Reactivity of antimony oxides and MSb2O6 (M = Cu, Ni, Co), trirutile-type phases with metallic lithium

Article abstract of DOI:10.1149/1.2219711

Isostrutural MSb₂O₆ (M = Cu, Ni, Co) trirutile-type phases were prepared by solid-state synthesis and their electrochemical activity towards lithium was investigated. NiSb₂O₆ and CoSb ₂O₆ showed similar electrochemical behavior, with an uptake of about 18-19 Li per formula unit along the first reduction and only 6-7 Li reversibly removed upon subsequent cycling. This totally differs from the behavior of CuSb₂O₆ that was found to react with only about 7 Li during the first reduction, without any capacity recovered on subsequent charge. From X-ray diffraction data, both Cu and Co phases lead to an amorphous composite down to 0 V. From high-resolution transmission electron microscopy observations, reduced CoSb₂O₆ consists of 10-50 nm Co particles dispersed in a Li₃Sb + Li₂O matrix. For CuSb₂O₆. our results and observations enlightened a two-step reduction: First, the formation of Cu clusters through an electrochemically-driven exchange reaction leading to Li₂Sb ₂^vO₆, and then reduction of this matrix into 5 nm Sb domains dispersed in an insulating amorphous Li-rich Li-Sb^v-O matrix, preventing any further alloying reaction and any charge reaction. The complete charge irreversibility observed on Li/Sb₂O₅ half-cells confirmed this point, while the T.i reduction of Sb₂O ₃ emphasized a close similarity with the reactivity of the (Ni/Co)Sb₂O₆ phases, suggesting a first reduction step of Sb⁵⁺ into Sb³⁺, assuring conduction and subsequent cycling capacity. Despite the composite nature of the as-formed electrode and the very fine Sb particles, the reversibility of the alloying reaction was not found to be satisfactory for Li-ion cells.

Full text of DOI:10.1149/1.2219711

A1778
Journal of The Electrochemical Society, 153 ͑9͒ A1778-A1787 ͑2006͒
0
013-4651/2006/153͑9͒/A1778/10/$20.00 © The Electrochemical Society
Reactivity of Antimony Oxides and MSb O „M = Cu,Ni,Co…,
2
6
Trirutile-type Phases with Metallic Lithium
a,z
a
a
b
b
D. Larcher, A. S. Prakash, L. Laffont, M. Womes, J. C. Jumas,
b
c
a,
J. Olivier-Fourcade, M. S. Hedge, and J.-M. Tarascon *
a
Laboratoire de Réactivité et Chimie des Solides, Université de Picardie Jules Verne, 80039 Amiens,
France
b
Laboratoire des Agrégats Moléculaires et Matériaux Inorganiques, Université Montpellier II, France
c
Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore, India, 560 012
Isostrutural MSb O ͑M = Cu,Ni,Co͒ trirutile-type phases were prepared by solid-state synthesis and their electrochemical activ-
2
6
ity towards lithium was investigated. NiSb O and CoSb O showed similar electrochemical behavior, with an uptake of about
2
6
2
6
1
8–19 Li per formula unit along the first reduction and only 6–7 Li reversibly removed upon subsequent cycling. This totally
differs from the behavior of CuSb O that was found to react with only about 7 Li during the first reduction, without any capacity
2
6
recovered on subsequent charge. From X-ray diffraction data, both Cu and Co phases lead to an amorphous composite down to
V. From high-resolution transmission electron microscopy observations, reduced CoSb O consists of 10–50 nm Co particles
0
2
6
dispersed in a Li Sb + Li O matrix. For CuSb O , our results and observations enlightened a two-step reduction: First, the
3
2
2
6
V
2
formation of Cu clusters through an electrochemically-driven exchange reaction leading to Li Sb O , and then reduction of this
matrix into 5 nm Sb domains dispersed in an insulating amorphous Li-rich Li–Sb –O matrix, preventing any further alloying
reaction and any charge reaction. The complete charge irreversibility observed on Li/Sb O half-cells confirmed this point, while
2
6
V
2
5
the Li reduction of Sb O emphasized a close similarity with the reactivity of the ͑Ni/Co͒Sb O phases, suggesting a first
2
3
2
6
reduction step of Sb5+ into Sb , assuring conduction and subsequent cycling capacity. Despite the composite nature of the
as-formed electrode and the very fine Sb particles, the reversibility of the alloying reaction was not found to be satisfactory for
Li-ion cells.
3+
©
2006 The Electrochemical Society. ͓DOI: 10.1149/1.2219711͔ All rights reserved.
Manuscript received March 14, 2006. Available electronically July 21, 2006.
Rechargeable lithium-ion batteries, first introduced in 1990, are
presently powering most of the portable electronic devices. In the
continuing endeavor to improve these systems, researchers have
shared their efforts between optimizing electrode materials reacting
with Li through the classical Li insertion/deinsertion mechanism and
searching for electrode materials displaying new reversible Li reac-
tivity mechanisms. Pursuing the latter path, it was recently shown
that, when working at the nanometer scale, Li electrochemically
driven conversion reactions ͑oxides ꢀ metals͒ can be achieved and
maintained for hundreds of cycles. Moreover, depending on the oxi-
investigation tools are also applied to Li Sb O ternary phase as
2 2 6
+
V
+III
well as to binary oxides containing either Sb ͑Sb O ͒, Sb ͑Se-
2
5
+
V
+III
narmontite, Sb O ͒, or a mixture of Sb and Sb
Sb O ͒. That way, we propose reaction mechanisms aimed at un-
͑Cervantite,
2
3
1
0
2
4
derstanding the drastic differences in electrochemical responses for
CuSb O and Li–͑Sb+ –O͒ phases on the one hand, and for
V
2
6
͑Co/Ni͒Sb O and trivalent binary oxides on the other hand.
