CHEMICAL RESISTANCE OF SIDELINING MATERIALS BASED ON SiC AND CARBON

IN CRYOLITIC MELTS – A LABORATORY STUDY

Egil Skybakmoen, Henrik Gudbrandsen and Lisbet I. Stoen

SINTEF Materials Technology, N – 7034 Trondheim, Norway

Abstract

A polarised laboratory test cell for sidelining materials has been developed, in order to simulate the extremely corrosive conditions prevailing in industrial aluminium cells when no sideledge is present. The test cell allowed for simultaneous testing of four different materials, and the materials were exposed to cryolitic melt, its vapour (mainly NaAlF₄), CO₂/CO from the anode, and liquid aluminium for up to 120 h. The chemical resistance of pure silicon carbide, silicon nitride bonded silicon carbide, as well as different carbon qualities, was investigated. The chemical resistance of SiC – based materials depended on the type of material, the amount of CO₂ evolved at the anode, and the time of exposure. The oxidation resistance of SiC – based materials was tested in flowing air (950 ^oC, 100 h). The chemical degradation mechanisms for SiC and carbon, as well as the correlation between oxidation resistance and chemical resistance for SiC – based materials, are discussed.

In recent years, sidelining materials for use in aluminium electrolysis cells have increasingly been based on silicon nitride bonded silicon carbide instead of carbon materials. During normal cell operation, a layer of frozen bath (sideledge) protects the sidelining. During anode effects, or cell instability, this sideledge may melt away, and the sidelining will be directly exposed to the bath, with chemical degradation as a result. The degradation mechanisms will be completely different for SiC-based materials and carbon-based materials.

Silicon nitride bonded silicon carbide materials have different properties compared to traditional carbon-based materials. Therefore, there is a need for information about the properties of these two types of materials as sidelining materials. The main differences between the two categories of materials, which may influence the chemical resistance as sidelining materials, are the chemical composition, the electrical resistivity, the oxidation resistance, the erosion resistance, and the thermal conductivity.

 The main reason for using SiC-based materials instead of carbon materials is the difference in the oxidation resistance. SiC-based materials undergo passive oxidation in air/oxygen at high temperatures, while carbon shows active oxidation. Since SiC-based materials have higher thermal conductivity than anthracitic carbon, the sideledge will be more stable, and the thickness of the sidelining can be reduced. This opens for the use of larger anodes and higher productivity. In contrast to carbon materials, silicon nitride bonded silicon carbide is a poor electrical conductor. This reduces the amount of current flowing to the sidelining walls, when no sideledge is present.

For the evaluation of the suitability of different types of SiC -based materials different types of laboratory-scale test methods have been developed, as reviewed by Pawlek [1].

The test methods can generally be divided into the following:

- Exposure to cryolitic melts [2 – 6].
- Exposure to cryolite and aluminium [7].
- Polarised test in cryolite melt [8].
- Oxidation in air/oxygen/carbondioxide/water vapour saturated air [9, 10,11].
- Oxidation in air/oxygen followed by exposure to cryolitic melt [12].

The problem is to develop a laboratory method, which as closely as possible simulates the chemical and physical conditions which exist in industrial cells. Laboratory test methods have a shortened test period compared with industrial cells, which may have a lifetime up to 10 years. The challenge is to develop a test method with controlled test parameters that gives realistic data within limited time.

The aim of the present work was to develop a realistic test method of SiC-based sidelining materials that allows us to make a quality ranking of different types of commercial materials. The first part of the work was concentrated on the development of the test method itself, and thereafter to make a standardised test procedure.

In the new set-up that was designed the materials were exposed to all corrosive environments: Cryolite melt, its vapour (mainly NaAlF₄), CO₂ / CO from the anode, and liquid aluminium.

Combined with examination of materials gathered from autopsies of industrial cells, our test cell can be useful in fundamental studies of the degradation mechanism of new materials. Some flexibility exists with respect to testing the materials under new or extreme electrolysis conditions, e.g. low-melting baths, or with inert oxygen-evolving anode materials.

