Professional Documents
Culture Documents
The Content of Sodium in Aluminium in Laboratory and in Industrial Cells
The Content of Sodium in Aluminium in Laboratory and in Industrial Cells
net/publication/286504861
CITATIONS READS
5 451
8 authors, including:
All content following this page was uploaded by V. Danielik on 09 May 2016.
Sodium and calcium are always present as undesired impurities in Boundary layer Bulk electrolyte
aluminium produced by the Hall-Héroult process [1]. When
lithium fluoride and magnesium fluoride are added to the cryolite- Distance from Al cathode
alumina-based electrolyte, the metal will also contain these
metals, due to exchange equilibria of the type,
Figure 1: Concentration gradients at the metal / electrolyte
MF + 1/3 Al = M(in Al) + 1/3 AlF3 M = Li, Na (1) interface.
MeF2 + 2/3 Al = Me(in Al) + 2/3 AlF3 Me = Ca, Mg (2)
Keller et al. [2] made estimates of the mass transfer coefficient for
A thermodynamic treatment of these equilibria is presented in the the transfer of Al(III) species through the boundary layer. The
Appendix. When molten aluminium is polarized, the content of composition of the melt at the interface and the sodium content in
sodium in the aluminium increases, and we observe a cathodic the metal could then be derived. By varying the mass transfer
overvoltage, which is mainly a concentration overvoltage [1]. The coefficient from 10×10-3 cm s-1 to 1×10-3 cm s-1 the sodium
primary cathode reaction is the reduction of Al(III) species, and it content varied from 46 ppm to 110 ppm at 0.5 A cm-2 cathodic
can be written schematically as follows, current density for an electrolyte with NaF/AlF3 molar ratio CR =
2.3 and 950 ºC.
AlF3 + 3Na+ + 3e-= Al + 3NaF (3) Tabereaux [3] showed that the average sodium level in the
aluminium in industrial cells depends on cell technology. Older
The electrical charge is transported mainly by Na+ cations (tNa+ = technology with less stable magnetics shows lower sodium
J.Thonstad et al. 1 of 7
contents than modern cells with well-balanced magnetic fields. As Results and discussion
an example, Soderberg cells can average about 70 ppm Na, while
modern prebakes can average 150 ppm. However, it can be a wide Equilibrium data The contents of alkali and alkaline earth metals
variation around these avarages. Modern stable cells also achieve in aluminium in equilibrium with cryolite melts have been
higher current efficiency, so Tabereaux concluded that there in an investigated by several authors [1, 5 - 12]. Tingle et al. [5] studied
indirect correlation between high current efficiency and high how the content of sodium in aluminium in equilibrium with NaF
sodium contents. – AlF3 melts varied with time, concentration and temperature.
Dewing and van der Kouwe [6, 7] derived a regression equation
Investigations of the content of sodium in aluminium and of the for the contents of sodium and magnesium in aluminium in
cathodic overvoltage, are much easier to perform in the laboratory equilibrium with cryolite-based melts. Peterson and Tabereaux [8]
than in industrial cells. Results obtained in laboratory cells are published a regression equation for the contents of sodium,
helpful in understanding the behaviour in industrial cells. In the calcium and magnesium in aluminium in equilibrium with
present paper we will present both laboratory and industrial data cryolite-based melts at 970 °C.
on the content of sodium in polarized aluminium. Cathodic
overvoltage data obtained in the laboratory will be correlated with
the contents of sodium in the aluminium, and sodium contents in
industrial cells will be related to operational parameters.
Laboratory Studies
Experimental
J.Thonstad et al. 2 of 7
value after 10 - 15 minutes of electrolysis. The results were found
to be reproducible. 110 350
100 300
300 90 250
w (Na) / ppm
80 200
- ηC / mV
w(Na) / ppm
70 150
200
60 100
50 50
100
40 0
0 5 10 15 20 25 30 35
time / min
0
1.5 2.0 2.5 3.0 3.5
CR Figure 5: The time dependence of the cathodic overvoltage at two
current densities (jc), and the time dependence of the content of
sodium in polarized aluminium. Na3AlF6 + 10 wt. % AlF3,
Figure 3: The equilibrium content of sodium in aluminium (CR = Al2O3(sat), 1000 ºC. ○ – jc = 0.375 A cm-2, □ – jc = 0.250 A cm-2,
nNaF / nAlF3). ○ - this work, Al2O3(sat), 1000 °C, ● - this work, ■ – the concentration of sodium at jc = 0.250 A cm-2.
