Reactions of sulphides in molten cryolite

Article abstract of DOI:10.1016/S0040-6031(03)00375-7

The freezing point depression of cryolite (Na₃AlF₆) by the addition of Al₂S₃ and FeS was investigated. It was found that for contents of up to 10wt.% Al₂S₃, it brings into the melt three new species. X-ray analysis of solidified melts of the system Na₃AlF₆-Al₂S₃ showed that it contained chiolite, Na₅Al₃F₁₄ and Na ₂S. Chiolite originates from a reaction between Na ₃AlF₆ and AlF₃. This suggests that the system Na₃AlF₆-Al₂S₃ is a part of the reciprocal system NaF, AlF₃//Na₂S, Al₂S ₃. The solubility of FeS in cryolite melt is so low that it cannot be determined by the thermal analysis. When FeO is added to the Na ₃AlF₆-Al₂S₃ melt, Fe²⁺ cations and S^2- anions react under the formation of solid FeS. A similar reaction was observed for Ni²⁺ and S^2- ions.

Full text of DOI:10.1016/S0040-6031(03)00375-7

Thermochimica Acta 410 (2004) 87–91
Reactions of sulphides in molten cryolite
P. Fellnera,∗, M. Korenkoa, M. Ambrováa, V. Danielika, J. Thonstadb
a
Department of Inorganic Technology, Slovak University of Technology, Radlinského 9, SK-812 37 Bratislava, Slovakia
Department of Materials Technology, Norwegian University of Science and Technology, N-7491 Trondheim, Norway
b
Received 21 May 2003; received in revised form 14 July 2003; accepted 14 July 2003
Abstract
The freezing point depression of cryolite (Na3AlF6) by the addition of Al2S3 and FeS was investigated. It was found that for contents of
up to 10 wt.% Al2S3, it brings into the melt three new species. X-ray analysis of solidified melts of the system Na3AlF6–Al2S3 showed that
it contained chiolite, Na5Al3F14 and Na2S. Chiolite originates from a reaction between Na3AlF6 and AlF3. This suggests that the system
Na3AlF6–Al2S3 is a part of the reciprocal system NaF, AlF3//Na2S, Al2S3. The solubility of FeS in cryolite melt is so low that it cannot be
determined by the thermal analysis. When FeO is added to the Na3AlF6–Al2S3 melt, Fe2+ cations and S2− anions react under the formation
of solid FeS. A similar reaction was observed for Ni2+ and S2− ions.
© 2003 Elsevier B.V. All rights reserved.
Keywords: Cryolite; Aluminium sulphide; Cryoscopy
1. Introduction
contain sulphur, mainly as sulphates (up to 1%). Sulphur
leaves the aluminium cell mainly as SO2, with minor
amounts of COS, CS2 and H2S [2,3]. The phase diagram
of the system Na3AlF6–AlF3–Al2O3–Na2SO4 was inves-
tigated by Delmárová and Malinovský [4]. Danielik and
Gabcova [5] studied the system Na3AlF6–NaF–Na2SO4.
Baimakov and Vetyukov [6] reported that metallic alu-
minium dissolved in electrolyte reduces sulphates to sul-
phides:
Aluminium is produced by the electrolysis of alumina dis-
solved in cryolite-based melts. Besides cryolite (Na3AlF6)
and alumina, the electrolyte contains an excess of AlF3 with
respect to Na3AlF6, and 3–6 wt.% CaF2. Other additives like
LiF and/or MgF2 can be present as well. The electrolyte
may also contain impurities, like iron, silicon, phosphorus,
sulphur, etc. [1]. The impurities are introduced into the elec-
trolyte with the alumina or fluoride salts or they originate
from the carbon anodes. The main constituents of the gas
leaving the cell are as follows: CO2, CO, CF4, C2F6, HF,
SiF4, H2O, hydrocarbons, SO2, H2S, COS, CS2, together
with some solid particulate emissions. In this paper, we
will discuss the behaviour of the sulphur compounds in alu-
minium cells, in particular, the solubility of Al2S3 in molten
cryolite.
3Na2SO4(l) + 8Al(l) → 3Na2S(l) + 4Al2O3(s),
ꢀrG◦1233 K −3271 kJ mol
(1)
−1
∼
=
Shvartsberg [7] studied losses of sulphur from cryolite melts.
He found that at a temperature of 1000 ◦C sulphate ions are
reduced by graphite. He explained his observations by the
following reaction schemes:
Sulphur originates mainly from two sources. Petroleum
coke used for the production of carbon anodes contains
0.7–4 wt.% sulphur. Cryolite and aluminium fluoride also
3Na2SO4(l) + 2Na3AlF6(l) + 3C(s)
→ 12NaF(l) + Al2O3(s) + 3SO2(g) + 3CO(g),
ꢀrG◦1233 K= − 173.057 kJ mol−1
(2)
(3)
∗
Corresponding author. Tel.: +421-2-59325523;
fax: +421-2-52920171.
Na2SO4(l) + 4C(s) → Na2S(l) + 4CO(g),
ꢀrG◦1233 K = −261.986 kJ mol−1
E-mail addresses: fellner@chtf.stuba.sk (P. Fellner),
danielik@chtf.stuba.sk (V. Danielik).
0040-6031/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0040-6031(03)00375-7

