See discussions, stats, and author profiles for this publication at: [Link]

net/publication/257347677

The glassy and supercooled state of elemental sulfur: Vibrational modes,

structure metastability, and polymer content

Article in The Journal of Chemical Physics · September 2013

DOI: 10.1063/1.4821592 · Source: PubMed

CITATIONS READS

37 1,025

4 authors, including:

Konstantinos Andrikopoulos Olga Falagara

Foundation for Research and Technology Hellas University of Ioannina
107 PUBLICATIONS 2,936 CITATIONS 1 PUBLICATION 37 CITATIONS

SEE PROFILE SEE PROFILE

Spyros N. Yannopoulos
University of Patras
223 PUBLICATIONS 5,573 CITATIONS

SEE PROFILE

All content following this page was uploaded by Spyros N. Yannopoulos on 16 May 2014.

The user has requested enhancement of the downloaded file.

The glassy and supercooled state of elemental sulfur: Vibrational modes,
structure metastability, and polymer content
K. S. Andrikopoulos, A. G. Kalampounias, O. Falagara, and S. N. Yannopoulos

Citation: J. Chem. Phys. 139, 124501 (2013); doi: 10.1063/1.4821592

View online: [Link]
View Table of Contents: [Link]
Published by the AIP Publishing LLC.

Additional information on J. Chem. Phys.

Journal Homepage: [Link]
Journal Information: [Link]
Top downloads: [Link]
Information for Authors: [Link]
 THE JOURNAL OF CHEMICAL PHYSICS 139, 124501 (2013)

The glassy and supercooled state of elemental sulfur: Vibrational modes,

structure metastability, and polymer content
K. S. Andrikopoulos,1,a) A. G. Kalampounias,1 O. Falagara,1,2 and S. N. Yannopoulos1,b)
1
Foundation for Research and Technology Hellas, Institute of Chemical Engineering Sciences
(FORTH-ICE/HT), P.O. Box 1414, GR – 26504 Rio-Patras, Greece
2
Department of Chemistry, University of Ioaninna, GR – 45110 Ioannina, Greece
(Received 21 June 2013; accepted 4 September 2013; published online 23 September 2013)

We report a detailed investigation of vibrational modes, structure, and dynamics of elemental sulfur
in the glassy and the supercooled state, using Raman scattering and ab initio calculations. Polarized
Raman spectra are recorded – for sulfur quenched from 473 K – over a broad temperature range
from 93 K to 273 K where the supercooled liquid crystallized. The temperature induced shifts of
the majority of the vibrational modes are determined and compared with the corresponding ones of
crystalline sulfur. Analysis of the reduced isotropic spectra showed that the structure of the quenched
product is composed of eight member rings (S8 ) and polymeric chains (Sμ ) with a relative fraction
comparable to that of the parent liquid at 473 K. Low temperature spectra, where spectral line broad-
ening due to thermal effects is limited, revealed that two different polymeric species are present in
the glass with distinct vibrational frequencies. Their interpretation was assisted by ab initio calcula-
tions used to simulate the vibrational frequencies of polymeric chains S8k (k = 1, . . . , 7). Theoretical
results exhibit an increasing breathing mode frequency for sulfur chains up to k = 2, although it
remains constant beyond the above value. The polymeric content is metastable; heating the glass
above its glass transition temperature, Tg , destabilizes the chains and drives them back to the more
thermodynamically stable rings. This bond interchange mechanism provides the structural origin of
a secondary relaxation process in supercooled sulfur reported long ago, which has been also consid-
ered as a complication in the correct fragility estimation of this material. Finally, the Boson peak of
the glass was found to exhibit strong temperature dependence even at temperatures below Tg . © 2013
AIP Publishing LLC. [[Link]

I. INTRODUCTION Details regarding the various crystalline phases of sul-

fur are rather well known. The interest in understanding some
Elemental sulfur has received great attention from both
aspects of sulfur’s LLPT has been renewed as evidenced from
the experimental and theoretical points of view.1 The interest
the number of recent investigations.3–10 The temperature de-
arises from the following main reasons. (i) Although being an
pendence of the extent of polymerization φ(T) expressed as
elemental solid, sulfur exhibits a plethora of stable allotropes
the relative weight fraction of the polymer content at T ≥ Tλ
including cyclic molecules of various sizes. Actually, sulfur is
is perhaps the most valuable parameter of sulfur’s living poly-
the element with the larger number of known solid allotropes
merization, and the central issue of several studies.3–10 φ(T)
than any other element.1 (ii) Liquid sulfur displays a remark-
has been used to extract many thermodynamic quantities of
able thermoreversible liquid-liquid phase transition (LLPT)
the LLPT. In view of the absence of a reliable method for an
which belongs to the λ-type transitions. Sulfur’s LLPT is es-
in situ determination of the extent of polymerization many re-
sentially a living polymerization transition,2 occurring at Tλ
searchers have attempted, in the past, to estimate φ(T) using
≈ 159 ◦ C with a subsequent viscosity increase of about four
ex situ methods, where the quenched product was dissolved
orders in magnitude over a temperature range of 20 ◦ C around
in proper solvents in order to isolate the polymeric content.
Tλ . In particular, molten sulfur contains S8 rings; at Tλ the
However, it has been shown that φ(T) can be extracted in
rings open to S8∗ diradicals and concatenate to form long poly-
situ with high accuracy by exploiting the distinctive Raman
meric chains or polymeric sulfur Sμ . (iii) Liquid sulfur can be
response of the symmetric S–S bond-stretching vibrational
obtained in the bulk glassy state once rapidly quenched from
mode situated at ∼472 cm−1 when the S–S bond is part of
any temperature Tq > Tλ .1(a) The presence of polymeric sul-
a S8 ring and at ∼461 cm−1 when S–S bond participates in
fur is the key factor that facilitates glass formation and the ob-
a chain. This advance made it possible to obtain accurate re-
tained glass contains both types of molecules, i.e., S8 and Sμ .
sults for φ(T) in several cases of sulfur polymerization, i.e., in
Although the formation of a glass composed of pure S8 ring
the bulk,4(a)–4(c) under nanoscopic confinement where round-
molecules, quenched from Tq < Tλ , has not yet been demon-
ing effects of the sharpness of the λ-transition was observed,5
strated, it cannot be excluded on the basis of sufficiently fast
and in the presence of dopants or impurity atoms.6 These
quenching.
results have proved useful to check the predictions of the-
a) E-mail: candrik@[Link]. oretical works7 and computer simulations.8 Additional ex-
b) E-mail: sny@[Link]. perimental studies complement our knowledge on dynamical

