ISIJ International, Vol.

56 (2016),
ISIJ International,
No. 11 Vol. 56 (2016), No. 11, pp. 1989–1995

Three-dimensional Observation of Nonmetallic Inclusion Clusters

in Solid Metal by X-ray Micro-CT

Tao LI,1,2)* Shin–ichi SHIMASAKI,1,3) Shoji TANIGUCHI,1) Shunsuke NARITA1,4) and Kentaro UESUGI5)

1) Graduate School of Environmental Studies, Tohoku University, 6-6-20 Aoba-yama, Aoba-ku, Sendai, 980-8579 Japan.
2) Norwegian University of Science and Technology, Kolbjørn Hejes v 1B, 7491, Trondheim, Norway. 3) Department of
Electro-Mechanical Systems Engineering, National Institute of Technology, Kagawa College, 335, Chokushi-cho, Takamatsu,
761-8058 Japan. 4) Daido Steel Co. Ltd., 1-10, 1-chome Higashisakura, Higashi-ku, Nagoya, Aichi, 461-8581 Japan.
5) Japan Synchrotron Radiation Research Institute (JASRI)/ Spring-8, 1-1-1 Kouto, Sayo-cho, Sayo-gun, Hyogo, 679-5198 Japan.
(Received on April 12, 2016; accepted on July 4, 2016)

In this study, we developed a method of observing three-dimensional (3D) inclusion clusters in metal.
The theory of X-ray micro-computed tomography (CT) is generally introduced. The criteria on determining
the sample size and energy of X-ray beam is set according to the characteristics of the beamline BL20XU
available at Spring-8, which is the largest synchrotron radiation facility. The Al–TiB2 and Al–SiC system were
measured by the X-ray micro-CT available in beamline BL20XU at Spring-8. The determination of the
sample size the energy of X-ray beam are discussed on basis of the X-ray linear attenuation coefficient
(LAC) and the transmission rate of Al–TiB2 and Al–SiC system. The limitation of the charge coupled device
(CCD) camera, such as spatial resolution and observing field, are taken into consideration as well. Phase
retrieval method is applied to reduce the noises of tomographic images for Al–SiC system due to its poor
contrast. The 3D particle clusters of TiB2 and SiC are reconstructed by self-developed program. The fractal
dimension of TiB2 and SiC clusters are calculated around 2.7. The feasibility of observing alumina cluster
in steel is discussed on basis of the latest capability of beamline BL20XU at Spring-8.

KEY WORDS: inclusion; steel; alumina; cluster; three-dimensional; X-ray; micro-CT.

sections of a sample from metal products. An inclusion

1. Introduction
cluster appears as a few of isolated particles on a 2D cross-
The inclusions in metal products largely decrease their sections of the samples. Tozawa et al.8) firstly investigated
mechanical properties. The fine inclusion particles coagulate the fractal dimension of alumina cluster in steel sample and
each other in molten metal and grow up to clusters, which its influence on floating velocity of clusters in liquid. The
are much more harmful. It is an everlasting topic to remove fractal dimension of an alumina cluster was calculated by
inclusions as much as possible and control morphology of 2D measurement of the cross-sections of a steel sample,
the remaining inclusions in metal products. which yielded a fractal dimension of 1.8. The model of
The three-dimensional (3D) structure of inclusion clus- cluster flotation velocity in liquid was modified using fractal
ters plays a key role in the inclusion behaviors in molten dimension; it agreed well with the cold model experiments.
metal. The 3D analysis of the inclusion cluster structure Doo et al.9) investigated the fractal dimension of the alumina
will provide significant information for understanding the cluster extracted from steel according to the projected area
mechanisms of inclusion behaviors in liquid metal, such as of the cluster. The 2D fractal dimension was reported in the
inclusion coagulation, flotation and bubble adhesion. Many range of 1.80–1.95 with average of 1.88.
modelling and experimental works1–5) have been performed However, those works above are all on basis on the obser-
to investigate the inclusion behaviors by using spherical vation on 2D cross section or projection area. It is very rare
particles. Nevertheless, the morphology of the inclusions is to find the experimental reports about 3D cluster structure
far from spherical when they coagulate each other and grow analysis, especially the nonmetallic particle clusters in metal
up into clusters. system because of its opaque and high melting point. X-ray
Fractal dimension6,7) is one of the most important param- Micro-CT is recently applied to observe the 3D pores and
eters describing the morphology of particle clusters, which cracks in metal and polymer,10,11) which is well known as
specifies the relationship between the volume (area) and the nondestructive inspection method.12–15) To achieve the 3D
size in length of a cluster. It is currently most common to observation, the contrast of between the materials in the
observe of the inclusions on two-dimensional (2D) cross- samples should be detectable, which is determined by the
sample size, X-ray power and the types of materials. Mostly,
* Corresponding author: E-mail: [email protected] the pores and crack in the solid have quite enough contrast
DOI: http://dx.doi.org/10.2355/isijinternational.ISIJINT-2016-218 of X-ray absorption rate to the solid material, particularly

