Amorphous and nanocrystalline luminescent Si and Ge obtained via a solid-state chemical metathesis synthesis route

Article abstract of DOI:10.1016/j.jssc.2004.12.040

A new solid-state metathesis synthesis route was applied to obtain bulk samples of amorphous or microcrystalline Si and Ge. The method involves reaction of Zintl phases such as NaSi or NaGe, with ammonium or metal (e.g., CuCl, CoBr₂) halides. The driving force for the solid-state reaction is provided by the formation of alkali halides and the transition metals or metal silicides, or gaseous ammonia and hydrogen. The semiconductors were purified by washing to remove other solid products. The amorphous semiconductors were obtained in bulk form from reactions carried out at 200-300°C. Syntheses at higher temperatures gave rise to microcrystalline semiconductors, or to micro-/nanocrystalline particles contained within the amorphous material. Similar crystalline/amorphous composites were obtained after heat treatment of bulk amorphous materials.

Full text of DOI:10.1016/j.jssc.2004.12.040

ARTICLE IN PRESS
Journal of Solid State Chemistry 178 (2005) 937–949
www.elsevier.com/locate/jssc
Amorphous and nanocrystalline luminescent Si and Ge obtained via a
solid-state chemical metathesis synthesis route
Ã
Paul F. McMillana,b, , Jan Grykoc, Craig Bulld, Richard Arledgec, Anthony J. Kenyone,
Barbara A. Cresseyf
aChristopher Ingold Laboratory, Materials Chemistry Centre, Department of Chemistry, University College London, 20 Gordon Street,
London WC1H 0AJ, UK
bDavy Faraday Research Laboratory, Royal Institution of Great Britain, London W1S 4BS, UK
cDepartment of Physical and Earth Sciences, Jacksonville State University, AL 36265, USA
dSchool of Physics and Centre for Science at Extreme Conditions, University of Edinburgh, Mayfield Road, Edinburgh EH9 3JZ, UK
eDepartment of Electronic and Electrical Engineering, University College London, Torrington Place, London WC1E 7JE, UK
fScience and Engineering Electron Microscopy Centre, School of Chemistry, University of Southampton, Highfield, Southampton SO17 1BJ, UK
Received 6 October 2004; received in revised form 27 December 2004; accepted 28 December 2004
Abstract
A new solid-state metathesis synthesis route was applied to obtain bulk samples of amorphous or microcrystalline Si and Ge. The
method involves reaction of Zintl phases such as NaSi or NaGe, with ammonium or metal (e.g., CuCl, CoBr2) halides. The driving
force for the solid-state reaction is provided by the formation of alkali halides and the transition metals or metal silicides, or gaseous
ammonia and hydrogen. The semiconductors were purified by washing to remove other solid products. The amorphous
semiconductors were obtained in bulk form from reactions carried out at 200–300 1C. Syntheses at higher temperatures gave rise to
microcrystalline semiconductors, or to micro-/nanocrystalline particles contained within the amorphous material. Similar
crystalline/amorphous composites were obtained after heat treatment of bulk amorphous materials.
r 2005 Elsevier Inc. All rights reserved.
1. Introduction
or microcrystalline silicon and germanium materials,
that can also yield luminescent nanoparticles embedded
within an amorphous solid matrix.
Silicon and germanium are important semiconducting
materials that are used in bulk or thin film form as both
crystalline and amorphous solids. The amorphous forms
have been developed particularly for use in photovoltaic
devices [1]. Nanocrystalline particles of the group IV
elements, either free-standing or incorporated in solid
matrices, are now under development because of their
useful photo- and electro-luminescent properties [2–7].
Here we report a solid-state chemical synthesis route
that can be used to obtain bulk samples of amorphous
The synthesis method involves reacting polyanionic
‘‘Zintl phases’’ (e.g., NaSi, NaGe) together with metal
halides (e.g., CuCl, CoBr2) or ammonium salts (NH4Cl,
NH4Br). Sodium silicide is an ionic Zintl phase that
contains isolated Si44ꢀ tetrahedral clusters surrounded by
Na+ ions [8] (Fig. 1). The formation reaction proceeds
via a chemical ‘‘metathesis’’ process, in which ionic salts
(alkali halides) are formed as products, together with
solid metals (Cu, Co) or gaseous NH3, as well as the
elemental semiconductors (Si, Ge); e.g.,
Ã
Corresponding author. Christopher Ingold Laboratory, Materials
CuBrðsÞ þ NaSiðsÞ ¼ NaBrðsÞ þ SiðsÞ þ CuðsÞ:
(1)
Chemistry Centre, Department of Chemistry, University College
London, 20 Gordon Street, London WC1H 0AJ, UK. Fax:
+44 207 679 7463.
Depending upon the transition metal used, metal
silicides (e.g., FeSi2) are also occasionally formed in the
E-mail address: p.f.mcmillan@ucl.ac.uk (P.F. McMillan).
0022-4596/$ - see front matter r 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jssc.2004.12.040

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
938
minimal washing in water to remove the alkali halide.
We found that the physical state of the semiconductor
produced (i.e., amorphous vs. crystalline material, as
determined by X-ray powder diffraction) depends both
upon the precursors used, and the reaction temperature.
Analogous metathesis reactions provide a general
approach to solid-state synthesis, that has been applied
to a wide range of materials including various nitrides,
tetrelides, and chalcogenide compounds [9,10]. Related
reactions using Zintl phase precursors have also recently
been used to give rise to Si- and Ge-containing
luminescent nanoparticles in organic liquid suspension
[11,12].
