Silicon Carbide-Silicon Nitride Materials: Part 2—Corrosion Resistance and Oxidation in Al Reduction Cells and at Lab Testing

Abstract

The key question for understanding the corrosion phenomena of Si₃N₄-SiC material in Al reduction cells is as follows: does the interaction with gases promote future step corrosion by molten cryolite (bath) or does cryolite interact with the Si₃N₄-SiC refractory and deteriorate the properties of the refractory material? More probably the reactions of silicon carbide and silicon nitride with gases, which result in the formation of silica, occur before the reactions of silica with molten cryolite. The corrosion of Si₃N₄-SiC material in the reduction cell may take place by “gas-solid” reaction and by “liquid-solid” reaction. There are several variants of lab corrosion tests for the evaluation of the corrosion resistance of Si₃N₄-SiC material to cryolite. The results of the investigation of Si₃N₄-SiC lab corrosion tests give no direct evidence of selective dissolution at a specific phase (Si₃N₄ or SiC, α-Si₃N₄ or β-Si₃N₄) in cryolite. The existing variants of lab corrosion testing require clarification.

1. Introduction

The corrosion of materials includes several processes. There are different mechanisms of corrosion. There are thermodynamical approaches, electrochemical approaches, the surface energy approach, structural approaches and others that can be used for the investigation of corrosion. Corrosion testing may be considered to be a separate approach from the investigation of corrosion; it may be subdivided into industrial testing and laboratory corrosion testing.

Silicon nitride-based, silicon carbide-based and silicon nitride-silicon carbide-based materials are considered to be advanced materials for modern challenging applications [1,2]. The investigations on SiC, Si₃N₄ and Si₃N₄-SiC materials include long-term tests at high temperatures [1,2] and corrosion and oxidation resistance investigations [3,4,5].

The main customer for Si₃N₄-SiC ceramic refractory material is the aluminum industry [6]. Si₃N₄-SiC material has been used for the side lining of aluminum reduction cells since the 80s of the last century.

During electrolysis the Si₃N₄-SiC refractory (Figure 1a) makes contact with molten cryolite (bath) [7]. The main point of contact takes place during the first 1.5–2 month of service (after start-up) of the reduction cell, when the side ledge (Figure 1b) has not formed on the side lining. Also, it may take place in the case of overheating of the cell and dissolution of the side ledge (which happens occasionally) [6]. In the upper part of the side wall (above the bath) (Figure 1a) the oxidation in the atmosphere of CO/CO₂ and of vapors of fluorine and sodium compounds takes place.

When in contact with the electrolyte (consisting mainly of cryolite Na₃AlF₆), the Si₃N₄-SiC refractory material is subjected to a liquid phase interaction. Also, it is necessary to remember that the bath (electrolyte) and liquid aluminum circulate in the reduction pot result in an erosion interaction with the refractory material. After the start-up period, the normal period of the service of the reduction cell will start. Beginning from that period onwards, the side lining of the refractory material is covered with side ledge (Figure 1b) and frozen cryolite.

Probably the key question for understanding the corrosion phenomenon of Si₃N₄-SiC is whether the priority of the gas phase interaction and whether it occurs before the liquid phase interaction.

The side ledge (Figure 1b) to some extent protects Si₃N₄-SiC material from corrosion. However, the chemical changes of Si₃N₄-SiC refractory material start from the very beginning of the service of the cell but continue during the whole service.

The quality of Si₃N₄-SiC materials is very important for the service lifetime of reduction cells. Laboratory corrosion testing is popular in the Al industry. Al producers introduced a sort of “in-going control” of corrosion resistance for side lining materials. Currently there are four variants of laboratory corrosion testing for Si₃N₄-SiC materials in order to evaluate the possibility of using the material as a side lining (

Table 1. Conditions of lab corrosion tests of Si₃N₄-SiC material to molten cryolite.

