Defect control in a selective reduction reaction to synthesize a metal/ceramic nanocomposite

Article abstract of DOI:10.1111/j.1551-2916.2007.01939.x

A metal/ceramic nanocomposite was synthesized by defect control in a selective reduction reaction of a complex solid solution. The defect type of a (Co, Mg)O solid solution was controlled by doping oxides with different valencies. The controlled defects strongly affected the reduction reaction behaviors as well as material morphologies after the reaction. In the case of scandium doping, the diffusion was rate controlling in the reduction reaction, and metals of a nanocolumn-like structure were highly dispersed into the matrix. On the other hand, in the case of lithium doping, the interface reaction was rate controlling in the reaction, and metal nanoparticles were highly dispersed into the matrix.

Full text of DOI:10.1111/j.1551-2916.2007.01939.x

J. Am. Ceram. Soc., 90 [11] 3441–3448 (2007)
DOI: 10.1111/j.1551-2916.2007.01939.x
r 2007 The American Ceramic Society
ournal
J
Defect Control in a Selective Reduction Reaction to Synthesize a
Metal/Ceramic Nanocomposite
w
Tomohiro Suetsuna, Takayuki Fukasawa, Seiichi Suenaga, and Koichi Harada
Functional Materials Laboratory, Corporate Research & Development Center, Toshiba Corporation,
Kanagawa 212-8582, Japan
A metal/ceramic nanocomposite was synthesized by defect con-
trol in a selective reduction reaction of a complex solid solution.
The defect type of a (Co, Mg)O solid solution was controlled by
doping oxides with different valencies. The controlled defects
strongly affected the reduction reaction behaviors as well as
material morphologies after the reaction. In the case of scandi-
um doping, the diffusion was rate controlling in the reduction
reaction, and metals of a nanocolumn-like structure were highly
dispersed into the matrix. On the other hand, in the case of
lithium doping, the interface reaction was rate controlling in the
reaction, and metal nanoparticles were highly dispersed into the
matrix.
important to increase variation of the morphologies in metal/
ceramic composite materials.
The objective of the present research is to investigate the re-
duction reaction behavior of the (Co, Mg)O solid solution sys-
tem in a hydrogen atmosphere, and to increase the variation of
the morphologies in the metal/ceramic nanocomposite material
by controlling defects in the solid solution. The defect type of
(Co, Mg)O solid solution was controlled by doping an oxide with
different valencies. It was found that the reduction behaviors of
(Co, Mg)O as well as the morphologies after the reduction were
drastically modified by controlling the defect in (Co, Mg)O.
II. Experimental Procedure
The raw materials used for the present research were MgO
(
I. Introduction
500A, Ube Material Industries, Ltd., Ube, Japan; 499.9% pu-
OMPOSITE materials, such as metal/ceramic and metal/
polymers, offering multifunctionality and superior physical
rity, o0.1-mm particle size), CoO (NanoTek CoO-2, C.I. Kasei
C
Co. Ltd., Tokyo, Japan; 499% purity (CoO:Co O 5 55:44), o
3
4
properties, are expected to be used for various applications.
For example, metal nanoparticles embedded in a ceramic can
improve or modify the mechanical, optical, electrical, and mag-
0
.1-mm particle size), Sc O (45940F, Kojundo Chemical Lab.
2 3
Co. Ltd., Saitama, Japan; 99.99% purity, 1–3-mm particle size),
and Li CO (126-01135, Wako Pure Chemical Industries Ltd.,
Osaka, Japan; 99% purity) powders. Sc O and Li CO were
2
3
1
–7
netic properties of the ceramic. The metal nanoparticles also
become stable in a matrix such as a ceramic and a polymer, al-
though their oxidation resistance and thermal stability are quite
low when their sizes are of a nanoscale order. Therefore, pecu-
liar physical properties of metal nanoparticle such as size effect
and excellent catalytic property can be obtained by embedding it
into the matrix. These advantages of composite materials can be
obtained by controlling the material morphology and interface
structure at the nanoscale, and process development of the com-
posite material is very important.