2. Experimental methods
All solid-state Zintl phase precursors were prepared
and handled using glove box techniques (o10 ppm O2/
H2O). Metathesis reactions were carried out inside
sealed SiO2 glass tubes, with samples contained within
Ta crucibles or wrapped in Ta/Mo foil. The synthesis
reactions were carried out at 200–500 1C, for periods
ranging between 10 and 20 h. The solid products were
then washed in water, dilute acid (e.g., HNO3) and/or
ethanol/methanol, to remove various alkali halides and
transition metals or metal silicides that were produced in
the reactions.
The products were initially characterized by powder
X-ray diffraction (Siemens D-5000: CuKa radiation),
and by Raman spectroscopy (514.5 nm excitation), using
a home-built high-throughput spectrograph based on
Kaiser super-notch filters and a CCD detector [13].
Photoluminescence data were obtained with Ar+
(476 nm) laser excitation using a Bentham M300
single-grating monochromator and photomultiplier de-
tection system [4]. PL spectra were corrected for the
optical response of the system, and temperature effects
due to heating by the incident laser (300–400 1C) were
determined by a Planck fit to the corrected intensity data
between 1000 and 1700 nm. Initial FTIR spectra were
obtained by powder transmission on samples pressed
into KBr discs: later samples were examined by
transmission/reflectance spectroscopy of pure materials,
that were either pressed into plates between diamonds in
a diamond anvil cell, or as free-standing powdered
samples, using a Bruker IFSv-66 microbeam FTIR
system.
Solid-state 29Si MAS-NMR spectra were obtained
using a Bruker MSL-300 (7 T), corresponding to a
Larmor frequency oL ¼ 59:60 MHz for 29Si; chemical
shifts were referenced to TMS ((CH3)4Si). Samples were
spun at 5.5 kHz, using a 301 pulse with a 2 min recycle
time. 1H–29Si cross-polarized (CP) spectra were ob-
tained using a decoupling field of 50 kHz and 10 ms
contact time. Selected samples were also run using
Fig. 1. The monoclinic Zintl phases NaSi and NaGe contain isolated
Si44ꢀ or Ge44ꢀ tetrahedral groupings, as in the isoelectronic and
isostructural elemental allotrope of white P, surrounded by alkali
cations (shown as gray balls). The compounds are highly reactive to
atmospheric oxidation and hydration reactions, because of the exposed
Na+ species, and also due to the exposed electron lone pairs on the
pyramidal 3-coordinated Si atoms, and they must be handled in strictly
controlled low O2/H2O environments via glove box or Schlenck line
techniques. They are thus also ideal chemical precursors for
metastable/metathesis synthesis routes, that require products with a
high value of the chemical potential, compared with the desired
products.
synthesis reaction. To purify the elemental semiconduc-
tors formed by reaction (1), other solid products are
readily removed by washing in water, dilute acid, or
methanol/ethanol. In order to minimize or avoid the
washing steps, we also investigated the use of ammo-
nium bromide or chloride as reagents in the related
metathesis reactions; e.g.,
NH4BrðsÞ þ NaSiðsÞ
¼ NaBrðsÞ þ SiðsÞ þ NH3ðgÞ þ 1=2H2ðgÞ:
(2)
In this case, NH3 and H2 gases are expelled from the
system during synthesis, and the amorphous semicon-
ductor is recovered (in nominally pure form) after

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
939
analogous single-pulse and 1H–29Si CP conditions using
3. Results and discussion
a
Varian-Unity spectrometer operating at 9.4 T
(oL ¼ 79:46 MHz for 29Si).
3.1. Products formed and X-ray diffraction studies
Samples were examined by scanning electron micro-
scopy (SEM), and chemical compositions were obtained
via energy dispersive analysis, using either the SEM or a
JEOL 8600 electron microprobe. Selected samples were
also studied by transmission electron microscopy (TEM)
via imaging and selected area diffraction techniques,
using a JEOL JEM 2000-FX instrument operating at
200 kV (Southampton). Samples were prepared by
standard techniques as grain dispersions supported on
carbon films on Cu TEM specimen holders.
The amorphous/crystalline semiconductor products
were prepared by reactions (1) or (2) at T ¼
2002500 1C; using NaSi/NaGe as precursors, mixed
with various transition metal (Cu, Fe, Mn) or ammo-
nium halides (Cl, Br). Using metal halide precursors,
reactions were driven by exothermic formation of alkali
halides and the solid elements. Mixtures treated at
T4300 1C resulted in formation of microcrystalline Si/
Ge samples, along with metal particles mixed with the
alkali halides according to reaction (1). In a few cases,
transition metal silicides (e.g., FeSi2, CoSi2) were also
formed in the reaction. After completion of the
metathesis reactions, the various solid halides, silicides
and metal particles were removed from the samples by
washing with H2O, dilute acid (HNO3), and/or alcohol
(ethanol/methanol).
During syntheses carried out at lower T (200–300 1C),
it was generally observed that no crystalline semicon-
ductors, but only amorphous Si or Ge, were obtained, as
revealed by X-ray diffraction and Raman scattering
(Fig. 2). The bandgapof a sample of a-Si synthesized via
reaction (1) was determined to be 1.129 eV via electrical
Fig. 2. (a) X-ray diffraction patterns of amorphous silicon samples
synthesized by reaction between metal halides or NH4Cl and NaSi at
200 and 300 1C, after washing to remove the other solid products. The
lines below the amorphous diffraction pattern indicate the positions
and intensities of the powder diffraction lines for crystalline Si. (b)
Raman scattering of a-Si synthesized via reaction (2) between NH4Cl
and NaSi at 200 1C. The features above 550 cmꢀ1 are due to
multiphonon scattering.