1989 © 2016 ISIJ

 ISIJ International, Vol. 56 (2016), No. 11

in polymers. It would be difficult to obtain an enough X-ray of BL20XU is 0.95 mm in width and 0.7 mm in height.
transmission rate for observation of metal samples. Ham- Therefore, the maximum diameter of the sample should be
ilton et al.16) made an effort to observe the TiB2 particles less than 0.95 mm in diameter. The sample size is set in the
by X-ray micro-CT. However, the resolution is only 5 μm, range of 0.5 mm–0.8 mm in this study; and the details of
which is not sufficient to derive quantitative description of determining the sample size is discussed below.
the 3D morphology of the TiB2 particles.
A beamline of X-ray micro-CT is available in Spring-8,
3. X-ray Tomography
which is a large synchrotron radiation facility available in
Japan. It yield a resolution of 0.5 μm, which is ten time 3.1. BL20XU at Spring-8
higher than Hamilton’s work.16) The current authors’ group Spring-8 is a large synchrotron radiation facility that
has applied the X-ray micro-CT to study the 3D structure delivers the most powerful synchrotron radiation available
of TiB2 and SiC particle clusters in aluminium.17,18) The in Japan. The beamline BL20XU is designed to the imaging
coagulation of SiC particles in molten aluminum was stud- technologies, which is the second medium-length beamline
ied based on the 3D observation of the particle clusters,5) in the Spring-8. Monochromator in beamline BL20XU is
which attracted quite a lot interests from researchers. In this “Spring-8 standard” double crystal monochromator. The
study, we disclose the details of the development of the 3D first crystal of monochromator is combination of Si (111)
observation method for Al–TiB2 and Al–SiC systems. The orientation crystal and Si (511) orientation. These two
sample size and energy of X-ray source are determined by crystals are interchangeable by using linear-translation stage
calculating the X-ray attenuation efficiencies and the trans- without venting the monochromator vacuum. The second
mission rate of the samples. This study opens an eye to crystal is Si 111–orientation. The energy region of X-ray
observe and measure the morphology of inclusion clusters in beam was 8–37 keV before and recently extended into
metal product. The feasibility of observing alumina clusters 7.62–113 keV.
in steel is discussed as well. Figure 2 shows the principle of X-ray micro-CT. The
parallel X-ray beam from source penetrates the sample and
is partly absorbed by the sample; the charge–coupled device
2. Experimental
(CCD) camera detects the received energy of X-ray and
The samples of aluminium with particle clusters were generates projection images. The raw data for the sample
prepared in the mechanical agitating crucible,5,17,18) which are composted by multiple projections of the object being
contains molten Al and particles (TiB2 or SiC). The molten scanned. Tomographic reconstruction, essentially involves
aluminium with particles was stirred in a crucible at 973 K solving the inverse Radon transformation,19) produces a
to prepare the particle clusters in aluminium. The size of the series of cross-sectionsal images.
TiB2 particle varies from 2–30 μm with peak diameter at 20
μm. The volume fraction of the particle is around 0.5 vol%. 3.2. X-ray Attenuation
While the size of SiC particles is in the range of 2–30 μm The contrast of the materials in the micro-CT images is
with peak diameter at 12 μm. The volume fraction of SiC determined by the X-ray linear attenuation rate of the differ-
is up to 2 vol%.The samples were sucked from the molten ent materials. The exponential attenuation law of a narrow
aluminium by the silica tube with diameter of 3 mm, and beam of monoenergetic photons (X–ray) with an incident
then processed by electron discharge machining to fit the intensity I0, penetrating a layer of sample is given by
requirement of the X-ray micro-CT.
I
Figure 1 shows the photo of the micro-CT in beamline = exp [ − µm x ] ............................. (1)
BL20XU at Spring-8 and the processed sample. It is difficult I0
to obtain a sample with diameter less than 0.5 mm by the where, I is the intensity of X-ray penetrating the sample, x is
electron discharge machining. On the other hand, the size the mass thickness of the sample and μm is the X-ray mass
of the charge coupled device (CCD) camera in the beamline attenuation coefficient (MAC).
The X-ray MAC μm is expressed by the X-ray linear
attenuation coefficient (LAC) μl,
µm = µl ρ ................................. (2)
Where, ρ is the density of material in the sample.
The mass thickness x is defined as the mass per unit area,

