www.afm-journal.

de

FULL PAPER
Silicon-Doped LiFePO4 Single Crystals: Growth,
Conductivity Behavior, and Diffusivity
By Ruhul Amin, Chengtian Lin, Jubo Peng, Katja Weichert, Tolga Acartürk,
Ulrich Starke, and Joachim Maier*

diffusion time by appropriate doping. Dop-

Single crystals of silicon doped LiFePO4 with a silicon content of 1% are ing is not only a powerful means to tune
transport properties, it is also an important
grown successfully by the floating zone technique and characterized by single-
means to investigate the underlying defect
crystal and powder X-ray diffraction, secondary ion mass spectroscopy, and chemistry. It should be mentioned that the
chemical analysis. Electron paramagnetic resonance demonstrates the ionic and electronic conductivities and Li-
presence of only Fe2þ; no traces of Fe3þ are found. Impedance spectroscopy diffusivity of doped samples of LiFePO4
as well as step-function polarization/depolarization (DC) measurements are have not been studied extensively, even in
carried out using the cells Ti/LiFe(Si)PO4/Ti and LiAl/LiI/LiFe(Si)PO4/LiI/LiAl. polycrystalline materials, let alone in single-
crystals, which have not been available in
The electronic and ionic conductivities as well as the Li-diffusivity of the
significant amounts.[18,21–23]
sample in the major crystallographic directions ([h00], [0k0], and [00l]) are
determined. Within experimental error the transport properties along the b-
and c-axes are found to be the same but differ significantly from the a-axis,
2. Results and Discussion
which exhibits lower values. Compared to undoped LiFePO4, Si-doping leads
to an increase of the ionic conductivity while the electronic conductivity
decreases, which is in agreement with a donor effect. The activation energies 2.1. Single-Crystal Growth and
of conductivities and diffusivities are interpreted in terms of defect chemistry Characterization
and relevant Brouwer diagrams are given.
The crystal grains were found to preferably
grow along the crystallographic b-axis,
parallel to the ingot axis (space group
1. Introduction [Pnma]). The ingot with dimensions of 5-mm diameter and 70-
mm length showed a crack-free and black appearance with a
Owing to environmental benignity, cost effectiveness, and quite metallic luster. The crystal wafers were cut and polished to a
ideal charge/discharge behavior, compounds of the olivine family mirror face, which exhibited a blackish-green color. Under a
LiMXO4 (M ¼ Fe, Mn, Ni, Co, V, and X ¼ P, As, Si, etc.,) are polarization light microscope only a few grains were found on the
promising materials as cathodes for Li-based secondary bat- periphery area. No inclusions, cracks, or grain boundaries were
teries.[1–3] The intercalation/deintercalation of Li occurs in a observed in the inner area of about 4-mm diameter. It should be
narrow range without changing the bulk structure while larger Li noted that constant heating temperature is necessary for the
variations involve formation of the MXO4 phase. The most growth of high quality single crystals and that a high quality
important member of the family is the iron-based member crystal emerges generally from an ingot with uniform diameter
LiFePO4 which has been studied extensively.[4–18] Despite its along the traveling direction of the crystal. This is indicative of a
numerous advantages, the main drawback of the material is its stable molten zone obtained by constant temperature during
intrinsically sluggish mass and charge transport. Therefore, growth (any fluctuation in the temperature may cause an unstable
various attempts have been made to improve the electrochemical molten zone, forming a non-uniform diameter ingot. Variation in
performance of the material.[4,5,15,17,19] One possibility is to the temperature leads to a super-cooling or super-saturation in the
reduce the diffusion time by size reduction of the particles (this molten zone and results in a poor quality crystal).
works particularly well if a mixed conducting nanometer-sized Single-crystal X-ray diffraction measurements were performed
network is superimposed);[20] another is to attempt to decrease on different orientations of a 2.5 mm
1.4 mm
0.85 mm
crystal. The representative pattern in Figure 1 clearly indicates
[*] Prof. J. Maier, Dr. R. Amin, Dr. C. T. Lin, J. Peng, Dr. K. Weichert, single-crystallinity of the olivine phosphate material.
T. Acartürk, Dr. U. Starke Powder X-ray diffraction analysis of the single crystals along
Max Planck Institute for Solid State Research different directions was performed after they were oriented using
D-70569 Stuttgart (Germany) the Laue X-ray technique (Fig. 2). The oriented crystal was tilted
E-mail: s.weiglein@fkf.mpg.de
with an angle of a few degrees during X-ray measurements. All
DOI: 10.1002/adfm.200801604 the peaks were indexed as [h00], [0k0], and [00l] of standard

