13

13

sensitized SUS316

598

84600

sensitized SUS316

568

148320

1:64  10

8:68  10
14

sensitized SUS316

548

316800

3:59  10
14

sensitized SUS316

518

321660

9:29  10
15

cold-worked SUS316

698

2400

5:26  10
12

12

cold-worked SUS316
cold-worked SUS316

668
652

3900
7200

2:42  10
1:96  10
12

cold-worked SUS316

598

84600

2:79  10
13

cold-worked SUS316

568

148320

1:61  10
13

14

cold-worked SUS316

548

316800

7:50  10

cold-worked SUS316

533

331200

4:37  10
14

cold-worked SUS316

518

321660

2:14  10
14

sub-grains and small angle grain boundaries. It should be

emphasized that the grain boundary diusion of chromium in
SUS316 is much faster than that recognized by the previous
works and the activation energy for it is much smaller than
that known so far.
Conclusion

Grain boundary diusion coecients of 51 Cr in SUS316

and 316L have been measured in the temperature range
between 518 and 698 K using the type C kinetics and between
993 and 1173 K using the type B kinetics. The present results
show remarkably high diusivity and low activation energy
in comparison with those obtained by previous authors,

10
10

10

T/K

1000

500

present work
cold-worked SUS316
solution-treated SUS316L
solution-treated SUS316
sensitized SUS316

Dgb /m2 s
1

solution-treated SUS316

4.

-7 1500

-8

-9

-10

-11

-12

-13

3 2 1

10
10

-14

1 Perkins et al. (1973)

2 Smith (1975)

(1990)
3 Cermk

-15

1.0

1.5

2.0

T-1 / 10-3K-1
Fig. 11 Arrhenius plot for grain boundary diusion coecients of 51 Cr in
solution-treated SUS316 and 316L, sensitized SUS316 and cold-worked
SUS316.

suggesting that the previous authors have observed grain

boundary diusion through relatively slow paths such as subgrains and small angle grain boundaries.
REFERENCES
1) S. H. Bush and R. L. Dillon: Stress Corrosion Cracking and Hydrogen
Embrittlement of Iron Base Alloys, (National Association of Corrosion
Engineers, Texas, 1977) p. 61.
2) K. Arioka, Y. Kaneshima and T. Yamada: J. Inst. Nuclear Safety
System 10 (2003) 125135.
3) K. Arioka, Y. Kaneshima, T. Yamada and T. Terachi: Proceedings of
the 11th International Symposium on Environmental Degradation of
Materials in Nuclear Power Systems, (Stevenson, USA, 2003).
4) I. Kaur, W. Gust and L. Kozma: Handbook of Grain and Interphase
Boundary Diusion Data, (Ziegler Press, Stuttgart, 1989) p. 658.
5) R. A. Perkins, R. A. Padgett, Jr. and N. K. Tunali: Metall. Trans. 4
(1973) 25352540.
6) A. F. Smith: Metall. Sci. 9 (1975) 375378.
7) J. Cermak: Z. Metallk. 81 (1990) 193195.
ermak, J. Ruzickova, A. Pokorna and B. Million: Mater. Sci. Eng.
8) J. C
A158 (1992) 7985.
9) L. G. Harrison: Trans. Faraday Soc. 57 (1961) 11911199.
10) T. Suzuoka: Trans. JIM 2 (1961) 2532.
11) T. Suzuoka: J. Phys. Soc. Japan 19 (1964) 839851.
12) A. P. Sutton and R. W. Ballu: Interfaces in Crystalline Materials,
(Clarendon Press, Oxford, 1995) p. 467.
13) T.-F. Chen, G. P. Tiwari, Y. Iijima and K. Yamauchi: Mater. Trans. 44
(2003) 4046.