Citrate-derived carbon nanocoatings for poorly conducting cathode: A detailed study using TiO2 substrate materials

Article abstract of DOI:10.1149/1.2221864

Preparation of conductive coatings is an effective way for improvement of kinetics in Li battery materials. Although carbon-based conductive coatings are most often used, they have been studied less systematically than, for example, phosphides or phospho-carbides. In this study we tried to find the most important correlations between the preparation conditions and the resulting composition, morphology, and electrical properties of selected carbon-coated materials. As a model substrate material we mainly used TiO₂ particles while as a carbon precursor we employed citric acid. For comparison, other substrates and precursors were occasionally used. It is shown that in the typical temperature range between 500 and 800°C, the carbon yield varies from 3 to 10 wt %, depending on the initial heating rate and on the duration of heat-treatment. Correlations between the composition, morphology, and the conductivity of final composites are thoroughly discussed. The steep increase in conductivity beyond the percolation threshold (between 1-2 wt % of carbon) can be correlated with the occurrence of a uniform, 1-5 nm thick carbon film that appears to be continuous on a micrometer scale. Finally, we show that the activation energies for conduction in various carbon-coated samples (0.04-0.13 eV) are strikingly similar to those found previously for phosphide- or phosphocarbide-based coatings and even for LiFePO₄ doped with supervalent ions.

Full text of DOI:10.1149/1.2221864

Citrate-Derived Carbon Nanocoatings for Poorly Conducting
Cathode : A Detailed Study Using Substrate Materials
J. Moskon, R. Dominko, M. Gaberscek, R. Cerc-Korosec and J. Jamnik
J. Electrochem. Soc. 2006, Volume 153, Issue 10, Pages A1805-A1811.
doi: 10.1149/1.2221864
Email alerting
service
Receive free email alerts when new articles cite this article - sign up in the box at the
top right corner of the article or click here
To subscribe to Journal of The Electrochemical Society go to:
http://jes.ecsdl.org/subscriptions
©
2006 ECS - The Electrochemical Society
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
A1805
0
013-4651/2006/153͑10͒/A1805/7/$20.00 © The Electrochemical Society
Citrate-Derived Carbon Nanocoatings for Poorly Conducting
Cathode
A Detailed Study Using TiO Substrate Materials
2
a
a
a, ,z
b
a
J. Moskon, R. Dominko, M. Gaberscek, * R. Cerc-Korosec, and J. Jamnik
a
National Institute of Chemistry, SI-1000 Ljubljana, Slovenia
b
Faculty of Chemistry and Chemical Technology, SI-1000 Ljubljana, Slovenia
Preparation of conductive coatings is an effective way for improvement of kinetics in Li battery materials. Although carbon-based
conductive coatings are most often used, they have been studied less systematically than, for example, phosphides or phospho-
carbides. In this study we tried to find the most important correlations between the preparation conditions and the resulting
composition, morphology, and electrical properties of selected carbon-coated materials. As a model substrate material we mainly
used TiO particles while as a carbon precursor we employed citric acid. For comparison, other substrates and precursors were
2
occasionally used. It is shown that in the typical temperature range between 500 and 800°C, the carbon yield varies from 3 to
1
0 wt %, depending on the initial heating rate and on the duration of heat-treatment. Correlations between the composition,
morphology, and the conductivity of final composites are thoroughly discussed. The steep increase in conductivity beyond the
percolation threshold ͑between 1–2 wt % of carbon͒ can be correlated with the occurrence of a uniform, 1–5 nm thick carbon film
that appears to be continuous on a micrometer scale. Finally, we show that the activation energies for conduction in various
carbon-coated samples ͑0.04–0.13 eV͒ are strikingly similar to those found previously for phosphide- or phosphocarbide-based
coatings and even for LiFePO doped with supervalent ions.
4
©
2006 The Electrochemical Society. ͓DOI: 10.1149/1.2221864͔ All rights reserved.
Manuscript submitted March 17, 2006; revised manuscript received April 24, 2006. Available electronically July 26, 2006.
Many materials for positive battery electrodes with interesting
thermodynamic properties ͑high voltage, high theoretical capacity͒
are by nature typical ceramic materials with quite low electronic
ity, claiming that in those materials the reason for conductivity en-
hancement was the presence of a percolated conducting network
consisting of elementary carbon. Another explanation for conductiv-
ity enhancement in certain LiFePO4-based materials refers to the
presence of small amounts of phosphides or phosphocarbides which
−
3
−1
conductivity ranging from about 10 S cm for LiCoO down to
2
−
9
−1
1-5
10
S cm for LiFePO at room temperature. To enhance the
4
5
insertion/deinsertion kinetics of such materials it is necessary to cre-
ate an efficient electron-conducting network among or around the
active particles. For active materials from the upper part of the con-
are known for their metallic-like conducting properties. In some
cases, in which authors do attribute the conductivity enhancement to
the presence of carbon, they assume that carbon is present in the
1
3
−
3
−5
−1
ductivity range ͑10 –10 S cm ͒ it is usually sufficient to create
percolated conduction pathways by adding a sufficient amount ͑sev-
eral wt %͒ of an electron conductor ͑typically carbon-based͒ in the
form of fine powder ͑Fig. 1a͒. To achieve a comparable performance
form of a dispersed nanophase rather than in the form of coatings.