2
6
Experimental
Samples syntheses.— MSb O ͑M = Cu, Ni, and Co͒ metal an-
2
6
dation state of the 3d-metal, these conversion reactions can either
1
timonates were synthesized by solid-state reaction from a stoichio-
metric mixture of MO ͑M = Cu, Ni, or Co͒ and Sb2O5 powders
involve 1 ͑Cu O͒, 2 ͑FeO, NiO, CoO, CuO͒, 2.66 ͑Co O ͒,
3
2
3
4
2
3
͑
Fe O ͒, or 4 ͑RuO ͒ electrons, thus offering the possibility of
2 3 2
͑
Aldrich, purities Ͼ99.9%͒. The mixtures were air-heated in an alu-
having compounds with large capacities as compared to only 0.5 Li
per metal in today’s manufactured intercalation materials. Soon af-
ter, many other sightings of conversion reactions processes in
mina crucible at 800°C for 2–3 days with intermediate grindings.
Afterwards, the sample was cooled down to room temperature and
ground to obtain fine powder. Phase purity and composition of pur-
chased Sb O ͑Senarmontite͒ and Sb ͑Fluka, purity Ͼ99.5%͒ were
4
5
6
7
nitride, fluoride, phosphide, or sulfide compounds were reported
2
3
+
to occur at potentials ranging from 0 to 3.0 V vs Li /Li°, depending
checked by XRD and Mössbauer spectroscopy, respectively. The
on the nature of the anion. Along that line, we decided to study the
electrochemical properties of compounds containing metal elements
with even higher oxidation states, and therefore decided to focus on
batch of Sb O was found badly crystallized but thermal analysis
2
5
͑
TGA͒ coupled with Mössbauer spectroscopy ascertained the for-
mula Sb O ·0.4 H O.
2
5
2
V
Sb -based oxides and more specifically on the trirutile-type com-
Powdered Li Sb O was synthesized by reacting Li CO ͑Ald-
2
2
3
2
3
pounds of general formula M+ Sb2 O , whose structure derives
II +V
6
rich, Ͼ99.9%͒ with Sb O under static air at 750°C for 24 h with a
2 3
8
from the rutile lattice by tripling the c-axis. Besides their ability to
heating rate of 5°/min. A slight excess of Li salt was necessary to
compensate heating-induced loss.
exist under high oxidation state, our choice for Sb is also nested in
9
its ability to alloy with Li leading to Li Sb, therefore potentially
Partially reduced ␣-Sb O ͑Cervantite͒ was synthesized by one-
3
2
4
offering an uptake of 8 Li per Sb+ via both conversion and alloy-
V
step heating of a batch of Sb O ·0.4 H O under static air at a rate of
2
5
2
11
ing.
5°C/min up to 800°C.
In the present study, we prepare cobalt, copper, and nickel-
containing trirutile materials and compare their respective reactivity
and reversibility through electrochemical titration, Mössbauer effect
Characterization.— Routine phase composition of powders was
achieved by XRD using a Philips diffractometer ͑PW 1710, Cu K␣
radiation, ␭ = 1.5418 Å͒ equipped with back monochromator. Lat-
tice parameters were refined using the Fullprof refinement
͑
ME͒ and electron dispersive X-ray ͑EDXS͒ spectroscopies, com-
bined with ex situ and in situ X-ray diffraction ͑XRD͒ and high-
resolution transmission electron microscopy ͑HRTEM͒. For com-
parison purposes and for the sake of references, the same set of
1
2
software.
The electrochemical reactivity of MSb O ͑M = Cu,Ni,Co͒, Sb
2
6
oxides and Sb, toward Li was performed in Swagelok cells as-
sembled in an argon-filled MBraun MB150B-G dry box. The posi-
tive electrode composite, consisting of 85 wt % of active material
and 15 wt % of conducting SP carbon, was separated from the nega-
*
Electrochemical Society Active Member.
Corresponding author: Dominique.Larcher@sc.u-picardie.fr
z
Downloaded on 2015-02-10 to IP 129.97.124.140 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, 153 ͑9͒ A1778-A1787 ͑2006͒
tive electrode ͑lithium foil͒ by two sheets of Whatman GF/D boro-
A1779
silicate fiber glass, the assembly being soaked by a 1 M solution of
LiPF in an ethylene carbonate ͑EC͒/dimethyl carbonate ͑DMC͒ sol-
6
vent mixture ͑1:1 v/v͒. Typically, the cells were cycled at 20°C in a
+
galvanostatic mode between 0 and 2–3 V vs Li /Li° using a Mac–
Pile automatic cycling data recording system ͑Biologic SA Claix,
France͒ and at rates ranging from 1 Li in 5 h ͑noted C/5͒ to 1 Li in
2
0 h ͑noted C/20͒. Galvanostatic intermittent titration technique
͑
GITT͒ experiment was also performed to determine the quasiequi-
librium open-circuit potential at various reaction levels. The open-
circuit voltage ͑OCV͒ was determined once the voltage evolution
remained under 3 mV/h, after reduction steps of 2 h at a C/10 rate
͑
i.e., 0.2 Li per step͒. Based on the high gravimetric capacities re-
leased by the tested samples, the contribution of the added SP car-
bon ͑60 mAh/g͒ onto the cells capacities is neglected.