Fig. 3. X-ray diffraction patterns of a-Si and a-Ge samples synthesized
by metathesis reactions (1) or (2), after washing to remove other solid
products. Weak crystalline features are sometimes apparent in the
spectra, as observed for the ‘‘a-Si-300’’ samples, dependent upon the
batch, especially for higher temperature syntheses. These additional
diffraction peaks can be assigned to the semiconductor clathrate phase
Na8Si46, that is usually obtained by controlled thermal decomposition
of the Zintl phase NaSi.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
940
conductivity measurements on pressed discs of pow-
dered material.
subtle differences were noted in the position,
width and relative amplitude of the ‘‘first sharp
diffraction peak’’ (FSDP) between different synthesis
runs (Figs. 2 and 3). It is known that the position and
appearance of the FSDP scales with the density among
amorphous materials [17]. The density of amorphous
silicon samples prepared by eꢀ-beam deposition on
crystalline Si, or by ion bombardment of c-Si, has been
determined to range between 2.26 and 2.29 g/cm3,
approximately 1.8–3.0% less dense than the crystalline
semiconductor (2.33 g/cm3) [18]. Our bulk samples
prepared by chemical synthesis exhibited an average
density slightly lower than this range of values,
approximately 2.0–2.1 g/cm3, as determined by sink-
float techniques.
In syntheses carried out via reaction (2), the (Si/Ge)
semiconductor products obtained generally remained
amorphous to X-ray diffraction, to higher reaction
temperatures than those formed by reaction (1). This is
explained by the formation of volatile products (NH3,
H2), that carry away part of the reaction enthalpy [10].
For reactions carried out at T4300 1C; weak additional
crystalline peaks occasionally appeared in X-ray diffrac-
tion patterns (Fig. 3). These peaks are identified as due
to ‘‘semiconductor clathrate’’ phases, e.g., Na8Si46
[14–16], that are formed metastably during the synthesis
process. These low-density clathrate materials are
usually made by controlled thermal decomposition of
the Zintl phase precursor (e.g., NaSi, NaGe). The
metathesis reactions could thus provide a new route to
the semiconducting (or semimetallic) clathrate-struc-
tured phases.
We also observed that most of the samples contained
fractions with a significant ‘‘tail’’ to lower densities,
down to o1.75 g/cm3. During our studies, we separated
some of the samples into density fractions for certain
physical and spectroscopic investigations, because of the
possibility of ‘‘polyamorphism’’ occurring within the
During X-ray characterization of the amorphous
semiconductors produced by reactions (1) and (2),
Fig. 4. (a) FTIR transmission and (b) reflectance transmittance spectra of an a-Si samples prepared via reaction (2), that did not contain any
significant impurities such as oxygen. The reflectance spectrum is referenced to Au, that exhibits 100% reflectance across the IR range.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
941
amorphous silicon, between a ‘‘normal’’ low-density a-
Si (LDA), and a ‘‘very low-density’’ form (VLDA).
The concept of density-driven ‘‘polyamorphism’’,
related to phase transitions occurring in simple liquids
at constant composition, has been described among
amorphous semiconductors, including Si and Ge, for
many years [19–22]. It is now gaining wide acceptance,
based on new experimental observations and results of
MD simulations [23–27]. A low–high density amor-
phous transition (LDA–HDA) is present under high-
pressure conditions, or during cooling from the metallic
liquid. In the present work, we initially thought that a
further polyamorphic transition could occur, between
LDA and VLDA polyamorphs, accessed metastably
during the solid-state metathesis synthesis. This was
based on a consideration of metastable phase relation-
ships occurring under ‘‘negative pressure’’ conditions
[28,29]. However, after a suite of characterization
studies on density-separated fractions of a-Si obtained
from various batches synthesized via reactions (1) and
(2), we concluded that the differences observed within
and between batches were most likely due to (a) partial
crystallization of the samples (to give Si clathrates or
nanocrystalline Si), or (b) to the presence of chemical
impurities (H, and also O, N), included within the
samples. However, the possibility of an LDA-VLDA
polyamorphic transformation is still not excluded, just
as in amorphous solid H2O, there is now evidence for an
additional very-high density (VHDA) form of amor-
phous ice [30,31], supplementing the already known
HDA and LDA polyamorphs [32,33].
3.2. Chemical characterization of the samples
Apart from H, the primary chemical impurity
detected within the samples was O, that was often
found to be present in concentrations up to 3–5%
by weight among earlier samples, but was reduced
Fig. 5. FTIR spectrum taken in reflectance mode of pressed discs of an early batch of a-Si sample containing substantial amounts of O (10–15 at%),
H (10 at%) and N (ꢁ8 at%) impurities, as determined for the bulk sample via chemical analysis. The spectra refer to different density fractions
separated from a single sample. Because Si–H stretching and other features appear as positive peaks in the spectra, these features are likely due to a
combination of reflectance and transmission effects at the sample surface. (a) Spectra from 700 to 2500 cmꢀ1 showing the detail of Si–H and Si–O(N)
stretching modes. (b) Full spectra between 430 and 7500 cmꢀ1
.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
942
to o1% in later synthesis batches. Because all
samples were handled and reactions carried out
in O-free environments, the oxidation reactions most
likely took place during washing, or while samples from
earlier batches were stored in air for upto several
months.
Hydrogen is a well-known impurity within a-Si; it is
important for passivation of broken bond defects, and
thus optimizing the electronic/photovoltaic properties of
the amorphous semiconductor. Initial batches of a-Si
prepared in this study were originally thought to contain
minimal amounts of H, mainly because no Si–H
vibrations could be detected by initial FTIR or Raman
scattering spectroscopy carried out on the sample.
