Abstract
C-free 
crystalline powders were prepared by a synthesis method based on direct precipitation under atmospheric pressure. The particle size distribution is extremely narrow, centered on ca. 
. A soft thermal treatment, typically at 
for 
under slight reducing conditions was shown to be necessary to obtain satisfactory electrochemical 
deinsertion/insertion properties. This thermal treatment does not lead to grain growth or sintering of the particles, and does not alter the surface of the particles. The electrochemical performances of the powders obtained by this synthesis method are excellent, in terms of specific capacity (
at 5C-rate) as well as in terms of cyclability (no significant capacity fade after more than 400 cycles), without the need of carbon coating.
Olivine-type 
, namely, triphylite, was first proposed by Padhi et al. in 1997 to be used as a positive electrode material for Li-ion batteries.1 
deinsertion/insertion of this compound occurs at a potential value of ca. 
vs 
with a high theoretical specific capacity of 
. The main drawback of this cheap and nontoxic material is its low gravimetric density besides a poor electrical conductivity which limits the 
deinsertion/insertion rates, and hence the practical specific capacity. To overcome this problem, several chemical routes so as to produce carbon coatings at the surface of 
particles were proposed.2 Besides increasing the overall conductivity of the material through an electronic percolating network around the particles, one generally agrees that such coating also prevents particle growth and sintering during annealing treatments.3–6
The presence of carbon has a dramatic impact though on the tap density of the powder: it is reduced by 2 when only 
carbon is present in the composite material, which gives energy densities only half of those of standard materials such as 
.7 Moreover, a recent comparative electrochemical study and kinetic modeling of 
powders from different sources concluded that the average particle size of the materials, as well as their distribution (namely, the psd), must be narrowed in order to obtain an ideal material.8
AC and dc conductivity measurements of C-free 
dense pellets suggest that carbon coating is far from being the only route to explore for producing a more efficient 
electrode. 
conductivity is indeed likely to be in the same order of magnitude or even lower than the electronic one.9 A recent DFT contribution of Maxisch et al. confirms our experimental data, as a transport mechanism involving (
, 
') and (
', 
) exciton-like quasi particles (in 
and 
, respectively) is envisaged.10 Limitations appear then to be both ionic and electronic which strengthen the importance of tailoring as small particles as possible so as to shorten both electronic and ionic paths within the particles.
Along that line, Chimie douce is a well-known method to prepare materials having controlled particle size with narrow size distributions. Among the numerous recent contributions on the precipitation of iron-based nanocompounds, one may quote 
and 
,11, 12 nanoparticles of amorphous hydrated iron phosphates of formula 
.13 Regarding 
, soft chemistry has been restricted so far to the precipitation under hydrothermal conditions.14–22 Tajimi and Nuspl demonstated first that carbon coating was not any longer essential for providing good electrochemical activity: through the use of surfactants such as PEG, carbon-free 
particles ranged between 0.5 and 
with a specific capacity of 
at a current density of 
.20 Nuspl et al. reported on a carbon-free powder with a narrow particle size distribution (average in the 400 to 
range) that could deliver 
at 8C rate.17–19
Experimental
Based on our recent findings on the direct precipitation of crystalline 
with particle size around 100 to 
, under atmospheric conditions,23 we used a similar approach for the synthesis of carbon-free 
particles at low temperature under atmospheric pressure.24 To this end, we first undertook a detailed thermodynamic study of the 
system following our extensive previous studies on the precipitation of 
allotropes.25 Precipitation of 
may occur at an optimal pH value close to neutrality as plotted in the solubility-pH diagram shown in Fig. 1. A mixture of 
of 
sulfate (
, Sigma Aldrich) and 
(Baker) was neutralized to this average pH value by slowly adding 
of 
(
, Alfa Aesar), leading to a green mixture which was subsequently kept under refluxing conditions for ca. 
under magnetic stirring. The green-grayish precipitate thus obtained was then recovered by centrifugation, washed several times with distilled water and acetone and finally dried in an oven at 
for one day (Fig. 2). For electrochemical measurements, the precipitate was previously annealed for 
at 
under a 
gas flow in order to fully dehydrate the material and reduce the possible 
-containing impurities.