The structural evolution of the electrode material during electro-
chemical reactions was monitored using in situ XRD measurements.
To carry out such an experiment, a home-designed XRD electro-
1
3
chemical cell was equipped with a Be window being transparent to
the X-rays while acting as a current collector. This cell was then
placed in a Scintag diffractometer ͑Cu K␣, ␭ = 1.5418 Å͒, and
XRD scans were collected as cell is cycled at a constant current rate
as low as 1 Li in 20 h.
Mössbauer and HRTEM ex situ measurements were performed
on the electrodes at several depths of discharge and charge. In order
to remove adsorbed species and soaking residual electrolyte, the
Swagelok cells were opened inside the glove box and the powders
washed several times with salt-free solvent and separated from the
liquid by centrifugation.
The 121Sb Mössbauer spectra were obtained with a classical EG
&
G constant acceleration spectrometer in transmission mode. The
121m
source of nominal activity 0.5 mCi was
Sn in a BaSnO matrix.
3
During the measurements, both source and absorbers were simulta-
neously cooled down to 4 K in order to increase the fraction of
recoil-free absorption and emission processes. The active material
was placed on a specific sample holder transparent to the ␥-rays.
The velocity scale was calibrated with the standard spectrum of an
iron absorber using a 57Co͑Rh͒ source. The zero isomer shift was
defined from the spectrum of InSb at 4 K ͓␦ = −8.70͑4͒ mm/s rela-
Figure 1. XRD patterns for ͑a͒ CuSb O , ͑b͒ CoSb O , and ͑c͒ NiSb O .
2
6
2
6
2
6
These phases having the same structure/space group, Bragg indexation is
only displayed for CoSb O .
2
6
The voltage-composition curves for the first galvanostatic cycles
of MSb2O6/Li cells are shown in Fig. 2a.
CoSb2O6 and NiSb2O6 show very similar behavior with a large
uptake of about 18–19 Li per formula unit ͑ca. 1200 mAh/g͒
through two distinct pseudovoltage plateaus. These large numbers of
Li are consistent with the full reduction of the cationic species plus
tive to the Ba1 SnO3 source͔. The data was processed using a
21m
14
transmission integral analysis software GM5SIT with the Möss-
bauer source fraction ͑fs͒ and source line width ͑⌫s͒ held constant at
0
.62 and 1.45 mm/s, respectively.
TEM and HRTEM imaging were performed using a FEI TEC-
the alloying reaction of Sb with Li leading to Li Sb, as is later
NAI F20 S-TWIN equipped with an EDXS detector. Ground
samples were deposited on copper or nickel grids coated with lacey-
carbon films. For sensitive-reduced materials, these preparations
were done within the argon-filled dry box. The diffraction patterns
were obtained in selected area diffraction ͑SAED͒ mode or by Fou-
rier transform of the HRTEM imaging. The SAED patterns obtained
on polycrystalline samples were analyzed using the process diffrac-
3
confirmed by HRTEM, and summarized as
͑
Ni/Co͒Sb O + 18Li → 2Li Sb + Co/Ni + 6Li O
͓1͔
2
6
3
2
The corresponding derivatives ␦x/␦V curves ͑Fig. 2b͒ exhibit a
well-defined peak located at 1.5 V vs Li /Li° ͑6–7 Li per formula
+
unit͒, while the second one ͑about 12.5 Li per formula unit͒ is much
1
5
+
tion software.
broader and spread from 1.1 to 0.5 V vs Li /Li°. The length of the
2
+
first voltage plateau rules out the sole reduction of the M cation ͑2
Li͒ and it seems difficult to argue a contribution of Li–Sb alloying
reaction occurring before the oxide decomposition. Going on, the
length of the low-voltage feature ͑12.5 Li͒ cannot match the Sb
Results and discussions
The XRD patterns of CuSb O , CoSb O , and NiSb O powders
2
6
2
6
2
6
→
Li Sb alloying reaction alone ͑6 Li͒ even be associated with the
3
͑
Fig. 1a-c͒ can be indexed on the basis of a tetragonal crystal cell. In
contrast to ͑Co/Ni͒Sb O6 that crystallized in the P42/nmn space
2
group, CuSb O is reported to adopt a lower symmetric P2 /n space
2
6
1
group at ambient temperature, as a consequence of the octahedral
Table I. Crystallographic parameters for NiSb2O6, CoSb2O6 and
CuSb O .
distortion due to the presence of Cu2+ ͑d ͒.
perature for the monoclinic → tetragonal transition is still the sub-
9
8,16
However, the tem-
2
6
17,18
Lattice parameters
ject of some controversies,
and our XRD pattern is found to
match the tetragonal lattice. Besides the peaks due to this trirutile
structure, only a small amount of NiO impurity could be spotted
Sample
a ͑Å͒
c ͑Å͒
͑
Fig. 1c͒. The lattice parameters were refined with the Fullprof soft-
NiSb O
4.657͑2͒
4.654͑3͒
4.647͑6͒
9.189͑4͒
9.283͑5͒
9.29͑1͒
2
6
ware ͑Table I͒ and were found to be in good agreement with those
reported in the literature.
CoSb O
2
6
6
8,17-21
CuSb O
2
Downloaded on 2015-02-10 to IP 129.97.124.140 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1780
Journal of The Electrochemical Society, 153 ͑9͒ A1778-A1787 ͑2006͒
Figure 2. First galvanostatic cycles for various Sb-based materials vs metallic lithium, recorded at 20°C and at a constant current of 1 Li/20 h. ͑left plots͒
Voltage-composition curves and ͑right plots͒ associated ␦x/␦V derivatives curves for ͑a,b͒ trirutile samples, ͑c,d͒ Sb, ͑e,f͒ Li Sb O , and ͑g,h͒ Sb O and Sb O .
