Alkali metals extraction reactions with the silicides Li15Si4 and Li3NaSi6: Amorphous Si versus allo -Si

Article abstract of DOI:10.1021/cm503371e

The silicides Li₁₅Si₄ and Li₃NaSi₆ were subjected to chemical extraction of the alkali metal component by liquid ammonia and ethanol, respectively, which after washing yielded black powders of amorphous silicon. The investigated reactions are interesting with respect to both the formation of novel Si modifications and the delithiation process in Si anode materials. The products termed a-Si (from Li₁₅Si₄) and a-allo-Si (from Li₃NaSi₆) were characterized by a combination of methodologies (X-ray diffraction, transmission electron microscopy, differential thermal analysis, Raman, atomic absorption, and energy-dispersive X-ray spectroscopy) which revealed (i) a porous microstructure for a-Si built from spherically shaped particles with sizes around 10 nm, (ii) partial surface oxidation of both materials and (iii) the presence of nanocrystalline Si in both materials. The result of the protic oxidation of Li₃NaSi₆ is at variance with earlier findings reporting the formation of a crystalline bulk allotrope of silicon (allo-Si) from the topotactic combination of silicon layers present as polyanions in Li₃NaSi₆. Additionally, quantum chemical calculations show that silicon layers in Li₃NaSi₆ cannot combine to energetically favorable allotropic forms of Si. This is different from Li₇Ge₁₂, where polyanionic germanium layers topotactically convert to the germanium allotrope m-allo-Ge upon oxidation.

Full text of DOI:10.1021/cm503371e

Article
pubs.acs.org/cm
Alkali Metals Extraction Reactions with the Silicides Li15Si4 and
Li3NaSi6: Amorphous Si versus allo-Si
Michael Zeilinger,† Laura-Alice Jantke,† Lavinia M. Scherf,† Florian J. Kiefer,† Gero Neubuser,‡
̈
Lorenz Kienle,‡ Antti J. Karttunen,§ Sumit Konar,∥ Ulrich Haussermann,∥ and Thomas F. Fassler*,†
̈
̈
†Department of Chemistry, Technische Universitat Munchen, Lichtenbergstraße 4, 85747 Garching b. Munchen, Germany
̈
̈
̈
‡Faculty of Engineering, Christian-Albrechts-University, Kiel, Kaiserstrasse 2, 24143 Kiel, Germany
§Department of Chemistry, Aalto University, FI-00076 Aalto, Finland
∥Department of Materials and Environmental Chemistry, Stockholm University, 10691 Stockholm, Sweden
S
* Supporting Information
ABSTRACT: The silicides Li15Si4 and Li3NaSi6 were subjected to chemical extraction of the alkali metal component by liquid
ammonia and ethanol, respectively, which after washing yielded black powders of amorphous silicon. The investigated reactions
are interesting with respect to both the formation of novel Si modifications and the delithiation process in Si anode materials.
The products termed a-Si (from Li15Si4) and a-allo-Si (from Li3NaSi6) were characterized by a combination of methodologies
(X-ray diffraction, transmission electron microscopy, differential thermal analysis, Raman, atomic absorption, and energy-
dispersive X-ray spectroscopy) which revealed (i) a porous microstructure for a-Si built from spherically shaped particles with
sizes around 10 nm, (ii) partial surface oxidation of both materials and (iii) the presence of nanocrystalline Si in both materials.
The result of the protic oxidation of Li3NaSi6 is at variance with earlier findings reporting the formation of a crystalline bulk
allotrope of silicon (allo-Si) from the topotactic combination of silicon layers present as polyanions in Li3NaSi6. Additionally,
quantum chemical calculations show that silicon layers in Li3NaSi6 cannot combine to energetically favorable allotropic forms of
Si. This is different from Li7Ge12, where polyanionic germanium layers topotactically convert to the germanium allotrope m-allo-
Ge upon oxidation.
a ∼10 times higher capacity in Li ion batteries when compared
to commercially used graphite.16−18 Generally, Li15Si4 and
amorphous Si phases are discussed during charging and
discharging processes. Although there are numerous reports
on the electrochemical lithiation of Si to form amorphous or
crystalline Li−Si alloys and delithiation to form amorphous Si, a
chemical delithiation of presynthesized Li15Si4 has not been
reported.19−22
1. INTRODUCTION
The search for new forms of semiconducting silicon and
germanium attracts increasing attention because novel or
superior electronic transport and optical properties−being
real or envisioned−will allow improving performance and
scalability of devices based on established semiconductor
technologies.1,2 Desired new crystalline forms range from
nanomaterials (particles, wires, tubes),1 nano- and mesostruc-
tures (e.g., nano- and mesoporous germanium)3−7 to new
allotropes.8−13 Amorphous forms, which are widely used in thin
film transistors and photovoltaic applications, are being tailored
by impurity doping or by compositing with nanocrystalline
particles.14,15
In general, Zintl phases comprising alkali or alkaline earth
metals and silicon or germanium represent versatile precursors
toward new forms of silicon and germanium. Metal extraction
Received: September 15, 2014
Revised: October 28, 2014
Nanosilicon is also in the midpoint of the search for
electrochemically active anode materials since silicon possesses
© XXXX American Chemical Society
A
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
from solid Zintl phases or the oxidation of Zintl anions in
solution are established routes. Examples include the crystalline
clathrate-II allotrope of silicon from simple thermal decom-
position of NaSi into NaxSi136 and subsequent Na extraction
with I2,23 the crystalline germanium clathrate from the
the chemical delithiation of Li15Si4. However, there are subtle
differences between both materials, which are introduced by the
simultaneous formation of nanocrystalline Si.
2. EXPERIMENTAL SECTION
oxidation of Ge9 Zintl ions in ionic liquids,24 and the
4−
Solid-State Synthesis. Starting materials were lithium rods (99%,
Rockwood-Lithium), sodium rods (99%, Chempur) and silicon
powder (99.999%, Sigma-Aldrich). All steps of solid state synthesis
and preparation of air- and moisture-sensitive samples were carried out
in a glovebox (MBraun, Ar-atmosphere, H2O and O2 levels <0.1 ppm).
Metal ampules (tantalum and niobium) for solid state synthesis were
thoroughly cleaned, heated to 1000 °C under dynamic vacuum (p < 1
× 10−3 mbar) for at least 2 h and transferred to the glovebox. An all-
glass Schlenk line supplied with Ar, which is dried over P2O5,
molecular sieve and heated titanium sponge (T = 750 °C), was used
for baking out ampules, encapsulating ampules in silica tubes, handling
air- and moisture-sensitive materials, and reactions in inert
atmosphere.
isolation of silicon sheets with a two-dimensional graphitic
structure by the redox assisted exfoliation of CaSi2.25−27 Metal
extraction allows also for bulk amorphous and nanocrystalline
samples. This was shown by McMillan et al.28 and Wang et
al.,14 respectively, by applying a solid state metathesis (SSM)
reaction between NaSi or NaGe and NH4X (X = Cl, Br), where
NaX is subsequently removed by washing with alcohol or water.
NaSi may also be directly oxidized by alcohols to yield
nanosized amorphous Si.29
Zintl phases with layered polyanions are especially interesting
for obtaining new allotropes because of their potential for
topotactic reactions. Ideally, metal extraction will lead to a
kinetically favored (low barrier) oxidative coupling of two-
dimensional polyanions, retaining (or only slightly rearranging)
their topology. Deintercalation reactions with Li7Ge12 and
Li3NaSi6, whose layered structures are shown in Figure 1, were
Li15Si4. Bulk samples of Li15Si4 were synthesized according to ref 33
by mechanical ball-milling of stoichiometric mixtures of elemental Li
and Si in a RETSCH PM 100 planetary ball mill. The procedure yields
large quantities (up to several grams) of X-ray pure Li15Si4.