Adv. Funct. Mater. 2009, 19, 1697–1704 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1697
 www.afm-journal.de

composition given is consistent with Si substituting Fe essentially

FULL PAPER

at the expense of a Li deficiency, which clearly leads to a donor

effect. However, other substitutions cannot be excluded. While
interstitial incorporation of Si is very unlikely owing to its
covalency, from a structure chemical point of view replacement of
4
PO34 by SiO4 is imaginable, but would lead to an acceptor effect.
An acceptor would be expected to be compensated by holes
(Fe:Fe ¼ Fe3þ ). This can be excluded in view of conductivity
(decreasing electronic conductivity) and electron paramagnetic
resonance (EPR) results (no paramagnetic species, i.e., Fe3þ, were
detected). However, another scenario is possible; that the donor
effect we observe results from SiO2 3
3 replacing PO4 . This would
0
be equivalent to Si P being strongly associated with an oxygen
vacancy VO yielding ðSiP VO Þ: or the more detailed ðSiO3 Þ:PO4 (note
that at the temperatures of interest oxygen vacancies may simply
be frozen and irreversibly trapped). To gain more details on this,
Figure 1. Single-crystal X-ray diffraction pattern of LiFe(Si)PO4 (crystal synchrotron and/or NMR measurements are necessary. Yet, for
size: 2.5 mm
1.4 mm
0.85 mm) with indexing of selected reflections. the defect chemical analysis and hence for the effect on the
In the pattern reflections for both Mo Ka1 and Mo Ka2 diffraction are visible. transport properties, only the effective sign of the impurity defect
is decisive, which ought to be positive.
The results obtained from secondary ion mass spectroscopy
LiFePO4. Lattice parameters were derived from both powder and (SIMS) analysis show that the silicon is not distributed
single-crystal X-ray diffraction followed by refinement. The a-, b-, homogeneously in the crystal. Rather, silicon rich areas and
and c-axis constants were estimated from single-crystal X-ray silicon depleted areas were found (Fig. 3a). Comparison of these
diffraction to be 10.3027(47) Å, 5.9916(36) Å, and 4.6909(22) Å, areas reveals a difference of about one order of magnitude in the
respectively. While the lattice constant along the a-axis (b-axis) was Si concentration. It is not expected that the inhomogeneities
only slightly increased (decreased), the lattice constant along the affect the below conclusion other than forcing us to refer to a
c-axis was considerably increased in comparison with the mean dopant level. The differently doped regions may
undoped single crystalline sample (undoped sample: contribute differently to the measured conductivities only as
a ¼ 10.2951(202) Å, b ¼ 6.0044(74) Å, and c ¼ 4.6727(62) Å, far as the anisotropy is concerned. In Figure 3b the depth profile
determined by single-crystal X-ray diffraction). analysis of the decisive elements (Li, Fe, P, Si) is shown. With
The exact composition of the single crystalline material the exception of lithium, the intensities are nearly constant in
obtained by inductively coupled plasma (ICP) analysis and the investigated area; small deviations were found only at the
hot carrier gas extraction method is Li0.98  0.02Fe0.98  0.02 surface, which presumably results from surface contamination.
Si0.009  0.009P0.98  0.02O4  0.01. In anticipation of results to be Other cations (i.e., Na, K, Mg, Ca, Al) were also detected within
described, it can be stated that Si exerts a donor effect. The the crystal, but their contributions seems negligible due to their
very weak intensity. They may stem from contamination of the
educts.

