Materials Transactions, Vol. 49, No. 3 (2008) pp. 522 to 526
#2008 The Japan Institute of Metals
Face-Centered Cubic Ti Cluster Assemblies
Prepared by Plasma-Gas-Condensation
Naokage Tanaka1; * , Dong-Liang Peng1;2 , Kenji Sumiyama1 and Takehiko Hihara1
1
2
Department of Materials Science and Engineering, Nagoya Institute of Technology, Nagoya 466-8555, Japan
Department of Materials Science and Engineering, College of Materials, Xiamen University, Xiamen 361005, P. R. China
Ti cluster assemblies have been prepared by a plasma-gas-condensation cluster deposition system, where the Ar gas flow rate (RAr ) was
fixed and the He gas flow rate (RHe ) was varied. The microstructure of Ti clusters has been observed by a transmission electron microscope
(TEM). The mean Ti cluster size, dm , decreases with increasing RHe . Ti clusters are hexagonal close-packed (hcp) for dm > 15 nm, while they
are face-centered cubic (fcc) for dm < 15 nm. The critical cluster size between the fcc and hcp phases is much larger than the theoretically
predicted value and that (5–7 nm) for mechanically milled Ti powders. The fcc phase formation is attributed to the negative internal pressure in
small Ti clusters. Moreover, Ti cluster assemblies have been prepared via the above-mentioned procedures with slight introduction of oxygen
gas in the deposition chamber, and investigated by high resolution TEM, electron diffraction, and optical transmittance measurements. No clear
core-shell morphology is observed in the partially oxidized Ti cluster assemblies. [doi:10.2320/matertrans.MRA2007174]
(Received July 25, 2007; Accepted December 5, 2007; Published February 14, 2008)
Keywords: titanium cluster, plasma-gas-condensation, nonequilibrium structure, transmission electron microscopy, optical property
1.
Introduction
Assembling of nanometer sized clusters is a vital method
to fabricate nanoscale structure-controlled materials, and
optimizations of their structure, concentration, stability as
well as suitable accumulations are demanded because their
sizes correspond to the elemental units of material functions.1,2) For this purpose, we have developed a plasma-gascondensation cluster deposition (PGCCD) system depicted
in Fig. 1.3–5) It is composed of three main parts: a sputtering
chamber, a growth room and a deposition chamber. Using
this system, we could prepare size-monodisperse Co clusters
whose mean diameter, dm , ranged between 5 and 15 nm with
the standard deviation less than 10% of dm . With increasing
the cluster deposition time, geometrical and electrical network formations were observed and discussed in terms of a
percolation6) as well as a transition from the superparamagnetic to ferromagnetic states in these assemblies.7)
In Co clusters prepared by slightly introducing oxygen gas
into the deposition chamber of the PGCCD system and
controlling the oxygen flow rate (R0 O2 ), their surfaces were
uniformly oxidized:8) uniform size Co cores were covered by
CoO layers with uniform thickness. The metallic cores were
stable against oxidation in ambient atmosphere and their
direct contacts were prohibited. In such core-shell type
monodispersive Co cluster assemblies, the electrical resistivity and the magnetoresistance were predominated by
electron tunneling between the Co cores,9) and the magnetic
relaxation by macroscopic quantum tunneling at low temperatures.10) These characteristic features were well distinguished from the normal semiconductor conductivity and the
classical magnetic relaxation at high temperatures.
Recently we have produced Ti clusters by introducing the
oxygen gas into the sputtering chamber of the PGCCD
system (RO2 6¼ 0 mol/s),11) where Ti clusters were more
*Graduate Student, Nagoya Institute of Technology, Corresponding author,
E-mail: [email protected]
Fig. 1 A plasma-gas-condensation cluster deposition system.
severely oxidized than those prepared by introducing the
oxygen gas into the deposition chamber (R0 O2 6¼ 0 mol/s).
