Melt casting LiFePO4: I. Synthesis and characterization

Article abstract of DOI:10.1149/1.3284505

A melt casting process to make an electrochemically active LiFeP O ₄ cathode material was explored. The melting of carbon-coated LiFeP O₄ powder at 1000°C followed by its cooling leads to a high purity LiFeP O₄ material wi

Full text of DOI:10.1149/1.3284505

Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
A453
0013-4651/2010/157͑4͒/A453/10/$28.00 © The Electrochemical Society
Melt Casting LiFePO4
I. Synthesis and Characterization
M. Gauthier,a C. Michot,a N. Ravet,a M. Duchesneau,a J. Dufour,a G. Liang,a,
*
,z
J. Wontcheu,b L. Gauthier,b and D. D. MacNeilb,
*
aPhostech Lithium, Inc., St-Bruno de Montarville, Québec J3B 6V7, Canada
bDépartement de Chimie, Université de Montréal, Montréal, Québec H3T 1J4, Canada
A melt casting process to make an electrochemically active LiFePO4 cathode material was explored. The melting of carbon-coated
LiFePO4 powder at 1000°C followed by its cooling leads to a high purity LiFePO4 material with excellent crystallinity and large
crystals. From a detailed study on the melting of naturally occurring triphylite, elements such as Mg and Mn were easily
incorporated into the olivine structure, while others such as Ca, Si, and Al were segregated into grain boundaries. Experiments also
demonstrated that a variety of synthetic precursors can be used to make high purity LiFePO4 materials through melt casting. The
phase purity, crystallinity, and microstructures of the final product depend on the reducing conditions and the solidification process
of the melt. The melt-casted material, after coarse grinding, shows good electrochemical activity.
© 2010 The Electrochemical Society. ͓DOI: 10.1149/1.3284505͔ All rights reserved.
Manuscript submitted October 15, 2009; revised manuscript received December 7, 2009. Published February 26, 2010.
The 1980 Goodenough discovery of the high voltage, lithium-ion
reversible, layered LiCoO2 material led, 10 years later, to the com-
mercial deployment of the lithium-ion battery.1 This battery was first
commercialized by Sony and is now used in most portable electronic
equipment.2,3 Twenty years after his discovery of LiCoO2, Good-
enough proposed the use of an olivine material, LiFePO4, as a low
cost, widely available, and environment-friendly substitute to the
cobalt-based cathode used at the time.4 What made this naturally
occurring compound especially attractive as a cathode material was
the unexpected high voltage of the Fe3+/Fe2+ redox couple at about
3.4 V vs a lithium anode. The higher than expected voltage was due
to the inductive effect of the P–O–M covalent bonding on the M3+/2+
redox couple. Furthermore, Goodenough and co-workers explained
the importance of the crystal field energy of certain polyanionic
structures, especially olivine and NASICON, toward electrochemi-
cal performance.5
With the recent emergence of large lithium-ion batteries for
power and energy storage applications, additional benefits of
phosphate-based cathode materials, as originally pointed out by
Goodenough in his patent application, became obvious to the scien-
tific community.4,5 This includes his suggestion of nanoscale par-
ticles to address the intrinsically limited mass transport properties
and mechanical stress within the polyanion structure. This sugges-
tion has led to the use of either pure or so-called doped LiFePO4
cathode formulations for applications that require high power charge
and discharge rates. Examples of such applications are portable
power tools, hybrid-electric vehicles, and military applications.6,7
Another property that can lead to the development of large LiFePO4
batteries rests in the covalent bonding of the oxygen with phos-
phorus within the phosphate polyanion. This bonding environment
provides high thermal stability. Tests on commercial 26650 cells
demonstrated that the strong covalent bonding within LiFePO4 sup-
presses the release of oxygen from the cathode at elevated tempera-
tures. This suppression leads to a cell with improved safety charac-
teristics when compared to cells which use metal oxide cathodes, as
oxygen released at elevated temperatures leads to thermal excursion
that can lead to thermal runaway in real cells.8,9
cursors, the final product, or during composite cathode formulation.
However, most electrode developers have found that optimal elec-
trochemical performance is obtained when an organic carbon pre-
cursor is used to generate a pyrolytic conductive carbon deposit via
heat-treatment during or after the LiFePO4 synthesis.13-15 Further-
more, when the organic precursor is pyrolyzed during the LiFePO4
synthesis, it brings the additional benefit of limiting particle growth
and sintering, especially when present at the submicrometer level.
Also, the pyrolytic reaction generates organic gases that help control
the reducing environment and maintain the iron at its 2+ oxidation
state in LiFePO4. This leads to a final carbon-coated LiFePO4
͑C-LiFePO4͒ product that can be present as an individual particle or
as an agglomerated form.16
Many types of syntheses have been explored in the last 10 years
to produce an optimal LiFePO4 powder for cathode formulation and
performance optimization. Table I is our attempt to regroup and
illustrate the different chemical routes into three process categories.
Each of these categories has intrinsic advantages and disadvantages
in view of their electrochemical performance, ease of processing,
and cost optimization. Comments are given for each process cat-
egory as to their impact on electrochemical performance ͑which is
strongly linked to the control of individual or agglomerated particle
size distribution͒, their purity, presence of secondary phases, their
conductivity, and finally, their manufacturing cost. An inherent dif-
ficulty with all solid-state processes rests in achieving a homoge-
neous powder mixture that avoids parallel reactions and inhomoge-
neities during the synthesis reaction. Trivalent ions, such as Fe3+ or
PO34−, have extremely low diffusion coefficients in the solid state.
Unless a temporary liquid phase exists during the reaction to im-
prove mass transport, there is a risk that an unreacted ͑or incom-
plete͒ product͑s͒ remains. Long dwell times and higher temperatures
may help push the reaction to completion, but they also favor par-
ticle sintering, especially for submicrometer particles without a pro-
tective coating ͑such as the product of carbon pyrolysis͒. Solvent-
assisted syntheses such as hydrothermal reactions, which include
precipitation and agglomeration steps, are a somewhat better solu-
tion to traditional solid-state reactions because the liquid media
maintain the reactants in close proximity, thus reducing the possibil-
ity of by-products or partially reacted precursors. In addition, hydro-
thermal reactions provide an improved particle size control and a
narrow particle size distribution compared to solid-state methods.
Synthesis in the molten state is also presented for the sake of com-
parison with solid-state and solvent-assisted processes. This paper
expands upon the unique characteristics of this synthesis method.