2.2. Impedance and DC Polarization Measurements on Ion

Blocking Cell Ti/LiFe(Si)PO4/Ti

Figure 4 represents impedance spectra taken on a single crystal

oriented along the three main crystallographic directions at
selected temperatures using the symmetric ion blocking cell Ti/
LiFe(Si)PO4/Ti. As can be seen from Figure 4, almost perfect
semicircles are obtained over the temperature range 40–230 8C.
The absence of a second semicircle is consistent with the absence
of twin/phase boundaries as confirmed by polarization micro-
scope and X-ray single-crystal diffraction.
The capacitance values are in the order of picofarads,
confirming that the impedance spectra observed are only due
to the bulk of the sample. The ideality of the semicircles enabled a
reliable deduction of the dielectric constant. The dielectric
Figure 2. XRD patterns of an oriented (by the Laue X-ray technique) constant as a function of temperature of the material is shown in
LiFe(Si)PO4 single crystal along the [00l], [0k0], and [h00] directions indexed Figure 5 for the three basic directions. It is almost constant within
in space group [Pnma]. the temperature range in each direction.

1698 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 1697–1704
www.afm-journal.de

FULL PAPER
Figure 4. Impedance spectra for a- and b-orientations at selected tem-
peratures (c-direction qualitatively the same) using ion blocking cell Ti/
LiFe(Si)PO4/Ti.

during polarization and, after having reached a steady state, a

symmetrical decay of the voltage during depolarization with
rate constants being determined by the chemical diffusion
Figure 3. SIMS analysis: a) intensity spectrum for Si (100
100 mm, sum coefficient of Li DdLi [24–26]). The same features have been found for
over 1 500 scans). b) Depth profile of the marked region in (a).
[h00] as well as [00l] oriented crystals during DC polarization/
depolarization measurements.

As the Ti electrode is reversible for electronic exchange but

blocks the lithium ions,[18] the fact that no additional polarization
process was observed at very low frequencies for all orientations
indicates that the conduction along the three orientations is
predominantly due to electronic carriers and not due to lithium
ions.
This conclusion is clearly confirmed by DC polarization and
depolarization measurements made on the same cell for the
different orientations. As can be seen from Figure 6, the above-
mentioned cell with a [0k0]-oriented crystal exhibits a near-zero
open circuit voltage while, during galvanostatic DC polarization,
the voltage increases in a step-function manner to a stationary
value which analogously decays in the form of a step-function on
reducing the DC current to zero. This behavior, which has also
been observed and described for undoped LiFePO4,[18] is
characteristic of an electronically dominated conduction process
and again shows the high reversibility of the Ti electrodes for
electron transfer (in the case of significant contributions by ionic Figure 5. Dielectric constant of the LiFe(Si)PO4 single crystal for the three
motion one would observe a continuous increase of the voltage main crystallographic directions (heating and cooling cycle).

Adv. Funct. Mater. 2009, 19, 1697–1704 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1699
 www.afm-journal.de

pure and doped samples had occurred, not only would the
FULL PAPER

electronic conductivity drop, but we would also have to expect a

drop in the ion conductivity. As will be reported below, the ion
conductivity, however, increases on Si-doping, which is perfectly
in agreement with the donor effect, leading to a decrease in the
hole concentration and an increase in the lithium vacancy
concentration.[26,28]
Now let us turn to the activation energy of 0.65 eV, which is
close to the value achieved for the pure single crystal. In Ref. [28]
this activation energy was successfully explained by the sum of
the energy necessary for dissociating holes out of neutral Li-
vacancies (0.6 eV) and the energy necessary for holes migration
(0.05 eV) from Fe3þ to Fe2þ.