With increasing RO2 , the crystal structures of Ti clusters
changed from a face-centered cube, via an NaCl-type to an
amorphous and rutile-type mixture, contradicting the fact that
there is no fcc phase in the equilibrium phase diagram.12)
The electrical resistivity was rather low (< 103 m m) for
RO2 < 4:8 108 mol/s, while it became about six orders of
magnitude larger for RO2 > 5:6 108 mol/s. Then, the
optical transmittance in a visible wavelength range was very
low (< 20%) for RO2 < 7:5 108 mol/s, while it became as
high as about 80% for RO2 ¼ 8:9 108 mol/s.11)
The high-resolution TEM images of Ti clusters prepared
with RO2 6¼ 0 mol/s revealed no wide range lattice fringe and
no core-shell morphology in contrast to those of Co clusters
prepared with R0 O2 6¼ 0 mol/s. Polymorphic structures have
been observed in transition metal ultrafine particles which
were prepared by an inert gas condensation method and
discussed in terms of the cluster size, internal pressure,
contamination of ambient gases, etc.13) In this context, we
describe the effects of cluster size and oxygen contamination
on forming abilities of the fcc phase and core-shell
morphology in Ti cluster assemblies, comparing with
previous results of Ti clusters prepared with varying RO2 .
Face-Centered Cubic Ti Cluster Assemblies Prepared by Plasma-Gas-Condensation
2.
Experimental Procedures
In the PGCCD system (see Fig. 1), Ti metal vapors were
generated from Ti target by dc sputtering. A mixture of argon
and helium gases was continuously injected in the sputtering
chamber through a nozzle. To control Ti cluster sizes, the
argon gas flow rate, RAr , was fixed at 1:9 104 mol/s
and the helium gas flow rate, RHe , was varied from 0 to
7:1 104 mol/s, where the total pressure of the sputtering
chamber was 1:5 103 Pa for RHe ¼ 0 mol/s and 6:1 103
Pa for RHe ¼ 7:1 104 mol/s. In order to study the oxygen
contamination effect, oxygen gas was introduced into the
deposition chamber from a gas inlet with the flow rate,
R0 O2 ¼ 0{2:2 106 mol/s, where the total pressure in the
deposition chamber varied from 7:1 103 to 1:5 102 Pa
with R0 O2 .
Vaporized atoms in the sputtering chamber were decelerated by collisions with a large amount of Ar and He atoms,
and were swept to the growth room which was cooled by
liquid nitrogen. Ti clusters formed in the growth room were
ejected by differential pumping through an aperture, and a
part of the cluster beam was intercepted by skimmers and
then deposited on a substrate set on a sample holder in the
deposition chamber. The thickness monitor was installed
behind the substrate and its value was used as a quantity
measure of deposited clusters on a given area (about 50 mm2 )
of the substrate. We also evaluated thicknesses of Ti clusterassembled films by a stylus instrument. It took about 60 s to
prepare Ti cluster assemblies whose effective thicknesses
were 10–40 nm under different RHe and R0 O2 . We deposited
Ti clusters onto TEM microgrids covered by carbon-coated
collodion films which are supported by Cu grids for TEM
observations, and quartz glass plates for optical transmittance
and thickness measurements. The samples thus obtained
were observed with the Hitachi HF-2000 TEM operated at
200 kV. The mean cluster size, dm , was estimated from about
two hundreds Ti clusters on the TEM microgrid. Their crystal
structures were investigated by selected area electron
diffraction (ED) patterns of Ti cluster assemblies deposited
on the TEM microgrids. The optical transmittance spectra in
the wavelength range between 200 and 800 nm were
measured by the ultra violet-visible (UV-VIS) spectrophotometer (Jasco V-570) and calibrated against a bare quartz
glass as a reference sample. The sample thicknesses were
about 3 nm for TEM observations, about 30 nm for ED
measurements, and about 120 nm for optical transmittance
measurements.
3.
3.1
523
Fig. 2 Bright field TEM images of Ti clusters at initial deposition stage
prepared with (a) RHe ¼ 0 mol/s, (b) 2:6 104 mol/s, (c) 4:1 104
mol/s and (d) 7:1 104 mol/s, where RAr ¼ 1:9 104 mol/s for (a)–
(d). The size distribution histogram (a0 )–(d0 ) are estimated from (a)–(d).
Results
Influence of inert gas flow rates in the sputtering
chamber
Figures 2(a)–(d) show bright field TEM images of Ti
clusters at initial deposition stages prepared with RHe ¼
0{7:1 104 mol/s, and Figs. 2(a0 )–(d0 ) their size distributions estimated from Figs. 2(a)–(d). As seen in these figures,
many Ti clusters are spherical, but some ones are polyhedral
in particular for RHe ¼ 0 mol/s (Fig. 2(a)). For RHe ¼ 0
mol/s, dm is 18 nm and the size distribution is rather wide: the
standard deviation, ¼ 14% (Fig. 2(a0 )). With increasing
Fig. 3 Electron diffraction patterns of Ti cluster assemblies prepared
with (a) RHe ¼ 0 mol/s, (b) 2:6 104 mol/s, (c) 4:1 104 mol/s and
(d) 7:1 104 mol/s, where RAr ¼ 1:9 104 mol/s for (a)–(d).