0021-9606/2013/139(12)/124501/11/$30.00 139, 124501-1 © 2013 AIP Publishing LLC

124501-2 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

aspects of sulfur’s LLPT.9, 10 Further, the establishment of an of the specific volume and the shear modulus of g-S produced
experimental methodology on the in situ investigations of by quick quenched sulfur from 200 ◦ C to dry ice was studied
monomer ↔ polymer equilibrium based on Raman scattering by Tobolsky et al.13(a) who suggested that g-S is a mixture of
has proved useful in understanding photo-induced structural polymeric chains and S8 rings, both being in the amorphous
changes in light-sensitive, sulfur-rich glasses.11 state. A qualitative analysis of quenched sulfur focusing on
The rest of the article is organized as follows. A brief sur- S7 , S12 , a-S18 , and S20 species and their preparation from S8
vey on g-S properties is presented in Sec. II. Section III con- is given elsewhere.1(d) Structural studies of g-S have been con-
tains the experimental and computational details. Section IV ducted by neutron diffraction experiments,15 which indicated
presents the results and their discussion and is subdivided into that the pair correlation function of liquid sulfur and quick
six subsections. In Sec. IV A we provide a brief overview quenched sulfur differ significantly in the intermediate-range
of the spectra manipulation and data analysis procedure. De- order. Some preliminary results on g-S have been obtained by
tails about the assignment of Raman bands in g-S are pro- Raman scattering experiments.16 Recently, a detailed study of
vided in Sec. IV B. Raman band anharmonicities are dis- the vibrational dynamics of g-S, including acoustic properties,
cussed in Sec. IV C. The structure of g-S as revealed from the non-ergodicity factor, and the fragility of the supercooled
the Raman spectra analysis and assisted by ab initio calcula- liquid has been reported.10(b)
tions is described in Sec. IV D with emphasis on the poly- Given the above context it is obvious that a limited ex-
mer fraction and the existence of two different polymeric tent of information exists concerning the glassy and super-
species. Structure metastability of the supercooled liquid and cooled liquid state of elemental sulfur. Based on our previ-
the chains-to-rings equilibrium is the subject of Sec. IV E. Fi- ous work4(a)–4(c) where Raman scattering was used to iden-
nally, the main conclusions of the article are summarized in tify and quantify the polymeric content and its temperature
Sec. V. dependence, in this paper we present a detailed temperature
dependent Raman scattering study focusing on several struc-
tural and dynamical aspects of g-S. Ab initio studies of sulfur
II. BRIEF HISTORY ON GLASSY ELEMENTAL
chains S8k of various lengths (k = 1, 2, . . . , 8) are employed
SULFUR PROPERTIES
to assist the interpretation of Raman data.
While several issues of the LLPT of elemental sulfur
have been extensively studied and adequately understood, the
same is not true for the nature of elemental glassy sulfur,
III. EXPERIMENTAL AND COMPUTATIONAL DETAILS
g-S. A fact that complicates the study of the glassy state
is that the glass itself also exhibits a kind of allotropy. In Elemental sulfur (99.9995% purity from Alfa Aesar) was
particular, the glass composition, and hence the glass struc- further purified by repeated distillations into evacuated sil-
ture and properties, depend strongly upon preparation con- ica tubes. The product obtained by distillation was loaded
ditions and especially on the quenching temperature, Tq . As under inert atmosphere into a carefully cleaned silica tube
mentioned above, the glassy state is obtained only if poly- with dimensions 4 mm outside diameter – 3 mm inside di-
meric species are present in the parent liquid to be quenched ameter. After loading with appropriate amounts of sulfur, the
and hence polymers are inevitably present in the glass.1(a) In silica tube was evacuated and flame sealed. The tube was
the ideal case of a sufficiently fast quenching rate the liq- previously cleaned with a dilute HF solution, rinsed many
uid would freeze to a glassy state with the same polymer times with triple distilled water, and baked and degassed at
content and structure as the liquid at Tq . Glassy sulfur, be- 1000 K under vacuum. Removing external “particles” is
ing a solid metastable configuration of the liquid, can main- important since the polymerization transition of sulfur is
tain the liquid’s features as long as its temperature does not strongly dependent upon such impurities. The cell was heated
exceed the glass transition temperature. At T > Tg kinetic at 473 K for at least 3 h and then was quenched in liq-
effects concerning the rings ↔ chains equilibrium become uid nitrogen and immediately placed in a home-made optical
important. cryogenic device which was already cooled at low temper-
The fact that sulfur can be obtained in the glassy phase ature. The fast transfer ensures that the temperature rise of
has already been known for more than 150 years; see, for ex- the quenched g-S was negligible. The cryogenic device is a
ample, Ref. 12, for early reviews on the subject. Since glass liquid-nitrogen-cooled Pyrex “Dewar type” cryostat. Within
structure depends strongly on Tq , evidently Tg should depend the cryostat a copper metal cylindrical block (10 mm diam-
on the preparation conditions. It has been reported that g-S eter; 4 cm long) with optical opening was used as sample
rapidly quenched from 200 ◦ C exhibits a glass transition tem- holder. The cylindrical block was wound with insulated wire
perature Tg ≈ −30 ◦ C.13 The glass transition temperature is (Ari Industries Inc., Illinois) resistance allowing temperature
expected to increase with increasing polymer content. It is in- variation and control in the temperature range 80-500 K. A
teresting to note that polymeric S (commercially known as temperature controller was used which allowed measurements
Crystex), obtained after extracting S8 molecules by dissolving with accuracy ±0.2 K.
the quenched glass in CS2 , is a semi-crystalline product with Raman spectra were in situ obtained from the transpar-
Tg ≈ 70 ◦ C.13 Structural investigations on g-S are sporadic ent sample. A He-Ne laser was used as the excitation source
through the literature. Early x-ray studies on g-S have served (632.8 nm); it is important to work at off-resonant conditions
to distinguish the structure of the amorphous sulfur from that so as to avoid possible photo-induced structural transforma-
of the corresponding crystal.14 The temperature dependence tion in g-S. The right-angle scattered light was collected and
124501-3 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

IV. RESULTS AND DISCUSSION

A. Data analysis
Because an important issue of the present work is the
quantification of the rings ↔ polymers transformation, the
raw Raman intensities must be corrected. In particular, for
an accurate determination of the relative Raman intensities
we have carried out two types of spectra corrections. First,
we have disentangled the scattered intensity caused by the di-
agonal terms of the Raman tensor (isotropic spectrum) from
the scattered intensity caused by the off-diagonal elements
(anisotropic spectrum); and focused our analysis on the for-
mer part which reflects purely vibrational motion. The exper-
imental Raman intensities I V V (ν̃) and I V H (ν̃) of the two po-
larization configurations (VV and VH) used for recording the
spectra can be used to calculate the isotropic and anisotropic
scattering intensities through the relations:
4
I iso (ν̃) = I V V (ν̃) − I V H (ν̃), (1a)
3

I aniso (ν̃) = I V H (ν̃), (1b)

