Metal oxide composites for lithium-ion battery anodes synthesized by the partial reduction process

Article abstract of DOI:10.1149/1.1500345

A thermochemical process based on the partial reduction of mixed oxides is used to create ultrafine metal-ceramic composites for Li-ion battery electrodes. Mixed oxides containing a more noble metal selected to be capable of alloying with lithium at potentials useful as a Li-ion battery anode are partially reduced to form electrochemically active metal-ceramic composites. Experiments show the differences in microstructure obtained in systems with slow oxygen diffusion (SbVO₄, AgVO₃, and Ag₂V₄O₁₁), fast oxygen diffusion (Sb₂Mn₂O₇ distorted fluorite), and microphase separation (Sn_0.5Ti_0.5O₂ rutile). Materials are characterized using X-ray diffraction, scanning electron microscopy, transmission electron microscopy, and scanning transmission electron microscopy, and electrochemical tests are presented. Reversible charge capacities of 200-350 mAh/g (1100-2200 mAh/cm³) have been obtained in experiments to date.

Full text of DOI:10.1149/1.1500345

Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
A1237
0013-4651/2002/149͑9͒/A1237/9/$7.00 © The Electrochemical Society, Inc.
Metal Oxide Composites for Lithium-Ion Battery Anodes
Synthesized by the Partial Reduction Process
*
**
**,z
Pimpa Limthongkul, Haifeng Wang, Eva Jud, and Yet-Ming Chiang
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, USA
A thermochemical process based on the partial reduction of mixed oxides is used to create ultrafine metal-ceramic composites for
Li-ion battery electrodes. Mixed oxides containing a more noble metal selected to be capable of alloying with lithium at potentials
useful as a Li-ion battery anode are partially reduced to form electrochemically active metal-ceramic composites. Experiments
show the differences in microstructure obtained in systems with slow oxygen diffusion ͑SbVO4 , AgVO3 , and Ag2V4O11͒, fast
oxygen diffusion ͑Sb2Mn2O7 distorted fluorite͒, and microphase separation ͑Sn0.5Ti0.5O2 rutile͒. Materials are characterized using
X-ray diffraction, scanning electron microscopy, transmission electron microscopy, and scanning transmission electron micros-
copy, and electrochemical tests are presented. Reversible charge capacities of 200-350 mAh/g ͑1100-2200 mAh/cm3͒ have been
obtained in experiments to date.
© 2002 The Electrochemical Society. ͓DOI: 10.1149/1.1500345͔ All rights reserved.
Manuscript submitted November 21, 2001; revised manuscript received March 19, 2002. Available electronically August 6, 2002.
Increased demand for high density portable energy storage has
driven searches for new storage materials with improved perfor-
mance. Several Li-metal alloy systems have been investigated as
alternatives to the carbon negative electrode in Li-rechargeable bat-
teries, due to their high theoretical energy densities.1-3 However, due
to large volume changes occurring upon Li insertion and removal
during cycling which lead to mechanical failure of the electrode, the
cyclability of bulk metal alloys is poor.
Many efforts have been directed toward improving the cycling
stability of these metal electrodes. The development of tin oxide-
containing glass anodes and the mechanistic investigation of this
class of materials have shown that improved cyclability is possible if
the metal phase is prepared as ultrafine particulates.4-6 These oxides
offer higher volumetric capacity than carbon ͑Ͼ1800 mAh/cm3͒ and
can exhibit good cyclability, but a major limiting factor is the large
irreversible capacity loss during the first cycle, due to irreversible
lithium consumption upon initial electrochemical reduction of tin
oxide to tin metal prior to lithium alloying.5 Reduction of the first-
cycle irreversibility while maintaining good cyclability has been de-
scribed for composites containing ultrafine dispersions of lithium-
active metals produced by mechanical milling or multiphase active
metal systems.6-8 These studies suggest that metal anodes may be
viable if stabilized fine dispersions of lithium-active metals are
attained.2-3,7-9
Here the partial reduction of several mixed oxides has been ex-
plored as an alternative approach to synthesizing metal-ceramic
nanocomposites with reduced first-cycle irreversibility loss, and in
which phase and volume changes during lithium insertion and re-
moval can be accommodated. This approach allows tailored phase
assemblages and microstructures to be prepared using simple, eco-
nomical thermochemical processes and inexpensive starting materi-
als. While partially reduced oxides have previously been investi-
gated for possible applications, including selective absorption of
solar energy, optical data storage, and toughening of ceramics,10-16
we show that this method can also be used to synthesize electro-
chemically active materials in several systems. Experiments are dis-
cussed for ternary oxides in the systems Sb-V-O, Sb-Mn-O, Ag-V-O,
and Sn-Ti-O.