RHe , the dm value monotonically decreases (see Fig. 4(a)):
dm ¼ 10 nm for RHe ¼ 7:1 104 mol/s (Fig. 2(d0 )). In
these Ti clusters the size-monodispersivity is improved with
adding He gas, where ¼ 5:7% for RHe ¼ 2:6 104 mol/s
(Fig. 2(b0 )).
524
N. Tanaka, D.-L. Peng, K. Sumiyama and T. Hihara
Fig. 5 Bright field TEM images of Ti clusters at initial deposition stages
prepared with (a) RHe ¼ 0 and R0 O2 ¼ 2:2 106 mol/s, and (b)
RHe ¼ 4:1 104 and R0 O2 ¼ 2:2 106 mol/s. Electron diffraction patterns of Ti cluster assemblies prepared with (a0 ) RHe ¼ 0 and R0 O2 ¼
2:2 106 mol/s, and (b0 ) RHe ¼ 4:1 104 and R0 O2 ¼ 2:2 106
mol/s. There, we fixed RAr ¼ 1:9 104 mol/s for (a)–(b0 ).
Fig. 4 (a) The mean cluster size, dm , and (b) the lattice constant, a, of
fcc Ti clusters as a function of RHe .
Figures 3(a)–(d) show electron diffraction (ED) patterns of
Ti cluster assemblies prepared with RHe ¼ 0{7:1 104
mol/s. In Fig. 3(a), the ED pattern of the Ti cluster assembly
prepared with RHe ¼ 0 mol/s displays a set of strong hcp
diffraction rings and a weak ring indicated by X (Fig. 3(a)).
The ED pattern for RHe ¼ 2:6 104 mol/s displays a set of
broad hcp diffraction rings and halo rings (Fig. 3(b)),
indicating presence of very small crystallites. With further
increasing RHe (¼ 4:1 104 and 7:1 104 mol/s), the ED
patterns consist of a set of fcc diffraction rings (Figs. 3(c) and
(d)). The lattice constant (a) of the fcc phase has been
estimated from the ED pattern. Figures 4(a) and (b) show dm
and a as a function of RHe . As shown in Fig. 4(b), a ¼
0:433 nm, being independent of RHe and dm . This value is in
agreement with the reported values,14–16) but much larger
than the calculated one (0.406 nm).17) The lattice spacing
estimated from the diffraction ring X is 0.154 nm, being
attributable to (220)fcc of the fcc phase with a ¼ 0:433 nm.
However, we cannot confirm presence of the fcc phase
because (111)fcc and (200)fcc rings are not detected at around
the (100)hcp and (002)hcp rings in that specimen.
3.2
Influence of oxygen flow rates in the deposition
chamber
Figures 5(a) and (b) show bright field TEM images of Ti
clusters at initial deposition stages prepared with RHe ¼ 0
and 4:1 104 mol/s, and R0 O2 ¼ 2:2 106 mol/s. Although dm decreases with increasing RHe , the dm value and
the cluster shape are not so different between R0 O2 ¼ 0 and
2:2 106 mol/s (compare them with Figs. 2(a) and (c)).
Figures 5(a0 ) and (b0 ) show ED patterns of Ti cluster
assemblies prepared with RHe ¼ 0 and 4:1 104 mol/s,
and R0 O2 ¼ 2:2 106 mol/s. In Fig. 5(a0 ), there is a set of
Fig. 6 High resolution TEM images of Ti clusters at initial deposition
stage prepared with (a) RHe ¼ 0 and R0 O2 ¼ 0 mol/s, (a0 ) RHe ¼ 0 and
R0 O2 ¼ 2:2 106 mol/s, (b) RHe ¼ 4:1 104 and R0 O2 ¼ 0 mol/s, and
(b0 ) RHe ¼ 4:1 104 and R0 O2 ¼ 2:2 106 mol/s, where RAr ¼ 1:9 104 mol/s for (a)–(b0 ).