2
3
2
3
selected as doping oxides with different valencies (trivalent and
monovalent) from those of CoO and MgO (divalent). Among
2 3
the various trivalent oxides, Sc O is distinguished by its small
ionic radius and the ease with which relatively large amounts of
1
3
Sc
MgO, Sc
2
O
3
can be solved into another oxide. These powders (CoO,
, and Li CO ) were mixed at a given composition in
O
2 3
2
3
ethanol for 24 h using a ball mill, and the mixed powders were
sieved to 200 mm after removing ethanol. These powders were
sintered in air at 14001C for 20 h in a furnace (MSTF-1530,
Yamada Denki Co. Ltd., Tokyo, Japan) to produce (Co, Mg)O
solid solution powders with and without Sc O or Li O as a
One of the effective methods to produce metal nanoparticles
embedded in a matrix is the selective reduction reaction method
from a complex solid solution, which is composed of a stable
2
3
2
doping oxide (Li CO
2
3
2
was changed to Li O during sintering in
1
–12
and unstable oxide in a reducing atmosphere.
In this meth-
air at 14001C for 20 h. Co
O
3 4
was also changed to CoO during
od, metal nanoparticles are deposited selectively from an unsta-
ble oxide in a complex solid solution in a reducing atmosphere.
Previous studies revealed that nickel or cobalt nanoparticles
could be deposited on the surface of an oxide matrix by reducing
the sintering). Each powder was ground and 50 mg of it was
reduced in flowing hydrogen (at a flowing rate of 100 sccm) us-
ing a thermogravimetric analyzer system (TG; TGD-9600 and
MTS-9000, ULVAC-RIKO Inc., Yokohama, Japan). The heat-
ing rate was 151C/min for the investigation of the reduction re-
action behavior over a wide range of temperatures up to 10001
and 2001C/min for the investigation of the reaction rate constant
at a specific temperature. It was investigated for the reaction
behavior from 7 to 13 wt% reduction in order to obtain a uni-
form sample temperature.
1
–12
(
Ni, Mg)O or (Co, Mg)O solid solutions.
However, almost
all the metal particles in these studies were deposited on the
surface of grain boundaries of the matrix, and not into the grain.
Therefore, the packing densities of metal particles into the oxide
were low, and insufficient for applications such as particulate
recording media, where the large packing density of metal par-
ticles is important. Also, the deposited metal particles in the
previous studies have either cubic or spherical structures.
Because the shape of metal particles generally affects physical
properties such as the electrical and magnetic properties, it is
2,3
The real densities of the sintered powders were examined by a
pycnometer (Micromeritics Gas Pycnometer Accupyc 1330-01,
Shimadzu Analytical & Measuring Center Inc., Tokyo, Japan)
using helium gas. About 7 g of each of the sintered powders was
used for the density measurements after they were ground and
heated in vacuum at 2001C for 1 h. Their compositions were
examined by an inductive-coupled plasma spectrometer (ICP;
ICPS-8000, Shimadzu Analytical & Measuring Center Inc.).
Crystalline phases were identified by X-ray diffractometry
1
–12
L. Klein—contributing editor
Manuscript No. 23016. Received April 2, 2007; approved June 11, 2007.
Author to whom correspondence should be addressed. e-mail: tomohiro.suetsuna@
22
w
(XRD; M18XHF -SRA, MacScience Co. Ltd., Yokohama,
Japan), using CuKa radiation at 40 kV and 100 mA. Lattice
toshiba.co.jp
3
441

3
442
Journal of the American Ceramic Society—Suetsuna et al.
Vol. 90, No. 11
Table I. Information on the (Co, Mg)O Samples Sintered at 14001C for 20 h
˚
Lattice constant (A)
ꢁ3
)
Sample No.
Density (g/cm
Co:Mg:Sc (or Li) ratio
Crystalline phase
1
2
3
4
5
6
4.2529870.00043
4.2542770.00072
4.2558170.00053
4.2527170.00024
4.2492970.00058
4.2469970.00047
5.857470.0038
5.843470.0023
5.749570.0032
5.847570.0030
5.819170.0040
5.716070.0036
4.02:1:0 (without additive)
3.98:1:0.00973 (Sc)
3.99:1:0.104 (Sc)
4.02:1:0.00497 (Li)
4.04:1:0.0557 (Li)
4.03:1:0.142 (Li)
Only solid solution
Only solid solution
Solid solution1Sc O (little)
3
2
Only solid solution
Only solid solution
Only solid solution
constants were examined by mixing the sample powders with sil-
icon standard powder (640c, National Institute of Standards &
(A) Without Dopant: Both CoO and MgO are p-type
oxides, which easily form metal cation vacancy. Assuming that
the synthesized (Co, Mg)O are also a p-type oxide, the defect
would be formed in accordance with the following equation:
Technology, Gaithersburg, MD; lattice constant at 22.51C is
˚
5
.431194670.0000092 A). The nanostructures and compositions
of both sintered and reduced powders were examined by scanning
electron microscopy (SEM; JSM-840F, JEOL Ltd., Tokyo, Japan)
and transmission electron microscopy (TEM; JEM-2000EX and
JEM-2000FX, JEOL Ltd.) with an accelerating voltage of 200 kV,
and energy-dispersive spectroscopy (EDS; JEM-2000FX, JEOL
Ltd., and Link-ISIS, Oxford Instruments plc, Abingdon, U.K.).
