S Kβ x-ray fluorescence spectra of the Tl2S-Sb2S3 chalcogenide system

Article abstract of DOI:10.1088/0953-8984/8/43/033

The sulphur 3p valence states of Sb₂S₃, Tl₃SbS₃, TlSbS₂, TlSb₃S₅, TlSb₅S₈, and Tl₂S of the Tl₂S-Sb₂S₃ system hav

Full text of DOI:10.1088/0953-8984/8/43/033

Home Search Collections Journals About Contact us My IOPscience
x-ray fluorescence spectra of the
-
chalcogenide system
This content has been downloaded from IOPscience. Please scroll down to see the full text.
1996 J. Phys.: Condens. Matter 8 8421
(http://iopscience.iop.org/0953-8984/8/43/033)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 155.69.4.4
This content was downloaded on 22/08/2015 at 04:16
Please note that terms and conditions apply.

J. Phys.: Condens. Matter 6 (1996) 8421–8428. Printed in the UK
S Kβ x-ray fluorescence spectra of the Tl2S–Sb2S3
chalcogenide system
S Dupont†, A Gheorghiu†, C Se´ne´maud†, J-M Mariot†, C F Hague†,
P-E Lippens‡, J Olivier-Fourcade‡ and J-C Jumas‡
† Laboratoire de Chimie Physique–Matie`re et Rayonnement (Unite´ associe´e au CNRS),
Universite´ Pierre et Marie Curie, 11 rue P et M Curie, 75231 Paris Cedex 05, France
‡ Laboratoire de Physicochimie des Mate´riaux Solides (Unite´ associe´e au CNRS), Universite´
Montpellier II – Sciences et Techniques du Languedoc, Place Euge`ne Bataillon, 34095
Montpellier Cedex 02, France
Received 12 February 1996, in final form 1 July 1996
Abstract. The sulphur 3p valence states of Sb2S3, Tl3SbS3, TlSbS2, TlSb3S5, TlSb5S8, and
Tl2S of the Tl2S–Sb2S3 system have been investigated by means of the S Kβ (3p → 1s)
x-ray emission band. TlSb3S5 and TlSb5S8 were also studied in the glassy phase. Synchrotron-
radiation-induced fluorescence was used. The spectra are discussed in terms of bonding between
the sulphur atoms and their nearest neighbours with the help of tight-binding calculations. We
observe significant differences among the S 3p states of all these materials, including between
the crystalline and glassy phase of TlSb3S5. The variations in the S Kβ bandwidth and shape
clearly reflect the sensitivity of the S 3p states to the local environment and indicate that the
valence densities of states are mainly influenced by local order.
1. Introduction
Antimony chalcogenides have broad-ranging applications in optoelectronics [1] and as such
represent an interesting class of compounds. The antimony atom, in its Sb3+ oxidation
state, is characterized by the existence of a 5s2 lone active electronic pair, which distorts
its local environment and leads to complex and widely differing local structures. For
example, in Sb2S3, Sb atoms exhibit two different sites with three and five S atoms as first
neighbours [2]. The addition of a third more electropositive element such as Tl leads to
changes in the environment of the Sb atoms and modifies the properties of the system. We
have isolated four crystalline phases belonging to the Tl2S–Sb2S3 system: Tl3SbS3 (25%
Sb2S3 and 75% Tl2S), TlSbS2 (50% Sb2S3 and 50% Tl2S), TlSb3S5 (75% Sb2S3 and 25%
Tl2S), and TlSb5S8 (83% Sb2S3 and 17% Tl2S).
In this class of compounds, a wide range of optical and electrical properties are
expected [3]. For instance, Tl3SbS3 is isostructural with Tl3AsSe3 [4, 5] which is a
semiconductor and exhibits nonlinear optical [6] and opto-acoustic [7] properties. It has a
large transparency in the 1.23–18 µm range and a high refractive index between 1.55 and
5.3 µm.
Chalcogenide glasses are of special interest because of their ease of fabrication, with the
possibility of being able to vary their composition to influence their properties [8]. In fact
glassy Sb2S3 can only be obtained by means of a high cooling rate [9] and many studies
have been concerned with solid solutions of Sb2S3 with other chalcogenides. Addition of
c
0953-8984/96/438421+08$19.50 ꢀ 1996 IOP Publishing Ltd
8421