X 0
VLi )VLi þ h
* (1)

Figure 6. DC polarization and depolarization measurement of the ion

blocking cell Ti/LiFe(Si)PO4/Ti. The value of 0.6 eV for the reaction energy connected with
Equation 1 is very close to the value calculated in Ref. [12]. In spite
of the same dopant concentration, unlike in Al-doping an ionic
association regime is not seen in the investigated temperature
The electronic conductivities of the doped material along regime, indicating low energies for the association of the silicon-
different directions are depicted in Figure 7. The conductivity defect with lithium vacancies.
along the b- and c-axes is almost the same within experimental
error, while it is significantly less along the a-axis. However, the
activation energy is the same for all these directions (a-axis:
2.3. Impedance and DC Polarization Measurements of the
0.64 eV, b-axis: 0.66 eV, c-axis: 0.66 eV). It appears that the
electronic conductivity is effectively two-dimensional, similar to Electron Blocking Cell LiAl/LiI/LiFe(Si)PO4/LiI/LiAl
the undoped single crystals. The electronic conductivity of the
doped sample is decreased in each direction for almost one order Impedance spectra of the electron blocking cell at selected
of magnitude in comparison with the undoped LiFePO4.[18] This temperatures along [00l] orientation are presented in Figure 8. In
is in agreement with p-type conduction, which we already contrast to the results from the ion blocking cell (Fig. 4), the
concluded from the dependence of s on the Li content.[27] impedance spectra now consist of one semicircle at high
Although it is not probable, we have to rule out the possibility that frequencies and a Warburg response at low frequencies. The
distinctly different Li stoichiometries in the pure and Si-doped high-frequency semicircle gives the total impedance due to
single crystal are responsible for the conductivity decrease, as electronic and ionic motion, including minor contributions from
different Li contents also lead to different ionic and electronic contact impedance and bulk conductivity of LiI, and the Warburg
carrier concentrations. If, however, strongly different Li losses in response is caused by the stoichiometric polarization as a
consequence of blocking electrons. The impedance spectra for
the other orientations, namely [h00] and [0k0], are qualitatively
similar. The conductivity data derived from the bulk resistance of
the semicircle at high frequencies agree with those obtained from
AC and DC experiments of the Ti/LiFe(Si)PO4/Ti cell. The

Figure 7. Electronic and ionic conductivities of the Si-doped sample along

the three main crystallographic directions. For comparison electronic and
ionic conductivity for nominally undoped LiFePO4 along b-axis (taken from Figure 8. Impedance spectra of a LiAl/LiI/LiFe(Si)PO4/LiI/LiAl cell along
Ref. [18, 22]) are given. the c-axis.

1700 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 1697–1704
www.afm-journal.de

Warburg response (see Fig. 8) indicates that the ionic conductivity

FULL PAPER
is less than the electronic conductivity along all three directions,
similar as for the pure LiFePO4 single crystal.[18,22]
For the purpose of quantification we performed DC polariza-
tion/depolarization measurements of the electron blocking cell
arrangement.[20,25] Immediately after switching on a constant
current I both ionic and electronic charge carriers flow
unimpeded according to their transference numbers while the
voltage jumps from zero to IReonRion/(Reon þ Rion). Rion and Reon
refer to the resistances due to Li ions and electronic carriers. With
increasing time, the partial current of the blocked electrons
decreases, eventually leading to a steady state (voltage being IRion)
in which electronic current has vanished and the total current is
carried by the non-blocked ions. The relaxation time of the
polarization process is td, from which the chemical diffusion
coefficient DdLi can be calculated via Figure 10. Extraction of ionic conductivity and diffusivity by fitting the
obtained cell voltage.

