331
§
E-mail address: paris@camars.unicam.it
The Canadian Mineralogist
Vol. 39, pp. 331-339 (2001)
THE VALENCE AND SPECIATION OF SULFUR IN GLASSES
BY X-RAY ABSORPTION SPECTROSCOPY
ELEONORA PARIS
§
, GABRIELE GIULI AND MICHAEL R. CARROLL
Dipartimento di Scienze della Terra, Università di Camerino and INFM
(Istituto Nazionale Fisica della Materia) – Unità di Camerino, I-62032 Camerino, Italy
IVAN DAVOLI
Dipartimento di Fisica , Università di Roma “Tor Vergata”, and INFM
(Istituto Nazionale Fisica della Materia) – Unità di Roma 2, I-00133 Roma, Italy
ABSTRACT
The geochemical behavior of sulfur in magmas depends strongly on the oxidation state of sulfur, but this is not easily deter-
mined by standard analytical methods. We have measured XANES absorption spectra at the sulfur K-edge and have found that
such measurements are useful to characterize the oxidation state and speciation of sulfur in silicate glasses of geological rel-
evance. Measured spectra of a set of reference minerals show the effects of different oxidation states and coordination numbers
of sulfur; there is a large shift in energy (~10–12 eV) of the sulfur K-edge between S
2–
and S
6+
. This large and easily detectable
difference makes possible the measurement of the valence of sulfur in unknown samples by measuring the shift in energy of the
absorption edge. This approach is applicable to both crystalline and glassy materials, and useful results have been obtained on
samples with as little as 450 ppm S. We have used XANES measurements to characterize oxidation state and speciation of sulfur
in a set of natural and synthetic sulfur-bearing glasses. The samples cover a range of composition from basaltic to almost rhyolitic,
and some were synthesized over a range of pressure, temperature and oxygen fugacity; glass S content varies between 450 and
3000 ppm. XANES analyses, carried out in fluorescence mode at LURE, allowed determination of the sulfur oxidation state in all
of the samples and clearly show that some samples contain a mixture of S
2–
and S
6+
; no other sulfur species were observed.
Quantitative determination of the abundance of sulfide and sulfate shows good agreement with independent measurements based
on electron-microprobe determination of the wavelength shift of sulfur K␣ X-rays.
Keywords: sulfur, silicate glasses, X-ray absorption spectroscopy, XANES.
SOMMAIRE
Le comportement géochimique du soufre dans les magmas dépend fortement du taux d’oxydation du soufre, mais cet aspect
n’est pas facile à documenter par les méthodes analytiques habituelles. Nous avons mesuré les spectres d’absorption XANES du
soufre, en particulier du seuil d’absorption K, et nous avons trouvé ces mesures utiles pour caractériser le taux d’oxydation et la
spéciation du soufre dans les verres silicatés d’importance géologique. Des spectres de référence ont été mesurés sur certains
minéraux choisis pour montrer les taux d’oxydation et les coordinences distincts du soufre; il y a un décalage important en énergie
(~10–12 eV) du seuil d’absorption K entre le S
2–
et le S
6+
. Cette différence majeure et facilement décelable rend possible la
mesure de la valence du soufre dans des échantillons méconnus en examinant le déplacement de l’énergie associée au seuil de
l’absorption. Ces mesures s’appliquent également à des matériaux cristallins ou vitreux, et mènent à des résultats utiles même
dans des cas où la teneur en soufre est très faible, par exemple 450 ppm. Nous avons utilisé des mesures XANES pour caractériser
le taux d’oxydation et la spéciation du soufre dans des verres naturels et synthétiques. Les échantillons correspondent à un
intervalle de composition entre basaltique et presque rhyolitique, et certains ont été synthétisés sur un intervalle de pression, de
température et de fugacité d’oxygène; la teneur en S des verres varie entre 450 et 3000 ppm. Les analyses XANES, faites en mode
fluorescence à l’emplacement LURE, nous ont permis de déterminer le taux d’oxydation de tous les échantillons, et montrent
clairement que certains échantillons contiennent un mélange de S
2–
et de S
6+
; aucune autre valence de soufre n’a été observée.