FIG. 1. Raman spectra of glassy and supercooled sulfur (quenched from

where ν̃ is the Raman shift in cm−1 . Representative VV and
473 K) at various temperatures. Solid (black) and dashed (red) lines repre- VH spectra of g-S at 93 K are shown in Fig. 2(a).
sent the VV and VH spectra, respectively. Second, the relative intensities of the vibrational lines in a
Raman spectrum depend – among other factors – on the pop-
ulation of the vibrational energy levels involved in the scat-
analyzed by a Jobin Yvon T64000 Raman system operating tering process, which are sensitive to temperature variations.
at the triple subtractive mode, with a spectral resolution of In order to remove thermal effects and to reveal the genuine
∼1.5 cm−1 . Both polarized (VV: vertical polarization of changes of “species” populations, the reduced representation
incident laser – vertical analysis of scattered light) and has been used; the Stokes-side reduced Raman intensity (I red )
depolarized (VH: horizontal polarization of incident laser – is related to the experimentally measured one (I exp ) with the
vertical analysis of scattered light) scattering geometries were
employed. Spectra of sulfur glass were recorded by heating
g-S in the temperature range 93–273 K with an incremental
step of 5 or 10 K. The system was equilibrated at each specific
temperature for a time period of ∼30 min before the acquisi-
tion of the corresponding spectrum. At 273 K the transparency
of the sample deteriorated significantly due to diffuse scat-
tering from the crystallites that began to develop. At 278 K
the acquired spectra matched those of the a-S8 crystal. Repre-
sentative spectra of glassy and supercooled sulfur at various
temperatures are shown in Fig. 1.
Ab initio calculations were performed using the Gaus-
sian 09 package.17 Optimization of molecular geometries
and calculation of vibrational frequencies were carried out
at Hartree-Fock (HF), density functional theory (DFT), and
Moeller-Plesset second order perturbation theory (MP2).
A series of functionals was used at the DFT level, in-
cluding the long range corrected CAM-B3LYP18(a) and hy-
brid functionals (B3LYP,18(b) a form of Becke Half and
Half18(c) and a form of Becke Half and Half LYP18(d) ). Ba-
sis sets used include split valence basis set 3-21G18(e) (aug-
mented with polarization and diffuse functions), correlation-
consistent polarized valence double zeta basis set (cc- FIG. 2. Polarized anisotropic and isotropic Raman spectra of g-S at 93 K.
Stars denote He-Ne laser plasma lines (they served for frequency calibra-
pVDZ)18(f) (augmented with polarization functions), and
tion). Almost all of the resolved bands are attributed to S8 species as well as
Goddard/Smedley effective core potentials18(g) (augmented polymeric sulfur. Frequency ranges for the torsional/librational, bending, and
with polarization functions). stretching intermolecular are also denoted.
124501-4 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

equation: polymerization temperature in the case of dopants,6 as well as

from the quantitative agreement of experimental data,5(a), 5(b)
I red (ν̃) = (ν̃0 − ν̃)−4 ν̃[n(ν̃, T ) + 1]−1 I exp (ν̃), (2)
and modeling of the polymerization transition in confined
where the term in the fourth power is the usual correction for geometries.5(c)
the wavelength dependence of the scattered intensity and ν̃0
denotes the wavenumber of the incident radiation. The term
n(ν̃, T ) = [exp(¯ν̃/kB T ) − 1]−1 is the Bose-Einstein occu- B. Identification of the Raman bands in g-S
pation number where ¯ and kB are the Planck and Boltz-
The fact that the laser beam could propagate undistorted
mann constants, respectively. A typical reduced isotropic
through the quenched product indicated that the sample was
spectrum is shown in Fig. 2(b). Fittings of reduced isotropic
a glass rather than a polycrystalline form of sulfur. This indi-
Raman spectra were carried out according to the procedure of
cation was furthermore supported by examination of the po-
Ref. 4(a) and the polymeric fraction φ(T) was calculated by
larized Raman spectra obtained. The depolarization ratio is
I Sμ /(I Sμ + I S8 ), where I Sμ and I S8 are the reduced isotropic
in accordance with the corresponding one of the melt,4 hence
Raman intensities of the bands assigned to polymeric, Sμ , and
providing evidence that the material is in its glassy state. Fur-
ring, S8 , species, respectively.
thermore, apart from the correct polarization behavior of the
The procedure, to extract φ(T) by in situ Raman scat-
most characteristic bands, the Boson peak (BP), which is a
tering, has been challenged in two main aspects:4(d) (a) The
well-known and genuine spectral feature of glassy materials,
assumption for proportionality between the ratio of Raman
is clearly observed at low wavenumbers, see Fig. 2. An ex-
peak intensities (for S8 and Sμ species) and the concentration
ample of fitting the reduced isotropic and anisotropic Raman
ratio of the corresponding chemical species was considered
spectra of g-S at 93 K is given in Figs. 3(a) and 3(b).
to introduce errors. (b) The existence of small ring molecules,
Early (indirect) evidence provided in previous
other than S8 , i.e., mainly S7 and S6 (ignored in our analysis),
works,12(a), 12(b), 13(a), 14(b), 14(c) that quick quenched sulfur
were considered to contribute cumulatively in the spectral re-
glass consists of both S8 rings as well as polymeric species,
gion of the polymer peak. As pointed out in Ref. 4(d) the two
is reconfirmed here by the vibrational modes appearing in
approximations mentioned above could cause overestimation
the Raman spectrum. The Raman spectrum of g-S can be
of φ(T) obtained by in situ Raman experiments in relation to
divided into three main spectral ranges as shown in Fig. 2. In
φ(T) extracted by the ex situ quench-and-dissolution method.
brief, the most prominent Raman bands in each range are the
This dispute was clarified in a subsequent article,4(b) where
following:
detailed arguments were provided to show that the above
limitations do not practically affect the Raman determination (i) Intramolecular bond stretching region: 400-500 cm−1 .
of the polymer fraction as a function of temperature. It was As previous studies have demonstrated,4 the dominant
shown that the presence of non-S8 rings is negligible under
the conditions that in situ Raman experiments are performed.
No evidence for any of small non-S8 ring molecules was
provided in the Raman spectra of the liquid. It was also
described that even in the presence of certain non-S8 rings
(at the concentrations indicated by quench and dissolution
experiments), their contribution to the isotropic Raman inten-
sity would be negligible and hence the accuracy of polymer
content determination is accurate. Further, the sole approxi-
mation in the Raman analysis is the assumed proportionality
between the ratio of Raman peak intensities and the ratio
of the concentration of the corresponding chemical species.
This approximation is related to the fact that the scattered
Raman intensity – apart from the phonon occupation number
dependence eliminated by the reduction scheme we use –
depends also on the Raman activity or Raman cross-section
of the vibrational mode in question. The activity of a
vibrational mode (i.e., the intensity of the corresponding
band) depends on the change of the polarizability during
the vibrational motion. Since the polarizabilities of sulfur
atoms in S8 and Sμ species have been experimentally found
to be comparable (see Ref. 1(a) for details), the assumed
proportionality between the ratio of Raman peak intensities
and the ratio of the concentration of S8 and Sμ species is a
valid hypothesis which does not add practically important
errors in the analysis. In summary, the validity of the analysis
FIG. 3. Representative fitting example of Raman spectra, (a) isotropic and
of the reduced isotropic Raman spectra used also here is tes- (b) anisotropic, restricted in the spectral range of S–S stretching modes.
tified from the correct determination of the change of sulfur’s (c) The depolarization ratio, ρ, of the high frequency Raman spectrum.
124501-5 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