Fig. 1. Micro-CT in beamline BL20XU at Spring-8. Fig. 2. Principle of X-ray micro-CT.

© 2016 ISIJ 1990

 ISIJ International, Vol. 56 (2016), No. 11

which is obtained by product of the material density and the of the CCD camera is 0.5 μm. Totally, 1 439 slices were
thickness L in length, obtained for each sample with the observed length of around
0.715 mm, which is limited by the CCD camera. ImageJ
x = ρ L .................................... (3)
was applied to process the 16 bit Micro-CT images, which
By combining Eqs. (1)–(3), the X-ray transmission rate is a public domain image processing software built up by
is given by the National Institutes of Health. A self-developed program
developed on ImageJ to extract 3D clusters from the micro-
I
= exp ( − µl L ) ............................. (4) CT image stacks, which was explained in detail in our
I0 previous work.17)
The attenuation law mentioned above is based on the
system including one type of materials. The X-ray contrast
4. Results and Discussions
of two different materials is estimated by the relative differ-
ence in linear attenuation. 4.1. Measurement of Aluminium Containing Particles
The X-ray MAC is a function of the atomic number of
µl ,1 − µl ,2 the elements included in the materials. The atomic number
Difference = ....................... (5)
µl ,1 of main elements in metal matrix and particles are listed in
Table 1. The atomic numbers of C and B are quite closed.
where, component–1 is dominated in the mixture. It is pos- However, the Si in SiC is adjacent to Al; while the Ti in
sible to distinguish the difference of materials when the TiB2 has large difference from Al in the atomic number that
relative difference in LAC exceeds 20%. would lead to a much larger contrast between aluminium
The volume fraction of each material should be taken into and TiB2.
account in a multiple materials system. The X-ray transmis- Figure 3 shows the X-ray MAC of Al matrix and parti-
sion rate of a multiple material is expressed by cles (TiB2 and SiC).20) The gap between Al matrix and TiB2
particle in X-ray MAC is quite large; while the difference of
I  n  that between Al matrix and SiC is very small, particularly
= exp  −∑ ϕi µl ,i L  ........................ (6)
I0  i  with increasing energy of X-ray beam.
The X-ray transmission rate is a function of the LAC and
where, φi is the volume fraction of the component–i. the size of the sample. Figure 4 shows the X-ray LAC of Al
To determine the size of samples and the energy of X-ray matrix and TiB2 particles and their relative difference. The
beam, the following criteria should be considered. X-ray LAC show the same tendency to the X-ray MAC. The
(1) Due to the small size of TiB2 and SiC particles, the difference of the LAC between Al and TiB2 is far greater
spatial resolution of the CCD camera should be less than than the criteria of 20% in the entire X-ray energy range of
1 μm. The maximum observing field then is 0.95 mm in from 8 keV up to 37 keV, though it decreases at 20 keV
width and 0.7 mm in height, which is limited by the CCD of X-ray energy beam. Due to the large difference in LAC
camera. Thus, the maximum diameter of the sample is set
as 0.8 mm for safety.
Table 1. Atomic number of main elements in metal matrix and
(2) The X-ray energy available at BL20XU is between
particles.
8–37 keV when the sample of Al–TiB2 and Al–SiC were
measured. Material
Density
Element
Atomic Atomic
(3) The acceptable X-ray transmission rate is in the range (g/cm3) Number Weight
is 0.2–0.8, with optimist value around 0.5. Metal matrix 2.7 Al 13 27
(4) For the sake of involving more particle clusters in the Si 14 28
sample, the sample volume should be as much as possible, SiC particle 3.21
C 6 12
on the premise of enough X-ray transmission rate and con-
trast in the X-ray micro-CT images. Ti 22 48
TiB2 particle 4.52
B 5 11
3.3. Micro-CT Image Reconstruction
The sample was settled between the X-ray source beam
and the CCD camera. Numbers of projection images were
taken by the CCD camera when the sample is rotating with
a constant angular speed. In this study, totally 1 891 pro-
jection images were taken for each sample of Al–TiB2, in
which the first 1–60 images are for background, 61–1 860
for the rotating samples (10 shots/degree), and 1 861–1 891
for background again. For the Al–SiC system, the projec-
tion number was set as 3 691 to deduce the noise, in which
1–60 are for background, 61–3 660 are the projections for
samples, and 3 661–3 691 are for background again to con-
firm (180 degree, 20 shots/degree).
The projection images were then transferred into cross-
sectionsal images of the sample. The spatial resolution Fig. 3. X-ray MAC of Al matrix and particles (TiB2 and SiC).