L2
td ¼ (2) direction the cell voltage reached a steady state situation (the same
p2 DdLi was observed for [00l] orientation). The cell voltage along [h00],
however, did not reached a steady state value even at higher
temperatures and for longer waiting times (see Fig. 9b). In all
Owing to the internal concentration profiles, sion and DdLi are cases the depolarization was analogous.
averaged over a Li composition range which corresponds to the The long-time polarization voltage depends on time according
polarization voltage 70 mV.[26] to[25,26]
Figure 9 displays the time dependence of the polarization
voltage (galvanostatic mode) for [h00] and [0k0] orientations at 

ip L s eon ip L 8 t
selected temperatures. We can see from Figure 9a that in the [0k0] Uion ¼ þ 1  2 exp  (3)
s s s ion p td

Accordingly, a plot of lnjU(t)  U(t ¼ 1)j versus time should

give a straight line with t d a D1d as a slope. Figure 10 shows
Li
that this is nicely fulfilled for the [00l] orientation. Analogously,
we evaluated DdLi and sion for the [0k0] direction. While for [00l]
and [0k0] we can determine U(t ¼ 1) experimentally (at least for
higher temperatures), it was fitted in the case of [h00]. Hence,
here the transport parameters are less accurate. Both DdLi
(calculated from the transient) and sion (calculated from
U(t ¼ 1)) are obtained from the long-time behavior. The short-
pffi
time response (ideally described by a t law) is more complex as it
includes also space charge polarization and polarization of other
interfaces. The same is true for the onset Warburg response
observed in the AC experiments (Laplace transform of short-time
DC transients).
In Figure 7 the ionic conductivity data of LiFe(Si)PO4 are
displayed and compared for all directions as a function of inverse
temperature. It can be seen that the electronic conductivity is
more than two orders of magnitude higher than the ionic
conductivity. The data are also compared with pure LiFePO4 (the
pure samples are only slightly donor doped owing to a small
fraction of Fe on Li sites (Fe:Li )). Figure 7 shows that upon Si-
doping the electronic conductivity decreases while the ionic
conductivity increases, compared to nominally undoped single
crystalline LiFePO4.[18,22] This is in line with lithium vacancies
0
(VLi ) and holes (h.) being the decisive carriers and rules out the
possibility that different frozen Li stoichiometries are the cause,
Figure 9. DC polarization measurements along different directions using as this would affect VLi 0
and h similarly. It is worth mentioning
electron blocking cell LiAl/LiI/LiFe(Si)PO4/LiI/LiAl. that the sion values along the b- and c-axes are only a little different

Adv. Funct. Mater. 2009, 19, 1697–1704 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1701
 www.afm-journal.de

in the measured temperature range. The ratio is approximately

FULL PAPER

1.3, while the ratio for the values along the b- and a-axes is about
half an order of magnitude higher (5.0). This is similar to what
was observed for the Al-doped[23] and nominally pure sam-
ple,[18,22] the first factor being even closer to unity for the pure
one. Obviously ionic conductivities, and even more so the
electronic conductivities, are effectively two-dimensional.

2.4. Transient and Chemical Diffusivity

Given predominant electronic conductivity, DdLi should be

proportional to
Figure 12. Defect concentration as a function of Li-activity (reversible
  conditions; jzj ¼ d for the silicon defect, jzj ¼ 1 otherwise).
xion xeon
s ion þ (4)
cion ceon
silicon impurity. In view of similar doping concentrations this
where xion and xeon refer to the contributions due to trapping of indicates a lower absolute value of the association energy. As we
0
ionic and electronic carriers and cion, ceon denote to the relevant always stay in the ionic disorder regime (D-regime), VLi / C (C:
ionic and electronic carrier concentrations.[29] Thus, it is expected total doping content) and the ion migration energies follow as
that the chemical diffusion coefficients should exhibit similar 0.70 eV (or 0.9 eV for the a-axis). This value is in approximate
agreement with modeling studies.[12] The coupling with the hole
orientational anisotropies to the ion conductivity. Indeed, the DdLi
concentration is provided by the ionization of neutral lithium
values presented for the [0k0] and [00l] orientations in Figure 11
vacancies.[28] Based on these simple mass action concentra-
show almost isotropic behavior and are distinctly greater than for
tions we derive the Brouwer diagrams, log[defect] vs. log aLi and
[h00] direction indicating a preferential chemical diffusion vs. log[D] (with D ¼ Si) which are given in Figures 12 and 13.
coefficient of Li in the b–c plane. It should be mentioned that As discussed previously, donor doping by Si is made
the diffusivity data obtained from the Al-doped sample[23] are possible through Si replacing Fe (Si::Fe ) or by SiO2 replacing
3
approximately one order of magnitude and from Si-doped sample D 0
PO3 ::
4 ððSiO3 Þ_PO4 ¼ ðSi P VO ÞÞ. As we do not know the dopant’s
are almost two orders of magnitude higher than for the pure nature we abbreviate it by Dd .
.
single crystal LiFePO4.[18,22] In Refs. [22] and [28] we gave a detailed analysis of DdLi . In the
extreme case of VLi X
 ½h: the activation energy for DdLi is given
by hV  DH1 þ DHLi while it is given by hV þ DHLi in the opposite
2.5. Defect Chemistry case of VLiX