Synthesis within a molten state combines ideal liquid-phase reaction
kinetics and short dwell times to provide a product with high mate-
rial density using a simple process method. With the increasing de-
mands for electric vehicles, a simple synthetic method using com-
In addition to particle size reduction, the high power capability
of polyanionic oxides, such as LiFePO4, was made possible through
the use of a conductive carbon additive, added either after or during
synthesis.10-12 The conductive additive was required because these
covalently bounded complex oxides ͑phosphates͒ are insulating ma-
terials. The carbon additive, such as carbon blacks or graphitic pow-
ders, can be added through a number of methods to either the pre-
*
Electrochemical Society Active Member.
z E-mail: dean.macneil@umontreal.ca
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A454
Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
Table I. LiFePO4 synthesis families, advantages, and limitations.
Family
Typical reactants
Reference
Advantages
Limitations
Solid-state
reactions
Li2CO3 + Fe͑CH3CO2͒ + ͑NH4͒H2PO4
4,17
14,18
12
16
11,19,20
• Simplicity: mixing + heating
• Low cost equipment ͑furnace͒
• C-LiFePO4 possible in one
step synthesis with an organic
reactant
• Depends on trivalent ion
diffusion
• Rests on solid reactant
mixing quality
• Incomplete reaction or
nonreversible side reaction
• Cost of precursors ͑FePO4,
FeC2O4, etc.͒
2
Li3PO4 + Fe3͑PO4͒2·xH2O
Fe2O3 + LiHPO4 + C͑s͒
FePO4·xH2O + Li2CO3 + copolymer
FeC2O4·xH2O + NH4H2PO4 + Li2CO3
• Possibility to use reactant
combinations, e.g., FePO4
Solvent-assisted
synthesis/
precipitation
FeSO4·xH2O + LiOH + H3PO4 ͑water͒
21-23
24
• Low cost reactants
• Complete reaction in liquid
phase
• Quality control and
chemical purity control
possible at every step
• Particle size, morphology, and
purity control by precipitation
• Submicrometer particles for
power applications
• Waste recycling cost
• Capital-intensive
manufacturing process
• Requires large-scale
installations
FeSO4·xH2O, H3PO4, H2O + DMSO, LiOH
Molten state
Fe2O3 + Fe + LiH2PO4
Fe3͑PO4͒2·2H2O + Li3PO4
Fe2O3 + Li2CO3 + NH4H2PO4
FeO + P2O5 + LiOH
25
26
• Process simplicity, no waste
water
• Rapid and complete reaction
at equilibrium in a liquid
phase
• Requires micrometer and
submicrometer powderization
to get product for use in
power applications
• Solid–liquid phase separation
for purity
• Possibility of using
commodity reactants
• High active material density
• Allows certain substitutions/
stoichiometries and
heat-treatment/cycles
modity precursors is required for the large-scale synthesis of
LiFePO4. It is believed that the molten synthetic method is a viable
alternative to the current methods of synthesis.
and 1050°C. The small laboratory samples ͑ Ͻ 1 kg͒ were cooled
to room temperature within the box furnace under the protection of
an Ar atmosphere. Casting of large ingots was performed by pouring
the molten liquid into a graphite mold in air. Numerous material
compositions have been used for synthetic tests and the specific
details are expanded within the Discussion section, but a summary
can be found in Table II.
X-ray analyses of the materials were performed on a D500 Si-
emens or a Bruker D8 advance diffractometer using Co K␣ radia-
tion. For X-ray analysis, the ingot obtained from the synthesis was
first crushed and then ground in a mortar with a pestle to a powder
with micrometer-sized diameters. The phase purity and lattice pa-
rameters of the sample were calculated using the Topas Rietveld
refinement program.
The microstructure of cast materials was characterized with an
S-4700 Hitachi scanning electron microscope ͑SEM͒ equipped with
an energy-dispersive X-ray microanalysis system and a field-
emission gun. Small samples were fixed in epoxy and then cut and
polished for SEM observation. Thermogravimetric analysis ͑TGA͒–
differential scanning calorimetry ͑DSC͒ analyses were performed
using an SDT600 from TA Instruments coupled with a mass spec-
trometer ͑Pfeiffer͒ for off-gas analysis. The heating rate was
10°C/min, and the experiments were performed with a flowing He
carrier gas. Carbon content for the samples was analyzed using an
LECO C, S analyzer ͑LECO-SC632͒.
Electrochemical activity of the LiFePO4-melt phase was tested in
the following manner. A 5 g LiFePO4-melt sample was crushed and
ground in an agate mortar and subsequently C-coated using a
1,4,5,8-naphthalenetetracarboxylic dianhydride treatment as de-
scribed in Ref. 28. A 3.19 g portion of the ground sample was dis-
persed in a 10 mL acetone solution containing 0.16 g of the dian-
hydride carbon precursor. After evaporation, the mixture was heated
Triphylite olivines, of the general formula Li͑Fe–Mn͒PO4, are
found in pegmatite rocks as one of the last fractions to crystallize
from molten basic magma. They also have electrochemical activity
similar to a synthetic one when properly ground and carbon
coated.27 However, the conditions of temperature, pressure, and
chemical environment ͑water and redox atmosphere͒ in which these
natural crystals are formed are not easily reproduced within the
laboratory. The results of Ref. 27 provide an indication that LiFePO4
can be produced at molten conditions without significant decompo-
sition. Thus, we undertook a program to explore this alternative
method to make LiFePO4 through a molten process. The investiga-
tion began by simply melting presynthesized C-LiFePO4 powder,
followed by casting. Versatility of this preparative technique was
also established using different chemical precursors.
In this first part of a two-part paper, we describe the synthesis
conditions and product characterizations which are required to un-
derstand the basic processes at play as well as to confirm the elec-
trochemical activity for LiFePO4 synthesized via a molten process.
In addition to the preparative advantage of the molten process ͑va-
riety of precursors, liquid media, fast kinetics, etc.͒, the technique is
also a tool to study the phase diagrams of pure and substituted
LiMPO4 compositions and stability domains of different Fe2+/Fe3+
species as well as their systematic electrochemical characterization
by the use of well-defined particulate products for comparative ki-
netic studies with or without the presence of a carbon coating.
Experimental
Melt casting was performed within an airtight box furnace under
an Ar or N2 protection in a graphite or alumina crucible between 950
Downloaded on 2014-10-30 to IP 128.148.252.35 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, 157 ͑4͒ A453-A462 ͑2010͒
A455
Table II. List of LiFePO4-melt syntheses.