1991 © 2016 ISIJ

 ISIJ International, Vol. 56 (2016), No. 11

Fig. 4. X-ray LAC of Al matrix and TiB2 particles and their dif- Fig. 6. X-ray LAC of Al matrix and SiC particles and their differ-
ference. ence.

Fig. 5. X-ray transmission rate of Al–TiB2 (0.5 vol% TiB2). Fig. 7. X-ray transmission rate of Al–SiC (2 vol% SiC).

between Al and TiB2, the problem of contrast could be emit- contrast between Al and SiC increases with X-ray energy.
ted from the consideration. The transmission rate is then the On the other hand, it is required to adjust the X-ray trans-
key factor on determining the energy of X-ray beam and the mission rate to be around 0.5 in order to achieve the best
sample size of Al–TiB2. contrast of micro-CT image. When the sample size is set at
Figure 5 shows the X-ray transmission rate for Al–TiB2, 0.8 mm in diameter, the energy of X-ray beam should be
which is calculated by Eq. (6) with varying sample sizes. around 21 keV according to Fig. 7. However, the thicker
Since the volume fraction of TiB2 particles in the sample the sample, the more noise comes from photon scattering,
is very small, it has little effects on the X-ray transmission which would practically decrease the contrast. In this way,
rate. It is acceptable to observe the samples with an X-ray a thinner sample should be used for Al–SiC system. The
transmission rate of 0.2–0.8, with optimized value at 0.5 sample size of Al–SiC is set at 0.5 mm in diameter and the
for the observation. In order to increase the particle clusters energy of the X-ray source is set as 18 keV to obtain an
containing in the observed sample, the sample size is deter- X-ray transmission rate of 0.50.
mined at the maximum (0.8 mm), which is limited by the
CCD camera. The energy of X-ray beam is around 20 keV 4.2. 3D Reconstruction of Particle Clusters in Alumin-
to achieve the transmission rate of 0.5; and was finally fixed ium
at 19.99 keV after testing the quality of the X-ray images. An example of 8-bit micro-CT cross-sectional image of
Figure 6 shows the change of LAC of Al and SiC with Al–TiB2 sample is show in Fig. 8(a). The white area is TiB2
X-ray energy. It is obvious that the difference in LAC particles; the grey area is Al matrix; and the black area is
between Al and SiC is much smaller than that between Al the background of air. It turns out that the contrast between
and TiB2. The SiC particles may be difficult to be distin- Al and TiB2 is quite large. Figure 8(b) shows the binary
guished by the CCD camera in the X-ray micro-CT images. image preprocessed image by setting a threshold of grey
Figure 7 shows the X-ray transmission rate of Al–SiC value in ImageJ.
containing 2 vol% particles. The X-ray transmission rate The particle clusters in the samples are reconstructed
of the sample with certain size increases with the increas- from the series of cross-sectional images by a self-devel-
ing energy of X-ray beam; whereas it decreases with the oped program based on ImageJ. Figure 9 shows an example
increasing size of samples exposed under the X-ray with of TiB2 clusters and its cross-sectional slices. It is obvi-
the same energy. It turns out that, on basis of Fig. 6, the ously difficult to identify the cluster when one of the slices

