Synthesis of ternary transition metal fluorides Li
3
MF
6
via a sol–gel route as
candidates for cathode materials in lithium-ion batteries†
Julia Kohl,
a
Dennis Wiedemann,
a
Suliman Nakhal,
a
Patrick Bottke,
b
Noel Ferro,
c
Thomas Bredow,
c
Erhard Kemnitz,
d
Martin Wilkening,
b
Paul Heitjans
e
and Martin Lerch*
a
Received 5th April 2012, Accepted 8th June 2012
DOI: 10.1039/c2jm32133e
A sol–gel route for ternary lithium fluorides of transition metals (M) is presented allowing the synthesis
of Li
3
MF
6
-type and Li
2
MF
5
-type compounds. It is based on a fluorolytic process using transition metal
acetylacetonates as precursors. The domain size of the obtained powders can be controlled by
modifying the conditions of synthesis.
6
Li and
7
Li magic angle spinning (MAS) nuclear magnetic
resonance (NMR) spectroscopy is used to study local environments of the Li ions in orthorhombic and
monoclinic Li
3
VF
6
as well as Li
2
MnF
5
. The number of magnetically inequivalent Li sites found by
MAS NMR is in agreement with the respective crystal structure of the compounds studied. Quantum
chemical calculations show that all materials have high de-lithiation energies making them suitable
candidates to be used as high-voltage battery cathode materials.
Introduction
The search for new cathode materials as part of lithium-ion
batteries is an important objective today.
1–5
Currently, oxides
such as LiCoO
2
, layered Li–Mn-spinels as well as olivine-type
structures such as LiFePO
4
are in the focus of interest.
6,7
Modern
cathode materials need to fulfil many different requirements.
Promising compounds are supposed to exhibit high concentra-
tions of lithium in order to achieve high energy densities and
capacities. In addition, besides a good electronic conductivity,
sufficiently high lithium-ion diffusivity is one of the prerequisites
for facile lithium insertion and removal. Finally, chemical and
thermal stability of the components directly affect the cyclability
of the lithium-ion battery and thus its lifetime.
Until recently, various oxide materials have been in the focus
of research.
6,7
Interestingly, theoretical calculations indicate a
large increase of the redox potential by substituting fluorine for
oxygen.
8
Consequently, ternary lithium fluorides are increasingly
considered for advanced electrochemical characterization. In
particular, the Li
3
MF
6
(M ¼transition metal) family exhibits
high lithium contents combined with variable oxidation
numbers. Recently, Gonzalo et al. reported Li
3
FeF
6
to be a
promising cathode material for lithium-ion batteries.
9
Further-
more, the electrochemical properties of Li
3
VF
6
prepared by
microwave synthesis have been studied quite recently.
10a
In
addition, Li
3
VF
6
is also in the focus of a low-temperature
precipitation route in aqueous solution using alcohols recently
reported by Basa et al.
10b
The relatively low capacity of the
compounds is assumed to increase with decreasing particle size.
Therefore, the preparation of Li
3
MF
6
phases with particle sizes
less than 50 nm, as successfully shown by Basa et al.,
10b
is of great
interest to improve the associated electrochemical performance.
The compounds Li
3
MF
6
(M ¼V, Cr, Fe) are known to exist
in two polymorphs. Whereas the monoclinic form (space group
C2/c) crystallizes isotypically with b-Li
3
AlF
6
, the orthorhombic
modification crystallizes with the space group Pna2
1
being
identical to that of the corresponding a-form of Li
3
AlF
6
. The
two polymorphs are structurally related to the cryolite type.
11,12
Usually, the monoclinic polymorph is obtained by slow
cooling down of a stoichiometric mixture of the binary fluorides
to room temperature, while the orthorhombic form can only be
prepared by quenching the samples from an elevated (T> 600 K)
to ambient temperature. Due to the poor electronic conductivity
of transition metal fluorides the design of particle morphology is
of great importance for potential electrochemical applications.
Since ternary transition metal fluorides are usually synthesized
by solid-state reactions at high temperatures and/or high pres-
sures, the obtained particles show diameters in the micrometer
a
Technische Universit€
at Berlin, Department of Chemistry, Straße des 17.
Juni 135, 10623 Berlin, Germany. E-mail: martin.lerch@tu-berlin.de;
Fax: +49 030 314 22740; Tel: +49 030 314 22603
b
Technische Universit€
at Graz, Institut f€
ur Chemische Technologie von
Materialien, Stremayrgasse 9, 8010 Graz, Austria. E-mail: wilkening@
tugraz.at; Fax: +43 316 873 32332; Tel: +43 316 873 32330
c
Universit€
at Bonn, Mulliken Center for Theoretical Chemistry, Department
of Physical and Theoretical Chemistry, Beringstr. 4, 53115 Bonn,
Tel: +49 0228 73 3839
d
Humboldt-Universit€
at zu Berlin, Department of Chemistry, Brook-
Taylor-Straße 2, 12489 Berlin, Germany. E-mail: erhard.kemnitz@
chemie.hu-berlin.de; Fax: +49 030 2093 7277; Tel: +49 030 2093 7555
e
Leibnitz Universit€
at Hannover, Insitute of Phyiscal Chemistry and
Electrochemistry, Callinstr. 3 – 3a, 30167 Hannover, Germany. E-mail:
† CCDC 868968. For crystallographic data in CIF or other electronic
format see DOI: 10.1039/c2jm32133e
This journal is ªThe Royal Society of Chemistry 2012 J. Mater. Chem., 2012, 22, 15819–15827 | 15819
Dynamic Article LinksC
<
Journal of
Materials Chemistry
Cite this: J. Mater. Chem., 2012, 22, 15819
www.rsc.org/materials PAPER
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
/ Journal Homepage
/ Table of Contents for this issue
range. Generally, nm-sized particles can be easily prepared by
high-energy ball milling of the coarse grained materials.
13–15
Besides activation, mechanochemistry has also successfully been
used for the synthesis of a variety of materials from precursors at
room temperature including oxides and fluorides.