8422
S Dupont et al
a modifier such as Tl2S decreases the cooling rate required to form the glass. We have
obtained glasses containing 5–35% Tl2S by simply quenching the melt in water [10]. In
this composition region, we synthesized two glass phases, g-TlSb3S5 and g-TlSb5S8, with
the same composition as the crystalline phases c-TlSb3S5 and c-TlSb5S8.
Up to now, the atomic structure and the electronic properties of these compounds
have been studied mainly by x-ray diffraction, Mo¨ssbauer spectroscopy, x-ray absorption
spectroscopy and extended x-ray absorption fine-structure analysis [11–13].
Little
experimental information exists concerning the valence densities of states (DOSs) [14, 15].
X-ray emission spectroscopy (XES) probes the occupied DOS with angular momentum and
element selectivity. Our purpose was to search for modifications in the partial S 3p DOS as
the structure varies in the Tl2S–Sb2S3 system using the S Kβ (3p→ 1s) emission from these
ternary chalcogenides and the two binary compounds Tl2S and Sb2S3. The S Kβ spectra of
the two glasses were also recorded in order to obtain structural information by comparison
with the crystalline phases. The XES data are discussed in relation to local environment
and to tight-binding DOS calculations performed for TlSbS2 and Tl3SbS3.
2. Experiments
The binary chalcogenides Tl2S and Sb2S3 were prepared by direct reaction of appropriate
quantities of high-purity elements (99.999%). Ternary compounds were synthesized by
mixing stoichiometric quantities of the binary compounds through solid state reaction at
various temperatures according to the compound desired. TlSb5S8 was synthezised by
recrystallizing amorphous TlSb5S8 [16]. The glasses were obtained by quenching the melt
of the binary compounds in water following the procedure indicated in [13]. The materials
were crushed into powders with about 5 µm granularity and stored in evacuated silica tubes.
The emission spectra were excited in the fluorescence mode with a polychromatic beam
from a bending magnet beamline at the Super-ACO storage ring (LURE, Orsay). This
excitation mode is specially suitable for reactive samples because it avoids contamination
and chemical modifications of the samples during the measurements. The samples were
prepared by depositing a thin uniform layer of chalcogenide powder onto adhesive tape. A
¯
specially designed bent crystal spectrometer [17] fitted with a 0.5 m radius (1011) quartz
crystal (2d = 0.66855 nm) and a multichannel-plate 2D position-sensitive detector was
used to record the data. Typical count rates were around 300 Hz over the whole spectrum.
The energy resolution was about 1 eV.
3. The theoretical approach
We have used a tight-binding method based on the approach of Slater and Koster to calculate
the DOS [18]. The choice was mainly dictated by the complexity of the crystal structures
(for example up to 81 atoms in the Tl2S unit cell). A previous tight-binding analysis of the
121Sb Mo¨ssbauer isomer shift for antimony chalcogenides had shown the validity of this
approach for these materials [19, 20]. Within the tight-binding approximation the S p-like
DOS may be written
X
ꢀ
ꢀ
< ϕp | 9k > 2 δ(E − Ek)
ꢀ
ꢀ
np(E) =
k
where ϕp is the 3p orbital of the sulphur atoms, Ek and 9k are the energies and the wave
functions at point k of the Brillouin zone. The summation is performed over the 3p orbitals
of each of the sulphur atoms and 64 k-points of the Brillouin zone. We checked that