hcp diffraction rings, but the ring X in Fig. 3(a) is not
observed, while a set of fcc diffraction rings exists in
Fig. 5(b0 ). The a value estimated from the diffraction rings in
Fig. 5(b0 ) is about 0.423 nm and smaller than that of fcc Ti
(0.433 nm), being attributable to that of a defective NaCl
structure.18,19)
Figures 6(a)–(b0 ) show high resolution TEM images of Ti
clusters at initial deposition stages prepared with (a) RHe ¼ 0
and R0 O2 ¼ 0 mol/s, (a0 ) RHe ¼ 0 and R0 O2 ¼ 2:2 106
mol/s, (b) RHe ¼ 4:1 104 and R0 O2 ¼ 0 mol/s, and (b0 )
RHe ¼ 4:1 104 and R0 O2 ¼ 2:2 106 mol/s. Clear lattice fringes are extended over a wide range in Figs. 6(a) and
(a0 ), while they are detectable only within a narrow range in
Face-Centered Cubic Ti Cluster Assemblies Prepared by Plasma-Gas-Condensation
Fig. 7 Optical transmittance, T, as a function of the optical wave length, ,
for Ti cluster-assembled films on quartz glass substrates prepared with (a)
RHe ¼ 0, and R0 O2 ¼ 0 or 2:2 106 mol/s and (b) RHe ¼ 4:1 104 , and
R0 O2 ¼ 0 or 2:2 106 mol/s where RAr ¼ 1:9 104 mol/s for (a) and
(b).
Figs. 6(b) and (b0 ). These figures indicate that the crystallinity is degraded with increasing RHe (the decrease in dm ).
Figures 7(a) and (b) show the optical transmittance, T, as
a function of the optical wave length, , for Ti clusterassembled films on quartz substrates prepared with (a)
RHe ¼ 0 mol/s and (b) RHe ¼ 4:1 104 mol/s, while
R0 O2 ¼ 0 and 2:2 106 mol/s. Within the observed spectral
range, T is low for Ti cluster-assembled films prepared at
R0 O2 ¼ 0 mol/s, although it become higher for those prepared
at R0 O2 ¼ 2:2 106 mol/s. The absorption edge, ed , is
estimated to be about 360 nm for R0 O2 ¼ 2:2 106 mol/s.
The increase in T is ascribed to formation of titanium
dioxides.
4.
Discussion
4.1 Cluster size effect
In the PGCCD system, both dm and the cluster deposition
rate increase with increasing RAr , which is roughly proportional to the partial pressure of Ar gas (PAr ) in the sputtering
chamber.4) This is ascribed to the promotion of clusternucleation and -growth because the Ti atoms sputtered out of
the Ti target increase and the collision probability between
them is enhanced.4,5) Since clusters nucleate and grow in both
the sputtering chamber and the growth room at high RAr
(PAr ), we choose an appropriate RAr (PAr ) value to confine the
cluster growth only in the growth room for suppressing the
size distribution.
With increasing RHe , dm decreases (see Fig. 4(a)) and the
cluster deposition rate increases though the sputtering yield
525
of a He ion is about one order lower than that of an Ar ion.20)
Since the thermal conductivity of helium gas is about one
order higher than that of argon gas,21,22) helium gas reduces
the kinetic energy of Ti atoms more effectively. When the
number of Ti cluster nuclei increases and they absorb the Ti
atoms, the Ti atom concentration decreases in the growth
room and they do not grow further. The viscosity of helium
gas is also lower than that of argon gas.21,22) The helium gas
and the Ti clusters are so effectively extracted from the
growth room through the aperture that the cluster deposition
rate is increased.
The nonequilibrium fcc phase formation is an interesting
aspect in these Ti cluster assemblies. In the equilibrium state
of bulk Ti metal, an hcp phase is stable below 1155 K and a
bcc phase above 1155 K.12) Within a simple Debye model of
Ti solids, the Debye temperature of the fcc phase is lowest
and the energy gain in the entropy term of its Helmholz free
energy is smallest among these three phases: an fcc phase is
absent.12,23) The fcc phase can be stabilized against the hcp
and bcc phases, when negative internal pressure is induced in
nanometer size crystals as a surface effect. The critical size
has been predicted to be dc ¼ 5{7 nm for the phase transition
from hcp to fcc.15,16)
In mechanically milled Ti powders, the significant reduction in the cluster size less than 10 nm is accompanied by
negative internal pressure.14–16) The resultant increase in the
molar volume of Ti lattices leads to a substantial increase of
the Gibbs free energy in ultrafine hcp crystals, and thus
causes a partial phase change from hcp to fcc, whereas the
increase in the stored strain energy enhances the phase
transition, leading to the complete fcc phase in well-milled Ti
powders.