16–18
However,
in contrast to such a top-down approach, the preparation of
nanostructured particles by precipitation from aqueous solution
is much more beneficial when the shape and surface morphology
of the crystallites have also to be controlled. By following such a
route, the corresponding salts, for example nitrates or oxides, are
reacted with hydrofluoric acid and subsequent dehydration is
carried out by annealing at elevated temperatures. In general,
solution-based syntheses offer many advantages. For instance,
low temperatures and good mixing of the precursors allow the
preparation of highly homogeneous compounds with a large
surface area. In contrast to aqueous routes, syntheses carried out
in organic solvents provide a large range of different media with
versatile characteristics. When HF is dissolved directly in an
organic solvent, competing reactions between HF and H
2
O with
the transition metal can be eliminated. When, for example,
alcoholates or acetylacetonates are used as starting materials the
corresponding alcohols or acetylacetone is formed which can
easily be removed under vacuum. Moreover, due to very fast
crystallization materials consisting of very small particle
diameters (low nm-range) can be obtained.
To our knowledge only few reports can be found in the liter-
ature which report on the synthesis of ternary fluorides from
these starting materials. Kemnitz et al. presented the synthesis of
aluminum and magnesium fluorides with high surface areas by a
fluorolytic sol–gel process.
19,20
These studies illustrate the
synthetic potential of organic hydrogen fluoride solutions.
Aluminum compounds crystallizing in the cryolite or elpasolite
type were prepared from the corresponding alcoholates. For
example, Li
3
AlF
6
can be obtained via the reaction of LiOtBu and
Al(OiPr)
3
with HF in isopropanol.
21
This method allows one to
work under water-free conditions. Concerning transition metal
fluorides only acetylacetonates, acetates and, in the case of iron,
alcoholates are obtainable as starting materials. The aim of our
present work is to develop a nonaqueous sol–gel route for the
preparation of Li
3
MF
6
compounds (M ¼V, Cr, Fe, Mn, Co) and
to elucidate the possibilities of controlling the size of the particles
synthesized.
Experimental section
X-ray powder diffraction
X-ray powder diffraction experiments were performed using a
PANalytical X’Pert PRO MPD diffractometer (CuK
a
-radiation,
2qrange 5 to 120, Bragg–Brentano (q–q) geometry) with PIXcel
detector (Si–Li-semiconductor with 255 measuring channels). All
samples were prepared on small Si-cavity mounts.
X-ray fluorescence
For X-ray fluorescence analysis, a PANalytical Axios PW4400/
24 X-ray fluorescence spectrometer with an Rh-tube and wave-
length-dispersive detection was used. Depending on the analyzed
elements a LiF single crystal (crystallographic orientation (220)
and (200)), a Ge single crystal (orientation (111)), a PE single
crystal (orientation (002)) and a PX1 multi-layer mono-
chromator were used together with an Si(Li) scintillation
detector.
NMR
Liquid-phase
1
H,
13
C,
51
V, and
19
F NMR spectra were recorded
on a Bruker Avance 200 and a Bruker Avance 400 NMR spec-
trometer. TMS, CFCl
3
and VOCl
3
served as references. Solid-
state high-resolution, i.e., magic angle spinning (MAS),
6
Li and
7
Li NMR spectra were acquired using Avance III NMR (Bruker
BioSpin) spectrometers connected to cryomagnets with nominal
fields of 7 T and 14.1 T. This results in
6
Li resonance frequencies
of 44 and 88 MHz and
7
Li resonance frequencies of 117 and
233 MHz, respectively. We used a standard (double-resonance)
2.5 mm-probe (Bruker) which can be operated at spinning speeds
of up to 30 kHz. Additionally, some
6
Li MAS NMR spectra were
recorded with an Avance NMR spectrometer being connected to
a cryomagnet with a nominal magnetic field of 17.6 T. The
spectrometer can be used in combination with an MAS NMR
probe allowing a maximum spinning speed of 15 kHz. The Li
MAS NMR spectra shown were referenced to LiCl (aq). They
were recorded using a single excitation pulse and recycling delays
of up to several seconds.
6
Li and
7
Li NMR spin–lattice relaxation
times T
1
of the paramagnetic Li
3
MF
6
and Li
2
MnF
5
samples were
measured with a conventional saturation recovery experiment
using up to 12 different delay times. As expected, the T
1
values
associated with the paramagnetic shifts do not exceed 100 ms.
Preliminary 2D exchange MAS NMR spectra of Li
2
MnF
5
were
recorded using a conventional NOESY pulse sequence. Time
domains of 512 data points in both f1 and f2 directions were
used. Processing of the data was carried out using TopSpin 3.1
software (Bruker) and Mnova7 (Mestrelab research).
FTIR
FTIR spectra were measured on a Varian 640IR FTIR spec-
trometer equipped with a Pike GladiATR device for measure-
ments in attenuated total reflectance mode. FTIR spectra of KBr
pellets and CsI pellets were measured with a Nicolet Series II
Magna-IR System 750 FTIR spectrometer in transmission mode.
Elemental analysis
The carbon and hydrogen contents were determined by
combustion analysis (Thermo Finnigan FlashEA 1112 NC
analyzer), the oxygen contents using a LECO EF-TC 300 N
2
/O
2
analyzer (hot gas extraction).
Materials and methods
V(acac)
3
(ABCR), Fe(acac)
3
, Cr(acac)
3
, Mn(acac)
3
,
Mn(OAc)
3
$2H
2
O, Co(acac)
3
and LiOtBu (Sigma-Aldrich) were
used as received. Solutions of HF in ethanol, THF and Et
2
O were
prepared by feeding gaseous HF into the solvent under cooling.
The solvents were dried according to standard literature proce-
dures. All reactions were carried out using standard Schlenk
techniques. Reagents and samples were stored in an Ar-filled
glove box.
