Bimodal phase percolation model for the structure of Ge-Se glasses and the existence
of the intermediate phase
Pierre Lucas,
1
Ellyn A. King,
1,2
Ozgur Gulbiten,
1
Jeffery L. Yarger,
3
Emmanuel Soignard,
3
and Bruno Bureau
2
1
Department of Materials Science and Engineering, University of Arizona, 4715 E. Fort Lowell Road, Tucson, Arizona 85712, USA
2
UMR CNRS 6226 Sciences Chimiques, Groupe Verres et Céramiques, Université de Rennes I, Campus de Beaulieu, 35042 Rennes Cedex,
France
3
Department of Chemistry and Biochemistry, Arizona State University, Tempe, Arizona 85287-1604, USA
Received 19 October 2009; published 18 December 2009
A detailed nuclear magnetic resonance and Raman study of Ge
x
Se
1-x
glasses indicate that the glass structure
is composed of intertwined microdomains of GeSe
2
and Se
n
. Static nuclear magnetic resonance spectra of
glasses ranging from 0 x
1
3
reveal the absence of Ge-Se-Se fragments in the structure. High temperature
nuclear magnetic resonance showing considerable line narrowing confirms this observation. More importantly,
the fraction of Se-Se-Se obtained by integration of nuclear magnetic resonance lines matches closely the
percentage predicted for a bimodal phase model and is not consistent with the existence of Ge-Se-Se frag-
ments. Raman spectra collected on the same glass also confirm the existence of GeSe
2
domains up to high
selenium concentrations. The mobility of the Se
n
phase observed at high temperature while the GeSe
2
phase
remains rigid is consistent with their respective underconstrained and overconstrained structural nature. The
proposed bimodal phase percolation model is consistent with the original Phillips and Thorpe theory however
it is clearly at odds with the intermediate phase model which predicts large amounts of Ge-Se-Se fragments in
the structure. A calorimetric study performed over a wide range of cooling/heating rates shows a narrow
composition dependence centered at r = 2.4 in contrast with the wide reversibility window observed by
Modulated Differential Scanning Calorimetry. This suggests that the observation of the reversibility window
associated with the intermediate phase in Ge-Se glasses could be an experimental artifact resulting from the use
of a single modulation frequency.
DOI: 10.1103/PhysRevB.80.214114 PACS numbers: 61.43.Dq, 81.05.Gc
I. INTRODUCTION
The structure of Ge
x
Se
1-x
glasses has long been a source
of controversy in the field of amorphous semiconductors.
The structure of Ge
x
Se
1-x
glasses in the range 0 x
1
3
was
initially described using the chain crossing model CCM,
whereby GeSe
4/2
tetrahedra are linked by Se chains whose
length increases with Se content.
1,2
However, this model
was quickly contradicted by the persistent observation of
a Raman line near 213 cm
-1
associated with Ge-Se-Ge
sequences
3
and later assigned to structural fragments com-
posed of corner- and edge-sharing tetrahedra.
4,5
The presence
of edge-sharing tetrahedra in Se-rich glasses is not consistent
with the CCM and in particular contradicts the existence of
an “ordered phase”
1
at x = 0.2 GeSe
4
where all Ge atoms
would be linked by a Se-Se doublet. More recently, the struc-
ture of Ge
x
Se
1-x
has been the object of intense modeling
efforts centered on topological arguments involving counting
constraints and atomic degrees of freedom.
6–14
An original
model by Phillips and Thorp predicts an optimal glass for an
average number of bonds per atom r = 2.4 corresponding to
a topology where bond and angular constraints exactly
equals the number of degrees of freedom in the structure.
6–8
This composition is characterized as the threshold between
an underconstrained floppy phase for r 2.4 and an over-
constrained rigid phase for r 2.4. The transition point
r = 2.4 is thought to correspond to the composition at which
the rigid phase percolates throughout the structure. This
model was later refined by invoking the “self-organization”
of the glassy network which results in a structure that re-
mains rigid but unstressed over a wider range of composi-
tions near r = 2.4.
15
The unstressed domain between the
floppy and rigid phases was then termed the “intermediate
phase.” Subsequently, models based on connecting structural
fragments of different connectivity were developed to ratio-
nalize the intermediate phase
11–14
and experimental measure-
ments were performed to corroborate its existence.
10
In par-
ticular, the nonreversible enthalpy obtained by Modulated
Differential Scanning Calorimetry MDSC was shown to
decrease in this domain.
10
In any case it was found that
the structural origin of the intermediate phase requires the
existence of Se-Se-Ge “isostatic” structural fragments in or-
der to account for the “rigid but unstressed” nature of the
network.
12–14
Among structural probes, solid state nuclear magnetic
resonance constitutes a unique technique for examining the
local environments of active elements in a qualitative and
quantitative manner. But while silicate and phosphate glasses
have been widely studied, chalcogenides and in particular
selenides have received little attention due to the low nuclear
magnetic resonance sensitivity of selenium.
16
Nevertheless,
high quality spectra can be obtained using extended acquisi-
tion times up to several days.
17,18
In this work, an extensive
high-temperature nuclear magnetic resonance study shows
evidence for a different structural model of Ge-Se glasses
based on a bimodal percolation of Se
n
and GeSe
2
phases. It
is shown that no significant amount of Se-Se-Ge fragments is
present in the structure. These results are at odds with the
existence of the intermediate phase. Calorimetric measure-
ments on a series of Ge-Se glasses further suggest that the
observation of the intermediate phase could be based on an
experimental artifact.
PHYSICAL REVIEW B 80, 214114 2009
1098-0121/2009/8021/2141148 ©2009 The American Physical Society 214114-1