S Kβ x-ray fluorescence spectra of Tl2S–Sb2S3
8423
taking more k-points did not modify the results. As in [19, 20], the present calculation
was performed within the two-centre approximation. The one-centre parameters were taken
as the energies of the neutral atoms calculated by Herman and Skillman [21]. For the
two-centre parameters empirical forms according to Harrison [22] and Robertson [23] were
considered and no significant differences were observed for the S p DOS. For simplicity
we did not include interactions between thallium and sulphur in the calculation. This point
will be discussed later.
The calculated p DOSs were convoluted with a Gaussian function of 2.5 eV full-width
at half-maximum (FWHM) for easier comparison with the experimental data. This was the
heuristic value used to broaden calculated molecular orbitals for comparison with very-high-
resolution experiments on S8 [24]. It is larger however than the combined effects of the
core-level lifetime broadening, which is about 0.6 eV [25], and the experimental broadening,
which is taken to be Gaussian in shape.
4. Results and discussion
S Kβ spectra of crystalline Tl2S and Sb2S3 are compared in figure 1; those of crystalline
Tl3SbS3, TlSbS2, TlSb3S5, TlSb5S8, glassy TlSb3S5 and glassy TlSb5S8 are given in figure 2.
All spectra are normalized with respect to the same maximum amplitude (peak A) and
plotted on an x-ray transition energy scale. The overall shape of the S Kβ spectrum changes
significantly from one compound to the other. In Tl2S the S Kβ band (FWHM 3.6 eV)
consists of a main band (A) and a well-resolved shoulder (C) located at about 4.1 eV to the
low-transition-energy side of band A. The main S Kβ band in Sb2S3 appears to be much
broader (FWHM 4.5 eV) due to the smaller energy separation between the peak (A) and the
shoulder (B) and a larger B/A intensity ratio. The S Kβ spectrum of the ternary compounds
also consists of the two well-resolved features. The spectrum for Tl3SbS3 is even narrower
(FWHM 3.2 eV) than that of Tl2S. The spectra of g-TlSb3S5, g-TlSb5S8 and c-TlSb5S8 are
very similar but the crystalline phase of TlSb3S5 shows a narrower main peak as a result of
a weaker lower transition energy structure. In table 1, we give the S Kβ peak and shoulder
positions. The shoulder positions were estimated by systematically fitting two Voigt profiles
only to the spectra (this is illustrated in figure 1). The procedure has no specific justification
in relation to the S 3p partial DOS but it is useful for assessing the relative influence of
bonding of the more tightly bound states in the various compounds.
Table 1. The energy position (eV) of the main peak (A) and its shoulder (B or C) given by the
decomposition of the S Kβ spectrum.
Sample
A
B or C
A − (B or C)
Tl2S
Tl3SbS3
TlSbS2
c-TlSb3S5
g-TlSb3S5
c-TlSb5S8
g-TlSb5S8
Sb2S3
2465.8
2465.8
2466.0
2466.1
2466.5
2466.4
2466.4
2466.2
2461.7
2462.0
2463.2
2463.4
2464.1
2463.9
2464.0
2463.7
4.1
3.7
2.8
2.8
2.4
2.5
2.4
2.6
The structural complexity of these compounds is of course closely related to the very
large number of allotropes of sulphur [26] and the Sb 5s2 lone pair. Sb2S3 is built from