Figure 1. (a) Distribution diagram of solid compounds in equilibrium with the solution (molar ratio). (b) Solubility diagram of phosphate. (c) Solubility diagram of 
. Conditions: 
, 
, an hypothetical pKs value of 12 was chosen for 
.
Figure 2. XRD diagram of the as-obtained precipitate after 
of reaction under refluxing conditions. Inset: SEM image of the particles, having size comprised between 100 and 
.
X-ray diffraction (XRD) experiments were performed on a Philips PW 1710 diffractometer (
, 
radiation, back monochromator). Powders' morphology was observed by scanning electron microscopy (SEM) by means of a Philips FEG XL-30 and by transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) by using a FEI Technai F20 S-Twin. Particle size distribution (psd) was determined from image analysis of SEM pictures of the 
powder. Electrochemical lithium deinsertion/insertion tests were performed either in Swagelok-type cells or in coin cells, which were both assembled in an argon-filled dry box. The negative electrode was a disk of lithium metal foil. A Whattman GF/D borosilicate glass fiber sheet, soaked with a 
salt dissolved in 1:1 EC/DMC solution (Merck, LP30), was placed between the two electrodes. The positive electrodes simply consisted in a mixture of the C-free 
obtained from the precipitation and ketjen black carbon (EC-600 JD, Akzo Nobel), by means of a mortar and a pestle. When used in a Swagelok, the positive electrode material thus obtained was directly led onto the current collector (typical loading of 
for 
current collector area). For coin cells, a slurry of the electrode material and N-methylpyrrolidone (NMP) was made and deposited onto the current collector. The electrodes were left for at least 
in an oven at 
in order to fully evaporate the NMP. The typical loading in this case was comprised between 2 and 
(for 
current collector area). Lithium deinsertion/insertion was monitored either with MacPile or VMP cycling/data recording systems (Biologic SA, Claix, France), operating in galvanostatic mode.
Results and Discussion
The XRD pattern of the precipitate obtained for 
aging under refluxing conditions is given in Fig. 2. It was entirely indexed in the space group Pnma reported for 
(ICSD no. 200155) with lattice constants 
Å, 
Å, and 
Å, i.e., smaller than usually encountered for 
prepared through more classic ceramic routes. We do not know yet what is at the exact origin of such behavior. Unlike 
,23 the nucleation and growth of the olivine-type phase 
occurs very quickly as crystalline samples were obtained after as small as 
under refluxing conditions. The nucleation step is thus likely to be predominant over the growth, which leads to very small particle size comprised between 100 and 
, as illustrated in the inset of Fig. 2.
Prior to perform electrochemical characterizations of the 
powder, a heat-treatment under reducing conditions (typically a 
gas flow) was ensured so as to reduce the small amount of 
together with the removal of parasitic "OH" groups through water departure. Indeed, Mössbauer spectroscopy revealed that the pristine 
powders may contain up to 
of 
. The presence of 
groups was attested by Fourier transform (FTIR) measurements: a weak absorption band was detected at 
and was related to 
bond stretching into the amorphous 
phase, from a comparison with the FTIR spectrum of crystalline 
(Tavorite, ICDD no. 41-1376) (not shown here). This also suggests that slight amounts of amorphous 
composition are present within the powder, which is concordant with the presence of 
. Figures 3a and 3b are SEM and TEM micrographs of a 
sample heat-treated at 
for 
under 
gas flow. Note that the temperature and the dwell time of the thermal treatment are significantly reduced compared to a ceramic synthesis process. Coarsening of 
particles is hence significantly reduced, and the average particle size remains in the range of 
. Note, however, that the particle morphology has changed during the thermal treatment, from parallelepipeds for the as-made precipitate to spheres for the annealed powders. From SEM analyses, the particle size distribution (psd) could be determined (Fig. 3c). The d50 value is 
, while the relative span, defined as (d90-d10)/d50, is about 0.50. The psd is much narrower and shifted toward smaller values than that reported by Nuspl et al. for 
hydrothermally synthesized.17, 18
Figure 3. (a) SEM, (b) TEM, and (c) volumetric particle size distribution (obtained from image analysis of SEM images) of the C-free 
powder obtained after a heat-treatment at 
under 
atmosphere for 
.