15820 | J. Mater. Chem., 2012, 22, 15819–15827 This journal is ªThe Royal Society of Chemistry 2012
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
General synthesis
Precursor synthesis, Li
3
VF
6
as example: 1 g (2.87 10
3
mol) of
V(acac)
3
and 0.6895 g (8.61 10
3
mol) of LiOtBu (Li : V ¼
3 : 1) were weighed into a Schlenk tube and suspended in 20 ml of
absolute ethanol at room temperature for 30 min. To the resul-
tant suspension 8.40 ml of a 10.25 M HF–EtOH solution was
added leading to a green solution (V : HF ¼1 : 30). Alterna-
tively, a 6.89 M HF–Et
2
O or a 10.68 M HF–THF solution was
used. The solution was stirred for 2 h at room temperature.
Afterwards, the solvent was removed under vacuum and the
resultant fine green powder was dried at 80 C for 2 h.
Synthesis of the Li
3
MF
6
samples: all syntheses were carried out
in sealed copper or monel capsules under nitrogen atmosphere.
100 to 200 mg of the precursor were filled into a one-side sealed
capsule. If not stated otherwise, all samples were slowly cooled
down to room temperature. In Table 1, the synthesis conditions
of the fluorides prepared are listed.
V(acac)
2
(CH
3
CN)
2
BF
4
: a saturated solution of the Li
3
VF
6
precursor was prepared in 10 ml of absolute CH
3
CN and stirred
overnight at 50 C. The resulting green-brown solution was filled
in small glass tubes. These glass tubes were positioned in a
Schlenk tube filled with 20 ml of absolute Et
2
O. After approxi-
mately 20 days dark red crystals were formed.
Results and discussion
Monoclinic and orthorhombic Li
3
VF
6
As described in the Experimental section, Li
3
VF
6
can be
prepared by a two-step synthesis. During the first step the so-
called precursor is synthesized. LiOtBu and V(acac)
3
react with
hydrogen fluoride in a dry solvent (Li : V : HF ¼3 : 1 : 30).
After evaporating the solvent and drying, the obtained Li
3
VF
6
precursor is calcined at 300 to 800 C to form Li
3
VF
6
. The
precursor can be described as a fine, green and mainly X-ray
amorphous powder. The corresponding X-ray powder pattern
shows that the precursor also contains poorly crystalline LiF and
Li
2
SiF
6
. The latter phase results from the reaction of HF with the
glassware used. However, Li
2
SiF
6
does not affect the subsequent
reactions because it decomposes into LiF and SiF
4
at tempera-
tures above 250 C.
22
As determined by X-ray fluorescence
analysis, all samples contain approximately 3 to 4.5% Si. Inter-
estingly, small amounts of Li
2
SiF
6
were also formed when the
synthesis was carried out in a glass Schlenk tube equipped with a
closed PTFE insert inside (Fig. 1b and c). This insert was opened
after synthesis for removing the solvent under vacuum.
Surprisingly, remaining HF seems to react immediately with the
glassware in which the PTFE tube is inserted. Note that from
X-ray diffraction no information on the nature of the vanadium
species can be obtained.
Li
3
VF
6
, which was prepared by decomposition of the
precursor, was also analyzed by X-ray powder diffraction.
Heating to 300 C leads to the formation of Li
3
VF
6
.Li
2
SiF
6
seems to be completely decomposed (see also Fig. 1a). From the
XRD patterns there are only vague indications of very small
amounts of remaining LiF not reacted with vanadium species of
the precursor. When the samples were cooled down to room
temperature slowly, monoclinic Li
3
VF
6
was observed. Heating
the precursor to 700 C and quenching the sample to room
temperature results in the formation of orthorhombic Li
3
VF
6
as
described in the literature (see Fig. 2). Both polymorphs were
obtained with more than 98% purity. Elemental analysis resulted
in a residual carbon content of approximately 2%.
Besides the successful preparation of highly pure a-Li
3
VF
6
and
b-Li
3
VF
6
, it is also possible to modify the domain size of the
samples prepared. For example, this can easily be achieved by
Table 1 Synthesis conditions for the Li
3
MF
6
compounds (orth. ¼
orthorhombic and mon. ¼monoclinic)
Compound Dwell time Temperature Further conditions
Li
3
VF
6
, orth. 2 h 700 C Quenching to room
temperature after 2 h
Li
3
VF
6
, mon. 4 h 150 to 600 C
Li
3
FeF
6
, orth. 2 h 800 C Quenching to room
temperature after 1 h
Li
3
FeF
6
, mon. 4 h 400 to 600 C
Li
3
CrF
6
, mon. 4 h 500 C
Li
2
MnF
5
4 h 400 C 30% Li
2
MnF
5
Fig. 1 X-ray powder diffraction patterns of different Li
3
VF
6
precursors:
(a) precursor decomposition at 300 C (Li
2
SiF
6
has been completely
decomposed), (b) synthesis in glassware, (c) synthesis with PTFE insert
(o ¼LiF, * ¼Li
2
SiF
6
).
Fig. 2 X-ray powder diffraction patterns of monoclinic b-Li
3
VF
6
synthesized at (a) 300 C and (b) 600 C. (c) Corresponding X-ray powder
pattern of orthorhombic a-Li
3
VF
6
which has been synthesized at 700 C
and by subsequent quenching to room temperature.
This journal is ªThe Royal Society of Chemistry 2012 J. Mater. Chem., 2012, 22, 15819–15827 | 15821
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
varying the synthesis temperature as presented in the following.
The precursors were heated for 4 h at 400, 500 and 600 C,
respectively. As shown in Fig. 3, the full width at half maximum
of the diffraction reflections decreases with increasing tempera-
ture. The domain sizes were calculated by methods based on the
Scherrer formula.
26
The results are depicted in Table 2. Here, the
domain size can be varied from approximately 30 to 200 nm
which is the range of interest for electrochemical applications. In
general, the domain size increases with increasing temperature.
For different solvents the absolute values as well as the temper-
ature dependence of the sizes differs significantly. The reason for
this is unclear. First investigations of the samples by means of
standard SEM resulted in an average particle size of 200 nm for
the samples synthesized at 400 C.
Typical
6
Li MAS NMR spectra of polycrystalline a-Li
3
VF
6
and b-Li
3
VF
6
are shown in Fig. 4. In agreement with the crystal
structure of the orthorhombic modification the NMR spectrum
of a-Li
3
VF
6
reveals three distinct lines which can be attributed to
the Li positions Li(1), Li(2), and Li(3) as shown recently by some
of us.