 Experimental

 Polarised test cell for SiC-based sidelining materials

The design of the test cell used in the beginning of the project is shown in Figure 1. The test cell consisted of a graphite crucible with an inner lining of sintered alumina (Alsint, Haldenwanger), the top of the cell was covered with a lid of graphite. A plate of TiB₂-graphite composite (TiB₂ - G, Great Lakes) was placed at the bottom of the crucible to ensure that the entire bottom of the cell was covered by aluminium (Al wets TiB₂-graphite well), and it also acted as a current collector. A cylindrical graphite anode (1 cm dia.) was placed in the centre of the cell. It was immersed about 2.5 - 3 cm into the bath. The anode was changed every 24 hours.

Figure 1. The preliminary test cell for SiC-based sidelining materials.

Two test pieces, with dimensions 1 x 1 x 10–11 cm, were placed symmetrically around the anode. The test cell was kept in a vertical tube furnace under argon atmosphere (outside the crucible). The temperature was measured by immersing a thermocouple in the middle of the bath while the anode was replaced. The test pieces were cathodically polarised during the test. Some general results from the preliminary test cell and the test method were published in 1995 [13].

Other important test parameters are shown in Table I.

Table I. Test parameters in preliminary tests.

The bath composition was 11.9 wt % AlF₃ and 8 wt % Al₂O₃ balanced with synthetic cryolite.

The degree of corrosion was observed by visual inspection of the test material after the test.

During the first project period, and during the preliminary tests, it was observed that the amount of CO₂ evolution (anode consumption) was an important parameter. The corrosion of the materials increased with increasing CO₂ evolution rate. It was difficult to obtain a stable cell voltage and a normal, stable anode consumption during the whole test period of 120 h. Because of the instability and problems in controlling all the test parameters, the test cell and test procedure were modified and improved as shown below:

- The anode diameter was increased from 1 cm to 1.5 cm.
- The current was increased from 1 A to 2 A.
- An insulating ring of BN was placed between the anode and the graphite lid.
- 4 test pieces were used instead of 2 test pieces.
- A new bath composition (table II) was used.
- The duration of the test was reduced from 120 h to 50 h.
- A thermocouple was placed in the crucible wall to obtain better control of the bath temperature during the test.
- A new procedure for measuring the degree of corrosion was established.

With the modified test cell and the test procedure we were able to make a quality ranking of different types of commercial silicon nitride bonded silicon carbide materials and also of other types of alternative materials.

The standard test cell for sidelining materials is shown in Figure 2.

Figure 2. The standard cell for testing SiC-based sidelining materials.

The design of the new test cell allowed us to study the test material in different zones in different chemical environment, as shown in Figure 3.

Figure 3. The chemical conditions in the test cell at different zones of the materials tested.

The chemical conditions in the test cell are very similar to those in industrial cells (when no sideledge is present), insofar as it provides the same or very similar oxidising conditions (CO₂ / CO) in the gas phase and at the gas/melt interface, the same bath composition (except for higher alumina content), and the same reducing conditions in the aluminium zone.

The composition of the bath at the start of the electrolysis is shown in Table II.

Table II. Bath composition of standard sidelining test.

* The content of alumina increased during the experiment as the alumina lining of the crucible dissolved, and a concentration of about 8 wt% was reached during the test. Other important test parameters of the standard test are shown in Table III.

Table III. Parameters of standard sidelining test.

After the test, the crucible with the test pieces was cooled for about 12 hours. The graphite lid was removed, and the rest of the test cell was heated up again in an open furnace, and the test pieces were removed from the crucible when the bath melted. Adhering bath was removed mechanically and by washing in an aqueous solution of AlCl₃.

The degree of corrosion was related to the volume loss of the pieces during the test. The volume of the test materials was measured before and after the test by using ISO 5017 standard.

The volume loss of each piece was estimated (volume before – volume after) and a scale describing the degree of corrosion was made (Table IV).

Table IV. The degree of corrosion based on volume loss.

Volume loss [%] = [(volume before – volume after) / volume before] x 100 %

Oxidation test of SiC-based materials in air

The oxidation resistance was measured on various commercial materials in air at 950 °C for 100 hours. The furnace was heated up at a rate of 300 °C / hour. Air at a flow rate of 50 ml/min was passed through the furnace. The test specimen had dimension about 3 x 3 x 3 cm and weighed from 74 – 94 g. The weight of the test piece was measured as a function of time.