½h: (hV : migration energy of lithium vacancies;
DH1: reaction enthalpy for the trapping of h by VLi
. 0
; DHLi:
0
Now let us focus on the activation energy of electronic and ionic reaction enthalpy of Li incorporation under consumption of VLi
.
and h ). Using 0.7 eV (b, c) for hV and 0.6 eV for DH1 (VLi 0
þ h:
transport. The activation energies of ion conduction along the b-
and c-axes are similar (0.70 and 0.73 eV) and less than the a-axis association energy), the experimental DdLi -activation energy of
(0.90 eV), whereas the activation energies of electron conduction 1 eV (a-axis: 1.0 eV, b-axis: 1.01 eV, c-axis: 1.02 eV) would then
are approximately 0.65 eV for all directions. Unlike with Al-doped correspond to DHLi ¼ 0.3 eV (first case) or DHLi ¼ þ 0.3 eV
samples,[23] we do not see an ionic association regime with the

Figure 13. Dependence of defect concentration on doping content

Figure 11. Lithium diffusion coefficient DdLi and activation energies E along 0
[D] ¼ [Si] (without association between VLi and Dd.; jzj ¼ d for the silicon
the three main crystallographic directions. defect, jzj ¼ 1 otherwise).

1702 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 1697–1704
www.afm-journal.de

(second case). The second, positive value (in any case referring to The chemical composition was determined by ICP analysis and optical

FULL PAPER
unrealistic conditions) can be discarded. But also the value of emission spectroscopy (SPECTRO CIROS) as well as a hot-carrier-gas
extraction method (for oxygen, ONH-2000 ALTRA). EPR (X-band, Bruker ER
0.3 eV seems too low in magnitude in view of stability
041 Microwave Bridge) was used to identify paramagnetic impurities and
questions.[28] More reasonable values would be achieved by Fe3þ.
using expressions of the strongly associated regime,[23] but this The SIMS profiles were taken using TOF-SIMS IV from Ion-Tof (primary
would not be in line with the interpretation of seon and sion. In gun: Gaþ, 15 keV, 1.8 pA, analysis area 100 mm
100 mm. Sputtering gun
any case, the analysis of DdLi is much less reliable than for seon Oþ2 , 2 keV, 400 nA, sputtering area 500 mm
500 mm). Only the positive
and sion owing to higher complexity, lesser accuracy and in ions were measured.
Transport measurements were performed on oriented, well-polished
particular due to the fact that aLi was not kept constant during the
single crystals by employing the ion blocking cell Ti/LiFe(Si)PO4/Ti as well
experiment as assumed for the above analysis.[23,28] as the electron blocking cell LiAl/LiI/LiFe(Si)PO4/LiI/LiAl, both of which
had already been used for pure LiFePO4 single crystals [18,22]. For the ion
blocking cell, titanium metal films were evaporated onto the polished
surfaces with thickness of 300 nm. To establish the electron blocking cell
3. Conclusions the LiFe(Si)PO4 single crystal was contacted by LiI (Aldrich, 99.99%)
Si-doped single crystals of LiFePO4 have been successfully grown. pellets. A LiAl alloy whose polarization is small and quickly saturated was
used as the electrical contact. The contact resistance was minimized by a
We measured their electronic and ionic conductivities as well as
heat treatment at 200 8C for 20 h. The cell assembly was done inside a glove
Li diffusivities and explained them in terms of our defect box with controlled atmosphere (H2O/O2 content < 1 ppm).
chemical model given earlier.[28] Si-doping exerts a donor effect All electrochemical experiments for both cells were carried out inside an
and hence increases ionic and decreases electronic conductivity air-tight quartz tube under dry argon atmosphere. The total conductivity
(which indicates p-type conduction). The interpretation fits the was measured by using a high-resolution dielectric analyzer (Novocontrol)
general picture of a potentially Li-deficient hole-conducting in the frequency range of 107 to 102 Hz. Two point galvanostatic step
function experiments were performed in order to quantitatively determine
material, in which,
owing to the low temperature, neutral lithium
X electronic/ionic conductivity (steady state response) and chemical lithium
vacancies VLi also play a significant role. As in our earlier diffusivity (transient response, electron blocking cell).
samples the transport properties are anisotropic.[18,23] We found
similar conductivity and diffusivity values in the b–c plane,
whereas the values for the a-direction are considerably less. The
activation energy for electronic conduction can be explained in a
straightforward way by hole hopping (0.05eV) and Acknowledgements
hole
0
association (0.6 eV), while the ionic association VLi Si:Fe seems We would like to thank Dr. Constantin Hoch and Gabi Götz for single-
negligible in the temperature range considered. crystal X-ray and powder X-ray measurements, respectively. We also
acknowledge Dr. D. P. Chen for helping with the crystal growth and Helmut
Wendel and Annette Zechmeister for crystal preparation.