Finally, we have shown in our own reports
9,14,15
that certain prepa-
ration procedures result in architectures where the active substance
does not appear in a particulate form but rather in a form of porous
monoliths. Although there are indications that the carbon formed in
this procedure is deposited on all surfaces of such a material in the
form of a ca. 1 nm thick coating, we have not been able to prove this
general picture unambiguously.
with the poorest conductors ͑e.g., LiFePO ͒, the presence of a
4
simple percolated conducting network seems to be insufficient. It
6-9
10,11
has been shown empirically and theoretically
that the overall
transport in electrodes based on poorly conductive active particles is
only fast enough if an electron conductor surrounds the whole sur-
face of each particle ͑typical particle size of practical materials is in
the order of 100 nm͒. Obviously, the conductive coatings ͑Fig. 1b͒
help create a close-to-optimal wiring situation where the mixed
transport of electrons and lithium ions proceeds more or less spheri-
cally into/out of active particles. Here it should be stressed that, in
fact, the impact of electron conducting coatings on the ionic trans-
port remains rather unclear. One can speculate that the coatings are
thin enough ͑they are reported to be in the order of 1 nm or even
The present study is a first attempt to find clear correlations
between the preparation conditions and the resulting composition,
9
less ͒ that ions can penetrate through them without significant po-
larization. Alternatively, ions could penetrate through defects in the
structure of coatings. In any case, it has been shown experimentally
that thicker coatings ͑with average thickness 2–3 nm or more͒ do
9
show a deteriorating effect on the electrochemical kinetics, prob-
ably reflecting a hindrance of the ionic transport in such cases.
Although the general picture concerning the role of conducting
coatings presented above might seem quite consistent, various alter-
native views may be found in literature, especially when considering
Figure 1. Schematic presentation of two limiting types of cathode compos-
ites consisting of poorly conductive electrochemically active particles and a
second, well-conductive phase: ͑a͒ The case where the well-conductive
phase forms percolating pathways within the composite. Although the aver-
age conductivity of such a composite may be very high, the electrochemical
5
specific practical cases. For example, Chung et al. reported that the
excellent electron conductivity of LiFePO doped with selected su-
4
pervalent ions was due to changes in the electronic structure of these
materials with respect to the electronic structure of poorly conduc-
͑
insertion͒ reaction is very fast only at selected places on a particle’s surface,
12
that is, where the conducting phase is in electronic contact with the substrate
particles. ͑b͒ The case where the conductive phase forms a coating around
each active particle. The coating must be permeable for the reacting ions ͑the
permeability is indicated by the cracks in the layer although permeation may
be realized via an arbitrary mechanism͒. Here the insertion reaction is
equally fast at all points on the particle’s surface.
tive pristine LiFePO . Later on, Ravet et al. rejected this possibil-
4
*
Electrochemical Society Active Member.
E-mail: Miran.Gaberscek@ki.si
z
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

A1806
Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
morphology, and electrical properties of selected carbon-coated
composites. To avoid the influence of substrate on electrical conduc-
tivity measurements, we chose among insulators with a conductivity
−
10
−1
less than 10 S cm . At the same time we wanted the substrate to
have a uniform size to allow for easier and more reproducible study
of morphological and electrical features of prepared composites. For
the purpose of generality, it was also desirable that the substrate was
available in different particle sizes, ranging from micro- to nanom-
eter range. A material that satisfied these criteria was TiO ͑mainly
2
in the anatase modification͒. Still, to rule out completely a possible
influence of the substrate composition and morphology on selected
properties of interest, we occasionally also used other substrates
͑
LiFePO , flat quartz͒. As a carbon precursor, we mostly used the
4
citrate anion. Namely, as we have shown in a range of previous
9
,14,15
papers,
this precursor yielded uniform films on LiFePO , the
4
thickness of which could even be tuned. To study the aforemen-
tioned correlations, the following preparation conditions were con-
trolled and changed in a systematic way: ͑i͒ the composition of the
initial dispersion containing the substrate and organic precursor, ͑ii͒
the homogeneity of the initial dispersion, ͑iii͒ the heating rate, ͑iv͒
the final heating temperature, and ͑v͒ the duration of heat-treatment.
The resulting materials composition was studied using mainly ther-
mogravimetric ͑TG͒ analysis partly supplemented with X-ray dif-
fraction ͑XRD͒ and energy dispersive X-ray analysis ͑EDS͒ analy-
sis, the morphological features were studied using scanning electron
microscopy ͑SEM͒ and transmission electron microscopy ͑TEM͒,
and the electrical conductivity measurements were performed using
impedance spectroscopy. To find the conduction mechanism in vari-
ous composites, temperature dependence of the samples’ conductiv-
ity was measured and commented on with respect to selected previ-
ous results.
Figure 2. TG curves for TiO ͑0.25 ␮m͒–carbon composites with different
2
initial contents of citric acid ͑as indicated beside the curves͒.
lected temperature, the samples were further heated for 10 h at that
temperature and then left to cool down to room temperature.