1
O2 ! V00 þ OO þ 2hꢀ
(1)
M
2
where M is the metal cation of the (Co, Mg)O solid solution.
Assuming that cobalt and magnesium equally have a metal
cation vacancy, the composition of the synthesized (Co, Mg)O
was (Co0.79748, Mg0.19817)O from the experimental composition,
lattice constant, and Eq. (1).
III. Results and Discussion
(
B) With Scandium as a Dopant: In the (Co, Mg)O solid
solution, the metal cations are divalent, whereas the scandium
ion of the doping oxide is trivalent. Therefore, when the scan-
dium ion is solved into the (Co, Mg)O, the defect would be
formed in accordance with the following Eqs. (2)–(4) or (5) be-
cause the total charge remains neutral:
(
1) Synthesized Solid Solution
The information on the (Co, Mg)O samples (named Sample No.
–6) sintered in air at 14001C for 20 h is shown in Table I. The
1
ratio of cobalt to magnesium in all the samples was about 4. The
crystalline phase of all the samples except Sample No. 3 was
only the (Co, Mg)O solid solution phase. In Sample No. 3, a
small amount of a Sc
solution phase was detected, because the doping concentration
of scandium ion in (Co, Mg)O was sufficient to exceed the solid
solubility limit. As the doping concentration of scandium ion
in (Co, Mg)O increased, the lattice constant of the (Co, Mg)O
solid solution increased and the real density decreased. On the
other hand, as the doping concentration of lithium ion in-
creased, both the lattice constant and real density decreased.
Figures 1 and 2 show the dependence of the doping ratio on the
experimental and calculated density of the sintered (Co, Mg)O
samples (Fig. 1 is for the case in which Sc O was the doping
Sc2O3 ! 2Scꢀ þ 3OO þ V00
(2)
(3)
(4)
(5)
O
2 3
phase as well as a (Co, Mg)O solid
M
M
Sc2O3 ! 2Scꢀ þ 2OO þ Oi
00
M
1
3
Sc2O3 ! 2Sciꢀꢀꢀ þ 3OO þ 3V0M0
Sc2O3 ! 2Scꢀ þ 3O0i0
ꢀꢀ
i
Equations (2) and (3) apply when a metal cation of (Co,
Mg)O is displaced by a scandium ion of the doping oxide. Metal
cation vacancy is formed in accordance with Eq. (2), and inter-
stitial oxygen is formed in accordance with Eq. (3). Eqs (4) and
(5) apply when the scandium ion of the doping oxide becomes an
interstitial ion. Metal cation vacancy is formed in accordance
2
3
oxide and Fig. 2 is for the case in which Li O was the doping
2
oxide). The calculated density was obtained from the experi-
mental composition (Co:Mg:Sc(or Li) ratio) and lattice con-
stant, considering the following model.
6.02
5.94
5.86
5.78
5.70
Calculated from equation (2)
Calculated from equation (3)
Calculated from equation (4)
Calculated from equation (5)
Experimental
0
0.005
0.010
0.015
0.020
0.025
Sc/(Sc+Co+Mg) ratio
Fig. 1. Sc/(Sc1Co1Mg) ratio dependence on the experimental and calculated density of the (Co, Mg)O samples sintered at 14001C for 20 h. The line of
Eq. (2) is the same as that of Eq. (4).