Samples were taken to the Institut Laue-Langevin
(ILL) in Grenoble for examination of their low-energy
excitations by inelastic neutron scattering experiments
(spectrometer stations IN6 and D7) [36]. However,
during the neutron measurements, it was found that all
of the samples contained substantial H, with concentra-
tions ranging between 5 and 12 at% [37]. Because no
Si–H species had been observed by the preliminary
FTIR measurements on these materials, it was initially
thought that the hydrogen might be present as
molecular H2 groups contained within the a-Si network
[38–41]; attempts were made to remove such H-bearing
species by annealing under high-vacuum conditions at
T ¼ 3002450 1C [41]. However, as was revealed by later
characterization experiments, that process resulted in
partial recrystallization of the samples, yielding nano-
crystalline Si particles embedded within the amorphous
matrix [42–46]. These crystalline nanoparticles were
later detected by TEM and Raman scattering spectro-
scopy: those results are described below.
Some of the earlier-synthesized batches were in fact
found to contain very high oxygen contents, upto
10–20% by weight; they thus corresponded to amor-
phous SiOx (x ꢁ 1) or to physical mixtures of a-Si/a-
SiO2. The oxide components gave rise to obvious
features in the 29Si NMR and also the FTIR spectra,
described below. Some nitrogen was also be detected in
some samples prepared by reaction (2), that we believe is
likely to be incorporated in the amorphous matrix. The
incorporation of O (or N) atoms into an amorphous Si
matrix to form ‘‘siloxene’’ materials is thought to
provide enhancement of the luminescence characteristics
[2,34]; also the presence of amorphous/nanocrystalline
Si, Ge particles in an O-, N-rich matrix is known to
provide useful enhancement of the luminescence effi-
ciency [4,6,34,35]. These points are discussed below.
They are raised here to note that the serendipitous
incorporation of O,N impurities in the a-Si (and a-SiOx)
materials obtained by chemical synthesis (and subse-
quent storage in air) is not necessarily a ‘‘bad’’ outcome:
if controlled, it could lead to a useful new class of
optoelectronic materials.
Fig. 6. (a) Single pulse 29Si (301 pulse; 2 min recycle time, spun at 5.5 kHz) MAS NMR spectra of an a-Si sample containing no detectable impurities
(O,N) by chemical analysis. (b) 1H–29Si CP spectrum of the same sample. The correspondence between the single pulse and 1H-decoupled spectra
indicates that the sample contains H impurities distributed randomly with respect to the Si sites.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
943
3.3. FTIR spectroscopy
The features associated with Si–H vibrations were
best revealed in reflectance spectra, obtained from a
single side of the same pressed disc samples (Fig. 4(b)).
The reflectivity rises generally at higher energy values
due to the impending band edge (1:129 eV ¼ 9100 cmꢀ1).
The dip in the reflectance spectrum at approximately
2000 cmꢀ1 corresponds to Si–H vibrations in the sample.
(The adjacent sharpdipat 1630 cm ꢀ1, along with that
near 3500 cmꢀ1, are due to H2O adsorbed on the KBr
disc used to support the sample.)
Some of the earlier batches of a-Si samples were
found to contain substantial O (and N) impurities, as
well as H. Later protocols that minimized uptake of O/
H during washing and storage resulted in nearly pure a-
Si and SiHx materials. FTIR reflectance spectra of
density-separated fractions of an ‘‘impure’’ sample are
shown in Fig. 5. These spectra clearly show features due
to Si–H stretching and bending vibrations, and Si–O
(and perhaps also Si–N) stretching [47–49]. The Si–H
and Si–O features appear in the ‘‘reflectance’’ spectra as
FTIR transmission and reflectance spectra were
obtained from samples that were found to contain
minimal O, N or H impurities (the H contents of
samples prepared later in the study did not exceed
approximately 2–5 at% H) (Fig. 4). Powdered materials
were pressed into flat plates ꢁ10–50 mm thick between
two diamonds in a diamond anvil cell, and then run in
air (mounted on a KBr plate). Above the transmission
cutoff due to phonon absorption (550–600 cmꢀ1), the
transmission through the pressed discs was low, and
only a few weak resonances were observed throughout
the 500–4000 cmꢀ1 region (Fig. 4(a)). (Fig. 4 contains a
superposition of several transmission spectra: that is
why the doublet near 2350 cmꢀ1 due to atmospheric
CO2 appears as both positive and negative features in
the plot.) Similar IR spectra were obtained for a-Si
colloidal samples prepared by thermolysis of silanes in
supercritical hexane [46].
Fig. 7. (a) Single pulse 29Si MAS NMR spectrum of a sample containing ꢁ10 at% O impurity. This spectrum (unlike the others presented here) is
referenced to crystalline Si rather than TMS, so the peaks are shifted by 76.5 ppm relative to the chemical shift scale used for the other spectra. (b)
single pulse 29Si MAS NMR spectra of a-Si/SiOx samples prepared via reaction (2) (i.e., NaSi+NH4Cl) at different temperatures (T ¼ 200; 300,
400 1C). (c) 29Si MAS NMR spectra of a-Si/SiOx materials within different density fractions separated from a sample prepared via reaction (2):
bottom—1.50–1.75 g/cm3; middle—1.86–1.95 g/cm3; top—1.95–2.10 g/cm3.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
944
positive features that resemble the IR absorption spectra
[46–49]. This could mean that they are due to absorption
within thin films of impurity-rich material at the surface
of the sample grains, or that their appearance in the
reflectance spectrum is due to subtle effects on the
optical constants, within a region of low reflectivity. We
have already observed both positive and negative
contributions to the IR reflectivity, from minor impurity
resonances associated with O–H and C–O species in
silicate glass samples [50,51]. The form of the reflectivity
function in the infrared region is quite different to that
recorded for a-Si samples that contained essentially no
O and minor H impurities (Fig. 4). For the H, O(N)-
containing samples (Fig. 5), broad absorption features
at ꢁ630 cmꢀ1 and in the 900–1000 cmꢀ1 region are
intrinsic to a-Si, and are due to combinations of
transverse acoustic/optic phonons in the amorphous
semiconductor [47–49]. In the Si–H stretching region,
the envelope under the ꢁ2100 cmꢀ1 band contains
contributions from SiH, ¼SiH2 and –SiH3 mono- and
polyhydride species, isolated or clustered within the
bulk, and at surfaces [4,43,47–49]. From the integrated
IR absorption band intensities in the Si–H stretching
region, the bound H content is estimated to be between
5 and 15 at%, consistent with the neutron scattering
experiments [36,37]. These samples that were determined
to contain impurity O also exhibited features in the
1000–1100 cmꢀ1 region and at ꢁ800 cmꢀ1, that are
assigned to Si–O bond stretching and SiOSi linkages,
respectively [4,34,46–48]. The samples also showed a
Fig. 8. (a) Raman spectra of various silicon samples prepared by metathesis synthesis. None of the materials exhibited crystalline peaks via X-ray
diffraction. Most samples examined in the study were fully amorphous, and exhibited only a broad Si–Si stretching mode at ꢁ450 cmꢀ1. Other
samples contained a mixture of amorphous material and nanocrystalline regions, with Raman peaks ranging between 490 and 510 cmꢀ1 (the zone
center phonon for crystalline Si lies at 521 cmꢀ1). (b) Samples containing large amounts of nanocrystalline material and/or O (N) impurities exhibited
a strong luminescence background in the 1500–4200 cmꢀ1 (l ¼ 5572656 nm) region. The spectra shown here are for different density-separated
batches from a single sample prepared from NaSi+NH4Br at 200 1C, that contained significant O, N and H impurities (10–15 at% by bulk analysis),
and also showed the presence of nanocrystals. The largest fluorescence is obtained for the fraction with greatest density (42.1 g/cm3).