Li3NaSi6. Li3NaSi6 was typically synthesized according to
Li12Si7 + 4NaSi + 13 α‐Si → 4Li3NaSi6
Stoichiometric mixtures of reactants with a total mass of 0.5 g were
homogenized in an agate mortar, pressed to a pellet and sealed in a
metal ampule which was then encapsulated in a silica jacket under
vacuum. Ampules were heated in a muffle furnace to 550 °C at a rate
of 5 K min−1 and annealed for 24 h. After slow cooling at 0.075 K
min−1 the samples were recovered inside the glovebox. The procedure
yields X-ray pure samples of Li3NaSi6 (see Figure S1a in the
Supporting Information). Li3NaSi6 may also be obtained from
annealing ball-milled mixtures of Li15Si4, Na, and α−Si (RETSCH
PM 100 planetary ball mill, duration: 70 min, 250 rpm, 10 min
interval, 5 min break, direction reversal, ball charge 3, ball size 15 mm
diameter) with molar ratios 1:5:26 at 600 °C (total mass, 2.0 g;
heating rate, 4 K min−1; dwell time, 7 days; cooling rate, 0.075 K min−1
to room temperature) or from annealing ball-milled stoichiometric
mixtures of the elements at 550 °C (total mass, 0.5−3.0 g; heating rate,
1−2 K min−1; dwell time, 5−7 days; cooling rate, 0.1 K min−1 to room
temperature). These samples always contained small amounts of
impurities (Li12Si7 and/or α−Si, cf. Figure S1b in the Supporting
Information). It was later recognized that when using Nb ampules,
Li3NaSi6 samples also contained small amounts of NbSi2 although this
impurity could not be detected in powder X-ray diffraction patterns.
Li12Si7 and NaSi. The binary precursors for Li3NaSi6 were
synthesized from stoichiometric mixtures of the elements with a
total mass of 2.0 g each. The reactions were carried out in tantalum
ampules which were encapsulated in silica jackets under vacuum.
Sample annealing was done in a muffle furnace. For Li12Si7, the
temperature was raised to 750 °C at 5 K min−1, followed by 1 h
dwelling, cooling to 400 °C at a rate of 0.5 K min−1, dwelling for 6 h
and subsequent quenching. For NaSi, the temperature was raised
stepwise to 600 °C: 5 K min−1 to 500 °C, 1 h dwelling, 5 K min−1 to
600 °C. After a dwell time of 30 h a cooling rate of 5 K min−1 was
applied. Samples were recovered inside the glovebox.
Figure 1. Zintl phases Li7Ge122 and Li3NaSi6 containing polyanionic
layers (a) 2∞[Ge12]7− and (b) ∞[Si6]4− (black) being separated by Li
(red) and Na (yellow) atoms.
already performed in the 1980s and yielded new allotropic
forms of Ge and Si, named allo-Ge and allo-Si, respectively.8−11
However, the structures of these allo forms eluded conclusive
characterization. Only recently, a bulk synthesis strategy for
microcrystalline allo-Ge (m-allo-Ge) was established and the
structure of allo-Ge could be determined by a combination of
experimental and quantum chemical methodologies.13,30 In-
2
deed, the five-ring topology of the initial ∞[Ge12]7− layers is
strictly retained in m-allo-Ge. From m-allo-Ge, another
allotrope of Ge−hexagonal 4H-Ge−can be obtained which is
exclusively built from six-rings.9,12 Subsequently, the thermo-
chemistry and optical properties of these Ge allotropes were
investigated in detail.31 The situation for allo-Si remains
unclear. In 2002, Conesa suggested structure models for allo-
Si based on the ∞[Si6]4− layers in Li3NaSi6,32 however,
2
experimental validation is still missing.
Extraction Reactions. One half gram of Li15Si4 was loaded into a
custom-built extractor for liquid ammonia (Westfalen AG; dried over
sodium). The apparatus, shown in Figure 2, is connected to a high-
vacuum glass line that is supplied with purified argon (see above). The
extraction process (taking place at −78 °C by using an isopropanol dry
ice bath) was stopped after the color of the liquid ammonia solution
had turned from deep to light blue after approximately 40 extraction
cycles. After evaporating ammonia from the extractor, the solid black
product consisting of amorphous silicon and LiNH2 was recovered in
the glovebox. To remove LiNH2, 157 mg of the mixture were filled
into a Schlenk flask. The powder was suspended in 20 mL of dry
toluene (dried over Na). Subsequently, 50 mL of ethanol (dried over
Here, we describe metal extractions with Li3NaSi6 and
Li15Si4. The former were aimed at renewed efforts into the
synthesis and structural characterization of allo-Si, the latter to
emulate by chemical means the delithiation process occurring
electrochemically in silicon based lithium batteries (cycling Si
against Li) where amorphous and crystalline Li15Si4 correspond
to the fully lithiated (discharged) state.19,20 We solve the
quarter-century old puzzle of allo-Si and show that allo-Si is not
accessible from Li3NaSi6. Instead, products essentially corre-
spond to amorphous Si. Amorphous Si is also the product of
B
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
reactions were modified using i-PrOH, air, H2O, HNO3, HOAc, and
Ph2CO instead of EtOH. For details, see ref 35 and Table 1.
Powder X-ray Diffraction (PXRD). For examining the phase
purity of precursor materials and phase analysis of products, laboratory
PXRD patterns were recorded on a Stoe STADI P diffractometer
(Ge(111) monochromator for CuKα1 radiation, λ = 1.54056 Å)
equipped with a Dectris MYTHEN DCS 1K solid state detector.
Samples were ground in agate mortar and sealed inside 0.3 mm glass
capillaries. All measurements were performed in a 2θ range of 5−90°
(PSD steps, 0.06−1.00°; time/step, 10−40 s). Rietveld refinements of
crystalline samples were carried out with the TOPAS 4.0 software.36
High-energy PXRD measurements for a-Si and a-allo-Si were carried
out at the beamline P02.1, at PETRA III, DESY (Hamburg). This
beamline operates at a fixed energy of approximately 60 keV. The
wavelength has been determined to be 0.20727(6) Å by using a LaB6
NIST standard. The beam size is approximately 0.8 mm in diameter.
The materials were enclosed in 0.7 mm glass capillaries and placed in
front of the detector. For both samples, 30 2D diffraction images, each
obtained through the accumulation of 20 frames with an exposure time
of one second per frame, were collected with a PerkinElmer
amorphous silicon area detector (XRD1621) placed at 329 mm
from the sample. The distance between the sample and the detector
was determined with a LaB6 NIST standard. The 2D diffraction images
were then integrated into a linear scattering signal with the software
Fit2D and averaged.37 Pair distribution functions (PDFs) were
extracted by using the PDFgetX3 software.38 The different steps of
the PDF extraction include the subtraction of the scattering signal
from the container, the correction for Compton scattering, absorption
effects, and multiple scattering, and normalization to obtain the
function S(Q). Finally, the pair distribution function, G(r), is obtained
by sine Fourier transformation of the normalized scattering intensity,
F(Q) = Q[S(Q)−1], to a maximum Q-value, Qmax, of 22.0 Å−1. In
PDFgetX3, the Fourier transformation is executed using exponential
damping so as to suppress termination ripples, at the cost of reduced
spatial resolution.
Figure 2. (a) Custom-built apparatus for extracting Li from Li15Si4
with liquid NH3. Li15Si4 powder is loaded onto the frit inside the
glovebox. When NH3 is condensed onto Li15Si4 (sample space is
cooled, −78 °C), a deep blue color is observed. (b) By closing valve 1,
the extract is transported into the cooled reservoir. (c) By opening
valve 1, and cooling the sample space, NH3 can be recondensed onto
the solid. (c) Thereafter, Li/LiNH2 remains in the reservoir. The color
is characteristic for a concentrated metallic bronze solution of Li in
NH3 denoted as Li(NH3)x.34
Na) was added slowly while stirring. After 5 min, the black suspension
(pH ∼8) was transferred into an Erlenmeyer flask followed by adding
50 mL ethanol and 1 mL acetic acid to adjust the pH value (pH ∼5).