23
The NMR shifts result from the Fermi-contact interac-
tion which is related to the extent of electron spin density
transferred from the V
3+
t
2g
orbital to the 2s orbital Li ion. In ref.
23 the assignment of the NMR lines presented has been based on
(i) results from temperature-variable 1D and 2D exchange NMR
experiments, (ii) the considerations of different mechanisms to
transfer electron spin density, and (iii) the connectivities of the
VF
6
and LiF
6
polyhedra in a-Li
3
VF
6
. Note that with increasing
temperature the lines first broaden and finally coalesce because of
Li ion exchange taking place on the timescale determined by the
distance of the NMR lines (see Fig. 1 in ref. 23). The beginning of
this process can already be recognized when the spectrum shown
in Fig. 4c is considered. At lower temperatures, note that NMR
spectra down to 277 K were recorded, no additional lines show
up indicating that the three lines observed are not affected by any
coalescence phenomena occurring at lower temperatures. The
three lines detected show approximately the same intensity which
is expected from the crystal structure where the three crystallo-
graphically inequivalent Li sites, residing on the same Wyckoff
position 4a, are fully occupied.
Interestingly, the individual paramagnetic NMR shifts d
depend on temperature. A linear relationship between dand 1/T
is expected for the Curie–Weiss behavior quantifying the
dependence of the magnetic susceptibility on temperature. The
larger the paramagnetic shift the steeper the slope of the corre-
sponding d(1/T) line: see, e.g., the recent study by Spencer et al.
24
This effect can be clearly seen in Fig. 4d–f showing the
6
Li MAS
NMR spectra of the monoclinic counterpart of Li
3
VF
6
.Upto
353 K no coalescence of the NMR lines is observed.
In b-Li
3
VF
6
the Li ions occupy five crystallographically
inequivalent sites whereby Li(2), Li(3), Li(4) and Li(5) reside on
the Wyckoff position 8f and Li(1) on 4e. Since all the sites are
fully occupied one might ascribe the NMR signal with the lowest
intensity and the largest paramagnetic shift (76 ppm at 333 K) to
the Li(1) ions. The asymmetric shape of this NMR peak might
indicate that the signal is composed of more than one line, i.e.,
the ions residing on the position 4e are crystallographically
equivalent but not magnetically so. The assignment of the other
NMR peaks, which do not differ in intensity as expected from the
crystal structure, requires mixing time dependent 2D exchange
NMR experiments and a careful analysis of the relevant transfer
mechanisms of electron spin density. Such a study is beyond the
scope of the present contribution and will be published elsewhere
together with an investigation of the Li hopping processes taking
place in b-Li
3
VF
6
. First results were shown in ref. 25. Comparing
the 1D
6
Li MAS NMR spectra shown in ref. 23 with those
obtained from a-Li
3
VF
6
and presented in Fig. 4, it is already
evident that Li jump diffusion in the monoclinic modification is
slower than that in the orthorhombic form. Interestingly, our
quantum-chemical calculations show (vide infra) that the
monoclinic modification is found to be more stable than
the orthorhombic one which reveals rapid Li exchange among
the three regularly occupied crystallographic positions (in
particular, see Fig. 2 in ref. 23).
Li
3
MF
6
(M ¼Cr, Fe, Co) and Li
2
MnF
5
The preparation of the corresponding compounds containing
chromium, iron, and manganese was performed in analogy to
that of Li
3
VF
6
. Unfortunately, and in contrast to Li
3
VF
6
, which
can be synthesized with high yields, the designated products
could not be prepared as single-phase powders. During the
synthesis of the precursor, the formation of larger quantities of
the corresponding difluorides occurs. For the preparation of a
Li
3
FeF
6
-precursor, Fe(acac)
3
and Fe(OEt)
3
were tested in
Fig. 3 Temperature-dependent broadening of the reflections at 2q¼
30.72and 2q¼49.45of monoclinic Li
3
VF
6
(400 C (black), 500 C
(red), 600 C (green)). All samples were slowly cooled down to room
temperature.
Table 2 Calculated domain sizes for monoclinic Li
3
VF
6
prepared at
different temperatures. The values listed were calculated with LaB
6
as
standard. The respective precursors were synthesized in different solvents
Decomposition temperatures
and corresponding
domain sizes
400 C 500 C 600 C
Li
3
VF
6
, precursor synthesized in THF 37 nm 107 nm 127 nm
Li
3
VF
6
, precursor synthesized in Et
2
O 33 nm 84 nm 160 nm
Li
3
VF
6
, precursor synthesized in EtOH 44 nm 80 nm 155 nm
15822 | J. Mater. Chem., 2012, 22, 15819–15827 This journal is ªThe Royal Society of Chemistry 2012
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
different solvents. Tempering of the obtained orange powder at
400 C always resulted in a mixture of monoclinic Li
3
FeF
6
as well
as LiF and FeF
2
. The orthorhombic modification of Li
3
FeF
6
was
also obtained with only a 50% yield. In the case of M ¼Cr,
monoclinic Li
3
CrF
6
could be obtained with approximately 85%
purity. The precursor was prepared from LiOtBu, Cr(acac)
3
and
a HF–EtOH solution (see Fig. 5). Unfortunately, by using
LiOtBu/Co(acac)
3
/HF, Co
3+
was completely reduced to Co
2+
and only LiF and CoF
2
were formed. Surprisingly, in the system
Li–Mn–F the formation of Li
2
MnF
5
was observed. Mn(acac)
3
and Mn(OAc)
3
$2H
2
O were used as starting materials. MnF
2
was
already formed during the synthesis of the precursor. The vari-
ation of the reaction time to prepare the precursor turned out to
have no effect on the amount of MnF
2
formed. Interestingly,
Li
2
MnF
5
with a maximum yield of 30% was only formed by
decomposing the precursor synthesized from Mn(OAc)
3
$2H
2
O
(Fig. 6). Precursors synthesized from Mn(acac)
3
yielded only
MnF
2
and LiF after decomposition. Compared to conventional
solid-state routes reported in the literature, which require
temperatures ranging from 700 to 800 C,
27
the two-step
synthesis route followed here allows the preparation of Li
2
MnF
5
at temperatures as low as 400 C.