thermochemical process in which a ternary or higher-order oxide ͑or
nitride, sulfide͒ is subjected to conditions that reduce the most noble
metal͑s͒ but leaves the less noble metal͑s͒ in an oxidized form10
partially
MeI MeIIO —— aMeI ϩ MeIIO ϩ c Ϫ y͒/2 O
͓1͔
→
͑
c
y
2
a
b
b
reduced
Here MeI is more noble than MeII ͑MeI lies above MeII on the
Ellingham diagram͒. Depending on the starting composition, the
relative diffusion rates of oxygen and the cations, and the phase
makeup ͑single or multiphase͒ of the material, either internal reduc-
tion wherein the reduced species precipitates inside an oxide matrix,
or external reduction wherein the reduced metal forms at the outer
surface, can occur.10
For use as a lithium-storage electrode, our objective is to produce
fine particles of a lithium-active metal, MeI, enclosed within or dis-
persed among particles of a metal oxide matrix MeIbIOy . The matrix
itself may or may not be lithium active but should be a good elec-
tronic and lithium-ion conductor. For high energy density as an an-
ode, alloying at a low voltage of MeI with Li and a high Li/MeI ratio
are also preferred.
One expects that the first-cycle irreversibility of the electrode can
be reduced in such a composite compared to an oxide that is elec-
trochemically reduced since the displacement reactions forming
Li2O as a by-product upon first lithium insertion5 can be avoided.
These displacement reactions are typically irreversible or poorly re-
versible. Direct lithiation of the metal phase is possible after partial
reduction
partially
xLi
MeI-MeII͒ O —— MeI ϩ MeIIO —— Li MeI ϩ MeIIO
͑
→
→
a
b
y
x
y
b
b
reduced
͓2͔
Furthermore, ultrafine metal particles, which appear to better toler-
ate cyclic volume changes than coarser particles or bulk metal,3,8,18
can be produced. Internal reduction where metal particles are em-
bedded in the oxide matrix can prohibit coarsening of the metal
particles during heat-treatment, and very fine particle sizes on the
order of a few nanometers have been obtained in previous
studies.10-16 The fact that almost all metal oxides undergo volume
shrinkage upon reduction can be advantageous by providing room
for the subsequent expansion of the metal upon lithiation. Finally, a
microstructure whereby the lithium-active metal is embedded within
a metal oxide matrix can improve the chemical compatibility be-
tween the storage material and the electrolyte system, especially
since passivating layers are well known to form between metal elec-
trodes and liquid electrolytes. Transition-metal oxides are among the
Partial Reduction Approach
A general introduction to the partial reduction approach for cre-
ating metal-metal oxide composite Li-storage materials is given
elsewhere.17 The main concepts are as follows. Partial reduction is a
* Electrochemical Society Student Member.
** Electrochemical Society Active Member.
z E-mail: ychiang@mit.edu
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1238
Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
Table I. Examples of mixed metal oxides that have been partially
reduced to form metal-ceramic composite anodes. Their struc-
ture, composition, and transport characteristics are relevant to
the microstructures obtainable through thermochemical
reduction.
Observed
characteristic and
Compound
SbVO4
Structure
Rutile
Diffusivity
interpretation
External reduction
as expected
Internal reduction
Ͼ
Dcations
Doxy
3ϩ Ͼ
Ͼ
Dv
Doxy
DSb5ϩ
Dcations
Distorted
fluorite
Perovskite
Sb2Mn2O7
AgVO3
External reduction
due to high Ag/V
ratio
Ͼ
1ϩ Ͼ
Dcations
DAg
Doxy
Dv5ϩ
Bronze
͑corner-shared
octahedra͒
Rutile
Internal reduction
Ͼ
Ͼ
Ag2V4O11
(Sn,Ti)O2
Dcations
Doxy
Dv5ϩ
1ϩ Ͼ
DAg
Internal reduction
due to phase
Dcations
Doxy
confinement
most attractive candidates for the host metal oxide MeIbIOy since
most are chemically compatible with organic liquid electrolytes and
also have high electronic and lithium ionic conductivity.
Since the electrochemical performance of such composites, espe-
cially upon extended cycling, depends critically on the microstruc-
ture that is obtained, within the general approach of partial reduc-
tion, we have also identified three specific strategies for obtaining
fine dispersions of the metal within an oxide matrix: ͑i͒ in oxides
with low oxygen diffusivity, the concentration and relative diffusiv-
ity of the lithium-active metal can be selected to obtain internal vs.
external precipitation of the reduced metal;10 ͑ii͒ systems of high
oxygen diffusivity, for which internal reduction is always expected,
can be selected; and ͑iii͒ phase separation can be conducted prior to
reduction, in order to confine the reduced metal phase. In this paper,
we illustrate these approaches using several examples.
Figure 1. XRD patterns of ͑a͒ SbVO4 partially reduced at 400°C for 4 h in
H2 and ͑b͒ Sb2Mn2O7 partially reduced at 360°C for 18 h in H2 .
potentially high storage capacities. In electrochemical tests, Ag
alloys with Li at a low voltage of ϳ0.02-0.1 V for low
concentrations;23 more highly lithiated alloys have not been ob-
tained. Both AgVO3 and Ag2V4O11 are compounds in which oxy-
gen diffusion is expected to be slow relative to cation diffusion, but
the lower Ag concentration in the latter should promote internal
reduction.10 AgVO3 is available commercially, and Ag2V4O11 is a
well-known cathode material for primary lithium cells used in medi-
cal implant applications.24 After partial reduction, the vanadium ox-
ide matrix is attractive due to its high lithium-ion and electronic
conductivity.