After stirring shortly, the black solid was separated via centrifugation
and washed three times with 20 mL ethanol. Finally, the powder was
dried under vacuum while periodically heated to 150−200 °C with a
heat gun. This procedure yielded 30−40 mg of a black product,
thereafter termed a-Si.
Prior to subjecting Li3NaSi6 to extraction reactions, the material was
ground in an agate mortar or ball-milled in a hermetically sealed
stainless steel jar (RETSCH PM 100 planetary ball mill) for 8 h at 300
rpm (ball charge, 3; ball size, 15 mm diameter). A typical reaction with
ball-milled starting material was carried out as follows: 135 mg of
Li3NaSi6 were shortly (5−10 min) treated with 30 mL of ethanol
under inert conditions. Subsequently, 4 mL of concentrated HCl
solution were added. The solid product was separated by
centrifugation, repeatedly washed (3−4 times) with 30 mL of a 1:1
ethanol−water mixture (V/V), dried under vacuum, and finally
retained as a black fine powder (in the following termed a-allo-Si). The
Differential Thermal Analysis (DTA). Differential thermal
analysis was carried out with a NETZSCH DSC 404 Pegasus
apparatus equipped with a DTA carrier system. Custom-built niobium
crucibles were thoroughly cleaned, heated to 1000 °C under dynamic
vacuum (p < 1 × 10−3 mbar) for 2 h and transferred to an Ar-filled
glovebox. Crucibles were loaded with 20−30 mg of sample. The open
Table 1. Summary of Alkali Metal Extraction Reactions with Li3NaSi6 Targeting the Synthesis of allo-Si; Crystalline Products
That Were Identified by Powder X-ray Diffraction Are Specified
additional
solvent
(mL)
crystalline phases
identified by
PXRD
Li3NaSi6
(mg)
no.
1
reagent
experimental details
135
(a) EtOH, 30 mL (dried
over Na); (b) HCl,
4 mL
ball-milled Li3NaSi6 and EtOH were stirred in a Schlenk flask (5−10 min), 4 mL α-Si, Na8Si46
of HCl was added, washed with EtOH, dried under vacuum
2
3
600
EtOH (dried over Na),
15 mL/20 mL
stirred in a Schlenk flask for 14 days followed by decanting EtOH, stirred
Li3NaSi6, α-Si
b
another day with 20 mL of EtOH, washed with EtOH, dried under vacuum
a
100
i-PrOH, 0.5 mL
Et2O, 5
stirred in a Schlenk flask, 3 months, washed with Et2O (4×), dried under
Li3NaSi6, NbSi2
b
vacuum
a
b
4
400
i-PrOH, 1 mL
i-PrOH, 0.75 mL
air
Et2O, 9
stirred in a Schlenk flask, 6 weeks, washed with Et2O (3×), dried under vacuum α-Si
a
c
5
100
Et2O, 14
stirred in a vial, 2 weeks, washed with Et2O (3×)
6 months
Li3NaSi6
Li3NaSi6
a
c
6
300
a
c
7
50
H2O, 5 mL
stirred in a vial, 2 weeks, washed with H2O (2×)
stirred in a Schlenk flask, 4 days, washed with H2O (3×), dried under vacuum Li3NaSi6
a
b
8
100
HNO3 (68%), 6 mL
HOAc, 3 mL
a
c
9
50
stirred in a vial, 4 weeks, washed with H2O (3×)
α-Si
a
c
10
30
buffer solution
NaHPO4/KHPO4, 5 mL
stirred in a vial, 4 days, washed with H2O
Li3NaSi6
a
11a 1000
Ph2CO 5 M, 3 mL
THF, 10
stirred in a Schlenk flask, 3 months, washed with THF (4×), dried under
Li3NaSi6, NbSi2
b
vacuum
c
11b 217 from H2O, 5 mL
11a
stirred in a vial, 2 months, washed with H2O (3×)
SiO2, NbSi2,
Li2SiO3,
Na2SiO3
a
b
c
Li3NaSi6 was synthesized in a Nb ampule Reacted, washed, and dried under inert conditions. Sample handling carried out in air.
C
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
end was closed by crimping and then sealed by arc-welding inside the
glovebox under cooling. All measurements were performed under an
Ar-flow rate of 60−70 mL min−1 and a heating/cooling rate of 10 K
min−1. An empty niobium crucible served as reference. Samples were
recovered after the measurement by opening the niobium crucibles
inside the glovebox. The Proteus Thermal Analysis program was used
for data handling.39
Transmission Electron Microscopy (TEM). Transmission
electron microscopy and selected-area electron diffraction (SAED)
measurements were performed in a Tecnai F30 STwin microscope
(300 kV, field emission gun (FEG) cathode, spherical aberration
coefficient Cs = 1.2 mm). EDX nanoprobe analysis as part of the TEM
examinations were carried out using a Si/Li detector (EDAX System).
To prevent oxidation and contamination of the samples the entire
procedure for TEM preparation was done in an Ar-filled glovebox.
Ground samples of a-Si and a-allo-Si were placed onto a lacey-carbon
grid and subsequently transferred to the TEM in an argon-filled bag.
Heating experiments (cf. Figures S9 and S10 in the Supporting
Information) were performed with a single-tilt heating stage to
tentatively prove crystallization and morphology changes at elevated
temperatures.
and 16 were used for the evaluation of the Coulomb and exchange
integrals (TOLINTEG). Default optimization convergence thresholds
and an extra-large integration grid (XLGRID) for the density-
functional part were applied in all calculations. Another computational
study concerned the full structural relaxation of the Si substructure of
11
Li3NaSi6. Here the experimental structure of Li3NaSi6 was first fully
optimized and the obtained lattice and Si atomic parameters used as a
starting model for subsequent relaxation of an elemental Si structure.
This step was performed without any symmetry constraints at DFT-
PBE41/SVP level of theory. At this point the system was metallic and
the Fermi surface was smeared with a width of 0.02 au. For the
evaluation of the tolerance and exchange integrals (TOLINTEG),
tight criteria of 7, 7, 7, 7, and 14 were used and for the density-
functional part, a large integration grid (LGRID) was applied. The
geometry optimization starting from the Si substructure in Li3NaSi6
naturally resulted in large changes in the geometry. Therefore, the
optimization process had to be carried out carefully and was manually
restarted several times to enable the reclassification of the two-electron
integrals as the geometry became very different from the original
geometry. Between the steps, the symmetry had to be lowered to space
group P1 (1) and too big displacements from one point of the
potential surface to the next had to be avoided by the keyword
MAXTRADIUS. The resulting minimum structure was subsequently
optimized at the DFT-PBE0/SVP level of theory and its space group
determined with FINDSYM.45 Stationary points were confirmed to be
minima for all optimized structures by applying frequency
calculations.46,47
Attenuated Total Reflectance Fourier Transformed Infrared
Spectroscopy. ATR-FT-IR spectra were recorded on a Varian 670 IR
FT-IR spectrometer equipped with a Pike GladiATR ATR accessory
and using a diamond ATR element. All measurements were performed
at room temperature. The Varian Resolution Pro software was used for
data handling.40
Raman Spectroscopy. Raman spectra for a-Si and a-allo-Si were
measured using a LabRAM HR 800 spectrometer. The instrument is
equipped with an 800 mm focal length spectrograph and an air-cooled
(−70 °C), back thinned CCD detector. Samples were sealed in 0.7
mm glass capillaries. The capillaries were placed on a glass slide and
samples were excited with an air-cooled double frequency Nd:YAG
laser (532 nm). The laser was focused onto the sample using a 50X
superlong working distance (WD = 20.5 mm) Mitutoyo objective.
Measurements were performed in two steps. The samples were excited
at 0.56 mW laser power (corresponding to a low power density of 5.5
× 10−6 mW μm−2) and spectra were recorded at different spots of the
sample. Next the samples were heated with the laser for 1 min by
increasing the power to 14 mW and a spectrum was measured at this
laser power and another one at the low laser power used in the initial
examination (0.56 mW). The laser power was adjusted by a software
controlled filter wheel with 6 ND filters. Spectra were collected with an
exposure time of 60 s (10 accumulations) and using a grating of 600
lines mm−1. All spectra were calibrated and normalized. The spatial
resolution of the instrument is specified as 1 μm in diameter.