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
945
weak feature at 2250 cmꢀ1, that can be assigned to
O3SiH species. In the high frequency region (Fig. 5(b)),
features at 3630, 3350 and 2935 cmꢀ1superimposed upon
high-order excitations from the a-Si matrix, reveal the
presence of O–H and perhaps N–H vibrations within the
sample. The data in Fig. 5 show superimposed the
spectra from different density fractions separated within
the sample (synthesized at 200 1C from NaSi+NH4Br).
The lightest fraction (o1.75 g/cm3) shows the least
amount of Si–H impurities (weakest feature at
Fig. 9. (a) Selected area microdiffraction patterns in TEM showing the predominantly amorphous nature of the matrix material in samples prepared
by reactions (1) and (2). (b) Selected area diffraction pattern showing the presence of crystalline material, within nanocrystalline areas as shown in the
TEM image below. The first diffuse ring occurs at 2.52 A, that corresponds well with the FSDP measured by X-ray diffraction of the amorphous
material. (c) TEM image showing the presence of 2–20 nm nanocrystalline regions (exhibiting lattice fringes) within the amorphous Si samples
prepared in this study. The SAED patterns were collected from areas approximately 100 nm in diameter. The measured positions of the spots
correspond to lattice spacings of diamond-structured Si.

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
946
2080 cmꢀ1), but otherwise the spectra are nearly
identical. Another difference between the density frac-
tions is in the amount of O impurities, revealed by 29Si
NMR spectroscopy discussed below: this difference does
not show upwell in the FTIR spectra.
among samples that were density-separated into frac-
tions, the lowest density fraction exhibited the largest
amount of O contamination (Fig. 7(c)).
3.5. Raman and photoluminescence spectra
3.4. NMR spectroscopy
Most Raman spectra recorded for samples prepared
within this study were identical with previous determi-
nations for amorphous silicon, and were dominated by a
broad asymmetric Si–Si stretching band with its max-
imum at ꢁ450 cmꢀ1 (Fig. 2). However, some samples
exhibited Raman features that indicated the formation
of nanocrystalline silicon particles within the amor-
phous matrix [49,55–57]. For example, a sample
prepared via reaction (2) at 300 1C showed a sharp
Raman peak at 508 cmꢀ1, in addition to the broad
amorphous feature near 450 cmꢀ1 (Fig. 8). The Raman-
active phonon mode of bulk crystalline Si occurs at
521 cmꢀ1: this is shifted to higher or lower frequency by
compressive or tensile stresses acting on crystalline
samples, or by surface oxidation; however, the max-
imum shifts associated with these effects are ꢁ1–3 cmꢀ1
[57–59]. Nanocrystalline Si (and Ge) particles typically
have their Raman modes shifted to lower wavenumbers
because of the reduction in q ¼ 0 selection conditions,
and also due to structural relaxation occurring within
the finite particles [54–56]. The Raman signal observed
here is thus due to the appearance of nanocrystalline Si
within the amorphous sample, occurring during the
The highest purity samples of a-Si synthesized in this
study (i.e., with minimal or undetectable quanitties of
O,N impurities, and with H contents ranging from 5 to
10 at%) exhibited a single broad Gaussian peak in the
single pulse 29Si NMR spectrum with its maximum
between ꢀ60 and ꢀ70 ppm (relative to tetramethyl
silane: TMS), and a peak width at half maximum
(FWHM) ꢁ70 ppm (Fig. 6). The corresponding peak for
crystalline Si occurs at ꢀ80 ppm, with FWHM ¼ 6 ppm.