Materials tested

The aim of the work was to make a quality ranking of different types of commercial silicon nitride bonded silicon carbide materials. We have tested 8 different types of materials, here denoted as A, B, C, D, E, F, G and H. The same materials were tested in the oxidation resistance apparatus, except material H. The test pieces were cut from bricks, using a diamond saw. The content of SiC varied typically from 74 – 78 wt %, the content of Si₃N₄ varied from 18 – 24 wt %, the open porosity varied from 15 – 19 % and the density varied from 2.58 – 2.65 g/cm³.

As a reference material, pure silicon carbide was tested, denoted SiC. Two tests were also performed with carbon-based materials denoted AC, BCS and CCS. AC was graphitised and BCS and DCS were carbon materials with some SiC added to reduce air oxidation at high temperatures.

Preliminary tests in the sidelining test cell

Based on visual observations of the pieces after the test, the results of the preliminary tests can be summarised as followed:

Si₃N₄ bonded SiC materials:

Corroded in the gas phase zone and at the interface electrolyte/ gas zone. The corrosion attack was dependent on the rate of CO₂ evolution from the anode (anode consumption), time, and type of material. In the preliminary tests, the anode consumption varied from test to test. Due to the fact that the degree of corrosion of the materials was dependent on CO₂-evolution, it was difficult to make comparisons between different types of materials from one test to another.

Self-bonded SiC:

- Pure SiC with porosity 16 – 18 % corroded little/somewhat in the gas zone above the electrolyte level.
- Dense SiC did not corrode at all.

Carbon-based materials:

- All the carbon materials tested (AC, BCS and CCS) were corroded in the electrolyte zone on the side facing the anode. Deposition of carbon on the anode was observed as shown in Figure 4.

Figure 4. The deposition of carbon at the anode.

After the test procedure was improved, it was possible to make a quality ranking of different types of commercial Si₃N₄ bonded SiC. The corrosion attack was strongly dependent on electrolysis time and total anode carbon consumption during the test. The tested materials, A, D, F and G, are shown after 50 h and 100 h in Figure 5 and Figure 6, respectively. All the four materials were more strongly corroded in the gas phase zone after 100 h compared with 50 h.

Figure 5. Materials tested 50 h. Total anode carbon consumption = 13.72 g / 0.274 g C /h.

Figure 6. Materials A, D, F and G tested for 100 h (bath not removed). Total anode carbon consumption = 24.26 g / 0.242 g C /h.

The results, including anode carbon consumption, volume loss and the degree of corrosion for 8 different types of commercial Si₃N₄ bonded SiC after 50 h test period are shown in Table V.

Table V. Test results listed by type of material. 50 h test period.

Explanation: A-1a is material type A, brick 1 and parallel a. A-2a is material type A, brick 2 and parallel a.

* Based on visual observations.

Pure silicon carbide (denoted SiC-1) was superior to all the Si₃N₄ bonded SiC materials with respect to chemical resistance. Among the Si₃N₄ bonded SiC materials materials, A, B and C had somewhat higher chemical resistance than material D. Materials E, F, G and H had the lowest chemical resistance. Material A-2 (block with thickness 120 mm) seemed to have a higher degree of corrosion than material A-1 (block with thickness 68 mm).

The porosity of the materials decreased typically from 14 – 18 % at the start to 4 – 6 % after the test. Electrolyte, aluminium and oxidation products penetrated the porous structure of the material during the test. The density increased from 2.6 – 2.7 g / cm³ at the start to 2.9 – 2.95 g / cm³ after the test. The weight of the test materials increased typically up to 16 %, but this was of course dependent on the grade of corrosion.

The correlation between open porosity and volume loss was investigated for some materials, as shown in Figure 7. A statistically significant correlation between volume loss and open porosity was not found.

Figure 7. The volume loss as a function of the open porosity of different types of materials tested in the sidelining test cell.

In Table VI is shown the total weight changes in %, mg / area of the surface and mg / time and area of the surface for each material after 100 h in flowing air at 950 °C.

Table VI. Weight changes after 100 h in flowing air at 950 °C.

The weight changes as function of time obtained with seven different Si₃N₄ bonded SiC materials kept in flowing air at 950 °C for 100 h are shown in Figure 8.