bands in this spectral region are the symmetric S–S bond process as unstable recombination products due to cy-
stretching of the S8 ring molecule at ∼474 cm−1 and the clization of corresponding chain fragments, thus being
symmetric S–S bond stretching of the polymeric chains unstable recombination products. Further, these species
which appears as shoulder of the former peak situated at are known to be unstable above ∼240 K (Ref. 1(b)) a
∼470 cm−1 , denoted as Sμb . Both modes are polarized, al- fact that is confirmed in the present work since their
beit with minor differences, as shown in Fig. 3(c). How- peak intensities in the Raman spectra of sulfur glass
ever, a careful inspection of the polarized spectrum and gradually decrease for temperatures above ∼243 K and
its comparison with the depolarized one reveals the exis- finally they are not observable at 273 K. Because of
tence of a third Raman peak, at ∼462 cm−1 , denoted as their initial low concentration and their instability, their
Sμa positioned between the two aforementioned modes. presence does not affect the arguments of this paper
The origin of this Raman mode will be discussed in and thus they will be neglected throughout the rest of
Sec. IV D. The visibility of this new mode in the present the manuscript. Working in the isotropic form of the
case arises from the better resolved Raman peaks in the spectra we also cancel out contributions from small
glassy states as compared with the spectra of the liquid ring molecules (S6 , S7 ) as has been discussed in detail
phase. elsewhere.4(b)
(ii) Intramolecular bond bending region: 100-300 cm−1 . The (iii) Intermolecular modes: 0-100 cm−1 . The bands located
dominant vibrational lines present in this spectral range below ∼100 cm−1 are characteristic of other internal vi-
are Raman modes of the S8 molecule. In particular, the brational modes of the S8 unit as well as of librational
bands located at ∼152 and ∼221 cm−1 modes are as- motions of this molecule. While in the Raman spectrum
signed to the antisymmetric bond-bending and the sym- of the crystal they may be resolved, in the glass spectrum
metric bond-bending modes of the S8 ring molecule, they merge into two, a rather featureless broad band at
respectively. Both Raman bands are composite consist- ∼55 cm−1 and a mode located at ∼90 cm−1 at 93 K.
ing of various vibrational modes. Apart from the strong In addition, the ubiquitous feature of non-crystalline
bands of S8 rings, weak bands assigned to S7 species1(b) phases, the Boson peak, is also clearly resolved at very
can also be identified at ∼360, 410, and 520 cm−1 . low frequency (∼23 cm−1 ) in both the polarized and de-
Bands of S7 species are absent from the Raman spec- polarized spectra. All the spectral features resolved in the
tra of the liquid phase at 474 K.4 Therefore, it is pos- Raman spectra of sulfur glass at 93 K are summarized in
sible that S7 ring molecules form during the quenching Table I.

TABLE I. Assignment of the vibrational bands observed in the Raman spectra of glassy sulfur along with the respective thermal coefficients. The corresponding
modes and thermal coefficients of α-S8 crystal are also given for comparison.

α-S8 crystal peak position α-S8 crystal S glass peak position S glass
Assignment (T = 0) ν [cm−1 ]a dν/dT [cm−1 /K]a (T = 93 K) [cm−1 ] dν/dT [cm−1 /K]

Boson peak 23.0 −0.073 ± 0.008

Librations and 31.4 − 0.020 ∼55 −0.038 ± 0.004
translations 45.1 − 0.012
54 − 0.020
60.4 − 0.025
64.3 − 0.011
69.7 − 0.027
(ν 9 ) 91.2 − 0.014 88.5 −0.027 ± 0.002
85.8 − 0.012
79 − 0.005
(ν 8 ) 148 − 0.003 7 152.1 −0.007 5 ± 0.000 9
151.6 − 0.003 2
156.5 − 0.012 7
158.6 − 0.009 7
(ν 6 ) 183.2 − 0.002 27 194.8 −0.025 ± 0.002
189 − 0.007 6
199.7 − 0.014 8
(ν 2 ) 214.4 − 0.000 96 221.0 −0.007 1 ± 0.000 8
218 − 0.000 42
(ν 4 ) 234.4 − 0.000 74 235.2 −0.013 ± 0.003
(ν 11 ) 246.4 − 0.005 242.1 −0.004 5 ± 0.002 5
250.8 − 0.011 249.1 −0.009 6 ± 0.003 4
Sμa stretching 461.5 0.004 6 ± 0.003
Sμb stretching and (ν 7 ) 470.7 − 0.009 1 469.9 0.001 5 ± 0.001
(ν 1 ) 476.6 − 0.012 475.6 −0.007 2 ± 0.000 5
a
Data from Ref. 19(a).
124501-6 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

is given by19(b)
 