Selected Systems
Table I lists example systems we have investigated. Each system
has different structural, composition, and transport characteristics
which result in different microstructures and electrochemical
characteristics.
SbVO4 and Sb2Mn2O7.—Sb can alloy up to three Li per Sb atom
at a voltage plateau of 0.94-0.96 V19 relative to Li metal. Interest-
ingly, the molar volumes of Sb2O5 and Li3Sb are almost equal,
resulting in nearly zero volume change upon reducing the oxide to
Sb metal followed by lithiation to Li3Sb. We compare the starting
oxides SbVO4 and Sb2Mn2O7 because of their different crystal
structures and transport characteristics. SbVO4 has the rutile struc-
ture in which cation diffusion is expected to be more rapid than
oxygen diffusion; hence, the kinetics of partial reduction analyzed
by Schmalzried and Backhaus-Ricoult10 should apply. Due to the
high Sb fraction, external reduction may occur. This compound has
been widely used in catalytic applications.20 In contrast, Sb2Mn2O7
has the distorted fluorite structure21 in which oxygen diffusion
should be much more rapid than cation diffusion once some reduc-
tion occurs, resulting in oxygen vacancies. In this case, internal
reduction is expected as long as the matrix remains intact. Some
results on this system have already been reported in Ref. 17. Here
we present additional results showing how control of the starting
oxide particle size can be used to increase the fraction of internal
metal. Analyses of the composite morphology before and after cy-
cling have also been conducted.
Sn1ϪxTixO2.—(Sn,Ti)O2 was studied as an example of metal con-
finement due to phase separation. SnO2 lies well above TiO2 on the
Ellingham diagram and can be preferentially reduced. Sn is a well-
studied metal anode that can alloy up to 4.4 Li per Sn, while TiO2 is
known to have good ionic and electronic conductivity.25 While a
complete rutile solid solution of TiO2 and SnO2 can be achieved at
26
Ͼ
high temperatures ͑T
1430°C for the equimolar composition͒,
at lower temperatures there is a miscibility gap that is nearly sym-
metrical in composition. Within the spinodal regime, finely divided
lamellar microstructures with layers as fine as 20 nm thick, alterna-
tively rich in SnO2 and TiO2 , can be obtained.27,28 Two-phase mi-
crostructures of varying length scales can be created, within which
the SnO2-rich phase can be preferentially reduced to Sn metal. Dop-
ing influences both decomposition kinetics27,28 and electrical prop-
erties; most subvalent cations in TiO2 form deep acceptor levels
resulting in electronically insulating oxides at room temperature,
while supervalent cations tend to form shallow donor levels yielding
good electronic conductivity.25,29 Therefore, both microstructure and
electronic conductivity can be widely manipulated in this system
through control of composition and processing conditions.
AgVO3 and Ag2V4O11.—Lithium is reported to alloy with Ag to
form compounds LiAg, Li10Ag3 , and Li12Ag,22 thereby yielding
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
A1239
Figure 4. Voltage vs. capacity curve of ͑a͒ Sb-V2O3 produced by partial
reduction of SbVO4 ͑400°C for 4 h in H2͒, tested at 10 mA/g, and ͑b͒
Sb-MnO produced by partial reduction of Sb2Mn2O7 ͑360°C for 18 h in H2͒,
tested at 14 mA/g.
Experimental
Figure 2. SEM images of ͑a͒ partially reduced SbVO4 ͑400°C for 4 h in H2͒
and ͑b͒ partially reduced Sb2Mn2O7 ͑360°C for 18 h in H2͒. Coarse exter-
nally precipitated Sb is seen for the SbVO4 derived sample, while much finer
internal precipitates of Sb are found for Sb2Mn2O7 , in which oxygen diffu-
sion is much faster than cation diffusion.
SbVO4 , Sb2Mn2O7 , and Ag2V4O11 were prepared using solid-
state reaction processes beginning with reagent grade oxides or car-
bonates. SbVO4 was prepared by ballmilling a 1:1 molar ratio of
Sb2O3 ͑99.3% Alfa Aesar͒ with V2O5 ͑99.6% Aldrich͒ in a polypro-
pylene bottle using Teflon balls for 24 h. The mixture was then fired
Figure 3. SEM image of partially re-
duced Sb2Mn2O7 ͑420°C for 24 h in H2͒
with initially large starting Sb2Mn2O7
particle size and its cross-sectional back-
scattering SEM image, showing inter-
nally precipitated Sb metal.
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1240
Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
Sb2O3 ͑99.3% Alfa Aesar͒ and MnCO3 ͑99.9% Alfa Aesar͒, and
then calcining and reacting in air at 600°C for 7 h, 650°C for 4 h,
750°C for 5.5 h, and 1100°C for 21 h.