Une détermination quantitative de l’abondance de sulfure et de sulfate montre une bonne concordance avec les mesures fondées
sur le décalage en longueur d’onde des rayons X K␣ du soufre, faites indépendemment avec une microsonde électronique.
(Traduit par la Rédaction)
Mots-clés: soufre, verres silicatés, spectroscopie d’absorption X, XANES.
331 39#2-avril-01-2189-07 3/07/01, 14:05331
332 THE CANADIAN MINERALOGIST
INTRODUCTION
An atomistic-scale understanding of the way sulfur
dissolves in silicate melts can help us to better under-
stand the geochemical behavior of sulfur in magmas.
For example, information about the parameters that af-
fect sulfur solubility in magmas is crucial in attempts to
use measurements of volcanic gas emissions to make
inferences about subsurface movements of magma and
probabilities of eruption. Sulfur is unusual among the
common magmatic volatiles because over the range of
typical magmatic oxidation states, the predominant form
of sulfur changes from S
2–
under reducing conditions to
SO
4
2–
under oxidizing conditions (e.g., Nagashima &
Katsura 1973, Katsura & Nagashima 1974, Carroll &
Rutherford 1988). The pattern of solubility of these two
sulfur species differs considerably (e.g., see review in
Carroll & Webster 1994). Thus an understanding of
variations in sulfur solubility and of volcanic degassing
of sulfur requires information concerning sulfur oxida-
tion state in magmas. In this paper, we show how X-ray
absorption spectroscopy (XAS) of S-bearing glasses can
provide information on the oxidation state of dissolved
sulfur. The samples studied had been previously used
to evaluate the oxidation state of sulfur by measuring
the wavelength shift of sulfur K␣ X-rays with the elec-
tron microprobe (Carroll & Rutherford 1988). More
importantly, we show that the XAS results are consis-
tent with the hypothesis that the S dissolved in silicate
melts (glasses) of geological interest occurs predomi-
nantly as SO
4
2–
and S
2–
(Carroll & Rutherford 1988),
with no evidence for significant quantities (<~5%) of
sulfur with an intermediate oxidation state.
BACKGROUND INFORMATION
In silicate melts, sulfur dissolves mainly as sulfide
under reducing condition and as sulfate under oxidizing
conditions. The relative abundance of these two species
in a melt is controlled by the equilibrium
S
2–
melt
+ 2O
2
gas
= (SO
4
)
2–
melt
(1)
with an equilibrium constant
K
1
= [SO
4
2–
] / [S
2–
] {f(O
2
)}
2
(2)
where f denotes fugacity and square brackets denote
activities. This expression indicates that a major factor
controlling sulfur speciation is oxygen fugacity, f(O
2
),
as long as the activity coefficients of (SO
4
)
2–
melt
and
S
2–
melt
, or their ratio, remain approximately constant.
Previous studies of the systematics of sulfur solubil-
ity provide information concerning the solution mecha-
nisms of sulfur in melts. For anhydrous systems, many
studies of melt compositions involved in commercial
processes such as steel production (e.g., Fincham &
Richardson 1954, Abraham et al. 1960a, b) show that
under reducing conditions, S dissolves as S
2–
, in a reac-
tion that can be considered as an exchange for O
2–
in
the melt
1/2 S
2
+ O
2–
= 1/2 O
2
+ S
2–
(3)
Since S
2–
is much larger than O
2–
, the oxygen atoms
available for exchange are likely those bound to net-
work-modifying cations, instead of those associated
with tetrahedrally coordinated Al
3+
or Si
4+
. This infer-
ence is supported by the observation that solubility of
sulfur is higher in more silica-poor melts, all other fac-
tors being equal (Fincham & Richardson 1954,
Abraham et al. 1960a, b). The positive correlation be-
tween sulfide solubility and FeO contents of the melt
suggests a strong affinity of S
2–
species with network-
modifying Fe
2+
cations (Haughton et al. 1974, Mathez
1976, Carroll & Rutherford 1985, Wallace & Carmichael
1992).