alumina-free melt, 1000 °C, ◊ - Tingle et al. [5], 1000 °C, ∆ -
Fellner and Lubyová [9], 1000 °C. Full line - calculated for
alumina-free melt, 1000 °C. Dashed line - calculated for 1/3 AlF3 (II) + NaF(I) = 1/3 AlF3(I) + NaF(II) (5)
Al2O3(sat) melt; 1000 °C. Dotted line - calculated for alumina-free
melt; 950 °C. where (I) and (II) represent the compositions at the interface and
in the bulk, respectively. The cell potential when the two
300
compartments have different concentrations, can be expressed as,
150
where a(i) are the activities. A cathodic overvoltage is always
negative. This also follows from the shift in the concentrations of
50 CR = 2 NaF and AlF3 (see Fig. 1).
The chemical reaction at the cathode surface determining the
equilibrium sodium content, can be written as,
Figure 4: The influence of calcium fluoride on the content of The two reaction schemes are equivalent. The thermodynamics of
sodium in aluminium. □ - data at 1030 °C, CR = 3, Al2O3(sat); ■ - these reactions is discussed in the Appendix. Owing to these
data at 970°C, CR = 3, Al2O3(sat); ○ - data at 970 °C, CR=2, equilibria, the potential of the concentration cell and hence the
Al2O3(sat); ● - data at 970 °C, CR = 2, alumina-free melt; full cathodic overvoltage, η, can be expressed as a function of the
lines - calculated according to the thermodynamic model. activity of sodium in aluminium, as follows,
J.Thonstad et al. 3 of 7
Fig. 6 shows semi-logarithmic plots of the cathodic overvoltage as solidifies very rapidly. The sampler is then removed from the cell;
a function of current density for melts having different excess the body (heat treated resin sand) is crushed and the aluminium
contents of AlF3 (10 and 20 wt. %, respectively) and containing 0 disk (12×35 mm, 31 g) can be removed.
or 5 wt. % CaF2. It can be seen that the addition of calcium
fluoride has only a minor influence on the cathodic overvoltage, 5.4
which is in agreement with the equilibrium studies. The lines in 5.3
Fig. 6 correspond to overvoltages calculated according to Eq. (9).
5.2
The data on the content of sodium in polarized aluminium were
taken from previous mesurements in the same cell [10-12]. 5.1
5.0
200 4.9
ln(ηC/mV)
4.8
180
4.7
160
4.6
140 4.5
-ηC / mV
120 4.4
100
4.3
4.2
80 7.70 7.75 7.80 7.85 7.90 7.95 8.00 8.05 8.10 8.15
-4
60
1/(T/K).10
40
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Figure 7: Cathodic overvoltage versus the inverse temperature for
j / A.cm-2 three current densities. Na3AlF6 + 10 wt. % AlF3 + 5 wt. % CaF2
+ Al2O3(sat). ∇ - jC = 0.75 A cm-2; * - jC = 0.50 A cm-2; □ - jC =
0.25 A cm-2. Full lines – calculated from Arrhenius plots.
Figure 6: The cathodic overvoltage vs. current density (semi-
logarithmic scale) obtained in cryolite-based melts, Al2O3(sat) at
1000 ºC. □ - Na3AlF6 + 10 wt. % AlF3; ∇ - Na3AlF6 + 10 wt. %
AlF3 + 5 wt. % CaF2; ○ - Na3AlF6 + 20 wt. % AlF3; full lines –
calculated; dotted lines - extrapolation of the full line.
ln η = A + B / T (10)
Sampling
J.Thonstad et al. 4 of 7
This equation shows that more than 30 % of the sodium was lost temperature, as expected when cells operate at constant so-called
during sampling by the conventional procedure. In the following superheat (temperature above the liquidus temperature). The
all results reported are taken with the Electro-Nite sampler (with approximate trends in equilibrium sodium contents (i.e.
the exception of Fig. 9a). equilibrium with the bulk of the bath) are also indicated on the
figures, and one sees that: (1) the real sodium levels are far above
Results the equilibrium data, and, (2) the trends are opposite. This will be
discussed in the following.