November 2007
Selective Reduction Reaction to Synthesize a Metal/Ceramic Nanocomposite
3443
6.00
5.92
5.84
5.76
5.68
Calculated from equation (6)
Calculated from equation (7)
Calculated from equation (8)
Calculated from equation (9)
Experimental
0
0.01
0.02
0.03
Li/(Li+Co+Mg) ratio
Fig. 2. Li/(Li1Co1Mg) ratio dependence on the experimental and calculated densities of the (Co, Mg)O samples sintered at 14001C for 20 h. The line
of Eq. (6) is the same as that of Eq. (9).
with Eq. (4), and interstitial oxygen is formed in accordance with
Eq. (5). The calculated densities were obtained from the exper-
imental composition, lattice constant, and Eqs. (2)–(4) or (5)
shown as dotted or solid lines in Fig. 1. From Fig. 1, the
experimental density is in relatively good agreement with the
calculated ones in the case of Eqs. (2) or (4); the metal cation of
Sample No.2 (700ºC)
2
m =0.129t+48.1
1
1
1
1
70
50
30
10
Sample No.3 (575°C)
m =0.155t+48.2
2
(
Co, Mg)O is displaced by the scandium ion of the doping oxide
or the scandium ion becomes an interstitial ion to form a metal
cation vacancy. It was found that the concentration of the metal
cation vacancy increased as the doping concentration of the
scandium ion increased.
Sample No.1 (900°C)
2
∆
m =0.0823t+48.8
9
7
5
0
0
0
(
C) With Lithium as a Dopant: When lithium ion is
solved into the (Co, Mg)O, the defect would be formed in
accordance with the following equations:
0
400
800
Time; t/sec
Fig. 4. Weight change behaviors of the scandium-doped and undoped
Co, Mg)O samples in the flowing hydrogen as a function of time at a
1200
1600
0
ꢀꢀ
i
Li2O ! 2Li M þ OO þ M
(6)
(7)
(8)
(9)
Li2O ! 2Li0 þ OO þ VOꢀꢀ
M
(
ꢀ
00
specific temperature. Regression lines are also shown as black lines.
Li2O ! 2Li þ O
i
i
ꢀ
00
M
Li2O ! 2Li þ OO þ V
in accordance with Eq. (9). The calculated densities were ob-
tained from the experimental composition, lattice constant, and
Eqs. (6)–(8), or (9) shown as dotted or solid lines in Fig. 2. From
Fig. 2, the experimental density was in relatively good agreement
with the calculated one in the case of Eq. (7); the metal cation of
(Co, Mg)O is displaced by the lithium ion of the doping oxide to
form an oxygen ion vacancy. It was found that the concentra-
tion of oxygen ion vacancy increased as the doping concentra-
tion of lithium ion increased.
i
Equations (6) and (7) apply when a metal cation of (Co,
Mg)O is displaced by the lithium ion of the doping oxide. In-
terstitial metal cation is formed in accordance with Eq. (6), and
an oxygen vacancy is formed in accordance with Eq. (7). Equa-
tions (8) and (9) apply when the lithium ion of the doping oxide
becomes an interstitial ion. An interstitial oxygen ion is formed
in accordance with Eq. (8), and a metal cation vacancy is formed
2
Sc-system
Sample No.3
5
1
0
0
Sample No.2
Sc-doping
–
5
Sample No.1
–1
–
–
–
10
15
20
–
–
2
3
Sample No.1
Sample No.2
Sample No.3
7
8
9
10
11
12
0
200
400
600
800
1000
Temperature /10 K–1
–1
–4
Temperature/°C
Fig. 5. Arrhenius plots of the hydrogen-reduction reaction of the scan-
dium-doped and undoped (Co, Mg)O samples: Sample No. 1 ( & ),
Sample No. 2 (J), Sample No. 3 (&).
Fig. 3. Weight change behaviors of the scandium-doped and undoped
Co, Mg)O samples at up to 10001C in flowing hydrogen.
(

3
444
Journal of the American Ceramic Society—Suetsuna et al.
Vol. 90, No. 11
0
1
2
Table II. Activation Energy and Frequency Factor (with 71
Standard Error) of Hydrogen-Reduction Reaction of the Doped
and Undoped (Co,Mg)O Samples Sintered at 14001C for 20 h
Li-system
kl
kp
Temper-
ature
Activation
energy
–
–
Sample No.6
Sample
No.