In present Ti clusters, the fcc phase formation is also
attributed to their small sizes. However, dc is roughly
estimated to be about 15 nm in Figs. 2 and 3, being two times
larger than those estimated in the previous reports.14–16) It is
difficult to estimate the stored strain energy in the present
Ti clusters.
4.2 Oxygen contamination effect
The careful structural and chemical analyses have demonstrated that contaminations (oxygen, nitrogen and transition metal impurities) play no crucial role for the fcc phase
formation in mechanically milled Ti powders.14) Although Ti
clusters prepared in the PGCCD system (in the high vacuum
system with introduction of pure argon and helium gases) are
also not severely oxidized, their surfaces are oxidized with
introducing oxygen gas into the deposition chamber. They
are also oxidized when they are exposed in ambient atmosphere for TEM observations.
In the equilibrium TiOx system,12,24) there is a random
solid solution of O in hcp Ti for x < 0:33. This is followed by
preferential occupancies of certain positions via a high
temperature heat treatment leading to the ordered structures:
Ti3 O and Ti2 O for x < 0:5. With increasing x, there is an "TaN-type structure for x ¼ 0:68{0:75, while a defective
NaCl-type structure which has a cubic symmetry with the
equal number of random vacancies in the cation and anion
positions for x ¼ 0:68{1:25 at high temperatures and for x ¼
1:0{1:25 at ordinal temperatures.
526
N. Tanaka, D.-L. Peng, K. Sumiyama and T. Hihara
In order to discuss the oxygen contamination effect, it is
worth summarizing the lattice constants of bulk TiOx alloys
and oxides18) in comparison with those of Ti cluster
assemblies. The lattice constant, a ¼ 0:424{0:416 nm for x ¼
0:68{1:25 in bulk NaCl-type TiOx oxides, while a ¼ 0:424
nm in NaCl-type Ti oxide cluster assemblies prepared with
R0 O2 ¼ 2:2 106 mol/s and a ¼ 0:433 nm in fcc Ti cluster
assemblies prepared with R0 O2 ¼ 0 mol/s. In bulk hcp TiOX
alloys, c ¼ 0:468{0:471 nm and a ¼ 0:295{0:297 nm for
x ¼ 0{0:33, while in Ti cluster assemblies c ¼ 0:469{
0:471 nm and a ¼ 0:295{0:296 nm. Since the lattice mismatch between the fcc Ti and the defective NaCl-type
structure is small, O atoms substitute Ti atom sites in fcc
clusters, and fcc and NaCl-type phases coexist in contrast to
the bulk equilibrium state.12,24) In addition, no diffraction ring
allotted to Ti oxide phases is detected in the hcp Ti clusters,
suggesting that the atomic fractions of O to Ti are less than
those in the fcc Ti clusters. In large Ti clusters, which absorb
O atoms at R0 O2 6¼ 0 mol/s, the hcp phase is formed probably
due to its high oxygen solubility.12) Therefore, core-shell
morphology is not observed in the Ti clusters prepared even
with introducing oxygen gas into the deposition chamber.
5.
Conclusion
Ti cluster assemblies have been prepared using the
PGCCD system with controlling the gas flow rates: RAr ,
RHe and R0 O2 . The TEM observations indicate the following
characteristics. The dm value decreases monotonically with
increasing RHe and fixing RAr . These Ti cluster assemblies
have an equilibrium hcp structure for dm > 15 nm, while they
have a nonequilibrium fcc structure for dm < 15 nm. The
critical size (dc ) of phase formation between the fcc and hcp
phases is about 15 nm, being larger than the reported ones
(5–7 nm). In small Ti clusters, the fcc phase formation is
attributed to the increase in the negative internal pressure.
When O atoms substitute Ti atom sites, the fcc Ti and
defective NaCl type phases presumably coexist in small Ti
clusters and only the hcp phase in large Ti clusters, where no
core-shell morphology is obtained.