After the heat-treatment, the carbon content in all composites
͑
except in the sample with flat quartz substrate͒ was determined
using TG analysis. The measurements were performed on a Mettler
Toledo TGA/SDTA 851e thermoanalyzer in a temperature range
−
1
from 30 up to 650°C at a heating rate of 10°C min . The TG
−
1
curves were recorded in an oxygen flow of 100 mL min . The ini-
tial mass of samples was approximately 50 mg. In all cases, the
baseline was subtracted.
The morphology of prepared composites was investigated using
a field-emission SEM—Supra 35 VP Carl Zeiss. For selected
samples EDS analysis was also performed. Selected powdered ma-
terial were spread on holey carbon-coated Cu grid and investigated
by a high-resolution field-emission TEM JEOL 2010F.
The X-ray powder diffraction pattern of the samples was col-
lected on a Philips PW 1710 diffractometer using Cu K␣1 radiation
Experimental
As substrate particles for preparation of carbon coatings, three
different sets of TiO particles were used. The most extensive stud-
2
ies were performed using anatase particles “Riedel-de Haen 14027”
which, according to the producer, have a mean particle size value
of 0.25 ␮m. For comparison, 1 ␮m Aldrich 248576-100G and
͑
␭ = 1,5418 Å͒. The data were collected in the range between 15
2
5–70 nm Aldrich 634662-25G TiO particles were used. The re-
2
and 70°2␪ in steps of 0.04°.
spective specific surface areas of these particles determined in our
For conductivity measurements, the obtained composite powders
were pressed in pellets using about 10 wt % of Teflon as a binder.
The diameter of pellets was always 8 mm while the thickness varied
from 1.75 to 2.65 mm. The contacts were made by painting both
basal planes using either a silver conductive paste ͑for room-
temperature measurements͒ or a Pt paste ͑for measurements of con-
ductivity as a function of temperature͒. The room-temperature im-
pedance measurements were performed in air conditions while the
temperature-dependent impedance spectra were recorded by putting
the sample in a gas-tight quartz tube equipped with appropriate
shielded Pt wires and thermocouples. Prior to and during heating, an
inert atmosphere was maintained within the tube using a 5.0 Ar. The
impedance spectra were recorded in a frequency range of 1 MHz to
laboratory using a Brunauer, Emmett, and Teller ͑BET͒ analysis
2
−1
͑
Micromeritics TriStar 3000͒ were: 9.6, 9.0, and 20 m g . In one
experiment, the carbon films were additionally prepared on a flat
quartz plate and on a LiFePO cathode material for Li batteries. In
4
9,14,15
the latter case, we used a special sol-gel procedure
in which in
the carbon precursor ͑citrate͒ and the material’s precursors were
mixed on the molecular level and both the LiFePO4 and carbon
phases were created and separated later on, during the heating pro-
cedure. In the case of TiO particles, carbon films were prepared by
2
adding a known amount of the particles into an aqueous solution of
citric acid. The mass ratio between TiO particles and citric acid was
2
varied in a wide range to obtain the final composites containing a
carbon content from 0 to more than 20 wt % ͑for comparison, pure
citric acid was also used in one experiment͒. The resulting disper-
sion was first homogenized using an ultrasound mixer, dried, and
finally heat-treated according to a selected procedure. In one experi-
2
0 Hz with a Hewlet Packard 4284A instrument.
Results and Discussion
ment, solid citric acid and substrate TiO particles in appropriate
Influence of synthesis parameters on carbon yield.— The first
2
mass ratio were simply ground together and the obtained solid mix-
tures were heat-treated according to a specified procedure. In the
case of the flat quartz substrate, we used for the carbon precursor
naproxen instead of citric acid. It appeared that upon heat-treatment,
naproxen formed more uniform carbon films on the flat quartz sur-
face which, later on, gave more reproducible conductivity measure-
ments. Naproxen was simply melted on the substrate surface and
then burnt according to a specified thermal procedure in inert atmo-
sphere.
series of samples was prepared by heat-treating the initial mixtures
containing a different precursor ͑citric acid͒: substrate ͑0.25 ␮m
anatase͒ ratios. In all cases the heating regime was the same: the
initial heating rate was 10°C/min while after reaching the final tem-
perature of 900°C the samples were maintained at that temperature
for 10 h. TG curves for these samples recorded in oxygen atmo-
sphere are shown in Fig. 2. In all cases it was assumed that the
carbon content was proportional to the weight loss between 300 and
600°C, respectively. A comparison between the initial materials
composition ͑established through precise weighting of reactants͒
and the composition of final composite shows that the carbon yield
is more or less independent of the content of citric acid in the initial
If not specified otherwise, the samples were heat-treated in a
gas-tight quartz tube with a moderate but constant flow of Ar 5.0.
−
1
The initial heating rate was 10°C min . After reaching the prese-
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
A1807
Figure 4. ͑a͒ Carbon yield as a function of carbonization temperature. ͑b͒
TG curves for the composite samples prepared at different carbonization
temperature; only the sample prepared at 960°C exhibits an increase in mass
above ca. 500°C.