Synthesis
Starting materials
XRD analysis
Comment
FePO4·2H2O + 1 Li2CO3 + 0.5 wt %
͑1͒
91.7% LiFePO4,
3.2% FePO4,
Graphite crucible,
5 g mixture,
2
PE/PEG copolymer
1.6% LiFeP2O7,
1.4% LiFe5O8
1 h and 30 min at 980°C, Ar
͑2͒
͑3͒
͑4͒
Fe2O3 + 2NH4H2PO4 + Li2CO3
94% LiFePO4
95% LiFePO4
Graphite crucible,
50 g, 1 h at 980°C, Ar
Fe3O4 + 3NH4H2PO4 + 3 Li2CO3
Graphite crucible,
50 g, 1 h at 980°C, Ar
2
FePO4·2H2O + 1 Li2CO3
92% LiFePO4,
Graphite crucible,
2
8% Li3Fe2͑PO4͒
10 kg, 2 h at 980°C, 15 h cooling under Ar
3
͑5͒
͑6͒
͑7͒
͑8͒
Fe2O3 + 2NH4H2PO4 + Li2CO3
Fe2O3 + Fe + 3LiH2PO4
Fe3O4 + Fe + 4LiH2PO4
Fe3͑PO4͒2·8H2O + Li3PO4
94% LiFePO4
LiFePO4
Same as ͑3͒ but 2 kg
Al2O3 crucible, 1000°C, Ar
Al2O3 crucible, 1000°C, Ar
Al2O3 crucible, 1000°C, Ar
LiFePO4
LiFePO4,
Li3Fe2͑PO4͒
3
FePO4·2H2O + 1 Li2CO3
͑9͒
94% LiFePO4,
3% LiFeP2O7,
2% Fe2P2O7,
1% Li4P2O7
Graphite crucible + graphite powder in excess
10 kg and 40 kg ingots, 1000°C, N2, 4 h
casting, and cooling under N2 for ϳ10 h
2
͑10͒
Fe2O3 + LiH2PO4 + MgO ͑2 wt %͒
+MnO ͑2 wt %͒
Li͑Fe,Mn,Mg͒PO4
Graphite crucible, 4 h plateau at 1000°C,
Ar, cooling for 12 h
in a flow of CO2/CO gas ͑50:50 ratio͒ within a rotary chamber
suspended in a tube furnace. The gas mixture was circulated for
20 min at ambient temperature to ensure a complete purge of the
system. The sample was then heated to 650°C in 100 min and main-
tained at that temperature for 1 h before cooling. The concentration
of carbon after pyrolysis was determined to be 0.33% via LECO
analysis.
A cathode slurry was made by dispersing 101 mg of the
C-LiFePO4-melt sample in acetonitrile with 82 mg poly͑ethylene
oxide͒ polymer ͑400,000 Mw͒ from Aldrich and Ketjenblack
͑Akzo-Nobel͒. The slurry was then coated on a stainless steel sup-
port ͑1.54 cm2 area͒ to give a solid composition of 41 wt % poly-
͑ethylene oxide͒ composition, 7.5 wt % Ketjenblack, and 51.5 wt %
C-LiFePO4 melt. This composition, although not ideal for commer-
cial applications, allows us to determine if electrochemical activity
is possible for a sample synthesized via the molten process. A coin
cell was then assembled within an Ar-filled glove box using 1.97 mg
of active cathode material, a 500,000 Mw poly͑ethylene oxide͒ ͑Al-
drich͒ containing 30 wt % LiTFSI ͑3M͒ electrolyte, and a lithium
foil anode ͑FMC͒. The battery was tested with a VMP2 multichanel
potentiostat ͑Bio-Logic Science Instruments͒ at 80°C, with a
20 mV/h scan rate between 3 and 3.7 V vs Li.
under 55,000 lbs for 5 min to give a pellet Ϸ3 in. in diameter with
a thickness of Ϸ8 mm. This pellet was then deposited on an alumina
ceramic plate and heated, in an airtight oven maintained under argon
with a continuous flow of gas, from ambient temperature to
950 Ϯ 30°C in 4 h. The product was left for 4 h at 950 Ϯ 30°C
and then cooled to ambient temperature in 8 h. As shown in Fig. 1,
a gray mineral phase with large crystals was obtained, in addition to
a black crust. The black crust was later confirmed ͑via LECO analy-
sis͒ to be carbon. This crust maintained the shape of the original
pellet but with a smaller diameter ͑ Ϸ 1 in.͒. The mineral phase was
separated from the attached carbon crust and ground in a mortar for
X-ray diffraction ͑XRD͒ analysis. Surprisingly, the mineral phase
was essentially LiFePO4 ͑identified as “LiFePO4 melt”͒ containing a
small amount of secondary phases of Li3PO4 and Li4P2O7 ͑visible in
the XRD pattern͒. The well-defined XRD spectra of the powdered
LiFePO4 melt are provided in Fig. 2 as compared to the XRD spec-
tra of the initial C-LiFePO4 powder used for the molten synthesis.
Remarkably, a high phase purity for LiFePO4 was obtained after
the melting treatment. No Fe2P phase could be detected by XRD
analysis of the bulk LiFePO4-melt sample. Achieving a high purity
of LiFePO4 after melting is likely linked to the purification through
liquid-phase separation and the expulsion of solid impurities ͑carbon
among others͒ from C-LiFePO4. An LECO experiment ͑C analysis͒
performed on the LiFePO4-melt phase showed no carbon present.
This confirms the separation of the LiFePO4 liquid-molten phase
from the carbon coating during the molten process. The XRD analy-
sis ͑not shown͒ of the C crust shown in Fig. 1 clearly identifies the
presence of C and Fe2P associated with some residual LiFePO4,
lithium pyrophosphate, and lithiated iron oxide species. This further
indicates that the formation of Fe2P occurs at the vicinity of the C
additive and was expelled from the LiFePO4 crystals during solidi-
fication. This example clearly demonstrates the feasibility of prepar-
ing LiFePO4 in the molten state without any significant decomposi-
Results and Discussion
Melting of C-LiFePO4 powders.— The first evidence of pure
crystalline LiFePO4 from a molten synthesis technique was obtained
by melting C-LiFePO4 in the following manner: A low carbon con-
tent C-LiFePO4 powder ͑Ϸ1.6 wt % pyrolytic carbon deposit͒
sample from Phostech Lithium Inc. was used as the starting mate-
rial. The powder was previously synthesized through a solid-state
reaction of FePO4·2H2O, Li2CO3, and an organic carbon precursor
in a reducing atmosphere, according to Ref. 14 and 16. About 30 g
of this C-LiFePO4 was mixed with 10 wt % glycerin and pressed
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A456
Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
Figure 3. ͑Color online͒ Slow scan cyclic voltammetry of the C-LiFePO4
͑solid line͒ precursor compared to that of the LiFePO4-melt phase ͑dashed
line͒.
Figure 1. ͑Color online͒ Photograph of a C-LiFePO4 pressed pellet after
melting at 1000°C. A clear phase separation between a liquid-molten phase
͑LiFePO4͒ and a carbon-containing solid crust is visible.
mmetries of the samples are nearly identical. The first discharge
coulombic efficiency of the molten sample ͑92.3%͒ is close to the
purity of the LiFePO4 phase ͑92%͒ as determined by XRD analysis.