88
P. Fellner et al. / Thermochimica Acta 410 (2004) 87–91
In most cases, no experimental evidence of the reactions
mentioned above has been given. The Gibbs energies of the
reactions were calculated using the JANAF tables [8]. No
data on the solubility of sulphides in cryolite-based melts are
available. In preliminary experiments on the reduction of sul-
phates dissolved in cryolite melts by metallic aluminium, we
found by X-ray diffraction analysis that the solidified melt
contained Al2S3. In this paper, we will present results of the
study of the phase diagram of the system Na3AlF6–Al2S3.
(BN) crucible (60 mm diameter, 55 mm height). Platinum
cannot be used because it reacts with Al2S3. The BN cru-
cible was placed in a vertical furnace, under argon atmo-
sphere. A sketch of the apparatus is shown in Fig. 1. The
temperature was measured with a Pt/Pt10Rh thermocouple
located inside a hot-pressed boron nitride protection tube.
The stirrer was made from hot-pressed boron nitride and
rotated with a speed of about 300 rpm.
The melt was heated about 20 ◦C above the liquidus
temperature when the stirring was started. The stirring
lasted for 60 min before cooling was initiated with a rate of
0.7 ◦C min−1. The slow cooling rate and the vigorous stir-
ring were applied in order to prevent or reduce super-cooling
of the melt. As a further measure to avoid super-cooling
was to add small crystal seeds of cryolite were added to
the melt during cooling. The precision of the temperature
measurement was 0.1 ◦C, the reproducibility was 0.4 ◦C.
A similar apparatus was used to investigate the influence
of the addition of FeS on the temperature of primary crys-
tallisation of cryolite.
2. Experimental
2.1. Chemicals
Handpicked Greenland cryolite, Al2S3 (99.9% metal ba-
sis, Alfa Aesar), FeS (99.9% metal basis, Merck) and FeO
(99.9%, Aldrich) were used. The other chemicals were of
the grade “pro analysis”.
2.2. Apparatus
The liquidus curve of cryolite (Na3AlF6) in the system
Na3AlF6–Al2S3 was determined by thermal analysis under
an argon atmosphere. Since Al2S3 reacts with moisture,
the use of inert atmosphere was of primary importance in
the preparation of the samples and during the measure-
ments. Mixtures of 150–200 g of handpicked and crushed
Greenland cryolite and Al2S3 powder with known compo-
sition were prepared in a glove box under low-humidity
atmosphere and transferred into the pyrolytic boron nitride
3. Results and discussion
In Fig. 2, the liquidus temperature of cryolite (Na3AlF6)
is plotted as a function of the addition of up to 0–1.5 mol%
Al2S3. By addition of small amounts of Al2S3, we can as-
sume ideal behaviour of molten Na3AlF6. The curves in
Fig. 2 correspond to the theoretical prediction of the tem-
perature of primary crystallisation of cryolite based on the
Le Chatelier–Schröder equation [9,10]:
ꢀ
ꢁ
ꢀH◦fus(Na3AlF6)
1
1
ln x(Na3AlF6) =
−
(4)
RkSt
Tfus
T
where x(Na3AlF6) is the mole fraction of cryolite in the
mixture, ꢀH◦fus(Na3AlF6) the enthalpy of fusion of cryo-
lite (=106.7 kJ mol−1 [1]), Tfus the temperature of fusion of
pure cryolite (1284 K) and T the temperature of its primary
crystallisation. R is the gas constant and kSt the Stortenbeker
factor [11], which equals the number of new foreign species
that Al2S3 brings into the solution. It can be seen that Al2S3
forms (when added in small amount) three new species in
cryolite melts.
By analogy with the system Na3AlF6–Al2O3 [12], we
may assume that the composition of these species is of the
form Al2SF62−, which is a complex of the bridging type, i.e.
Al–S–Al. However, it can also be explained by the formatio−n
of any other species containing only one S atom, e.g. AlSF2
species by analogy with the species that Hjuler et al. [13]
assumed to be formed in NaCl–AlCl3–Al2S3 melts.
In Fig. 3, the liquidus temperature for higher additions
of Al2S3 is shown. For additions above 15 mol% Al2S3,
the decrease in the temperature of primary crystallisation
of cryolite is surprisingly large. This observation can be
explained by the metathetical reaction, where one product
Fig. 1. Experimental apparatus: (1) Pt/Pt10Rh thermocouple; (2) stirrer;
(3) Alsint tube for adding small crystal seeds; (4) radiation shields; (5)
hot-pressed boron nitride protection of stirrer and thermocouple; (6) melt;
(7) pyrolytic BN crucible.