© 2016 ISIJ 1992

 ISIJ International, Vol. 56 (2016), No. 11

(a). 8-bit image from X-ray Micro-CT. (b). Processed image by ImageJ.
Fig. 8. Cross-sectional image from X-ray Micro-CT and processed image.

Fig. 9. Example of TiB2 cluster and its cross sectional slices.

Fig. 10. X-ray micro-CT image obtained by ordinary reconstruc-
tion method.
appears independently. The fractal dimension of the cluster
is 2.74, which is calculated by
π Df
V= dave .................................. (7)
6
where, the V is the volume of the cluster; dave is the aver-
age diameter of the cluster in x, y and z direction; Df is the
fractal dimension.
When it comes to the Al–SiC system, the contrast of the
X-ray micro-CT images becomes poor. Figure 10(a) shows
an 8-bit micro-CT image of Al–SiC sample. A magnified
area is shown in Fig. 10(b) where the fine dots in the Al
matrix have the similar grey values as the SiC particles. It
is difficult to identify the SiC particle from the Al matrix e
by setting a threshold of grey value, as done for Al–TiB2. Fig. 11. Normalized X-ray micro-CT image obtained by phase
retrieval.
The quality of the tomographic images were largely
improved by applying the phase retrieval methods21–23)
implemented by Transport-of-Intensity equation22) using phase retrieval based on FFT algorithm. As seen from Fig.
fast Fourier transform (FFT) method.23) The phase-retrieval 11(b), the noise of dots shown in Fig. 10 has been largely
method was developed to obtain the projected phase and removed. The phase retrieval images are then and converted
absorption information of the object in multi-material into 8-bit images; and a threshold is set to convert the
objects with distinct complex refractive indexes. The algo- images into binary, which are shown in Fig. 12. Figure 13
rithm was implemented using phase contrast data acquired gives an example of SiC cluster whose fractal dimension is
at the Spring-8 Synchrotron facility in Japan. The 3D 2.7 calculated by Eq. (7). It shows a similar contracture as
complex refractive index distribution of a multi-material TiB2 clusters. On basis of statistical analysis of thousands
test object was quantitatively reconstructed using a single of TiB217) and SiC5) clusters, it turned out that their fractal
X-ray phase contrast image per projection. The technique is dimension is around 2.7. The fractal dimension of the 2D
robust in the presence of noise, compared to conventional cross-sectional particles is around 1.8, which agreed well
absorption based tomography. Figure 11(a) shows a nor- with Tozawa’s work for alumina cluster in steel.8)
malized X-ray micro-CT image of Al–SiC obtained by the

1993 © 2016 ISIJ

 ISIJ International, Vol. 56 (2016), No. 11

Fig. 12. X-ray micro-CT images processed by ImageJ.

Fig. 14. X-ray LAC of Fe and Al2O3 particles and their relative
difference.

Fig. 13. Example of SiC cluster and its cross sectional slices.