Figure 8. Weight changes of 7 different types of Si₃N₄ bonded silicon carbide materials in flowing air at 950 °C for 100 h as a function of time.

Materials A, B and C had the lowest weight increase, followed by materials D and E, while materials F and G had the highest weight increase.

It was expected that increasing porosity should give more oxidation due to higher surface area exposed. The relationship between open porosity and oxidation is shown in Figure 9.

Figure 9. Weight increase for different types of materials as a function of open porosity. Oxidation in air at 950 °C for 100 h.

The results obtained regarding chemical resistance and oxidation resistance clearly showed that higher oxidation resistance gave higher chemical resistance. Materials A, B and C showed both high oxidation resistance and high chemical resistance, while materials F and G were more oxidised and were also more corroded in the sidelining test. The relationship of the oxidation resistance and the chemical resistance for different types of Si₃N₄ bonded SiC materials is shown in Figure10.

Figure 10. The chemical resistance as a function of the oxidation resistance for different types of Si₃N₄ bonded SiC materials.

Carbon-based materials tested showed wear in the electrolyte zone facing the anode. At the same time, deposition of carbon was observed at the graphite anode. In contrast to SiC-based materials, carbon acts as an electrical conductor, this means that the carbon materials in the sidelining test cell act as a cathode.

The observations of the behaviour of the carbon materials in the sidelining test cell is a good example of so-called cathodic dissolution of carbon, as previously described by Gudbrandsen et al. [14].

It has been shown by Oedegaard [15] that Na₃Al₃CF₈ is the dominating dissolved carbide species in cryolite melts according to the equilibrium,

Al₄C_3(s) + 5 AlF_3(l) + 9 NaF_(l) = 3 Na₃Al₃CF_8(l) (1)

In the same work, it was also demonstrated that anodic deposition of carbon might occur according to the reaction,

Na₃Al₃CF_{8 (l)} + NaF = C_(s) + 3 AlF_{3 (l)} + 4 Na⁺ + 4 e^- (2)

which may proceed reversibly at low current densities on, for instance, a graphite substrate.

The cathodic dissolution of carbon into cryolite melts takes place by reaction (2) in the opposite direction,

C_(s) + 3 AlF_{3 (l)} + 4 Na⁺ + 4 e^- = Na₃Al₃CF_{8 (l)} + NaF (3)

The total reaction will then be the sum of (2) and (3),

C _(cathode) = C _(Anode) (4)

This means that cathodically polarised samples will corrode in the bath phase zone, and carbon (from this cathode) will be deposited on the anode, as observed in the experiments. Laboratory experiments [14] have shown that the rate of cathodic dissolution of carbon increased with increasing cathodic current density until a steady state loss was reached at a cathodic current density of about 0.1 A/cm² (dependent on melt composition and convection).

The cathodic current density on the test samples in the sidelining test cell was about 0.06 A / cm². As the cathodic current density must be higher at the side of the test samples facing the anode, it seems reasonable that the corrosion was highest in this area of the carbon-based samples.

Several papers have been published concerning oxidation of non-oxide ceramics [10,11,16 - 21]. Generally, it can be concluded that oxidation of non-oxide ceramics at high temperatures is a complex topic with many factors included. Oxidation of SiC– based materials involves transport of oxygen through surface films of silica / oxide layers. The properties of such an oxide layer are of great significance. The oxidation is dependent on the nature of the oxide layer, its density, and its thickness.

The polarised laboratory tests have shown that Si₃N₄-SiC materials typically degrade in the gas zone and at the interface electrolyte/gas zone. The degradation was dependent on the type of material, the amount of CO₂ evolution, and the time of exposure. Unpolarised Si₃N₄-SiC materials exposed to aluminium and cryolite at 1000 ° C for 720 h in argon atmosphere showed no corrosion at all [7].

It has been published that the silicon carbide tiles in the upper sidewall in an industrial cell were severely eroded [22].