Ag2V4O11 powder was prepared by ballmilling a 1:2 molar mix-
ture of Ag2O ͑99% Alfa Aesar͒ and V2O5 ͑99.6% Aldrich͒. Single-
phase Ag2V4O11 was obtained after reacting the mixture in O2 at-
mosphere at 500°C for 10 h. The AgVO3 starting powder was
purchased from Aldrich ͑Aldrich Chemical Co., Milwaukee, WI͒.
Phase-separated SnO2-TiO2 composites were obtained by first
synthesizing a (Sn,Ti)O2 solid solution powder and then carrying
out thermal decomposition. The compositions studied had a 1:1 mo-
lar ratio of Sn:Ti, to which Al3ϩ ͑0.8 mol %͒ and Ta5ϩ ͑0.6-1.0 mol
%͒ were added as the acceptor and donor dopants, respectively, to
influence decomposition and electrical properties. The starting pow-
der was prepared via either a coprecipitation28 or a solid-state reac-
tion method. In the solid-state reaction method, stoichiometric
amounts of SnO2 ͑99.9% Alfa Aesar͒ and TiO2 ͑99.9% Alfa Aesar͒
doped with Al(OH)3 ͑Alfa Aesar͒ or Ta2O5 ͑99.993% Puratonic͒
were ballmilled as described previously, then heated in air to
1475°C,
a temperature above the miscibility gap maximum
͑1430°C͒, for 12 h to obtain a solid solution oxide. For starting
powder prepared by either method, decomposition heat-treatments
were subsequently carried out between 900 and 1000°C for 12-880 h
to allow phase separation. The rate of phase separation is highly
sensitive to composition and temperature in this system. X-ray dif-
fraction ͑XRD͒ was used to confirm that phase separation had oc-
curred prior to partial reduction in each instance.
For partial reduction, a small amount of each powder ͑ϳ0.4-2 g͒
was heated in an alumina boat ͑powder depth ϳ2 mm͒ in a tube
lined furnace ͑temperature and atmospheric conditions are given in
the Results and Discussion section for each material͒. Both buffered
CO/CO2 gas mixtures and unbuffered pure H2 gas at 1 atm total
pressure were used. For the CO/CO2 gas mixture, temperature and
oxygen activity can be selected to reduce only the more noble metal.
For hydrogen gas, although the specific oxygen partial pressure
could not be determined, we could carry out the reduction process at
low temperatures ͑Ͻ500°C͒ without the deposition of carbon ob-
served for CO/CO2 . XRD ͑Rigaku RU300, Cu K␣ radiation͒ and
weight loss were conducted for each sample before and after partial
reduction heat-treatments to determine the phases present and the
extent of reduction. For brevity, the detailed data are not included
here but are available in Ref. 30. JEOL 2000FX and 2010FX trans-
mission electron microscopes ͑TEM͒, a Fisons HB603 scanning
transmission electron microscope ͑STEM͒, a Philips XL30 environ-
mental scanning electron microscope, and a JEOL JSM-6320FV
field emission gun scanning electron microscope were used to char-
acterize the sample microstructures and identify phases.
Electrodes consisting of 70-88 wt % of the sample, 5-10 wt %
carbon additive ͑Super P, Chemetals Inc.; acetylene black, 99.9%,
Alfa Aesar; or graphite, 1-2 ␮m, Aldrich͒, and 10-20 wt % of poly-
vinylidene fluoride binder ͑Aldrich͒ were prepared. These were
mixed with ␥-butyrolactone solvent ͑98%, Alfa Aesar͒ and either
cast on a sheet of Cu foil, or dried and then pressed at 4000 kg/cm3
into 1/4 in. diam pellets. All electrodes were assembled in stainless
steel cells using Li foil ͑Alfa Aesar͒ as the reference electrode, and
ethylene carbonate:dimethyl carbonate ͑1:1 by weight͒ with 1 M
LiPF6 ͑EM Science͒ as the electrolyte, and Celgard 2400 as the
separator. Cycling was conducted at room temperature. The Super P
carbon additive made a significant contribution to the first-cycle
irreversible capacity loss and reversible capacity of the electrodes.
When formulated into electrodes and tested separately, up to 550
mAh/g of carbon and up to 80 mAh/g irreversible capacity was seen
on the first discharge at 0.7-1.3 V. The capacity data was corrected
by subtracting the capacity contribution of the added carbon from
the total electrode capacity.
Figure 5. Capacity vs. cycle number. ͑a͒ Sb-V2O3 sample with large exter-
nal Sb precipitates ͑0.5-1.5 V, 10 mA/g͒. Note the rapid fade in the first few
cycles. ͑b͒ Sb-MnO ͑sample from Fig. 2b͒ with finer external precipitates as
well as internally reduced Sb ͑0.7-1.3 V, 14 mA/g͒. ͑c͒ Sb-MnO ͑sample
from Fig. 3͒ with predominantly internal Sb particles ͑0.7-1.3 V, 6 mA/g͒.
in air at 700°C for 24 h, resulting in a single phase of SbVO4 .
Sb2Mn2O7 was prepared by ballmilling a 1:2 molar ratio mixture of
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
A1241
Figure 6. Cross-sectional backscattering SEM image of the Sb-MnO com-
posite after electrochemical cycling ͑50 cycles between 0.7 and 1.3 V͒. The
microstructure of Sb embedded in MnO is retained.