4.3. Alumina Cluster in Steel

The atomic number of Fe (26) is much larger than that Fig. 15. X-ray transmission rate of Fe–Al 2O3 (1 vol% Al2O3).
of Al; thus it would be much difficult for the X-ray to
penetrate the steel sample. It thus requires a much higher
energy of X-ray to observe the steel sample than that for
aluminium samples. Figure 14 shows the LAC of Fe and
Al2O3 and their relative difference. An energy absorption
edge25) appears at 9 keV for Fe, due to the photoelectric
effect, which is caused by the removing of electrons from
the shells around the nuclear in the atoms. It is common to
set the energy of X-ray beam around the absorption edge,
which results in the maximum contrast of the micro-CT
image. However, the X-ray transmission rate of the sample
is another critical factor of observation. Figure 15 shows the
X-ray transmission rate for Fe–Al2O3 samples containing 1
vol% Al2O3. The volume fraction is calculated by assuming
that all the inclusions in the steel sample (T[O] = 20 ppm)
appearing in term of Al2O3. The X-ray transmission rate
is extremely small with photon energy of 9 keV, though
the contrast between Fe and Al2O3 is large. Actually, the Fig. 16. Alumina extracted from steel sample.
minimum X-ray energy is around 30 keV even for observ-
ing a sample with diameter of 200 μm. Figure 16 shows
an alumina cluster extracted from a steel sample.9) The size observing field of the CCD camera. Therefore, the energy
of alumina clusters reaches several tens microns, even a of X-ray beam should be around 60 keV for to achieve an
hundred microns. Therefore, it is difficult to guarantee that X-ray transmission rate around 0.5. The measurable sample
alumina clusters are involved in a sample with diameter of size would theoretically reach to 3 mm if a CCD camera
0.2 mm. Fortunately, the energy range of X-ray up to 113 with enough spatial resolution is available.
keV is currently available at beamline BL20XU at Spring-8, Though the behavior of alumina inclusions has been
which makes it possible to observer Fe–Al2O3 samples with attracting much interest, the other types of inclusion have
larger size. On the other hand, the spatial resolution should bad influence to the steel products as well. The feasibility
be less than 1 μm due to the small size of alumina particles of 3D observation of other inclusion would be considered
contained in the cluster. We set the maximum diameter of in the future. The contrast is low when the elements in the
the sample at 0.8 mm as before due to the limitation of the inclusions have similar atomic number to Fe, such as FeO,