Reduction
behavior
Sample No.5
range (1C)
(kJ/mol)
Frequency factor
7
1
1
2
900–1000 Parabolic 204729 9.75ꢂ 10 71.66ꢂ 10
2
wt% /s)
(
7
1
Sample No.1
700–775 Parabolic 166720 9.65ꢂ 10 71.07ꢂ 10
Sample No.4
2
wt% /s)
(
10
2
8
00–850 Parabolic 199754 1.06ꢂ 10 73.53ꢂ 10
2
wt% /s)
–3
(
7
8
9
10
11
12
10
3
4
575–700 Parabolic 186712 5.02ꢂ 10 75.12
Temperature /10 K–1
–
1
–4
2
wt% /s)
(
4
850–950 Parabolic 12776 9.30ꢂ 10 71.89
Fig. 8. Arrhenius plots of the hydrogen-reduction reaction of the lith-
ium-doped and undoped (Co, Mg)O samples: Sample No. 1 ( & ), Sam-
ple No. 4 (J), Sample No. 5 (&), Sample No. 6 (&).
2
wt% /s)
(
5
6
700–800 Linear
600–700 Linear
9.69 1.19 (wt%/s)
12.7 2.39 (wt%/s)
out the scandium-doping system (including the sample without
doping oxide) obeyed a parabolic law as follows:
5
0
5
2
Dm ¼ kpt
(10)
Li-doping
where Dm is the weight loss per unit surface area, k is the par-
p
–
abolic reduction reaction rate constant, and t is the reduction
time. This parabolic law indicates that diffusion was rate
controlling in the reaction. The reduction reaction rate constant
at each temperature was also examined to obtain Arrhenius
plots, as shown in Fig. 5. Table II summarizes the activation
energy and frequency factor from the Arrhenius plots (the data
of lithium-doped samples are also shown in Table II). From
Fig. 5 and Table II, the activation energy for reduction reaction
decreased slightly and the frequency factor increased as the dop-
ing concentration of scandium ion increased. The diffusion rate
of the metal cation of the (Co, Mg)O solid solution increased
on increasing the metal cation vacancy, which was due to the
increasing doping concentration of scandium ion. However, the
oxygen ion diffusion or hydrogen diffusion would be rate con-
trolling in the reduction reaction of (Co, Mg)O in this study,
because the oxygen ion diffusion is much slower than that of
metal cation diffusion if the metal cation vacancy is the dom-
inant defect. Therefore, the increased metal cation diffusion rate
would have little effect on the activation energy for the reduction
reaction, and would affect the frequency factor because the in-
creased metal cation diffusion rate could increase the possibility
of starting a reduction reaction. This increase in the frequency
factor could be a reason for the decreased temperature from
which reduction started.
Sample No.1
Sample No.4
–
–
–
10
15
20
Sample No.5
Sample No.6
0
200
400
600
800
1000
Temperature/°C
Fig. 6. Weight change behavior of the lithium-doped and undoped (Co,
Mg)O samples at up to 10001C in flowing hydrogen.
(
2) Reduction Reaction Behavior
The (Co, Mg)O solid solution samples were reduced in flowing
hydrogen. The hydrogen-reduction reaction behaviors of the
scandium-doped (Co, Mg)O solid solution are shown in Fig. 3.
The behavior of the undoped (Co, Mg)O solid solution is also
shown in Fig. 3. From Fig. 3, it can be seen that as the doping
concentration of scandium ion increased, the temperature from
which reduction started drastically decreased. Figure 4 shows
the weight change behavior as a function of reduction time at a
specific temperature. The reduction reaction behaviors through-
1
1
1
1
70
50
30
10
14
Sample No.4 (850°C)
∆ m =0.115t+44.9
Sample No.6 (600°C)
2
|
∆ m|=0.416t+7.79
1
2
10
8
Sample No.5 (700°C)
∆ m|=0.358t+7.09
|
Sample No.1 (900°C)
9
7
5
0
0
0
2
∆
m =0.0823t+48.8
6
0
400
800
Time; t/sec
1200
1600
0
5
10
15
20
Time; t/sec
Fig. 7. Weight change behaviors of the lithium-doped and undoped (Co, Mg)O samples in the flowing hydrogen as a function of time at a specific
temperature. Regression lines are also shown as black lines.