These electrochemical tests, although performed on dense, nonopti-
mized particle sizes, demonstrate that LiFePO4 via a molten process
behaves essentially like LiFePO4 made by solid-state syntheses as
referred to in Table I.
tion of the material. Furthermore, the phase separation, which was
observed during the melting process, has the beneficial effect of
separating LiFePO4 from residual impurities or decomposition prod-
ucts associated with the thermal treatment. Another practical benefit
of this molten process is that it leads to LiFePO4 in a dense crystal-
line form.
Metallographic studies reveal LiFePO4 dendrites with a crystal-
lite size larger than 200 ␮m wide and 3000 ␮m long. In between
these dendrites, secondary phases are observed, which likely form
from low melting point ͑mp͒ species, which solidify during the last
stage of cooling. For instance, Li4P2O7 has an mp of 885°C, which
is about 80°C lower than that of LiFePO4.29 These details are ad-
dressed in the latter part of this paper.
Voltammetry scans are reported in Fig. 3 for the LiFePO4-melt
sample as compared to the starting material ͑C-LiFePO4͒ that was
tested under similar conditions. Clearly, the initial slow scan volta-
Melting of a natural triphylite Li͑Fe–Mn͒PO4 mineral
.—
Naturally occurring triphylite ore is scarce, but there is some interest
in evaluating the melting of this ore to increase the purity of
LiFePO4. Consequently, triphylite ore was purchased from Excali-
bur Mineral Corp. ͑Peekskill, NY͒. The XRD analysis indicates that
the ore contains approximately 65% LiFePO4, 15%
Fe3͑PO4͒2͑OH͒2, 7% Fe3͑PO4͒2, and other minor impurity phases.
A piece of this triphylite ore ͑ Ϸ 1 cm3͒ was deposited on an alu-
mina ceramic plate and heated under a flow of argon from ambient
temperature to 980°C in 100 min and maintained at that temperature
for 1 h followed by cooling to ambient temperature in 3 h. After
cooling, an Ϸ1 in. diameter glossy ore deposit was obtained in the
form of a deep green mound. Interestingly, the top crust seemed to
indicate a different composition, so it was decided to determine the
distribution of the main elements on its surface by energy-dispersive
X-ray analysis ͑EDX͒ on a SEM. The melted triphylite ore was
sliced and mounted in a resin for microstructure analysis. Cartogra-
phies of the elements on the top of the melted triphylite sample are
shown in Fig. 4. Oxygen is not shown in Fig. 4, but its presence was
seen equally throughout the entire sample without any segregation
into discrete phases or areas. Figure 4 clearly indicates that the
melting of the ore induced a liquid–liquid and liquid–solid phase
separation visible at the surface. Well-defined square crystals of FeO
͑upper right corner͒ are surrounded by an immiscible silicon-based
molten phase. In addition, an immiscible calcium-based phase ͑cen-
ter left͒ can be clearly distinguished from the bulk. The main phase
consists of Li͑Fe,Mn,Mg͒PO4 with relative 25:9:1 atomic ratio of
Fe, Mn, and Mg ͑not shown͒, as a consequence of Mn and Mg
miscibility within both the LiFePO4 liquid and solid phases.
A subsequent analysis was performed with the sample embedded
in epoxy, cut diagonally using a diamond-cutting tool. The sample
͑cross section͒ was again analyzed under the SEM and X-ray detec-
tor, the results of which are shown in Fig. 5a and b. The figure
provides mappings of the cut sample slice for Fe, P, Mn, Si, and Al
Figure 2. ͑Color online͒ XRD powder pattern of the initial C-LiFePO4 pow-
der as compared to the LiFePO4-melt phase.
Downloaded on 2014-10-30 to IP 128.148.252.35 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, 157 ͑4͒ A453-A462 ͑2010͒
A457
Mn
SEM
P
Fe
(a)
Si
Ca
Figure 4. ͑Color online͒ SEM micrographic cartographies of the top surface
of the melted triphylite ore showing the elemental distributions for Mn, P, Fe,
Si, Ca, and O.
elements and some bielemental distributions. There are clear ͑Fig.
5b͒ Si and Al inclusion phases dispersed within the main
Li͑Fe,Mn,Mg͒PO4 solid solution that were likely segregated and
solidified during the cooling of the sample. The silicon-rich phase is
composed of Si, Fe, O ͑not shown͒, Mn, and P with a relative atomic
ratio between Si, O, and P of 68:21:5, indicating both silicate and
phosphate phases dispersed within Li͑Fe,Mn,Mg͒PO4 with the
same composition as observed at the surface. The calcium-rich
phase ͑not shown in Fig. 5͒ is composed of a phosphate containing
Ca, Mn, and Fe with a relative atomic ratio of 25:25:10, possibly as
an olivine phase, dispersed within Li͑Fe,Mn,Mg͒PO4. The
aluminum-rich phase is composed of a phosphate containing Fe and
(b)
Al with
a relative atomic ratio of 14:7, dispersed within
Figure 5. ͑Color online͒ ͑a͒ SEM micrograph of a cross section of the
melted triphylite ore with elemental distributions for P, Fe, and Mn. ͑b͒ SEM
micrographs showing the elemental ͑Fe, Al, and Si͒ and bielemental ͑Fe–Al,
Si–Al͒ distributions using X-ray analysis.
Li͑Fe,Mn,Mg͒PO4. This confirms the reorganization of the ore dur-
ing melting and the separation of “impure” phases as inclusions.
These inclusions could in principle be separated from the main
Li͑Fe,Mn,Mg͒PO4 solid solution phase directly from the melt or
after cooling by various metallurgical processes ͑crushing, grinding,
screening, leaching/washing, etc.͒. This experiment confirms that
pure or substituted olivines can be prepared by melting whenever a
solid solution is possible in the liquid state and that it can be pre-
served upon cooling. In addition, Mg and Mn elements can be easily
substituted in the olivine structure, while Ca, Si, and Al are segre-
gated into secondary phases. Further melting and cooling experi-
ments could be performed to evaluate the effect of immiscible ele-
ments or phases on the structure and performance of this material.
for 1 h. The sample was then cooled naturally to room temperature,
and when it was cool, the sample was retrieved, ground, and ana-
lyzed by XRD.