8424
S Dupont et al
Figure 1. S Kβ spectra of Tl2S and Sb2S3. The continuous line through the points is constructed
from two Voigt profiles to indicate the shoulder position and a nonlinear background (not shown).
˚
(Sb4S6)n double chains with Sb–S distances varying from 2.38 to 2.83 A. As for Tl2S,
it has a layered structure, similar to that of CdI2. In CdI2, the layers are connected by
anion–anion bonds, whereas the layers are connected by cation–cation bonding in Tl2S.
The thallium atoms lie at six different pyramidal sites. Each sulphur atom is surrounded
by six Tl atoms but lies at five distinct crystallographic sites. The Tl–S distances vary
˚
from 2.5 to 3.3 A [27]. Details of the crystallographic structure of the ternary materials
have been published elsewhere [4, 28–30]. Tl3SbS3 has the simplest structure with just one
˚
crystallographic site for each element. The Tl–S bond length is 3.03 A. Sulphur has only
˚
one antimony nearest neighbour at 2.43 A, but each Sb atom has three nearest neighbour
˚
sulphur atoms [4]. TlSbS2 also has relatively short Sb–S distances (2.5 A). This time Sb
˚
has two and S has four different sites. The Tl–S distance of 3.1 A is similar to that of
Tl3SbS3. In TlSb3S5, sulphur has five different crystallographic sites. Each S has two Sb
˚
nearest neighbours on average at an average Sb–S distance of 2.54 A. Each S atom has one
˚
Tl nearest neighbour at an average distance of 3.17 A [28]. Finally there are 16 different
˚
crystallographic sites for sulphur in TlSb5S8 with large Tl–S bond lengths (3.25 A) for
an average of 0.75 nearest neighbour. The 2.2 Sb nearest neighbours, on average, have
˚
covalent Sb–S bonds (2.56 A). It is interesting to note that the sum of the Sb–S covalent
˚
˚
radii is 2.4 A whereas the sum of the Tl–S ionic radii is 3.3 A. Information concerning the
local environment of sulphur atoms is given in table 2.
Moving down the group IIIB elements from aluminium to thallium, the overlap between
the s and p valence states diminishes to the point at which a gap opens between the s and p
bands in Tl. Note that the s-like states are centred around 6 eV so that the overall valence
band width is similar to that of aluminium. The Tl 6p states form a narrower peak close to
the Fermi energy (EF ). Going across the periodic table from the group IIIB to group VIB
elements, the overlap between s and p states also decreases. This time it is a consequence
of the increased binding energy of the s-like states. The most studied allotrope of sulphur is
S8. X-ray photoelectron spectroscopy (XPS) experiments [31] indicate that the S 3s states
contribute to the bonding but are tightly bound (10–20 eV binding energy). This has also

S Kβ x-ray fluorescence spectra of Tl2S–Sb2S3
8425
Figure 2. S Kβ spectra of crystallized and glassy compounds of the Tl2S–Sb2S3 system.
Table 2. Local environments of sulphur atoms.
Number of
S sites
Average number of
first Sb neighbours
Average Sb–S
Average number of
first Tl neighbours
Average Tl–S
bond length (A)
˚
˚
Sample
bond length (A)
Tl2S
5
1
4
5
16
3
3.0
3.0
2.5
1.0
0.75
2.87
3.03
3.17
3.17
3.25
Tl3SbS3
TlSbS2
TlSb3S5
TlSb5S8
Sb2S3
1.00
1.75
2.0
2.2
2.7
2.43
2.55
2.54
2.56
2.64
been confirmed by S Kβ and L2,3 XES spectra [24]. Antimony occupies an intermediate
position between sulphur and thallium with a relatively narrow 5p band and with 5s states
centred at about 7 eV binding energy.
˚
With some Tl–S distances as short as 2.5 A in Tl2S, this compound should show
some overlap between the respective p orbitals. In fact what is unambiguously observed is
hybridization betwen S 3p and Tl 6s states (peak C). Tl2S XPS measurements [32] confirm
that the S p-like band has a FWHM of about 3.5 eV and an s-like band situated about
4.5 eV below the p-peak, namely 6 eV below EF . The origin of shoulder B in Sb2S3, on
the other hand, must be the result of hybridization between S 3p and Sb 5p states, whereas
the main peak belongs mainly to non-bonding S 3p states.