Elemental Analysis. A JEOL-JSM 7500F scanning electron
microscope (SEM) equipped with an Oxford X-Max EDX (energy-
dispersive X-ray spectroscopy) analyzer with Mn as internal standard
was used to determine elements with atomic numbers larger than four.
Samples were handled inside an Ar-filled glovebox or outside if not air-
sensitive and fixed on a graphite platelet which was attached to an
aluminum stub. The lithium content determination was carried out by
atomic absorption spectroscopy using a Varian AA280FS device. C, H,
N contents were determined by a Hekatech Euro EA CHNSO-
Analyzer.
Computational Details. Quantum chemical calculations with the
Perdew−Burke−Ernzerhof hybrid density functional method (DFT-
PBE0)41,42 were performed using the CRYSTAL09 software pack-
age.43 Split-valence + polarization (SVP) quality basis sets were
employed for Si, Na, and Li. The basis set for Si was taken from ref 44.
For the alkali metals they were derived from the molecular def-SVP
basis set (see the Supporting Information).
Both the lattice and the atomic parameters of Si trial structures
(according to Conesa)32 were allowed to relax within the constraints
imposed by the space group symmetry for structural optimization. The
shrinking factor (SHRINK) used for the generation of a Monkhorst−
Pack-type grid of k-points in reciprocal space was 4 for lattice sizes
larger 10 Å, 8 for lattices smaller or equal 10 Å, but larger than 5 Å and
16 for smaller lattice parameters. Tight tolerance factors of 8, 8, 8, 8,
3. RESULTS AND DISCUSSION
Chemical Extraction Experiments with Li15Si4 (Syn-
thesis of a-Si). Because the structure of the highly air- and
moisture-sensitive silicide Li15Si4 comprises isolated Si atoms
that are embedded in a Li matrix,33 nanoparticulate forms of Si
could be obtained by a soft extraction of Li. Hence, chemical
delithiation of Li15Si4 was realized using liquid ammonia. The
extraction apparatus is shown and described in Figure 2. The
blue color of liquid ammonia when condensed onto Li15Si4
powder clearly indicates the formation of solvated electrons
(Figure 2b). The formation of a metallic bronze solution,34 as
shown in Figure 2c, typically indicates that Li is extracted. After
several extraction cycles solely LiNH2 was identified as
crystalline phase in the PXRD pattern of the product (see
Figure S2 in the Supporting Information; residual Li content:
21.9(1) wt % Li vs 48.1 wt % in Li15Si4). The IR spectrum
essentially only shows NH stretching and HNH deformation
modes at 3259 and 1560 cm−1/1538 cm−1, respectively (see
Figure S3 in the Supporting Information).48 Si−O vibrational
bands that typically occur at around 800−1100 cm−1 and would
indicate O impurities are not observed.14 As the last step,
LiNH2 is removed by slowly reacting it with ethanol in a
toluene suspension. Acetic acid was used to adjust to acidity as
silicon dissolves under alkaline conditions. Alternatively, a
buffered solution can be used to remove LiNH2. The
delithiation of Li15Si4 is formulated as follows
NH3
Li15Si4 ⎯⎯⎯→ 15(1 − x)LiNH + 4Si + 15xLiNH2 + 7.5xH2
3
As can be seen from the PXRD pattern in Figure S2 in the
Supporting Information, the dried black product is amorphous
and will be termed a-Si in the following.
Chemical extraction experiments with Li3NaSi6 (syn-
thesis of a-allo-Si). As described in the introduction section,
the Zintl phase Li3NaSi6 consists of polyanionic layers 2∞[Si6]4−
which are separated by Li and Na ions.11 Previous attempts to
deintercalate Li3NaSi6 and combine the Si layers topotactically
employed protic solvents according to
D
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
Li3NaSi6 + 4ROH → 6 allo‐Si + 2H2 + 3LiOR + NaOR
partially oxidized samples of Li3NaSi6 at 300 °C for 2 weeks did
not give rise to any new reflections. A closer inspection of the
PXRD pattern of allo-Si reported in ref 11 and the database
entry PDF 41−1111 reveals strong analogies to NbSi2 (see
Figure S6 in the Supporting Information). Li3NaSi6 employed
in ref 11 was synthesized in Nb ampules at temperatures up to
800 °C, which readily explains the formation of NbSi2.
According to our experience, the presence of minor
concentrations of NbSi2 may not be recognizable in the
diffraction pattern of crystalline Li3NaSi6 but will uncover in the
PXRD pattern of the amorphous product after alkali metal
extraction (i.e., reactions 3 and 11a, Table 1).
With water, an almost complete conversion (97%) to allo-Si
was reported.10,11 Crystallites of α-Si were found concom-
itantly.10,11 Tentative unit-cell parameters for allo-Si were
reported as orthorhombic, a = 13.29 Å, b = 3.79 Å, c = 12.63
Å.10 However, the reflection positions calculated from those
parameters differ substantially from the diffraction pattern of
allo-Si compiled in the Powder Diffraction File (PDF 41−
1111).
We diligently tried to reproduce the reaction of Li3NaSi6 with
protic solvents to allo-Si. Generally, we found that the reactivity
of Li3NaSi6 strongly depends on the degree of crystallization
(consistent with ref 11). For instance, a thoroughly powdered
crystalline sample of Li3NaSi6 (see Figure S4a in the Supporting
Information) was allowed to react with ethanol for 15 days
(Table 1, reaction 2). Thereafter, reflections from Li3NaSi6 are
still recognizable in the PXRD pattern and the increased
background with maxima at 2θ ≈ 28° and ≈53° indicates the
presence of an amorphous phase (see Figure S4b in the
Supporting Information). In contrast, when using ball-milled
Li3NaSi6 as starting material (see Figure S5b in the Supporting
Information) reactions were completed within several minutes
(Table 1, reaction 1). The PXRD pattern of the product shows
an amorphous background as well as Bragg reflections of α-Si
and Na8Si46 (see Figure S5c in the Supporting Information).
This product is termed a-allo-Si in the following. The small
clathrate impurity may have already been introduced in the
synthesis of Li3NaSi6 although it is not recognizable in the
PXRD pattern of the starting material (overlap of Bragg
reflections). It may also be that ball-milling induces a partial
decomposition of Li3NaSi6 leading to the formation of NaSi,
which could be oxidized to form Na8Si46 when treated with
protic reagents.
Characterization of a-Si and a-allo-Si. Elemental
Analysis, FT-IR. IR spectroscopic investigations clearly reveal
Si−O vibrational modes for a-Si, in particular the Si−O
stretching band at around 1069 cm−1 (Figure 3), indicating a
Figure 3. ATR-FT-IR spectra of a-Si and a-allo-Si.
Extraction reactions employing mixtures of 2-propanol and
diethyl ether (Table 1, reactions 3−5) yielded amorphous
products beside α-Si and unreacted Li3NaSi6. In the case of
reaction 3 the Li3NaSi6 starting material was synthesized in Nb
ampules and contained a marginal NbSi2 impurity (confirmed
by EDX analysis). Crystalline NbSi2 could not be observed in
the PXRD pattern of the starting material but is clearly
identified in the PXRD pattern of the oxidation product.