Test	Electrolysis	Pre Oxidation	Atmosphere	Movement of Specimen Rods
SINTEF Skybakmoen [8,9]	yes	no	Air. Some vapors (mainly NaAIF4) and CO2/CO from the burning anode.	No
Lacournet [10]	no	yes	Air. Some vapors (mainly NaAIF4)	No
LIRR [11,12,13]	yes	no	Some vapors (mainly NaAIF4). CO2 flow	Yes. Rotation of rods
RUSAL Proshkin [14]	no	no	Some vapors (mainly NaAIF4). Air	Yes. Specimen rods are dipped in the cryolite and taken out

): the test of SINTEF (Skybakmoen) [8,9], the test of Lacournet [10], the test of LIRR (Luoyang Institute of Refractory Research, China) [11,12,13] and the test of RUSAL [14].

The SINTEF test models the behavior of Si₃N₄-SiC materials in the reduction cell, yet it does not consider the partial reduction atmosphere above the melt and the change levels of cryolite in the service during the day. The test of Lacournet does not take into account the procedure of electrolysis, but it considers the oxidation of Si₃N₄-SiC materials during long services. The LIRR test adds the flow of CO₂ to SINTEF test (models the partly reduced atmosphere in the reduction cell due to the burning of carbon anode), and the RUSAL test considers the permanent change of the level of cryolite in the cell during the service.

The RUSAL corrosion test was chosen for the evaluation of corrosion resistance and chemical changes in tested materials. The deterioration of Si₃N₄-SiC materials in this case, the level of cryolite (and consequently the atmosphere and the conditions of interactions) looks the most severe and provides reasonable results within rather a limited amount of time.

The aim of the investigation is to summarize the data on corrosion and oxidation of Si₃N₄-SiC materials in the service from the smelters, to compare the existing lab corrosion tests of Si₃N₄-SiC materials and to analyze the results of lab corrosion tests according to the RUSAL variant of testing [14].

2. Materials and Methods

Fabrication and properties of Si₃N₄-SiC materials, the analysis of its properties and the analysis of its structure and chemical composition are described in Part 1 of the paper [15].

Laboratory corrosion resistance testing of Si₃N₄-SiC materials was performed in Engineering technological center of RUSAL (Krasnoyarsk, Russia) and in the laboratory of Voljsky Abrasive Plant, Voljsky, Russia), according to methodic of RUSAL—the test specimen in the form of the rods (Figure 2) 10 × 10 × 150 mm were dipped and taken out from the melt of cryolite in 10 min intervals for 8 h. The corrosion resistance was calculated from the change in the volume of the samples after corrosion of rods (Figure 2).

The XRD and SEM analysis was performed in three zones–the zone above the melt, the zone of reaction and the zone below the melt.

During the lab testing Si₃N₄-SiC rods were moving up and down in the molten cryolite, so there was no strict border between the Si₃N₄-SiC material oxidized by air and the Si₃N₄-SiC material immersed in cryolite. The Si₃N₄-SiC rod during certain times was immersed in cryolite and during certain times was exposed to oxygen with vapors of sodium and fluorine compounds. Only a small part of the rod remained in cryolite the entire time.

XRD analysis was performed by Rigaku Ultima IV (Tokyo, Japan) diffractometer (2θ = 10–120°); SEM images of microstructures were obtained on TESCAN MIRA 3 (Brno, Czech Republic) and JEOL 6510LV (Tokyo, Japan) scanning electron microscopes.

3. Results

3.1. Corrosion and Oxidation of Si₃N₄-SiC Materials in Industrial Al Reduction Cells

The normal service time of industrial reduction cells is 60–84 months. As was said earlier, the chemical changes in the side lining (and in all cathode materials as well) starts from the very beginning of the service (Figure 3). In the contact with cryolite Si₃N₄-SiC side lining may dissolute and become thinner (Figure 3a); above the level of the bath (molten cryolite) it may become thinner (Figure 3b) and (occasionally) the upper part of the side lining may crack (Figure 3c).