P. Fellner et al. / Thermochimica Acta 410 (2004) 87–91
89
1016
1014
1012
1010
1008
1006
1004
1002
1000
1×
2×
3×
0.0
0.5
1.0
1.5
2.0
Na3AlF6
X(Al2S3) / mole %
Fig. 2. Liquidus temperatures of cryolite as a function of the amounts of Al2S3 added (0–1.5 mol%). Full lines: calculated according to Eq. (4) for new
species 1–3.
is AlF3:
ified melts in the system Na3AlF6–Al2S3 showed that they
contained Na2S and chiolite (Na5Al3F14), which is an incon-
gruently melting compound in the system Na3AlF6–AlF3.
This observation indicates that the system Na3AlF6–Al2S3
is part of the reciprocal system NaF, AlF3//Na2S, Al2S3.
The excess of AlF3 (in the form of chiolite) in the samples
with the addition of 10, 20, 30, and 40 mol% Al2S3 was
determined by a standard X-ray method. It was found that the
content of AlF3 increased from 3.6 wt.% for the addition of
2Na3AlF6(l) + Al2S3(l) = 3Na2S(l) + 4AlF3(s),
ꢀrG◦1273 K = 72.633 kJ mol−1
ꢀrG◦973 K = 54.313 kJ mol−1
,
(5)
The standard Gibbs energy of this reaction is positive. How-
ever, the activity of AlF3 in cryolite melts is very low [14]
because of complex formation. X-ray analysis of the solid-
m.p.(Al2S3) 1100˚C
1000
900
800
700
600
500
0
10
20
30
40
Na3AlF6
X(Al2S3) / mole %
Fig. 3. The cryolite-rich part of the phase diagram of the system Na3AlF6–Al2S3. Open squares: primary crystallisation of Na3AlF6. Full line: calculated
(formation of AlF3 was taken into account).