2
2
6
2
5
2
3
reduction of the M2+ species ͑8 Li͒. Therefore, the respective ca-
pacities observed on these two plateaus cannot be trivially attributed
to simple and separated reactions, and the oxide reduction is obvi-
ously contributing to both plateaus.
͑iv͒ The high voltage charge reaction ͑1.4 V͒ is therefore corre-
sponding to a small fraction of the reaction occurring aside from the
alloying process along the low voltage discharge plateau, either a
partial re-oxidation of M° and/or back formation of Sb-based oxides.
͑v͒ Based on the total charge capacity ͑7 Li͒ and on the respec-
tive contributions of the 1.1 and 1.4 V reactions, the de-alloying
reaction is not totally reversed.
On charge, both CoSb O and NiSb O materials also behave in
2
6
2
6
a very comparable way with about 7 Li extracted per formula unit
through three distinct voltage features: a very small signal at around
0.9 V at the early stage of the process and not identified so far, one
For CuSb O , the situation is drastically different ͑Fig. 2a and b͒.
2
6
broad and intense at 1.1 V, and finally a much broader signal of
lower intensity centred at around 1.4 V. From these electrochemical
data and comparison with those of a Li/Sb cell ͑Fig. 2c and d͒,
several comments and observations can be made.
It reacts with only 6–7 Li on first discharge through a well-defined
initial plateau at 2.2 V vs Li /Li° corresponding to 2 Li per
+
CuSb O , followed by a voltage decrease down to 0 V mainly as-
2
6
sociated with two broad derivative signals at 1.2 and 0.9 V. The
subsequent charge surprisingly shows a complete irreversibility of
the reaction.
͑
i͒ The total absence of charge capacity corresponding to the first
discharge reaction ͑1.5 V͒ indicates its total irreversibility.
ii͒ The voltage of the most intense charge signal ͑1.1 V͒
͑
In order to grasp additional data in view of definitive identifica-
tion of the involved mechanisms, we performed in situ XRD experi-
ments. The evolutions in XRD patterns collected during the dis-
charge of CoSb O /Li ͑Fig. 3͒ and CuSb O /Li ͑Fig. 4͒ cells
matches that of the Li Sb → Sb dealloying reaction ͑see Fig. 2c and
3
d͒.
͑
iii͒ The total capacity associated to the Sb → Li Sb reaction
3
2
6
2
6
͑
0.8 V, Fig. 2c and d͒ is contributing to the low voltage discharge
revealed a complete and progressive amorphization of the materials
plateau.
finally leading to featureless pattern at 0 V. For CoSb O , the oxide
2
6
Downloaded on 2015-02-10 to IP 129.97.124.140 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, 153 ͑9͒ A1778-A1787 ͑2006͒
A1781
Figure 3. In situ XRD patterns collected
on a Li/CoSb O cell discharged in a gal-
2
6
vanostatic mode at a C/20 rate and at
0°C.
2
reflections are still vanishing beyond the first plateau ͑7 Li͒, con-
firming the contribution of the oxide decomposition onto both pla-
teaus. When one looks carefully at specific depths of discharge,
important differences are spotted ͑Fig. 5͒. For instance, at x = 2, the
tion processes by means of HRTEM. Bright-field images ͑Fig. 6͒
indicate that CoSb O and CuSb O initial materials consist of
2
6
2
6
1
00–200 nm and agglomerated 10–50 nm well-crystallized par-
ticles, respectively, as deduced from the sharp spots displayed on
their corresponding SAED patterns. Note that the smaller size of the
CuSb O crystallized domains agrees with its broader XRD reflec-
reflections due to CuSb O6 are replaced by mainly two slightly
2
shifted and broader peaks. At the same level of reduction, only a
2
6
slight decrease in reflections intensities is observed for CoSb O
2
6
tions ͑Fig. 1͒. After full reduction, CoSb O initial particles trans-
2
6
without any new peak growth or shift. Upon charge after full dis-
charge, XRD patterns are not noticeably changed and only a small
amount of unreacted CoSb O is detected.
formed into a composite material made of 10–50 nm fairly crystal-
lized Co nanoparticles ͑Fig. 7a and d͒ in a matrix of Li O and Li Sb,
2
3
2
6
Due to the incapacity for XRD experiments to provide further
information about the structural/textural aspects of the electrode ma-
terials during discharge, besides its apparent amorphization, we
studied both CoSb O and CuSb O at various stages of their reduc-
as deduced from the line profile of the SAED pattern of the bright-
field image ͑Fig. 7b and c͒. These observations are in complete
agreement with the overall discharge capacity of about 18–19
2
6
2
6
Figure 4. In situ XRD patterns collected
on a Li/CuSb O cell discharged in a gal-
2
6
vanostatic mode at a C/20 rate and at
0°C.
2
Downloaded on 2015-02-10 to IP 129.97.124.140 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1782
Journal of The Electrochemical Society, 153 ͑9͒ A1778-A1787 ͑2006͒
Figure 5. Selected in situ XRD patterns
collected along the first galvanostatic
cycle for ͑left panel͒ Li/CuSb O and
2
6
͑
right panel͒ Li/CoSb O cells, after reac-
2 6
tion of 2 Li per formula unit, full dis-
charge down to 0.0 V, and after full
+
charge at 3.0 V vs Li /Li°. In the bottom
panels, XRD patterns of the starting mate-
rials are displayed for references.