Fig. 4
6
Li MAS NMR spectra of a-Li
3
VF
6
(a–c), 12 kHz spinning speed and b-Li
3
VF
6
(d–f), 30 kHz spinning speed recorded at the temperatures
indicated. Spectra have been referenced to aqueous LiCl. Note that in each case the number of NMR signals detected is in agreement with the
crystallographic data of the two polymorphs. The temperature dependence of the NMR shifts points to Curie–Weiss behaviour as expected.
Fig. 5 X-ray powder diffraction pattern of monoclinic Li
3
CrF
6
prepared at 500 C. The sample was slowly cooled down to room
temperature (* ¼LiF).
Fig. 6 X-ray powder pattern of Li
2
MnF
5
(*). The sample was prepared
at 400 C from Mn(OAc)
3
$2H
2
O. Other products found are LiF and
MnF
2
.
This journal is ªThe Royal Society of Chemistry 2012 J. Mater. Chem., 2012, 22, 15819–15827 | 15823
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
Note that the procedure described in ref. 27 is useful to prepare
single crystals. However, no information on the yields is given by
the authors although preliminary experiments carried out by our
group resulted in a maximum yield of 95% when a stoichiometric
mixture of LiF and MnF
3
was used in a solid state route (60 h,
480 C, monel capsule) to obtain Li
2
MnF
5
.
Preliminary
7
Li MAS NMR spectra of Li
2
MnF
5
, synthesised
via the above-mentioned solid state route (95% yield), recorded
at spinning speeds ranging from 15 to 35 kHz (ambient bearing
gas temperature) comprise rather broad resonance lines. The
MAS NMR spectrum recorded at 15 kHz (see Fig. 7) reveals an
intense line (A) at approximately 180 ppm and another one of
lower intensity (B) at 0 ppm (the dashed line is to guide the eye),
both with spinning sidebands. The intense line spreads over a
range of some hundreds of ppm. The isotropic resonance can be
distinguished from the spinning sidebands by recording MAS
spectra at different rotation frequencies (see Fig. 7b). As can be
clearly seen in Fig. 7b, the heat development, which increases
with increasing spinning speed when ambient bearing gas pres-
sure is used, is directly reflected by the shift of the isotropic
resonance. As an example, using room temperature bearing gas
the isotropic resonance of the main signal shows up at 172 ppm.
Reducing the spinning speed by 5 kHz shifts the line towards
positive ppm values. Additional heating of the bearing gas causes
the line to show up at 162 ppm, as expected.
Thus, the main line clearly reveals a temperature-dependent
NMR shift (see the corresponding MAS NMR spectra labeled
(ii) and (iii) in Fig. 7b); the associated spin–lattice relaxation time
T
1
turns out to be of the order of 2 ms only. This is in contrast to
Fig. 7
7
Li MAS NMR spectra of Li
2
MnF
5
recorded at (a) 15 kHz spinning speed and (b) at 25 and 35 kHz, respectively. The resonance frequency used
was 116 MHz. Except for the spectrum labeled with (iii) ambient bearing gas temperature has been used. The dashed line is to guide the eye. Spinning
sidebands, clearly shifting with rotation frequency, are marked with stars.
Fig. 8 2D
7
Li exchange MAS NMR spectrum (116 MHz) of poly-
crystalline Li
2
MnF
5
recorded at a spinning speed of 30 kHz and heated
bearing gas pressure. The mixing time (1 ms) was chosen to be as large as
possible; note that the corresponding NMR spin–lattice relaxation time
of the main component (A) is approximately 2 ms. No off-diagonal
intensities show up in the areas where the solid lines do cross. The dashed
line simply connects the diagonal intensities.
15824 | J. Mater. Chem., 2012, 22, 15819–15827 This journal is ªThe Royal Society of Chemistry 2012
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
the line at 0 ppm which is characterized by T
1
¼1.9 s. Further-
more, no temperature dependence of this minor NMR compo-
nent could be observed as shown in Fig. 7b. These findings
indicate that the two spin ensembles seem to be magnetically
decoupled.
This is supported by the observation that no cross-peaks show
up between the isotropic resonances in a 2D exchange MAS
NMR experiment (see the signals labeled A and B in Fig. 8),
which was carried out at 116 MHz, with a heated bearing gas
temperature of 333 K and a mixing time of 1 ms. Off-diagonal
intensities are expected to be caused by, e.g., chemical exchange
or spin-diffusion between Li sites in the same lattice. The absence
of any cross-peaks might indicate that the two signals simply
stem from two different phases present whereby the signal at
0 ppm might reflect a (diamagnetic) impurity formed during
synthesis (see above) and/or NMR sample preparation, see also
ref. 28. In conclusion, at least at ambient temperature and above,
the 1D MAS NMR spectra of Li
2
MnF
5
reveal a single resonance.
This is in good agreement with the crystal structure reported for
Li
2
MnF
5
showing only one Li site. Further measurements, which
are intended to be performed at much lower temperatures, might
help in finding out whether the main resonance detected is a
coalesced one being the result of fast Li exchange processes
between magnetically slightly differing Li positions. Previously,
this was reported for Li intercalated Cr
2
Ti
3
Se
8
studied by
Wontcheu et al.
28
Crystal structure determination of the Li
3
VF
6
precursor
As already described above, X-ray powder diffraction gave no
satisfying information on the crystal structure and the species the
Li
3
VF
6
precursor is composed of. Therefore, further analyses
were performed to collect information on the vanadium species
and to answer the question of whether Li
3
VF
6
is already formed
during the synthesis of the precursor or is generated during the
annealing step. Such information helps optimize the synthesis
conditions of the related precursors used to prepare Li
3
FeF
6
and
Li
2
MnF
5
, for example. The Li
3
VF
6
precursor described above
was complementarily investigated by NMR and FTIR
spectroscopy.