Figure 3. Carbon yield as a function of precursor ͑citric acid͒-substrate mass
ratio for ͑a͒ different substrate particles and for different methods of com-
posite preparation and ͑b͒ for different initial heating rates ͑i.e., heating rates
before reaching the final carbonization temperature of 900°C͒.
dispersion ͑Fig. 3a͒ and amounts on average 3.4 ± 0.5%. One of the
reasons for the comparatively low yield is the fact that citric acid
contains oxygen atoms. Indeed, it has been reported that the yield
can be maximized by maximizing the ratio between the content of
different processes taking place at high temperatures. The first is
indicated already in the TG curve for the sample prepared at 960°C,
which shows an increase in mass between 500 and 650°C ͑Fig. 4b͒.
1
6
aromatic fragments and the content of heteroatoms. When slightly
This increase is attributed to changes in the substrate ͑TiO ͒ rather
2
larger TiO particles were used ͑1 ␮m͒, the yield remained similar
2
than changes in the organic precursor. Indeed, XRD patterns of un-
but dropped to only about 2.3% if 70 nm particles were employed
treated TiO -citric acid sample and two samples treated at 960°C for
2
͑
Fig. 3a͒. The reason for this decrease is not clear. More expected is
different periods of time show important structural changes in the
the decrease observed in samples where the initial mixture was less
homogeneous, i.e., prepared by simple “dry mixing” of solid reac-
tants rather than dissolving or dispersing and mixing them in appro-
priate solvents ͑Fig. 3a, full triangles͒. We believe that in such less
homogeneous samples a larger portion of organic precursor is de-
tached from the substrate during heating ͑probably carried away by
the large amount of gases evolved during carbonization͒ and depos-
ited elsewhere in the heating chamber. In a way, this assumption is
consistent with the next set of experiments in which the initial heat-
ing rate was gradually decreased from 10 to 1°C/min ͑Fig. 3b͒.
Apparently, with lower heating rates, the process of carbonization
becomes less vigorous, that is, it involves lower rates of gas evolu-
tion, which results in smaller losses of solid carbonized material.
substrate ͑Fig. 5͒. Apart from the expected partial transformation
17,18
from anatase to rutile,
shows the presence of Ti O . It is assumed that during subsequent
the diffractogram at 960°C ͑10 h͒ also
17
9
recording of TG curves ͑in oxygen͒, this phase is transformed back
into TiO , hence the observed increase in mass. The second, and, as
2
shown later, important reason for understanding the conductivity
measurements, may be the low carbon content determined for the
high-temperature samples could be entrapment of a significant
amount of carbon into agglomerates. To confirm the possibility of
such an entrapment, we additionally heat-treated a sample prepared
at 900°C and a sample prepared at 960°C for 1 h at 800°C in air
conditions. While the former sample turned into a whitish nuance
typical for carbon-free TiO , the latter retained a brown-grayish hue
2
Anomalies due to interaction between carbon precursor and sub-
strate during carbonization.— One of the crucial parameters affect-
ing the process of carbonization is expected to be the plateau value
of carbonization temperature. Figure 4a shows that the yield remains
quite constant within the range 500–750°C but then starts to de-
crease significantly, especially above 900°C. At 960°C the carbon
yield is only about 2%. Based on several supplementary tests we
assume that this low yield can be a consequence of at least three
indicating the remainders of carbon within the sample, similar to
1
9
that reported previously for carbon coatings on Al O . The pres-
2
3
ence of carbon in this sample was additionally confirmed using an
EDS analysis before and after the additional heat-treatment ͑not
shown͒. This result implies that the sample prepared at 960°C con-
tains a higher total content of carbon than observed with TG and
displayed in Fig. 4a. We take advantage of this implication when
explaining an anomaly observed in conductivity measurements ͑c.f.,
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

A1808
Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
Figure 5. XRD patterns for three selected samples prepared using the
.25 ␮m TiO substrate particles: ͑bottom͒ uncarbonized sample containing
0
2
citric acid, ͑middle͒ sample carbonized at 960°C for 1 h, and ͑top͒ sample
carbonized at 960°C for 10 h. The peaks pertaining to different phases are
marked as explained in the legend.
Fig. 9 and the corresponding discussion͒. A third reason for the low
carbon yield at 960°C could be due to further carbonization pro-
cesses expected at higher temperatures.
Figure 7. SEM micrographs of selected carbonized samples prepared with
0.25 ␮m TiO2 particles: ͑a͒ untreated TiO2 particles, ͑b͒ carbon content
Ͻ0.02%, ͑c͒ carbon contentϭ0.80%, ͑d͒ carbon contentϭ2.66%, ͑e͒ carbon
contentϭ4.42%, and ͑f͒ carbon contentϭ20.53%.