Li/CoSb O . Similarly to observations made on fully discharged
CoO/Li cells, the composite agglomerates are surrounded by a
thin polymeric-like coating ͑Fig. 7d͒.
that of the parent CuSb O material with only some slight modifi-
cations in cell parameters ͑Fig. 9c͒, suggesting the following ex-
change reaction
2
6
2
6
2
2
Similar HRTEM experiments were conducted on fully reduced
CuSb O ͑Fig. 8͒. Bright-field images show the initial particles have
transformed into a composite consisting of 5 nm Cu and Sb nano-
particules embedded in an amorphous matrix, as deduced from in-
V
V
2
6
CuSb O + 2Li → Li Sb O + Cu
͓2͔
2
6
2
2
6
Interestingly, the formation of ilmenite Li2Sb2O6 through cat-
ionic exchange reaction has already been demonstrated. This in-
termediate formation of a Li2Sb O6 matrix was confirmed by com-
2
3
dexation of SAED patterns. Surprisingly, no trace of Li O and/or
2
V
2
Li–Sb alloys could be detected. These points were confirmed by
heating this discharged material at 350–500°C under inert atmo-
paring the electrochemical behavior of CuSb2O6 ͑Fig. 2a and b͒
with that of a homemade Li2Sb2O6 powder ͑Fig. 2e and f͒. Only 4–5
Li per formula unit can be incorporated in this material, while an
uptake of 16 Li is expected for its full reduction into a mixture of
Li2O and Li3Sb. In addition, this small capacity is released through
two features corresponding to two broad derivative peaks perfectly
matching those observed for a Li/CuSb2O6 cell. Finally, the total
amount of 6–7 Li reacting per CuSb2O6 is in good agreement with
sphere, leading to a mixture of crystallized Cu Sb and Sb together
2
with a small amount of Li CO likely resulting from the decompo-
2
3
sition of residual salts, without any evidence for Li–Sb alloys. Even
at 500°C, no crystallization of the matrix occurred.
According to the evolution of the XRD patterns and the peculiar
behavior at x = 2, HRTEM observations coupled with EDS analyses
were conducted at this specific depth of discharge ͑Fig. 9͒. It re-
vealed the presence of crystallized Cu nanograins ͑Fig. 9b͒, and the
Fourier transform of the HRTEM image of the embedding crystal-
lized matrix lead to a reconstructed diffraction pattern very close to
2
+
the reduction of Cu ions followed by the reduction of the as-
induced Li Sb O . The nature of the amorphous matrix formed at
2
2
6
the bottom of this discharge being still out of reach, we collected the
Figure 6. Bright-field TEM images for ͑a͒ CoSb O and ͑b͒ CuSb O powders with ͑insets͒ corresponding indexed SAED patterns.
2
6
2
6
Downloaded on 2015-02-10 to IP 129.97.124.140 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, 153 ͑9͒ A1778-A1787 ͑2006͒
A1783
Figure 7. ͑a͒ Bright-field TEM and ͑d͒ HRTEM images, ͑b͒ corresponding SAED pattern, and ͑c͒ phase composition as deduced from indexation of the SAED
pattern, for a powder recovered after full discharge ͑18 Li per formula unit, 20°C͒ of a Li/CoSb O cell.
2
6
Figure 8. ͑a͒ Bright-field TEM images
and ͑b,c͒ SAED patterns and HRTEM im-
ages showing the presence of both Cu°
and Sb° metals, for the powder recovered
after full discharge ͑7 Li per formula unit,
2
0°C͒ of a Li/CuSb O cell.
2 6
Downloaded on 2015-02-10 to IP 129.97.124.140 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1784
Journal of The Electrochemical Society, 153 ͑9͒ A1778-A1787 ͑2006͒
Figure 9. ͑a͒ Bright-field TEM image and ͑b,c͒ two selected HRTEM pictures with corresponding EDS spectra and either indexed SAED pattern or recon-
structed electron diffraction pattern obtained by Fourier transform of the bright-field image, for the powder recovered after partial discharge ͑2 Li per formula
unit, 20°C͒ of a Li/CuSb O cell.
2
6
Mössbauer signal. As shown on Fig. 10 and Table II, about half of
reduced LiSbO ͑Li Sb O ͒ ͑Fig. 10a͒. Interestingly, for a fully dis-
3 2 2 6
the antimony ͑43%͒ is still at the +V state after full discharge ͑Fig.
charged CuSb O /Li cell, aside from the Sb° Mossbauer signal
2
6
V
1
0c͒, and very similar Sb contents ͑46%͒ were measured for fully
Figure 10. Mössbauer spectra for ͑a͒ ini-
tial and reduced LiSbO at 0.0 V, ͑b͒ ini-
3
tial and reduced So O at 0.0 V, ͑c͒ re-
2
5
duced CuSb O at 0.0 V, and ͑d͒ reduced
2
6
CoSb O at 0.0 V. Discharge of the cells
2
6
was done at 20°C and at a 1 Li/20 h rate.
Spectra were collected at 4 K and the zero
isomer shift was defined from the spec-
trum of InSb at 4 K ͓␦ = −8.70͑4͒ mm/s
1
21m
relatively to the Ba
SnO source͔.
3
Downloaded on 2015-02-10 to IP 129.97.124.140 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, 153 ͑9͒ A1778-A1787 ͑2006͒
A1785
1
21
+
Table II.
isomer shift relative to InSb (−8.70 mm/s relatively to Ba
Misfit = comparative goodness of fit criterion which gives the fraction of the experimental signal that remains unfitted.