Previously 29Si MAS NMR spectra have been recorded
for a-Si samples prepared by sputtering methods [52]:
these samples exhibited an asymmetric NMR peak with
FWHM ꢁ70–80 ppm, with its maximum between ꢀ40
and ꢀ50 ppm. The reason for the difference between the
two results is not yet understood. The r.f. sputtered
samples certainly exist in a different, probably higher
energy, state of structural relaxation than the samples
prepared by chemical synthesis at lower temperature,
and they likely contain a smaller concentration of Si–H
species. It is well known that the 29Si spectra and
chemical shifts among Si polymorphs and other Si-
containing species are sensitive to small structural
variations in the bond length and angle distribution, as
well as to the nature of ligands attached to the Si atoms
[53]. It is quite likely that differences in the Si–Si bond
lengths and SiSiSi angles between the r.f. sputtered
samples and our materials prepared by chemical
precursor synthesis. It is also possible that a difference
in H content between our and the previous samples
could influence the NMR spectra. In our study, we also
1
obtained H–29Si CP spectra for various a-Si samples:
these gave rise to nearly identical spectra to those for the
single-pulse 29Si MAS studies (perhaps shifted to slightly
less negative ppm values), indicating that the H atoms
present within the samples are distributed approxi-
mately homogeneously relative to the different Si sites in
the amorphous structure (Fig. 6(b)).
Among samples that showed clear presence of O
impurities, the NMR spectra usually showed an addi-
tional feature that could be correlated with the presence
of SiOx species in the sample (Fig. 7). An ‘‘extra’’ peak
was often present at approximately ꢀ104 ppm, that lies
close to the 29Si NMR peak of amorphous SiO2
(ꢀ110 ppm [53]). This peak was usually least prominent
among samples synthesized at higher temperatures. (Fig.
7(b)), indicating that the lowest T samples were most
prone to oxidation during washing and subsequent
exposure to air (O2/H2O-rich environments). Further,
Fig. 10. Enlarged TEM image of a region within one sample of a-Si
that contained nanocrystalline material, showing lattice fringes within
the nanoparticles. The field of view is approximately 100 nm. The
distance between lattice planes is measured to be 2.75 A, that
corresponds to the spacing between Si atoms in a slightly expanded
(
diamond lattice (a0 ¼ 5:4301 A for Z ¼ 2 in the face-centered cubic
cell of bulk crystalline material).

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
947
solid-state syntheses. The presence of Si nanocrystals
was also detected by Raman spectroscopy in samples
that had been treated post-synthesis by heating in
vacuum in order to reduce the hydrogen content: this
procedure is known to induce crystallization in a-Si:H
materials [43–46,60,61]. The presence of nanocrystalline
particles 2–20 nm across within an amorphous matrix
was detected in selected samples studied by TEM and
selected area diffraction (Figs. 9 and 10).
During our Raman spectroscopic studies that were
originally designed to probe the possible existence of
Si–H and/or H–H vibrations, it was noted that some
samples exhibited a strong fluorescence background
(Fig. 8). Room temperature photoluminescence (PL)
spectra were then recorded with the same spectrometer
system from 1.35 to 2.4 eV at using 514.5 nm excitation.
Different batches of sample exhibited either a strong
luminescence with maximum at 1.78 eV (696 nm), or a
weaker broad luminescence band maximized at 2.1 eV
(590 nm) (Fig. 11).
spectrometer response and temperature effects (black-
body) induced by the incident laser. It was found that
samples were locally heated during the experiments to
approximately 650 K, by absorption of the incident
laser. Samples from batches that contained little or no
oxygen (or nitrogen) impurities, and that were found to
be fully amorphous from Raman and TEM studies,
showed no observable PL signal. It is known that only
weak PL at ꢁ1.6 eV is obtained from amorphous Si [39].
The strong red luminescence band observed at
1.65–1.78 eV has been associated with the presence of
nc-Si particles, that are embedded here within the
amorphous matrix [3,4,48,62]: it is thought to arise
from quantum confinement excitonic effects within the
clusters [2,63]. It is also thought that the PL efficiency
may be enhanced by the presence of O and/or N
impurities within the matrix, or concentrated at the
surface of the nanoclusters [4,6,34,44,48,62,63]. The
weaker, broad PL band in the green range (ꢁ2 eV) could
be associated with oxygen-related defects in the amor-
phous matrix, such as paramagnetic ꢂSi–Od centers
[34]. However, discrete Si nanoparticles 1.5–4.0 nm in
Further PL measurements were carried out using
476 nm excitation, carefully correcting the spectra for
Fig. 11. (a) Room temperature photoluminescence spectra of selected amorphous/nanocrystalline samples of silicon prepared in this study, excited
by an Ar+ laser (l ¼ 514:5 nm; 2.41 eV). The spectra are not corrected for spectrometer response or for thermal excitation effects. The spectra shown
here were collected from different density-separated fractions from a single batch of sample, that was shown to contain O, N and H impurities by
chemical analysis. The Raman spectra (weak peaks at right) also indicate the presence of nanoparticles. The sample was prepared from
NaSi+NH4Cl at 300 1C. Here the strong luminescence at 1.78 eV is observed for the lowest density fraction (o1.75 g/cm3); fractions with densities
1.9–2.2 g/cm3 show weaker luminescence at 2.1 eV (cf. spectra in Fig. 8). (b) PL spectrum excited by the l ¼ 476:5 nm line of an Ar+ laser. The
sample is similar to the low-density fraction that gave rise to the strong luminescence at 1.78 eV in the spectra above. The spectrum in (b) has been
corrected for spectrometer response and for thermal excitation effects (Planck function: T ¼ 650 K) [4,34].

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
948
dimension have recently been produced via thermolysis
in supercritical hydrocarbon fluids [46,64]. These mate-
rials exhibited clear size-dependent PL at 300–320 nm,
that is thought to be intrinsic to the unsupported
nanoparticles [64]. The origin of the PL in the solid-state
samples prepared by metathesis from reactive Zintl
phase precursors remains to be determined.
NMR spectra were obtained by J. Diefenbacher, at
Arizona State University.
References
[1] A. Shah, P. Torres, R. Tscharner, N. Wyrsch, H. Keppner,
Science 285 (1999) 692–698.
[2] L. Brus, J. Phys. Chem. 98 (1994) 3575–3581.
[3] E. Edelberg, S. Bergh, R. Naone, M. Hall, E.S. Aydil, Appl. Phys.