As a rule, the porosity of Si₃N₄-SiC side lining material diminishes (

Table 2. The porosity and density of Si₃N₄-SiC refractories of different producers (1 and 2) at service in the reduction cells (in the upper part—above the level of molten cryolite).

No	Apparent Density (Initial), g/cm3	Apparent Density (After 180 Days), g/cm3	Open Porosity (Initial), %	Open Porosity (After 180 Days), %
1	2.68	2.75	15.8	10.4
2	2.68	2.77	15.6	7.5

), especially in the upper part of the lining, where the material has no contact with molten cryolite. Only industrial tests may give an answer regarding in which material the porosity diminishes quicker. A lower porosity may have a positive effect, because the area of possible reactions diminishes. However, in case of lower porosity the thermal strains may become bigger, which may cause cracking.

In the structure of Si₃N₄-SiC materials the grains of silicon carbide are surrounded by small grains of silicon nitride and usually these crystals have strict shapes (Figure 4a). Various authors [16,17,18] consider that α-silicon nitride is in the form of needle-like grains, while β-silicon nitride crystals are more uniform in length and width and tend to be short and prismatic in shape.

At oxidation above the molten cryolite (the structure of material #2,

Table 3. Chemical composition of Si₃N₄-SiC materials after service in reduction cells–current research (#2–5) and according to [14] (#1) (results of chemical analysis, wet chemistry).

	Composition					Service Time, Months	Comments
	SiC wt.%	Si3N4, wt.%	SiO2, wt.%	Si, wt.%	Oxides, Including Na2SiO3, wt.%
1 [11]	50.7	16.62	11.1	-	21.58	46	-
2	73.1	15.4	7.3	-	2.2	39	Upper part
3	68.1	18.2	7.2	-	6.5	39	Lower part
4	73.7	23.3	1.65	0.34	0.98	36	Upper part
5	71	25.3	2.3	0.3	1.1	36	Lower part

) silicon nitride crystals lose their crystal shapes and become more rounded (Figure 4b), and silicon oxide appears on lumpy silicon carbide crystals.

In the course of the interaction of silicon nitride (the microstructure of material #3, Table 3) with molten cryolite silicon nitride crystals disappear, and silicon carbide crystals become smooth and round shaped (Figure 4c).

In pioneer papers on the application of Si₃N₄-SiC materials in reduction cells [19,20,21], Jorge, Marquin and Temme took a generalized picture of the growth of silica content in Si₃N₄-SiC over time, not giving details and values of time and concentrations. According to these authors, after some time the silica content in the material reaches 7–8 wt.% and after this remains permanent for a certain time.

The exact data on Silica content is Si₃N₄-SiC side lining in industrial reduction cells is limited (due to the high cost of industrial cells). Our data from smelters show considerable silica content (Table 3), from 1.65% to 7–11 wt.%.

3.2. Corrosion and Oxidation of Si₃N₄-SiC Materials at Laboratory Corrosion Testing

It is well known that in different zones Si₃N₄-SiC material is exposed to different kinds of chemical attacks. According to the current research, the specimen after testing were analyzed in three zones (to the extent that was possible to do on small rods). The investigated zones were as follows: the zone above the cryolite melt; a rather broad zone (because the specimen was moving up and down) of the “triple point”—the contact of the material, air and cryolite—and a zone that remained in the cryolite melt. The most intensive corrosion (and volume loss) was found in the zone of the “triple point”.

During lab testing the Si₃N₄-SiC rods were moving up and down in the molten cryolite, so there was no strict border between the Si₃N₄-SiC material oxidized in air and the Si₃N₄-SiC material immersed in cryolite. The Si₃N₄-SiC material during certain time was immersed in cryolite and during certain time was exposed to oxygen with vapors of sodium and fluorine compounds. The results of SEM and XRD analysis are shown in Figure 5 and Figure 6 and in

Table 4. The composition of Si₃N₄-SiC materials before and after lab corrosion testing according to XRD.