Sb Mössbauer data for initial LiSbO , Sb O , CuSb O , CoSb O and after reduction down to 0.0 V vs Li /Li° (20°C). ␦
3 2 5 2 6 2 6
1
21m
2
=
SnO source), ⌬ = quadrupole splitting, ⌫ = line width, X = chi-squared, and
3
Sample
␦ ͑mm/s͒
⌬ ͑mm/s͒
⌫ ͑mm/s͒
C ͑%͒
X2
Misfit ͑%͒
Ref.
a
LiSbO3
LiSbO3
+8.69͑1͒
−0.28͑6͒
−3.02͑6͒
—
+10.2͑9͒
+8.9͑7͒
—
1.28͑2͒
1.46͑5͒
1.46͑5͒
1.46͑5͒
1.63͑2͒
1.42͑6͒
1.42͑6͒
1.42͑6͒
100
28
26
46
100
26
34
40
100
27
30
43
100
37
47
16
1.41
1.14
0.07
0.07
a
0
V
+
8.58͑1͒
a
a
Sb O
+9.04͑1͒
−1.3͑4͒
−3.7͑1͒
—
2.88
1.14
0.19
0.11
2
5
5
Sb O
+11͑4͒
+13͑1͒
—
2
0
V
+
+8.8
8.61͑2͒
CuSb O
+1.3
17
2
6
6
a
CuSb O
−0.45͑7͒
−3.5͑1͒
+5͑1͒
+13͑1͒
—
1.51͑2͒
1.51͑2͒
1.51͑2͒
0.96
1.06
0.17
0.08
2
0
V
+
+8.6
8.55͑2͒
CoSb O
+2.7
17
2
6
6
a
CoSb O
+0.39͑8͒
−2.27͑7͒
+7͑1͒
+10.2͑9͒
—
1.98͑2͒
1.98͑2͒
1.98͑2͒
2
0
V
+
8.64͑5͒
a
This work.
2
4
26-28
͑
͑
30%, ␦ = −3.5 mm/s͒, another contribution at ␦ = −0.45 mm/s
ides, some being even used as dielectric additives.
In addition to
27%͒ may match intermetallic phase such as Cu Sb, clueing some
the composition of the composite recovered at the bottom of the
discharge ͑Cu/Sb metals and Sb -based matrix͒, this electronic iso-
lating effect can explain the total absence of charge capacity. Back
to ͑Co/Ni͒Sb O , the fully discharged material was confirmed by
Mössbauer to have a limited 16% Sb content ͑Fig. 10d, Table II͒,
confirming the large extent of the reaction up to the alloying reac-
tion.
The conductivity issue can also be illustrated by the evolutions in
the OCV and GITT traces, compared in Fig. 11. The polarization of
a Li/CuSb O cell ͑Fig. 11a͒ slowly decreases during the reaction
with the first 2 Li, consistently with the in situ formation of metallic
Cu acting as a conductive additive. Then, beyond this point, the
polarization continuously increases ͑up to about a value of 1.2 V͒,
as expected for the continuous formation of an isolating Li-rich
2
V
interaction between the two expulsed metals. This intimate metal
mixing could also account for the easy formation of crystallized
intermetallics upon mild heating, as previously described. From all
these results, the complete reaction of Li with CuSb O leads to the
formation of a mixture of nano-Sb and Cu particles expulsed from,
and embedded in, an amorphous Li–Sb –O matrix, summarized as
2
6
2
6
V
V
follows
V
2
V
CuSb O + 7Li → Li Sb O + Cu + Sb
This scenario implies that the Sb particles are not able to undergo
any alloying reaction with Li at low voltage, suggesting an ionic
and/or electronic isolating role of the Li–Sb –O matrix. Actually,
the lithium-rich Sb oxides are well known to be very fast ionic
͓3͔
6
7
6
2
6
V
V
25
conductor materials, and many reports indicated very poor elec-
tron conduction and large bandgap for Li-containing Sb -based ox-
V
V
Sb –O matrix. Worth being pointed out is the absence of constant
Figure 11. Voltage-composition curves
in GITT mode ͑1 Li/10 h, 20°C͒ for ͑a͒
Li/CuSb O , ͑b͒ Li/CoSb O , and ͑c͒
2
6
2
6
Li/Sb cells. The quasiequilibrium OCVs
were recorded when the time evolution
of the voltage was lower than 3 mV/h.
Downloaded on 2015-02-10 to IP 129.97.124.140 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1786
Journal of The Electrochemical Society, 153 ͑9͒ A1778-A1787 ͑2006͒
Figure 14. Cycling galvanostatic voltage-composition curve ͑1 Li/20 h,
0°C͒ for a Li/CoSb O6 cell. Beyond the third cycle, note the temporary
2
2
increase in polarization on charge.
2
4
−
͑
0.45 mm/s͒, and the partial back formation of trivalent oxide
71%͒ which amount is totally consistent with the charge capacity
Figure 12. ͑top panel͒ First galvanostatic cycle for a Li/Sb O cell ͑20°C,
2
3
1
Li/20 h͒ and ͑bottom panel͒ selected Mössbauer spectra. Spectra were col-
recovered at 3.0 V ͑Fig. 12, Table III͒. The effect of the Sb oxida-
tion state could also be illustrated by the drastic increase in capacity
on the lower discharge plateau and therefore improvement of the
lected at 4 K and the zero isomer shift was defined from the spectrum of
1
21m
InSb at 4 K ͓␦ = −8.70͑4͒ mm/s relatively to the Ba
SnO source͔.