In contrast to alcohols, H2O and acids rapidly oxidize
Li3NaSi6 under formation of an amorphous phase. In some
cases α-Si and unreacted Li3NaSi6 are identified as crystalline
phases (Table 1, reactions 7-9). Because all reactions
employing protic solvents as oxidative reagents are accom-
panied by the formation of LiOH and NaOH, the pH value is
critical regarding the dissolution of Si. In reaction 10 a
NaHPO4/KHPO4 buffer solution was employed to compensate
for the pH increase. We also tried to oxidize Li3NaSi6 using a
solution of benzophenone in THF. After three months the
PXRD pattern of the product showed an increased background
and Bragg reflections of NbSi2 and residual Li3NaSi6 (Table 1,
reaction 11a). Further oxidation of residual Li3NaSi6 with H2O
yielded Li2SiO3, Na2SiO3, and SiO2 (11b).
slight partial surface oxidation.14 This is additionally strength-
ened by SEM EDX examinations (see Figure S7 in the
Supporting Information) and nanoprobe EDX analysis in the
TEM (described below). The oxidation occurs during the
removal of LiNH2 with ethanol. According to SEM EDX, the
average O content of a-Si samples is 12.8 1.9 at. %. A slightly
visible band at around 2100 cm−1 may indicate the presence of
Si−H. However, this band is significantly weaker than observed
for hydrogen-capped silicon nanoparticles.49 Furthermore,
traces of Na (<1 at. %) were detected by SEM EDX. The
residual Li content in a-Si (∼1 wt %) was determined by
atomic absorption spectroscopy. C, H, N impurities of a-Si
samples are marginal (C, H, N < 1.2 wt %). In contrast to a-Si,
the IR spectrum of a-allo-Si is essentially featureless. Si−O
bands are barely visible. However, according to SEM EDX
analysis a-allo-Si samples contain 7.4 1.9 at. % O.
TEM Analysis. Representative results of the TEM inves-
tigations are shown in Figures 4 and 5. The morphology of a-Si
and a-allo-Si samples differs remarkably. Samples a-Si show a
highly porous microstructure built of agglomerated particles
with sizes around 10 nm (Figure 4a). The electron diffraction
pattern (Figure 4b) corresponds to faint rings of diffuse
intensity which indicates marginal sizes of coherently scattering
domains and thus short-range ordering. The diameters of these
rings correlate with the d-values of α-Si. HRTEM micrographs
(Figure 4c) show the presence of α-Si nanocrystallites with
sizes below 5 nm within the thin and electron transparent areas.
The oxygen content as determined by nanoprobe EDX was
In summary, deintercalation of Li3NaSi6 with protic solvents
leads inevitably to the formation of amorphous products. In
most cases, significant amounts of Li3NaSi6 remained
unreacted. Complete reactions were only achieved with ball-
milled starting material. Regardless of reaction conditions for
the deintercalation of Li3NaSi6, we never observed the
reflections reported for allo-Si (PDF 41−1111). Also, annealing
E
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
Figure 4. (a) Bright-field micrograph of a-Si particles, (b) electron
diffraction pattern from a-Si sample containing diffuse concentric rings
matching with nanocrystalline α-Si, (c) high-resolution micrograph
showing crystalline nanoparticle (circled) in amorphous matrix and
respective FFT-image showing sharp 111 reflections.
Figure 5. (a) Bright-field micrograph of a-allo-Si, (b) electron
diffraction pattern from the circled area showing microcrystalline
nature of the sample, (c) high-resolution micrograph showing
crystalline nanoparticles (circled) in amorphous matrix.
Figure 6. (a) High-energy X-ray diffraction patterns and (b) PDFs for
the a-Si and a-allo-Si materials (λ = 0.20727(6) Å, Qmax = 21.0 Å−1).
about 8 at. % on average and up to 40 at. % in thin areas close
to the boundary of particles. This agrees roughly with the result
of the SEM EDX analysis. The oxygen containing areas do not
show any structural ordering, thus, they are not indicated by
any prominent intensity within the electron diffraction patterns.
Samples a-allo-Si consist of larger particles with minimum
sizes of 50 nm (see Figure 5a). These particles contain
crystallites of α-Si with sizes large enough to yield coherent
scattering (cf. the fully indexed diffraction pattern in Figure 5b).
Consequently, the width of the diffraction circles is
considerably lower compared to a-Si (cf. Figure 4b). Moreover,
the diffuse intensities clearly segregate into Bragg intensities as
expected for rotationally disordered microcrystals of α-Si.
Indeed, HRTEM confirms the presence of α-Si crystallites with
sizes up to 10 nm in a-allo-Si (Figure 5c). For a-allo-Si the
average oxygen content as established by nanoprobe EDX was
comparable to a-Si (ca. 10 at. %), but the maximum value
determined on the edges was only 15 at. %.
PXRD Investigations. The high energy diffraction patterns
and PDFs for a-Si and a-allo-Si are compiled in Figure 6. In
agreement with the TEM investigations the diffraction pattern
of a-Si is void of sharp Bragg peaks. Broad maxima are located
around 2θ = 4, 7, and 10.5° corresponding to d = 3.28, 1.75,
and 1.19 Å, respectively.
The amorphous contribution to the pattern of a-allo-Si is
modulated by the Bragg peaks of crystalline Si, which again
reflects the finding from the TEM analysis. The minimum
crystallite size yielding sharp Bragg peaks in X-ray diffraction is
4−5 nm.50 The diffraction pattern of a-allo-Si contains also
Bragg peaks from the Na8Si46 impurity. The positions of the
peaks in the PDF for a-allo-Si closely correspond to the
separation of atom pairs in crystalline Si. The PDF for a-Si
resembles amorphous Si with a sharp first neighbor peak at 2.35
Å and a broad second neighbor peak at 3.8 Å.51,52 In contrast
with amorphous Si this peak is clearly asymmetric. This
asymmetry is robust with respect to variations of Qmax and
indicates that the second peak is actually split. An analogous
phenomenon has been recently observed and analyzed for
amorphous Ge and attributed to a fraction of third neighbors
being closer than some second ones, indicating structural
relaxation (i.e., deviation from a continuous random net-
work).53 This interpretation is not applicable for our a-Si
material which−according to the TEM analysis−represents a
mixture of nanocrystalline Si embedded in an amorphous
environment. Rather, the PDF for a-Si appears as a convolution
of amorphous Si and sub-Bragg sized crystallites (2−4 nm).
The contribution from the latter dominates the medium range
of r (4−8 Å). Indeed the third peak centered at 5.7 Å for
amorphous Si51,52 is clearly split according to crystalline Si
(5.47 and 5.86 Å), whereas the faint fourth peak (centered at
around 7 Å for amorphous Si)51,52 may indicate fine structure,
mirroring distances from the center of crystallites to their
(relaxed) interface with amorphous environment. It is tempting
to interpret a-Si as paracrystalline Si where the crystallite size is
large enough to express with distinct pairs in the PDF.54
However, because of the partial oxidation of a-Si such
conclusions remain highly ambiguous. The peaks in the PDF
for a-allo-Si essentially correspond to crystalline Si, which is
expected from the presence of crystallites larger than 5 nm.
Raman Spectroscopy. Figure 7 shows Raman spectra for a-
Si and a-allo Si, respectively. Focusing first on a-Si (Figure 7a),
the majority of observed spectra correspond to amorphous Si
characterized by a broad maximum at 460−470 cm−1 (the TO1
band), weaker bands at 380 and 300 cm−1 assigned to LO and
LA, respectively, and a pronounced TA like band at low
wavenumbers. Occasionally spectra characteristic of nano-
crystalline (nc) Si (2−4 nm crystallite size) could also be
obtained. Those display a sharp and asymmetric TO2 peak at
500 cm−1.55 It is important to note that spectra of amorphous
and nc-Si can only be observed when applying the lowest
F
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
Figure 8. First DTA heating traces of thermograms recorded for
Li3NaSi6, partially oxidized Li3NaSi6 (Li3NaSi6/a-allo-Si mixture), a-
allo-Si, and a-Si (Tmax = 750 °C, 10 K min−1 heating/cooling rate,
heat effects marked with * could not be assigned).