Under oxidizing conditions, such as those used in
commercial glass-making and in some subduction-re-
lated magmas (e.g., Pinatubo, El Chichón), sulfur dis-
solves mainly as sulfate, as indicated by the reaction
1/2 S
2
+ 3/2 O
2
+ O
2–
= (SO
4
)
2–
(4)
Sulfur solubility depends therefore on the fugacity of
oxygen f(O
2
), the fugacity of sulfur f(S
2
), and on the
activity of O
2–
anions in the melt. At intermediate oxi-
dation states, S can occur as sulfide and sulfate species,
with the relative proportions depending on oxidation state
(Carroll & Rutherford 1988, Wallace & Carmichael
1994, Nilsson & Peach 1993). Therefore, to understand
the behavior of sulfur in melts, we need information on
S speciation.
Sulfur speciation can be measured by wet-chemical
analysis, but the methods are not simple; a significant
mass of sample is required, and glass must be carefully
separated from crystals in order to get reliable informa-
tion on content and speciation of sulfur in the melt. The
electron microprobe can also be used to determine S
speciation by measuring the wavelength of the S K␣ X-
rays. This is done by scanning the crystal spectrometer
of the probe over the region of the S K␣ peak, which
shifts in wavelength by about 1.4 eV between S
2–
and
S
6+
; the peak shifts can typically be measured to a pre-
cision of ±10–15% (Carroll & Rutherford 1988, Wallace
& Carmichael 1992, 1994). It is possible to use the
measured shifts in wavelength to estimate the propor-
tions of reduced and oxidized sulfur species in natural
glasses if we assume that the S K␣ peak shifts linearly
with S oxidation state (Carroll & Rutherford 1985,
Wallace & Carmichael 1994, Nilsson & Peach 1993,
Matthews et al. 1999). This assumption is the key to
using measurements of wavelength shift to determine
sulfur speciation (or oxidation state of the magma), and
our new XANES measurements support the validity of
this assumption.
331 39#2-avril-01-2189-07 3/07/01, 14:05332
THE VALENCE AND SPECIATION OF SULFUR IN GLASSES 333
XAS ANALYSIS FOR SULFUR
In this paper, we address the topic of determination
of oxidation state and speciation of sulfur in silicate
glasses by XANES (X-ray absorption near-edge struc-
ture) spectroscopy. X-ray absorption spectroscopy
(XAS) is a non-destructive, element-specific technique
that is sensitive to the electronic structure, oxidation
state, coordination number and local bonding geometry
of a selected element, independent of the crystallinity
of the material under investigation, and down to low
concentrations of the absorber (Bianconi 1988). Al-
though this is the first application of XANES to the
study of sulfur in silicate glasses, X-ray absorption spec-
troscopy has been used previously to study S in a vari-
ety of natural materials, including minerals, coal,
petroleum and related hydrocarbons, following many
studies concerning the investigation of sulfur K-edge in
organic and inorganic materials (e.g., Spiro et al. 1984,
Huffman et al. 1995, Waldo et al. 1991). Li et al. (1995)
have recently reported results of a comprehensive study
of sulfur-bearing minerals with varying oxidation states
of sulfur. They showed that a significant energy-shift
occurs at the sulfur absorption K and L edges as a func-
tion of S oxidation state [as observed by Frank et al.
(1987) and George & Gorbaty (1989) in studies of or-
ganic materials]. This shift can be used to evaluate the
valence of unknown sulfur compounds, as the shift in
energy of the absorption edge can be related to the dif-
ference in electron density between different oxidation
states. The removal of a valence electron produces a
decrease of the screening of core electrons, which in turn
strengthens the core energy-levels. Therefore, an in-
crease of oxidation states produces an increase of the
binding energy of inner-shell 1s and 2p electrons and a
shift of the sulfur K-edge toward higher energy.