d ν̃ ∂ ν̃ α d ν̃
= − , (3)
dT ∂T V β dP
where α is the material’s coefficient of volume thermal expan-
sion, β is its compressibility, and d ν̃/dP denotes the variation
of the vibrational mode with pressure. Equation (3) states that
the variation of a mode with temperature can be separated
into two components an “explicit” or dynamic one caused by
the effect of temperature on a constant volume system and
an “implicit” or static one caused by the system’s volume
change. A complete Raman study of crystalline orthorhom-
bic sulfur (α-S8 ) and the variation of its vibrational modes
with temperature as well as pressure19(a) revealed that the ther-
mal coefficients of almost all crystal’s internal modes possess
values between −2.3 × 10−3 and −1.5 × 10−2 cm−1 K−1 .
Exceptions are the ν 2 and ν 4 modes for which the correspond-
ing values were around −7 × 10−4 cm−1 K−1 . The corre-
sponding values of the external (lattice modes) modes were
slightly higher (1.1–2.7 × 10−2 cm−1 K−1 ). The fact that
temperature had no observable effect on the energy of the
ν 2 mode was explained by suggesting that the anharmonic-
ities due to phonon-phonon interactions and the thermal ex-
pansion compensated for one another.19(a) Besides, by taking
into account that the shift due to thermal expansion is nega-
tive, the authors argued that for this particular mode the shift
due to anharmonic phonon interactions had to be equal and
opposite in sign.19(a) Whereas a discrepancy in the temper-
ature dependence of the frequency of the ν 2 mode with re-
spect to the other internal modes does indeed exist in α-S8
crystal, this is not the case for g-S. The thermal variation of
this mode in the glass follows the variation of the other vibra-
tional modes (Fig. 4) which is roughly equal to the crystal
one. Thus, the cause responsible for the compensation be-
tween implicit and explicit terms in Eq. (3) does not exist
FIG. 4. (a) Temperature dependence of the ν 1 , ν 2 modes of S8 rings as well in the case of glass and the explicit term contribution dimin-
as of the modes related to Sμa and Sμb species. (b) Ratio of the thermal coeffi-
cients of glassy and a-S8 sulfur for representative S8 normal modes.
ishes compared to the contribution of implicit term. The above
discussion focuses on the comparison of the crystal and the
glassy sulfur with respect to the temperature dependence of
the S8 vibrational modes in either case. The extracted val-
ues of (d ν̃/dT )glass for both the additional vibrational bands,
Sμa and Sμb , appearing in the glassy sulfur spectrum are non-
C. Temperature effects of the Raman bands in g-S
negative in opposition to those of the vibrational modes of S8
The temperature dependence of all vibrational modes ob- species.
served in the Raman spectrum of sulfur glass is compiled in Having provided a brief description of the origin and tem-
Table I. The corresponding temperature dependencies of crys- perature dependence of the main vibrational modes of g-S in
talline orthorhombic sulfur19(a) (α-S8 ) are also listed for com- comparison to the crystalline form we turn our attention to
parison. The influence of temperature on the Raman modes the issue of the polymer content in g-S and its temperature
of the glass and the crystal is graphically illustrated in Fig. 4. dependence.
With respect to the values of this ratio, three spectral regions
can be identified: (i) the low wavenumber region where the
D. The structure of g-S: Ab-initio-assisted
glass thermal coefficients are greater than those of the corre-
interpretation of vibrational modes
sponding crystal by a factor of two; (ii) the region of the ν 2
normal modes, located at ∼220 cm−1 where (d ν̃/dT )glass is As stated briefly in the introductory part of this paper,
very high; and (iii) the high wavenumber region where the the determination of the temperature dependence of the poly-
thermal coefficients of the glass are systematically lower than meric content φ(T) in liquid S at T > Tλ is a long standing
the corresponding ones of the crystal. issue in studies of elemental sulfur. φ(T) denotes the ratio
The variation of the frequency of a vibrational mode with of the number of S atoms participating in polymer chains
respect to temperature is determined by its anharmonicity and over the total number of S atoms in the sample. A detailed
124501-7 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

discussion of Fig. 3. In order to obtain an accurate fitting

result an extra line has to be incorporated in the spectral
range of the polymeric chains. There are therefore two Ra-
man lines associated with the vibrational modes of chains
at ∼462 (Sμa ) and ∼470 cm−1 (Sμb ). Evidence about a com-
mon origin of these two lines is provided by their similar
depolarization ratios as shown in Fig. 3(c). The frequency
of the Raman peak associated to Sμb can be more precisely
extracted by fitting the reduced anisotropic spectrum, see
Fig. 3(b). As is evident from the Raman spectra shown in
Figs. 1 and 2, ring molecules other than S8 are practically in-
visible in the Raman spectra. Since their formation is not ev-
ident under the conditions of the present experiment, Raman
modes of such molecules do not interfere in the proposed fit-
ting analysis. Further, even if non-S8 rings were present in the
glass at small concentrations, the isotropic form of the Raman
spectra used here to perform data analysis ensures that their
weak bands do not affect the results for reasons discussed
elsewhere.4(b) Alternatively, we suggest that these two Raman
lines represent vibrational modes of two types of Sn chains,
FIG. 5. Reduced isotropic intensity of the S–S bond stretching spectral re- i.e., low molecular weight Sn species (n < ncr ) (oligomers)
gion for the parent liquid (solid red curve) at 473 K and the quenched prod- and high molecular weight Sn species (n > ncr ) (polymers).
uct (dashed blue curve) at 93 K. The upper spectrum (dashed-dotted green Our interpretation is in agreement with experimental investi-
curve) is the theoretically calculated spectrum representing the vibrational
gations of the structure of glassy sulfur carried out with neu-
modes of a mixture of short and long chains and S8 units (see text for
details). tron diffraction experiments15(a) where it was shown that the
diffraction pattern is different between the two states, i.e.,
the melt (240 ◦ C) and the quenched product. Further neu-
reference to the existing ex situ and in situ methods for the tron diffraction studies have also demonstrated structural dif-
determination of the polymer extent has been given in Ref. 4. ferences between the liquid and the quenched solid.15(b) The
Ex situ approaches, widely known as quench-and-dissolution structural differences have been considered to arise from the
methods, are based on the reasonable assumption that fast existence of broken bonds in the quenched state. These exper-
cooling of the melt can lead to the formation of a solid with imental findings unequivocally show that fragmentation of the
the same polymer content as that of the parent liquid. How- large polymer chains takes place during the quenching proce-
ever, the subsequent treatment with various solvents as well as dure and thus short polymeric species with different vibra-
the storage at ambient temperature influence considerably the tional properties can form.
breakage of long polymer chains and the reduction of the es- The question to be answered is “what is the critical value
timated polymer content.14(c) In our case, the quenching of S of ncr above which a chain behaves as a polymer in the case of
inside the ampoule and its immediate storage at temperatures Raman scattering?” This issue relates to the fact that changes
much lower than Tg made it possible to carry out a more re- in S–S bonding occur as the length of the chain increases and
liable comparison of the polymer content between the parent these changes would affect the frequency of the S–S vibra-
liquid and the quenched non-crystalline solid. tional mode. To answer this question and to assist the above-
A comparison between the Raman spectrum of liquid S at mentioned hypothesis about the existence of short and long
473 K and its quenched product at 93 K is shown in Fig. 5. It (Sμb ) chains in glassy sulfur we performed ab initio calcula-
is obvious that g-S at 93 K preserves approximately the struc- tions of the crown-shaped S8 molecule and several di-radical
ture of the melt from which it was produced. The spectrum helical sulfur chains S8k for k = 1, 2, . . . , 7. For each chain, ge-
of the liquid is broader due to thermal effects. The polymer ometry optimization was followed by vibrational analysis, in-
content of liquid S at 473 K which has already been deter- cluding calculation of vibrational frequencies, depolarization
mined elsewhere4 by in situ Raman experiments is φ(473 K) ratios, and Cartesian displacements for each of the vibrational
≈ 0.45.20 Therefore, one expects that the polymeric frac- modes.
tion of the corresponding glass quenched sufficiently fast Before presenting our results it would be instructive to
from 473 K would be comparable with the polymer content mention that a number of publications have appeared dur-
in the liquid. However, since the quenching rate is not in- ing the last two decades regarding theoretical studies of sul-
finitely fast, partial depolymerization may occur. It is rea- fur clusters, mainly dealing with sulfur clusters up to 20
sonable that the higher the quenching rate, the closer the atoms and particularly focusing on structural rather than vi-
polymeric content of the quenched sample to the respec- brational properties.21 A wide variety of methods were ap-
tive one of the melt would be. The spectral range of the plied including semi-empirical,21(a) ab initio and high level
characteristic S–S stretching bands in the reduced isotropic ab initio,21(b)–21(h) Monte Carlo,21(i) and molecular dynam-
Raman spectrum of the glass cannot be satisfactorily fitted ics simulations.21(j) The geometric structures of the clusters
by two Lorentzian line profiles, as mentioned above in the studied are mostly rings. Open (chain) structures are rare
124501-8 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