Several syntheses were performed using an FePO4·2H2O precur-
sor from Fabrik Budenheim KG, Germany and Li2CO3 from
Limtech ͑Québec, Canada͒. Each of these samples required a spe-
cies for iron reduction, and thus, some samples used a polyethylene-
block-poly͑ethylene glycol͒ 50% copolymer ͑PE/PEG͒ from Aldrich
͓synthesis ͑1͔͒. Others used a graphite powder ͓synthesis ͑9͔͒, while
in other cases, simply a graphite crucible ͓syntheses ͑4͒ and ͑5͔͒ was
used. To further investigate the reactions that occur at elevated tem-
perature, a DSC-TGA run ͑SDT600, TA instruments͒ coupled with a
mass spectrometer ͑Pfeiffer͒ for off-gas analysis was performed for
synthesis ͑9͒ using graphite powder as the reducing agent, and the
results are presented in Fig. 6. The DSC-TGA-MS experiment
shows the following reactions: loss of water ͑m/z = 18͒ at ϳ200°C
and loss of CO2 ͑m/z = 44͒ from either Li2CO3 decomposition or a
LiFePO4-melt synthesis from different precursors.— After hav-
ing shown that an electrochemically active LiFePO4 can be obtained
via melting under an inert atmosphere at 1000°C, followed by cool-
ing and powderization, an investigation into the use of synthetic
precursors was initiated. Several syntheses were performed using
different iron, phosphorus, and lithium precursors. These are sum-
marized in Table II. The samples were mixed well in a mortar, and
then they were transferred to the indicated crucible. The sample was
placed in a furnace and heated to 1000°C under a flow of Ar or N2
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A458
Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
analysis reveals the presence of about 8% monoclinic Li3Fe2͑PO4͒3
in addition to LiFePO4. A metallographic examination of a cut
sample from the ingot is shown in Fig. 7. This analysis confirms the
two phases identified in the XRD profile in addition to an Fe–O
phase in smaller quantity ͑dispersed white phase seen in Fig. 7b͒. A
slow scan cyclic voltammetry ͑Fig. 8͒ of the sample powder made
by ballmilling ͑particle size Ͻ7 ␮m͒ confirms the electrochemical
activity of the two observed phases. A monoclinic Li3Fe2͑PO4͒3
phase is visible at about 2.85 and 2.7 V on discharge, and a
LiFePO4 phase is visible at 3.4 V.4 Such a large quantity of Fe3+ in
LiFePO4 is probably the result of an incomplete reaction during
scale-up: FePO4 + Li2CO3 → Li3Fe2͑PO4͒3 + FeOx → LiFePO4
that occurs in less reducing synthetic conditions.
FePO4·2H2O is a preferred starting material for solid-state syn-
theses of LiFePO4 because it chemically combines the two trivalent,
slow diffusing ions. However, FePO4 is complex to synthesize and
more expensive than the basic chemicals used in the phosphorus and
steel industry. For cost optimization, the molten synthesis of
LiFePO4 from ferric or mixed Fe3+/Fe2+ oxides, mono- or diammo-
nium acid phosphates, and lithium carbonate or lithium hydroxide
are preferred. This was performed in syntheses ͑2͒, ͑3͒, and ͑5͒
using either a thermal reduction ͓synthesis ͑5͔͒ or carbothermal re-
duction to Fe2+ ͓syntheses ͑2͒ and ͑3͔͒. In syntheses ͑6͒ and ͑7͒,
metallic iron was added as the reducing agent for the ferric oxide,
while synthesis ͑8͒ uses a previously synthesized iron ͑2 + ͒ vivi-
anite precursor with lithium orthophosphate.
Figure 6. ͑Color online͒ DSC-TGA-MS thermal study of the
FePO4·2H2O + Li2CO3 + graphite solid mixture. The solid line ͑black͒ indi-
cates the TGA profile, while the short dashed curve ͑blue͒ provides the DSC
profile. Mass spectra for an m/z of 18 ͑H2O͒, 44 ͑CO2͒, and 28 ͑CO͒ are
shown as long dashed curves and labeled in the figure.
reaction initially at 400°C and then from the carbothermic reduction
of Fe3+ to Fe2+ ͑from ϳ650 to 900°C͒. This was followed by the
melting endotherm of LiFePO4 at 960°C, and a further reduction by
C forming CO ͑m/z = 28͒ is visible at 1100°C.
When a test run for possible scale-up ͑large 10 kg synthesis͒ is
performed in synthesis ͑4͒ using FePO4·2H2O and Li2CO3 precur-
sors in a graphite crucible by heating to 980°C for 2 h, the XRD
These results ͑Table II͒ demonstrate that LiFePO4, with high
(a)
Figure 7. ͑Color online͒ ͑a͒ Microscopic
examination of the cut and polished
LiFePO4 ingot of synthesis ͑4͒ ͑Table II,
10 kg scale-up͒. ͑b͒ Chemical analysis of
the two visible phases in ͑a͒ indicate ei-
(b)
ther
a LiFePO4 ͑light͒ phase or a
Li3Fe2͑PO4͒ ͑dark͒ phase. Dispersed
3
phases of FeO are clearly visible ͑white
inclusions͒ in ͑b͒, but were not analyzed
here.
Downloaded on 2014-10-30 to IP 128.148.252.35 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, 157 ͑4͒ A453-A462 ͑2010͒
A459
Figure 10. ͑Color online͒ XRD profiles of LiFePO4 synthesized using dif-
ferent stoichiometric concentrations. Sample “crystal” is the pure single crys-
tal LiFePO4 from Reaction ͑9͒, while samples “DM4a” to “DM4d” are de-
scribed in Table III, where DM4a = Li1.05FePO4, DM4b = Li0.95FePO4,
DM4c = LiFePO4, and DM4d = Li1.05Fe͑PO4͒1.05. The symbols within the
figure are used to represent the location of possible impurities for the
samples ͑open squares = Li4P2O7, filled triangle = Li3PO4, and filled
circle = Fe2P2O7͒.
Figure 8. Slow scan voltammetry of the sample from synthesis ͑4͒ ͑Fig. 7͒,
showing electrochemical activity at low potentials for Li3Fe2͑PO4͒ and at
3
3.4 V for LiFePO4.
purity, can be obtained by the melting of various precursors of iron
and phosphorus as long as there is no loss of Li, Fe, and P species
before melting and in the molten liquid state. Surprisingly, no obvi-
ous Li loss is observed while holding the molten liquid at 1050°C
for 1–2 h. Similarly, no significant reduction of LiFePO4 by carbon
or CO to Fe2P is observed via XRD in any of the solidified ingots.