Lett. 68 (1996) 1415–1417.
4. Conclusion
[4] A.J. Kenyon, P.F. Trwoga, C.W. Pitt, G. Rehm, Appl. Phys. Lett.
73 (1998) 523–525.
We have prepared bulk samples of the amorphous
semiconductors Si and Ge using a new solid-state
chemical ‘‘metathesis’’ approach, using Zintl phases
(e.g., NaSi, NaGe) mixed with metal or ammonium
halides as precursors. At sufficiently high synthesis
temperatures, depending upon the precursors used and
the metathesis procedure followed, the bulk microcrys-
talline semiconductors were produced. At lower synth-
esis temperatures, for example in the 200–300 1C range,
the materials obtained generally showed only ‘‘amor-
phous’’ features in their X-ray diffraction patterns. It
was later established by Raman scattering and TEM
studies that some of these samples in fact contained
nanocrystalline semiconductors. Some synthesis runs
yielded small amounts of metastable crystalline ‘‘semi-
conductor clathrate’’ phases, such as Na8Si46. Some
oxidation of the amorphous semiconductor samples
occurred during washing steps to remove unwanted
products and unreacted starting materials, and also
during long-term storage in air. Such oxidation pro-
cesses could enhance the luminescence efficiency asso-
ciated with the embedded nanoparticles (red
luminescence), and also the intrinsic luminescence
associated with broken-bond defects in the amorphous
Si/SiOx matrix. Si nanoparticles were also obtained
within the amorphous matrix by annealing to remove
hydrogen introduced during synthesis, or were appar-
ently formed directly during the synthesis process, in
some cases. All of these parameters could be adjusted by
optimizing and controlling the synthesis procedure and
post-synthesis treatment steps, to yield a range of
amorphous or nano- to micro-crystalline Si and Ge
bulk materials, containing variable amounts of O, H
and N impurities. These should be further explored for
photo- and electro-luminescent or photovoltaic applica-
tions.
[5] C.-S. Yang, S.M. Kauzlarich, Y.C. Wang, Chem. Mater. 11
(1999) 3666–3670.
[6] S. Takeoda, M. Fujii, S. Hayashi, Phys. Stat. Sol. (b) 224 (2001)
229–232.
[7] L.H. Lie, M. Duerdin, E.M. Tuite, A. Houlton, B.R. Horrocks, J.
Electroanal. Chem. 538–539 (2002) 183–190.
[8] R. Schafer, W. Klemm, Z. Anorg. Allg. Chem. 312 (1961)
214–220.
[9] I.P. Parkin, Chem. Soc. Rev. (1996) 199–207.
[10] I.P. Parkin, Trans. Met. Chem. 27 (2002) 569–573.
[11] R.A. Bley, S.M. Kauzlarich, J. Am. Chem. Soc. 118 (1996)
12,461–12,462.
[12] B.R. Taylor, S.M. Kauzlarich, H.W.H. Lee, G.R. Delgado,
Chem. Mater. 10 (1998) 22–24.
[13] E. Soignard, P.F. McMillan, Defect chemistry in g-Si3N4 and g-
Ge3N4 spinel nitride phases probed by Raman scattering in the
laser-heated diamond anvil cell, Chem. Mater. 16 (2004)
3533–3542.
[14] C. Cros, M. Pouchard, P. Hagenmuller, J. Solid State Chem. 2
(1970) 570–581.
[15] G.K. Ramachandran, J. Dong, J. Diefenbacher, J. Gryko, R.F.
Marzke, O.F. Sankey, P.F. McMillan, J. Solid State Chem. 145
(1999) 716–730.
[16] E. Reny, P. Gravereau, C. Cros, M. Pouchard, J. Mater. Chem. 8
(1998) 2839–2844.
[17] D.L. Price, S.C. Moss, R. Reijers, M.-L. Saboungi, S. Susman, J.
Phys. C: Solid State Phys. 21 (1988) L1069–L1072.
[18] J.S. Custer, M.O. Thompson, D.C. Jacobson, J.M. Poate, S.
Roorda, W.C. Sinke, F. Spaepen, Appl. Phys. Lett. 64 (1994)
437–439.
[19] L.I. Aptekar, Sov. Phys. Dokl. 24 (1979) 993–995.
[20] O. Shimomura, S. Minomura, N. Sakai, K. Asaumi, K. Tamura,
J. Fukushima, H. Endo, Philos. Mag. 29 (1974) 547–558.
[21] E. Rapoport, J. Chem. Phys. 46 (1967) 2891–2894.
[22] E.G. Ponyatovsky, O.I. Barkalov, Mater. Sci. Rep. 8 (1992)
147–191.
[23] S.K. Deb, M.C. Wilding, M. Somayazulu, P.F. McMillan, Nature
414 (2001) 528–530.
[24] S. Sastry, C.A. Angell, Nat. Mater. 2 (2003) 739–743.
[25] N. Jakse, L. Hennet, D.L. Price, S. Krishnan, T. Key, E. Artacho,
B. Glorieux, A. Pasturel, M.-L. Saboungi, Appl. Phys. Lett. 83
(2003) 4733–4736.
[26] P.F. McMillan, J. Mater. Chem. 14 (2004) 1506–1512.
[27] E. Principi, A.D. Di Cicco, F. Decremps, A. Polian, S. De
Panfilis, A. Filipponi, Phys. Rev. B 69 (2004) 201201(4).
[28] M. Wilson, P.F. McMillan, Phys. Rev. Lett. 90 (2003)
135,703–135,707.
Acknowledgments
[29] P.F. McMillan, Nat. Mater. 1 (2002) 19–25.
[30] J.L. Finney, D.T. Bowron, A.K. Soper, T. Loerting, E. Mayer, A.