3
III
V
charge reversibility when moving from Sb O to Sb O ͑Sb , Sb ͒
2
5
2
4
by a simple air-heating treatment ͑Fig. 13͒. The same trends were
OCV ranges, generally clueing biphasic domain such as Li–Sb al-
noticed by Courtney et al. for amorphous Sb–O-based glasses with
2
9
loying reaction ͑Fig. 11c͒. For a Li/CoSb O6 cell ͑Fig. 11b͒, the
various Sb/O ratios. By the way, the complete irreversibility noted
for Li/Sb2O5 cell rules out the formation of Sb2O5 + Ni° at the early
stage of the reduction of ͑Co/Ni͒Sb2O6.
2
polarization is roughly constant during the first plateau ͑about 6 Li
in length͒ while it slightly and progressively increases along the
second one while still remaining far smaller than values measured
for a Li/CuSb O cell. Beyond this value of 6 Li reacted, the per-
The first reduction step for ͑Co/Ni͒Sb O is thus believed to
2
6
2
6
consist in the reduction of pentavalent antimony, while that of
V
V
fectly constant OCV values indicate at least one biphasic process.
CuSb O leads to Li Sb O and finally isolating Li–Sb –O. The
2 6 2 6
2
By looking at the reactivity of Li with Sb O and Sb O and
intimate process linked to this aspect and the specific role of the
divalent metal are still not fully understood, but this point is pres-
ently being studied by in situ Mössbauer spectroscopy.
2
5
2
3
associated derivative curves ͑Fig. 2g and h͒, it is clear that
Co/Ni͒Sb O behaves as a trivalent oxide and that the short capac-
͑
2
6
ity and complete irreversibility noted for Sb O mimic the behavior
Finally, despite the very small size of the Sb and Sb alloys
formed during the first discharge, the long-term cycling behavior of
͑Co/Ni͒Sb O ͑2,0–0,0 V͒ was not satisfactory. After a few cycles,
2
5
of Li Sb O and CuSb O . For both binary oxides, the first dis-
2
2
6
2
6
charge voltage plateau is still found totally nonreversible, and amor-
phous materials were formed at bottom discharge as deduced from
in situ XRD measurements. Mössbauer analyses further confirmed
the similarity with Cu-based trirutile with a large amount of remain-
2
6
the voltage-composition curve developed an uncommon large and
temporary rise in polarization at the early stage of the charge and
was responsible for the final drop in capacity ͑Fig. 14͒. This point
has not been figured out so far but an overlapping of the charge
processes ͑dealloying plus back formation of oxides͒ may result in a
complex sequence of reactions and evolving conductivity.
V
ing unreduced Sb ͑40%͒ for fully discharged Li/Sb O cells
2
5
͑
Fig. 10b͒. For a Li/Sb O3 cell, Mössbauer spectra showed the
2
complete transformation at 0.0 V into mainly Li Sb alloy ͑␦ =
3
Figure 13. ͑right panel͒ XRD patterns for
Sb oxides with various metal oxidation
states, and ͑left panel͒ corresponding first
galvanostatic cycles vs Li° ͑20°C,
1
Li/20 h͒.
Downloaded on 2015-02-10 to IP 129.97.124.140 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, 153 ͑9͒ A1778-A1787 ͑2006͒
A1787
Table III. 121Sb Mössbauer data for initial Sb O , after reduction down to 0,0 V and after subsequent oxidation at 3.0 V vs Li /Li° (20°C).
+
2
3
1
21m
2
␦
= isomer shift relative to InSb (−8.70 mm/s relatively to Ba
SnO source), ⌬ = quadrupole splitting, ⌫ = line width, X = chi-squared, and
3
Misfit = comparative goodness of fit criterion which gives the fraction of the experimental signal that remains unfitted.
Sample
Sb O
␦ ͑mm/s͒
⌬ ͑mm/s͒
⌫ ͑mm/s͒
C ͑%͒
X2
Misfit ͑%͒
−3.38͑1͒
+1.57͑2͒
−0.50͑5͒
+18.3͑1͒
—
+10.9͑5͒
—
+10.0͑2͒
—
—
1.63͑3͒
1.47͑5͒
1.47͑5͒
1.47͑5͒
1.68͑1͒
1.68͑1͒
1.68͑1͒
100
65
32
3
71
26
3
1.85
1.19
0.11
0.05
2
3
3
Sb O
2
0
V
+
9.0͑2͒
Sb O
−2.89͑2͒
+0.1͑1͒
1.12
0.03
2
3
3
.0 V
+
8.9͑2͒
don), 407, 496 ͑2000͒.
. D. Larcher, C. Masquelier, D. Bonnin, Y. Chabre, V. Masson, J.-B. Leriche, and
Conclusions
2
The electrochemical reactivity of MSb O6 ͑M = Cu, Ni Co͒
J.-M. Tarascon, J. Electrochem. Soc., 150, A133 ͑2003͒.
3. P. Balaya, H. Li, L. Kienle, and J. Maier, Adv. Funct. Mater., 13, 621 ͑2003͒.
2
phases with Li were investigated in order to understand their poten-
tiality as negative electrodes. The study revealed drastically different
behaviors and reaction mechanisms depending on the nature of the
4
. N. Pereira, L. C. Klein, and G. G. Amatucci, J. Electrochem. Soc., 149, A262
2002͒.
. H. Li, G. Richter, and J. Maier, Adv. Mater. (Weinheim, Ger.), 15, 736 ͑2003͒.