	SiC, wt.%	β-Si3N4, wt.%	α-Si3N4, wt.%	α/β	Σ Si3N4, wt.%	Si3N4/SiC	Si2ON2, wt.%	Si, wt.%	SiO2, wt.%
before corrosion test	81.6	5	10.1	2.02	15.1	0.185	3	0.3	-
after corrosion test
upper part	73.3	6.7	10.5	1.57	17.2	0.23	4.5	5	-
corrosion zone	81.6	9	3.1	0.34	12.1	0.148	6.3	-	-
lower part, dipped in cryolite	76.3	6.7	11.1	1.66	17.8	0.23	3.5	-	2.4

.

In our research, in the zone above the level of cryolite (with a partly reduced atmosphere due to the influence of the vapors of sodium and fluorine compounds) the silicon carbide concentration diminishes more quickly than the silicon nitride concentration. It looks like disappearing silicon carbide gives free silicon (Figure 6, Table 3). Yet the open porosity of the specimen after 8 h exposure above the melt of cryolite (in the partly reduced atmosphere with the vapors of fluorine and sodium compounds) reduces from 12.3 wt.% to 9.2 wt.%. In the zone above the melt, the ratio of α/β modifications of silicon nitride remains approximately the same.

In the corrosion zone the silicon carbide content remains the same, the total silicon nitride content diminishes from 15.2% to 12.1 wt.%, while α-silicon nitride content decreases from 10.1% to 3.1 wt.%. This behavior is very close to what is described in the literature [16,22,23,24]. It is rather unexpected that silicon oxynitride content grows from 3% to 6.3 wt.%.

In the zone below the level of cryolite, silicon carbide is also less corrosion-resistant compared to silicon nitride, but here it transforms to silicon oxide. In the zone below the melt the ratio of α/β modifications of silicon nitride remains approximately the same, as in the starting material. Open porosity of the specimen after 8 h exposure below the melt of cryolite reduces from 12.3% to 4.7%.

4. Discussion

4.1. On Corrosion and Oxidation of Si₃N₄-SiC Materials in Al Reduction Cells and Lab Testing

According to Wang and Skybakmoen [25,26], thermodynamically the potential for silicon nitride and silicon carbide to react directly with liquid cryolite Na₃AlF₆ is minimal due to the positive Gibbs free energies of reactions (ΔG = +500 kJ/mole).

However, we know perfectly well that the corrosion process occurs in SiC-S₃N₄ refractories of the side lining in the Al reduction cells (Figure 3a). This may take place due to complex reactions, mainly due to pre-oxidation.

The reactions of silicon carbide and silicon nitride in the reduction cells may take place in the gaseous phase and in the liquid phase. A major part of oxidation reactions [7] proceed with positive volume transformation. The products of reactions occupy more space than the reactants.

Usually, the temperatures of 800–900 °C (normal in the side lining) are not considered to be critical for oxidation of silicon carbide and silicon nitride. Yet in presence of fluorine compounds and vapors of alkali compounds the mechanism of oxidation changes from passive oxidation with appearing silica protective film to active oxidation (gaseous reaction products are removed from the reaction zone).

The picture with the specimen from the lab reduction cells is a little bit different. In the zone above the melt of cryolite, the content of silicon nitride remains almost unchanged (Table 4); there is a little growth of silicon oxynitride content, but there appears a 5% of free silicon. The amount of silicon carbide decreases from 81.6% to 73.3%, so formal analysis of the lab corrosion experiment suggests that in the partly reduced atmosphere above the melt of cryolite, silicon nitride content remains unchanged while silicon carbide may be reduced to free silicon and carbon oxide.