Another effect observed in TiO substrate, which can be attrib-
2
uted to the presence of carbonized phase, is a decrease in the rate of
anatase → rutile transformation. The XRD diffractogram in Fig. 6a
shows that in the case of 70 nm TiO particles, already the untreated
2
sample contained a mixture of anatase and rutile phases in an ap-
proximate ratio of 3:1. As expected, upon heat-treatment an addi-
tional amount of anatase transformed into rutile. However, at equal
heating conditions, the sample containing more carbonized phase
particles” was significantly smaller ͑Ͻ200 nm͒ than the nominal
value. This smaller value is consistent with the “average particle
size” that can be evaluated from BET measurements: taking the
2
−1
measured value of specific surface area ͑9.6 m g ͒ and the theo-
−
3
͑
ca. 1%͒ retained a larger portion of anatase structure than the one
retical density for anatase ͑3.82 g cm ͒, one gets for average par-
ticle size a value of ca. 160 nm ͑if assumed that the particles are
spherical͒. In carbonized composites ͑Fig. 7b-e͒ containing low
carbon contents, the carbon phase cannot be observed as a separate
morphological feature. We assume that at low contents the carbon
phase is rather homogeneously ͑uniformly͒ distributed around and
between the substrate particles. The typical dimensions of carbon
phase ͑either thickness, if a film, or radius, if aggregates͒ should
be less than ca. 5–10 nm ͑i.e., less than the resolution of SEM
at present conditions͒. A good uniformity on a scale from ca. 1 to
with the smaller ͑0.31%͒ carbon content ͑see Fig. 6b and c, respec-
tively͒. This result is consistent with several earlier reports on the
2
0-23
role of carbon in anatase → rutile transformation.
The traces of
unidentified phase in Fig. 6b and c could be related to a reduced
TiO phase, similar to that discussed earlier ͑c.f., Fig. 5 and the
2
corresponding explanation in text͒.
Morphology of TiO -carbon composites.— The effect of carbon-
2
ization on materials’ morphology was studied using SEM ͑Fig. 7 and
8͒ and TEM ͑Fig. 9͒. First it should be noted that the typical particle
size of as-received sample designated by the producer as “0.25 ␮m
Figure 6. XRD patterns for three selected samples prepared using the 70 nm
TiO2 substrate particles: ͑bottom͒ initial TiO2 substrate particles, ͑middle͒
carbonized sample with 0.99 wt % of carbon, and ͑top͒ carbonized sample
with 0.31 wt % of carbon. The peaks pertaining to different phases are
marked as explained in the legend.
Figure 8. EDS line analysis of the sample prepared with 0.25 ␮m TiO and
containing 3.62 wt % of carbon.
2
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
A1809
Impact of carbon phase on the electrical properties of compos-
ites.— The main feature of the impedance spectra of pellets pre-
pared from different TiO -carbon composites was an arc, or at high
2
carbon contents a part of an arc, in complex plane. It is reasonable to
attribute such an arc to a parallel combination of a capacitance due
to bulk dielectric properties of the pellet and a resistor due to aver-
age resistivity ͑or, alternatively, average conductivity͒ of this pellet.
The average pellet conductivity as a function of carbon content for
differently prepared TiO -carbon composites is displayed in Fig. 10.
2
The basic curve ͑Fig. 10a͒ obtained for the most investigated
2
4
0
.25 ␮m anatase sample shows a typical percolation shape with a
roughly estimated percolation threshold between 1–2 wt % ͑in good
agreement with TEM observations, see the previous section͒. The
maximum ͑plateau͒ conductivity of composites is between 0.1 and
−
1
1
.0 S cm , a value commonly obtained for insulators or poor con-
2
4
ductors doped with carbon. For battery purposes, the materials
slightly above the percolation threshold are the most interesting:
there the best compromise between the requirement of high enough
a conductivity ͑e.g. Ͼ0.01 S cm ͒ and as low as possible a carbon
content ͑i.e., to avoid unnecessary decrease of energy density͒ is
achieved. In the present case, the samples with about 2.5 wt % of
carbon could represent such a compromise. Decreasing the heating
rate moderately improves the conductivity at a given carbon content
−
1
͑
Fig. 10b͒. Based on this, and on some previously shown results ͑see
Fig. 3 and the corresponding discussion͒, we speculate that a less
vigorous carbonization process with a lower rate of gas evolution
can result in a slightly more uniform carbon distribution on the
surface of substrate particles and/or between them.
Figure 9. TEM micrographs of carbon-TiO composites: ͑a,b͒ examples of
2
two different morphologies found in sample with 0.80 wt % of carbon; ͑c,d͒
two typical morphologies found in sample with 2.66 wt . % of carbon. The
Knowing the basic percolation curves, we were interested in how
conductivity develops if a selected composite with a given starting
precursor content is heated to different final temperatures. The re-
sults in Fig. 10c show that heating up to 500°C gives a very poor
conductivity of the composite ͑comparable to the conductivity of
crystal planes of TiO can be clearly distinguished from the more disordered
2
and transparent surface film and patches due to carbon.
pure TiO ͒. This indicates that up to this temperature no conducting
2
carbon phase has been developed. By heating to higher tempera-
tures, the conductivity improves significantly. Such an increase in
conductivity above 500°C has been observed previously for other
ca. 100 ␮m was indirectly confirmed by performing EDS analysis
on selected samples. A typical concentration profile for carbon ob-
tained by EDS line scan is shown in Fig. 8. It can be seen that along
a line of ca. 50 ␮m in length, carbon is distributed quite uniformly,
that is, it forms a continuous path, although its concentration fluc-
tuates on a micrometer scale. The continuity of carbon on these
scales is very important for understanding why good electrical be-
havior can be achieved already at relatively low carbon content ͑see
next section͒.
24
carbon materials. At 700°C the carbon phase achieves a maximum
conductivity ͑within the conditions tested in the present study͒.