Figure 7. Raman spectra for (a) a-Si and (b) a-allo-Si. The spectra in
the upper panel in a) were obtained from different parts of the a-Si
sample after increasing the laser power.
after DTA investigations (see Figures S4c and S8 in the
Supporting Information), the latter clearly reveals larger
portions of α-Si. The heating trace of the a-allo-Si sample
only contains an irreversible exothermic effect that is markedly
shifted to higher temperatures (effect 3 at 565 °C). The
subsequently recorded PXRD pattern corresponds to α-Si (see
Figure S5d in the Supporting Information). Thus, effect 1 and 3
are attributed to the amorphous-to-crystalline transition of a-
allo-Si. The temperatures for the Li3NaSi6/a-allo-Si mixture and
the a-allo-Si sample embrace the reported phase transition of
allo-Si at ∼527 °C.11
The variation in temperature is ascribed to different
concentrations of sub-Bragg sized nanocrystals (2−4 nm),
that is the presence of different concentrations of crystallization
nuclei in both samples. We assume that in a-Si the
concentration of sub-Bragg sized crystallites is considerably
smaller than in a-allo-Si. The phase transition of a-Si occurs at
658 °C (effect 4) which is similar to amorphous silicon
obtained from ball milling at a heating rate of 40 K min−1, but
lower than the crystallization temperature for ion-implanted
amorphous silicon (around 690 °C).58,59 Thus, for a-Si,
nucleation (i.e., the formation of a critical concentration of
critically sized nuclei) has to be induced prior to the transition,
which explains the shift to higher temperatures. It is interesting
to note that when heating individual particles of both a-Si and
a-allo-Si in the TEM, crystallization is not observed below 700
°C (cf. Supporting Information, Figures S9 and S10).
possible power density (0.56 W/5.5 × 10−6 mW μm−2).
Amorphous spectra did not change with time whereas nc-Si
spectra showed well-known variations attributed to temperature
increase.56,57 When increasing laser power, amorphous and nc-
Si spectra convert into a spectrum characteristic for (bulk)
crystalline Si. The sharp and symmetric peak of the TO2 band
shifts to 519−521 cm−1 and only varies slightly with
temperature. The two phonon contributions at lower wave-
numbers are absent.
Raman spectra for the a-allo-Si sample (Figure 7b) mirror
the three constituents of this sample. First there are spectra
reflecting a solely amorphous state. However, these spectra
were frequently convoluted with contribution from nc-Si. Note
that the shape of such spectra has been considered to be typical
of paracrystalline Si.50 Vice versa, spectra of nc-Si with a
pronounced asymmetry stemming from an amorphous silicon
contribution were recorded. Finally, we also obtained spectra
corresponding to bulk crystalline Si. Those spectra are
attributed to Si crystallites with a size larger than 5 nm.
Accordingly, a-allo-Si contains α-Si crystallites with a large size
variation (2 to ∼10 nm), whereas α-Si crystallites in a-Si have
sub-Bragg size (2−4 nm).
DTA Investigations. The amorphous-to-crystalline phase
transition of a-Si and a-allo-Si was investigated by differential
thermal analysis. Figure 8 shows the first heating traces for (i)
Li3NaSi6, (ii) partially deintercalated Li3NaSi6 (i.e., a mixture of
Li3NaSi6 and a-allo-Si), (iii) a-allo-Si, and (iv) a-Si. Li3NaSi6
decomposes at around 600 °C (effect 2) into Li12Si7, NaSi, and
α-Si (see Figure S8 in the Supporting Information). For the
Li3NaSi6/a-allo-Si mixture, an additional exothermic effect is
observed at 487 °C (effect 1). Comparing the PXRD patterns
of the pure Li3NaSi6 sample and the Li3NaSi6/a-allo-Si mixture
Allotropism of Si (Or How Feasible is allo-Si?). As
described above, the existence of a crystalline allotropic form of
silicon obtained from Li3NaSi6 through oxidation is highly
questionable. This is in contrast to m-allo-Ge, which is readily
prepared from delithiating Li7Ge12. Both Li7Ge12 and Li3NaSi6
possess two-dimensional polyanions with five-membered rings,
yet their topotactic combination leads to different branches of
three-dimensional structures. According to Conesa, the
G
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
combination of layers in Li7Ge12 results in “G” structures,
whereas “S” structures are derived from Li3NaSi6.32
types four structure variations exist. The fourth group SD is
built by connecting red Si2− atoms (2b) to both Si− atoms (3b)
of the next layer. For such a network, only one structure type is
possible, the TON zeolite-type network (ZSM-22).
2
The two-dimensional slabs ∞[Si6]4− in Li3NaSi6 are built of
four (4b), three (3b), and two (2b) bonded Si atoms,
respectively (cf. Figure 9). Based on this structure, Conesa’s
A comparison of total energies of the computationally
relaxed Conesa S-type Si structures (Figure 10) shows that SD
is the favored structure with an energy difference to α-Si of 13.7
kJ (mol Si)−1 (Si−Si distances, 2.34−2.40 Å; angles, 90.7−
142.9°; crystallographic density, 2.011 g cm−3, Figure 11a). Its
Figure 11. (a) Most stable network (SD) derived from the Si
substructure of Li3NaSi6 with view along the c-direction emphasizing
the one-dimensional channel structure built from 10-, 6-, and 5-
membered rings; (b) network of the structure SC1 with view along the
a-direction. It is built of 6- and 5-membered rings.
Figure 9. Si substructure in the phase Li3NaSi6. The four (Si0), three
(Si−), and two (Si2−) connected Si atoms are shown in black, light/
dark blue, and red, respectively. For the four different groups of allo-Si
networks derived from this layered structure, the charged atoms are
connected differently.
structure has the lowest density of all G and S group networks
and is equivalent to the zeolite network TON. As shown in
Figure 11a it contains 10-, 6-, and 5-membered rings which
form parallel channels proceeding along the c-axis. The
diameter of the largest channel ranges from 6.6 to 7.4 Å. In
energy, this framework is followed by SC1 (Figure 11b) with a
relative energy of 14.0 kJ (mol Si)−1. Si−Si distances vary from
2.31 to 2.43 Å and angles range from 92.0 to 128.4°. The
crystallographic density of this structure is 2.292 g cm−3 which
is close to that of α-Si (2.290 g cm−3). SC1 also displays one-
dimensional channels, which run along the a direction and are
based on 6- and 5-membered rings (Figure 11b).
To simulate computationally the topotactic recombination of
Si layers in Li3NaSi6 (cf. Figure 9) to allo-Si, we carried out a
full relaxation of the Si substructure in Li3NaSi6 without any
symmetry restrictions. This resulted in a true local minimum
with Cmcm space group symmetry being the TON (SD)
S group follows, which can be divided into four smaller
structure groups. The SA group is built by connecting 2b-Si
atoms (red) of neighboring layers and bonding them to 3b-Si
atoms of their own layers (dark blue). The light blue 3b-Si
atoms are bonded to the light blue ones of the neighboring
layer. SB is a connection of red Si2− atoms (2b) with light blue
Si− atoms (3b) of the neighboring layer and dark blue Si−
atoms (3b) of their own layer. For structures SC, the 2b-Si2−
atoms are connected to dark blue 3b-Si− atoms and to the red
Si2− of the neighboring layer. The 3b-Si− atoms (light blue) are
connected to the same kind of the next layer. For each of these
Figure 10. Energies relative to α-Si for Conesa’s S-type structures (derived from the Si substructure of Li3NaSi6). The four different network classes
are colored differently; SA in red, SB in green, SC in orange, and SD in blue. The energy difference to well-known open tetrahedral frameworks60 is
depicted as well (violet). All relative energies are also listed in Table S1 in the Supporting Information.
H
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
structure, which was also the favored of all “handmade”
structures of Conesa, highlighting the key role of this network.