Studies of sulfide minerals by X-ray absorption spec-
troscopy (e.g., Li et al. 1994a, b, c, Charnock et al. 1990,
Sugiura 1981, Sugiura & Muramatsu 1985) show that
the edge shift depends not only on the binding energy
of the inner shells, but also on the final states (i.e., the
first unoccupied states). The electronic structure of 3d-
element sulfides affects the edge chemical shift, with
lower values for the compounds characterized by ele-
ments with partially occupied 3d orbitals (metal 3d – S
3s- and 3p-like states in the band gap) compared with
those with fully occupied 3d orbitals (metal 3d – S 3s-
and 3p-like states at the conduction-band minimum) (Li
et al. 1995). Values of the energy-band gap are also
positively correlated with the edge position.
XANES analysis has been applied successfully to the
determination of the forms of sulfur in coal, where sul-
fur is a critical element controlling industrial use, and
sulfur-removal techniques comprise a major research
topic (Spiro et al. 1984, Huffman et al. 1991). These
studies have shown that the oxidation state of sulfur in
coal ranges from sulfide to sulfate, and XAS has been
used to quantify the abundances of organic and inor-
ganic forms of sulfur (Huffman et al. 1991, 1995,
Premuzic et al. 1994).
XANES studies of sulfur in bitumens and asphal-
tenes (Kasray et al., 1994) and petroleums or petroleum
source-rocks have been used to identify and quantify
various compounds of sulfur (Waldo et al. 1991) and to
document the correlation between sulfur speciation and
depositional environment, thermal maturity and aroma-
ticity. The same technique has been used successfully to
investigate sulfur in biogenic and diagenetic carbonates;
such studies show that S in the sulfate form substitutes
for carbonate (Pingitore et al. 1995). Also, sulfonates
have been detected in some organic-matter-rich marine
sediments and found to be a novel and major compo-
nent of organic sulfur in sedimentary rocks (Vairava-
murthy et al. 1994).
METHODS AND MATERIALS
The S K-edge spectra have been collected at LURE,
Orsay (France) on the SA32 beam line of the Super–
ACO storage ring. This ring operates at 800 MeV with
an average stored current of 200 mA, the actual value
depending on the bunch mode (2 or 24). The SA32 beam
line uses an upstream toroidal mirror and a two-crystal
monochromator with an exit beam kept fixed by means
of a cam. In the present experiment, we used two Ge
(111) crystals, which allow a resolution of 0.9 eV at
3000 eV. The incoming flux of photons has been moni-
tored by the drain current from a 0.8 m aluminum foil,
and the absorption spectra, by the fluorescence yield
(FY) using a one-element germanium detector mounted
in the horizontal plane oriented at 90° to the incident X-
ray beam. The FY spectra have been obtained by moni-
toring the number of counts within the energy window
of the detector output centered on the S K emission line.
In the present work, no correction for self-absorption
was necessary because the dilute nature of the glass
samples produces a FY signal directly proportional to
the absorption coefficient.
The samples (~50 mg total) were prepared by finely
grinding glasses and crystal chips to powder and mount-
ing them on an Al plate (over an area of several mm
2
).
All of the ground glass samples were treated multiple
times in an ultrasonic bath filled with warm distilled
water to remove any possible trace of crystalline sulfate
[anhydrite, present as phenocrysts (< 2 vol.%) in some
of the experimental samples], and washed with warm
distilled water prior to drying for analysis. The high
solubility of anhydrite in water makes this an effective
procedure for removing crystalline sulfate; optical ex-
amination of the resulting powders verified the effi-
ciency of this treatment. The sample of back-arc basalt
glass (BAB) was not sulfide-saturated, so no treatment
of the glass was necessary. Experimental glasses pro-
duced at more reducing conditions could not be ana-
331 39#2-avril-01-2189-07 3/07/01, 14:05333
334 THE CANADIAN MINERALOGIST
lyzed because all had dispersed grains of FeS, and we
could not find an effective way to separate sulfide-free
glass from this material.
Energy calibration was carried out using crystalline
sulfur, routinely measured during collection of the spec-
tra. Standard data-reduction procedures were used to
extract the XANES signals: the background has been
subtracted from the spectra using a Victoreen function,
and the spectra have been normalized to the high-en-
ergy part of the spectrum. The energy position of the
edge has been determined from the first maximum of
the spectrum derivative. Fitting of the XANES spectra
was carried out using an arctangent baseline and mixed
lorentzian–gaussian peak functions, and the areas of the
peaks were used to quantify the proportions of sulfide
and sulfate.