Heating to higher temperatures decreases the carbon content, con-
sistent with the results shown in Fig. 4a. Very interestingly, this loss
of carbon shifts the conductivity points exactly along the percolation
curve obtained in a completely different way ͑see discussion of Fig.
1
0a͒. In other words, it seems that, at least for samples prepared
TEM micrographs ͑Fig. 9͒ reveal the carbon distribution on a
nanometer scale. It seems that at very low carbon contents ͑up to ca.
between ca. 700 and 900°C, the conductivity of the composites is
determined mainly by the carbon content but is quite insensitive to
the preparation procedure. This result can be explained by a good
uniformity of the composites. The only exception is the point at
1
wt %͒ the substrate surface is mostly carbon-free ͑Fig. 9a͒. A
thorough inspection, however, reveals the presence of isolated car-
bon patches at certain spots ͑Fig. 9b͒. Based on a large number of
inspections, the general impression is that, at low concentrations
9
4
60°C. Based on the previous results, especially those shown in Fig.
and 7, we explain this significant deviation by at least three pos-
͑
below ca. 1 wt %͒, carbon tends to agglomerate at selected spots
sible effects: ͑i͒ the carbon content for this point is not determined
correctly ͑due to possible carbon entrapment within sintered ag-
glomerates, see explanations for Fig. 4a and 7b͒, ͑ii͒ the conductiv-
ity of the carbon phase is further increased due to a possible further
process of carbonization/graphitization ͑see discussion on the devia-
tion of the point at 960°C in Fig. 4a͒, and ͑iii͒ the conductivity of
substrate itself is significantly increased, thus contributing to the
total composite conductivity. The last mechanism refers to formation
rather than spread over the entire substrate surface to form films.
Such films, however, are clearly observed at concentrations above
ca. 2 wt % ͑Fig. 9c͒. Although a major part of carbon now deposits
in the form of a quite uniform film, it should be stressed that larger
agglomerates ͑patches͒ can still be observed ͑Fig. 9d͒. A similar
morphology development was observed in our previous studies
9
,15
where carbon coatings were prepared on LiFePO4.
It can be
expected that the transition from the mostly carbon-free surface at
ca. 1 wt % of carbon to the uniformly carbon-covered surface at ca.
of a reduced TiO phase, for example, such as observed in Fig. 5,
2
and to literature data which show that reduced TiO phases may
2
wt % of carbon will lead to a corresponding transition in electrical
2
2
5
properties, i.e., the conductivity percolation threshold should lie be-
exhibit metallic properties.
We also checked the possible impact of the substrate size on
composite conductivity ͑Fig. 10d͒. While the larger particles with
similar sizes ͑0.25 and 1 ␮m͒ show similar conductivities, the
smaller particles ͑70 nm͒ exhibit somewhat higher conductivities at
a given carbon content. The reason for this effect is unclear. In
principle, it could be related to the incorrect determination of carbon
content using TG, as a consequence of carbon entrapment into ag-
tween 1 and 2 wt % of carbon.
Another well-seen feature in Fig. 7 is the decreasing average
particle size with increasing carbon content. This is consistent with
22,23
some previously published results
in which authors reported that
the presence of carbon phase prevented sintering of particles during
heat-treatment. Analogous trends in sample morphology to those
shown in Fig. 7 were observed for the 70 nm particles ͑not shown͒.
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

A1810
Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
Figure 10. Conductivity as a func-
tion of carbon content for different
samples prepared in this study: ͑a͒
A basic curve obtained with the
samples containing the 0.25 ␮m
TiO2 substrate particles and car-
bonized at 900°C for 10 h with the
initial heating rate of 10°C/min;
average values of up to 10 indi-
vidual measurements are shown as
points, and the bars represent stan-
dard deviations. ͑b͒ The effect of
initial heating rate on the final con-
ductivity of selected composites.
͑
c͒ The effect of carbonization
temperature; the samples denoted
by triangles were obtained by car-
bonizing
a set of equal initial
samples with given precursor con-
tent at different carbonization
temperatures. ͑d͒ The effect of
particle size and the preparation
method on the conductivity of final
composites.
glomerates ͑see Fig. 4a͒. We hoped that a more detailed analysis of
the percolation curves for different samples could give additional
information in this respect. For analysis we took the frequently
throughout the present study, the citric acid. To check for the gen-
erality of these curves, we prepared two further samples. One was a
carbon-coated LiFePO material where, as the carbon precursor, a
4
2
6,27
28
used,
though simplified, scaling equation which is only appro-
mixture of citric acid and gelatine was used. The last sample was a
carbon-coated flat quartz substrate where the coating was prepared
by heat-treatment of naproxen. Despite the big differences in the
selected systems, one can observe several common features in the
measured Arrhenius plots in Fig. 11. The most important is that all
systems exhibit a gradual increase in activation energy with increas-
ing temperature but, also, similar values of the slopes at lower tem-
peratures. This type of Arrhenius curve has been previously reported
priate for phenomenological description of the percolation curve at
the volume fractions of conductive phase, p, higher than the perco-
lation threshold, pc
t
p − p
c
␴
= ␴0ͩ ͪ
͓1͔
1
− p
c
where ␴ is the sample conductivity, ␴ is the bulk conductivity of
the conducting phase, and t is an exponent related to dimensionality
0
30
for amorphous and partly graphitized carbons and has been attrib-
26,27
of the conducting network.