November 2007
Selective Reduction Reaction to Synthesize a Metal/Ceramic Nanocomposite
3445
undoped (Sample No.1)
Sc-doped (Sample No. 2)
Li-doped (Sample No. 5)
(
a)
b)
(
(
a)
b)
c)
(a)
(b)
(c)
(
(c)
(
3
0
40
50
60
70 30
40
50
60
70 30
40
50
60
70
2
ꢀ/ °
2ꢀ/ °
2ꢀ/ °
Fig. 9. X-ray diffraction patterns of the scandium-doped, lithium-doped, and undoped (Co, Mg)O samples before and after the hydrogen reduction: (a)
before reduction, (b) after 2.5–3 wt% reduction, (c) after 5.5–6 wt% reduction (reduction temperature was 10001C in the case of the undoped sample,
and 7501C in the case of the doped ones)
ꢀ
, (Co, Mg)O solid solution; ., Co (15-0806).
The reduction reaction behaviors of the lithium-doped (Co,
Mg)O solid solution are shown in Fig. 6. The behavior of the
undoped (Co, Mg)O solid solution is also shown in Fig. 6. From
Fig. 6, as the doping concentration of the lithium ion increased,
the temperature from which reduction started drastically
decreased, except that of Sample No. 4. The reduction behav-
ior also drastically changed, except that of Sample No. 4; the
weight loss progressed rapidly at the temperature from which
reduction started. Figure 7 shows the weight change behavior as
a function of time. In the case of Sample No. 4, the reduction
behavior obeyed a parabolic law, and in the case of Sample
No. 5 and No. 6, the behavior obeyed a linear law as follows:
Dm ¼ klt
(11)
where k is the linear reduction reaction rate constant. The re-
l
duction reaction rate constant at each temperature was exam-
ined to obtain Arrhenius plots, as shown in Fig. 8 (Table II
summarizes the activation energy and frequency factor from the
Arrhenius plots). The linear law of the reduction reaction is
Fig. 10. Scanning electron microscopy micrographs of the scandium-doped, lithium-doped, and undoped (Co, Mg)O samples after the hydrogen
reduction; (a) after 2.5–3 wt% reduction, (b) after 5.5–6 wt% reduction at 10001C for the undoped sample, and at 7501C for the doped samples.

3
446
Journal of the American Ceramic Society—Suetsuna et al.
Vol. 90, No. 11
cobalt of
column-like
structure
5
00nm
200nm
100nm
(
111)matrix
(111)Co
(131)Co
(
131)matrix
(022)Co
(
022)matrix
Fig. 11. Transmission electron microscopy micrographs of the scandium doped (Co, Mg)O (Sample No. 2) after the hydrogen reduction at 7501C (after
.5–6 wt% reduction). White arrows indicate the deposited cobalt of a nano column-like structure. Diffraction pattern in the [ 2¯ 1¯ 1¯ ] zone axis is also shown
in the Ø400 nm areas including the deposited cobalt particles and matrix.
5
thought to be realized when interface reaction controls the re-
action, while parabolic law is thought to be realized when diffu-
sion controls the reaction. In the case of the quite small doping
concentration of lithium ion, the reduction behavior obeyed a
parabolic law, and diffusion was rate controlling. On the other
hand, in the case of the large doping concentration of lithium
ion, the reduction behavior obeyed a linear law, and the inter-
face reaction was rate controlling. In this case of the linear law,
the reduction reaction progressed more rapidly than in the case
of the parabolic law. This rapid reduction in the linear law could
have been caused by the following reason: as the doping
concentration of lithium ion increased, oxygen ion vacancy
increased and oxygen ion may have diffused faster. The diffu-
sion of cobalt ion of (Co, Mg)O would also be relatively fast,
because the concentration of cobalt ion into (Co, Mg)O was
high (the ratio of cobalt to magnesium was about 4) [Correction
added after online publication 10 August 2007; ‘‘metal cation’’
and ‘‘since it could diffuse through the oxygen ion vaccancy
because of its small ion radius compared with that of oxygen
ion.’’ deleted]. Therefore the diffusions of both metal cation
before the reduction. Figures 9(b) and (c) are after 2.5–3 and
5.5–6 wt% reductions, respectively). In all cases, a metallic
cobalt phase as well as (Co, Mg)O solid solution phase after
the reduction was observed, and the cobalt phase increased as
the amount of reduction increased.
Figure 10 shows SEM micrographs of the doped and
undoped (Co, Mg)O samples after the hydrogen reduction.