1
H and
13
C NMR spectra of the removed solvents
clearly show the signal of acetylacetone which indicates that
acetylacetonate was abstracted. Samples of the Li
3
VF
6
precursor
were measured in MeOH-d
4
and D
2
O, respectively.
1
H NMR
signals of acetylacetonate bonded to V
3+
were observed at 44
ppm. The NMR signals are very broad and have a pronounced
low-field shift which is caused by the V
3+
ion (vide infra). The
chemical shift corresponds to the values known for V(acac)
3
.
29,30
The
19
F NMR spectra reveal intensities in the range from 124
ppm up to 145 ppm (D
2
O). When the samples have been solved
in MeOD-d
4
NMR signals at 132 and at 154 ppm show up.
NMR lines with chemical shifts ranging from 124 to 135 ppm
can be ascribed to Li
2
SiF
6
.
31
Note that the
19
F NMR shifts of
[SiF
6
]
2
depend on the pH-value.
32
Because of exchange reactions
they shift to lower fields with increasing pH-value. In the
51
V
NMR spectra signals at 657 and 896 ppm were observed. The
51
V shift for V(acac)
3
in MeOD-d
4
shows up at 899 ppm.
FTIR experiments were carried out in transmission mode (50
to 4000 cm
1
) and by using an ATR-module (400 to 4000 cm
1
,
attenuated total reflectance mode). We used samples which were
prepared as KBr and CsI pellets. First, let us discuss the far IR
range (50 to 400 cm
1
). The transmission spectrum shows one
band at 295 cm
1
which is in good agreement with the n
4
-band of
[VF
6
]
3
.
33
Further bands are visible in the organic fingerprint
range.
34
The C]O– and C]C combination modes of acetyla-
cetonate appear at 933, 1532 and 1586 cm
1
. Further bands can
be found at 1033 cm
1
(r
r
(CH
3
)), 1366 + 1388 cm
1
(d
s
(CH
3
)) and
at 1290 cm
1
(n(C–CH
3
)+n(C:::C)). Depending on the prepa-
ration conditions of the sample, the IR spectra show a broad
band in the range from 714 to 741 cm
1
which agrees with the
n
3
-band of [SiF
6
]
2
.
35
In summary, it might be concluded that the
Li
3
VF
6
-precursor contains several vanadium species including
acetylacetonate as well as [VF
6
]
3
.
In addition to the spectroscopic analyses, attempts were made
to prepare single crystals. Saturated solutions of the precursor in
CH
3
CN or MeOH were filled into small glass tubes. These glass
Fig. 9 Crystal structure of [V(acac)
2
(CH
3
CN)
2
]BF
4
as derived from X-
ray diffraction (ORTEP representation with 50% probability
ellipsoids).
37,38
Table 3 Crystal data of [V(acac)
2
(CH
3
CN)
2
]BF
4
Chemical formula C
14
H
20
BF
4
N
2
O
4
V
Formula mass 418.07
Crystal system Monoclinic
a/
A 11.9556(10)
b/
A 5.7003(4)
c/
A 15.8568(19)
a/90.00
b/119.774(7)
g/90.00
Unit cell volume/
A
3
937.99(17)
Temperature/K 150(2)
Space group P2
1
/c
Number of formula units per unit cell, Z2
Radiation type MoKa
Number of reflections measured 3866
Number of reflections observed 1372
R
s
0.0635
Number of independent reflections 1832
R
int
0.0390
Final R
1
values (I>2s(I)) 0.0450
Final wR(F
2
) values (I>2s(I)) 0.1044
Final R
1
values (all data) 0.0633
Final wR(F
2
) values (all data) 0.1093
This journal is ªThe Royal Society of Chemistry 2012 J. Mater. Chem., 2012, 22, 15819–15827 | 15825
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
tubes were placed inside a Schlenk tube filled with Et
2
Oor
pentane. From the CH
3
CN solutions, dark red crystals of the
compound [V(acac)
2
(CH
3
CN)
2
]BF
4
were obtained (Fig. 9). This
compound shows the existence of acac-containing vanadium
species in the precursor. The current work of our group indicates
that the here-presented species is just the first one in a sequence of
more and more fluorinated species. More details will be given in a
forthcoming contribution.
36
BF
4
is formed by the reaction of
HF with the borosilicate glassware during the synthesis of the
precursor. V
3+
is coordinated octahedrally by two acetylaceto-
nate ions and two molecules of CH
3
CN. The BF
4
ion is severely
disordered over a center of inversion. Crystal data for
[V(acac)
2
(CH
3
CN)
2
]BF
4
are given in Table 3.
Quantum-chemical calculations
In order to investigate the suitability of the synthesized Li
3
MF
6
compounds as potential cathode materials in lithium-ion
batteries, we calculated the corresponding de-lithiation energies
at density-functional theory (DFT) level. Based on our experi-
ence with open-shell transition metal systems
39
we chose the
DFT/Hartree–Fock hybrid functional PW1PW.
40
The calcula-
tions were performed with the crystalline-orbital program
package CRYSTAL09.
41
The standard atomic basis sets for the
elements Li, O, V, Cr, Mn and Fe were taken from the
CRYSTAL homepage
42
and augmented where possible with
diffuse shells and polarization functions. A detailed description
of the optimized solid-state basis sets will be given elsewhere.
43
The computational accuracy parameters were set to rather strict
values as described previously.
39
In Table 4 the calculated lattice
parameters are compared with our measured values obtained
from X-ray powder diffraction where available (vide supra). The
agreement found between theory and experiment for the struc-
ture parameters is encouragingly good. In all cases the deviations
are below 1% except the band clattice vectors of orthorhombic
Li
3
FeF
6
. This indicates that our theoretical approach is suffi-
ciently accurate to describe the transition metal fluorides. For
Li
3
VF
6
and Li
3
FeF
6
we computed the relative stability of the
orthorhombic and monoclinic phases. In both cases the mono-
clinic polymorph was more stable so that only this phase was
considered for the calculation of the de-lithiation energy E
d
. For
all compounds we also investigated the magnetic structure.