Hallbrucker, Phys. Rev. Lett. 89 (2002) 205,503–205,507.
[31] C.A. Tulk, C.J. Benmore, J. Urquidi, D.D. Klug, J. Neuefeind, B.
Tomberli, P.A. Egelstaff, Science 297 (2002) 1320–1324.
[32] O. Mishima, L.D. Calvert, E. Whalley, Nature 314 (1985) 76–78.
PFM acknowledges support from the Wolfson-Royal
Society Research Merit Award scheme. R. Arledge and
also Ms. Ja-Mia Ward were associated with this project
as undergraduate students at JSU. Some preliminary

ARTICLE IN PRESS
P.F. McMillan et al. / Journal of Solid State Chemistry 178 (2005) 937–949
949
[33] O. Mishima, H.E. Stanley, Nature 396 (1998) 329–335.
[34] A.J. Kenyon, P.F. Trwoga, C.W. Pitt, J. Appl. Phys. 79 (1996)
9291–9300.
[48] E. Edelberg, S. Bergh, R. Naone, M. Hall, E.S. Aydil, J. Appl.
Phys. 81 (1997) 2410–2417.
[49] D.C. Marra, E.A. Edelberg, R.L. Naone, E.S. Aydil, J. Vac. Sci.
Technol. A 16 (1998) 3199–3210.
[35] N.-M. Park, C.-J. Choi, T.-Y. Seong, S.-J. Park, Phys. Rev. Lett.
86 (2001) 1355–1357.
[50] A. Grzechnik, H.D. Zimmerman, R. Hervig, P. King, P.F.
McMillan, Contrib. Mineral. Petrol. 125 (1996) 311–318.
[51] G. Moore, P.F. McMillan, A. Chizmeshya, Geochim. Cosmo-
chim. Acta 64 (2000) 3571–3579.
[36] The ILL experiment resulted from ideas developed between C.A.
Angell and H. Schober regarding the likely low frequency (‘‘boson
peak’’) characteristics of an ‘‘ideal glass’’ such as a-Si, within a
collaborative programme developed with A. Toelle and M. Koza.
J. Gryko and PFM provided a-Si samples for the study, and
assisted with the neutron experiments.
[52] W.-L. Shao, J. Shinar, B.C. Gerstein, F. Li, J.S. Lannin, Phys.
Rev. B 41 (1990) 9491–9494.
[53] G. Engelhardt, D. Michel, High-Resolution Solid-State NMR of
Silicates and Zeolites, Wiley, New York, 1987.
[37] The high H contents in our a-Si samples were determined using
ILL instrument D7, developed and directed by Dr. R. Stewart,
that permits spin-polarised neutron scattering spectroscopy of
materials, to independently examine coherent- and incoherently
scattered channels.
[54] A.V. Andrianov, G. Polisski, J. Morgan, F. Koch, J. Lumin. 80
(1999) 193–198.
[55] S.K. Deb, in: M. Yousuf, N. Subramian, K. Govind Rajan (Eds.),
Adv. High Pressure Sci. Technol. Proceedings of the IV NCHST,
IGCAR, Kalpakkam, India, 1997, pp. 147–152.
[38] Y.J. Chabal, C.K.N. Patel, Phys. Rev. Lett. 53 (1984) 210–213.
[39] Y.J. Chabal, C.K.N. Patel, Phys. Rev. Lett. 53 (1984) 1771–1774.
[40] C.A. Guy, A.C. Wright, R.N. Sinclair, R.J. Stewart, F. Jansen,
J. Non-Cryst. Solids 196 (1996) 260–266.
[56] A.V. Kolobov, J. Mater. Sci. 12 (2004) 195–203.
[57] Th. Englert, G. Abstreiter, J. Pontcharra, Solid State Electron. 23
(1980) 31–33.
[41] M. Vergnat, S. Houssaıni, G. Marchal, Ph. Mangin, Phys. Rev. B
47 (1993) 7584–7587.
[58] E. Anastassakis, A. Pinczuk, E. Burstein, F.H. Pollak, M.
Cardona, Solid State Commun. 8 (1970) 133–138.
[59] J.T. Fitch, C.H. Bjorkman, G. Lucovsky, F.H. Pollak, X. Yin,
J. Vac. Sci. Technol. B 7 (1989) 775–781.
[42] N.H. Nickel, W.B. Jackson, Phys. Rev. B 51 (1995) 4872–4881.
[43] I. Kaiser, N.H. Nickel, W. Fuhs, W. Pilz, Phys. Rev. B 58 (1998)
R1718–R1721.
[60] J.J. Boland, G.N. Parsons, Science 256 (1992) 1304–1306.
[61] P. Roca i Cabarrocas, J. Non-Cryst. Solids 266–269 (2000) 31–37.
[62] E. Edelberg, S. Bergh, R. Naone, M. Hall, E.S. Aydil, Appl. Phys.
Lett. 68 (1996) 1415–1417.
[44] A. Fontcuberta i Morral, J. Bertomeu, P. Roca i Cabarrocas,
Mater. Sci. Eng. B 69–70 (2000) 559–563.
[45] M. Katiyar, J.R. Abelson, Mater. Sci. Eng. A 304–306 (2001)
349–352.
[63] M.J. Sailor, E.J. Lee, Adv. Mater. 9 (1997) 783–793.
[64] J.D. Holmes, K.J. Ziegler, R.C. Doty, L.E. Pell, K.P. Johnston,
B.A. Korgel, J. Am. Chem. Soc. 123 (2001) 3743–3748.
[46] L.A. Pell, A.D. Schricker, F.V. Mikulec, B.A. Korgel, Langmuir
20 (2004) 6546–6548.
[47] G. Lucovsky, R.J. Nemanich, J.C. Knights, Phys. Rev. B 19
(1979) 2064–2073.