Analyzing the measured percolation curves using Eq. 1, one
finds that, unfortunately, uncertainty of parameters determination
is too high to make a conclusive comparison between different
samples. It can, however, be generally concluded that the maximum
−
1
value of conductivity of the carbonized phase itself ͑ca. 100 S cm ͒
is much lower than the maximum conductivity of graphite
−
1
͑
1
26,000 S cm
along the a-axis in single-crystalline graphite,
−
1
29,24
300 S cm for polycrystalline graphite͒.
The value of perco-
lation threshold varies quite a lot but it is safe to claim that it does
not exceed 4 vol % ͑about 2 wt %͒. The dimensionality factor, t,
also shows large variations. The most reliable results were obtained
for the 0.25 ␮m TiO ͑t = 1.9 ± 02͒. This indicates that the carbon
2
phase is more or less isotropically distributed within the composite.
Namely, the theoretical value of t for ideal 3D networks of conduc-
tive phase is reported to be 1.94, while for 2D structures this value
2
6,27
is 1.3.
In the last set of experiments we measured the activation energy
for electron conduction in selected carbonized composites ͑Fig. 11͒.
The results for the 0.25 ␮m TiO -carbon system are compared with
2
a system of high interest for battery research, namely LiFePO deco-
4
9-11,14,15
rated with a carbon coating.
Most importantly, the carbon
Figure 11. Arrhenius plots for different samples measured in inert atmo-
coating on LiFePO was prepared using the same precursor as used
sphere in temperature range 25–500°C.
4
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 153 ͑10͒ A1805-A1811 ͑2006͒
A1811
2
uted to a wide distribution of barrier heights and widths at sp sites.
A larger degree of graphitization ͑e.g., due to different heat-
treatment or a different substrate͒ can bring about weaker tempera-
terials prepared in the present study: LiFePO -citrate, LiFePO -
4 4
citrate + gelatine, and SiO -naproxene. Interestingly, this is the
2
same range as previously reported for various LiFePO - based ma-
4
3
0
ture dependence and, hence, a smaller average slope. This means
that most differences in the observed curves can be explained by the
slightly different degree of graphitization of these samples. Having
this in mind, we find it quite remarkable that the difference between
the highest ͑0.13 eV͒ and the lowest ͑0.04 eV͒ activation energy
observed for these systems is relatively small. In any case, this brief
analysis and comparison with literature data indicates, with a high
probability, that the electron conduction in all these systems is pri-
marily determined by the carbon phase and is essentially indepen-
dent of the substrate and even of the precursor type. Although such
a conclusion might seem quite obvious in the present context, it
should be reminded that authors often seek for alternative explana-
tions. As mentioned in the introduction, the best-known example is
terials in which, however, the conduction was attributed to noncar-
4
,5
bon phases.
Acknowledgments
Financial support from The Ministry of Higher Education, Sci-
ence, and Technology of the Republic of Slovenia is gratefully ac-
knowledged. Support from the European Network of Excellence
‘
ALISTORE’ network is also acknowledged.
National Institute of Chemistry assisted in meeting the publication costs
of this article.
References
1. J. Molenda, A. Stoklosa, and T. Bak, Solid State Ionics, 36, 53 ͑1989͒.
4
12
2
3
. Y. Shimakawas, T. Numata, and J. Tabuchi, J. Solid State Chem., 131, 138 ͑1997͒.
. H. Kawaia, M. Nagatab, H. Kageyamac, H. Tukamoto, and A. R. West, Electro-
chim. Acta, 45, 315 ͑1999͒.
the dispute between Chung et al. and Ravet et al. about the nature
of electron conduction in doped LiFePO materials. Comparing the
4
4
present data with those published by Chung et al., one may find it
4
. S. Chung, J. T. Bloking, and Y.-M. Chiang, Nat. Mater., 1, 123 ͑2002͒.
very indicative that the activation energies for the materials prepared
by Chung et al. ͑0.062–0.075 eV, see Ref. 4͒ lie in the middle of the
range determined for the present materials ͑0.04–0.13 eV͒ and that
the room temperature values are of the same order of magnitude
5. P. S. Herle, B. Ellis, N. Coombs, and L. F. Nazar, Nat. Mater., 3, 147 ͑2004͒.
6
. N. Ravet, J. B. Goodenough, S. Besner, M. Simoneau, P. Hovington, and
M. Armand, Abstract 127, The Electrochemical Society and the Electrochemical
Society of Japan Meeting Abstracts, Vol. 99–2, Honolulu, HI, Oct. 17–22, 1999.
7. H. Huang, S. C. Yin, and L. F. Nazar, Electrochem. Solid-State Lett., 4, A170
2001͒.
8. Z. H. Chen and J. R. Dahn, J. Electrochem. Soc., 149, A1184 ͑2002͒.
−
1
͑
͑
0.01 S cm ͒. The values of activation energies determined by
5
Herle et al. for various iron phosphates ͑0.03–04 eV͒ are also simi-
lar to the ones determined in Fig. 11. Of course, a mere comparison
of activation energies is insufficient for reliable determination of a
conduction mechanism. However, given the present results and their
comparison with literature data, one should consider the possibility
that in all these materials, which may contain various interesting
phases, the phase which determines the overall conductivity is al-
ways the carbon phase.