The stability of SiC-materials against oxidation in general is due to the formation of a protective layer of SiO₂. SiO₂ will be dissolved into the cryolite melt, and in the gas phase zone it can react with the main vapour compound, NaAlF₄,

2SiO₂ + NaAlF_4(g) =SiF_4(g) + NaAlSiO_4(s) G⁰_1200K = -21.64 kJ (5)

The degradation in the gas phase zone is enhanced by the presence of NaAlF_{4 (g)} / Na _(g). Both the Si₃N₄ -and SiC-phases are unstable when CO₂ and NaAlF₄ are present (CO₂ is saturated with NaAlF₄) and the following reactions may take place:

2 Si₃N₄ + 3 NaAlF_4(g) + 6 CO₂ =
3 SiF_4(g) + 3 NaAlSiO_4(s) + 6 C + 4N₂ G⁰_1200K = -1166.82 kJ (6)

2 SiC + NaAlF_4(g) + 2 CO₂ =
SiF_4(g) + NaAlSiO_4(s) + 4 C G⁰_1200K = - 491.66 kJ (7)

At the gas/melt interface zone, one may expect an enhanced removal of SiO₂ because both mechanisms may be active. The NaAlSiO₄ will be dissolved in the melt. Kvam and Oye [23] have also suggested several other possible degradations reactions.

Thermodynamically the Si₃N₄ phase is more unstable than the SiC-phase, so the binder (Si₃N₄) will be attacked more easily than SiC. It has been claimed that the Si₃N₄-phase is oxidising first in a Si₃N₄ bonded SiC material [11].

This is in agreement with visual observations of the materials from the tests. The surface of the materials became porous and grains of SiC seemed to be detached. Pure SiC-materials were also less corroded than materials with a binder of Si₃N₄.

The results obtained with oxidation of different types of commercial silicon nitride bonded silicon carbide samples in an air flow at 950 °C for 100 h showed passive oxidation (weight increased due to the formation of SiO₂) for all the materials tested, as expected. But the differences in oxidation rate between the materials were considerable, i.e. material types A, B and C had about 4 times less weight increase than material types F and G.

A comparison of chemical degradation and the oxidation resistance of the different types of materials tested clearly indicate that higher oxidation resistance gave better chemical resistance. Materials of types A, B and C had the highest oxidation resistance and the best chemical resistance, while materials of types E, F and G had the lowest oxidation resistance and the lowest chemical resistance. The chemical resistance is obviously dependent on the degree of oxidation. The degree of oxidation is again dependent on the type of Si₃N₄ bonded SiC material.

The homogeneity of two types of commercial Si₃N₄ bonded SiC materials has recently investigated by Kvam and Oye [23]. The homogeneity was found by measuring the bulk density, porosity and the contents of nitrogen and oxygen at different positions in the bricks. These investigations revealed inhomogeneities regarding porosity, density and content of nitrogen and oxygen. The nitrogen content was found to be higher at the corners compared to the centre of the blocks, while oxygen showed the opposite trend. These inhomogeneities can be of importance regarding the oxidation resistance and chemical resistance of the materials tested.

Other properties that can affect the oxidation and chemical resistance of silicon nitride bonded materials can be as follows:

- The pore size and pore size distribution.
- The relative amount of Si₃N₄ and SiC.
- The particle size and particle size distribution of SiC.
- The amount of free silicon, the degree of nitration.
- The amounts of SiO₂ , Si₂ON₂ and SiAlON.
- The amounts and types of sinter additives (Al₂O₃, CaO, MgO, Fe₂O₃ etc).
- The amounts and types of impurities.
- The type of organic binder used before compression.
- The homogeneity of the brick.

In spite of improved resistance to degradation of the SiC-based materials compared to carbon materials, it will still be essential to maintain a stable layer of frozen bath, sideledge, during the electrolysis.

Acknowledgement

Financial support from the Research Council of Norway and from the Norwegian aluminium industry is gratefully acknowledged. Thanks are due to Stein Julsrud, Stanislaw Jarek and Ole Jacob Siljan, Hydro Research, for initiating the construction of the sidelining test cell. Thanks are also due to Alton Tabereaux, Reynolds Metals Company, for supplying TiB₂-graphite composite materials. Christian Schoning and Ove Darell, SINTEF Materials Technology, performed the oxidation resistance measurements.