90
P. Fellner et al. / Thermochimica Acta 410 (2004) 87–91
Fig. 4. Cooling curves in the system Na3AlF6–FeS: (1) pure Na3AlF6; (2) Na3AlF6 + 0.3 wt.% FeS; (3) Na3AlF6 + 0.5 wt.% FeS.
10 mol% Al2S3 up to 6.0 wt.% for the addition of 30 mol%
Al2S3, while it remained constant by higher additions of
Al2S3.
in cryolite melts. These observations may contribute to the
understanding of the influence of investigated impurities on
the current efficiency in aluminium electrolysis.
This is in accordance with the predictions resulting from
a thermodynamic model based on this reaction scheme
(Eq. (5)). This model gives only a qualitative explanation of
the formation of AlF3, since we do not know the activities
of Al2S3 and Na2S. The activities of AlF3 can be estimated
on the basis of data published by Solheim and Sterten [14].
The model fails when it is used to calculate the liquidus
temperature in the system Na3AlF6–Al2S3 containing more
than 15 mol% Al2S3.
Acknowledgements
This work was supported by courtesy of the Slovak
Grant Agency (VEGA 1/9406/02). The authors wish to ac-
knowledge financial support from the Research Council of
Norway. Moreover, the authors would also like to thank Slo-
valco, a.s., for determination of AlF3 content in investigated
samples.
The present paper explains the behaviour of sulphide as
impurity in cryolite melts. However, further work is needed
in order to understand the whole phase diagram for the sys-
tem Na3AlF6–Al2S3. This is the reason why a part of the
liquidus curve in Fig. 3 is plotted by a dotted line.
In Fig. 4, the cooling curves obtained in the system
Na3AlF6–FeS are plotted. It follows that within the exper-
imental error FeS has no influence of the temperature of
primary crystallisation of cryolite, so it is practically insol-
uble in this melt. Thus, it was of interest to find out if FeS
precipitates when both Fe2+ and S2− ions are presents in
the melt. According to Jenttoftsen et al. [15], up to 15 mol%
“FeO” can dissolve in the cryolite melt at 1000 ◦C (FeO
is a non-stoichiometric compound.). We added 5 mol%
Al2S3 and a corresponding stoichiometric amount of FeO
to cryolite and kept the system at the temperature of
1000 ◦C for 30 min in a BN crucible. When the system was
cooled down to ambient temperature, we found a small
sphere having diameter of ∼5 mm on the bottom of the
crucible. X-ray analysis showed that the sphere consisted
of FeS.
References
[1] J. Thonstad, P. Fellner, G.M. Haarberg, J. H´ıveš, H. Kvande, Å.
Sterten, Aluminium Electrolysis, Fundamentals of the Hall–Heroult
Process, third ed., Aluminium Verlag, Düsseldorf, 2001, 279 pp.
[2] K. Tveito, J. Tonheim, K. Paulsen, J. Thonstad, in: Greenhouse
Gases in Metallurgical Industries, Proceedings of the Conference of
Metallurgists, COM 2001, Toronto, 26–29 August 2001, 291 pp.
[3] S.H. Hay, M.M. Hyland, J.B. Metson, O.M. Bade, in: Proceedings of
the 11th Slovak–Norwegian Symposium on Aluminium Electrowin-
ning, Trondheim–Bergen–Trondheim, 19–22 September 2001, 231
pp.
[4] V. Delmárová, M. Malinovský, Chem. Zvesti 21 (1967) 826.
[5] V. Danielik, J. Gabcova, Thermochim. Acta 366 (2001) 79.
[6] Yu.V. Baimakov, M.M. Vetyukov, Molten Salts Electrolysis, Metal-
lurgia, Moscow, 1966, 412 pp. (in Russian).
[7] M.B. Shvartsberg, in: G.A. Abramov, et al. (Eds.), Theoretical Princi-
ples of the Electrometallurgy of Aluminium, Metallurgizdat, Moscow,
1953, p. 556 (in Russian).
[8] JANAF, Thermochemical Tables, second ed., NSRDS-NBS 37, US
Department of Commerce, Washington, 1971.
[9] H. Le Chatelier, C.R. Acad. Sci. (Paris) 100 (1885) 441.
[10] I. Schröder, Z. Phys. Chem. 11 (1893) 449.
A similar experiment was carried out with the admixture
of NiO. In this case, a ball of NiS was found on the bottom
of the BN crucible. This indicates that NiS does not dissolve
[11] W.Z. Stortenbeker, Phys. Chem. 10 (1892) 183.

P. Fellner et al. / Thermochimica Acta 410 (2004) 87–91
91
[12] J. Thonstad, P. Fellner, G.M. Haarberg, J. H´ıveš, H. Kvande, Å.
Sterten, Aluminium Electrolysis, Fundamentals of the Hall-Heroult
Process, third ed., Aluminium Verlag, Düsseldorf, 2001, Chapter 4,
p. 69.
[13] A.H. Hjuler, R.W. Berg, K. Zacharlassen, N.J. Bjerrum, J. Chem.
Eng. Data 30 (1985) 203.
[14] A. Solheim, Å. Sterten, in: Proceedings of the Ninth International
Symposium on Light Metal Production, Tromsø–Trondheim, Norway,
18–21 August 1997, 225 pp.
[15] T.E. Jenttoftsen, O.A. Lorentsen, E.W. Dewing, G.M. Haarberg, J.
Thonstad, Metall. Mater. Trans. 33B (2002) 901.