Ferromagnetic and antiferromagnetic couplings were consid-
ered. For example, for Li
2
MnF
5
the previously reported anti-
ferromagnetic chains of Mn
3+
ions were considered.
44
The de-lithiation energy was computed by removing all Li
atoms occupying a certain Wyckoff position. As discussed above
there are five crystallographically inequivalent Li atoms in the
asymmetric unit of monoclinic Li
3
MF
6
.InLi
2
MnF
5
the two Li
atoms of the primitive unit cell are crystallographically equiva-
lent. In the antiferromagnetic ground state they are not elec-
tronically equivalent anymore. However, the differences between
local electrostatic potentials and electric field gradients are rather
small. In Table 3 we only report the smallest values of the
de-lithiation energies for each compound according to the
following equation:
Li
3n
M
n
F
6n
(s) ¼xLi(s) + Li
3nx
M
n
F
6n(s)
with lithium metal as reference according to the electrochemical
reference. xcorresponds to the number of Li ions occupying the
particular Wyckoff position. E
d
is then normalized to one Li
atom. Since only one electron is transferred in this process, E
d
in
eV directly corresponds to a battery voltage in V (for approxi-
mations see below). It has to be noted that in the present lithium-
ion batteries graphite is used as the counter anode rather than
metallic lithium, but the potential differences are rather small.
The present values can only be regarded as zero-order approxi-
mations since we did not take into account possible phase tran-
sitions after Li removal. The change in Li concentration is rather
large in our models due to the small unit cell size. This means that
we calculate an average value of the potential for the corre-
sponding xrange rather than E
d
(x) as in electrochemical
measurements. Furthermore, enthalpy and entropy are not taken
into account due to the high computational effort. Nevertheless,
we have shown in a parallel study that the simplistic approach
described here can indeed reproduce measured cell potentials of a
wide range of battery materials with surprisingly high accuracy.
45
The resulting E
d
values presented in Table 3 may therefore have
some significance. As they range from 4.7 to 6.7 eV, not too far
from the initially discussed theoretical result (6 eV) for LiCa-
CoF
6
or LiCdCoF
6
,
8
all materials synthesized in this work can be
regarded as promising high-voltage cathode materials provided
that stability issues are resolved.
Conclusions
Ternary lithium fluorides of transition metals were synthesized
via a sol–gel route from transition metal acetylacetonates as
precursors and hydrogen fluoride in organic solvents. The two
polymorphs of Li
3
VF
6
were obtained with approximately 98%
purity. Local environments probed by high-resolution solid-state
6
Li and
7
Li NMR are in very good agreement with expectations
from the crystal structure of the compounds. The synthesis of the
monoclinic phase is already possible at temperatures as low as
300 C. Its domain size can be controlled by modifying the
temperature used for the synthesis. This may be of importance in
Table 4 Calculated lattice parameters a,b,c,band de-lithiation energy E
d
of the compounds Li
3
MF
6
and Li
2
MnF
5
(method PW1PW). Values in
parentheses correspond to measured lattice parameters (orth. ¼orthorhombic, mon. ¼monoclinic)
Compound a/
Ab/
Ac/
Ab/E
d
/eV
Li
3
VF
6
, orth. 9.53 (9.59) 8.45 (8.49) 5.03 (5.04)
Li
3
VF
6
, mon. 14.30 (14.39) 8.68 (8.69) 10.05 (10.06) 96.4 (95.9) 4.73
Li
3
FeF
6
, orth. 9.58 (9.53) 8.43 (8.24) 5.01 (4.88)
Li
3
FeF
6
, mon. 14.34 (14.41) 8.64 (8.67) 10.02 (10.05) 95.6 (95.3) 6.69
Li
3
CrF
6
, mon. 14.32 (14.43) 8.59 (8.61) 10.02 (10.04) 94.9 (94.6) 6.18
Li
2
MnF
5
10.02 (10.02) 4.97 (4.95) 7.39 (7.41) 112.4 (112.2) 5.88
15826 | J. Mater. Chem., 2012, 22, 15819–15827 This journal is ªThe Royal Society of Chemistry 2012
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
the preparation of cathode materials for lithium-ion batteries.
Monoclinic Li
3
CrF
6
was also prepared with a phase purity of
approximately 85%. For manganese, iron and cobalt the corre-
sponding difluorides were formed in addition to the ternary
fluorides. Quantum chemical calculations show that all materials
have high de-lithiation energies making them suitable candidates
to be used as high-voltage battery cathode materials.
Acknowledgements
We thank the German Ministry of Education and Research
(BMBF) for financial support within the LIB 2015 initiative,
project HE-Lion. We thank D. Freude and E. Romanova
(Leipzig) for recording of some of the NMR measurements
presented. M.W. gratefully acknowledges financial support by
the Deutsche Forschungsgemeinschaft (DFG). We would also
like to thank Mr Manfred Detlaff (NMR service group, TU
Berlin) and Ms Paula Nixdorf (XRD service group, TU Berlin)
for their experimental help.
Notes and references
1 J. M. Tarascon and M. Armand, Nature, 2001, 414, 359.
2 L. F. Nazar, G. Goward, F. Leroux, M. Duncan, H. Huang, T. Kerr
and J. Gaubicher, Int. J. Inorg. Mater., 2001, 3, 191.
3 M. S. Whittingham, Chem. Rev., 2004, 104, 4271.
4 A. S. Aric
o, P. Bruce, B. Scrosati, J.-M. Tarascon and
W. V. Schalkwijc, Nat. Mater., 2005, 4, 366.
5 P. G. Bruce, B. Scrosati and J.-M. Tarascon, Angew. Chem., Int. Ed.,
2008, 47, 2930.
6 B. L. Ellis, K. T. Lee and L. F. Nazar, Chem. Mater., 2010, 22, 691–
714.
7 J. B. Goodenough and Y. Kim, Chem. Mater., 2010, 22, 587.
8 Y. Koyama, I. Tanaka and H. Adachi, J. Electrochem. Soc., 2000,
147, 3633.