Tests were made on 84 kA Soderberg cells, one cell-line of
modernized 160 kA prebake cells and one modern 180 kA It was found that the sodium content went down with increasing
prebake (AP18) cell-line. Typical sodium levels were 60-100 ppm noise. This can easily be explained by assuming that a high noise
in the Soderberg cells and 60-220 ppm in the prebake cells. It was level reflects fluid dynamic instability in the cell, e.g. a disturbed
found to be difficult to correlate sodium levels to operational metal pad, uneven current distribution, anode spikes etc. The
parameters, because the sodium levels in individual cells varied in following regression equation was derived,
a seemingly erratic manner that could not always be related to
recorded changes in operational parameters. This is demonstrated c(Na) / ppm = - 3.87 * t /°C + 6.5 * I /kA -
in Fig. 9, where the variation with time is demonstrated. Over - 827 * El.noise /µΩ + 2872 (12)
short periods of time (Fig. 9a) the sodium remained relatively
stable, while for longer periods (b and c) the variations were large. As shown in Fig. 2, the alumina content in the bath has a minor
Fig. 9c shows that the cell under study after having stayed at a effect on the equilibrium sodium content. Nevertheless, tests in
high sodium level for a period of 20 days, dropped gradually to pointfed, prebake cells of different technology showed
quite low levels. The reason for this behaviour was not clear. consistently that the sodium content varied during the normal
overfeeding-underfeeding cycle (e.g. 130 % and 70 % of the
The following parameters were believed to have an influence on actual consumption rate). As shown in Fig. 12, the sodium content
the sodium content: temperature, acidity (excess AlF3), electrical decreased during overfeeding and increased during underfeeding;
noise (current oscillations), and the alumina content. or in other words: the sodium content decreased with increasing
alumina content.
The effects of bath temperature and excess AlF3 are depicted in
Figures 10 and 11. These two parameters are interrelated, since
there was a clear correlation between low excess AlF3 and high
Figure 9: Sodium content in individual prebake cells versus time. In (a) also data obtained with conventional sampling are shown.
J.Thonstad et al. 5 of 7
Discussion of industrial data
240 6.0
230 Na Al2O3 5.5
Although the average sodium contents for entire cell-lines can
220
c (Na) / ppm
remain very stable over long periods of time, individual cells 5.0
c (Al2O3) / wt %
show, as mentioned above, large and seemingly erratic 210
4.5
variations in the sodium levels. It is clear that we do not have 200
4.0
sufficient information about the forces that govern the sodium 190
3.5
level to be able to explain some of the variations. It is 180
particularly puzzling that the trends with respect to temperature 170 3.0
and AlF3 (Figs. 10 and 11) are opposite to what one would 160 2.5
expect from thermodynamics. Many possible explanations might 150 2.0
be proposed. However, as long as they have not been tested and 0 1 2 3 4 5 6 7 8 9
verified, such speculation may be of little value. Only a few time / hours
tentative ideas will be presented here.
Figure 12: Variation in the sodium content in the aluminium and
the alumina content in the bath in prebake cells during
350 overfeeding (0 – 4 hours) and underfeeding (4 – 8 hours).
300
Overheating can be used as one example. In case the cell gets
c (Na) / ppm
250 overheated, the sideledge will start melting. If the sodium in the
200 metal reacts with this melt, the metal may loose or gain sodium,
depending on the initial level and on the temperature. The
150 composition of the side ledge is close to that of cryolite [16], so
100 the metal will be exposed to two bath phases with different
compositions. Reaction with the near neutral molten sideledge
50
may then drive the sodium content down, in spite of the fact that
0 equilibrium with the bath phase at high temperature should have
945 950 955 960 965 970 975 980 985
the opposite effect.
cell temperature / oC
Reaction with sludge may have a similar effect as reaction with
Figure 10: The variation in sodium content in prebake cells molten sideledge. It is known that the bath phase in the sludge
versus bath temperature. Full line: linear regression line. Dotted tends to contain less excess AlF3 than the bath [24]. Keller et
line: trend in sodium contents that would be in equilibrium with al. [2] attributed this fact to a reaction between the sodium in the
the bulk of the bath. The regression is valid for cell current aluminium and the bath phase in the sludge. If equilibrium is
186.1 kA and electrical noise 0.13 µohm. established according to Eq. (7), where “melt” represents the
bath phase in the sludge, it is clear that the sludge will acquire a
considerably higher NaF/AlF3 ratio than the bulk of the bath.
350
Freshly formed sludge containing bulk bath may then extract
300 sodium from the aluminium and thus lower its sodium content.
c (Na) / ppm
250
If a cell cools off, some cryolite freezes out, and the content of
200 excess AlF3 in the bath increases. In addition to growth of the
150 sideledge, the cooling may lead to the formation of a slurry
containing solid cryolite at the bath-metal interface. Such a
100
slurry will inhibit mass transfer and thus raise the sodium
50 content in the metal.