͑
5
divalent ion. While CuSb O reacts with only 6–7 Li during the first
2
6
6. R. Alcantara, J. L. Tirado, J. C. Jumas, L. Monconduit, and J. Olivier-Fourcade, J.
discharge and there is no subsequent charge capacity, NiSb O and
Power Sources, 109, 308 ͑2002͒.
. P. Poizot, S. Laruelle, S. Grugeon, and J. M. Tarascon, J. Electrochem. Soc., 149,
A1212 ͑2002͒.
. A. Bystrom, B. Hock, and B. Mason, Ark. Kemi, Mineral. Geol., B15, 1 ͑1941͒.
. W. Weppner and R. A. Huggins, J. Electrochem. Soc., 125, 7 ͑1978͒.
10. P. S. Gopalakrishnan and H. Manohar, Cryst. Struct. Commun., 4, 203 ͑1975͒.
11. W. Gruender, H. Paetzold, and H. Strunz, Neues Jahrb. Mineral., Monatsh., 1962,
2
6
7
CoSb O react with 18–19 Li and 6 Li could be removed on first
2
6
cycles. By coupling electrochemical analyses with in situ XRD mea-
surements and HRTEM/Mössbauer data, we could propose a sce-
nario accounting for these surprising redox behaviors. The reduction
of CuSb O proceeds via a two-step process initially involving the
8
9
2
6
9
3–8.
expulsion of metallic copper clusters, leading to amorphous pen-
V
2
12. J. Rodriguez-Carvajal, Physica B, 192, 55 ͑1993͒.
13. M. Morcrette, Y. Chabre, G. Vaughan, G. Amatucci, J.-B. Leriche, S. Patoux, C.
tavalent Li Sb O that can secondly further react with Li by expul-
2
6
sion of metallic nano-Sb embedded in the resulting Li-rich
Masquelier, and J.-M. Tarascon,. Electrochim. Acta, 47, 3137 ͑2002͒.
V
2
1
1
1
1
4. K. Rubenbauer and T. Birshall, Hyperfine Interact., 7, 125 ͑1979͒.
5. J. L. Lábár, Microscopy and Analysis, 75, 9 ͑2002͒ ͑Euro. edit.͒.
6. V. Propach and D. Reinen, Z. Anorg. Allg. Chem., 369, 278 ͑1969͒.
7. J. D. Donaldson, A. Kjekshus, D. G. Nicholson, and T. Rakke, Acta Chem. Scand.,
Ser. A, A29, 803 ͑1975͒.
Li Sb O amorphous pentavalent matrix. The high electronic insu-
7
6
lating effect of this matrix prevents alloying of Li with extruded Sb
and causes the complete irreversibility of the reaction. In contrast,
͑
Co/Ni͒Sb O are totally reduced into a mixture of Li O, Co/Ni,
2 6 2
and Li Sb and partial reversibility is thus recovered. A close parallel
1
1
8. E.-O. Giere, A. Brahimi, H. J. Deiseroth, and D. Reinen, J. Solid State Chem., 131,
3
2
63 ͑1997͒.
can be made between these results and the compared reactivity of
9. J. N. Reimers, J. E. Greedan, C. V. Stager, and R. Kremer, J. Solid State Chem.,
83, 20 ͑1989͒.
Sb O and Sb O , suggesting that the nature of the divalent cationic
2
5
2
3
species plays a crucial role in determining the nature and oxidation
state of the intermediates phases formed along the first discharge,
directly determining the charge yield. The present study exemplified
that reduction process and therefore cycling performances in the
Li–Sb–O system can be drastically and unexpectedly altered by the
initial oxidation state of Sb and by its evolution upon first reduction.
Although not exhibiting satisfactory cycling behavior, many aspects
of the reactivity of these binary and ternary materials are still ques-
tioned and worth further investigations.
20. A. M. Nakua, H. Yun, J. N. Reimers, J. E. Greedan, and C. V. Stager, J. Solid State
Chem., 91, 105 ͑1991͒.
2
2
1. A. M. Nakua and J. E. Greedan, J. Cryst. Growth, 154, 334 ͑1995͒.
2. M. Dolle, P. Poizot, L. Dupont, and J.-M. Tarascon, Electrochem. Solid-State Lett.,
5
, A18 ͑2002͒.
2
3. N. Kumada and K. Nobukazu, Mater. Res. Bull., 28, 849 ͑1993͒.
24. L. Aldon, A. Garcia, J. O. Fourcade, J. C. Jumas, F. J. F. Madrigal, P. Lavela, C. P.
Vicente, and J. L. Tirado, J. Power Sources, 119, 585 ͑2003͒.
2
2
2
5. C. Muehle, R. E. Dinnebier, L. van Wuellen, G. Schwering, and M. Jansen, Inorg.
Chem., 43, 874 ͑2004͒.
6. H. Bourset, P. Salles, and J.-M. Savariault, Solid State Ionics, 106, 315–320
͑1998͒.
Université de Picardie Jules Verne assisted in meeting the publication
costs of this article.
7. J. M. S. Skakle, M. A. Castellanos, R. S. T. Tovar, and A. R. West, J. Solid State
Chem., 131, 115 ͑1997͒.
2
2
8. W. W. Coffeen, J. Am. Chem. Soc., 39, 154 ͑1956͒.
9. I. Courtney, Ph.D. Thesis, Dalhousie University, Halifax, Nova Scotia, Canada
͑1999͒.
References
1
. P. Poizot, S. Laruelle, S. Grugeon, L. Dupont, and J. M. Tarascon, Nature (Lon-
Downloaded on 2015-02-10 to IP 129.97.124.140 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).