9
. R. Dominko, M. Bele, M. Gaberscek, M. Remskar, D. Hanzel, S. Pejovnik, and J.
Jamnik, J. Electrochem. Soc., 152, A607 ͑2005͒.
10. M. Gaberscek, R. Dominko, M. Bele, and J. Jamnik, in NewTtrends in Intercalation
Compounds for Energy Storage and Conversion, K. Zaghib, C. M. Julien, and J.
Prakash, Editors, PV 2003-20, pp. 309–315, The Electrochemical Society, Proceed-
ings Series, Pennington, N.J. ͑2003͒.
11. M. Gaberscek and J. Jamnik, Accepted for publication in Solid State Ionics ͑special
issue of 14th SSI Meeting in Baden Baden, Germany͒.
1
2. N. Ravet, A. Abouimrane, and M. Armand, Nat. Mater., 2, 702 ͑2003͒.
1
3. S. T. Yang, N. H. Zhao, H. Y. Dong, J. X. Yang, and H. Y. Hue, Electrochim. Acta,
5
1, 166 ͑2005͒.
Conclusions
14. R. Dominko, J. M. Goupil, M. Bele, M. Gaberscek, M. Remskar, D. Hanzel, and J.
Jamnik, J. Electrochem. Soc., 152, A858 ͑2005͒.
15. M. Gaberscek, R. Dominko, M. Bele, M. Remskar, D. Hanzel, and J. Jamnik, Solid
State Ionics, 176, 1801 ͑2005͒.
6. I. L. Spain, Chemistry and Physics of Carbon, P. L. Walker, Jr., and P. A. Thrower,
Editors, Marcel Dekker, Inc., New York ͑1981͒.
7. H. Zhang and J. F. Banfield, Am. Mineral., 84, 528 ͑1999͒.
8. J. M. G. Amores, V. S. Escribano, and G. Busca, J. Mater. Chem., 5, 1245 ͑1995͒.
19. M. Inagaki, H. Miura, and H. Konno, J. Eur. Ceram. Soc., 18, 1011 ͑1998͒.
0. T. Tsumura, N. Kojitani, I. Izumi, N. Iwashita, M. Toyoda, and M. Inagaki, J.
Heat-treatment of citric acid in the presence of TiO substrate
2
particle leads to relatively low carbon yields ͑typically 3.4 ± 0.5%
after 10 h of heating at 900°C in Ar atmosphere͒. This relatively
low yield is quite suitable for battery purposes where a compromise
between the content of inactive material and the total conductivity is
required. In any case, the yield can be increased up to 10%, if the
initial heating rate is decreased from 10 to 1°C. Between 500 and
1
1
1
2
2
Mater. Chem., 12, 1391 ͑2002͒.
1. B. Tryba, T. Tsumura, M. Janus, A. W. Morawski, and M. Inagaki, Appl. Catal., B,
50, 177 ͑2004͒.
8
00°C the yield is independent of the heating temperature, but at
higher temperatures it starts to decrease. Various mechanisms for
this behavior are proposed. Above ca. 2 wt % the carbon phase
forms more or less homogeneous coatings around the substrate par-
ticles, such as those observed in our previous work on olivines.
Conductivity as a function of carbon content exhibits a typical per-
22. T. Tsumura, N. Kojitani, H. Umemura, M. Toyoda, and M. Inagaki, Appl. Surf. Sci.,
96, 429 ͑2002͒.
23. M. Inagaki, Y. Hirose, T. Matsunaga, T. Tsumura, and M. Toyoda, Carbon, 41,
619 ͑2003͒.
1
9
2
24. K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, John
Wiley & Sons, New York ͑1988͒.
25. B.-S. Jeong, D. P. Norton, and J. D. Budai, Solid-State Electron., 47, 2275 ͑2003͒.
colation behavior with a percolation threshold of about 1–2 wt %
−
1
and a conductivity plateau within the range of 0.1–1 S cm . De-
creasing the heating rate moderately improves the conductivity at a
given carbon content. To achieve a full conductivity of the carbon-
ized phase, the composite material has to be treated at a temperature
of at least 700°C. The Arrhenius plot for conductivity of the heat-
treated TiO -citric acid sample shows a similar shape and a similar
value for activation energy ͑0.04–0.13 eV͒ as a range of other ma-
2
6. S. Rul, F. Lefevre-Schlick, E. Capria, C. Laurent, and A. Peigney, Acta Mater., 52,
061 ͑2004͒.
1
27. J. K. W. Sandler, J. E. Kirk, I. A. Kinloch, M. S. P. Shaffer, and A. H. Windle,
Polymer, 44, 5893 ͑2003͒.
8. D. McLachlan, J. Electroceram., 5, 93 ͑2000͒.
9. R. O. Grisdale, J. Appl. Phys., 24, 1288 ͑1953͒.
30. M. Koos, S. H. S. Moustafa, E. Szilagyi, and I. Pocsik, Diamond Relat. Mater., 8,
1919 ͑1999͒.
2
2
2
Downloaded on 2012-11-11 to IP 132.236.27.111 address. Redistribution subject to ECS license or copyright; see www.esltbd.org