8426
S Dupont et al
Figure 3. The S Kβ spectrum of TlSbS2 and partial densities of states for S and Sb.
The analogy between these binary compounds and the ternary compounds is clear.
Starting with Tl3SbS3 which has the highest thallium content, we observe a well-defined
shoulder at C which we attribute to the S 3p–Tl 6s hybridization. Peak A is narrower than
that in Tl2S which could mean that covalent bonding between Tl 6p and S 3p states is
weaker due to the absence of short Tl–S bond lengths. The single Sb atom has little impact
on the shape of the spectrum.
Compared to Tl3SbS3, the thallium content in TlSbS2 is roughly halved but the Sb
content increased. Each S has 2.5 nearest-neighbour Tl sites. This gives rise to a fairly
intense shoulder B situated closer to peak A. The Tl 6s–S 3p hybridization has obviously
weakened while the Sb–S covalent bonding contributes to the structure. The width of peak
A is little changed despite the four distinct sulphur sites. This is not unexpected because
the x-ray emission from each type of site measures the difference in energy between the
S 1s core hole and the valence states. A change in core-level binding energy involves a
compensating change in screening by valence electrons at each type of site, thus cancelling
out broadening effects. The same can be said for TlSb3S5 with its five different sulphur
sites. Shoulder B in TlSb3S5, however, is weaker than expected compared to the glassy
phase and to TlSbS2 with its lower Sb content. c-TlSb5S8 again shows a strong shoulder at
B. In this case the glassy and crystalline compounds are indistinguishable. As pointed out
above, even the crystalline compound has 16 different crystallographic sites so it is hardly
surprising that the glassy state does not have a different spectrum. Coming back to TlSb3S5,
it is interesting to note that this compound can be described roughly as consisting mainly
of (SbS4)n ribbons. 121Sb Mo¨ssbauer measurements [33] indicated that the glass is made
up of fragments of ribbon. The effect of disorder or more generally increasing the number
of different crystallographic sites according to our XES measurements is to increase the
covalent bonding between the Sb and S p-like states.
The model calculations performed for TlSbS2 (figure 3) and Tl3SbS3 (figure 4) tend
to confirm our qualitative interpretations since they indicate that the Sb–S interaction is
dominated by p-bonding. Sb s-like states are not predicted to hybridize with the S p-like

S Kβ x-ray fluorescence spectra of Tl2S–Sb2S3
8427
Figure 4. The S Kβ spectrum of Tl3SbS3 and partial densities of states for S and Sb.
states. The main peak in the calculated DOS corresponds to non-bonding S states. Lone
electron pairs are expected in chalcogenides with twofold coordination. They have indeed
been observed in XPS spectra of S8 thin films [31] as a weakly bound peak about 2 eV
below EF . The S p-like bonding states lie about 5 eV below EF . The complex bonding of
these compounds certainly explains that no clearly indentifiable structure attributable to lone
pairs is observed in these S Kβ spectra. The main difference between the S p-like states in
Tl3SbS3 and those in TlSbS2 according to the tight-binding calculation is the absence of a
shoulder to the low-binding-energy side of A. This was not confirmed by the shape of the S
Kβ spectrum but, as mentioned in section 3, the calculation neglected Tl 6p states. These
very probably contribute to the signal at low binding energies, further masking any lone-
pair-related structure. The structure to the low-energy side of peak A in TlSbS2 (figure 3)
is echoed by small features in the Sb s-like and p-like states according to the calculation.
Though not resolved experimentally, its presence may explain that the DOS are not steeply
sloped towards EF .
The optical gap in the Tl-rich compounds is in the range 1.2–1.3 eV but in the other
compounds it increases to about 1.5 eV. This is no doubt as a result of the increased Sb-S
covalency observed here.
5. Conclusions
The S Kβ emission band of crystalline and glassy materials of the Sb2S3–Tl2S system
measured in the fluorescence mode using synchrotron radiation reveals significant changes
in the S 3p electronic structure as relative Tl and Sb contents are modified. A tight-binding
calculation shows that the main features of the spectra of the antimony compounds can be
assigned to non-bonding S 3p states. Bonding states, essentially consisting of hybridized S
and Sb p-like states, are very sensitive to local environment and to disorder. In the case
of TlSb5S8, which has the most complex structure with 16 different sulphur sites, there is