0
7.5 10.0 12.5 15.0 17.5 In newly started cells the carbon lining absorbs appreciable
excess AlF3 / wt % quantities of sodium, which is taken from the aluminium. During
the initial period of cell life, lasting about two months, the cells
Figure 11: The variation in sodium content in prebake cells may tend to have low sodium contents while the lining acts as a
versus excess AlF3. Full line: linear regression line. Dotted line: sodium sink [25].
trend in sodium contents that would be in equilibrium with the
bulk of the bath. Appendix
We can suppose that a given cell has a stable operational mode Thermodynamic model
where it runs quietly (low noise), the temperature and the bath
composition are stable, the current efficiency is high and the The model is based on the following assumptions and
sodium level is high. If such a cell makes an “excursion” into an approximations:
unstable regime, this steady state is upset, the metal / bath 1. In the first step the composition co-ordinates of the system
interface may be disturbed as well, and the sodium level may NaF — AlF3 are transformed to the system NaF — NaAlF4.
drop. This transformation does not reflect the real ionic
J.Thonstad et al. 6 of 7
composition of the melt, but it is suitable to describe the 11. V. Danielik, P. Fellner and J. Thonstad, J. Appl.
thermodynamic properties. When alumina or CaF2 are Electrochem. 28 (1998) 1265-1268.
present in the melt, we assume that these additions have only 12. V. Danielik and P. Fellner, Chem. Papers 52 (1998) 195-
a diluting effect. 198.
2. The content of sodium in aluminium, which is in 13. J. Zorić, I Roušar and J. Thonstad, Light Metals 1997, 449-
equilibrium with the melt, can be described by the reaction, 456.
14. J. Thonstad and S. Rolseth, Electrochim. Acta 23 (1978)
Al(l) + 4 NaF(melt) = 3 Na(in Al) + NaAlF4(melt) (13) 223–243.
15. P.V. Polyakov, V.M. Mozhaev, V.V. Burnakin, S.S.
The equilibrium constant of this reaction was calculated Romanchenko, Electrochimiya 17 (1981) 1454-1459.
from the thermodynamic data published in the JANAF 16. J. Thonstad and S. Rolseth, Light Metals 1983, 415-423.
tables [17]. The standard Gibbs energy of formation of 17. D.R. Stull and H. Prophet, ‘JANAF Thermochemical
NaAlF4(l) was taken from Sterten et al. [18]. At 1000 °C the Tables 2nd’, ed. Nat. Bur. Stand. (USA), coden NSRDA
value of the equilibrium constant of reaction (13) was found 1971.
to be K = 3.25×10-4. The activity coefficients of sodium (and 18. Å. Sterten and I. Mæland, Acta Chem. Scand. A39 (1985)
of the other impurities) in aluminium were reviewed by 241-257.
Sigworth and Engh [19]. The Henrian activity coefficient of 19. G.K. Sigworth and T.A. Engh, Scand. J. Met. 11 (1982)
sodium in aluminium at 1000 °C in terms of mole fractions 143-149.
equals 175. 20. J. Guzman, K. Grjotheim and T. Østvold, Light Metals
3. The excess Gibbs energy of the system NaF — NaAlF4 is 1986, 425-429.
expressed by the following equation (x(i) represents mole 21. A. Solheim and Å. Sterten, Proc. of Ninth Int. Symp. on
fractions of the model components), Light Metals Production Tromsø – Trondheim Norway
August 19 – 21 1997, 225-234.
∆GE / (J mol-1) = x(NaF)x(NaAlF4) {a0 + a1x(NaAlF4) + 22. J.L. Holm, ‘Thermodynamic properties of molten cryolite
and other fluoride mixtures’, Institute of Inorganic
+ a2 [x(NaAlF4)]2 +a3 [x(NaAlF4)]3} (14)
Chemistry, NTH, Trondheim, Norway, 1971.
23. K.C. Hong and O.J. Kleppa, High Temp. Science 8 (1976)
where : a0 = - (956 ± 34)×102 + (43.97 ± 0.72) T / K; 299-307.
a1 = - (249 ± 25)×103 + (202.3 ± 3.2) T / K; 24. J. Thonstad, F. Nordmo and J.B. Paulsen, Met. Trans. 3
a2 = (659 ± 56)×103 – (502.1 ± 6.7) T / K; (1972) 403-408.
a3 = - (457 ± 40)×103+ (347.9 ± 5.0) T / K. 25. H.A. Øye, J. Thonstad, K. Dahlqvist, S. Handå, V. De
Nora, Aluminium 72 (1996) 918-924.
4. The parameters of this equation were determined by a non-
linear regression analysis (the criterion of least squares,
confidence level 99 %) using available experimental data for
the content of sodium in aluminium [1, 5-12], the vapour
pressure of NaAlF4 [20] and the solid - liquid equilibrium of
NaF with the melt [21]. For the enthalpy of mixing of the
system NaF – NaAlF4 the data by Holm [22] and by Hong
and Kleppa [23] were used.
References
J.Thonstad et al. 7 of 7