Results and Discussion
SbVO4 and Sb2Mn2O7.—Partial reduction was successfully car-
ried out at very low temperatures; complete reduction of the anti-
mony oxide fraction to Sb metal could be accomplished at tempera-
tures as low as 400°C ͑4 h, H2͒ for SbVO4 , and 360°C ͑18 h, H2͒
for Sb2Mn2O7 . Figure 1a and b ͑from Ref. 17͒ show XRD patterns
of the partially reduced SbVO4 and Sb2Mn2O7 , respectively. SEM
and TEM analyses showed a dramatic difference in Sb precipitate
size between the two. Reduced SbVO4 showed external Sb metal
Figure 8. ͑a͒ STEM image of internally precipitated Ag for Ag2V4O11 par-
tially reduced at 500°C for 4 h in H2 . ͑b͒ Cycling data for the partially
reduced Ag2V4O11 ͑20 mA/g͒.
precipitates of octahedral morphology as large as 20 ␮m mixed with
vanadium oxide phase, while the reduced Sb2Mn2O7 exhibited
much finer Sb particles, ranging from ϳ2 ␮m to ϳ20 nm as ob-
served by SEM and TEM analysis, mixed with MnO phase ͑Fig. 2a,
b͒.17 The finer precipitate size occurred despite a longer reduction
time for Sb2Mn2O7 . We believe that Sb metal precipitation occurs
internally in the distorted fluorite due to its high oxygen diffusivity,
causing Sb coarsening to be inhibited. In contrast, external reduction
appears to occur throughout for the slow oxygen diffusion system
SbVO4 , leading to coarse metal. For Sb2Mn2O7 , even though the
precipitated metal is primarily internal, a certain amount of Sb lying
within a characteristic diffusion length of the surface can diffuse to
the surface and form external precipitates of Sb metal.
To obtain samples with a finer and higher fraction of Sb internal
precipitates, the Sb2Mn2O7 particle size prior to partial reduction
was increased through thermal coarsening. This lowers the specific
surface area and is expected to decrease the fraction of external
precipitates. The as-synthesized Sb2Mn2O7 was further heat-treated
for various times and temperatures, after which the particle size was
examined using SEM. Here we present results for heat-treatment in
air at 1300°C for 8 h. A significant increase in the average particle
size from Ͻ1 ␮m in the starting material to 6.5 ␮m after coarsening
Figure 7. XRD patterns of ͑a͒ AgVO3 , partially reduced at 360°C for 6 h in
H2 , and ͑b͒ Ag2V4O11 , partially reduced at 500°C for 4 h in H2 .
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1242
Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
Figure 9. Bright-field TEM images of doped Sn0.5Ti0.5O2 . ͑a͒ 0.8 mol % Al-doped phase separated at 900°C, 12 h, and ͑b͒ partially reduced at 360°C, 15 h in
H2 . ͑c͒ 1 mol % Ta-doped composition phase separated at 1000°C, 325 h, and ͑d͒ partially reduced at 450°C, 24 h in H2 .
sembled and reassembled using a new lithium electrode and electro-
lyte. The charge ͑lithium extraction͒ capacities were then mostly
restored. Furthermore, the cumulative inefficiencies found in these
samples are much larger than the irreversible storage capacities pos-
sible from the materials in the electrode. The inefficiency also in-
creases again after the rebuilding of the cells ͑Fig. 5b, c͒. This in-
dicates that the coulombic inefficiency cannot be attributed to the
tested material alone but to some other side reaction due to the Li
metal electrode or the electrolyte, such as solid electrolyte inter-
phase ͑SEI͒ formation or electrolyte decomposition.
SEM examination of the microstructure of the internally reduced
Sb-MnO composite after electrochemical cycling reveals that the
initial morphology of Sb metal embedded in the MnO matrix was
retained ͑Fig. 6͒, although some fracturing of the matrix does occur.
We attribute the improved capacity retention upon cycling of this
material to the preservation of the internal reduction morphology,
which helps to maintain connectivity between the active metal and
the oxide matrix through cyclic volumetric expansion and contrac-
tion of the metal. In a separate study,31 we showed that Sb metal
powder cycled in the same electrolyte system undergoes systematic
disintegration into nanocrystalline metal particles. This is clearly
avoided when the Sb metal is contained within an oxide matrix, as
in the internally reduced Sb-MnO composites.
was observed. The coarsened Sb2Mn2O7 was then subjected to par-
tial reduction at 420°C for 24 h in H2 . SEM images of the coarsened
oxide and partially reduced Sb-MnO composite ͑Fig. 3͒ show that a
high fraction of Sb internal precipitates is obtained.