The feasibility of scaling up the synthesis of LiFePO4 by melt
casting ͑Fig. 9͒ was used to cast 40 kg ingots within a graphite
mold. During the slow cooling of the large LiFePO4 ingots, a pocket
was created inside the core of the ingot. Within these pockets, other
LiFePO4 crystals were observed to grow. These crystals were iso-
lated and characterized. The 5–10 mm long crystals did not contain
any other secondary phase and were ϳ99% pure LiFePO4 as deter-
mined by XRD and shown in Fig. 10 ͑labeled as crystal͒. Rietveld
refinement of the pure LiFePO4 crystals gave a unit cell volume of
291.5 Å3, which is approximately the same as that found in the
literature.30 Antisite defects are produced when materials are synthe-
sized under low hydrothermal reaction temperatures31 and low sin-
tering temperature of solid-state reactions,32 and a cell volume as
large as 294.5 Å3 has been reported. In the present case of large-
scale molten synthesis, the cooling rate is very slow ͑on the magni-
tude of ϳ0.1 C/s͒ in the core of the large ingot; thus, the LiFePO4
is expected to crystallize into a well-ordered olivine structure. Test-
ing by atomization of molten LiFePO4, using a high pressure water
spray to provide a rapid cooling rate ͑on the order of 1000°C/s͒,
leads to a partially amorphous phase ͑not shown͒, indicating the
influence of the solidification process on the formed crystal struc-
ture, microstructures, and composition segregation within the
sample.
Characterization of LiMPO4 compositions and the effect of sto-
ichiometry on mp and secondary phases.— Composition analysis
of the melted triphylite mineral showed that the main component
found upon cooling had a nominal Li͑Fe,Mn,Mg͒PO4 composition
in which Mn and Mg were homogeneously dispersed within the
olivine phase. An attempt was made to prepare such a LiFePO4
synthetic composition by using Fe and Li precursors in a graphite
crucible and by heating to 1000°C under argon with the addition of
2% Mg as MgCO3 and 2% MnO. After synthesis, a metallographic
examination of the ingot was performed to characterize the phases
which were observed upon cooling. In Fig. 11a not only are Mn and
Mg cations uniformly present in the olivine phase as found in nature
͑or in melted minerals͒, but two additional secondary phases ͑Fig.
11b and c͒ can also be observed. These secondary phases are local-
ized between the large Li͑FeM͒PO4 crystals. X-ray analysis of these
secondary phases indicates that these two phases ͑Fig. 11b and c͒
have a lower Fe and a higher P content than the olivine phase in Fig.
11a. One of these phases can be attributed to Li4P2O7, because XRD
made on ground samples of the bulk material confirms the presence
of a crystalline Li4P2O7 phase. At present it is difficult to interpret
the remaining phase as either a finely dispersed discrete phase or a
possible artifact from the EDX. It could very well be an artifact
given the small size of the observed area and the background inter-
ference from the major olivine component. However, some areas of
the secondary phases show different compositions and contrast on
the backscattering images, thus it is highly possible that other
lithium polyphosphate phases containing Fe or Mn are present.
More detailed analyses are currently under way and will be the
subject of a follow-up paper from our group.
The observation of these secondary phases in both XRD and
X-ray microstructural chemical analysis prompted us to characterize
LiFePO4 prepared by melting different nonstoichiometric mixtures
of precursors. All samples were prepared using identical methods
Figure 9. ͑Color online͒ Photograph of large-scale LiFePO4 casting from
synthesis ͑9͒ and the resultant 10 kg ingot.
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A460
Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
Figure 11. ͑Color online͒ Micrographic
examination of a Mg and Mn partially
substituted LiFePO4 ͓synthesis ͑10͔͒
showing ͑a͒ the bulk LiMPO4 phase,
while ͑b͒ and ͑c͒ show the presence of two
additional phases free of Mg and Mn with
varying O/P:Fe ratios.
following the composition in Table III. Each sample was mixed with
graphite powder, heated to melting at 1000°C for 1 h under an Ar
flow, and allowed to cool overnight. The major composition, as de-
termined from XRD ͑Fig. 10͒, was LiFePO4, while the possible
secondary phases ͑Li3PO4, Li4P2O7, and Fe2P2O7͒ are shown in
Table III and identified in Fig. 10. In all cases a slightly smaller unit
cell volume than that of the pure LiFePO4 crystals obtained from
synthesis ͑9͒ is observed. In addition to XRD analysis, a metallo-
graphic examination ͑Fig. 12͒ of each ingot after cooling was per-
formed to confirm secondary crystalline phases. Figure 12 clearly
shows the localization of the secondary ͑impurity͒ phases between
large LiFePO4 crystallites ͑1–3 mm by 0.2–5 mm͒. Qualitative in-
formation on the chemical composition of several crystallized
phases is given in Fig. 13 for sample DM4D ͑“LiFe0.95PO4” in Table
III͒. The analysis suggests a fractional crystallization from melt of
the remaining liquid phase present during LiFePO4 crystallization.
Many phases with different Fe–P–O ratios from LiFePO4 ͑Fig. 13a͒
are visible, some Fe-rich ͑Fig. 13b and c͒ and others made essen-
tially of P–O–Li ͑Fig. 13d and e͒. It is difficult to make an exact
attribution from EDX data within this sample because these impu-
rity phases are present only in a limited quantity vs the olivine
phase. This is in sharp contrast to that of the phases present during
synthesis ͑4͒ ͑Table II͒, in which a monoclinic phase Li3Fe2͑PO4͒3
could be easily identified electrochemically by XRD and microstruc-
ture chemical analysis ͑Fig. 7͒. These metallographic characteriza-
tions suggest that LiFePO4 crystallization from melt is controlled by
phase diagram rules and by the purity and stoichiometry of the re-
actant used. It appears that LiFePO4 crystals, which are formed
upon slow cooling, reject any excess nonstoichiometric reactants
into grain boundaries.
A DSC analysis, on the synthesized products, to confirm this
observation is shown in Fig. 14. All samples from Fig. 10 were
heated at 10°C/min under a He flow. Their mp’s were compared to
that of the pure LiFePO4 ͑single crystal͒ sample obtained as an
inclusion from synthesis ͑9͒. The highest melting temperature of
985°C ͑endothermic peak value͒ was found for pure LiFePO4 crys-
tal and the second highest mp was for LiFePO4 with the next highest
purity ͑DM4C͒ that contained higher mp impurities of Li3PO4 ͑mp
of 1225°C͒.33 This is in contrast to the mp of the samples which
contain a higher concentration of impurities with lower mp compo-
sitions. One can also infer from this preliminary work that when
Table III. LiFePO4 synthesis using nonstoichiometric Li–Fe–P reactant stoichiometry.
Chemical reactant ratios ͑mole ratio to Fe͒
Global
formula
Secondary phases
from XRD
Sample
DM4-A
FePO4·2H2O + 41 C
1.0
1 Li2CO3
LiH2PO4
0
Volume
291.28
2
Li1.05FePO4
1.05
0.95
1.0
1.8%
1.0%
Li3PO4
Li4P2O7
DM4-B
DM4-C
DM4-D
Li0.95FePO4
LiFePO4
1.0
1.0
1.0
0
0
1%
Li4P2O7
291.43
291.21
291.39
2.8%
1.8%
Li3PO4
Li4P2O7
Li1.05Fe͑PO4͒
or LiFe0.95PO4
1.0
0.05
1.1%
Li4P2O7
1.05
Downloaded on 2014-10-30 to IP 128.148.252.35 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, 157 ͑4͒ A453-A462 ͑2010͒
A461
Figure 14. ͑Color online͒ DSC of LiFePO4 samples synthesized in Table III.