© 2016 ISIJ 1994

 ISIJ International, Vol. 56 (2016), No. 11

MnO, and MnS. The phase retrieval method would then be numbers of Spring-8 are 2011B1395 and 2012B1448 on
applied as done for Al–SiC system. beamline BL20XU. The authors wish to express thanks
to Dr. A. Takeuchi and Dr. Y. Suzuki for their help dur-
ing the operation of the X-ray micro-CT at Spring-8. We
5. Conclusions
also thanks Dr. M. Iguchi of Nihon Ceratec Co., Ltd. for
In this study, the method of observing 3D inclusion the cooperation on preparing the Al–TiB2 metal matrix
clusters in metal was disclosed in details. The Al–TiB2 composite.
and Al–SiC system were measured by the X-ray micro-CT
available at Spring-8. The 3D reconstruction of the X-ray REFERENCES
micro-CT images was presented, with the calculation of 1) H. Arai, K. Matsumoto, S. Shimasaki and S. Taniguchi: ISIJ Int., 49
the fractal dimension. The feasibility of observing alumina (2009), 965.
2) T. Nakaoka, K. Matsumoto, H. Arai, S. Shimasaki and S. Taniguchi:
cluster in steel is discussed on basis of the latest capability Prog. Comput. Fluid Dyn., 10 (2010), 62.
of beamline BL20Xu at Spring-8. 3) T. Li, R. Oka, S. Taniguchi and S. Shimasaki: CAMP-ISIJ, 24 (2011),
(1) The contrast of Al–TiB2 system is sufficiently large 811.
4) T. Li, S. Shimasaki, S. Taniguchi and K. Uesugi: 13th Int. Conf.
that it does not effect on determining the sample size. The Aluminum Alloys, TMS, Warrendale, PA, (2012), 1337.
sample size is then limited by the observing field of the 5) T. Li, S. Shimasaki, S. Taniguchi, S. Narita and K. Uesugi: ISIJ Int.,
53 (2013), 1958.
CCD camera. The energy of X-ray beam is set at 19.99 keV 6) K. Falconer: Fractal Geometry, Wiley, New York, (2003), 308.
to achieve an X-ray transmission rate of 0.55. 7) H. Sagan: Space–Filling Curves, Springer–Verlag, Berlin, (1994),
(2) The difference in LAC between Al and SiC is pretty 156.
8) H. Tozawa, Y. Kato, K. Sorimachi and T. Nakanishi: ISIJ Int., 39
close that the contrast on the micro-CT images is bad. In (1999), 426.
order to reduce the noise coming from photon scattering, the 9) W. C. Doo, D. Y. Kim, S. C. Kang and K. W. Yi: ISIJ Int., 47 (2007),
1070.
sample size is set to be the minimum (0.5 mm) processable 10) J. Y. Buffiere, E. Maire, Cloetens, G. Lormand and R. Fourgeres:
by electron discharge machining. And the energy of X-ray Acta Mater., 47 (1999), 1613.
11) A. Elmoutaouakkil, G. Fuchs, Bergounhon, R. Peres and F. Peyrin:
beam is set at 18 keV to achieve an X-ray transmission rate J. Phys. D: Appl. Phys., 36 (2003), A37.
of 0.50. 12) R. Jullien, M. Kolb and R. Rotet: J. Phys. Lett., 45 (1984), 211.
(3) The X-ray micro-CT images of Al–TiB2 is easier 13) T. A. Witten and L. M. Sander: Phys. Rev. Lett., 47 (1981), 1400.
14) P. Meakin: Phys. Rev. A, 27 (1983), 604.
to be processed due to its large contrast; while the phase 15) P. Meakin and H. E. Stanley: Phys. Rev. Lett., 51 (1983), 1457.
retrieval method is applied to the tomographic images of 16) R. W. Hamilton, M. F. Forste, R. J. Dashwood and D. Lee: Scr.
Mater., 46 (2002), 25.
Al–SiC to reduce the noises. The fractal dimension of both 17) T. Li, S. Shimasaki, S. Taniguchi, S. Narita and K. Uesugi: ISIJ Int.,
TiB2 and SiC clusters in aluminium is around 2.7. 53 (2013), 1943.
(4) The alumina clusters in steel is theoretically able 18) T. Li, S. Shimasaki, S. Taniguchi, K. Uesugi and S. Narita: Metall.
Mater. Trans. B, 44 (2013), 750.
to be observed currently due to the larger range of X-ray 19) S. R. Deans: The Radon Transform and Some of Its Applications,
energy available in beamline BL20XU at Spring-8. The John Wiley & Sons, New York, (1983).
20) NIST: XAAMDI, X-ray Attenuation and Absorption for Materials of
sample should be processed into cylinder with diameter Dosimetric Interest Database, NIST, Gaithersburg, MD, (2013).
around 0.8 mm. The energy of X-ray should be around 60 21) T. E. Gureyev and K. A. Nugent: Opt. Commun., 133 (1997), 339.
keV. 22) D. Paganin and K. A. Nugent: Phys. Rev. Lett., 80 (1998), 2586.
23) M. A. Beltran, D. M. Paganin, K. Uesugi and M. J. Kitchen: Opt.
Express, 18 (2010), 6423.
Acknowledgement 24) Duhamel and M. Vetterli: IEEE Signal Proc. Soc., 19 (1990), 259.
25) J. W. Robinson, E. M. Skelly Frame and G. M. Frame, II: Under-
This work was partly supported by Grant–in–Aid for Sci- graduate Instrumental Analysis, Ed. 6, Marcel Dekker, New York,
entific Research (A) (No. 22246097) which is provided by (2005), 543.
Japan Society for the Promotion of Science (JSPS). Project

1995 © 2016 ISIJ