Light Metals 2004 Edited by Alton T.

Tabereaux TMS (The Minerals, Metals & Materials Society), 2004

Wear of Silicon Nitride Bonded SiC Bricks in Aluminium Electrolysis Cells

Flemming Bay Andersen1, Guido Dörsam1, Marco Stam2, Marcel Spreij1
1
Corus Research, Development & Technology, Ceramics Research Centre, PO Box 10000, 1970 CA IJmuiden, The Netherlands
2
Aluminium Delfzijl bv., P.O. Box 133, 9930 AC Delfzijl

Keywords: Aluminium Electrolysis, Refractory materials, Wear mechanism, Silicon nitride bonded silicon carbide, Sidewall blocks

Abstract 3
Samples from a sidewall of an aluminium electrolysis cell were Sample locations
submitted for study of the wear mechanism of the Si3N4 bonded
SiC-brick. Samples were studied using a combination of chemical 2
analysis, microscopy, SEM-EDS and powder-XRD. Unused
bricks from two suppliers were used as reference materials.
1
The unused bricks are very similar. Differences are found in the
grain size of the SiC and in the binder (E-Si3N4 content varies
between bricks in also internally in individual samples). All bricks
have a coarse matrix and higher porosity in the core.

The wear mechanism of the bricks seems a combination of attack

by Na, HF, oxidation and moisture. Below bath level oxidation
and Na play a major role but bricks are least altered. At bath level
attack is severe but observations were complicated through a
repair, which caused secondary reactions in addition to oxidation.
Figure 1. Schematic view of a cell (anode block omitted) with
Above bath level the bricks disintegrate but reaction products are
sample locations marked.
almost absent. Here we hypothesize that HF-vapor transformed
the oxidation products into volatile fluorides and other gases.
Observations on autopsy materials
This study shows that the corrosion tests currently used for Si3N4
bonded SiC-bricks, do not simulate the wear mechanism Samples from the e-cell were taken at three locations as seen in
described in this paper. In order to evaluate the performance of Figure 1, and sealed in plastic bags.
these bricks it is important to develop a new corrosion test, where The SiC material at location 3 has a grey colour and disintegrates
water, Na, HF and oxidation play a role. upon touching. The samples from location 2 were severely
contaminated with cryolite. At location 1 some wear is observed
Introduction but generally (visually) the sample looks unattacked.
SEM, XRD and light microscopy were used to investigate the
The aim of this study is to get a better understanding of the actual wear mechanism.
wear mechanisms of the sidewall lining in aluminium electrolysis
cells at Corus smelters. We intend to use this knowledge to design Samples of unused material from two different suppliers were
an improved test program including a new corrosion test. examined as well.
A lot of work on this subject has been published earlier but we
believe a new and fresh look at the problems will increase Autopsy sample location 1
knowledge about what really goes on in the sidewall. The Visually samples from this location are in much better conditions
reference list in the paper by Skybakmoen [1] gives a good than samples from the other locations.
overview of what has been published previously. The micrograph in figure 2 shows damaged SiC grains in a highly
hygroscopic matrix. SiC grains have irregular shapes possibly due
The paper describes the post-mortem examination of SiC-bricks to oxidation. The figure shows hydration products in the matrix
used in the sidewall of aluminium electrolysis cells. Figure 1 close to a porous part, which is filled with epoxy.
shows a schematic view of the cell with sample locations. The cell
was 1986 days old when it was cut out. The condition of the
sidewall was representative of the average cell at the Corus
Aluminium smelter at Delfzijl, the Netherlands. Sidewall
temperatures in operation differ at the three locations with the
highest temperature being at location 2.

413
 Figure 4. Formation of SiC and Si3N4 on a big Si-metal grain.
Figure 2. Sample location 1; Hydration products in the matrix Sample location 1
(reflected light microscopy)
The SEM – Element distribution map with composite images for
In general the matrix-material is sub-microscopic in grain size. different combinations of elements is given in figure 5. Sodium
Figure 3 shows that SiC is more stable than the matrix of Si3N4. and oxygen could only be found in the matrix between SiC and
The epoxy in the top left-hand corner represents the outside of the Si3N4. Fluoride is absent in the mapping. Other areas in the
brick. The SiC-grains that stick out are oxidized and the matrix sample were also examined for fluoride but without success.
has disappeared.