Furthermore, the structure with the smallest density is favored
in the present structural relaxation without any initial
specification of Si−Si bonds.
small uncapped Si nanoparticles of 5−20 nm can be obtained
without aggregation using a reverse micelles technique also
showing a SiO2 coverage. The decreased particle size of Si
enhances the cycling stability of Si anodes. Using conductive
additives, reversible capacities above 2000 mAh g−1 were
achieved with particles in the range of 5 to 20 nm. Smaller
particles, however, increase the irreversible capacity losses.66
Importantly, the most favorable structures that follow from
2
the combination of two-dimensional slabs ∞[Si6]4− are rather
high in energy, around 14 kJ (mol Si)−1 with respect to the
ground state structure of Si (α-Si). Several clathratic and
zeolitic open tetrahedral frameworks have a considerably lower
energy and are only by about 7 kJ (mol Si)−1 less stable than α-
Si. Such frameworks are also included in Figure 10. As already
shown earlier the clathrate-II framework mtn and the chiral unj
framework appear most favorable as Si allotropes.61,62 The
crystallographic densities of those frameworks are lower
compared to the “S” structures and range from 1.979 g cm−3
(mtn) to 2.180 g cm−3 (tum) (see Table S1 in the Supporting
Information). In contrast the total energies of Conesa “G”
structures (derived from Li7Ge12 layers) compare closely with
the clathratic and zeolitic frameworks (cf. Figure S11, Table S1
in the Supporting Information) and are favored over the “S”
structures derived from Li3NaSi6. This strongly supports the
experimental observation that Li3NaSi6 is not a suitable
precursor for topotactic formation of allo-Si, whereas allo-Ge
is readily obtained from Li7Ge12.
ASSOCIATED CONTENT
■
S
* Supporting Information
PXRD patterns of Li3NaSi6 samples from different reactions
(Figure S1), PXRD patterns of (a) Li15Si4 bulk material, the
sequential products after (b) Li-extraction with liquid ammonia,
(c) subsequent treatment with ethanol, and (d) differential
thermal analysis (Figure S2), ATR-FT-IR spectra of the a-Si/
LiNH2 mixture obtained after Li extraction from Li15Si4 (Figure
S3), PXRD patterns of Li3NaSi6, the product after partial
oxidation with ethanol and subsequent DTA-investigation
(Figure S4), PXRD patterns of Li3NaSi6 before and after ball-
milling, the product after treatment with ethanol/HCl and
subsequent DTA investigation (Figure S5), PXRD pattern of a
NbSi2 sample referenced with the database entry for allo-Si and
NbSi2 (Figure S6), results from EDX investigations of a-Si
(Figure S7), Rietveld fit to the PXRD pattern of Li3NaSi6 after
DTA (Figure S8), results from TEM heating experiments of a-
Si and a-allo-Si (Figures S9 and S10), comparison of the total
energies of networks optimized on the basis of Conesa’s G-type
structures (Figure S11), DFT-PBE0 results in tabular format
(Table S1), and basis sets for Na and Li employed in the
theoretical calculations. This material is available free of charge
via the Internet at http://pubs.acs.org.
CONCLUSION
■
Alkali metal extraction reactions with the silicides Li15Si4 and
Li3NaSi6 lead to amorphous products termed a-Si and a-allo-Si,
respectively. a-Si possesses a porous microstructure which is
built from spherically shaped particles with a diameter of
around 10 nm. The amorphous particles contain sub-Bragg
sized (below 5 nm) Si crystallites which are revealed in TEM
investigations and Raman spectroscopy. They have a partially
oxidized surface, which is introduced in the purification process
of a-Si. Extraction reactions with Li15Si4 chemically emulate the
delithiation process occurring in silicon-based lithium batteries.
The porous microstructure of a-Si might bear close
resemblance with electrochemically obtained amorphous Si.
a-allo-Si does not display a distinguished microstructure but
consists of irregularly shaped particles with sizes larger than 50
nm. These particles contain sub-Bragg-sized and Bragg-sized
crystallites with sizes up to 10 nm. The considerably lower
crystallization temperature of a-allo-Si (around 570 °C)
compared to a-Si (around 660 °C) suggests that the
concentration of sub-Bragg-sized particles (acting as crystal-
lization nuclei) is significantly larger in the former material. The
formation of a-allo-Si from Li3NaSi6 contradicts earlier findings,
reporting the formation of a crystalline bulk allotrope of silicon
(allo-Si) from the topotactic combination of silicon layers
present as polyanions in Li3NaSi6. Quantum chemical
calculations show that the combination of Li3NaSi6 silicon
layers does not lead to energetically favorable allotropic forms
of Si. This is different to Li7Ge12, where polyanionic germanium
layers topotactically convert to the germanium allotrope m-allo-
Ge upon oxidation.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: Thomas.Faessler@lrz.tum.de.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work has been funded by the Deutsche Forschungsge-
meinschaft project number FA 198/11−1, Academy of Finland
(Grant 138560/2010), as well as the National Science
Foundation through Grant DMR-1007557. M.Z. and L.M.S.
thank the Fonds der Chemischen Industrie for their fellow-
ships. L.A.J. is grateful for her Christiane Nusslein-Volhard
̈
Stiftung fellowship. M.Z. thanks F. Kraus for providing the
custom-built extractor. This work was also supported by the
Swedish Research Council through grants 2010-4827 and 2011-
6512. The authors acknowledge PETRA III, DESY (Hamburg),
for provision of synchrotron radiation beamtime and thank Drs
Hanns-Peter Liermann and Jozef Bednarcik for assistance in
using beamline P02.1.
REFERENCES
■
(1) Okamoto, H.; Sugiyama, Y.; Nakano, H. Chem.Eur. J. 2011, 17,
The alkali metal extraction reactions reported herein expand
the methods for the fabrication of small nanoparticles. Further
examples include the first commonly recognized successful
production of Si nanoclusters capped with alkyl groups63 via
SiCl4 + RSiCl3 + Na → Si + NaCl and the magnesiothermic
reduction of silica64 to form porous nanostructured Si, which
was investigated as potential anode material.65 In addition,
9864.
(2) Miles, R. W.; Zoppi, G.; Forbes, I. Mater. Today 2007, 10, 20.
(3) Sun, D.; Riley, A. E.; Cadby, A. J.; Richman, E. K.; Korlann, S. D.;
Tolbert, S. H. Nature 2006, 441, 1126.
(4) Armatas, G. S.; Kanatzidis, M. G. Science 2006, 313, 817.
(5) Armatas, G. S.; Kanatzidis, M. G. Nature 2006, 441, 1122.
(6) Armatas, G. S.; Kanatzidis, M. G. Adv. Mater. 2008, 20, 546.
I
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX

Chemistry of Materials
Article
(7) Armatas, G. S.; Kanatzidis, M. G. J. Am. Chem. Soc. 2008, 130,
(42) Adamo, C.; Barone, V. J. Chem. Phys. 1999, 110, 6158.
(43) Dovesi, R.; Orlando, R.; Civalleri, B.; Roetti, C.; Saunders, V. R.;
11430.
Zicovich-Wilson, C. M. Z. Kristallogr. 2005, 220, 571.
(8) Gruttner, A. Dissertation, Universitat Stuttgart, Stuttgart,
Germany, 1982.
̈
̈
(44) Karttunen, A. J.; Fassler, T. F.; Linnolahti, M.; Pakkanen, T. A.
̈
Inorg. Chem. 2011, 50, 1733.
(9) Gruttner, A.; Nesper, R.; von Schnering, H. G. Angew. Chem., Int.
̈
(45) Stokes, H. T.; Hatch, D. M. J. Appl. Crystallogr. 2005, 38, 237.
(46) Zicovich-Wilson, C. M.; Pascale, F.; Roetti, C.; Saunders, V. R.;
Orlando, R.; Dovesi, R. J. Comput. Chem. 2004, 25, 1873.
(47) Pascale, F.; Zicovich-Wilson, C. M.; Gejo, F. L.; Civalleri, B.;
Orlando, R.; Dovesi, R. J. Comput. Chem. 2004, 25, 888.
(48) Bohger, J. P. O.; Essmann, R. R.; Jacobs, H. J. Mol. Struct. 1995,
348, 325.
(49) Neiner, D.; Kauzlarich, S. M. Chem. Mater. 2010, 22, 487.
(50) Voyles, P. M.; Zotov, N.; Nakhmanson, S. M.; Drabold, D. A.;
Gibson, J. M.; Treacy, M. M. J.; Keblinski, P. J. Appl. Phys. 2001, 90,
4437.
Ed. 1982, 21, 912.
(10) Schwarz, M. Dissertation, Universitat Stuttgart, Stuttgart,
̈
Germany, 1988.