9 E. Gonzalo, A. Kuhn and F. Garcia-Alvarado, J. Power Sources,
2010, 195, 4990–4996.
10 (a) I. Gocheva, K. Chihara, S. Okada and J. Yamaki, Lithium
Batteries Discussion 2011 – Electrode Materials – Archacon, France
(extended abstract);(b) A. Basa, E. Gonzalo, A. Kuhn and
F. Garcia-Alvarado, J. Power Sources, 2012, 207, 160–165.
11 W. Massa, Z. Kristallogr., 1980, 153, 201–210.
12 W. Massa and W. Ruedorff, Z. Naturforsch., B, 1971, 26, 1216–1218.
13 P. Heitjans and S. Indris, J. Mater. Sci., 2004, 39, 5091–5096.
14 M. Wilkening, V. Epp, A. Feldhoff and P. Heitjans, J. Phys. Chem. C,
2008, 112, 9291.
15 P. Heitjans, M. Masoud, A. Feldhoff and M. Wilkening, Faraday
Discuss., 2007, 134, 67.
16 E. Avvakumov, M. Senna and N. Kosova, Soft Mechanochemical
Synthesis: A Basis for New Chemical Technologies, Kluwer
Academic Publishers, Boston, 2001.
17 K. L. Da Silva, D. Menzel, A. Feldhoff, C. K€
ubel, M. Bruns,
A. Paesano, A. D€
uvel, M. Wilkening, M. Ghafari, H. Hahn,
F. J. Litterst, P. Heitjans, K. D. Becker and V.
Sepel
ak, J. Phys.
Chem. C, 2011, 115, 7209–7217.
18 A. D€
uvel, M. Wilkening, R. Uecker, S. Wegner, V.
Sepel
ak and
P. Heitjans, Phys. Chem. Chem. Phys., 2010, 12, 11251–11262.
19 E. Kemnitz, U. Groß, S. R€
udiger and C. S. Shekar, Angew. Chem.,
2003, 115, 4383–4386.
20 St. Wuttke, A. Vimont, J.-C. Lavaley, M. Daturi and E. Kemnitz,
J. Phys. Chem. C, 2010, 114, 5113–5120.
21 M. Ahrens, G. Scholz, M. Feist and E. Kemnitz, Solid State Sci.,
2006, 8, 798–806.
22 (a) D. L. Deadmore, J. S. Machin and A. W. Allen, J. Am. Ceram.
Soc., 1962, 45, 120–122; (b) Yu. N. Moskvich, B. I. Cherkasov and
G. I. Dotsenko, Zh.Strukt. Khim., 1979, 20, 348–357.
23 M. Wilkening, E. Romanova, S. Nakhal, D. Weber, M. Lerch and
P. Heitjans, J. Phys. Chem. C, 2010, 114, 19083.
24 T. L. Spencer, A. Ramzy, V. Thangadurai and G. R. Goward, Chem.
Mater., 2011, 23, 3105–3113.
25 P. Bottke, S. Nakhal, M. Lerch, M. Wilkening and P. Heitjans, Diff.
Fundam., 2011, 16, 40.
26 T. Ressler, WINXAS v3.1 2011.
27 J. Pebler, W. Massa, H. Lass and B. Ziegler, J. Solid State Chem.,
1987, 71, 87–94.
28 J. Wontcheu, W. Bensch, M. Wilkening, P. Heitjans, S. Indris,
P. Sideris and C. P. Grey, J. Am. Chem. Soc., 2008, 130, 288.
29 A. Johnson and G. W. Everett, J. Am. Chem. Soc., 1972, 94, 1419–
1425.
30 D. R. Eaton, J. Am. Chem. Soc., 1965, 87, 3087–3102.
31 P. M. Borodin and N. Kim Zao, Russ. J. Inorg. Chem., 1972, 16,
1720–1722.
32 W. F. Finnery, E. Wilson, A. Callender, M. D. Morris and
L. W. Beck, Environ. Sci. Technol., 2006, 40, 2572–2577.
33 R. Becker and W. Sawodny, Z. Naturforsch., 1973, 28b, 360.
34 M. Mikami, I. Nakagawa and T. Shiraanouchi, Spectrochim. Acta,
1967, 23A, 1037.
35 C. Naulin and R. Bougon, J. Chem. Phys., 1976, 64, 4155.
36 J. Kohl, D. Wiedemann, S. Troyanov and M. Lerch, in preparation.
37 Crystal data for [V(acac)
2
(CH
3
CN)
2
]BF
4
were collected on an Oxford
Diffraction Xcalibur S diffractometer equipped with an Enhance Mo-
K
a
source (graphite monochromated, l¼0.71073
A) and a Sapphire
3 detector at T¼150 K. The structure was solved with SHELXS-97
and refined against all F
2
data using the full-matrix least-squares
method of SHELXL-97. The tetrafluoroborate moiety is disordered
over an inversion centre. The positional disorder was not further
resolved resulting in somewhat higher displacement parameters.†
38 G. M. Sheldrick, Acta Crystallogr., Sect. A: Found. Crystallogr., 2008,
64, 112–122.
39 M. M. Islam, T. Bredow and A. Gerson, ChemPhysChem, 2011, 12,
3467–3473.
40 T. Bredow and A. R. Gerson, Phys. Rev. B: Condens. Matter, 2000,
61, 5194–5201.
41 R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R. Saunders and
C. M. Zicovich-Wilson, Z. Kristallogr., 2005, 220, 571–573.
42 http://www.crystal.unito.it/Basis_Sets.
43 N. Ferro, C. Reimann and T. Bredow, in preparation.
44 T. Roisnel, P. Nu~
nez, W. Massa and A. Tressaud, Phys. B, 1997, 234–
236, 579–581.
45 N. Ferro, M. Lerch, R. Glaum, H. Ehrenberg and T. Bredow, in
preparation.
This journal is ªThe Royal Society of Chemistry 2012 J. Mater. Chem., 2012, 22, 15819–15827 | 15827
Published on 05 July 2012. Downloaded by TU Berlin - Universitaetsbibl on 30/03/2016 14:15:57.
View Article Online