Na
Artifacts Na SiC
Figure 3. Direct melt attack on Si3N4 bonded bricks, sample SiO2
location 1 Na Na
Na
In the matrix some brightly shining Si-metal could be seen, which Na Si3N4
had partly reacted to SiC and Si3N4 (Figure 4.) In the matrix, red
shining grains were found, which were iron oxide (hematite). Na Na
Several bright white isotropic grains of FeSi metal were found in a. b.
the matrix.

O O
SiC
O
O O
Si3N4
O
O O
c. d.
Figure 5. Sample location 1, Elemental distribution.

414
The picture on top in figure 5 shows a back scatter image with
three round Si-Fe metal crystals in a SiC grain. On the right hand
side of the image silicon nitride and smaller SiC grains make up sodium silicate SiC
the matrix. Based on element mappings four composite images are
shown (figure 5a-d.) The aluminium mapping (not shown) gives a
signal below background level which means that no Al-containing
phase is present in the mapped area.
SiO2
The element mapping shows that sodium and oxygen are
concentrated in the matrix as a replacement of the Si3N4 (Figure
5a and 5c.) Some artifacts from sample preparation are observed
and they do not indicate the presence of F. In figure 5a the dark
blue gives a clear indication of the Na-distribution, in figure 5c,
the light green shows the oxygen to correlate with Na. Figure 5b Figure 7. SEM micrograph showing reaction order of attacked
shows Si3N4 as purple and the green areas the oxidized matrix. In SiC. Location 2.
figure 5d Si3N4 is marked with yellow and pink represents SiC. 1. Oxidation SiC – SiO2
Iron, nickel and zircon were found as impurities in the Si metal. 2. Reaction of Na2O or NaOH with SiO2

XRD analysis was performed and the detected phase content are Figure 7 also shows pure SiO2 in contact to SiC and sodium
given in table 1. silicate, which serves as evidence for a primary oxidation of SiC
and a secondary reaction with sodium oxide. The only fluorides
Phase content (XRD) sample location 1. found were NaF and CaF2. Closer to the shell iron was found
Į-SiC, Na2SiO3, Si2N2O, ȕ-Si3N4, Į-Si3N4 together with calcium as a brighter separate phase with a
Table 1. Phase content of used brick from location 1. consistent stoichiometry corresponding to the chemical
composition of FeCa2Na10O28.
Autopsy sample location 2 SiC
The material looks porous and partly reacted. One side of the
sample was impregnated with epoxy before cutting. sodium silicate
The sample shows rimmed remnants of SiC grains in a sodium (Na : Si = 1 : 10)
silicate matrix with some Si3N4 and some NaF and SiO2, see
figures 6 and 7. Single point analysis was carried out to determine
the chemical composition at positions of interest. Sodium silicates sodium silicate
(Na : Si = 1 : 3)
were found with different sodium content, depending on the
location in the matrix. The SiO2 content of the sodium silicates
was much higher in the reaction zone around the SiC than it was
in the matrix where the sodium silicate glasses were found, see Rest Si3N4
figure 8.

SiC Figure 8. SEM micrograph showing increasing sodium content

(weight ratio) into the matrix of sodium silicates from rimmed SiC
grain. Remaining Si3N4 shows remainder of previous matrix.
sodium silicate Location 2.

NaF FeCa2Na10(Si10O28)
pore

NaF

Figure 6. SEM micrograph showing sodium silicate rimmed SiC

grains in matrix of sodium silicate with sporadic NaF. Location 2. Sodium silicate
(Na : Si = 3 : 4)

Figure 9. SEM micrograph showing higher sodium content of

sodium silicates than in figure 8 and silica bounded iron (FeO)
and calcium (CaF2). Location 2.