Minerals containing sulfur in different oxidation
states have been used as reference compounds: they in-
clude native sulfur, gypsum, apatite, galena, pyrrhotite,
sphalerite. All the reference materials have been previ-
ously characterized by powder X-ray diffraction. De-
tails of the synthesis procedures of the experimental
glasses studied here are reported in Carroll & Ruther-
ford (1985), and essential data on composition, sulfur
content, and conditions of synthesis of the samples are
summarized in Table 1. The glass samples analyzed in
this work have a limited range of bulk compositions
(Table 1), and their total sulfur contents range between
450 and 3000 ppm, as determined by electron-micro-
probe analysis.
RESULTS AND DISCUSSION
Spectra of reference compounds
Measured XANES spectra showing the K-edge ab-
sorption of sulfur in sulfur-bearing minerals used as ref-
erence materials are shown in Figure 1. They cover a
range of oxidation states (from S
2–
to S
6+
) and coordi-
nation geometries. The overall shape of the spectra and
the energy position of S K-edge show significant vari-
ability among different types of samples. The number,
position and intensity of the spectral peaks depend
mainly on the number and kind of first neighbors around
the scattering element, but also on effects due to next-
nearest-neighbor atoms and their back-scattering power.
For example, in the spectra of galena (PbS) and sphaler-
ite (ZnS) (Fig. 1), the shape of the spectra and also the
energy position of the edge are affected by the different
structure of these species, coordination numbers (sulfur
is in 4-fold coordination in ZnS and in 6-fold coordina-
FIG. 1. Sulfur K-edge spectra of some S-bearing minerals
used as reference in a range of oxidation state (from S
2–
to
S
6+
). Edge position and spectral features change dramati-
cally with oxidation state, coordination number, local ge-
ometry around the absorber, and electronic structure.
331 39#2-avril-01-2189-07 3/07/01, 14:05334
THE VALENCE AND SPECIATION OF SULFUR IN GLASSES 335
tion in PbS), type of bond, and type of cation present.
Similar observations can be made for sulfate minerals
and to explain the differences observed between the
apatite and gypsum spectra.
The edge shifts we observe depend strongly on the
oxidation state and the electronic structure. For example,
among the spectra for the various minerals shown in
Figure 1, the S K-edge of sulfide minerals (PbS, ZnS,
Fe
1–x
S) lies at energy values near that for elemental S,
whereas the edge for gypsum and sulfate-bearing apa-
tite occurs at a higher energy (Table 2). The difference
in energy observed between the spectra of sulfide and
sulfate minerals is ~10–12 eV, and the position of the
edge can be evaluated precisely from first derivatives
of the fluorescence signal (Fig. 2). The data on the en-
ergy of the absorption edge for the standards versus S
oxidation state are shown in Figure 3, which can be used
to estimate the oxidation state of an unknown compound
by looking at the energy shift of the edge.
The spectra of silicate glasses
Sulfur XANES spectra of natural and synthetic
glasses with sulfur contents ranging from 450 to 3000
ppm (Table 1) are shown in Figure 4. Besides the
XANES results, sulfur speciation in all of the samples
(Table 1) has been estimated using the electron-micro-
probe-based peak-shift method described in Carroll &
Rutherford (1988). In spite of the low sulfur content of
FIG. 2. First-derivative spectra, used to evaluate the energy
position of the edge. Here, a comparison is shown between
the various reference materials, illustrating the large differ-
ence in edge position and the similarities of spectra for sam-
ples with the same oxidation state of sulfur.
FIG. 3. The relation between energy of the sulfur K-edge and
oxidation state of sulfur for a series of reference com-
pounds. The large difference in energy between sulfide and
sulfate (~12 eV) allows use of the energy of the absorption
edge to estimate the oxidation state of sulfur.
331 39#2-avril-01-2189-07 3/07/01, 14:05335