Acknowledgments
This work was supported by (1) Intellectual Cluster Project
given by the Ministry of Education, Science, Culture and
Sports, Japan, Aichi Prefecture, Nagoya City and Aichi
Science and Technology Foundation, and (2) a Grant-in-Aid
for Scientific Research given by the Ministry of Education,
Science, Culture and Sports, Japan.
REFERENCES
1) P. Melinon, V. Paillard, V. Dupuis, A. Perez, P. Jensen, A. Hoareau, J.
P. Perez, J. Tuaillon, M. Broyer, J. L. Vialle, M. Pellarin, B. Baguenard
and J. Lerme: Int. J. Mod. Phys. B9 (1995) 339–397.
2) K. Sumiyama, D. L. Peng, S. Yamamuro, T. Hihara and K. Suzuki:
Proc. Int. Symp. on Cluster Assembled Mater. IPAP Conf. Series 3
(2001) p. 22–27.
3) H. Haberland, M. Karrais, M. Mall and Y. Thurner: J. Vac. Sci.
Technol. A10 (1992) 3266–3271.
4) S. Yamamuro, K. Sumiyama, M. Sakurai and K. Suzuki: Supramol.
Sci. 5 (1998) 239–245.
5) S. Yamamuro, K. Sumiyama and K. Suzuki: J. Appl. Phys. 85 (1999)
483–489.
6) S. Yamamuro, K. Sumiyama, T. Hihara and K. Suzuki: J. Phys.
Condens. Matter. 11 (1999) 3247–3257.
7) S. Yamamuro, K. Sumiyama, T. J. Konno and K. Suzuki: Mater. Trans
JIM. 40 (1999) 1450–1455.
8) D. L. Peng, K. Sumiyama, T. J. Konno, T. Hihara and S. Yamamuro:
Phys. Rev. B60 (1999) 2093–2100.
9) D. L. Peng, K. Sumiyama, S. Yamamuro, T. Hihara and T. J. Konno:
Appl. Phys. Lett. 74 (1999) 76–78.
10) D. L. Peng, K. Sumiyama, T. Hihara and S. Yamamuro: Appl. Phys.
Lett. 74 (1999) 76–78.
11) N. Tanaka, D. L. Peng, K. Sumiyama and T. Hihara: Thin Solid Films
516 (2008) 1682–1687.
12) P. G. Wahlbeck and P. W. Gilles: J. Amer. Ceram. Soc. 49 (1966) 180–
183.
13) R. Uyeda: Prog. Mater. Sci. 35 (1991) 1–96.
14) I. Manna, P. P. Chattopadhyay, P. Nandi, F. Banhart and H. J. Fecht:
J. Appl. Phys. 93 (2003) 1520–1524.
15) T. Tepper, D. Shechtman, D. van Eleerden and D. Josell: Mater. Lett.
35 (1998) 100–107.
16) J. B. Lai, L. J. Chen and C. S. Liu: Micron. 30 (1999) 205–211.
17) Powder Diffraction File, PDF-2, Joint Committee on Powder Diffraction Standard, JCPDS-ICDD, 2003, file 88-2321.
18) S. Anderson, B. Collén, U. Kuylenstierna and A. Magneli: Acta
Chem.Scand. 11 (1957) 1641–1652.
19) K. Suzuki and S. Takeuchi: Y. Hoshino, S. Iida, M. Sugino (Eds),
International Conference on Ferrite, Kyoto, Japan, July 6–9, 1970,
Ferrites; Proceeding of the International Conference, (University of
Tokyo Press, Tokyo, 1971), p. 568–572.
20) J. L. Vossen and J. J. Cuomo: Thin Film Processes, ed. J. L. Vossen and
W. Kern (Academic Press, 1978) pp. 15–18.
21) S. Iida, K. Ohno, H. Kanzaki, H. Kumagai and S. Sawada: Tables of
Physical Constants (Asakura, 1979) p. 30 (in Japanese).
22) S. Iida, K. Ohno, H. Kanzaki, H. Kumagai and S. Sawada: Tables of
Physical Constants (Asakura, 1979) p. 123 (in Japanese).
23) Z. P. Chen, Z. Wen and Q. Jiang: Solid Satate. Comm. 132 (2004) 747–
750.
24) A. F. Wells: Structural Inorganic Chemistry 5th Editions, (Clarendon
Press, Oxford, 1984) p. 562.