415
The XRD sampling represents a profile through the wall and sampling allowed moisture of normal atmosphere to react with
results are given in table 2. The “Towards joint” sample mainly primary reaction products in the sidewall of the e-cell.
consists of products from reaction of cryolite with Si-compounds.
The “Center” sample is highly porous (compared to the unaltered Na2O + H2O Æ 2 NaOH (1)
brick) and contains oxidation products and matrix phases with
partial oxidation (SiAlON). The “Towards shell” sample is the NaF + H2O Æ NaOH + HF (2)
least altered and shows only the initial stages of attack. In this
sample both reaction with cryolite and oxidation play a role, 2 NaOH + SiO2 Æ Na2SiO3 + H2O (3)
because the Al-oxide bearing phase Nepheline was detected.
Cryolite presence must be due to penetration through joints. The NaOH + CO2 + H2O Æ Na2CO3 · H2O (4)
steel shell shows some rusting behind the SiC bricks and this
confirms the presence of oxygen. NaOH + CO2 Æ NaHCO3 (5)

Sample phase-signals General view of wear mechanism in the sidewall

Towards Į-SiC, Į-SiO2, Nephelin, NaAlSiO4, SiAlON, NaF, The lowest part of the cell wall (location 1) shows the least attack.
joint ȕ-Si3N4, Į-Si3N4 Attack is limited to some oxidation of the matrix and minor
Center Į-SiO2, Į-SiC, NaF, SiAlON, ȕ-Si3N4, Į-Si3N4 oxidation of the SiC grains. The matrix was subsequently
hydrated and now contains Na-silicate and NaOH and other
Towards Cryolite, Nephelin, CaF2, Į-SiC
secondary products in its pores. The elements F and Al are not
shell
present. The only primary alteration processes that played a role
Shell NaF, Fe2O3, Į-SiC
are Na-infiltration and oxidation.
Table 2. Phase content of sub-samples at location 2.
For the middle part of the sidewall (location 2) the severe
Autopsy sample location 3. oxidation is a main aspect of the wear mechanism. At this location
This very porous material disintegrates into powder upon SiO2 rims on SiC were found (figure 10.)
touching. Macroscopic observations reveal blueish colored
particles and thin metallic looking crystals in a grey matrix.
Microscopic investigation was hampered by the fragile nature of
the samples. The phases identified with powder-XRD are
presented in table 3 and 4.
Because of the unusual incoherent character of the material, both
the bulk sample and the below 32 Pm fraction were X-rayed
separately. No oxidation products were identified in the bulk
sample. Figure 10. Oxidation of SiC grain and Si3N4 matrix followed by
secondary hydration. Location 2.
phase-signals
Į-SiC, Si3N4, ȕ-Si3N4, Į-Si3N4, + amorph From the top part of the wall (location 3) a lot of brick material is
lost during operation of the cell. The crumbed appearance of the
Table 3. Phase-content location 3 large grains. brick seems to indicate a different wear mechanism. The bulk
sample contains exactly the same phases as the original material.
After sieving, the fraction below 32µm shows the same phases, However, in the fine fraction the presence of silicates
but some weak Į-cristobalite and Į-tridymite peaks were also (Cristoballite and Trydimite) was confirmed together with minor
found. Some other reaction products, NaF and NH4F were present F-bearing phases (NaF and ammonium fluoride). The very minor
in very small proportions. presence of silicates (oxidation products) and the strong alteration
of the brick is in strong contrast with the other parts of the wall
Weak phase-signals where predominantly Na-silicates are found. It is conceivable that
Villaumite (NaF), Cristobalite low(SiO2), in the upper wall HF vapors played a role. The HF could derive
Tridymite(SiO2), Ammonium from the bath through interaction with moisture from for instance
Fluoride(NH4F) the alumina feed. It acts as a destructive agent by forming Si-
Table 4. Additional phases below 32 µm grain size at location 3. fluoride vapors from the Si-oxides. A brick with a very high
porosity will remain, what can be seen in the sample above the
In several instances fluids were observed to form on the brick bath level. The mechanism is illustrated in figure 11.
samples. On the whole surface of the slice from cut 1, a yellow
liquid emerges from the brick. Over 5 days, the liquid gets more
and more solid. This stuff appears soapy. It was separated and
dried at 80 °C in a stove for one week. The white rest was
powdered for Powder-XRD. It contains mostly Sodium Carbonate
Hydrate, Sodium Silicate and Sodium bicarbonate (NaHCO3).