Figure 10(a) shows SEM micrographs after 2.5–3 wt% reduc-
tion and Fig. 10(b) shows them after a 5.5–6 wt% reduction. In
the case of the undoped sample, cobalt nanoparticles 20–100 nm
in diameter were dispersed on the surface of the ceramic after
2.5–3 wt% reduction, and more particles of larger sizes were
dispersed and the surface morphology was rough after 5.5–6
wt% reduction. The rough surface was thought to be caused by
the drastic reduction at the surface of the ceramic. This drastic
reduction was also observed in both cases of (Co, Mg)O with a
scandium dopant and a lithium dopant even after 2.5–3 wt%
reduction. This is supported by the drastic reduction behavior
(Figs. 3 and 6). Figures 11 and 12 show TEM micrographs of the
scandium doped and lithium doped (Co, Mg)O after the hydro-
gen reduction (after 5.5–6 wt% reduction), respectively. In the
case of the scandium doping, the reduced cobalt was grown
from the surface toward the inside of the ceramic, and cobalt of
a nanocolumn-like structure was formed (Fig. 11). The diameter
of the column was about 100 nm. The scandium ion of the dop-
ing oxide increased the metal cation diffusion rate toward the
surface of the ceramic and columnar growth likely occurred
from the surface. On the other hand, in the case of the lithium
doping, the reduced cobalt particles were dispersed from the
(cobalt ion) and oxygen ion would be relatively fast, and the
interface reaction but not diffusion would be rate controlling
in the reduction reaction, which caused the rapid reduction
reaction.
(
3) Morphology After Reduction
Figure 9 shows the XRD patterns of the doped and undoped
Co, Mg)O samples before and after the reduction (Fig. 9(a) is
(

November 2007
Selective Reduction Reaction to Synthesize a Metal/Ceramic Nanocomposite
3447
cobalt particle
5
00nm
200nm
100nm
(002)Co
(
002)matrix
(111)Co
(
111)matrix
(
111)matrix
(111)Co
Fig. 12. Transmission electron microscopy micrographs of the lithium doped (Co, Mg)O (Sample No.5) after the hydrogen reduction at 7501C (after
.5–6 wt% reduction). White arrows indicate the deposited cobalt particles. The diffraction pattern in the [110] zone axis is also shown in the Ø400 nm
areas including the deposited cobalt particles and matrix.
5
surface toward the inside without excessive growth although
cobalt had a relatively large size difference (Fig. 12). The diam-
eters of the deposited cobalt particles were 10–200 nm. The lith-
ium ion of doping oxide increased the oxygen ion vacancy, and
the interface reaction (not diffusion) was rate controlling in the
hydrogen-reduction reaction. Therefore, rapid reduction from
the surface toward the inside occurred and many cobalt particles
were dispersed toward the inside.
magneto-resistance behavior and magnetization curve, and may
be favorable for some applications; for example the magnetic
anisotropy can be increased by the inverse magnetostrictive
effect.
IV. Conclusions
Although the structures realized in the present research
were not so refined, there is a possibility to control the varia-
tion of the structures of metal/ceramic nanocomposites in the
selective reduction reaction of the solid solution. This variation
of the structures could control the electric and magnetic
properties such as magneto-resistance and magnetization
behavior; for example, a nano-granular structure of ferro-
magnetic particles and antiferromagnetic matrix could enhance
the magneto-resistance effect, or the columnar structure
could enhance the magnetic anisotropy by its large shape
anisotropy.
A novel metal/ceramic nanocomposite was synthesized by defect
control in a selective reduction reaction of a complex solid
solution. The defect type of a (Co, Mg)O solid solution was
controlled by doping oxides with different valencies. Doping of
scandium in (Co, Mg)O increased the metal cation vacancy, and
doping of lithium changed the dominant defect from a metal
cation vacancy to an oxygen ion vacancy. These changes in the
dominant defects strongly affected the hydrogen-reduction
reaction behaviors as well as material morphologies after the
reaction. In the case of scandium doping, the diffusion was rate
controlling in the reduction reaction, and metals of a nanocol-
umn-like structure were highly dispersed into the matrix. On the
other hand, in the case of lithium doping, the interface reaction
was rate controlling in the reaction, and metal nanoparticles
were highly dispersed into the matrix.