The indicated temperature represents the mp determined from the large en-
dothermic peak during the heating of the sample.
Figure 12. Micrographic examination of nonstoichiometric samples of the
Li–Fe–P system described in Table III: ͑a͒ Li0.95FePO4 and ͑b͒ LiFe0.95PO4.
thesis of LiFePO4 from different Li–Fe–P–O reactants, although
with different kinetics and experimental conditions. This may ex-
plain the presence of minor secondary phases that have been ob-
served within laboratory and commercial LiFePO4 synthesized by a
variety of solid-state processes. In fact, several minor impurity
phases of the general Li–Fe3+/Fe2+–P–O composition are present in
small amounts in commercial or laboratory synthesis of LiFePO4,
including Li3PO4,34 Li4P2O7,25 Li3Fe2͑PO4͒3,17,34-36 LiFeP2O7,35,37
LiFe5O8,25 Fe2O3/Fe3O4,34,38 Fe2P2O7,38,39 and Li8P8O4·6H2O.25
Many of these compounds are present as distinct phases and can be
detected, when their concentration is sufficient, by their XRD peaks.
In other cases, these impurity phases were confirmed by spectro-
using nonstoichiometric mixtures of Li–Fe–PO4 precursors to con-
trol the final LiFePO4 stoichiometry from a molten synthesis, one
has to take into account the phase diagram rules between LiFePO4
and the other phases that might coexist within a certain temperature
range and reducing conditions ͑oxygen chemical potential͒.
Metallographic examination of various molten LiFePO4 compo-
sitions after slow cooling suggests that LiFePO4 crystallizes with a
defined composition, while the excess Li–Fe–P–O reactants are ex-
pulsed during the stable olivine crystal structure growth and crystal-
lize in separate phases at lower temperatures. One can assume that
somewhat similar processes are at play during the solid-state syn-
Figure 13. ͑Color online͒ Detailed micro-
graphic and X-ray chemical analysis of
the LiFe0.95PO4 sample of Table III
͑DM4D͒. The analysis shows ͑a͒ the bulk
LiFePO4 phase, ͓͑b͒ and ͑c͔͒ an Fe-rich
impurity, and ͓͑d͒ and ͑e͔͒ an Fe-poor im-
purity.
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A462
Journal of The Electrochemical Society, 157 ͑4͒ A453-A462 ͑2010͒
scopic techniques such as Fourier transform IR or magnetic mea-
surements. However, amorphous or glassy phases are not excluded
as possibilities for some of these impurities. This has been men-
tioned recently in the case of glassy phosphates with the NASICON-
like structure Li3M3͑PO4͒3,38 and more recently reformulated by
Kang and Ceder to explain high rate LiFePO4.40
References
1. K. Misushima, P. C. Jones, P. J. Wisemann, and J. B. Goodenough, Mater. Res.
Bull., 15, 783 ͑1980͒.
2. Y. Nishi, A. Hideto, and O. Atsuo, U.S. Pat. 4,959,281 ͑1990͒.
3. H. Takeshita, in The 25th International Battery Seminar and Exhibit, Florida Edu-
cational Seminar ͑2008͒.
4. J. B. Goodenough, A. K. Padhi, K. S. Nanjundaswamy, and C. Masquelier, U.S.
Pat. 5,910,382 ͑1996͒.
Practically, the analysis of secondary phases present with
LiFePO4 becomes complex especially with submicrometer LiFePO4
particles, such as that used in power cells ͑large surface to bulk
ratio͒, when made by solid-state processes, not to mention the high
reactivity of these powders to their environment after synthesis.35 In
the absence of direct experimental data to ascertain the stability
domains of several possible phases under variable experimental con-
ditions, Ong et al. estimated by calculation some of the phases that
could be present in the Li–P–O–Fe2+/Fe3+ system.41 They presented
an interesting phosphate version of the Ellingham representation;
however, more experimentation is needed before a conclusion, on
the presence and role of the secondary phases which could coexist
within LiFePO4 when used as a cathode material, can be made. The
present work does not support the expectation that controlling the
reactant stoichiometry has a significant impact on solid LiFePO4
crystal composition, because nonstoichiometric reactant mixtures
seem to result in secondary phases that are found outside of the
olivine structure. Work is ongoing on both phase diagram and sta-
bility domain studies of the Li–Fe–P–O systems in equilibrium with
LiFePO4 to understand and optimize LiFePO4 as a cathode material
for both power and energy. LiFePO4 synthesis from the molten state
appears not only as an attractive synthesis route to make well-
defined LiFePO4, but also as a useful tool to study and optimize
polyanion cathode systems.
5. A. K. Padhi, K. S. Nanjundaswamy, and J. B. Goodenough, J. Electrochem. Soc.,
144, 1188 ͑1997͒; A. K. Padhi, K. S. Nanjundaswamy, and J. B. Goodenough,
Abstract 58, The Electrochemical Society Extended Abstracts, Vol. 175, Los An-
geles, CA, May 7–12, 1989.
6. B. Deveney, K. Nechev, T. Guseynov, R. Jow, and K. Xu, in 10th Electrochemical
Power Sources R&D Symposium, U.S. Department of the Navy, Aug 20–23, 2007.
7. A. Chu, in A123 Presentation at the 2009 ZEV Technology Symposium, Environ-
mental Protection Agency, Sept 21–22, 2009.
8. E. P. Roth, in First International Conference on Advanced Lithium Batteries for
Automobile Applications, Argonne National Laboratories, Sept 15–17, 2008.
9. T. D. Hatchard, D. D. MacNeil, A. Basu, and J. R. Dahn, J. Electrochem. Soc.,
148, A755 ͑2001͒.
10. M. Armand, J. B. Goodenough, A. K. Padhi, K. S. Nanjundaswamy, and C. Mas-
quelier, U.S. Pat. 6,514,640 ͑2003͒.
11. L. Guohua, A. Yamada, and H. Azuma, EP 1,094,532 A1 ͑2001͒.
12. J. Barker, M. Y. Saidi, and J. L. Swoyer, Electrochem. Solid-State Lett., 6, A53
͑2003͒.
13. N. Ravet, J. B. Goodenough, S. Besner, M. Simoneau, P. Hovington, and M.
Armand, Abstract 127, The Electrochemical Society Meeting Abstracts, Vol. 99-2,
Honolulu, HI, Oct 17–22, 1999.