Discussion of observations on autopsy materials

Secondary reactions took place after the pot was switched off and
cooled down. The period of storage between switch off and Figure 11. HF-gas attack on Si3N4 matrix. Location 3.

416
 between these parts. The black areas, representing porosity, are
Figure 12 below shows the cell with the assumed attack more frequent inside than outside. The dark crystals in the matrix
mechanism. With SiC bricks oxidation is a significant problem are ȕ-Si3N4 and the bright ones Į-Si3N4. The ȕ-Si3N4 crystals in
involving an exchange mechanism of oxygen possibly with part A have irregular shapes and needle shaped crystals of ȕ-Si3N4
fluorides that destroy the sidewall as time passes especially at and were observed in part B, especially in Material 2 (observed with
SEM). Part B of Material 2 shows larger areas of connected ȕ-
above the bath level. Figure 12 shows the role of the critical Si3N4 in the matrix than in Material 1.
agents O2, H2O, Na2O, and HF).
O2 Material 1, outside (part A) Material 1, inside (part B)

High SiF4
porosity and
SiO2 H2O
HF
Bath level
Porosity,
nepheline,
albite and Cryolite compounds
Material 2,outside (part A) Material 2, inside (part B)
SiO2
Na
SiC
ȕ-Si3N4
Na2SiO3 porosity
Na2O
Į-Si3N4
O2

Figure 14. Micrographs with reflected light microscopy of unused

SiC bricks. (Frame is 0,25 mm wide.)
Figure 12. Suggested wear mechanism on Si3N4 bonded SiC in
aluminium electrolysis cells at Corus Aluminium Delfzijl. Part B of Material 2 contains more ȕ-Si3N4 than the outer part A,
consistent with the microscopic observations. ȕ-Si3N4 is the high
It seems that moisture plays an important role. This moisture can temperature phase of silicon nitride. The chemical compositions
originate from the alumina and/or the refractory materials in the of the two parts of both bricks are listed in table 5 and 6.
lower sidewall and bottom of the cell.
Oxygen can penetrate from the outside of the cell to the inside Sample SiC Si3N4 SiO2* C< 850°C
through the collector bar windows or from above (due to poor Material 1 A 78,3 17,0 1,1 1,8
covering). Maybe a chimney effect is causing oxygen to circulate Material 1 B 79,9 15,0 1,1 0,2
from collector bar windows through the porosity of the refractory Material 2 A 72,5 22,8 2,1 0,3
material and access the SiC via the porosity in the mortar used Material 2 B 76,5 16,8 1,9 0,0
between SiC and steel shell. Sample Fe2O3 CaO Al2O3 Na2O
Material 1 A 0,19 0,17 <0,1 0,027
Observations on unused materials Material 1 B 0,16 0,14 <0,1 0,047
Material 2 A 0,41 0,21 0,18 <0,001
Cut surfaces of unused material 1 (same quality as in the Material 2 B 0,27 0,15 0,13 <0,001
autopsied cell) and material 2 (similar quality from another Table 5. Chemical Composition in [weight%]. *calculated
supplier) show that the inner core of the brick is darker than the
outer shell (see figure 13 below). The samples A (bright part) and Sample N [weight%] O
B (dark part) were taken for chemical analysis and Powder-XRD. Ccalc >850¡C
[weight%]
Pieces of parts A and B were also used for light-microscope [weight%]
observations. Material 1 A 6,8 0,6 23,5
Material 1 B 6,0 0,6 24,0
Material 2 A 9,1 1,1 21,8
Material 2 B 6,7 1,0 22,9
Table 6. Measured nitrogen, oxygen and calculated carbon
Figure 13. Sampling of unused SiC bricks
The phases SiC and Si3N4 and SiO2 are calculated from the N and
The four pictures in figure 14 were selected to show cross sections O analyses. The calculated SiO2 represents oxidized Si3N4 and
of the inner and the outer parts of Material 1 and Material 2. SiC. A crystalline phase of SiO2 was not found either by XRD or
Under the microscope both qualities have a coarser matrix inside by SEM-analysis. The XRD pattern of Material 2 shows weak
(part B) than outside (part A), shown by the difference in contrast signals of Si2N2O but these could not be found in other samples or