Surface reactivity is generally enhanced for divided powders, especially during thermal treatments. To identify the possible existence of parasitic phases such as iron phosphides at the surface of the heat-treated 
powders, HRTEM observations coupled with EDX analyses in STEM mode across 
particles were carried out (not shown here). The absence of any amorphous layer at the surface of the particles together with the constancy of the 
atomic ratio from the particle core to the surface clearly accounts for the absence of an iron phosphide layer around the particles. This conclusion was strengthened by ac conductivity measurements on dense sintered pellets of the material which show that the electrical conductivity value of the bulk is comparable to that reported in Ref. 9 for pure 
, prepared through a ceramic route (not shown here).
As shown in Fig. 4, 
deinsertion/insertion of the 
powder mixed with only 
of ketjen black carbon proceeds with an extremely small polarization, even under high current rates (C/2 during discharge), leading to a high reversible specific capacity of ca. 
. This excellent performance is mostly due to the lowering of the particle size, and hence we wish to stress that carbon coating is no longer required for such a divided material. Note also that no significant capacity fade was observed, even after more than 400 charge/discharge cycles: this may be due to the observed fact that 
particles are highly monodisperse (very narrow psd centered on 
) leading to an homogeneous current distribution within the electrode. Figure 5 shows signature curves obtained by varying several parameters, such as the amount of carbon in the positive electrode, as well as the type of positive electrode and cell configuration used. At low charge/discharge rates, an increase of the carbon amount from 5 to 16.7% leads to a constant increase of the specific capacity of 
, whatever the C-rate used. For higher charge/discharge rates (typically 
for cells made with powders), the influence of the carbon amount becomes more important. Even at 
rate, discharge specific capacities as high as 135 and 
are obtained for 
electrodes made from NMP-based slurries, and containing 5 and 
ketjen black carbon, respectively.
Figure 4. Galvanostatic charge/discharge profiles at 
of an electrode composed of C-free 
(heat-treated powder) mixed with 
of ketjen black carbon. The charge rate (
per formula unit exchanged in 
) is lower that than of discharge (
per formula unit exchanged in 
). Inset: Specific capacity retention of the active material.
Figure 5. Evolution of the specific capacities relative to the active material as a function of C-rate, at 
. Electrodes made from NMP slurries are composed of 
of active material 
ketjen black carbon and are cycled in coin cells. Electrodes directly made from powders of AM 
ketjen black carbon are loaded with 
and are cycled in Swagelok-type cells.
Conclusion
A soft chemistry method allowed the preparation of C-free 
particles in the range 
, with a very narrow particle size distribution. This material, after a heat-treatment under slightly reducing conditions, exhibits very satisfactory electrochemical properties in terms of specific capacity and capacity retention upon cycling. These properties are directly linked to the small particle size, which lowers both ionic and electronic transport within the particles. The performances reported here, in terms of specific capacity related to the active material, are identical or slightly higher than those reported for C-coated 
.3, 5, 26 They are therefore superior in terms of energy density due to the absence of C-coating. To conclude, besides being very attractive on a practical aspect, 
–type electrodes bring us new insights about the importance of size effects on the electrochemical activity. By transposing the same concepts to more insulating materials such as 
, one would expect that a huge decrease of the particle size should be the key point for an enhancement of its electrochemical activity.
Acknowledgments
The authors gratefully acknowledge J.-M. Tarascon and P. Gibot for fruitful discussions, as well as L. Laffont for TEM observations.
Université de Picardie Jules Verne assisted in meeting the publication costs of this article.