References

1. R.P. Pawlek, "SiC in aluminium electrolysis cells", Light Metals 1996, pp 527 – 33.

2. M.V. Swain and E.R. Segnit, "Reaction between cryolite and silicon carbide refractories", J.Austr. Ceram. Soc. 20, 9 – 12 (1984).

3. M.W.A. Stewart, D.S. Perera, and S.H.F. Leung, "Reaction of molten cryolite with refractory ceramics", Int. Ceram. Monogr. 1, 2^nd Proceedings of the int. ceramic conf., 984 – 91 (1994).

4. A.T. Tabereaux, "Silicon carbide bricks in aluminium reduction cell cathodes", Light metals proc. and appl., Canadian Inst. of Mining, 149 – 163 (1993).

5. A.T. Tabereaux and A. Fickel, "Evaluation of silicon carbide bricks", Light Metals 1994, pp 483 – 491.

6. A.F. Fickel, J.S. Kramss and P.W. Temme,"Silicon Carbide Refractories for Aluminium Reduction Cell Linings", Light Metals 1987, pp 183-188.

7. E. Skybakmoen, " SiC-materialer for elektrolyseceller. En undersøkelse av kjemisk bestandighet", Ildfaste materialer i aluminiumindustrien 13. - 14. nov., Trondheim, Norway, (1995).

8. B.J. Welch and A.E. May, "Materials problem in Hall-Heroult cells", 8. Int. Leichtmetalltagung, Leoben-Vienna, 120 – 125 (1987).

9. A.F. Fickel, B. Rolofs and, P.W. Temme, "On the oxidation behaviour of some pot lining refractories and the consequences for their practical application", Light Metals 1991, pp 405 - 412.

10. A.F. Fickel, H. Friedli, J.S. Kramms and P.W.Temme, "Mutual effects on board-lining metal and melt in aluminium-electrolysis", Light Metals 1989, pp 191 – 194.

11. M. Thorley and R. Banks,"Kinetics and mechanism of oxidation of silicon nitride bonded silicon carbide ceramics", J. Thermal Anal., 42, 811 – 822 (1994).

12. Unpublished information, Comalco.

13. E. Skybakmoen, L.I. Stoen, L. Delmas and H.A. Oye, " A new corrosion test method for sidewall materials used in aluminium electrolysis", VIII. Al-symposium. Slovak-Norwegian Conf. on Al Smelting Techn., Donovaly – Ziar Nad Hronom, 195 - 200 (1995).

14. H. Gudbrandsen, A. Sterten and R. Oedegaard, "Cathodic dissolution of carbon in cryolitic melts", Light Metals 1992, pp 521-528.

15. R. Oedegaard, " Solubility and electrochemical behaviour of aluminium and aluminium carbide in cryolitic melts", Dr.tech thesis, The University of Trondheim, Department of Electrochemistry (1986).

16. K.G. Nickel, " Corrosion of Non-Oxide Ceramics", Ceramic Int. 23, 127-133 (1997).

17. F.L. Riley, " The Corrosion of Ceramics: Where Do We Go from Here?", Key Engineering Materials Vol. 113, 1 - 4 (1996).

18. R.G.O’Donnel and M.B. Trigg, "Environmental Degradation of Silicon Based Non-Oxide Ceramics", Materials Forum 19, 91 – 100 (1995).

19. R.G. Munro and S.J.Dapkunas,"Corrosion Characteristic of Silicon Carbide and Silicon Nitride", J. Res. Natl. Inst. Stand. Technol. 98, 607 – 631 (1993).

20. N.S. Jacobson, E.J.Opila, D.S.Fox and J.L.Smialek, "Oxidation and Corrosion of Silicon-Based Ceramics and Composites", Mat.Science Forum Vols. 251 –254, 817 – 832 (1997)

21. N.S. Jacobson, "Corrosion of Silicon-Based Ceramics in Combustion Environments", J.Am.Ceram.Soc. Vol. 76, No.1, 3 - 28 (1993).

22. G. Bearne and A. Jenkin," The impact of cell geometry on cell performance", Light Metals 1995, p. 378.

23. K.R. Kvam and H.A. Oye,"Homogenity and degradation of SiC sidelinings", Ninth Int. symp. on Light Metals Prod., Trondheim, ed. J. Thonstad, 313 – 320 (1997).

Return to article index