14. N. Ravet, S. Besner, M. Simoneau, A. Vallée, M. Armand, and J. F. Magnan, U.S.
Pat. 6,855,273 ͑2005͒, 6,962,666 ͑2005͒, and 7,344,659 ͑2008͒.
15. H. Huang, S. C. Yin, and L. F. Nazar, Electrochem. Solid-State Lett., 4, A170
͑2001͒.
16. M. Armand, M. Gauthier, J. F. Magnan, and N. Ravet, WO 02/27823 A1 ͑2002͒.
17. A. Yamada, S. C. Chung, and K. Hinokuma, J. Electrochem. Soc., 148, A224
͑2001͒.
18. G. Arnold, P. Axmann, E. Wohlfahrt-Mehrens, and J. Garche, J. Power Sources,
119–121, 247 ͑2003͒.
19. S.-Y. Chung and Y.-M. Chiang, Electrochem. Solid-State Lett., 6, A278 ͑2003͒.
20. W. Ojczyk, J. Marzec, K. Swierczek, W. Zajac, M. Molenda, R. Dziembaj, and J.
Molenda, J. Power Sources, 173, 700 ͑2007͒.
21. S. Yang, P. I. Zavalij, and M. S. Whittingham, Electrochem. Commun., 3, 505
͑2001͒.
22. N. Ravet, J.-F. Magnan, M. Gauthier, and M. Armand, in International Conference
on Materials for Advanced Technologies (ICMAT 2001), MRS Singapore, July 1–6,
2001.
23. G. Nuspl, C. Vogler, M. Eisgruber, L. Wimmer, and N. Schall, Abstract 293, 12th
International Meeting of Lithium Batteries ͑2004͒.
24. C. Delacourt, P. Poizot, and C. Masquelier, WO 2007/000251 A1 ͑2007͒.
25. L. Gauthier, M. Gauthier, D. Lavoie, C. Michot, and N. Ravet, U.S. Pat. 7,534,408
͑2009͒.
26. Y. Okazaki, S. Okada, J. Yamaki, and T. Nishida, Abstract 2B02, 45th Conference
of the Battery Society of Japan ͑2004͒.
27. N. Ravet, Y. Chouinard, J.-F. Magnan, S. Besner, M. Gauthier, and M. Armand, J.
Power Sources, 97–98, 503 ͑2001͒.
28. M. M. Doeff, Y. Hu, F. McLarnon, and R. Kostecki, Electrochem. Solid-State Lett.,
Conclusions
LiFePO4 can be melted or synthesized through a molten process
to obtain a well-defined structure and an electrochemically active
LiFePO4. In some cases, partial iron substitution is possible from
the melt to make Li͑Fe–Mg–Mn͒PO4 crystals similar to what is
found in natural triphylite. However, other chemical elements or
compounds are immiscible in the molten phase at about 1000°C
͑e.g., C, Fe2P, silicates, aluminum, or calcium phosphate phases͒,
and they separate into discrete liquid or solid phases.
Many different precursors such as Fe, Fe2O3, and FePO4 can be
used to synthesize high purity LiFePO4. The phase purity of the final
products depends not only on the stoichiometry of Li, Fe, P, and O,
but also on the reducing conditions and the solidification process.
With less reducing conditions, a NASICON-type impurity is ob-
served.
Due to the slow cooling process in the ingot, LiFePO4 crystal-
lites grow into large oriented crystals with excellent crystallinity.
The low conductivity of LiFePO4 necessitates the reduction of the
particle size of these large crystals, and although this requires an
additional processing step, it is not an insurmountable task. A
follow-up paper from our group will detail particle size reduction
methods to generate the optimal electrochemical performance for
the electrode. In between the crystallites, secondary phases such as
Li4P2O7 and other low mp polyphosphates form in the last stage of
solidification. An electrochemical test shows good electrochemical
activity for the melt-cast LiFePO4.
6, A207 ͑2003͒.
29. J. Nakano, T. Yamada, and S. Miyazawa, J. Am. Ceram. Soc., 62, 465 ͑1979͒.
30. A. S. Andersson, B. Kalska, L. Haggstrom, and J. O. Thomas, Solid State Ionics,
130, 41 ͑2000͒.
31. S. Yang, Y. Song, P. Y. Zavalij, and M. S. Whittingham, Electrochem. Commun., 4,
239 ͑2002͒.
32. P. Jozwiak, J. Garbarczyk, F. Gendron, A. Mauger, and C. M. Julien, J. Non-Cryst.
Solids, 354, 1915 ͑2008͒.
33. D. R. Lide, CRC Handbook of Chemistry and Physics, pp. 4–66, CRC Press, Boca
Raton, FL ͑2004͒.
34. A. Ait-Salah, A. Mauger, K. Zaghib, J. B. Goodenough, N. Ravet, M. Gauthier, F.
Gendron, and C. M. Julien, J. Electrochem. Soc., 153, A1692 ͑2006͒.
35. N. Ravet, C. Michot, G. Nuspl, G. Liang, and M. Gauthier, Abstract 225, The
Electrochemical Society Meeting Abstracts, Vol. 2006-02, Cancun, Mexico, Oct
29–Nov 3, 2006.
36. P. Jozwiak, J. Garaczyk, F. Gendron, and C. M. Julien, Abstract 1385, The Elec-
trochemical Society Meeting Abstracts, Vol. 2005-01, Québec, Canada, May 15–
20, 2005.
Acknowledgments
The authors thank NSERC and Phostech Lithium, Inc. for fund-
ing this work under the auspices of the Industrial Research Chair
program. The authors are grateful to P. Hovington and M. Lagacé of
IREQ for several microscopic and metallurgical analyses, and to M.
Preda from Université du Québec for some of the XRD analysis.
37. N. Ravet, M. Gauthier, K. Zaghib, J. B. Goodenough, A. Mauger, F. Gendron, and
C. M. Julien, Chem. Mater., 19, 2595 ͑2007͒.
38. K. Zaghib, N. Ravet, M. Gauthier, F. Gendron, A. Mauger, J. B. Goodenough, and
C. M. Julien, J. Power Sources, 163, 560 ͑2006͒.
39. M. Maccario, L. Croguennec, A. Wattiaux, E. Suard, F. Le Cras, and C. Delmas,
Solid State Ionics, 179, 2020 ͑2008͒.
Université de Montréal assisted in meeting the publication costs of this
article.
40. B. Kang and G. Ceder, Nature (London), 458, 190 ͑2009͒.
41. S. P. Ong, L. Wang, B. Kang, and G. Ceder, Chem. Mater., 20, 1798 ͑2008͒.
Downloaded on 2014-10-30 to IP 128.148.252.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).