(11) von Schnering, H. G.; Schwarz, M.; Nesper, R. J. Less-Common
Met. 1988, 137, 297.
(12) Kiefer, F.; Hlukhyy, V.; Karttunen, A. J.; Fassler, T. F.; Gold, C.;
̈
Scheidt, E. W.; Scherer, W.; Nylen, J.; Haussermann, U. J. Mater.
Chem. 2010, 20, 1780.
̈
(13) Kiefer, F.; Karttunen, A. J.; Doblinger, M.; Fassler, T. F. Chem.
̈
̈
Mater. 2011, 23, 4578.
(14) Wang, J.; Ganguly, S.; Sen, S.; Browning, N. D.; Kauzlarich, S.
(51) Laaziri, K.; Kycia, S.; Roorda, S.; Chicoine, M.; Robertson, J. L.;
Wang, J.; Moss, S. C. Phys. Rev. B 1999, 60, 13520.
(52) Laaziri, K.; Kycia, S.; Roorda, S.; Chicoine, H.; Robertson, J. L.;
Wang, J.; Moss, S. C. Phys. Rev. Lett. 1999, 82, 3460.
(53) Roorda, S.; Martin, C.; Droui, M.; Chicoine, M.; Kazimirov, A.;
Kycia, S. Phys. Rev. Lett. 2012, 108, 255501.
(54) The paracrystalline model is a revision of the continuous
random network model and is defined as highly strained, micro-
crystalline grains surrounded by fully amorphous material. In contrast
to a microcrystalline phase, the paracrystalline model lacks defined
grain boundaries. See ref 50.
(55) Duan, Y.; Kong, J. F.; Shen, W. Z. J. Raman Spectrosc. 2012, 43,
756.
(56) Mishra, P.; Jain, K. P. Phys. Rev. B 2000, 62, 14790.
(57) Nikolenko, A. S. Semicond. Phys. Quantum Electron. Optoelectron.
2013, 16, 86.
M. Polyhedron 2013, 58, 156.
(15) Meillaud, F.; Shah, A.; Droz, C.; Vallat-Sauvain, E.; Miazza, C.
Sol. Energy Mater. Sol. Cells 2006, 90, 2952.
(16) Iwamura, S.; Nishihara, H.; Kyotani, T. J. Power Sources 2013,
222, 400.
(17) Szczech, J. R.; Jin, S. Energy Environ. Sci. 2011, 4, 56.
(18) Kim, H.; Seo, M.; Park, M.-H.; Cho, J. Angew. Chem., Int. Ed.
2010, 49, 2146.
(19) Obrovac, M. N.; Christensen, L. Electrochem. Solid-State Lett.
2004, 7, A93.
(20) Hatchard, T. D.; Dahn, J. R. J. Electrochem. Soc. 2004, 151, A838.
(21) Key, B.; Bhattacharyya, R.; Morcrette, M.; Seznec, V.; Tarascon,
J. M.; Grey, C. P. J. Am. Chem. Soc. 2009, 131, 9239.
(22) Key, B.; Morcrette, M.; Tarascon, J. M.; Grey, C. P. J. Am. Chem.
Soc. 2011, 133, 503.
(23) Ammar, A.; Cros, C.; Pouchard, M.; Jaussaud, N.; Bassat, J. M.;
Villeneuve, G.; Duttine, M.; Menetrier, M.; Reny, E. Solid State Sci.
2004, 6, 393.
(58) Shen, T. D.; Koch, C. C.; McCormick, T. L.; Nemanich, R. J.;
Huang, J. Y.; Huang, J. G. J. Mater. Res. 1995, 10, 139.
(59) Poate, J. M. Nucl. Instrum. Methods 1983, 209/210, 211.
(24) Guloy, A. M.; Ramlau, R.; Tang, Z.; Schnelle, W.; Baitinger, M.;
Grin, Y. Nature 2006, 443, 320.
(60) Zeilinger, M.; van Wullen, L.; Benson, D.; Kranak, V. F.; Konar,
̈
S.; Fassler, T. F.; Haussermann, U. Angew. Chem., Int. Ed. 2013, 52,
5978.
̈
̈
(25) Kautsky, H.; Haase, L. Z. Naturforsch. 1953, B8, 45.
(26) Kautsky, H.; Haase, L. Chem. Ber. 1953, 86, 1226.
(27) Tchalala, M. R.; Ali, M. A.; Enriquez, H.; Kara, A.; Lachgar, A.;
Yagoubi, S.; Foy, E.; Vega, E.; Bendounan, A.; Silly, M. G.; Sirotti, F.;
Nitshe, S.; Chaudanson, D.; Jamgotchian, H.; Aufray, B.; Mayne, A. J.;
Dujardin, G.; Oughaddou, H. J. Phys.: Condens. Matter 2013, 25,
No. 442001.
(28) McMillan, P. F.; Gryko, J.; Bull, C.; Arledge, R.; Kenyon, A. J.;
Cressey, B. A. J. Solid State Chem. 2005, 178, 937.
(29) Annou, K.; Pelosi, M.; Gershinsky, G.; Favier, F.; Cuminal, Y.;
Tillard, M.; Zitoun, D. Mater. Renew. Sustain. Energy 2014, 3, 32.
(61) Pickard, C. J.; Needs, R. J. Phys. Rev. B 2010, 81.
(62) Zwijnenburg, M. A.; Jelfs, K. E.; Bromley, S. T. Phys. Chem.
Chem. Phys. 2010, 12, 8505.
(63) Heath, J. R. Science 1992, 258, 1131.
(64) Bao, Z.; Weatherspoon, M. R.; Shian, S.; Cai, Y.; Graham, P. D.;
Allan, S. M.; Ahmad, G.; Dickerson, M. B.; Church, B. C.; Kang, Z.;
Abernathy, H. W., III; Summers, C. J.; Liu, M.; Sandhage, K. H. Nature
2007, 446, 172.
(65) Yu, Y.; Gu, L.; Zhu, C.; Tsukimoto, S.; van Aken, P. A.; Maier, J.
Adv. Mater. 2010, 22, 2247.
(30) Kiefer, F.; Fassler, T. F. Solid State Sci. 2011, 13, 636.
̈
(66) Kim, H.; Seo, M.; Park, M.-H.; Cho, J. Angew. Chem., Int. Ed.
2010, 49, 2146.
(31) Zaikina, J. V.; Muthuswamy, E.; Lilova, K. I.; Gibbs, Z. M.;
Zeilinger, M.; Snyder, G. J.; Fassler, T. F.; Navrotsky, A.; Kauzlarich, S.
M. Chem. Mater. 2014, 26, 3263.
̈
(32) Conesa, J. C. J. Phys. Chem. B 2002, 106, 3402.
(33) Zeilinger, M.; Baran, V.; van Wullen, L.; Haussermann, U.;
̈
̈
Fassler, T. F. Chem. Mater. 2013, 25, 4113.
̈
(34) Zurek, E.; Edwards, P. P.; Hoffmann, R. Angew. Chem., Int. Ed.
2009, 48, 8198.
(35) Kiefer, F. Dissertation, Technische Universitat Munchen,
̈
̈
Munich, Germany, 2010.
(36) TOPAS−Rietveld Software, version 4.0; Bruker AXS: Madison,
WI, 2009.
(37) Hammersley, A. P.; Svensson, S. O.; Hanfland, M.; Fitch, A. N.;
Hausermann, D. J. High Pressure Res. 1996, 14, 235.
̈
́
(38) Juhas, P.; Davis, T.; Farrow, C. L.; Billinge, S. J. L. J. Appl.
Crystallogr. 2013, 46, 560.
(39) Netzsch Proteus Thermal Analysis, version 4.8.2; Netzsch-
Geratebau: Selb, Germany, 2006.
̈
(40) Resolution Pro, version 5.1.0.822; Varian: Santa Clara, CA, 2009.
(41) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,
3865.
J
dx.doi.org/10.1021/cm503371e | Chem. Mater. XXXX, XXX, XXX−XXX