417
confirmed by SEM-analysis. The only significant difference and between the shell and the bricks. Good covering practice is
between the four sample compositions is found in the N-content also important. Improved materials without Si3N4 may also be an
of the outer part (A) of Material 2, which is about 30-50% higher option.
than all other N-values.
Oxidation resistance

The differences between the core and the shell of Material 1 and 2.5
Material 2 lies in the coarser matrix and higher porosity of the
2

Weight increase (%)

core of both bricks. Moreover Material 2 contains more ȕ-Si3N4 in
the core. The ȕ-Si3N4 represents the high temperature polymorph 1.5

(stable above 1400°C) indicating that conditions during 1

manufacturing were favorable to transformation of Į-Si3N4 into ȕ- 0.5

Si3N4. Since ȕ-Si3N4 does not revert easily to the D-modification, 0
these conditions were not met in Material 1 to the same extent. B D F
Product

However, the used (from autopsy) bricks of Material 1 (location

Polarised corrosion test
1) show also a higher ȕ-Si3N4 content in the middle part like the
2.5
unused Material 2. It seems that the ȕ-Si3N4 content is not a
tightly controlled product parameter and can vary from brick to 2

Volume loss (%)

brick. 1.5

Discussion and conclusion 0.5

In earlier work by several people NaAlF4 is described as being a
0
main agent in the wear mechanism of nitride bonded SiC in the B D F

gas phase[1]. Reaction products must then contain aluminium. Product

However we do not observe any aluminium containing phases in

SiC and Si3N4 content in sidewall blocks
the autopsy samples taken above bath level. The polarised test
presented by Skybakmoen et al. does not simulate a wear 100
90
mechanism where these agents are involved. The main attack 80
agents in this polarised test in general are cryolite, liquid 70
60
Mass %

Si3N4
aluminium, NaAlF4, CO and CO2. 50
SiC
40
30
20
Instead we find sodium hydroxide and assume that sodium metal 10
or sodium oxide combined with moisture is the cause of the 0
A B C D E F
chemical attack. HF gas is also assumed to cause chemical attack. Product
In [3] it is shown that a significant amount of HF is generated in a
modern aluminium electrolysis cell. Figure 15. Oxidation resistance, chemical resistance and chemical
Only at the bath level (location 2) Na-Al-silicates were found. composition (partly) of nitride bonded SiC products [2.]

A more realistic test on Si3N4 bonded SiC sidewall bricks to study Acknowledgements
the chemical attack in an electrolysis cell system should involve The authors wish to thank Sieger van der Laan, Jan Trouw, Jurgen
Na-metal together with oxygen, moisture and NaF in a Bakker, Siegfried Benninghoff and the people at Aluminium
temperature range between 300 and 1000°C. This will imitate the Delfzijl BV for their invaluable help.
attack mechanisms we observe and would allow a better
distinction between good and poor products. Determining the Literature
amount of ȕ-Si3N4 or the ratio to Į-Si3N4 is also an important [1] Skybakmoen et al., “Chemical Resistance of Sidelining
quality criterion. We intend to implement this in our evaluation of Materials”, Light Metals 1999, p215-222.
sidewall products and we hope to be able to report on this in the [2] Selected supplier information and data sheets.
future. [3] Patterson, Hyland, Moxnes, Welch, “Reducing HF emissions
from Al electrolysis cells,” Proceedings of Seventh Australasian
Today product quality of sidewall bricks is mainly evaluated by Aluminium Smelting Technology Conference and Workshops
three parameters: chemical composition, chemical resistance as (11-16 November 2001) ISBN 073341851.
determined by the polarised test and oxidation resistance. For a
number of SiC products from various suppliers these parameters
are compared in figure 15. It is obvious that only in the oxidation
test a significant difference is observed.

The Si3N4 bonded SiC brick is not resistant to the combination of

moisture, Na-metal, HF and oxidation. The combination of these
agents is one of the most aggressive reactants for oxides,
especially for SiO2.
Stopping air circulation is an important step to improve the
lifetime of the SiC brick in the E-cell. This can be achieved by
using denser SiC-bricks and by blocking the openings in the shell

418