Figures 4a and b show typical voltage profiles for the Sb-V2O3
and Sb-MnO composites, respectively. The average voltage between
charge and discharge plateaus was ϳ0.97 V, and closer examination
revealed two closely spaced plateaus upon discharge, corresponding
to the formation of Li2Sb and Li3Sb, respectively.19 Upon cycling,
the coarse-grained Sb-V2O3 composite behaves like a mixture of the
two phases without a tailored microstructure. The composite has an
initial capacity of 415 mAh/g that is close to the theoretical value of
a 1:1 molar ratio of Sb:V2O3 ͑409 mAh/g͒, assuming V2O3 to be
inactive in this voltage range. However, rapid capacity fade in the
first several cycles was observed for the coarse Sb metal in this
sample ͑Fig. 5a͒. The Sb2Mn2O7-derived material with its finer
metal dispersion exhibits much better cyclability ͑Fig. 5b͒. The
sample with a still higher fraction of fine internal Sb precipitates,
shown in Fig. 3, exhibited even better capacity retention ͑Fig. 5c͒.
These cells did show lower coulombic efficiency and a surprising
feature of sudden fade after about 20 cycles ͑Fig. 5b, c͒. Further
investigation indicated that the sudden capacity fade was a cell fail-
ure and not an intrinsic property of the tested material. After the
sudden capacity fade of cell Sb-MnO no. 1 at the 20th to 25th cycle
and Sb-MnO no. 2 at the 20th to 22nd cycle, the cells were disas-
The theoretical charge capacity for Sb-MnO obtained by reduc-
ing Sb2Mn2O7 is 417 mAh/g ͑2570 mAh/cm3, assuming the theo-
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
retical density of 6.16 g/cm3͒. The two Sb2Mn2O7-derived samples
A1243
in Fig. 5b and c show reversible capacities of ϳ300 mAh/g ͑1850
mAh/cm3͒ and ϳ225 mAh/g ͑1385 mAh/cm3͒, respectively. In the
latter case, the reduced capacity is attributed to incomplete reduction
of the Sb oxide in the starting material. This was clearly evident in
the amount of weight loss and the presence of retained fluorite phase
in the reduced sample observed by XRD. ͑From the measured
weight loss before and after reduction, it was calculated that 76% of
the starting Sb-O was reduced to Sb metal, assuming that every 1
mol of the partially reduced Sb2Mn2O7 gives 2 mol of Sb metal and
2 mol of MnO upon reduction.͒ With further development, a revers-
ible gravimetric capacity of ϳ350 mAh/g corresponding to a volu-
metric capacity of ϳ2150 mAh/cm3 should be possible for this sys-
tem. The comparison between SbVO4 and Sb2Mn2O7 clearly
demonstrates the influence of the structure and transport properties
of the starting compound on the microstructure obtainable through
partial reduction, and subsequently on the electrochemical
properties.
AgVO3 and Ag2V4O11.—XRD patterns showing AgVO3 partially
reduced at 360°C for 6 h in H2 and Ag2V4O11 reduced at 500°C for
4 h in H2 are presented in Fig. 7. Only peaks for Ag metal are seen;
the vanadium oxide is present as an amorphous phase. Externally
precipitated Ag particles were observed using TEM and STEM for
the partially reduced AgVO3 , whereas the partially reduced
Ag2V4O11 exhibited fine internal Ag precipitates Ͻ20 nm thick
within the vanadium oxide ͑Fig. 8a͒. The occurrence of internal
reduction in the latter case may be due to the lower Ag concentra-
tion, as expected from Schmalzreid’s model.10 Cycling data for
Ag2V4O11 reduced at 500°C for 4 h in H2 are shown in Fig. 8b. The
reversible capacity of ϳ50 mAh/g is close to the theoretical capacity
of the composite if it is assumed that only the Ag alloys with Li, and
that alloying up to one Li per Ag atom occurs. This indicates that the
higher lithium alloys (Li10Ag3 ,Li12Ag) are not formed within the
range 0.005-2 V ͑to our knowledge, this is the first time this has
been demonstrated͒. A large first-discharge capacity is also observed
in this instance, which is probably due to electrochemical activity of
the highly disordered vanadium oxide phase between 1.75 and 3 V.
Ag2V4O11 is thus another example of a system with good cyclability
achieved due to internal reduction of the metal. Note that pure Ag
metal powder by itself does not cycle reversibly to the composition
LiAg.31 However, due to the low overall charge capacity of this
composite system, it was not pursued further.
(Sn0.5Ti0.5)O2.—Figure 9 shows TEM images of phase-separated
microstructures in the Sn0.5Ti0.5O2 system, showing the SnO2-rich
͑dark͒ and TiO2-rich ͑light͒ lamellae. The Al3ϩ doped sample has
lamellar spacing of 10-20 nm after being decomposed at 900°C for
12 h in air, while the Ta5ϩ doped sample has a finer spacing of Ͻ10
nm despite being decomposed at a higher temperature and longer
time of 1000°C for 325 h in air. This is consistent with doping
effects seen in previous studies of this material.26-28 Figure 9 also
shows corresponding images after partial reduction ͑at 360°C for 15
h in H2 for Al-doped samples, and 450°C for 4 h in H2 for Ta-doped
samples͒. Much of the reduced Sn metal is captured between
TiO2-rich platelets, at a particle size of ϳ10 nm. However, a signifi-
cant fraction of the reduced Sn also formed external spherical par-
ticles with a distribution of sizes up to 0.5 ␮m diam. Unlike the
cases of Sb and Ag, partial reduction was carried out above the
melting point of Sn metal ͑232°C͒. The coarser Sn particles may
have coalesced from smaller droplets or coarsened through vapor
phase transport. Thus these composites contain a distribution of Sn
particle sizes ranging from ϳ10 nm to ϳ0.5 ␮m, with a larger
average size occurring in the Al-doped samples. Based on the
weight loss during partial reduction, it was determined that only
33% of the total SnO2 had been reduced to Sn metal for the Al-
doped sample ͑partially reduced at 360°C for 15 h in H2͒, and 56%
for the Ta-doped sample ͑partially reduced at 450°C for 4 h in H2͒.