FIG. 6. Structural parameters, i.e., bond lengths, bond angles, and dihedral angles, of the helical S8k (k = 1, 2, . . . , 7) chains as calculated at the HF/cc-pVDZ
level. Occurrence of values is depicted instead of actual values.

but when studied, triplets are more stable and more open
than their singlet counterparts21(i) . Interactions between sul-
fur clusters were studied with density functional and Monte
Carlo techniques,21(i) as, for example, the equilibrium poly-
merization of liquid sulfur.21(k)
Figure 6 compiles the structural parameters (bond
lengths, bond angles, and dihedral angles) obtained from the
geometry optimization of the helical chains. For the needs
of the present study we concentrate on the breathing mode
of the chain. This mode was selected based on two main cri-
teria, namely, the magnitude of the depolarization ratio and
the accompanying displacement vectors. In Fig. 7(a), the dis-
placement vectors for the breathing mode of the S24 chain
are shown as an example. The theoretical harmonic vibra-
tional frequencies calculated for various S8k chains, using the
HF/cc-pVDZ level of theory, are depicted in Fig. 7(b). This
diagram depicts the bond stretching vibrational frequency
changes along with the chain length. With increasing k the fre-
quency increases systematically from the value of the S8 chain
towards the frequency of the S8 ring molecule. The increase
saturates at some certain length k* = 2, above which the fre-
quency remains practically constant irrespective of the chain
length. Therefore, sulfur chains with k ≥ 2, i.e., S16 chains
and longer ones, can be considered as the long polymer con-
tent while shorter chains represent the oligomeric, non-cyclic
species in the glass and the liquid. The theoretical vibrational
spectrum, shown in Fig. 5, was calculated from the computed
harmonic frequencies using Lorentzian line broadening. Fair
agreement between the calculated and experimental spectra FIG. 7. (a) Displacement vectors for the breathing mode of the S24 chain.
(b) Variation of the frequency of the breathing mode of helical sulfur chains
is observed. Theoretical frequencies depicted in Fig. 7(b) are of various lengths, calculated at the HF/cc-pVDZ level. The frequency of the
scaled by a factor of 0.93. Using this value, the calculated S8 ring is given for comparison.
124501-9 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

symmetric stretch of the crown shaped S8 ring at the HF/cc-

pVDZ level is normalized to 474 cm−1 .
A comment should be made here on the fact that calcu-
lations are made for helical chains while the glass chains ex-
ist in disordered conformations. It should be remarked that
ab initio calculations are not employed in order to accurately
determine the absolute values of the chains vibrations. In-
stead, they were rather used as an indicator to understand
how the frequencies depend on chain length and this trend
will not depend on the chain conformation. Although the
absolute frequencies may indeed depend upon chain length,
this does not affect our interpretation. In addition, the vibra-
tional frequencies of ordered (trigonal-like) and disordered
(melt/glass) chains differ only by ∼3% (i.e., ∼450 cm−1 for
ordered chains and ∼465 cm−1 for disordered chains). This
means that even the absolute value of ordered and disordered
chains depends very little on chain conformation. FIG. 9. “Phase diagram” of the polymer content of elemental sulfur. Data
at T > Tλ are from Ref. 4(a). The polymer content at Tq is slightly higher
than the polymer content of the quenched product. At temperatures below
243 K φ(T) maintains a constant value of ∼0.43, while decrease of the poly-
E. Structure metastability of supercooled S: mer content commences above Tg .
Chains ↔ rings transformation
Having discussed the structural origin of the vibrational
modes in g-S we proceed now to the study of the temperature effect observed here demonstrates the metastability of the
dependence of the glass and supercooled liquid structure. g-S structure. As long as the structure is frozen at T < Tg
Figure 8 shows a collection of the reduced isotropic Raman there are no structural changes. When diffusion processes
spectra of sulfur recorded at temperatures spanning the glassy are activated at T > Tg , the thermodynamically less stable
and the supercooled regime. For the lowest temperature, sulfur chains depolymerize reverting back to the more stable
i.e., from 93 K to 243 K, representing the glassy state, S8 rings. This important finding shows that the ex situ
the spectra exhibit a perfect coincidence over the whole (quench-and-dissolution) methods employed for the polymer
wavenumber range, indicating the absence of structural fraction φ(T) determination are not as accurate as the in situ
changes in the glass. On the contrary, the polymeric content spectroscopic methods.4 Figure 9 presents the complete cycle
seems to decrease in a systematic and directional way at of the monomer-to-polymer transformation of elemental
temperatures higher than 243 K. It is not surprising that sulfur over a very broad temperature range including the
this temperature matches the Tg of g-S. This unprecedented glass, the supercooled liquid, and the melt. Notice that the
polymer fraction at Tq = 473 K obtained from in situ high
temperature Raman spectra is 0.45,4(a) while the respective
fraction obtained from the quenched product (from this Tq )
sulfur sample obtained by in situ low temperature Raman
spectra is 0.42. This finding not only indicates consistency of
the analysis procedure, but also shows that minimal structural
changes occur during the quenching process. The difference
in the polymer fraction of the parent mother liquid and
the glass appear to be within the error bar of the analysis.
We should stress that even though the polymer fraction is
maintained during our quenching process, the molecular
weight distribution may have been altered; the latter cannot
be directly observed by Raman spectroscopy. Solid squares
in Fig. 9 represent part of the data concerning the polymer
content at T > Tλ ; the full curve was shown in Ref. 4(c).
These structural changes, i.e., bond interchanges between
chains and rings, are expected to generate sub-Tg relaxation
processes. Indeed, such a secondary relaxation process has
already been hypothesized by Eisenberg and Teter13(c) in their
seminal studies on the mechanical relaxation properties of
glassy sulfur and sulfur-rich As-S alloys. They reached to the
conclusion that both simple diffusional motion and bond in-
FIG. 8. Raman spectra of the spectral range representative of the S–S stretch-
terchange take place in glassy sulfur. S–S bond interchange
ing modes obtained at various temperatures. Intensities are normalized with
respect to the maximum of the S8 band. For visual inspection the spectra have in small segments of the polymeric chains was invoked as
been shifted (along the x-axis) so that the S8 peak maxima coincide. the structural mechanism. Therefore, Eisenberg and Teter
124501-10 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

considered that the viscosity, η, of the medium has a contri-

bution from both processes, the molecular flow (process α)
and the bond interchange mechanism (process χ ), exempli-
fied as 1 / η = 1 / ηα + 1 / ηχ .13(c) For a long time, the origin
of the process χ was speculative. As our Raman data show,
the bond interchange is related to the S8 ↔ Sn transformation
which becomes active at T > Tg . The accurate description of
the relaxation mechanism in glassy and supercooled sulfur is
essential for the correct determination of the fragility of the
supercooled liquid; for a discussion about the fragility of sul-
fur see Ref. 10(b).