8428
S Dupont et al
no difference between the ordered and disordered phases. However, TlSb3S5 with only five
different sulphur sites does show a significant change in its electronic structure. This means
that quenching a particular composition may lead to modified physical properties relative
to the equivalent crystalline phase.
Acknowledgments
We would like to thank the staff of Super-ACO. The experiments were performed on
beamline SB3 belonging to the Commissariat a` l’Energie Atomique, De´partement des
Mesures, Centre d’Etudes de Bruye`res-le-Chaˆtel.
References
[1] Goodman C H L 1985 Mater. Res. Bull. 20 237
[2] McKee D O and McMullan J T 1975 Z. Kristallogr. 142 447
[3] Iban˜ez A, Fourcade J, Jumas J C, Philippot R and Maurin M 1986 Z. Anorg. Chem. 540/541 106
[4] Rey N, Jumas J C, Olivier-Fourcade J and Philippot E 1984 Acra Crystallogr. C 40 1655
[5] Hong H Y-P and Mikkelsen Jr J C 1974 Mater. Res. Bull. 9 365
[6] Feichtner J D and Roland G W 1972 Appl. Opt. 11 993
[7] Gottlieb M, Isaacs T J, Feichtner J D and Roland G W 1974 J. Appl. Phys. 45 5145
[8] Dubois B, Portier J and Videau J-J 1984 J. Optique 15 351
ˇ
[9] Cervinka L and A Hruby´ A 1982 J. Non-Cryst. Solids 48 231
[10] Durand J M 1994 Thesis Montpellier
[11] Olivier-Fourcade J, Iban˜ez A and Jumas J C 1990 J. Solid State Chem. 87 336
[12] Olivier-Fourcade J, Iban˜ez A, Jumas J C, Dexpert H, Blancard C, Esteva J-M and Karnatak R C 1991 Eur.
J. Solid State Inorg. Chem. 28 409
[13] Durand J M, Lippens P E, Olivier-Fourcade J, Jumas J C and Womes M 1996 J. Non-Cryst. Solids 194 109
[14] Lampre I, Gheorghiu A, Se´ne´maud C, Lippens P E, Durand J M, Olivier-Fourcade J, Lefebvre I and Lannoo
M 1993 Proc. 16th Int. Conf. on X-Ray and Inner-Shell Processes (Debrecen, 1993) ed L Sarkadi and D
Berenyi
[15] Gheorghiu A, Lampre I, Dupont S, Se´ne´maud C, El Idrissi Raghni M A, Lippens P E and Olivier-Fourcade
J J. Alloys Compounds 1995 228 143
[16] Jumas J C, Olivier-Fourcade J, Rey N and Philippot E 1985 Rev. Chim. Mine´rale 22 651
[17] Hague C F and Avila A 1996 to be published
[18] Slater J C and Koster G F 1955 Phys. Rev. 94 1498
[19] Lefebvre I, Lannoo M, Allan G, Iban˜ez A, Fourcade J, Jumas J C and Beaurepaire E 1987 Phys. Rev. Lett.
59 2471.
[20] Lefebvre I, Lannoo, M, Allan G and Martinage L 1988 Phys. Rev. B 38 8593
[21] Herman F and Skillman S 1963 Atomic Structure Calculations (Englewood Cliffs, NJ: Prentice-Hall)
[22] Harrison W A 1980 Electronic Structure and Properties of Solids (San Francisco: Freeman)
[23] Robertson J 1983 Phys. Rev. B 28 4671
[24] Kortela E-K, Suoninen E, Karras M and Mannes R 1972 J. Phys. B: At. Mol. Opt. Phys. 5 2032
[25] Krause M O and Oliver J H 1979 J. Phys. Chem. Ref. Data 8 329
[26] Meyer B 1976 Chem. Revs. 76 367
[27] Man L I Sov. Phys. Crystallogr. 15 399
[28] Gostojic M, Nowacki N and Engel P 1982 Z. Kristallogr. 159 217
[29] Rey N, Jumas J C, Olivier-Fourcade J and Philippot E 1983 Acta Crystallogr. C 39 971
[30] Engel P 1980 Z. Kristallogr. 151 203
[31] Salaneck W R, Lipari N O, Paton A, Zallen R and Liang K S 1975 Phys. Rev. B 12 1493
[32] Porte L and Tranquard A 1980 J. Solid State Chem. 35 59
[33] Olivier-Fourcade J, Jumas J C, Rey N, Philippot E and Maurin M 1985 J. Solid State Chem. 59 174