4. Experimental Received: October 29, 2008

Published online: April 14, 2009
Single crystals of Si-doped LiFePO4 were grown using the optical
floating zone technique. The source materials Li2CO3 (p.a. Merck),
FeC2O4  2H2O (98%, Fluka), NH4H2PO4 (97%, Aldrich), and [1] A. K. Padhi, K. S. Nanjundaswamy, J. B. Goodenough, J. Electrochem. Soc.
Si(OOCCH3)4 (Aldrich) were mixed with acetone and milled for 4 h using 1997, 144, 1188.
zirconia-milling media. Afterwards, the mixture was dried in an argon flow [2] M. E. A. Dompablo, U. Amador, M. Alvarez, J. M. Gallardo, F. G. Alvarado,
for 24 h. The typical product of 14 g is yielded from the final powder mixture. Solid State Ionics 2006, 177, 2625.
It is emphasized that the mixture has to be stored in a glove box (<3 ppm [3] M. Ram, R. N. P. Choudhary, A. K. Thakur, Adv. Appl. Ceram. 2006, 105, 140.
H2O and O2) to avoid oxidation and water absorption in air. After drying, [4] S.-Y. Chung, J. T. Bloking, Y.-M. Chiang, Nat. Mater. 2002, 1, 123.
the mixture was calcined in an Ar atmosphere at 350 8C for 10 h and then at [5] S. P. Herle, B. Ellis, N. Coombs, L. F. Nazar, Nat. Mater. 2004, 3, 147.
600 8C for a day. Finally it was slowly cooled down to room temperature at a [6] C. Delacourt, P. Poizot, J.-M. Tarascon, C. Masquelier, Nat. Mater. 2005, 4,
rate of 120 8C h1. The obtained polycrystalline powder is composed of 254.
single phase LiFe(Si)PO4 as indicated by X-ray powder diffraction and ICP [7] P. P. Prosini, M. Lisi, D. Zane, M. Pasquali, Solid State Ionics 2002, 148, 45.
analysis. The polycrystalline material was reground and shaped into [8] F. Sauvage, E. Baudrin, L. Gengembre, J.-M. Tarascon, Solid State Ionics
cylindrical bars of 6 mm in diameter and 100 mm in length by pressing at 2005, 176, 1869.
isostastic pressure of 70 MPa. The feed rod was sintered at 810 8C for [9] V. Srinivasan, J. Newman, Electrochem. Solid-State Lett. 2006, 9, A110.
10 h in an Ar flow. After sintering the rod was cut to a length of 20 mm to [10] S.-Y. Chung, Y.-M. Chiang, Electrochem. Solid-State Lett. 2003, 6, A278.
serve as a first seed and hereafter the grown crystal was used as a seed. The [11] C. Delacourt, L. Laffont, R. Bouchet, C. Wurm, J.-B. Leriche, M. Morcrette,
feed rod and the growing crystal were rotated in opposite directions at a
J.-M. Tarascon, C. Masquelier, J. Electrochem. Soc. 2005, 152, A913.
rate of 15 rpm. We applied a variety of growth rates between 2 and 4 mm