F. The Boson peak of g-S: Extraordinary

temperature dependence
All non-crystalline substances exhibit ubiquitously in the FIG. 11. Temperature dependence of the Boson peak maximum, BP .

low frequency part of their spectrum the Boson peak. This

peak represents the low energy excitations of the glass ac- isolate the function g(ν̃)/ν̃ 2 . It should be noticed that the Bo-
counting for the difference in vibrational modes between the son peak, or equivalently the peak in the g(ν̃)/ν̃ 2 representa-
glass and the crystal at sufficiently low energies, i.e., in the tion, does not correspond to a real, strong peak in the VDoS,
interval 0.3–2 THz (1 THz = ˆ 33.3 cm−1 = ˆ 4.13 meV).22(a) rather it appears in VDoS as a bump of negligible intensity at
These excitations are manifested as non-Debye scatter- considerably higher energy. The use of the frequency-reduced
ing mechanisms in inelastic neutron and Raman scattering. representation, Eq. (4), has several shortcomings that have
Figure 10 shows representative reduced, low-frequency Ra- been summarized elsewhere.22(b) More specifically, analyzing
man spectra of g-S at selected temperatures in the glassy state. a number of experimental data for several glasses it was found
Spectra reduction has been carried out using the following that the spectral changes of the Boson peak under the applica-
equation: tion of certain external stimuli, such as temperature, pressure,
I exp t (ν̃) g(ν̃)C(ν̃) etc., can be understood by considering the corresponding de-
I red (ν̃) = ∝ , (4) pendence of the true excess of the vibrational density of states,
ν̃[n(ν̃, T ) + 1] ν̃ 2
i.e., g(ν) − gDebye (ν).
where g(ν̃) the vibrational density of states (VDoS) and C(ν̃) The temperature dependence of the BP frequency, BP ,
is the Raman coupling coefficient. Equation (4) is a partic- appears to follow an almost linear relation with a nega-
ular expression of the reduced spectrum frequently used to tive slope, providing a BP anharmonicity, d BP /dT ≈ −6.6
× 10−2 cm−1 K−1 , as illustrated in Fig. 11. This is very high
in comparison to corresponding values of BP anharmonici-
ties of other typical glass-formers,23 which exhibit a mild T-
dependence with negative slope of the BP for temperatures
up to Tg , and a strong dependence at higher temperatures.
The Boson peak frequency decreases by a factor of two over
the temperature range 93–213 K, which is still well below
Tg . Unfortunately, the intervention of crystallization slightly
above Tg and the onset of strong quasi-elastic light scatter-
ing mechanisms do not allow the resolvability of the Boson
peak in the supercooled regime. It is not easy to account for
the peculiarly strong temperature dependence of BP of g-S.
At T > Tg the thermal expansion of the supercooled liquid is
usually adopted to explain the BP frequency softening via a
Grüneisen-type mechanism; however, this mechanism cannot
be considered for T < Tg . The structural heterogeneity in the
nanoscale could possibly be the origin of the intense sub-Tg
BP softening.

V. CONCLUSIONS
Vibrational modes, structure, and species metastability
of elemental sulfur in the glassy and the supercooled state
where investigated in detail using polarized Raman scatter-
FIG. 10. Low frequency depolarized Raman spectra of glassy sulfur as a ing. Molten sulfur was quenched from 473 K where the
function of temperature, showing the evolution of the Boson peak. liquid contains moderate polymer content. The temperature
 124501-11 Andrikopoulos et al. J. Chem. Phys. 139, 124501 (2013)

dependence of the main vibrational modes of the glass were 7 J. Dudowicz, K. F. Freed, and J. F. Douglas, J. Chem. Phys. 113, 434
determined; their comparison with the corresponding anhar- (2000).
8 P. Ballone and R. O. Jones, J. Chem. Phys. 119, 8704 (2003).
monicities of the crystal showed that the compensation of the 9 T. Scopigno, S. N. Yannopoulos, F. Scarponi, K. S. Andrikopoulos, D.
anharmonicities due to phonon-phonon interactions and the Fioretto, and G. Ruocco, Phys. Rev. Lett. 99, 025701 (2007).
10 (a) G. Monaco, L. Crapanzano, R. Bellissent, W. Crichton, D. Fioretto,
thermal expansion, existing in the crystal for the ν 2 mode, is
altered in the glass resulting in negative values for the ther- M. Mezouar, F. Scarponi, and R. Verbeni, Phys. Rev. Lett. 95, 255502
(2005); (b) B. Ruta, G. Monaco, V. M. Giordano, F. Scarponi, D. Fioretto,
mal coefficients similar to the remaining S8 modes participat- G. Ruocco, K. S. Andrikopoulos, and S. N. Yannopoulos, J. Phys. Chem.
ing in the glass structure. The thermal coefficients of the vi- B 115, 14052 (2011).
11 (a) F. Kyriazis and S. N. Yannopoulos, Appl. Phys. Lett. 94, 101901 (2009);
brational modes associated to the polymeric fragments of the
sulfur glass encompass non-negative values. Analysis of the (b) S. N. Yannopoulos, F. Kyriazis, and I. P. Chochliouros, Opt. Lett. 36,
534 (2011).
Raman spectra showed that the polymer content is preserved 12 (a) A. Smith, W. B. Holmes, and E. S. Hall, J. Am. Chem. Soc. 27, 797
fairly well in the quenched glass. However, the absence of (1905); (b) A. Smith and W. B. Holmes, ibid. 27, 979 (1905); (c) H. R.
thermal broadening effects, deep in the glassy state, allowed Kruyt, Z. Phys. Chem. 64, 513 (1908).
13 (a) A. V. Tobolsky, W. MacKnight, R. B. Beevers, and V. D. Gupta, Poly-
for the resolution of two types of polymer content, which were
mer 4, 423 (1963); (b) G. W. Miller, J. Appl. Polym. Sci. 15, 1985 (1971)
hypothesized to represent short and long polymer chains. Ab (c) A. Eisenberg and L. A. Teter, J. Phys. Chem. 71, 2332 (1967).
initio calculations were employed to support the interpreta- 14 (a) S. R. Das and K. Gosh, Indian J. Phys. 13, 91 (1939); (b) J. A. Prins, J.