Figure 10. Comparison of the cycling behavior of ͑a͒ phase-separated un-
reduced, ͑b͒ Al3ϩ doped partially reduced, and ͑c͒ Ta5ϩ doped partially
reduced (Sn0.5 ,Ti0.5)O2 . Rapid capacity fade occurred in the unreduced and
reduced Al3ϩ doped with coarse Sn precipitates, while the reduced Ta5ϩ
doped partially reduced sample exhibits better cyclablity due to finer Sn
microstructure and lower polarization from better conductivity.
This may be due to slow reduction kinetics at the temperatures used,
as well as retention of ϳ10% SnO2 in solid solution in the TiO2 , as
indicated by the equilibrium phase diagram.27
Figure 10 compares the electrochemical behavior of the unre-
duced Al-doped sample, its partially reduced Sn-TiO2 counterpart,
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1244
Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
Figure 11. Cycle vs. capacity of phase-separated unreduced, Al3ϩ doped
partially reduced, and Ta5ϩ doped partially reduced (Sn0.5 ,Ti0.5)O2 . The
best cyclability is found in the Ta5ϩ doped sample with fine Sn particles.
Figure 12. Extended cycling of Ta-doped Sn-TiO2 . Capacity drops from
200 mAh/g to 100 mAh/g at approximately 12th cycle due to coarse Sn
particles with fine Sn particles continuing to cycle reversibly.
and the partially reduced Ta-doped Sn-TiO2 composite. The unre-
duced (Sn,Ti)O2 exhibits the largest first-cycle irreversibility ͑78%͒.
Improvement in the first-cycle irreversibility was observed for both
of the partially reduced samples ͑64% for Al-doped and 55% for
Ta-doped samples͒. We interpret most of the irreversibility as being
due to the electrochemical reduction of the residual SnO2 to Sn. SEI
formation probably also contributes to the irreversible capacities ob-
served during the first cycle since the capacity loss is much larger
than the unreduced SnO2 alone can produce. The larger plateau at
ϳ1 V in the unreduced sample ͑Fig. 10a͒ is consistent with the
reduction of SnO2 by Li,6 and the charge capacities were lower than
the theoretical value of 594 mAh/g ͑3360 mAh/cm3͒ expected for
Sn0.5Ti0.5O2 , which has been fully reduced to a 1:1 molar ratio of
Sn:TiO2 .
externally precipitated metal. In these materials, improved cycling is
attributed to both a reduction in metal particle size and the passiva-
tion of the metal particles by the surrounding oxide matrix. While
first-cycle irreversibility due to the presence of residual lithium-
active oxide phases must be improved, reversible gravimetric ca-
pacities up to 350 mAh/g and volumetric capacities up to 2000
mAh/cm3 have been demonstrated in the systems studied to date.
From the processing viewpoint, this approach is attractive since in-
expensive starting materials and simple, scalable, thermochemical
processes can be used.
Acknowledgment
For the Al-doped composition, both the starting oxide and the
partially reduced composite exhibit poor cyclability, as seen in Fig.
10a and b. The rapid capacity fade is attributed to most of the re-
duced Sn being in the form of coarse particles. Note also that a large
polarization develops during the first few cycles. In comparison, the
Ta-doped sample exhibits lower first-cycle irreversibility, better ca-
pacity retention, and lower polarization ͑Fig. 10c͒. The lower polar-
ization compared to Fig. 10b is attributed to the higher electronic
conductivity of donor-doped TiO2 .25,29 Upon narrowing the voltage
range of the Ta-doped sample to 0.3-1.2 V, a decrease in capacity
from 300 to 200 mAh/g ͑1000 mAh/cm3͒ was seen, but the cycla-
bility further improved ͑Fig. 11͒. After about 50 cycles, the capacity
stabilized at ϳ100 mAh/g ͑Fig. 12͒. The earlier cycles appear to
cause the systematic loss of capacity from larger Sn particulates,
while the finer particles in the particle size distribution continue to
cycle reversibly.
This example demonstrates the concept of partial reduction of a
microphase-separated mixture of oxides to produce an electrochemi-
cally active composite. While the storage properties for the Sn-Ti-O
system are not yet sufficient for practical use, we believe that further
development of this concept in related systems can narrow the size
distribution of active metal particles and result in improved
properties.
Funding and instrumentation in the Shared Experimental Facili-
ties at MIT were supported by NSF grant no. 9400334-DMR.