From TEM diffraction patterns of Figs. 11 and 12, almost all
of the deposited cobalt particles in one matrix grain were ori-
ented in the same direction as that of the matrix like (100)Co//
(
100)matrix and (010)Co//(010)matrix. The lattice constants of
MgO, CoO, and cobalt are 4.21, 4.26, and 3.54, respectively.
The difference in the lattice constants between the (Co, Mg)O
solid solution and the deposited cobalt may induce a strain in
the interface between the matrix and the deposited cobalt. This
strain could modify the electric and magnetic properties such as
It was found that material morphologies in the metal/ceramic
nanocomposite could be varied by controlling the defect of the
solid solution. A more refined morphology could be obtained by
controlling the composition of the solid solution, grain size of
the solid solution oxide, and reduction condition.

3
448
Journal of the American Ceramic Society—Suetsuna et al.
Vol. 90, No. 11
7
T. Suetsuna, S. Suenaga, T. Fukasawa, and K. Harada, ‘‘Development of
Metal/Ceramic Nanocomposite,’’ Bull. Jpn. Electron. Mater. Soc., 37, 40–3 (2006)
(in Japanese).
References
1
J. Narayan and Y. Chen, ‘‘Physical Properties of Oxides Containing Metal
Precipitates Part II,’’ Philos. Mag. A, 49 [4] 475–92 (1984).
J. A. Smith, P. Limthongkul, L. Hartsuyker, S. Y. Kim, and S. L. Sass, ‘‘Mag-
neto-Optical Properties of Ni–MgO and Co–MgO Nanocomposite Films
8
J. Narayan, Y. Chen, R. M. Moon, and R. W. Carpenter, ‘‘Characterization of
Metal Precipitates in Magnesium Oxide Part I,’’ Philos. Mag. A, 49 [2] 287–300
(1984).
2
9
Obtained by Partial Reduction Reactions,’’ Mat. Res. Soc. Symp. Proc., 475,
2
D. L. Ricoult and H. Schmalzried, ‘‘Internal Reactions in the (Mg, Me)O Sys-
tem,’’ J. Mater. Sci., 22, 2257–66 (1987).
45–50 (1997).
J. A. Smith, P. Limthongkul, L. Hartsuyker, S. Y. Kim, and S. L. Sass, ‘‘Mag-
3
10
M. Backhaus-Ricoult and D. Ricoult, ‘‘Electron Microscopy of Internally
Reduced (Mg, Ni)O,’’ J. Mater. Sci., 23, 1309–16 (1988).
neto-Optical Properties of Nanocomposite Films Obtained by Partial Reduction
of (Ni, Mg)O and (Co, Mg)O Solid Solutions,’’ J. Appl. Phys., 83 [5] 2719–26
11
J. A. Smith, P. Limthongkul, and S. L. Sass, ‘‘Microstructures and Mechanical
Properties of Ni–MgO Composites Formed by Displacement and Partial Reduc-
(
1998).
4
T. Suetsuna, S. Suenaga, and T. Fukasawa, ‘‘Monolithic Cu–Ni-Based Cata-
lyst for Reforming Hydrocarbon Fuel Sources,’’ Appl. Catal A, 276, 275–9 (2004).
T. Fukasawa, T. Suetsuna, K. Harada, and S. Suenaga, ‘‘Catalytic Perfor-
mance of Metal Nanoparticles Supported by Ceramic Composite Produced by
tion Reactions,’’ Acta Mater., 45 [10] 4241–50 (1997).
S. Suenaga, T. Suetsuna, K. Harada, and T. Fukasawa, ‘‘In Situ Formation
12
5
of Metal Nanoparticles on Ceramic Surface,’’ J. Am. Ceram. Soc., 87 [5] 963–6
(2004).
13
Partial Reduction of Solid Solution with Dopant,’’ J. Am. Ceram. Soc., 88 [10]
2
F. Henriksen and W. D. Kingery, ‘‘The Solid Solubility of Sc O , Al O ,
2
3
2
3
938–41 (2005).
K. Harada, T. Fukasawa, T. Suetsuna, and S. Suenaga, ‘‘Small Reformer for
Cr O , SiO , and ZrO in MgO,’’ Ceramurgia Int., 5 [1] 11–7 (1979).
2
&
3
2
2
6
Steam Reforming of Methanol Using Ni Nanoparticles on a Ceramic Surface,’’
Trans. Mater. Res. Soc. Jpn., 30 [4] 987–9 (2005).