[12] T. Maxisch, F. Zhou, G. Ceder, Phys. Rev. B 2006, 73, 104301.
h1 under pure argon flow of 300 mL min1 for optimizing the total growth
[13] D. Morgan, A. Van der Ven, G. Ceder, Electrochem. Solid-State Lett. 2004, 7,
time and to minimize the volatilization. A rate of 2 mm h1 was applied at
an initial state of an approximately 8 mm long rod grown and thereafter a 30.
faster rate of 4 mm h1 was used to complete the growth. Details of the [14] C. Ouyang, S. Shi, Z. Wang, X. Huang, L. Chen, Phys. Rev. B 2004, 69,
general growth technique are described elsewhere [21]. 104303.
The crystal orientations were controlled by Laue X-ray technique and [15] M. S. Islam, D. J. Driscoll, C. A. J. Fisher, P. R. Slater, Chem. Mater. 2005, 17,
additionally confirmed by X-ray powder diffraction (using Cu-Ka radiation in 5085.
a Philips PW 3020 diffractometer). Furthermore single-crystal X-ray [16] N. Ravet, A. Abouimrane, M. Armand, Nat. Mater. 2003, 2, 702.
diffraction was performed (Diffractometer: IPDSII (Stoe, Darmstadt), [17] R. Dominko, M. Bele, M. Gaberscek, M. Remskar, D. Hanzel, S. Pejovnik, J.
radiation: Mo-Ka, graphite monochromator). Jamnik, J. Electrochem. Soc. 2005, 152, A607.

Adv. Funct. Mater. 2009, 19, 1697–1704 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1703
 www.afm-journal.de

[18] R. Amin, P. Balaya, J. Maier, Electrochem. Solid-State Lett. 2007, 10, A13. [24] C. Wagner, Proceedings of 7th Meeting of the International Committee on
FULL PAPER

[19] C. Delacourt, P. Poizot, S. Levasseur, C. Masquelier, Electrochem. Solid- Electrochemical Thermodynamics and Kinetics, Lindau 1955, Butterworth,
State Lett. 2006, 9, A352. London 1957.
[20] Y.-G. Guo, Y.-S. Hu, W. Sigle, J. Maier, Adv. Mater, 2007, 19, 2087. [25] I. Yokota, J. Phys. Soc. Jpn. 1961, 16, 2213.
[21] D. P. Chen, A. Maljuk, C. T. Lin, J. Cryst. Growth 2005, 284, 86. [26] J. Maier, Physical Chemistry of Ionic Materials: Ions and Electrons in Solids,
[22] R. Amin, P. Balaya, C. T. Lin, D. P. Chen, J. Maier, Solid State Ionics 2008, Wiley, Chichester 2004.
179, 1683. [27] R. Amin, J. Maier, Solid State Ionics 2008, 178, 1831.
[23] R. Amin, C. T. Lin, J. Maier, Phys. Chem. Chem. Phys. 2008, 10, 3519 (I), [28] J. Maier, R. Amin, J. Electrochem. Soc. 2008, 155, A339.
3524 (II). [29] J. Maier, J. Am. Ceram. Soc. 1993, 76, 1212.

1704 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 1697–1704