The Massachusetts Institute of Technology assisted in meeting the publi-
cation costs of this article.
References
1. D. Fauteux and P. Koksbang, J. Appl. Electrochem., 23, 1 ͑1993͒.
2. R. A. Huggins, J. Power Sources, 26, 109 ͑1989͒.
3. J. Yang, M. Winter, and J. O. Besenhard, Solid State Ionics, 90, 281 ͑1996͒.
4. Y. Idota, T. Kubota, A. Matsufuji, Y. Maekawa, and T. Miyasaka, Science, 276,
1395 ͑1997͒.
5. I. Courtney and J. Dahn, J. Electrochem. Soc., 144, 2045 ͑1997͒.
6. O. Mao, R. L. Turner, I. A. Courtney, B. D. Fredericsen, M. I. Buckett, L. J.
Krause, and J. R. Dahn, Electrochem. Solid-State Lett., 2, 3 ͑1999͒.
7. G. M. Ehrlich, C. Durand, X. Chen, T. A. Hugener, F. Spiess, and S. L. Suib, J.
Electrochem. Soc., 147, 886 ͑2000͒.
8. J. Yang, Y. Takeda, N. Imanishi, and O. Yamamoto, J. Electrochem. Soc., 146, 4009
͑2000͒.
9. I. Courtney and J. R. Dahn, J. Electrochem. Soc., 144, 2942 ͑1997͒.
10. H. Schmalzried and M. Backhaus-Ricoult, Prog. Solid State Chem., 22, 1 ͑1993͒.
11. J. Narayan and Y. Chen, Philos. Mag. A, 49, 475 ͑1984͒.
12. K. M. Ostyn, C. B. Carter, M. Koehne, H. Falke, and H. Schmalzried, J. Am.
Ceram. Soc., 67, 679 ͑1984͒.
¨
¨
13. R. Subramanian, E. Ustundag, S. L. Sass, and R. Dieckmann, Mater. Sci. Eng., A,
195, 51 ͑1995͒.
¨
¨
14. E. Ustundag, R. Subramanian, R. Dieckmann, and S. L. Sass, Acta Metall. Mater.,
43, 383 ͑1995͒.
Conclusions
15. J. A. Smith, P. Limthongkul, L. Hartsuyker, S. Y. Kim, and S. L. Sass, J. Appl.
Phys., 83, 2719 ͑1998͒.
The partial reduction of mixed oxides has been studied in several
systems as a new approach to synthesizing electrochemically active
metal-ceramic composites. By choosing oxide compounds with suit-
able crystal structures and active cation concentrations, fine internal
dispersions of lithium-active metals can be obtained which show
markedly improved resistance to cycling fade compared to coarser,
16. D. L. Ricoult and H. Schmalzried, J. Mater. Sci., 22, 2257 ͑1987͒.
17. P. Limthongkul, H. Wang, and Y.-M. Chiang, Chem. Mater., 13, 2397 ͑2001͒.
18. R. A. Huggins and W. D. Nix, Ionics, 5, 57 ͑2000͒.
19. J. Wang, I. D. Raistrick, and R. A. Huggins, J. Electrochem. Soc., 133, 457 ͑1986͒.
20. G. Centi, F. Marchi, and S. Perathoner, Appl. Catal., A, 149, 225 ͑1997͒.
21. H. G. Scott, J. Solid State Chem., 66, 171 ͑1987͒.
22. P. M Hansen, Constitution of Binary Alloys, McGraw-Hill, New York ͑1958͒.
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑9͒ A1237-A1245 ͑2002͒
A1245
23. J. O. Besenhard, J. Yang, and M. Winter, J. Power Sources, 87, 68 ͑1997͒.
24. C. A. Vincent and B. Scrosati, Modern Batteries, p. 121, John Wiley & Sons, New
York ͑1997͒.
28. H. Maruyama, Y.-M. Chiang, M. Kawamoto, Y. Seki, S. Arakawa, K. Kikuta, and
S. Hirano, Ceramic Transactions, Vol. 24, T. O. Mason and J. L. Routbort, Editors,
p. 367, The American Ceramic Society, Westerville, OH ͑1991͒.
25. J. F. Baumard and E. Tani, J. Chem. Phys., 67, 857 ͑1977͒.
26. R. S. Roth, T. Negas, and L. P. Cook, Phase Diagrams for Ceramists, Vol. 4, G.
Smith, Editor, The American Ceramic Society, Westerville, OH ͑1981͒.
27. T. C. Yuan and A. V. Virkar, J. Am. Ceram. Soc., 71, 12 ͑1988͒.
29. J. A. S. Ikeda and Y.-M. Chiang, J. Am. Ceram. Soc., 76, 2437 ͑1993͒.
30. P. Limthongkul, Ph.D. Thesis, MIT, Cambridge, MA ͑2002͒.
31. P. Limthongkul, E. Jud, and Y.-M. Chiang, Paper 255 presented at The Electro-
chemical Society Meeting, San Francisco, CA, Sept 2-7, 2001.
Downloaded on 2015-03-13 to IP 129.93.24.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).