tion based on the hypothesis of short and long polymeric Schenk, and L. H. Watchers, Physica 23, 746 (1957); (c) J. Schenk, Physica
23, 546 (1957).
species, which is in agreement with neutron diffraction data 15 (a) R. Winter, P. A. Egelstaff, W. C. Pilgrim, and W. S. Howells, J. Phys.:
of quenched sulfur. The metastability of the polymeric con- Condens. Matter 2, SA215 (1990); (b) M. Stolz, R. Winter, W. S. Howells,
tent was investigated in detail by comparing Raman spectra R. L. McGreevy, and P. A. Egelstaff, ibid. 6, 3619 (1994).
16 A. G. Kalampounias, D. Th. Kastrissios, and S. N. Yannopoulos, J.
recorded at closely spaced temperature intervals. It has been
Non-Cryst. Solids, 326-327, 115 (2003).
revealed that heating the glass above its glass transition tem- 17 M. J. Frisch, G. W. Trucks, H. B. Schlegel et al., Gaussian 09, Revision
perature is the cause that forces chains to the more thermo- B.1, Gaussian, Inc., Wallingford, CT, 2009.
18 (a) T. Yanai, D. Tew, and N. Handy, Chem. Phys. Lett. 393, 51 (2004);
dynamically stable S8 rings. Therefore, apart from the typical
atomic diffusion for all liquids in the supercooled regime as- (b) A. D. Becke, J. Chem. Phys. 98, 5648 (1993); (c) 0.5×EXHF
+ 0.5×EXLSDA + ECLYP; (d) 0.5×EXHF + 0.5×EXLSDA
sociated with the structural relaxation, the chain-to-ring bond + 0.5× EXBecke88 + ECLYP; (e) W. J. Pietro, M. M. Francl, W.
interchange mechanism in supercooled elemental sulfur signi- J. Hehre, D. J. Defrees, J. A. Pople, and J. S. Binkley, J. Am. Chem. Soc.
fies the onset of a secondary relaxation mechanism. This sec- 104, 5039 (1982); (f) D. E. Woon and T. H. Dunning, Jr., J. Chem. Phys.
ondary relaxation has to be considered as an important factor 98, 1358 (1993); (g) A. K. Rappé, T. Smedly, and W. A. Goddard III, J.
Phys. Chem. 85, 1662 (1981).
which complicates in the accurate estimation of the fragility 19 (a) B. Eckert, H. O. Albert, H. J. Jodl, and P. Foggi, J. Phys. Chem. 100,
of supercooled sulfur. It is still not clear if this relaxation is 8212 (1996); (b) R. Zallen and M. L. Slade, Phys. Rev. B 18, 5775 (1978).
20 Its determination took place by fitting the reduced isotropic Raman spec-
also responsible for the unexpectedly strong sub-Tg softening
tra of the liquid using two Lorentzian peaks for the polymer and monomer
of the Boson peak frequency.
contents. For better accuracy the fitting was performed over the region 300–
600 cm−1 taking into account two additional weak vibrational lines at ener-
gies below 420 cm−1 . This was indispensable for the increased background
ACKNOWLEDGMENTS of the liquid S spectra at the low wavenumber side of the polymer peak. In-
specting the low temperature Raman spectrum of g-S at 93 K it becomes
The authors would like to thank Dr. G. A. Voyiatzis and obvious that more than one peaks are hidden below the profile of the Raman
Professor G. N. Papatheodorou for providing experimental fa- band. The obvious fine structure is now evident due to the Raman peaks nar-
cilities and useful discussions. O.F. would like to thank Pro- rowing at this low temperature, whilst thermal effects at high temperature
fessor V. S. Melissas (University of Ioannina) for providing contribute to bands broadening, thus obscuring the existence of the various
contributions.
computational facilities and for useful discussions. 21 (a) H. Yilmaz and S. Erkoc, J. Mol. Struct.: THEOCHEM 231, 63 (1991);

(b) D. Dixon and E. Wasserman, J. Phys. Chem. 94, 5772 (1990); (c) K.
1 (a) B. Meyer, Chem. Rev. 76, 367 (1976); (b) R. Steudel and B. Eckert, Raghavachari, C. Rohlfing, and J. S. Binkley, J. Chem. Phys. 93, 5862
Top. Curr. Chem. 230, 1 (2003); (c) R. Steudel, ibid. 230, 81 (2003); (d) (1990); (d) J. Cioslowski, A. Szarecka, and D. Moncrieff, J. Phys. Chem. A
R. Steudel and H.-J. Mäusle, Z. Anorg. Allg. Chem. 478, 156 (1981). 105, 501 (2001); (e) M. Chen, M. Liu, H. Luo, Q. Zhang, and C. Au, J. Mol.
2 (a) S. C. Greer, J. Phys. Chem. B 102, 5413 (1998); (b) Ann. Rev. Phys.
Struct.: THEOCEM 548, 133 (2001); (f) S. Millefiori and A. Alparone, J.
Chem. 53, 173 (2002). Phys. Chem. A 105, 9489 (2001); (g) M. Wong, Y. Steudel, and R. Steudel,
3 V. F. Kozhevnikov, J. M. Viner, and P. C. Taylor, Phys. Rev. B 64, 214109
Chem. Phys. Lett. 364, 387 (2002); (h) R. Ludwig, J. Behler, B. Klink, and
(2001); V. F. Kozhevnikov, W. B. Payne, J. K. Olson, C. L. McDonald, and F. Weinhold, Angew. Chem. Int. Ed. 41, 3199 (2002); (i) R. Jones and P.
C. E. Inglefield, J. Chem. Phys. 121, 7379 (2004). Ballone, J. Chem. Phys. 118, 9257 (2003); (j) D. Hohl, R. Jones, R. Car,
4 (a) A. G. Kalampounias, K. S. Andrikopoulos, and S. N. Yannopoulos, J.
and M. Parrinello, ibid. 89, 6823 (1988); (k) P. Ballone and R. Jones, ibid.
Chem. Phys. 118, 8460 (2003); (b) S. N. Yannopoulos, K. S. Andrikopou- 119, 8704 (2003).
los, and A. G. Kalampounias, ibid. 121, 6575 (2004); (c) K. S. Andrikopou- 22 (a) J. Jäckle, in Amorphous Solids: Low-Temperature Properties, edited
los, A. G. Kalampounias, and S. N. Yannopoulos, ibid. 124, 146101 (2006); by W. A. Phillips (Springer, Berlin, 1981), p. 135; (b) S. N. Yannopou-
(d) R. Steudel and B. Eckert, ibid. 121, 6573 (2004). los, K. S. Andrikopoulos, and G. Ruocco, J. Non-Cryst. Solids 352, 4541
5 (a) A. G. Kalampounias, K. S. Andrikopoulos, and S. N. Yannopoulos, J.
(2006).
Chem. Phys. 119, 7543 (2003); (b) K. S. Andrikopoulos, A. G. Kalampou- 23 S. N. Yannopoulos and K. S. Andrikopoulos, J. Chem. Phys. 121, 4747
nias, and S. N. Yannopoulos, Soft Matter 7, 3404 (2011); (c) F. Begum, R. (2004); S. N. Yannopoulos, A. G. Kalampounias, A. Chrissanthopou-
H. Sarker, and S. L. Simon, J. Phys. Chem. B 117, 3911 (2013). los, and G. N. Papatheodorou, ibid. 118, 3197 (2003); A. G. Kalampou-
6 K. S. Andrikopoulos, A. G. Kalampounias, and S. N. Yannopoulos, Phys.
nias, S. N. Yannopoulos, and G. N. Papatheodorou, ibid. 125, 164502
Rev. B 72, 014203 (2005). (2006).

View publication stats