In the corrosion zone in the lab corrosion test (Table 4), where the material is exposed to cryolite and gases, there is almost no changes in the concentrations of silicon carbide, while the content of silicon oxynitride increases considerably, and there are no traces of silicon oxide (it likely dissolved in the cryolite). In the corrosion zone the content of α-silicon nitride decreases three-fold (compared to the initial material). The microstructures of this zone are on Figure 5c. Some small needle-like crystals (that may be attributed to α-Si₃N₄) are seen on the silicon carbide crystals.

In the zone below cryolite the content of silicon nitride and silicon oxynitride remains almost the same; the same may be said about the ratio of α/β modifications of silicon nitride. According to XRD analysis silicon carbide oxidizes to silicon oxide. This probably takes place in the pores of the Si₃N₄-SiC material (in cases without the penetration of cryolite in the pores)

So, the general considerations suggest that the more probable mechanism of decay for the Si₃N₄-SiC side lining is partial pre-oxidation followed by an interaction of the appearing silica with components or electrolytes, in the gas or liquid phases, and which culminates in evaporation or dissolution. However, the picture is complex and needs future investigations.

4.2. On Priority of Corrosion Resistance of Silicon Carbide over Silicon Nitride and of α-Silicon Nitride over β-Silicon Nitride

In the literature it is mentioned [16,22,23] that the silicon nitride phase is less corrosion-resistant to the cryolite melt compared to the silicon carbide phase. There are also suggestions that the α–modification of silicon nitride is less corrosive-resistant to cryolite melt when compared with the β-modification. These assumptions are made according to microstructure analysis of Si₃N₄-SiC material after contact with liquid electrolyte. According to the profound investigations of Skybakmoen [25,26,27], which were performed on a big number of samples, α–modification of silicon nitride is less corrosion-resistant to cryolite when compared to β-modification. The conclusions are made by analysis of the microstructure of the corroded zone after lab corrosion tests.

According to the current research in lab testing there is no direct indication of the influence of modifications on the bulk volume loss of Si₃N₄-SiC materials; yet, according to SEM and XRD, α-Si₃N₄ dissolves a little bit quicker and Silicon Nitride disappears more quickly also in the corrosion zone itself. In our research on the zone above the level of cryolite (with a partly reduced atmosphere due to the influence of the vapors of sodium and fluorine compounds) silicon carbide concentration diminishes more quickly than silicon nitride concentration. It looks like the disappearing silicon carbide gives free silicon (Table 4). In the zone above the melt, the ratio of α/β modifications of silicon nitride remains approximately the same.

In the zone below the level of cryolite, silicon carbide is also less corrosion-resistant compared to silicon nitride, but here it transforms to silicon oxide. In the zone below the melt, the ratio of α/β modifications of silicon nitride remains approximately the same, as in the starting material.

There are four known kinds of lab corrosion testing of Si₃N₄-SiC material [8,9,10,11,12,13,14] to molten cryolite (and aluminum). The criterion of testing is the volume change in the tested samples due to dissolution. Yet, it is well known that in different zones Si₃N₄-SiC material is exposed to different kinds of chemical attacks.

In the current research there was an attempt to investigate the zones of Si₃N₄-SiC material after lab corrosion testing. It looks like more profound research is required for the clarification of testing of Si₃N₄-SiC materials.

5. Conclusions

• More probably the reactions of silicon carbide and silicon nitride with liquid cryolite take place after the stage of oxidation. In the gas phase the greater part of the oxidation reactions of silicon carbide and silicon nitride proceed with positive volume transformation (decreasing the porosity of material).
• In the current research the results of chemical and phase analyses of Si₃N₄-SiC materials after lab corrosion testing with molten cryolite differed according to the different zones (above the level of molten cryolite, in the zone of the level of the melt and below the level of the melt of cryolite).
• There is no direct indication of the influence of silicon nitride modifications on the bulk volume loss in Si₃N₄-SiC materials, yet, according to SEM and XRD, α-Si₃N₄ in the zone of reaction (maximal exposure) dissolves a little bit quicker than β-Si₃N₄ and silicon nitride disappears in the course of the chemical interactions a little bit more quickly than silicon carbide.